article stringlengths 1.98k 169k | summary stringlengths 1.01k 4.15k | section_headings listlengths 2 38 | keywords listlengths 0 12 | year stringclasses 11 values | title stringlengths 30 189 |
|---|---|---|---|---|---|
Animals perform many stereotyped movements , but how nervous systems are organized for controlling specific movements remains unclear . Here we use anatomical , optogenetic , behavioral , and physiological techniques to identify a circuit in Drosophila melanogaster that can elicit stereotyped leg movements that groom the antennae . Mechanosensory chordotonal neurons detect displacements of the antennae and excite three different classes of functionally connected interneurons , which include two classes of brain interneurons and different parallel descending neurons . This multilayered circuit is organized such that neurons within each layer are sufficient to specifically elicit antennal grooming . However , we find differences in the durations of antennal grooming elicited by neurons in the different layers , suggesting that the circuit is organized to both command antennal grooming and control its duration . As similar features underlie stimulus-induced movements in other animals , we infer the possibility of a common circuit organization for movement control that can be dissected in Drosophila .
An animal may perform a particular movement in response to its environment and internal state , and many movements are selected from a repertoire of stereotyped motor patterns . This repertoire can include movements that serve particular purposes , such as feeding , grooming , song production , locomotion , and even coordinated facial poses for expressing different emotions ( Grillner and Wallén , 2004; Grillner et al . , 2005 ) . Many of these movements are produced by neural networks called pattern generators , which control the precise timing of motor neuron activity to coordinate stereotyped patterns of muscle contractions ( Pearson , 1993; Kiehn and Kullander , 2004 ) . These neural networks are localized to distinct regions of the central nervous system ( CNS ) , such as the hindbrain and spinal cord in vertebrates and ventral nerve cord in arthropods , and are capable of producing their respective movements even when experimentally isolated from the brain and sensory inputs ( Grillner et al . , 2005; Marder et al . , 2005; Büschges et al . , 2011 ) . The activity of pattern generators can be initiated , or adapted to particular circumstances , in response to inputs from proprioceptive neurons , neuromodulators , or command-like neurons ( Marder et al . , 2005; Grillner , 2006; Ritzmann and Büschges , 2007; Blitz and Nusbaum , 2011; Harris-Warrick , 2011 ) . The proximal trigger of pattern generator activity arises from the command-like neurons , which can consist of individual neurons , or populations of neurons that induce or ‘command’ specific movements ( Kupfermann and Weiss , 1978; Jing , 2009 ) . Such neurons are also implicated in controlling different parameters of the movements that they induce , such as their speed or duration ( Kupfermann and Weiss , 1978; Pearson , 1993; Kristan , 2008; Jing , 2009 ) . However , the organizational principles underlying how such command circuitry can both initiate movements and control their parameters remain unclear . Moreover , in many cases in which the activity of command-like neurons has been experimentally manipulated ( Jing , 2009 ) , the behavioral impact of the manipulations was not assessed in intact and freely moving animals . The development of neurogenetic tools in Drosophila has led to rapid progress in identifying command-like neurons . This progress has been enabled by the use of optogenetic and thermogenetic activation of specific , genetically targeted populations of neurons in freely moving adult flies ( Flood et al . , 2013a; Owald et al . , 2015 ) . For example , activation of specific neuronal types can elicit stereotyped movements such as feeding , locomotion , courtship song , or escape ( Lima and Miesenböck , 2005; von Philipsborn et al . , 2011; Flood et al . , 2013b; Gao et al . , 2013; Inagaki et al . , 2014; Bidaye et al . , 2014; von Reyn et al . , 2014 ) . Although some of these studies have led to the identification of individual neurons and groups of neurons that command their respective behaviors , the anatomical organization of and functional connections among such groups of neurons that control these movements has been largely unexplored . Grooming movements ( a . k . a . cleaning , scratching , or wiping reflexes ) can be studied to determine the neuronal mechanisms by which specific movements are initiated and controlled . Grooming is ubiquitous among limbed animals as a means of protecting the body surface from different types of mechanical or chemical irritants ( Sachs , 1988 ) . Such sensory stimuli induce the movement of a limb to the irritated body part , which then scratches or wipes the surface ( Stein , 1983; Dürr and Matheson , 2003 ) . Because grooming movements can be predictably elicited by defined stimuli , they offer a way to access the sensory-connected neural circuitry that commands precisely targeted limb movements . In addition , grooming movements exhibit differing response durations , limb trajectories , stimulus-induced habituation , and can be suppressed , suggesting that the neural circuitry underlying these movements is subject to regulation and flexible control ( Sherrington , 1906; Stein , 1983; Corfas and Dudai , 1989; Page and Matheson , 2009; Seeds et al . , 2014 ) . Therefore , the study of grooming movements may reveal basic principles of movement control , but little is known about the neuronal mechanisms governing their initiation and modulation . We previously discovered that activating different neuronal populations in the fly nervous system could induce distinct grooming movements , such as grooming of the eyes , antennae , wings , thorax , or legs ( Seeds et al . , 2014 ) . This raised the possibility that the functional organization of the neural circuitry controlling specific grooming movements could be defined . Here we examine a circuit that commands one of these grooming movements—antennal grooming . This movement involves the grasping and brushing of the antennae with the legs in response to different types of irritants ( Robinson , 1996; Böröczky et al . , 2013; Seeds et al . , 2014 ) . To deconstruct the neural circuitry underlying antennal grooming , we isolated a small number of GAL4-expressing transgenic lines that could elicit the appropriate leg movements when driving expression of the temperature-gated neuronal activator dTrpA1 ( Hamada et al . , 2008; Seeds et al . , 2014 ) . However , these lines expressed GAL4 in multiple neuronal subsets , making it difficult to determine which neurons were responsible for the grooming movement . In this work , we refine these GAL4 lines to identify the specific neurons that elicit antennal grooming . We show that these neurons are functionally connected to form a circuit that detects displacement of the antennae via mechanosensory neurons and then commands grooming through three different interneuronal classes . Our analysis of the complex organization of this circuit provides insight into how stereotyped movements are controlled .
Given that grooming is induced by stimulation of the body surface , we first sought to identify sensory neurons that could relay such stimulation from the antennae to the brain . To this end , we revisited the GAL4 lines we identified in our previous behavioral screen and examined them for expression in the antennae ( Seeds et al . , 2014 ) . One line expressed GAL4 in mechanosensory chordotonal neurons of the Johnston's Organ ( JO , see below ) , and elicited antennal grooming when the targeted neurons were thermally activated by dTrpA1 ( Figure 1—figure supplement 1A , B ) . Because this line also expressed in interneurons in the CNS ( Figure 1—figure supplement 1C , D ) , we identified four additional GAL4 lines that target these sensory neurons by visually screening an image database of GAL4 line expression patterns ( Jenett et al . , 2012 ) . Each of these lines elicited antennal grooming when used to thermogenetically activate the targeted neurons , further implicating this population of JO neurons in the behavior ( Figure 1—figure supplement 1G–J , behavior not shown ) . While our results were consistent with these sensory neurons being responsible for eliciting antennal grooming , the expression of GAL4 in other neurons allowed the possibility that they were contributing to this behavior . To examine whether the JO neurons elicit antennal grooming , we used the intersectional Split GAL4 ( spGAL4 ) approach in an effort to restrict GAL4 activity to these neurons ( Luan et al . , 2006; Pfeiffer et al . , 2010 ) . Specifically , the activation domain ( AD ) of GAL4 is expressed in the genomic enhancer-driven pattern of one identified GAL4 line , while the DNA binding domain ( DBD ) of GAL4 is expressed in the pattern of another line ( see ‘Materials and methods’ ) . When both halves are co-expressed in the same cell , the activity of GAL4 is reconstituted . By co-expressing the AD and DBD in the patterns of different enhancer pairs , we observed reconstitution of GAL4 activity in the JO neurons ( Figure 1—figure supplement 2A–E′ ) . Five of these different AD/DBD pairs showed increased antennal grooming with thermogenetic activation ( Figure 1A , pairs referred to as aJO-spGAL4-1 through aJO-spGAL4-5 , Supplementary file 1 shows enhancer pairs used ) . 10 . 7554/eLife . 08758 . 003Figure 1 . Sensory neurons that elicit antennal grooming . ( A ) Grooming movements performed by flies in which aJO spGAL4 pairs drove expression of thermally activated dTrpA1 . Movements were manually scored from 2 min of recorded video per fly ( n ≥ 17 flies per spGAL4 ) . Colors correspond to the percent of total time spent performing each movement . ( B ) Percent time flies spent antennal grooming with thermogenetic activation of neurons targeted by spGAL4 pairs , with or without their antennae ( filled or open boxes , respectively ) . Bottom and top of the boxes indicate the first and third quartiles respectively; median is the red line; whiskers show the upper and lower 1 . 5 IQR; red dots are data outliers ( n ≥ 17 for each box; asterisks show p < 0 . 0001 , Kruskal–Wallis and post hoc Mann–Whitney U pairwise tests with Bonferroni correction ) . Dotted line marks the median of the intact control . ( C–F ) aJO-spGAL4-1 driving expression of green fluorescent protein ( GFP ) . Maximum intensity projections are shown . ( C ) Frontal view of the head ( native GFP fluorescence , green; cuticle autofluorescence , magenta ) . Left bracket shows the third antennal segment . Right bracket marks the second antennal segment , which is shown in ( D ) . Scale bar , 100 μm . ( D ) Second antennal segment co-stained with anti-GFP ( green ) and anti-Elav ( magenta , marks neuronal nuclei ) antibodies . White arrows show the ventral and dorsal aJO clusters . Scale bar , 25 μm . ( E , F ) Central nervous system ( CNS ) co-stained with anti-GFP ( green ) and anti-Bruchpilot ( magenta ) to visualize the aJO afferent projections into the ventral brain neuropile ( E ) and their specific targeting of the indicated antennal mechanosensory and motor center ( AMMC ) and subesophageal zone ( SEZ ) regions ( arrows shown in F ) . Box in ( E ) indicates region shown in F . Scale bars , ( E ) 100 μm and ( F ) 25 μm . Prothoracic neuromeres ( ProNm ) . Ventral nervous system ( VNS ) . See also Figure 1—figure supplement 1 and Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 00310 . 7554/eLife . 08758 . 004Figure 1—figure supplement 1 . GAL4 lines that target expression to sensory neurons from the antennae and elicit grooming . ( A ) Percent of total time three GAL4 lines expressing dTrpA1 spent antennal grooming , compared with controls . Box plots , statistics , and experimental conditions are as described in Figure 1A , B ( n ≥ 10 , asterisks represent: ***p < 0 . 001 , *p < 0 . 01 ) . R26B12-GAL4 and R18C11-GAL4 expression patterns are shown in Figure 2—figure supplement 1B , C . ( B ) Head GFP expression pattern of R39A11-GAL4 ( native GFP fluorescence shown , green ) . The cuticular autofluorescence is shown in magenta . Scale bar , 100 μm . ( C , D ) Expression in ( C ) the CNS and ( D ) SEZ of R39A11-GAL4 . Brains were co-stained with anti-GFP ( green ) and anti-bruchpilot ( magenta ) . Scale bar , ( D ) 50 μm . ( E , F ) CNS images of control lines used in this study: ( E ) GAL4 and ( F ) spGAL4 controls do not show significant expression . Scale bars , 100 μm . ( G , J ) GFP expression patterns of the CNS of GAL4 lines that had expression in sensory neurons projecting from the antennae and displayed increased antennal grooming with dTrpA1 . ( G ) R25F11-GAL4 , ( H ) R52F12-GAL4 , ( I ) R60E02-GAL4 ( aJO-GAL4-1 ) and ( J ) R27H08-GAL4 ( aJO-GAL4-2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 00410 . 7554/eLife . 08758 . 005Figure 1—figure supplement 2 . spGAL4 pairs that target expression to sensory neurons in the antennae and elicit grooming . ( A–E ) CNS expression patterns of spGAL4 line pairs that displayed increased antennal grooming with dTrpA1 . ( A′–E′ ) Native GFP expression ( green ) in the antennae of corresponding spGAL4 lines shown in A–E . The cuticular autofluorescence is shown in magenta . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 00510 . 7554/eLife . 08758 . 006Figure 1—figure supplement 3 . JO neurons projecting to zone C/E elicit antennal grooming . ( A , B ) Co-expression of JO sensory neurons using aJO-LexA to express GFP ( green ) and ( A ) JO4-GAL4 or ( B ) JO31-GAL4 lines to express tdTomato ( magenta ) . See Figure 4—figure supplement 1 for more details about aJO-LexA . aJO-LexA neurons have distinct projections in the ventral SEZ ( middle and right panel , red arrows ) , but show overlapping arborizations with zone C/E neurons in the AMMC region ( right panel , white arrows ) driven by JO4-GAL4 . Note in A ( left and right panel ) that aJO-LexA does not overlap with the zone A projections ( blue arrow ) . Scale bar , 50 μm . ( C–G ) CNS expression patterns of JO-GAL4 lines that target different zones in the AMMC . ( C ) JO4-GAL4 ( zones A and C/E ) ; ( D ) JO31-GAL4 ( zones C/E ) ; ( E ) JO3-GAL4 ( zones A , B , C/E , D ) ; ( F ) JO22-GAL4 ( zone A ) , and ( G ) JO15-GAL4 ( zone A and B ) . Scale bar , 100 μm . ( H ) Number of bouts per minute of antennal grooming that each JO-GAL4 line expressing CsChrimson performed with optogenetic activation , with or without their antennae ( filled or open boxes , respectively ) . Box plots and statistics are described in Figure 1B . ( n ≥ 4 flies per JO-GAL4 , asterisks represent: *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . ) DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 00610 . 7554/eLife . 08758 . 007Figure 1—figure supplement 4 . Most stochastically labeled aJO neurons show projections to both the AMMC and ventral SEZ . ( A ) Overview of the aJO neuron population with AMMC , posterior , and ventral SEZ projections ( white , yellow , and red arrows point to each projection respectively ) . ( B–F ) Multicolor stochastic labeling of aJO neurons in five different brains . The full patterns shown in panels B , C , D , E , F , whereas the separated channels are shown in panels B′ , B′′ , C′ , C′′ , D′–D′′′ , E′–E′′′ . Cases where single neuron clones were obtained using this method are labeled with red numbers . ( G , H ) Lateral view of two neurons labeled with the same color from panel B′′ in ( G ) and from panel C in ( H ) show projections to the AMMC ( white arrow ) , ventral SEZ ( red arrow ) and posterior SEZ ( yellow arrow ) . ( I ) Table shows the analysis of the projection branches of individual neurons corresponding with the red numbers in the figure panels . The + or − signs indicate that the neurons either do or do not project to the indicated region respectively . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 007 Confocal imaging confirmed that each spGAL4 pair targeted neurons within the second antennal segment , as revealed by staining for the neuronal protein Elav ( Figure 1C , D , Figure 1—figure supplement 2A′–E′ ) ( Kamikouchi et al . , 2006 ) . The anatomy of these neurons identifies them as subsets of the approximately 500 chordotonal neurons within the JO , a mechanosensory structure that detects antennal movements . All five pairs were expressed in two distinct clusters of 40–50 neurons each in the dorsal and ventral regions of the JO , and we designate these clusters collectively as the antennal grooming JO ( aJO ) ( Figure 1D , Figure 1—figure supplement 2A′–E′ ) . Although four of the pairs still targeted GAL4 activity to neurons outside of the JO ( Figure 1—figure supplement 2B–E ) , aJO-spGAL4-1 expressed almost exclusively in the aJO , providing strong evidence that these sensory neurons elicit antennal grooming ( Figure 1D , E ) . To independently evaluate the necessity of the aJO in antennal grooming , we tested whether amputation of the antennae , and thus removal of the JO , would abolish the thermogenetically elicited behavior of the spGAL4 pairs . Indeed , antennal amputation abolished grooming at temperatures that would normally activate dTrpA1 and elicit antennal grooming ( Figure 1B ) . These experiments demonstrate that the aJO induces antennal grooming . The JO comprises different subgroups of chordotonal neurons whose axons project to distinct zones ( zones A–E ) in the brain antennal mechanosensory and motor center ( AMMC ) ( Kamikouchi et al . , 2006; Yorozu et al . , 2009; Matsuo et al . , 2014 ) . aJO axons enter the AMMC with zone C/E neurons ( Figure 1—figure supplement 3A–B ) . However , unlike previously described C/E neurons that terminate in the AMMC , the aJO have three apparent projections within the ventral brain: the AMMC , the ventral subesophageal zone ( SEZ ) , and the posterior SEZ ( Figure 1E , F ) . To test whether each neuron within the aJO has all three projections , we performed multicolor stochastic labeling ( Nern et al . , 2015 ) , which allows for visualization of individual neurons within the aJO population . Because each of the single neurons that we isolated projects from the AMMC to the ventral SEZ ( Figure 1—figure supplement 4A–F , I ) , it would appear that the majority of neurons within the aJO have similar projections . A smaller subset appears to project from the AMMC to the posterior SEZ; however , we were unable to isolate individual cells to definitively show this ( Figure 1—figure supplement 4G , H ) . In contrast to previously identified zone C/E-projecting neurons , we found no evidence of aJO neurons that project only to the AMMC . Because none of the previously described JO neurons project to the ventral SEZ ( Kamikouchi et al . , 2006; Matsuo et al . , 2014 ) , the aJO corresponds to a previously unrecognized set of neurons . Given that aJO neurons project to zone C/E before passing to the SEZ , we tested whether activation of previously described C/E neurons could elicit antennal grooming . Indeed , activation of zone C/E neurons using published GAL4 drivers ( Kamikouchi et al . , 2006 ) elicited antennal grooming ( Figure 1—figure supplement 3C–E , H ) . In contrast , activation of zone A or B neurons did not elicit antennal grooming ( Figure 1—figure supplement 3F , G , H ) , indicating that only the zone C/E subpopulation elicits the behavior . Therefore , our data indicate that at least two types of zone C/E-projecting neurons are sufficient to induce grooming . The first terminates within the AMMC and constitutes a previously described set of JO neurons ( Kamikouchi et al . , 2006 ) , whereas the second type are the aJO neurons that project into the AMMC and then ventrally to the SEZ . The muscles that control front leg movements necessary for antennal grooming are innervated by neurons residing in the most anterior region of the ventral nervous system ( VNS ) , the prothoracic neuromeres ( ProNm ) ( Figure 1E ) ( Burrows , 1996; Brierley et al . , 2012 ) . Because JO afferent projections terminate in the brain and do not project to the ProNm , where they would be positioned to activate leg movements , we reasoned that additional neurons must project to the ProNm to command grooming behavior . Therefore , we sought to identify interneurons that transmit the sensory signal to the ProNm . Of the GAL4 lines that elicited antennal grooming in our previous screen ( Seeds et al . , 2014 ) , two lack expression in antennal sensory neurons but have expression in interneurons within the brain ( lines R26B12 and R18C11 , behavioral analysis in Figure 1—figure supplement 1A , expression patterns in Figure 2—figure supplement 1A–C ) . A visual screen of the GAL4 line expression pattern database ( Jenett et al . , 2012 ) identified additional lines with interneuron projection patterns in the AMMC and SEZ that made them candidates for associating with the JO projections . Five of these lines elicited antennal grooming when thermogenetically activated ( Figure 2—figure supplement 1A ) . As each GAL4 line expressed in other neuronal populations ( Figure 2—figure supplement 1D–H ) , we again generated spGAL4 lines using their respective enhancers to further restrict GAL4 activity to the behaviorally relevant interneurons . With this approach , we identified spGAL4 pairs that elicited antennal grooming with thermogenetic activation , and targeted interneurons that we designated aBN1 , aBN2 , and antennal descending neuron ( aDN ) for reasons described below ( Figure 2A , Supplementary file 1 ) . Of note , there were striking differences in the amounts of grooming elicited by these pairs ( Figure 2A ) . For example , aDN1-spGAL4-1 flies spent 4 . 8% of their time grooming their antennae vs 66 . 1% for aBN2-spGAL4-2 ( Figure 2B , black boxes ) . 10 . 7554/eLife . 08758 . 009Figure 2 . Interneurons that elicit antennal grooming . ( A ) Grooming movements performed by interneuron spGAL4 pairs expressing thermally activated dTrpA1 . Data was obtained and displayed as described in Figure 1A . ( B ) Percent time flies spent antennal grooming with thermogenetic activation of interneurons targeted by spGAL4 pairs , with or without their antennae ( filled or open boxes respectively ) . Box plots and statistics are described in Figure 1B . Asterisks represent the following p values: *p < 0 . 01 , **p < 0 . 001 , ***p < 0 . 0001 ( n ≥ 9 flies per spGAL4 ) . Black p value statistics show differences between control and spGAL4 flies with their antennae . Gray statistics show differences between each spGAL4 with and without the antennae . ( C–F ) GFP expression patterns of spGAL4 lines: ( C ) aBN1-spGAL4-1 , ( D ) aBN2-spGAL4-2 , ( E ) aDN1-spGAL4-1 , ( F ) aDN2-spGAL4-4 . Images show maximum intensity projections of co-staining with anti-GFP ( green ) and anti-Bruchpilot ( magenta ) . White arrows show cell bodies . Scale bars , 100 μm . ( G ) antennal descending neurons ( aDNs ) targeted by each spGAL4 pair . Circles represent one of three neurons in aDN-GAL4 . Filled circles show which neurons are targeted by each spGAL4 ( enhancer pairs listed ) . ( H ) Two aDN neurons are targeted in a triple spGAL4 combination expressing GFP ( white arrows ) . No spGAL4 combinations were identified that exclusively target aDN3 . Scale bar , 25 μm . ( I ) Graphical summary of neuronal expression patterns of the spGAL4 pairs . Green boxes indicate expression of the pair on the left ( rows ) in the indicated neurons or region listed above the grid ( columns ) . Figure 2—figure supplement 2 shows the locations of these neurons . Black framing highlights antennal grooming neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 00910 . 7554/eLife . 08758 . 010Figure 2—figure supplement 1 . GAL4 lines that elicit antennal grooming . ( A ) Grooming movements performed by GAL4 lines expressing thermally activated dTrpA1 . Data was obtained and displayed as described in Figure 1A ( n = 10 per line ) . Note: R11B11-GAL4 did not perform significantly increased antennal grooming , however , this line was able to target antennal grooming interneurons when its enhancer was used for generating spGAL4 intersections . ( B–H ) Images of antennal grooming GAL4 lines expressing GFP . CNSs were co-stained with anti-GFP ( green ) and anti-bruchpilot ( magenta ) . ( B ) R26B12-GAL4 , ( C ) R18C11-GAL4 ( aDN-GAL4 ) , ( D ) R34C03-GAL4 , ( E ) R11B11-GAL4 , ( F ) R70H02-GAL4 , ( G ) R71D01-GAL4 , and ( H ) R76F12-GAL4 . Inset in C shows higher magnification of three candidate antennal grooming interneuron cell bodies . White arrow in ( B , C ) show where projections from JO sensory neurons would enter the brain if the GAL4 line targeted expression to them . Scale bars , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 01010 . 7554/eLife . 08758 . 011Figure 2—figure supplement 2 . spGAL4 lines with interneuron expression that elicit antennal grooming . ( A–I ) spGAL4 lines expressing GFP ( green ) are co-stained with anti-GFP ( green ) with anti-bruchpilot ( magenta ) . The activation domains ( ADs ) and DNA binding domains ( DBDs ) used are shown below each corresponding spGAL4 pair name . Arrows or lines indicate neurons or populations of neurons targeted by each pair . Summary expression data for all spGAL4 pairs are shown in Figure 2I . ( A ) aBN1-spGAL4-1 , ( B ) aBN1-spGAL4-2 , ( C ) aBN1/aDN1-spGAL4 , ( D ) aBN2-spGAL4-1 , ( E ) aBN2-spGAL4-2 , ( F ) aDN1-spGAL4-1 , ( G ) aDN2-spGAL4-2 , ( H ) aDN2-spGAL4-3 , and ( I ) aDN2-spGAL4-4 . Insets in D , E show aBN2 cell bodies at higher magnification; ( D ) shows three cell bodies and ( E ) shows five . Scale bars , 100 μm . ( J ) Summary of neuronal types targeted by interneuron spGAL4 pairs . Abbreviations: abdominal ganglion cluster ( AG-Cl ) , ascending cluster ( AscCl ) , metathoracic neuromere cluster ( MetaNmCl ) , mesothoracic neuromere cluster ( MesaNmCl ) , ProNm cluster ( ProNmCl ) , posterior cluster ( pCl 1 + 2 ) , dorsolateral cluster ( dLatCl ) , lateral cells ( latCells ) , VNS , central complex ( CX ) , optic lobes ( OL ) , aDN , antennal local interneurons ( aBN ) , pars intercerebralis cluster ( PI ) . Dashed lines show neuronal population on posterior side of the brain . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 011 To validate that the spGAL4 lines target interneurons , we removed sensory neurons by amputating the antennae , and then measured the time spent grooming during thermogenetic activation . As anticipated , amputation did not abolish movements directed towards the antennal region , and some pairs significantly increased such movements relative to intact antennae ( Figure 2B , aBN1/aDN1-spGAL4 , aBN2-spGAL4-1 , and aDN2-spGAL4-2/3/4 show increased grooming ) . Thus , we conclude that these spGAL4 pairs target interneurons that induce antennal grooming when activated . Moreover , the increased grooming associated with antennal amputation raises the possibility that grooming might be negatively influenced by sensory feedback from the antennae . We next examined the anatomy of these interneurons in greater detail . Two classes that induce antennal grooming are located entirely in the brain , and we designate them antennal grooming brain interneurons 1 and 2 ( aBN1 and aBN2 ) . aBN1-spGAL4-1 targeted expression to aBN1 , a single interneuron in the ventral brain ( Figure 2C ) . Two other spGAL4 pairs also targeted aBN1 and were able to elicit antennal grooming with thermogenetic activation ( aBN1-spGAL4-2 , aBN1/aDN1-spGAL4 , Figure 2I , Figure 2—figure supplement 2A–C ) . aBN2 was targeted by two spGAL4 pairs that each used the R26B12 enhancer , which itself expresses in a cluster of eight neurons with cell bodies in the posterior and ventrolateral brain . Each pair targets expression to either three or five of these neurons ( aBN2-spGAL4-1 , aBN2-spGAL4-2 , Figure 2D , I , Figure 2—figure supplement 2D , E , Figure 4—figure supplement 2C , D ) . The other interneuron class comprises aDNs that project from the brain to the VNS . Five spGAL4 pairs targeted single descending neurons with cell bodies located in the posteroventral SEZ ( Figure 2E , F , I , Figure 2—figure supplement 2C , F–I ) . Given that the R18C11 enhancer was used to generate several of these pairs , and by itself targets expression to three aDNs when driving GAL4 ( Figure 2—figure supplement 1C ) , we asked whether each spGAL4 pair targeted the same neuron or distinct neurons . By simultaneously combining the GAL4 DBD , expressed under control of the R18C11 enhancer , with two versions of the GAL4 AD , one expressed under control of the R76F12 enhancer and the other controlled by the R34C03 enhancer , we observed two aDNs ( Figure 2G , H ) . This demonstrates that the spGAL4 pairs target distinct descending interneurons , and we named these aDN1 and aDN2 . However , we did not identify a spGAL4 combination that exclusively expresses in the third R18C11-targeted aDN ( named aDN3 ) . The identified neuronal classes all have projections in the ventral brain ( Figure 3A–D ) . aBNs project to the AMMC and SEZ , following the aJO projections ( Figure 3B , C ) , whereas two aDNs ( aDN1 and aDN2 ) , and likely aDN3 project both to the SEZ and through the cervical connective to the ProNm ( Figure 3D , only aDN1 shown ) . Manual and computational alignment of the projections from these classes suggested intimate associations and the potential to form a neural circuit ( Figure 3E , Video 1 , Video 2 ) . Thus , we explored their functional relationships . 10 . 7554/eLife . 08758 . 008Figure 3 . Neurons that elicit antennal grooming have neurites in the AMMC and/or SEZ . ( A–D ) spGAL4 pairs targeting each neuronal class in the ventral brain: ( A ) aJO-spGAL4-1 , ( B ) aBN1-spGAL4-1 , ( C ) aBN2-spGAL4-2 , and ( D ) aDN1-spGAL4-1 . aDN1 is shown as an example in ( D ) , but there are additional aDNs ( aDN2 and aDN3 , see Figure 2F and Figure 2—figure supplement 2G–I ) . CNSs stained with anti-GFP ( green ) and anti-bruchpilot ( magenta ) . Maximum intensity projections are shown from frontal and lateral views . Arrows show the different projection regions: AMMC ( white ) , posterior SEZ ( yellow ) , and ventral SEZ ( red ) . Scale bars , 25 μm . ( E ) Traced neurons in different colors manually aligned ( also shown in Video 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 00810 . 7554/eLife . 08758 . 012Video 1 . Traced antennal grooming circuit . Traced and manually aligned neurons are shown in different colors . aJO neurons ( blue ) project from the second antennal segment into the anterior brain . aBN1 ( red ) has cell bodies in the dorsal and posterior brain . aBN2 is shown in yellow . aDN1 is shown in green and sends descending projections to the VNS . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 01210 . 7554/eLife . 08758 . 013Video 2 . Computationally aligned antennal grooming circuit . Computationally aligned neurons are shown in different colors . aJO neurons ( blue ) , aBN1 ( red ) , aBN2 ( yellow ) and aDN1 ( green ) . The neuropil was stained with anti-bruchpilot ( grey ) . See ‘Materials and methods’ for a description of how the computational alignment and rendering of images were done . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 013 We first asked whether the grooming elicited by aJO thermogenetic activation required the activity of the aBNs and aDNs by performing a behavioral epistasis test in which we genetically silenced the activity of the interneurons while activating the aJO ( Figure 4A ) . For silencing the interneurons , we used the aBN spGAL4 pairs and aDN-GAL4 to target expression of the synaptic transmission blocker , tetanus toxin ( TNT ) ( Sweeney et al . , 1995 ) . To genetically access the aJO independent of the interneurons , we employed the LexA binary transcriptional system so that the aJO enhancer R27H08 directed expression of LexA ( Lai and Lee , 2006; Pfeiffer et al . , 2010 ) . The resulting aJO-LexA line targeted aJO neurons and also elicited antennal grooming with thermogenetic activation ( Figure 4—figure supplement 1A , B ) . Although aJO-LexA targets more JO neurons than the aJO spGAL4 pairs ( Figure 4—figure supplement 2A , B ) , expression of TNT in the aJO by aJO-spGAL4-1 significantly decreased antennal grooming in response to thermogenetic activation by aJO-LexA . This showed that the aJO-spGAL4-1 neurons constituted a major portion of the grooming elicited with aJO-LexA ( Figure 4B , blue boxes ) . However , given that aJO-LexA targets additional zone C/E neurons that could induce grooming ( Figure 1—figure supplement 3C–E , H ) , it is possible that they are responsible for the residual grooming that occurs with expression of TNT in aJO-spGAL4-1 ( Figure 4B , blue boxes ) . Therefore , we hereafter refer to the neurons targeted by aJO-LexA as aJO+C/E neurons . We next found that TNT expression in the interneurons aBN1 and aBN2 suppressed grooming when aJO+C/E neurons were thermogenetically activated ( Figure 4B , red and yellow , respectively ) . However , grooming was not suppressed when TNT was expressed in the aDNs ( Figure 4B , green boxes ) . We conclude that the aBNs are necessary for the grooming response to aJO+C/E activation , while additional descending neurons important for mediating antennal grooming may remain to be identified . 10 . 7554/eLife . 08758 . 014Figure 4 . Functional relationships among putative antennal circuit components . ( A ) Overview of experiments shown in ( B , E ) . Grooming was induced by thermogenetic activation of Johnston's Organ ( JO ) neurons ( dTrpA1 ) or by imposed displacements of the antennae . Synaptic release was blocked in different neuronal classes expressing tetanus toxin ( TNT ) . ( B ) Antennal grooming performed by flies with thermogenetic activation of the aJO while inhibiting synaptic release in interneuron classes . The experiment was performed and data is displayed as described in Figure 1B ( n ≥ 11 flies per spGAL4 ) . ( C–E ) To displace the antennae , iron powder was glued to the third antennal segment and the flies were tethered within an electromagnet . ( C ) Image of the electromagnet apparatus . ( D ) Tethered fly with iron powder on its antennae . Magnetic pulses were delivered to displace the third antennal segment at 1 Hz for 4 × 10 s periods , with 30 s rests between stimulations . Flies were recorded and their grooming movements were manually scored ( see Figure 4—figure supplement 3D for ethograms ) . ( E ) The percent time that flies spent grooming their antennae within the total assay time is shown . The grooming responses to antennal movements were also tested while blocking synaptic release in the different neuronal types with TNT . Box plots and statistics are shown as described in Figure 1B ( n ≥ 11 flies per line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 01410 . 7554/eLife . 08758 . 015Figure 4—figure supplement 1 . aJO- , aBN2- , and aDN-LexA lines . ( A ) Percent of total time that LexA lines expressing thermally activated dTrpA1 spent antennal grooming . Box plots , statistics , and experimental conditions are as described in Figure 1A , B ( n > 10; ***p < 0 . 001 ) . ( B–D ) CNSs co-stained with anti-GFP ( green ) and anti-bruchpilot ( magenta ) . LexA lines are as follows: ( B ) aJO-LexA , ( C ) aBN2-LexA , and ( D ) aDN-LexA . Scale bar , 100 μm . Box insets show higher magnification of cell bodies of interneurons involved in antennal grooming; ( C ) shows four cell bodies and ( D ) shows three . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 01510 . 7554/eLife . 08758 . 016Figure 4—figure supplement 2 . Co-expression of LexA lines with selected spGAL4 pairs . ( A–F ) Co-expression of LexA and spGAL4 lines . Left column: spGAL4 expression patterns . Middle column: LexA expression patterns . Right column: Merged expression patterns . ( A , B ) Co-expression of aJO-LexA ( magenta ) with ( A ) aJO-spGAL4-1 or ( B ) aJO-spGAL4-3 ( green ) . Frontal views: Arrows point to JO projections that do not co-localize with aJO projections . ( C , D ) aBN2-LexA ( magenta ) expressed with ( C ) aBN2-spGAL4-1 or ( D ) aBN2-spGAL4-2 ( green ) . ( E , F ) Co-expression of aDN-LexA with ( E ) aDN1-spGAL4-1 or ( F ) aDN2-spGAL4-2 . Scale bar , 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 01610 . 7554/eLife . 08758 . 017Figure 4—figure supplement 3 . Testing of stimulus parameters for the antennal displacement assay . ( A–C ) Flies were prepared as shown in Figure 4C , D ( see ‘Materials and methods’ ) and tested for their grooming responses when different magnetic field frequencies ( A ) or voltages ( B ) were applied to displace their antennae . The grooming responses were recorded and manually scored . Plots are displayed as described in Figure 1B . ( A ) Box plots show the percent time that flies groomed their antennae when the magnetic field that was turned on and off at different frequencies ( magnetic strength was set at 10 Volts ) . ( B ) Percent time that flies groomed their antennae when the magnetic field was applied at different voltages ( frequency set at 1 Hz ) . Raw data used to generate the red box is shown in D . ( C ) Side view of a fly head showing the trajectories of the antennae before ( grey ) and after ( green ) the magnetic field was applied . Red asterisks and arrow show the measured distances . The values shown below the image represent the measured distances that the antennae moved at the indicated voltages . ( D ) Ethograms of grooming movements performed by 16 control flies ( w1118; UAS-TeTxLC . tnt; pBPGAL4U ) whose antennae were stimulated in 4 × 10 s periods by a magnetic field of 570 gauss ( 10 Volts ) at a frequency of 1 Hz . The magnet was left off for 30 s between periods . These flies are the controls used for the experiment shown in Figure 4E . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 017 We next sought to identify a sensory stimulus upstream of this putative grooming circuit . Because JO neurons detect antennal movements ( Kamikouchi et al . , 2009; Yorozu et al . , 2009; Matsuo et al . , 2014 ) , we reasoned that the role of the aJO might be to elicit grooming in response to physical displacements of the antennae . To test this , we glued iron powder to the third antennal segments and tethered the flies within an electromagnetic behavioral set up ( Figure 4C , D ) . Application of a magnetic field at 1 Hz caused visible displacements of the antennae ( 31 ± 6 μm ) and elicited grooming of both the antennae and other head parts ( Figure 4E , white boxes , Figure 4—figure supplement 3C , D , Video 3 ) . The aJO was critical for this response because expression of TNT in these neurons significantly reduced antennal and head grooming in response to the magnetic field ( Figure 4E , blue boxes , head grooming not shown ) . Given that activation of aJO elicits antennal grooming almost exclusively ( Figure 1A , Figure 7—figure supplement 1A , B ) , the role of the aJO in head grooming is unclear . 10 . 7554/eLife . 08758 . 018Video 3 . Grooming movements performed in response to displacements of the antennae . The third antennal segments of a control fly were coated with iron powder , and the fly was tethered within the electromagnetic apparatus shown in Figure 4C , D . The infrared light positioned behind the fly shows when the magnetic field was applied to displace the antennae . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 018 We next examined the functional necessity of the different interneuron classes for grooming in response to antennal displacement . aBN1 was found to be necessary , as expression of TNT in aBN1-spGAL4-1 or aBN1-spGAL4-2 both reduced the grooming response ( Figure 4E , red boxes ) . Intriguingly , expression of TNT targeted by the two different aBN2 spGAL4 pairs gave opposing results in this assay ( Figure 4E , yellow boxes ) : aBN2-spGAL4-1 activity showed necessity for antennal grooming while aBN2-spGAL4-2 did not . This contrasted with our earlier finding that both spGAL4 pairs disrupted grooming when the aJO+C/E neurons were thermogenetically activated ( Figure 4B , yellow boxes ) . This may reflect a functional difference between the specific subsets of aBN2 neurons targeted by these two spGAL4 pairs in the context of antennal displacement . Similar to the results for thermogenetic activation of aJO+C/E neurons , expression of TNT ( or the inward rectifying potassium channel Kir ) in all three aDNs did not block the grooming response to antennal displacement ( Figure 4E , green box , Kir data not shown ) . Taken together , our results demonstrate that all identified neurons within the putative circuit elicit antennal grooming when activated , but not all are necessary for the grooming response to antennal displacement . To address the potential for connectivity among these neurons , we first examined the relative proximities of their projections in the brain by visualizing expression of LexA in one neuronal class and spGAL4 in another . We found that the aBN1 co-localized with all major aJO projections , whereas aBN2 co-localized with the aJO , AMMC , and ventral SEZ projections ( Figure 5A , B , arrows ) . Interestingly , co-visualization of aJO with aDN1 or aDN2 revealed that the aDNs have distinct projections: aDN1 co-localizes with the aJO in the AMMC and ventral SEZ , whereas aDN2 only associated with the most ventral SEZ projections ( Figure 5C , D , arrows ) . We confirmed that aDN1 projects more dorsally than aDN2 by examining their relative projections within the aDN-LexA pattern ( Figure 4—figure supplement 2E , F ) . In support of our conclusion that the aBNs and aDNs are in close proximity with the JO projections , GFP-positive staining indicates reconstitution across synaptic partner experiments reported membrane contact among these neurons ( Figure 5—figure supplement 1A–D ) ( Feinberg et al . , 2008; Gordon and Scott , 2009 ) . 10 . 7554/eLife . 08758 . 019Figure 5 . Antennal grooming neurons are in close proximity . ( A–H ) Co-expression in neuronal pairs using two binary expression systems ( LexA and spGAL4 ) to express tdTomato or GFP in each neuronal class . Processed maximum intensity projections of frontal and lateral views are shown . See ‘Materials and methods’ about how images were processed ( unprocessed images in Figure 5—figure supplement 2 ) . Scale bars , 25 μm . ( A–D ) Proximity between aJO-LexA targeted sensory projections ( magenta ) and the following interneuron spGAL4 pairs ( green ) : ( A ) aBN1-spGAL4-1 , ( B ) aBN2-spGAL4-1 , ( C ) aDN1-spGAL4-1 and ( D ) aDN2-spGAL4-2 . ( E–G ) Proximity between aBN2-LexA targeted neurons ( magenta ) and the following interneuron spGAL4 pairs ( green ) : ( E ) aBN1-spGAL4-1 , ( F ) aDN1-spGAL4-1 , ( G ) aDN2-spGAL4-2 . ( H ) Proximity between aDN-LexA targeted neurons aDN ( green ) and aBN1-spGAL4-1 targeted neurons ( magenta ) . Overlap between different projections of the LexA and spGAL4-targeted neurons is indicated by different colored arrows: ( A–C , E , F ) AMMC projections white arrows , ( A ) posterior SEZ projection ( yellow arrow ) , ( B–H ) ventral SEZ projections ( red arrows ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 01910 . 7554/eLife . 08758 . 020Figure 5—figure supplement 1 . GFP-positive staining indicates reconstitution across synaptic partner ( GRASP ) staining indicates close proximity of neurons involved in antennal grooming . ( A–H ) Expression of different GFP halves ( spGFP11 and spGFP1-10 ) in putative connected neuronal pairs . aJO-LexA drove expression of spGFP11 in the JO neurons and spGFP1-10 expression was driven by the following spGAL4 lines: ( A ) aBN1-spGAL4-1 , ( B ) aBN2-spGAL4-1 , ( C ) aDN1-spGAL4-1 , ( D ) aDN2-spGAL4-2 , or ( E ) spGAL4 control . Brains were stained with an antibody that recognizes reconstituted GFP ( green ) ( Gordon and Scott , 2009 ) . The neuropil was stained with an anti-bruchpilot antibody ( magenta ) . GRASP where neurons contacted each other . ( F–H ) Controls where only GFP1-10 was expressed by: ( F ) aBN1-spGAL4-1 or ( G ) aBN2-spGAL4-1 , or ( H ) aDN1-spGAL4-1 . ( I ) White box marks the area shown in A–H . Scale bar , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02010 . 7554/eLife . 08758 . 021Figure 5—figure supplement 2 . Co-staining indicates close proximity of neurons involved in antennal grooming . ( A–H ) Unprocessed maximum intensity projections of co-stained images shown in Figure 5 ( see ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 021 We next examined projections of the interneuron types . For this , lines targeting LexA to either aBN2 or aDN were generated using the appropriate enhancers ( Supplementary file 1 ) . Both of these lines elicit antennal grooming with thermogenetic activation , and aDN-LexA targets three aDN neurons , while aBN2-LexA targets a cluster of four aBN2 neurons ( Figure 4—figure supplement 1A , C , D , Figure 4—figure supplement 2C–F ) . Projections of aBN1 and aBN2 were closely associated in both the AMMC and ventral SEZ but lacked associations in the posterior SEZ ( Figure 5E , arrows ) . Both aBN classes projected near the aDNs in the SEZ , but aDN1 also arborized near the dorsal projections of aBN2 ( Figure 5F–H , arrows ) . We conclude that the projections of these sensory neurons and interneurons are in close proximity with each other , and potentially form a functionally connected circuit . To test for functional connectivity among the neuronal classes , we measured calcium responses of the different interneurons when other neurons in the putative circuit were optogenetically activated . In isolated CNSs , we activated aJO+C/E neurons expressing the red light-inducible neuronal activator CsChrimson , while measuring fluorescence of the calcium responder GCaMP6s in the interneurons ( Chen et al . , 2013; Klapoetke et al . , 2014 ) . We detected significant calcium responses in aBN1 , aBN2 , and aDN1 ( Figure 6A–C , Figure 6—figure supplement 1A–D , Figure 6—figure supplement 2 , Figure 6—figure supplement 3 ) , but only a weak response in aDN2 even with high-intensity red light ( Figure 6D ) . Thus , aBN1 , aBN2 , aDN1 are likely downstream of aJO+C/E neurons , while aDN2 may be weakly or indirectly downstream of these sensory neurons . 10 . 7554/eLife . 08758 . 022Figure 6 . Different antennal grooming neurons are functionally connected . ( A–I ) Dissected CNSs with different neuronal classes expressing CsChrimson ( magenta ) were activated with red light while changes in calcium in their putative downstream partners expressing GCaMP6 ( green ) were imaged ( ΔF/F ) . Each tested neuronal pair is shown using circles and as traced pairs . The direction of the connection and whether it is excitatory or inhibitory is depicted with an arrow ( excitatory ) or ball and stick ( inhibitory ) . Changes in fluorescence of GCaMP6s of multiple flies under similar stimulus conditions are shown on the right ( average ± s . e . m . , 3–5 flies tested with 9–21 trials per trace ) . Arrow below each trace shows when the red light pulse was delivered . Black traces show flies that were imaged without drug treatment , whereas orange and blue traces were imaged while the nervous system was bathed with mecamylamine or picrotoxin respectively . See ‘Materials and methods’ , Figure 6—figure supplement 1 , Figure 6—figure supplement 2 , Figure 6—figure supplement 3 , and Supplementary file 2 for detailed ‘Materials and methods’ , stimulus conditions , and controls . ( A–D ) aJO-LexA tested with the following interneuron spGAL4 pairs: ( A ) aBN1-spGAL4-1 , ( B ) aBN2-spGAL4-1 , ( C ) aDN1-spGAL4-1 , and ( D ) aDN2-spGAL4-2 . ( E , F ) aBN2-LexA tested with aBN1-spGAL4-1 . ( G ) aBN1-spGAL4-1 tested with aDN-LexA . ( H , I ) aBN2-LexA tested with either ( H ) aDN1-spGAL4-1 or ( I ) aDN2-spGAL4-2 . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02210 . 7554/eLife . 08758 . 023Figure 6—figure supplement 1 . Functional connectivity: controls , technical details , and raw data . ( A ) Control experiments for CsChrimson/GCaMP6s activation . LexAop-CsChrimson was crossed with the control LexA driver , and GCaMP6s was expressed with aBN2-spGAL4-1 ( top ) or aDN2-spGAL4-2 ( bottom ) . Flies were tested and imaging results are displayed as described in Figure 5 . Average changes in fluorescence ±s . e . m . of multiple flies . ( B–D ) Representative examples of the regions imaged for ( B ) aBN1 , ( C ) aBN2 , and ( D ) aDNs . Top panel: example average projections of an experimental run . Middle panel: regions of interest used for analysis ( see ‘Materials and methods’ ) . Lower panel: black rectangles show the approximate positions of the fields of view in each whole pattern . Scale bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02310 . 7554/eLife . 08758 . 024Figure 6—figure supplement 2 . Raw data for functional connectivity experiments ( at low intensity red light ) . Raw data for experiments shown in Figure 6 . All experiments shown were done in the absence of drugs . Red light intensity was set at 50 μW/mm2 . Each column corresponds to the number of light pulses delivered , where each light pulse was 2 ms and the interpulse intervals were 18 ms . Each trace shown is the average of four responses recorded at ∼20 s intervals . Each row represents a different genotype and colored traces in a given row correspond to an individual CNS . Multiple runs were often performed for a given CNS and set of conditions ( shown by the same colored traces in a row ) . Black boxes show which conditions were used to generate the average traces displayed in Figure 6 . All traces within the black boxes were used to generate traces shown in Figure 6A–D . For Figure 6E–I , colored dots correspond to samples that were used to generate pre-drug treatment average traces and then used for the pharmacology experiments ( traces with drugs are not shown ) . Traces of samples used for mecamylamine experiments are marked using blue , green , red dots , whereas , those for or picrotoxin are marked using black , grey , orange dots . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02410 . 7554/eLife . 08758 . 025Figure 6—figure supplement 3 . Raw data for functional connectivity experiments ( at high intensity red light ) . Raw data for experiments shown in Figure 6 . All experiments shown were done in the absence of drugs . Red light intensity was set betwen 290 to 700 μW/mm2 . Each column corresponds to the number of light pulses delivered , where each light pulse was 2 ms and the interpulse intervals were 18 ms . Each trace shown is the average of four responses recorded at ∼20 s intervals . Each row represents a different genotype and colored traces in a given row correspond to an individual CNS . Multiple runs were sometimes performed for a given CNS and set of conditions ( shown by the same colored traces in a row ) . Black boxes show which conditions were used to generate the average traces displayed in Figure 6 . All traces within the black boxes were used to generate traces shown in Figure 6A–D . For Figure 6E–I , colored dots correspond to samples that were used to generate pre-drug treatment average traces and then used for the pharmacology experiments ( traces with drugs are not shown ) . Traces of samples used for mecamylamine experiments are marked using blue , green , red dots , whereas , those for or picrotoxin are marked using black , grey , orange dots . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 025 We next tested whether CsChrimson-mediated optogenetic activation of either aBN1 or aBN2 induced a calcium response in the other . Activation of aBN1 induced a calcium response in aBN2 ( Figure 6E ) , whereas only high-intensity red light activation of aBN2 could cause a weak calcium response in aBN1 ( Figure 6F ) . The latter response was inconsistent , as only three out of five flies tested showed increases in calcium in aBN1 ( average trace of three trials are shown in Figure 6F , raw traces of all five trials shown in Figure 6—figure supplement 3 ) . Further , because the aBN2-LexA driver used to express CsChrimson in aBN2 also expresses in neurons in other parts of the brain ( Figure 4—figure supplement 1C ) , we cannot rule out these other neurons as causing the aBN1 calcium response . Therefore , our results strongly support excitation from aBN1 to aBN2 , but we only find weak functional imaging-based evidence supporting the reverse connection . Nevertheless , the excitatory responses that we observed are cholinergic because they were suppressed by the cholinergic receptor antagonist mecamylamine ( Figure 6E , F ) . We next tested whether aBNs are upstream of the aDNs . Activation of aBN1 induced a calcium response in the aDNs , and this excitatory response is cholinergic as it was suppressed by mecamylamine ( Figure 6G ) . Interestingly , activation of aBN2 caused a mecamylamine-sensitive excitatory response in aDN1 ( Figure 6H ) , yet decreased the basal fluorescence of GCaMP6s in aDN2 in a manner that was alleviated by mecamylamine ( Figure 6I ) . The latter observation leads us to propose that aBN2 excites an unidentified neuron that then inhibits aDN2 . We tested this possibility by applying picrotoxin , a GABA and inhibitory glutamate receptor blocker ( Cleland , 1996; Liu and Wilson , 2013 ) . In the presence of picrotoxin , activation of aBN2 abolished the decreased calcium responses in aDN2 ( Figure 6I ) . This corroborated the presence of an inhibitory neuron ( IN ) that impinges on aDN2 that is stimulated by the activity of aBN2 . Thus , the aDNs are subject to both feedforward excitation and indirect inhibition from aBN2 . Taken together , these data demonstrate functional connectivity between the JO , aBNs , and aDNs and imply that they form a circuit that relays antennal sensory information through the brain and conveys it to the VNS . Our results motivated a circuit model that depicts the functional connectivity among the JO neurons , aBNs , and aDNs ( Figure 7A , B ) . The model includes putative reciprocal excitation between the aBNs ( Figure 5E , Figure 6E , F ) and feedforward inhibition of aDN2 mediated by aBN2 and an unknown IN ( Figure 6I ) . The proposed outputs of this circuit are the parallel descending commands to the VNS ( Figure 2 ) , where the antennal grooming pattern-generating circuitry is expected to reside ( see ‘Discussion’ ) . We propose that the aDNs can act in parallel because experiments to thermogenetically activate aDN1 or aDN2 alone indicate that they are each sufficient to induce antennal grooming ( Figure 2B ) . The presence of these features within the circuit suggests complex processing and raises the question of whether they influence the grooming output . Given that aBNs and aDNs induce different amounts of grooming with thermogenetic activation ( Figure 2A , B ) , we postulated that one feature of the circuit is to control the duration of antennal grooming . 10 . 7554/eLife . 08758 . 026Figure 7 . A circuit whose components elicit different antennal grooming durations . ( A ) The antennal grooming circuit ( lateral view of tracings ) . Specific colors represent each neuron type shown in B . ( B ) Wiring diagram of the circuit . Lightning bolt represents mechanical stimulation of the antennae . Arrows represent excitatory cholinergic functional connections and the ball and stick indicates an inhibitory ( picrotoxin sensitive ) connection from an unidentified inhibitory neuron ( IN ) . Note: JO neurons were previously reported to be cholinergic ( Yasuyama and Salvaterra , 1999; Salvaterra and Kitamoto , 2001 ) . Arrow to the gray oval surrounding the aDNs indicates that aBN1 provides excitatory input for aDN , but it is not known for which aDN ( s ) . Gray dashed arrows indicate relatively weak and/or inconsistent connections ( JO to aDN2 and aBN2 to aBN1 ) . Text on the left highlights putative circuit connectivity features . Dashed arrow from aDN3 depicts presumed descending command . ( C ) The different neuronal classes induce distinct grooming responses . Ethograms of manually scored video showing antennal grooming induced with red light sensitive CsChrimson expressed in different spGAL4 pairs ( aJO-spGAL4-1 , aBN1-spGAL4-1 , aBN2-spGAL4-1 , aDN1-spGAL4-1 , and aDN2-spGAL4-2 ) . Ethograms of individual flies are stacked on top of each other . The gray bars indicate presentation of red light . Colors correspond to the wiring diagram ( B ) and indicate which neuronal class expressed csChrimson . Control flies did not perform antennal grooming ( Figure 7—figure supplement 1 ) . See Video 4 , Video 5 , and Video 6 for representative examples . ( D ) Histograms representing the fraction of flies that were performing antennal grooming in C within one-second time bins . ( E ) The proposed organization of antennal grooming circuitry . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02610 . 7554/eLife . 08758 . 027Figure 7—figure supplement 1 . CsChrimson activation of different neuronal classes . ( A ) Stacked ethograms of grooming movements performed by the spGAL4 pairs indicated expressing CsChrimson ( 10–15 flies shown for each spGAL4 pair ) . Gray bars indicate when the red light was on . Colors correspond to grooming movements performed . ( B ) Percent of total time that each spGAL4 line spent grooming their antennae . Box plots and statistics are described in Figure 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 027 Our results raise the possibility that aBN1 and aBN2 are reciprocally excitatory , although the aBN2 to aBN1 excitatory connection was only weakly supported . One prediction of reciprocal excitation between the aBNs is that activation of either neuronal class will induce grooming that persists in the absence of stimulation ( Major and Tank , 2004; Li et al . , 2006 ) . To test this , we exploited CsChrimson's tight temporal control to activate these neurons briefly with red light , and then examined the dynamics of the grooming responses after red light cessation . Activation of either aBN1 or aBN2 elicited long antennal grooming durations that could persist for tens of seconds beyond red light cessation ( Figure 7C , D , Figure 7—figure supplement 1A , Video 4 ) . In contrast , activation of the aDNs , which are downstream of potential aBN-induced reciprocal excitation , should not induce persistent grooming ( Figure 7B ) . Indeed , activation of aDN1 or aDN2 caused grooming that did not outlast the red light activation ( Figure 7C , D , Figure 7—figure supplement 1A , Video 5 ) . We observed similar results using additional spGAL4 pairs that target aBN1 , aBN2 , and aDN2 ( not shown ) . Of note , activation of aDN2-elicited antennal grooming that lasted until the stimulus ended ( Figure 7C , D ) , whereas activation of aDN1 elicited about 50% fewer grooming bouts that terminated prior to the red light turning off ( Figure 7C , D , Figure 7—figure supplement 1A , B ) . This raises the intriguing possibility that the parallel aDNs induce different durations of antennal grooming . We next found that red light activation of the aJO , which is upstream of the aBNs , could also elicit grooming that persisted after red light cessation ( Figure 7C , D , Video 6 ) . This suggests that the aJO might excite the aBNs , which then induce persistent grooming . Taken together , our data indicate that the circuit produces persistent antennal grooming through aBN1 and aBN2 . Future experiments will test whether the persistence is caused by reciprocal excitation or alternate mechanisms ( see ‘Discussion’ ) . We concluded that our data describe an antennal grooming circuit whose components have the potential to modulate the duration of grooming , possibly mediated by their connectivity . 10 . 7554/eLife . 08758 . 028Video 4 . Grooming in response to red light stimulation of CsChrimson-expressing aBN1 neurons . CsChrimson was expressed in aBN1 using aBN1-spGAL4-1 . The infrared light in the bottom right hand corner shows when the red light was on to activate aBN1 . Note that grooming persists upon cessation of the red light . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02810 . 7554/eLife . 08758 . 029Video 5 . Grooming in response to red light stimulation of CsChrimson-expressing aDN2 neurons . CsChrimson was expressed in aDN2 using aDN2-spGAL4-2 . The infrared light in the bottom right hand corner shows when the red light was on to activate aDN2 . Note that grooming does not persist upon cessation of the red light . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 02910 . 7554/eLife . 08758 . 030Video 6 . Grooming in response to red light stimulation of CsChrimson-expressing aJO neurons . CsChrimson was expressed in the aJO using aJO-spGAL4-1 . The infrared light in the bottom right hand corner shows when the red light was on to activate the aJO . Note that grooming persists upon cessation of the red light . DOI: http://dx . doi . org/10 . 7554/eLife . 08758 . 030
This work provides the first description of a neural circuit that evokes a grooming movement: antennal displacements are detected by specific mechanosensory neurons that excite different interneuronal classes to elicit the front leg movements that constitute antennal grooming . We specifically identified JO neurons that project from the second antennal segment into the ventral brain , two brain interneuron classes ( aBNs ) that project within this ventral brain region , and descending neurons ( aDNs ) that project both within the ventral brain and descend to the VNS ( Figure 7A , B , E ) . These neuronal classes are functionally connected with the common purpose of specifically inducing antennal grooming . How does activation of this circuit elicit grooming of the antennae ? Grooming movements are ultimately produced by pattern generators in the VNS that control the front and hind legs ( Berkowitz and Laurent , 1996; Burrows , 1996 ) . The aDNs project to the region of the VNS that should control front leg grooming movements , the ProNm . This suggests that the aDNs might connect to and activate neural circuits that generate coordinated antennal grooming movements ( Figure 7B , E ) , but such circuitry remains to be identified . The tools for targeting aDNs acquired here may provide a means to identify this antennal grooming circuitry by examining their synaptic partners . Identification of this circuitry will provide a valuable example of how nervous systems achieve such exquisite specificity in targeting grooming movements to the site of a stimulus . In this work , we examined a role for the identified circuit in detecting displacements of the antennae and eliciting grooming . This was based both on our discovery that the JO induces grooming , and that different JO neurons can detect antennal movements to elicit specific behavioral responses ( Kamikouchi et al . , 2009; Yorozu et al . , 2009; Matsuo et al . , 2014 ) . Further , insects can encounter conditions in their natural environments , such as static electricity or unexpected mechanical disruptions that can move the antennae and cause different aversive behavioral responses ( Hunt et al . , 2005; Newland et al . , 2008; Jackson et al . , 2011; Matsuo et al . , 2014 ) . We have observed that flies show increased grooming of their antennae in chambers with high levels of static , and that this grooming was ablated using antistatic agents ( unpublished observations ) . This led us to find that imposed deflections of the antennae induce grooming through the JO . Because such deflections likely disrupt normal antennal functions , such as hearing ( Göpfert and Robert , 2002 ) , grooming may provide a means of restoring them to their proper position and function . Another function of grooming is to remove debris , parasites , or other substances from the body surface ( Sachs , 1988; Böröczky et al . , 2013; Zhukovskaya et al . , 2013 ) . We have shown that dust on the antennae induces grooming in flies ( Seeds et al . , 2014 ) , however , it seems unlikely that small particles of dust could cause the large displacements that induce grooming in this work . This is supported by our observations that expression of TNT in the aJO does not cause defects in antennal dust removal ( unpublished data ) , but does disrupt the grooming response to antennal displacements ( Figure 4E ) . Therefore , dust might be detected by different sensory neurons , such as the mechanosensory bristle neurons found on the surface of the antennae . Bristle neurons are good candidates for stimulating grooming , given that mechanical stimulation of bristles on other body parts can induce site directed grooming responses ( Vandervorst and Ghysen , 1980; Corfas and Dudai , 1989 ) . Like JO neurons , the antennal bristle mechanosensory neurons send projections to the AMMC , and possibly to the SEZ ( Homberg et al . , 1989; Melzig et al . , 1996 ) , where they could induce grooming by interacting with the interneurons described in this work . This raises the possibility that multiple different sensory neuron types feed into the circuit to induce grooming . However , we have not yet identified spGAL4 lines that specifically target the antennal bristle mechanosensory neurons to allow us to test this hypothesis . A scenario whereby different types of stimuli , such as antennal deflections or dust can cause grooming could explain some of the complexity of the circuit that we observe . Our data suggest multiple possible routes of feedforward excitation through the circuit layers to induce antennal grooming . For example , the function of at least a subset of the aBN2 neurons was not necessary for grooming in response to physical displacement of the antennae , whereas they were necessary in response to thermogenetic activation of the JO . Thus , feedforward excitation likely has multiple routes through the antennal grooming circuit , but why would this be ? One possibility is that each route reflects specific features of the sensory stimulus . Different stimuli might be sensed by distinct JO neurons or mechanosensory bristle neurons on the antennae that engage different ensembles of neurons within the circuit to produce specific grooming responses . Therefore , it may be revealing to test whether different stimuli coopt specific circuit components to fine–tune parameters of grooming , such as short vs persistent grooming durations . The finding that the antennal grooming circuit can induce persistent grooming raises the question as to its biological role . Our results do not reveal a natural stimulus that induces persistence , as displacements of the antennae rarely induced persistent grooming ( Figure 4—figure supplement 3 ) . Persistence was previously described in the grooming responses of other animals , where it was found to be at least partially dependent on the nature of the stimulus ( e . g . , stimulus strength , duration , or frequency ) ( Sherrington , 1906; Stein , 2005 ) . Thus , future work will explore the possibility that specific stimulus parameters can evoke persistent antennal grooming in flies . Additionally , persistence indicates the presence of a mechanism that produces long lasting neural activity within the grooming sensorimotor response . Such persistent activity could provide a mechanism by which temporal summation of successive stimuli is achieved . That is , successive , sub-threshold stimuli have been shown to be ‘remembered’ or summed to elicit grooming , although the neuronal mechanisms have not been identified ( Sherrington , 1906; Stein , 2005; Guzulaitis et al . , 2013 ) . Therefore , future experiments will examine the role of specific features of the antennal grooming circuit in producing persistent grooming , and possible temporal summation ( Figure 7B , possible circuit features discussed below ) . aJO mechanosensory neurons are critical for the grooming response to antennal movements and sufficient to induce grooming . This reveals a new role for the JO , which was previously implicated in hearing , gravitaxis , and wind-induced suppression of locomotion ( Kamikouchi et al . , 2009; Yorozu et al . , 2009 ) . These behaviors are mediated by JO neurons that are localized to distinct regions of the mechanosensory structure and respond to high frequency vibrations ( e . g . , sounds ) or more tonic movements ( e . g . , wind ) ( Kamikouchi et al . , 2006; Kamikouchi et al . , 2009; Yorozu et al . , 2009; Matsuo et al . , 2014 ) . Thus , the JO is functionally organized such that different groups of mechanosensory neurons mediate different behaviors . Our finding that the JO elicits antennal grooming supports this idea , and demonstrates an even greater diversity of JO functions than previously thought . Our data indicate that JO-induced antennal grooming is likely restricted to neurons projecting to zone C/E . Previously characterized JO neurons projecting to this zone are less sensitive to smaller movements of the antennae and adapt slowly to mechanical stimuli ( Kamikouchi et al . , 2009; Yorozu et al . , 2009; Matsuo et al . , 2014 ) . The antennal displacement distance that induces grooming is within the range shown to induce calcium responses in previously described zone C/E neurons ( Yorozu et al . , 2009 ) . This suggests that aJO neurons might also be activated by such large displacements , given that they are critical for the subsequent grooming response and project to zone C/E . Two different populations of C/E-projecting neurons can induce antennal grooming; one that projects only to the AMMC , and the other ( aJO ) that projects to the AMMC and SEZ . This raises the question of whether both populations interact with the antennal grooming circuitry described in this work . The aBNs and aDN1 have neurites within the AMMC where they could plausibly receive excitatory inputs from both C/E neuron populations . Additionally , because neurons within zone C/E are implicated in other behaviors like wind-induced suppression of locomotion , it remains to be determined whether these JO neurons are multifunctional , or whether specific subpopulations within this group are responsible for distinct behaviors . Future work will be required to determine how these two populations of JO sensory neurons interact with the antennal grooming circuitry . The antennal grooming circuitry consists of at least three different neuronal classes . As our extensive screening efforts identified multiple GAL4 lines for each class , we presume that we have uncovered a major portion of the neurons that elicit antennal grooming . Indeed , these neurons are sufficient to form a functionally connected circuit that extends from the JO to the VNS . However , additional descending neurons are likely involved given that TNT expression in the aDNs failed to disrupt the grooming response to aJO activation . We also found evidence for an unidentified neuron downstream of aBN2 that inhibits aDN2 . The neuronal classes in the circuit are each sufficient to elicit antennal grooming , similar to command-like neurons that evoke specific movements ( also termed decision neurons or higher order neurons ) ( Kupfermann and Weiss , 1978; Pearson , 1993; Kristan , 2008; Jing , 2009 ) . aBN1 and a subset of aBN2 neurons could potentially be more specifically termed command neurons , which are defined as being necessary and sufficient for initiating a specific movement , and fire in response to the movement-initiating sensory input ( Kupfermann and Weiss , 1978 ) . However , in order to definitively call these neurons command neurons , we need to test whether they are active in response to imposed movements of the antennae . Another term to describe collections of neurons that induce a behavior is a command system ( Kupfermann and Weiss , 1978; Jing , 2009 ) . However , a command system does not necessarily consist of functionally connected neurons . Therefore , given that these neurons constitute a functionally connected circuit , it may be appropriate to refer to them collectively as a command circuit for antennal grooming . When the antennal grooming command circuit is compared with those that were previously identified in other systems , it emerges that circuits can consist of different layers that each elicit the specific movement , but with different durations ( Frost et al . , 2001; Kristan et al . , 2005; Pirri and Alkema , 2012 ) . In such cases , neurons have been found within particular layers that either elicit a movement that persists beyond their initial activation , or must be continually activated to continue the movement . These neuronal types have been termed trigger and gating neurons respectively ( Stein , 1978 ) . In the marine mollusc ( Tritonia diomedia ) and in the leech ( Hirudo medicinalis ) , trigger neurons immediately downstream of sensory neurons induce persistent swimming motor patterns ( named Tr1 in both animals ) ( Frost et al . , 2001; Kristan et al . , 2005 ) . Gating neurons then induce swimming downstream of Tr1 in both animals ( named DRI in the mollusc and 204/205 in the leech ) . This organization is strikingly similar to the layers of neurons that induce antennal grooming identified here; the aBNs induce persistent grooming and might be considered trigger neurons , whereas aDN2 could serve as a gating neuron . Taken together , the use of trigger and gating neurons might constitute a common organization in circuits that command specific movements downstream of a sensory stimulus . The layered and complex organization of the antennal grooming circuit raises the question as to its function . We discussed above the possibility that the circuit may have different possible routes of feedforward excitation . Such organization may provide different points at which the circuit can be modulated , thus allowing for flexible control of different movement parameters . In the antennal grooming circuit , flexible control of movement duration could be provided by the parallel aDNs ( Figure 7B ) , which appear to elicit different amounts of antennal grooming when activated . Artificially activated aDN1 elicits isolated grooming bouts that terminate despite continued activation , whereas activated aDN2 elicits grooming throughout the activation period . These observations are reminiscent of two descending neurons in molluscs that initiate biting movements during feeding , as one induces longer protraction durations whereas the other induces shorter durations ( Jing and Weiss , 2005 ) . This ability to generate distinct movement parameters , such as duration , may be a general feature of descending neurons that induce the same movement . Other parameters that might be differentially controlled are illustrated by locomotor systems , wherein descending neurons can elicit movement while controlling parameters such as speed and direction ( Dubuc et al . , 2008; Roberts et al . , 2010; Mullins et al . , 2011; Puhl et al . , 2012; El Manira and Grillner , 2014 ) . The aDNs are also impinged upon by feedforward excitation and inhibition ( Figure 7B ) , possibly to control which ones are active . For example , aBN2 provides differential control by exciting aDN1 and inhibiting aDN2 . As there are multiple aDNs , it may be that feedforward excitation could impinge on them at the same time to induce hybrid durations of antennal grooming . Such an effect has been described for the two descending neurons controlling mollusc bite protraction , as activating the two together produces intermediate durations ( Jing and Weiss , 2005 ) . Thus , we will further test whether the aDNs similarly produce flexible durations of antennal grooming . The persistent grooming induced by brief excitation of the aBNs provides another mechanism for controlling grooming duration ( Figure 7B ) . This is reminiscent of previously described grooming responses that were sustained despite stimuli cessation , which Sherrington referred to as afterdischarge ( Sherrington , 1906; Stein , 2005 ) . Persistence is not limited to grooming responses , as it has been described in behaviors as diverse as locomotion and courtship song ( Stein , 1978; Kristan et al . , 2005; Inagaki et al . , 2014; Gao et al . , 2015 ) . In locomotion , reciprocal excitation among reticulospinal neurons has even been implicated in persistent tactile-induced swimming ( Li et al . , 2006 ) . Although our work only provides weak evidence for reciprocal excitation between the aBNs , a mechanism whereby reciprocal connections maintain their excitation offers a plausible explanation for the persistence that we observe . Alternatively , aBN1 and aBN2 could have intrinsic membrane properties that allow them to produce prolonged responses to brief excitation ( Major and Tank , 2004 ) . Thus , future experiments will examine whether reciprocal excitation or intrinsic membrane properties of the aBNs produce persistent grooming . In this section we have discussed how the complex organization of the antennal grooming circuit might control movement duration , however , it could also control parameters that our current analysis methods cannot detect . The level of behavioral analysis presented in this work reveals that the induced trajectories of the legs are specific for the antennae rather than other head parts; however , it does not allow for the detection of finer antennal grooming movements . Grooming is characterized by an initial targeting of the legs to the stimulated body part , followed by cyclic movements that groom the region ( Dürr and Matheson , 2003 ) . Higher resolution analysis of the leg kinematics could help resolve the boundary limits on the head to which the antennal grooming movements are confined during these two phases , and allow for determining how stereotyped these movements are . Furthermore , such analysis could elucidate how the legs interact with the antennal region to perform the grooming movements . Thus , higher resolution analysis would facilitate testing possible roles for the circuit in controlling different variables of antennal grooming , such as the limb trajectories or speed . Recent work indicates that command-like neurons may constitute a common means of eliciting specific movements in insects . Tools that allow for both acute control of neuronal activity and precise genetic targeting of specific neurons in fruit flies have enabled experiments that merge behavioral analyses with real time neuronal manipulations . Such experiments demonstrate that activation of specific neurons can elicit distinct movements , such as escape , locomotion , or courtship song ( Lima and Miesenböck , 2005; von Philipsborn et al . , 2011; Gao et al . , 2013; Flood et al . , 2013b; Inagaki et al . , 2014; Bidaye et al . , 2014; von Reyn et al . , 2014 ) . Additionally , work in crickets , grasshoppers , and locusts has revealed specific neurons that elicit courtship stridulation or flight ( Pearson et al . , 1985; Hedwig , 1994 , 1996 , 2000 ) . Given our discovery of a circuit that specifically elicits antennal grooming , it would appear that the use of dedicated neurons to command specific movements is a common mechanism of behavioral control in insects . However , these command-like neurons may be embedded in larger , more dynamic neural networks in which neurons that elicit one movement also participate in other movements ( Kupfermann and Weiss , 1978; Kristan , 2008 ) . For example , in the leech several of the neurons involved in commanding swimming were found to also be excited during stimulus-induced shortening movements ( Shaw and Kristan , 1997 ) . Such findings indicate that neurons that can command one movement might also participate in the production of other movements , with their specific output subject to the collective activities of multiple different neurons within a given network ( Kristan et al . , 2005; Kristan , 2008 ) . Thus , future experiments that examine whether the antennal grooming command circuit participates in controlling additional movements , may further elucidate the properties of neural networks that enable the performance of specific movements .
Flies were reared on cornmeal and molasses food at 21°C and 50% relative humidity on a 16/8 hr light/dark cycle . 5–8 day old males were used for all experiments , except for those in Figure 6 that were performed with 2–8 day-old flies . Stocks used in this study are listed in Supplementary file 1 and Supplementary file 3 . Three GAL4 lines that induced antennal grooming with thermogenetic activation using dTrpA1 were identified by screening over 1500 randomly selected GAL4 lines ( R39A11 , R26B12 , and R18C11-GAL4 ) ( Seeds et al . , 2014 ) . To identify additional lines , we visually screened through an image database of GAL4 expression patterns ( Jenett et al . , 2012 ) for those with expression in neurites close to aJO projections . Selected lines were crossed to UAS-dTrpA1 and screened for increased antennal grooming at 30–32°C . For pattern refinement and co-expression studies , the enhancers of GAL4 lines that exhibited increased antennal grooming were used to generate spGAL4 and LexA reagents , which were constructed as described previously ( Pfeiffer et al . , 2010 ) and produced by Gerald Rubin's lab . DBDs were inserted into the attP2 landing site ( on chromosome 3 ) , ADs were inserted into attP40 ( on chromosome 2 ) , and LexAs were inserted into attP40 . Supplementary file 1 lists the enhancer identities ( Jenett et al . , 2012 ) used to generate the spGAL4 , LexA , and GAL4 lines . Control flies contain the DNA elements used for generating the different GAL4 , spGAL4 halves , or LexA collections , but lack enhancers to drive their expression ( images of control lines crossed to UAS-GFP are shown in Figure 1—figure supplement 1E , F ) . R27H08 ( aJO-LexA ) was identified in a screen of existing LexA lines ( Pfeiffer et al . , 2010 ) that targeted the aJO and could induce antennal grooming ( Figure 4—figure supplement 1A , B ) . The camera setup and methods for recording the behavior of flies expressing dTrpA1 in different neuronal classes were described previously ( Seeds et al . , 2014 ) . Behavior was recorded at 35 frames per second for 2 min at 30–31°C . For amputation experiments , the entire antennae of 2 day-old males were severed with forceps and flies were allowed to recover for 4 days . The amputation did not damage the rest of the head or lead to mortality of the animals . CsChrimson experiments were performed in the dark , and flies were visualized for recording using an 850-nm infrared light source at 2 mW/cm2 intensity ( Mightex , Toronto , CA ) , which flies cannot see . For CsChrimson activation , we used 656-nm red light at 27 mW/cm2 intensity ( Mightex ) . The red light stimulus parameters were delivered using a NIDAQ board controlled through Labview ( National Instruments , Austin , TX ) . Red light frequency was 5 Hz for 5 s ( 0 . 1-s on/off ) , and 30-s interstimulus intervals ( total of 3 stimulations ) . Grooming movements were manually scored as previously described ( Seeds et al . , 2014 ) , with the exceptions listed below . Manual scoring of prerecorded video was performed with VCode software and the data was analyzed in MATLAB ( MathWorks Incorporated , Natick , MA ) . Modifications of previously described scoring of grooming movements: Antennal grooming: The legs grasp and brush the antennae , with the head often tilted forward . When the antennae were amputated , ‘antennal grooming’ was scored when the legs were directed towards the area where antennae used to be . Proboscis grooming: The legs sweep down the proboscis when it is extended , or the tip when it is retracted . Ventral head: Legs sweep the area between the antennae and the ventral side of the head , as well as below the eyes towards the ventral bottom of the head . Leg rubbing: This movement was not included in this study; however , leg rubbing was often associated with the activated movements . Movement start times were scored one frame before the specific body part was first touched and ended two frames after that body part was last touched . The time interval between the previous and the next movement was scored as standing . Third antennal segments were coated with a mixture of iron powder ( Atlantic Equipment Engineers , Upper Saddle River , NJ , 325 mesh ) and UV cured glue ( Kemxert Corp . , York , PA ) and then left to recover for 12 hr . The presence of the powder on the antennae did not cause increased antennal grooming after the recovery period . Flies were tethered using a pin that was glued to their thoraces and then positioned on a 6 mm diameter air-supported ball ( Seelig et al . , 2010 ) within a custom made electromagnet ( Figure 4C , D , Supplementary file 4 ) . The automated stimulation parameters of the electromagnet ( frequency , on/off durations ) were delivered via a NIDAQ board controlled through MATLAB . Voltage was controlled using a Power Supply ( B&K Precision Corporation , Yorba Linda , CA ) . The magnetic field was applied to move the antennae while the induced movements were recorded . Movement responses were manually scored as described above . The percent time the flies spent grooming their antennae was calculated through the duration of the experiment . The stimulus parameters used for our experiments were determined using control flies ( w1118; UAS-TNT; pBPGAL4U ) . The percent time that control flies groomed when the magnetic field was applied was measured at different frequencies and voltages to determine the optimal stimulus conditions ( Figure 4—figure supplement 3A , B ) . Flies performed an intermediate amount of grooming at 1 Hz and 10 V . Based on these results , antennal deflection experiments were performed at 1 Hz for 10 s ( 0 . 5 s on/off ) at 10 V , with a 30 s wait time between stimulations ( total of 4 stimulations ) . The magnetic field strength at 10 V was measured at 570 Gauss . Grooming was not induced with flies that lacked iron powder on their antennae and were exposed to the magnetic field ( data not shown ) . The antennal displacement caused by the magnetic field was measured by recording video of the antenna from a side view ( Figure 4—figure supplement 3C ) . Antennal displacement was calculated by measuring movement of the distal tip of the third antennal segment before and after the magnetic field was applied . Pixels were calibrated to physical distances using a known standard . Behavioral data was analyzed with nonparametric statistical tests . First , we performed a Kruskal–Wallis ( ANOVA ) test to compare more than three genotypes with each other . Next we performed a post-hoc Mann–Whitney U test and applied Bonferroni correction . Dissection and staining was performed as previously described ( Hampel et al . , 2011 ) , with modifications listed below . 5–8 day-old males were used for all dissections , except for the stochastic labeling experiments using MCFO-1 ( Nern et al . , 2015 ) in Figure 1—figure supplement 4A–H , where 1–2 day-old flies were dissected . For Figure 1E , F , Figure 2C–F , Figure 3A–D , Figure 5A–H , Figure 1—figure supplement 1F , Figure 1—figure supplement 2A–E , Figure 2—figure supplement 2A–I , Figure 4—figure supplement 1B–D , Figure 4—figure supplement 2A–F , and Figure 5—figure supplement 2A–H additional treatment was performed to clear the tissue: After immunohistochemistry , tissue samples were post-fixed in 4% paraformaldehyde in PBS for 4 hr at room temperature , followed by four 30 min washes in PBT . Before mounting the CNS on a poly-L-lysine ( P1524; Sigma , St . Louis , MO ) coated cover slip , the tissues were washed in PBS for 15 min to remove Triton . After mounting the tissues on the poly-L-lysine-coated cover slip , they were dehydrated through a series of ethanol dilutions ( 30% , 50% , 75% , 95% , and 3 × 100% ) for 10 min each , followed by an incubation series in 100% xylene ( Fisher Scientific , Fair Lawn , NJ ) three times for 5 min each in Coplin jars . Afterwards tissues were embedded in DPX , a xylene-based mounting solution ( Electron Microscopy Sciences , Hatfield , PA , Cat#13512 ) and allowed to dry for 48 hr before imaging . For Figure 2H , Figure 1—figure supplement 1C–E , G–J , Figure 1—figure supplement 3A–G , Figure 1—figure supplement 4A–H , Figure 2—figure supplement 1B–H , and Figure 5—figure supplement 1A–I stained tissue samples were mounted after the immunohistochemistry in Vectashield ( Vector Laboratories , Inc . Burlingame , California ) and allowed to incubate for an hour before imaging . Antibodies used: rabbit anti-GFP ( 1:500 , Thermo Fisher Scientific , Waltham , MA , #A11122 ) , chicken anti-GFP ( 1:2000 , Abcam , Cambridge , MA , #ab13970 ) , mouse anti-GFP ( 1:200 , Sigma , #G6539 ) , mouse mAb anti-nc82 ( 1:50 , Developmental Studies Hybridoma Bank , University of Iowa ) , rat anti-DN-cadherin ( 1:20 , Developmental Studies Hybridoma Bank ) , rabbit anti-RFP ( to detect tdTomato; 1:1000 , Clontech Laboratories , Inc . , Mountain View , CA , #632496 ) , rat anti-flag ( Novus Biologicals , LLC , Littleton , CO , #NBP1-06712 ) , rabbit anti-HA ( Cell Signaling Technology , Danvers , MA , #3724S ) , mouse anti-V5 ( AbD Serotec , Kidlington , England , #MCA1360 ) , AlexaFluor-488 ( 1:500; goat anti-rabbit , goat anti-chicken , goat anti-mouse; Thermo Fisher Scientific ) , AlexaFluor-568 ( 1:500; goat anti-mouse , goat anti-rat; Thermo Fisher Scientific ) , AlexaFluor-633 ( 1:500; goat anti-rat; Thermo Fisher Scientific ) . Confocal stacks of stained CNS and antennae were imaged on a Zeiss LSM710 confocal microscope with a Plan-Apochromat 20×/0 . 8 M27 objective and a Plan-Apochromat 63×/1 . 4 oil immersion objective . To visualize the neuronal classes together as shown in Video 2 , confocal images of different split-GAL4 lines were computationally aligned from individual specimens to one brain sample of our own collection with the Computations Morphometry Toolkit CMTK ( https://www . nitrc . org/projects/cmtk/ ) ( Jefferis et al . , 2007 ) and assembled in FluoRender , ( Wan et al . , 2009 ) a suite of tools for viewing and analyzing image data . Image preparation , analysis of overlap , and adjustment of brightness and contrast were performed with Fiji ( http://fiji . sc/ ) . Figure 5 shows maximum projections of confocal stacks that were modified using the 3D viewer plugin in Fiji to crop contaminating neurons and background noise . The same images are shown in Figure 5—figure supplement 2A–H with only brightness and contrast adjusted . Confocal stacks of brains imaged with a 63× objective were used to reconstruct each neuronal class shown in Figure 3E . We traced each neuron from different brains with neuTube software ( Feng et al . , 2015 ) and assembled the neuronal circuit manually . Male flies were dissected in saline containing 103 mM NaCl , 3 mM KCl , 5 mM TES , 8 mM trehalose dihydrate , 10 mM glucose , 26 mM NaHCO3 , 1 mM NaH2PO4 , 2 mM CaCl2 , 4 mM MgCl2 , and bubbled with carbogen . The brain and VNS were placed on a poly-lysine-coated coverslip . Dissections were visualized using minimal illumination to avoid activation of CsChrimson . The preparation was continuously perfused in saline at 60 ml/hr . Imaging was done using a two-photon scanning microscope ( Bruker , Billerica , MA ) with an excitation wavelength of 920 nm . Imaging fields of view were chosen as to contain a distinctive process of the candidate post-synaptic neuron , which included processes at the midline for the aDNs and aBN2 , and vertical running processes close to the midline for aBN1 . CsChrimson was excited with 2 ms pulses of 590-nm light via an LED shining through the objective . Instantaneous powers measured out of the objective ranged between 50 μW/mm2 and 800 μW/mm2 . Trains were delivered at 50 Hz . Experiments usually started with a 50-pulse train . If no response was observed , the power was raised progressively until one occurred or the maximum power was reached . Each experimental run consisted of 4 repeats lasting for approximately 20 s . Runs were repeated approximately every 2 min . When postsynaptic responses were observed , we did not see any desensitization . See Supplementary file 2 for stimulus conditions used for data shown in Figure 6 . For blocking nicotinic or GABAergic/glutamatergic transmission , mecamylamine ( 50 μM ) or picrotoxin ( 30 μM ) were administered through a perfusion line for 3–10 min , followed by a drug-free wash ( drugs from Sigma ) . Pharmacology experiments were run using samples that showed consistent responses . Stimulus conditions were chosen that elicited reliable transient responses . Average traces shown in Figure 6E–I are from those trials where pharmacology experiments were done ( trials used to generate average traces shown in Figure 6—figure supplements 2 , 3 ) . Average traces shown in Figure 6A–D correspond to all flies and trials at the indicated stimulus conditions shown in Figure 6—figure supplements 2 , 3 . ΔF/F0 was calculated ( F0 is the average signal before the stimulation ) for each video in a region of interest ( ROI ) obtained as follows: the average projection of the video was calculated and pixels of the projection were clustered by a k-means algorithm between ROI and non-ROI pixels . Of note , the selection method relies only on average intensity and not activity , because we wanted to use the same detection method for responsive and non-responsive runs . This also relies on selecting fields of view that unambiguously contain only the neuron of interest . Analysis was run in Julia ( http://julialang . org/ ) . | Many movements that animals perform regularly—including walking and grooming—consist of stereotyped sequences of muscle contractions . For example , a dog may scratch its side in response to a fleabite or because it is itchy . But how does the nervous system trigger such specific movements from among the repertoire of different movements that the animal could perform ? It also remains unclear how such movements can be produced in a reliable , yet flexible manner . Hampel et al . have now described the neural circuit that triggers and controls the stereotyped leg movements that the fruit fly Drosophila uses to groom its antennae . Such grooming movements are stereotyped yet have certain degrees of flexibility , which makes them ideal to study the neural circuits that underlie specific movements . Grooming further lends itself to this kind of investigation because it can be triggered by irritating the surface of the fly's body . There is also an extensive genetic toolkit that can be used to manipulate and observe the fruit fly's nervous system in detail . Hampel et al . first identified sensory neurons in the flies' antennae that were needed to elicit grooming in response to irritating displacements of the antennae . Once these neurons were found , techniques—including those that allow specific neurons in the fly's brain to be precisely controlled—were then used to find other neurons that participate in the grooming process . This approach highlighted three groups of interneurons: two in the brain and one in the fly's equivalent of the spinal cord . Together these layers of sensory neurons and interneurons formed a circuit that triggered grooming whenever the antennae were disturbed . Notably , activating different sensory or interneurons triggered bouts of antennal grooming of differing durations . This shows that the same neural circuit can both produce highly specific movements and modify the movements to provide flexibility . Neural circuits with similar features have been observed previously to induce other animal behaviors , for example , swimming in leeches . This suggests that this organization may be common in circuits that elicit movements . Additional experiments are now needed to validate whether similar circuits underlie other stereotyped movements in fruit flies and other animals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | A neural command circuit for grooming movement control |
How single cells in a mitotic tissue progressively acquire hallmarks of cancer is poorly understood . We exploited mitotic recombination in developing Drosophila imaginal tissues to analyze the behavior of cells devoid of the tumor suppressor PTEN , a negative regulator of PI3K signaling , under varying nutritional conditions . Cells lacking PTEN strongly overproliferated specifically in nutrient restricted larvae . Although the PTEN mutant cells were sensitive to starvation , they successfully competed with neighboring cells by autonomous and non-autonomous mechanisms distinct from cell competition . The overgrowth was strictly dependent on the activity of the downstream components Akt/PKB and TORC1 , and a reduction in amino acid uptake by reducing the levels of the amino acid transporter Slimfast caused clones of PTEN mutant cells to collapse . Our findings demonstrate how limiting nutritional conditions impact on cells lacking the tumor suppressor PTEN to cause hyperplastic overgrowth .
Clinically detectable cancer cells carry a multitude of mutations and chromosomal aberrations , and they display an enormous genetic heterogeneity ( Salk et al . , 2010; Wong et al . , 2011; Brosnan and Iacobuzio-Donahue , 2012; Marusyk et al . , 2012; Turner and Reis-Filho , 2012 ) . It is therefore desirable to target earlier tumorigenic stages but we know comparatively little about how pre-cancerous cells progressively develop into tumors ( Moreno , 2008 ) . The model system Drosophila allows analyzing the behavior of cells lacking particular tumor suppressor functions . During the growth phase ( larval instars ) , the cells of the imaginal discs ( that will eventually give rise to adult appendages ) remain diploid and proliferate until the discs have reached an appropriate size . The simple architecture of the imaginal discs ( the disc proper consists of a single cell-layered epithelium and is covered by the peripodial epithelium ) enables the labeling and tracking of cell populations . These cell populations can be genetically manipulated with the help of sophisticated tools . Finally , since the larvae live in their food , cellular stress situations can be imposed by controlling the food source . We have focused our analysis on cells lacking the tumor suppressor PTEN ( phosphatase and tensin homolog deleted on chromosome 10 ) . PTEN is well conserved from flies to humans , and it is the second most frequently mutated tumor suppressor found in many types of human cancers ( Goberdhan and Wilson , 2003; Salmena et al . , 2008; Hollander et al . , 2011; Song et al . , 2012 ) . PTEN antagonizes the function of the lipid kinase Phosphatidylinositide 3-kinase ( PI3K ) ; therefore , in the absence of PTEN , high levels of the lipid second messenger PIP3 result in an increased membrane recruitment and activation of the serine/threonine kinase PKB ( protein kinase B , also known as Akt ) , which leads to enhanced cellular growth , proliferation , and survival ( Altomare and Testa , 2005; Georgescu , 2010; Song et al . , 2012 ) . The consequences of activating PI3K signaling due to PI3K overexpression or loss of PTEN function has been extensively studied in Drosophila ( Leevers et al . , 1996; Goberdhan et al . , 1999; Huang et al . , 1999; Weinkove et al . , 1999; Gao et al . , 2000; Britton et al . , 2002 ) . Cells overexpressing Dp110/PI3K are enlarged and , in the fat body , increase their nutrient storage . This stockpiling of nutrients helps them to cell-autonomously bypass the nutritional requirements for cellular growth and DNA replication during amino acid deprivation ( Britton et al . , 2002 ) . In mitotic tissues , clones of PTEN mutant cells are enlarged , which is mainly caused by an increase in cell size ( Leevers et al . , 1996 ) . However , given the importance of PTEN as a tumor suppressor , the overgrowth caused by the loss of PTEN is rather mild ( Goberdhan et al . , 1999; Huang et al . , 1999; Gao et al . , 2000 ) . Recently , it has been demonstrated that tumors lacking PTEN or with increased PI3K activity are resistant to dietary restriction ( Kalaany and Sabatini , 2009 ) . This observation underscores the importance of understanding the intrinsic changes in early tumors caused by the microenvironment . Furthermore , it remains largely unknown how a growing tumor impacts on its environment . In this study , we attempted to mimic early events in tumor development by inducing clones of PTEN mutant cells under conditions in which nutrients become limiting . We show that cells lacking PTEN switch from hypertrophic growth to hyperplastic growth under nutrient restriction ( NR ) . This hyperproliferation occurs at the expense of neighboring wild-type cells , probably by competition for local and systemic pools of nutrients and other growth-promoting factors .
To assess the impact of starvation on survival , developmental timing , and weight , we reared larvae on food with varying yeast content . Yeast is the main source for micronutrients and amino acids in standard fly media . We observed a three-phase starvation response ( Figure 1A , B ) . Our standard medium contains 100 g/l yeast . Down to 50 g/l yeast , larvae can be regarded as fully fed since the developmental time , weight , and survival were not affected . At a yeast concentration range between 50 g/l and 10 g/l , the development was gradually delayed and the weight decreased . This phase can be regarded as mild starvation , since survival was not compromised down to 25 g/l yeast concentration . Reduction of the yeast concentration to and below 10 g/l increased the developmental time from first instar to pupa and severely decreased body weight . Since mortality was also increased , this phase is regarded as severe starvation . Based on these results , the turning point between mild starvation and severe starvation is approximately at 10 g/l concentration of yeast . 10 . 7554/eLife . 00380 . 003Figure 1 . Response to yeast starvation in wild-type and insulin signaling defective Drosophila . ( A and B ) Control flies reared on standard culture medium with varying yeast concentrations . ( A ) Weight reduction in males ( blue line ) and females ( red line ) . ( B ) Time from first instar to pupariation ( black line ) and the survival rates of first instar larvae to pupariation ( green line ) . ( C ) Adult dry weight of wild-type , PKB and PTEN mutants reared on 100 g/l and 10 g/l yeast food , respectively . Weight difference ( in % ) with respect to control ( y w ) on 100 g/l yeast food is depicted in each column . ( D ) Size comparison of female control flies ( y w ) and PKB hypomorphic mutants reared on 100 g/l and 10 g/l food , respectively . ( E ) Larvae of y w and a PTEN hypomorphic combination ( PTEN117/PTEN100 ) reared on 100 g/l ( late L3 ) and 10 g/l ( mid L3 ) . ( F ) Eye and wing discs of the larvae from ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 00310 . 7554/eLife . 00380 . 004Figure 1—source data 1 . Adult dry weight; survival L1 to pupariation; pupariation time; weight analysis of IIS mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 004 We next subjected hypomorphic PKB and PTEN mutants to 100 g/l and 10 g/l yeast food , respectively ( Figure 1C–F ) . Interestingly , fully fed hypomorphic PKB mutants weighed approximately the same as wild-type flies reared on 10 g/l yeast food . When reared on 10 g/l yeast food , PKB mutants were further delayed but only mildly decreased in size and weight ( Figure 1C , D ) . In contrast , hypomorphic PTEN larvae and adult flies were larger than wild-type controls under normal food conditions ( Figure 1E ) . The prospective adult tissues , the eye and wing imaginal discs , were larger compared to control discs under both conditions ( Figure 1F ) . PTEN hypomorphic larvae were highly sensitive to a reduction in yeast content , and although still larger , they did not survive to pupariation when reared on 10 g/l yeast food ( Figure 1C ) . This is consistent with the findings that larvae with randomly induced PI3K overexpression clones are starvation sensitive ( Britton et al . , 2002 ) . Thus , under NR conditions , insulin signaling needs to be dampened to allow for survival of the organism . In contrast to the situation at the organismal level , tumors with elevated PI3K activity are starvation resistant ( Kalaany and Sabatini , 2009 ) . To mimic the clonal nature of cancer , we analyzed the loss of PTEN function in clones generated in mitotic tissues ( imaginal discs ) by hsFlp/FRT-mediated mitotic recombination ( Golic and Lindquist , 1989; Xu and Rubin , 1993 ) . This system enables the generation of single genetically marked epithelial cells mutant for a defined tumor suppressor . Subsequent mitoses give rise to a patch of cells ( ‘clone’ ) devoid of the tumor suppressor . The clones can be tracked throughout development due to genetically encoded markers ( usually GFP ) . In a typical experiment , we induced clones by a heat shock 24 hr after egg laying . Subsequently , the larvae were split into two populations and transferred to vials with yeast concentrations of 100 g/l and 10 g/l , respectively . The discs were then dissected to analyze the clones either at the same time point or at the same developmental stage ( Figure 2A ) . 10 . 7554/eLife . 00380 . 005Figure 2 . PTEN mutant cells are resistant to starvation and have a growth advantage upon NR . ( A ) Schematic drawing of the clonal induction with hsFlp/FRT-mediated mitotic recombination . After egg laying , larvae were allowed to grow for 24 hr on yeast paste . After the heat shock , larvae were distributed on different food conditions ( e . g . , 100 g/l and 10 g/l yeast ) . The time points of dissection are depicted in green . ( B ) Third instar eye discs bearing hsFlp/FRT PTEN mutant and control clones ( marked by the absence of GFP ) of larvae reared under varying yeast concentrations . ( C ) Eye discs with hsFlp/FRT PTEN mutant and control clones ( marked by the absence of GFP ) induced at the same time and conditions , reared on 10 g/l yeast food , and dissected at the indicated age of the clones . ( D ) Eyes bearing eyFlp/FRT PTEN mutant and control clones ( marked by the absence of pigmentation ) of animals reared on 100 g/l and 10 g/l yeast food , respectively . ( D′ ) Quantification of the respective eye sizes . ( E ) Scanning electron micrographs of eyes almost exclusively composed of PTEN mutant or control tissue of animals reared on 100 g/l and 10 g/l food , respectively , and the quantification of ommatidia size ( E′ ) and number ( E′′ ) from PTEN mutant and control eyes . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 00510 . 7554/eLife . 00380 . 006Figure 2—source data 1 . Eye size; eye measurements of SEM pictures . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 00610 . 7554/eLife . 00380 . 007Figure 2—figure supplement 1 . Severe starvation induces malformations in adult eyes bearing PTEN clones . The structure of adult eyes carrying PTEN clones from animals raised on 10 g/l yeast food is unaltered . By contrast , severe distortions are observed in animals raised on 5 g/l yeast food . Magnifications of distorted parts of the eye are indicated with red boxes . The clones are marked by the absence of pigmentation . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 00710 . 7554/eLife . 00380 . 008Figure 2—figure supplement 2 . Severe starvation affects architecture of discs bearing PTEN clones . Discs bearing PTEN clones ( marked by the absence of GFP ) have a lobed appearance in larvae on 10 g/l yeast food , and they are severely distorted with polyp-like outgrowths of hollow appearance ( white arrow ) in larvae on 5 g/l yeast food . On the right: close-ups of polyp-like structures indicated with white arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 00810 . 7554/eLife . 00380 . 009Figure 2—figure supplement 3 . Severely overgrown PTEN clones tend to collapse . Melanized scars ( indicated with white arrows ) are sometimes observed in adult structures , probably as remnants of collapsed PTEN clones in animals reared on 5 g/l yeast food . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 00910 . 7554/eLife . 00380 . 010Figure 2—figure supplement 4 . PTEN mutant overgrowth is cell autonomous and does not affect differentiation . Eye sections of the adult eyes bearing control and PTEN clones from animals reared on 100 g/l and 10 g/l yeast food , respectively . The clones are marked by the absence of pigmentation . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01010 . 7554/eLife . 00380 . 011Figure 2—figure supplement 5 . PTEN clones have a growth advantage in wing discs under starvation . The starvation-dependent overgrowth of PTEN clones ( marked by the absence of GFP ) is not restricted to the eye imaginal disc but is also observed in wing imaginal discs . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01110 . 7554/eLife . 00380 . 012Figure 2—figure supplement 6 . PTEN mutant cells rapidly respond to the yeast content in the food . Food switch experiments reveal that a short starvation period early after clone induction is sufficient to result in the overgrowth phenotype of PTEN mutant tissue under starvation . The time points of the shifts are indicated; clones are marked by the absence of pigmentation . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 012 We monitored the growth behavior of clones of PTEN mutant cells ( henceforward called PTEN clones ) in larvae reared under decreasing yeast concentration in the culture medium . As previously published , PTEN clones were enlarged in fed larvae but they did not severely impact on the structure of the imaginal discs and of the adult eyes ( Goberdhan et al . , 1999; Gao et al . , 2000 ) ( Figure 2B and Figure 2—figure supplement 1 ) . Under mild starvation conditions ( 30 g/l yeast ) , PTEN clones were further increased in size , resulting in an overall increase in disc size as compared to discs harboring control clones . Reducing the yeast concentration to 20 g/l and below led to a massive size increase of the PTEN clones and to a concomitant reduction of the surrounding tissue . Furthermore , the PTEN clones had fused and rendered the discs lobed in appearance ( Figure 2B and Figure 2—figure supplement 2 ) . At 5 g/l yeast concentration , the eye discs with PTEN mutant tissue were severely overgrown with polyp-like structures protruding from the discs ( Figure 2—figure supplement 2 ) . The resulting adult eyes displayed strong malformations and outgrowths , indicating a loss of epithelial integrity ( Figure 2—figure supplement 1 , red squares ) . In some cases , no clones could be recovered in the adult eyes , probably due to a collapse of the PTEN clones resulting in melanized scars ( Figure 2—figure supplement 3 , white arrow ) . Tangential eye sections did not reveal any structural defects in ommatidial arrangement ( Figure 2—figure supplement 4 ) . PTEN mutant cells overgrow in a cell-autonomous manner as the enlarged ommatidia were exclusively composed of mutant cells . The overgrowth phenotype of PTEN clones is not specific to the eye; it was also observed in other imaginal tissues like the wing disc ( Figure 2—figure supplement 5 ) . Since starvation causes a developmental delay , the overgrowth of PTEN clones could be a consequence of the prolonged growth period . We therefore analyzed the PTEN clones at earlier time points ( 36 hr , 48 hr , and 60 hr after induction; Figure 2C ) . Already 36 hr after clone induction , the PTEN clones were increased in size , and the overgrowth of PTEN clones and the reduction in wild-type tissue were fully apparent after 60 hr ( Figure 2C ) , excluding the developmental delay as a main cause of the overgrowth . To investigate whether the early stages of starvation are crucial for the PTEN clones to overgrow , we performed ‘food switch’ experiments by transferring larvae with PTEN clones from starvation ( 10 g/l ) to normal ( 100 g/l ) food and vice versa at given time points . We used the eyeless- ( ey ) Flp/FRT system to induce clones specifically in the eye imaginal tissues early during larval development . When the transfer from starvation to normal food occurred within the first 72 hr , the overgrowth of the PTEN clones was efficiently rescued . Later shifts were no longer effective in suppressing the overgrowth . Consistently , it was sufficient to transfer larvae within the first 48 hr from normal to starvation food to produce the overgrowth phenotype . After this time point , PTEN clones did not acquire the full growth advantage over the surrounding tissue , confirming the importance of the initial stages for PTEN clones to develop the overgrowth phenotype ( Figure 2—figure supplement 6 ) . We quantified the PTEN induced overgrowth by measuring the eye size of eyFlp/FRT-induced PTEN clones ( Figure 2D , D′ ) . Similar to the hsFlp/FRT-induced clones , the eyFlp/FRT-induced PTEN clones and the entire eyes were slightly enlarged in flies raised on normal food . Under starvation conditions , the mutant tissue occupied most of the adult eye and the wild-type tissue was severely reduced . This overrepresentation of the mutant tissue resulted in an absolute increase of eye size under starvation as compared to normal food conditions ( Figure 2D , D′ ) . To investigate the effects of reduced nutrition on ommatidia size and number , we generated eyes almost completely composed of PTEN mutant tissue by means of the eyFlp/FRT cell lethal system and analyzed them by scanning electron microscopy ( Figure 2E–E′′ ) . Under normal food conditions , the eye size increase was mainly caused by larger ommatidia ( +27% ) , and to a minor extent by more ommatidia ( +14% ) . Under 10 g/l yeast food , the ommatidial size was roughly proportionally reduced in wild-type and PTEN mutant eyes . Intriguingly , whereas the ommatidia number was decreased by 6% in wild-type eyes , it was massively increased in PTEN mutant eyes ( +44% ) . Since the composition of the PTEN mutant ommatidia remained unchanged , the size and number of ommatidia reflects cell size and cell number . Thus , the PTEN mutant tissue displays a switch from hypertrophy to hyperplasia . Cell competition , a process that results in the elimination of suboptimal cells from a growing tissue ( Morata and Ripoll , 1975; Simpson , 1979; Simpson and Morata , 1981 ) , could contribute to the overgrowth of PTEN clones . ‘Winner’ cells actively eliminate ‘loser’ cells by inducing apoptosis and thereby take over the tissue ( Moreno et al . , 2002; Moreno and Basler , 2004; de la Cova et al . , 2004 ) . If PTEN mutant cells acted as supercompetitors , the surrounding tissue would suffer from apoptosis . We therefore monitored apoptosis in eye imaginal discs bearing PTEN clones by cleaved Caspase-3 and TUNEL staining ( Figure 3A , B ) . Whereas only few apoptotic cells were detected in the heterozygous tissue under both fed and starving conditions , increased levels of apoptosis were observed within the overgrowing PTEN clones ( Figure 3A , A′ ) . 10 . 7554/eLife . 00380 . 013Figure 3 . PTEN mutant cells are susceptible to cell death . ( A ) Cleaved Caspase-3 antibody staining ( in red ) of hsFlp/FRT PTEN clones ( marked by the absence of GFP ) in eye imaginal discs of larvae reared under the indicated conditions . The yellow dashed line indicates the morphogenetic furrow . About one third of the PTEN clones contain apoptotic cells . ( A′ ) Quantification of the apoptotic tissue area ( positive for cleaved Caspase-3 immunostaining ) in the PTEN mutant and the surrounding tissue ( PTEN117/+ , +/+ ) , respectively , relative to the total area of the respective cell populations . ( B ) TUNEL staining ( in red ) on eye imaginal discs harboring 48 hr control and PTEN clones reveals high apoptotic levels within the PTEN clones but only residual apoptosis in surrounding tissue or control discs . ( C ) Expression of anti-apoptotic p35 within PTEN clones ( marked by GFP ) enhances the overgrowth potential of PTEN mutant tissue . ( D ) Inhibition of apoptosis in the dorsal half of the eye discs by expression of p35 under the control of DE-Gal4 ( marked by RFP , red ) with randomly induced PTEN clones ( marked by the absence of GFP ) does not rescue the surrounding ( GFP positive ) tissue . ( E ) Inhibition of JNK-mediated apoptosis in the sister clones of the PTEN clones does not rescue the surrounding tissue . The different clones are marked as depicted with the colors in the labeling ( green: GFP positive , white: GFP negative ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01310 . 7554/eLife . 00380 . 014Figure 3—source data 1 . Tissue size; apoptotic area; ratio apoptotic area/tissue size . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01410 . 7554/eLife . 00380 . 015Figure 3—figure supplement 1 . Blocking cell death in PTEN clones enhances the overgrowth . Cleaved Caspase-3 antibody staining ( in red ) of PTEN and control clones expressing anti-apoptopic p35 ( positively marked by GFP ) in eye imaginal discs of larvae reared under the indicated conditions . Cleaved Caspase-3 signal is exclusively observed in the PTEN mutant clones and efficiently suppressed by expression of p35 . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01510 . 7554/eLife . 00380 . 016Figure 3—figure supplement 2 . The massive overgrowth of PTEN clones under harsh starvation conditions correlates with high levels of apoptosis . Outgrowing structures observed on 5 g/l yeast food are exclusively composed of PTEN mutant tissue ( marked by the absence of GFP ) and show high levels of cell death . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 016 Since the overgrowth is fully apparent at 60 hr after clone induction , we investigated PTEN clones at 48 hr to exclude the possibility that cell competition eliminates the surrounding tissue at this early stage ( Figure 3B ) . Again , most of the apoptosis was observed in the mutant tissue , arguing against a major role of cell competition in the overgrowth phenotype . Expression of p35 or of dominant negative JNK ( bskDN; Adachi-Yamada et al . , 1999 ) exclusively in the PTEN clones ( by means of the MARCM system ) efficiently blocked apoptosis and enhanced the overgrowth ( Figure 3C , Figure 3—figure supplement 1 , and not shown ) . To further address the role of apoptosis in the surrounding tissue , we expressed p35 in the dorsal half of the eye and randomly induced PTEN clones throughout the disc ( Figure 3D ) . Consistent with the above results and the apoptosis pattern , PTEN clones overgrew even more in the dorsal compartment , and the surrounding tissue still got strongly reduced . Thus , the cells neighboring PTEN clones are not eliminated by apoptosis . We also tested for the requirement of JNK signaling in the tissue neighboring the PTEN clones . hsFlp/FRT clones of bsk1 over a chromosome carrying a GFP marker and a PTEN mutation were induced . In this way , mitotic recombination events generate two adjacent twin spots mutant for bsk1 and PTEN , respectively . If PTEN cells attained the growth advantage by inducing JNK-mediated cell death in the twin spot , the overgrowth would be suppressed and the twin spot should grow larger . However , PTEN clones were overgrown irrespective of blocking JNK signaling in neighboring cells ( Figure 3E ) . Under severe starvation ( 5 g/l yeast content ) , the PTEN mutant polyp-like structures outgrowing from the eye discs exhibited very high levels of apoptosis ( Figure 3—figure supplement 2 ) . In the corresponding adult eyes , we often observed scars that were probably left behind by the collapsing clones ( Figure 2—figure supplement 3 ) . Those eyes contained more wild-type cells , indicating that the lost PTEN mutant tissue was compensated for . Thus , PTEN mutant cells are running at the edge of survival under limited nutrient conditions . They start to disobey the organ boundaries but are prone to apoptosis . To gain a better understanding of the downstream events contributing to the overgrowth of PTEN clones , we monitored the levels of the second messenger PIP3 in discs containing PTEN clones by means of the tGPH reporter , a GFP–PH domain fusion protein with high affinity for PIP3 ( Britton et al . , 2002 ) ( Figure 4A ) . The GFP–PH domain fusion protein was strongly localized at the membrane under both fed and NR conditions in PTEN clones . In contrast , the reporter signal was diffuse in the cytoplasm in the surrounding tissue under both conditions . Removal of Insulin receptor ( InR ) function from the PTEN mutant tissue only slightly reduced the overgrowth . Thus , in the absence of PTEN , the cellular PIP3 levels are sufficient to sustain the overgrowth regardless of the upstream signaling ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 00380 . 017Figure 4 . Interaction of insulin and TOR signaling with PTEN clones under starvation . ( A ) tGPH reporter ( green ) reveals high levels of PIP3 in PTEN clones ( clones marked by absence of LacZ in red ) under food conditions indicated . ( B ) P-PKB ( red ) is strongly increased in PTEN clones ( marked by the absence of GFP ) under normal conditions as well as under starvation . ( C ) PTEN overgrowth under normal and starvation conditions is strictly dependent on PKB activity . ( D ) Overexpression of FoxO results in a small eye under starvation conditions . This phenotype is suppressed by co-knockdown of PTEN . ( D′ ) Quantification of eyes from ( D ) . ( E ) Cherry-tagged FoxO localizes to the cytoplasm under normal food conditions and moves to the nucleus under starvation . This nuclear shuttling is prevented in PTEN clones . Clones are positively marked by Cherry ( in red ) . ( F ) Inhibiting TORC1 activity by overexpression of Tsc1/2 in PTEN clones ( positively marked by GFP ) suppresses the overgrowth . The suppression is evident already at 72 hr after clone induction . ( G ) Adult eyes showing the suppression of the overgrowth associated with PTEN clones by overexpression of Tsc1/2 under starvation under standard conditions and under starvation ( G′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01710 . 7554/eLife . 00380 . 018Figure 4—source data 1 . Eye size; shoulder size; eye/shoulder ratio; ommatidia number . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01810 . 7554/eLife . 00380 . 019Figure 4—figure supplement 1 . Genetic interactions between PTEN and IIS/TOR signaling components . Comparison of adult eyes bearing clones mutant for InR , FoxO , Rheb , PI3K92E , PTEN or combinations of PTEN with InR/FoxO/Rheb/PI3K92E under 100 g/l and 10 g/l yeast food conditions . Removal of FoxO slightly suppresses the overgrowth phenotype of PTEN mutant tissue under starvation . Similarly , a partial suppression is observed in PTEN InR and PTEN PI3K92E double mutants . Complete suppression is observed in PTEN Rheb double mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 01910 . 7554/eLife . 00380 . 020Figure 4—figure supplement 2 . Insulin and TOR signaling is modulated in PTEN clones in response to starvation . Western blots on eye imaginal disc tissue . On the left: PTEN mutant discs have strongly elevated P-PKB levels . Upon starvation , P-PKB levels are reduced but still much higher than in the control under normal conditions . Surprisingly , PKB levels are strongly reduced in the PTEN mutant tissue under starvation . On the right: PTEN mutant discs display high levels of P-S6K that strongly drop in starved animals . β-Tubulin served as loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02010 . 7554/eLife . 00380 . 021Figure 4—figure supplement 3 . The effects of PTEN knockdown and FoxO overexpression neutralize each other . GMR-Gal4 mediated expression of PTEN-RNAi , FoxO or both under 100 g/l yeast food conditions and quantification of the eye size normalized to shoulder size . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02110 . 7554/eLife . 00380 . 022Figure 4—figure supplement 4 . Overexpression of FoxO does not suppress PTEN mutant overgrowth . Overexpression of FoxO in PTEN clones ( MARCM system , positively marked by GFP ) does not suppress the PTEN mutant tissue overgrowth under both under normal and starvation conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02210 . 7554/eLife . 00380 . 023Figure 4—figure supplement 5 . PTEN clones require TORC1 activity to overgrow . Size measurements of eyes bearing PTEN and control clones co-expressing Tsc1/2 of animals reared on 100 g/l and 10 g/l yeast food . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02310 . 7554/eLife . 00380 . 024Figure 4—figure supplement 6 . Reducing TORC1 activity suppresses PTEN mutant overgrowth . Overexpression of Tsc1/2 in PTEN clones ( positively marked by GFP ) suppresses the overgrowth phenotype without reducing the P-PKB levels ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02410 . 7554/eLife . 00380 . 025Figure 4—figure supplement 7 . PTEN mutant cells require TORC1 activity to acquire a proliferative advantage under starvation . Scanning electron micrographs of eyes generated with the eyFlp/FRT cell lethal system and almost exclusively composed of PTEN , TOR , PTEN TOR mutant or control tissue of animals reared on 100 g/l and 10 g/l yeast food , respectively . On the right side: Quantification of eye size , ommatidia number and ommatidia size . Note that the PTEN TOR double mutant eyes are very convex . As the projection of the eye onto the plane of the picture is measured , the indicated eye size is an underestimate . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 025 Similarly , elevated levels of phosphorylated PKB ( P-PKB ) were found at the membrane in PTEN clones at 100 g/l and 10 g/l yeast content ( Figure 4B ) . These results were further confirmed by a Western blot analysis comparing discs exclusively composed of PTEN mutant tissue with discs bearing wild-type clones under normal food conditions ( Figure 4—figure supplement 2 ) . Due to the extremely small size of the wild-type discs under NR , we were not able to generate enough tissue for the analysis . Although the total levels of PKB were decreased in PTEN mutant tissue under both conditions as compared to the control under normal food conditions , the P-PKB levels were elevated . Thus , high PIP3 levels result in robust PKB activation under standard and starvation conditions . Consistently , reducing or removing PKB function by using a PH-domain mutant ( Stocker et al . , 2002 ) and a kinase-dead PKB ( Staveley et al . , 1998 ) , respectively , completely suppressed the PTEN overgrowth under normal as well as under starvation conditions ( Figure 4C ) . The transcription factor FoxO is an important target of PKB . Upon high PI3K signaling activity , FoxO is phosphorylated by PKB and sequestered in the cytoplasm , and FoxO-mediated transcription of growth-suppressing genes is thus inhibited ( Brunet et al . , 1999; Kops et al . , 1999; Burgering and Kops , 2002 ) . FoxO was localized in the cytoplasm in PTEN mutant tissue under both conditions , whereas starvation induced the shuttling of FoxO into nuclei in the control tissue ( Figure 4E ) . It has been shown that overexpression of FoxO decreases eye size , which is exacerbated under starvation conditions ( Junger et al . , 2003; Kramer et al . , 2003 ) . Consistent with the inactivation of FoxO in PTEN mutant cells , the FoxO-induced eye size reduction was not observed in PTEN clones ( Figure 4D and Figure 4—figure supplements 3 and 4 ) . Signaling via Target of Rapamycin complex 1 ( TORC1 ) promotes growth in response to nutrients . High TORC1 activity boosts cellular growth , at least in part via the phosphorylation of S6K and 4E-BP ( reviewed in Hietakangas and Cohen , 2009 ) . PTEN mutant cells did respond to starvation by reducing the levels of S6K and of phosphorylated S6K ( P-S6K ) , as revealed by Western blot analysis on discs bearing eyFlp/FRT cell lethal clones of PTEN ( Figure 4—figure supplement 2 ) . PKB signaling activity was also slightly reduced as compared to standard conditions ( Figure 4B and Figure 4—figure supplement 2 ) . We next inhibited TORC1 activity selectively in PTEN clones by co-overexpression of Tuberous Sclerosis Complex-1 and -2 ( Tsc1/2 ) —which together form a complex with GTPase activating protein ( GAP ) activity towards the small GTPase Rheb , an essential activator of TORC1 ( Garami et al . , 2003; Zhang et al . , 2003 ) —or by removing Rheb function . Reducing TORC1 activity suppressed the overgrowth of the PTEN mutant tissue ( Figure 4F , G , 4G′ and Figure 4—figure supplement 5 ) . P-PKB levels , however , were increased in the mutant tissue under both fed and NR conditions ( Figure 4—figure supplement 6 ) , which is in agreement with the negative feedback regulation of TORC1 on the activation of PKB ( Radimerski et al . , 2002; Kockel et al . , 2010 ) . Eyes mutant for PTEN and a hypomorphic allele of TOR were reduced to the size of control eyes at both fed and starved conditions , indicating that the overgrowth of PTEN clones strictly depends on normal TORC1 function ( Figure 4—figure supplement 7 ) . Interestingly , whereas the size reduction of PTEN clones caused by impaired TORC1 function was primarily due to smaller ommatidia under normal conditions , ommatidia size was only slightly affected under starvation . By contrast , the hyperproliferation observed in PTEN clones under NR was almost completely abolished ( Figure 4—figure supplement 7 ) . Thus , TORC1 is indispensable for the switch to hyperproliferation observed in PTEN mutant tissue upon NR . The cationic amino acid transporter Slimfast ( Slif ) has been shown to activate TOR signaling in the fat body , and it appears to be an important component of a systemic nutrient sensor mechanism ( Colombani et al . , 2003 ) . Slif function is also required in mitotic tissues , as clones in which Slif has been knocked down ( Slifanti ) are reduced in size , possibly due to reduced TORC1 activation . Since TORC1 is required for the overgrowth of PTEN clones , knocking down Slif should suppress this phenotype in a similar way as observed for Tsc1/2 overexpression . Whereas expressing Slifanti in PTEN clones resulted in a slight reduction of the mutant tissue under normal conditions , it caused a nearly complete elimination of the PTEN mutant tissue under starvation conditions , while not affecting the control clones ( Figure 5A ) . The loss of PTEN mutant tissue was accompanied by a massive increase in apoptosis , and GFP-positive ( and thus PTEN mutant ) cellular remnants were scattered throughout the disc ( Figure 5B ) , suggesting that PTEN clones collapsed after an initial overgrowth . Indeed , early overgrowth of PTEN clones and high P-PKB signals were observed under starvation ( Figure 5C upper panel ) . Later , Slif function became limiting in PTEN clones that were subsequently eliminated ( Figure 5A , C lower panel ) . Intriguingly , the structure of the resulting adult eyes that had lost the PTEN clones was completely normal , indicating that the surrounding tissue was able to compensate for the loss of the clones ( Figure 5D , D′ ) . The effects that the reduction of Slimfast exerted on PTEN clones are distinct from those resulting from inhibiting TORC1 function , as overexpression of Tsc1/2 suppressed the overgrowth of PTEN clones already at an early stage ( Figure 4F ) . This may indicate that , rather than the activation of TORC1 , another aspect of Slif function—probably the amino acid influx itself—is critical for the survival of the PTEN mutant tissue . 10 . 7554/eLife . 00380 . 026Figure 5 . PTEN mutant tissue critically depends on the amino acid transporter Slimfast . ( A ) PTEN clones co-expressing Slifanti collapse and disappear under starvation , whereas Slif reduction does not affect control clones . ( B ) TUNEL staining reveals high apoptosis levels ( in red ) in cells with Slifanti expression . Reducing the levels of Slif eliminates PTEN mutant cells by apoptosis . ( C ) Dying 110 hr-old PTEN clones expressing Slifanti lose the P-PKB signal ( red ) , although they initially overproliferate ( 72 hr-old clones ) indistinguishably from PTEN mutant tissue . ( D and D′ ) Adult eyes with unmarked control , PTEN and PTEN plus Slifanti clones under standard ( D ) and starvation conditions ( D′ ) . The reduction of Slif levels rescues the overgrowth associated with PTEN mutant eyes , especially under starvation conditions . All clones in the discs are positively marked by GFP . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02610 . 7554/eLife . 00380 . 027Figure 5—figure supplement 1 . Autophagy does not contribute to the initial survival of PTEN clones with reduced Slif levels under starvation . Inhibiting autophagy ( by means of Atg5-RNAi ) in PTEN clones with reduced Slif function ( positively marked by GFP ) neither impacts on their initial overgrowth nor on their later collapsing under starvation . Blocking apoptosis ( by means of p35 expression ) in PTEN clones with reduced Slif ( positively marked by GFP ) prevents the PTEN mutant cells from being eliminated . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02710 . 7554/eLife . 00380 . 028Figure 5—figure supplement 2 . Inhibition of cell death in PTEN clones with reduced Slif prevents the initial collapsing in eye discs . Higher magnification pictures of the undead PTEN mutant cells . Some large ( and thus PTEN mutant ) cells have lost GFP expression ( white arrows ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 02810 . 7554/eLife . 00380 . 029Figure 5—figure supplement 3 . Blocking autophagy and apoptosis in PTEN clones with reduced Slif does not prevent them from dying . Adult eye phenotypes of the clones shown in figure supplement 1 . PTEN clones with reduced Slif function are not recovered in adult eyes upon inhibition of autophagy or apoptosis under starvation conditions . White arrows indicate malformations reminiscent of the collapsed clones . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 029 We also assessed whether the initial overgrowth of the PTEN clones with reduced Slif function could be attributed to autophagy that maintains the levels of amino acids and thereby sustains TORC1 activity . However , reducing Atg5 , a key component of the autophagic machinery , did not impact on the initial growth of PTEN clones with reduced Slif function ( Figure 5—figure supplements 1 and 3 ) . Blocking apoptosis ( by means of p35 expression ) in PTEN clones with diminished Slif function delays the collapse of the clones as the clones can still be detected in imaginal discs ( Figure 5—figure supplements 1 and 2 ) but are absent from adult eyes ( Figure 5—figure supplement 3 ) . We also observed non-autonomous effects of the PTEN clones on the surrounding tissue . P-PKB levels in the wild-type tissue neighboring with PTEN clones were reduced as compared to the tissue neighboring with control clones ( Figure 6—figure supplement 1 ) . This reduction is more apparent under normal conditions , which can be attributed to the already low P-PKB levels in discs of starved animals . We were wondering whether this cell non-autonomous phenomenon only affects closely neighboring tissue or whether it acts at a longer range . Knocking down PTEN in the dorsal part of the eye not only resulted in its size increase , but also reduced the size of the ventral part as compared to control eyes ( Figure 6A , A′ ) . This growth-reducing effect within the eye was already visible under standard conditions but further enhanced on 10 g/l yeast food . In contrast to the null mutant situation , RNAi-mediated knockdown of PTEN did not result in enhanced overgrowth under starvation as compared to standard conditions , probably due to residual PTEN in the knockdown situation . 10 . 7554/eLife . 00380 . 030Figure 6 . Systemic effects of PTEN clones on peripheral tissues . ( A ) PTEN was knocked down in the dorsal part of the eye by means of DE-Gal4 . ( A′ ) Quantification of the sizes of the ventral and dorsal halves . Percentage indicated in the bars represents the size reduction with respect to the respective part of control eyes under normal conditions . ( B ) Flies with PTEN mutant heads have smaller bodies as judged by shoulder area ( s . a . ) under starvation conditions . ( C ) Wings of flies with PTEN mutant heads under starvation are smaller than wings of control flies . ( C′ ) Quantification of the wings from ( C ) . ( D ) Size of fat body nuclei of larvae with PTEN mutant eye discs is decreased under starvation as compared to the control . ( D′ ) Quantification of nuclear size from ( D ) . PTEN mutant heads were generated by the eyFlp/FRT cell lethal system . Statistical analyses were done with Student’s t-test ( two tailed ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03010 . 7554/eLife . 00380 . 031Figure 6—Source data 1 . Dorsal eye size; shoulder size; wing size; nuclear size in fat bodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03110 . 7554/eLife . 00380 . 032Figure 6—figure supplement 1 . PTEN mutant tissue influences PKB signaling in the neighboring tissue . The tissue surrounding PTEN clones ( marked by the absence of GFP ) displays a reduction in P-PKB levels ( red ) as compared to the tissue surrounding wild-type clones . Note that the nuclei of the cells adjacent to PTEN clones under starvation are more densely packed , indicative of a reduced cell size . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 032 The analysis of flies with heads mostly composed of PTEN mutant tissue revealed similar non-autonomous effects on the sizes of other organs and of the entire body . A decrease in shoulder area ( as a measure for body size ) was observed under starvation but not under normal conditions ( Figure 6B ) . Consistently , the wing size and the fat body nuclear size were reduced , especially in the animals that were raised under starvation ( Figure 6C , C′ , D , D′ ) . Thus , the growth-reducing non-autonomous effect of PTEN mutant tissue acts systemically . The observed non-autonomous effects suggest that the PTEN mutant tissue efficiently competes with other larval tissues for common resources to support their massive growth , thereby launching a vicious cycle in the neighboring tissue that gets further starved and reduced . If our interpretation was correct , PTEN clones should overgrow less under starvation when growth signaling is maintained in neighboring cells . We attempted to promote IIS by expressing the ligand Dilp-2 in eye imaginal discs ( Brogiolo et al . , 2001; Ikeya et al . , 2002 ) . Whereas expressing Dilp-2 under ey-Gal4 control did not affect control eyes , it suppressed the overgrowth caused by the loss of PTEN under starvation conditions . Furthermore , it restored the growth of both the tissue surrounding the PTEN clones and peripheral tissue ( Figure 7A ) . 10 . 7554/eLife . 00380 . 033Figure 7 . A reduction in growth signaling in the direct neighborhood and in peripheral tissues is required for the overgrowth of PTEN mutant cells . ( A ) Restoring insulin signaling by means of Dilp-2 expression ( driven by ey-Gal4 ) suppresses the overgrowth of PTEN mutant cells under starvation . The eyes get smaller and the growth of the surrounding and the peripheral tissue is restored . The expression of Dilp-2 has no effect on control eyes . Quantification of eye size ( A′ ) and shoulder area ( A′′ ) from ( A ) . ( B ) Systemic reduction of growth by ubiquitous expression of Imp-L2 with arm-Gal4 under standard conditions decreases the size of the control eyes but enhances PTEN mutant overgrowth . ( B′ ) Quantification of eye size and shoulder area from ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03310 . 7554/eLife . 00380 . 034Figure 7—Source data 1 . Eye size , shoulder size; ommatidia number; ommatidia size . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03410 . 7554/eLife . 00380 . 035Figure 7—figure supplement 1 . Growth properties of PTEN mutant tissue are influenced by the neighboring tissue . PTEN clones ( marked by the absence of GFP ) overgrow less under starvation when neighboring with Tsc1 clones ( marked by the absence of LacZ , pink ) , indicating that both clonal populations compete for common resources . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03510 . 7554/eLife . 00380 . 036Figure 7—figure supplement 2 . Autonomous and non-autonomous effects of reducing TORC1 activity on the overgrowth of PTEN clones . Reducing TORC1 signaling ( using the hypomorphic allele TOREP2353 ) specifically in PTEN mutant cells suppresses their proliferative advantage and restores growth of the surrounding normal tissue on both 100 g/l and 10 g/l yeast food . Conversely , when neighboring with sister clones mutant for TOR , PTEN clones outcompete the surrounding tissue already at standard conditions and overgrow more under starvation . The different clones are marked as depicted with colors in the labeling ( orange: containing red/orange pigmentation , black: lacking red/orange pigmentation ) . Below: Quantification of eye sizes . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03610 . 7554/eLife . 00380 . 037Figure 7—figure supplement 3 . PTEN mutant cells neighboring with TOR mutant twin clones acquire a proliferative advantage early in larval development . Eye discs with eyFlp/FRT PTEN and PTEN over TOR mutant clones ( PTEN cells are positively marked by GFP ) , dissected at the indicated time points after egg deposition . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03710 . 7554/eLife . 00380 . 038Figure 7—figure supplement 4 . Wild-type cells neighboring with TOR or PKB mutant cells gain a growth advantage . Eye discs with hsFlp/FRT TOR and PKB clones ( marked by the absence of GFP ) . Note that the wild-type twin clones ( bright green due to two copies of ubiGFP ) are strongly overgrown . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03810 . 7554/eLife . 00380 . 039Figure 7—figure supplement 5 . PTEN mutant cells sense the growth reduction in their neighborhood and proliferate faster under starvation . Scanning electron micrographs of eyes generated with the eyFlp/FRT ubiGFP system bearing TOR , PTEN , PTEN over TOR mutant or control clones of animals reared on 10 g/l yeast food . When neighboring with sister clones mutant for TOR , PTEN cells overgrow more under starvation exclusively due to an increase in cell number . On the rights side: Quantification of eye size , ommatidia number and ommatidia size . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 03910 . 7554/eLife . 00380 . 040Figure 7—figure supplement 6 . Systemic reduction of growth signaling induces a starvation-like response in PTEN mutant cells . Systemic reduction of IIS ( achieved by arm-Gal4 UAS-Imp-L2 ) enhances the overgrowth of PTEN clones that are neighboring with TOR mutant sister clones . The bigger eyes are composed of more but smaller ommatidia . Below: Quantification of eye size , ommatidia number and ommatidia size . Note that the increase in cell number in PTEN clones neighboring with TOR mutant cells is an underestimate because the reference eyes contain many very small ( TOR mutant ) ommatidia . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 040 We also analyzed the consequences of introducing cells with deregulated TORC1 activity—and thus enhanced cellular growth–in the neighborhood of PTEN clones . To this end , we simultaneously generated PTEN clones and clones of Tsc1 mutant cells ( Figure 7—figure supplement 1 ) . Interestingly , the mutant clones did not have additive effects on overgrowth . The introduction of Tsc1 clones into the neighborhood of PTEN clones rather reduced their overgrowth under starvation . The discs were completely composed of either Tsc1 or PTEN mutant tissue , as the clones did not heavily overlap . Thus , when neighboring with a tissue that also has a growth advantage , PTEN clones themselves experience competition for common resources . To investigate whether the overgrowth of PTEN clones is enabled by a non-autonomous influence from the starved neighboring tissue , we genetically starved the surrounding tissue of PTEN clones by reducing TORC1 activity . We induced eyFlp/FRT clones of PTEN with sister clones mutant for TOR ( Figure 7B , Figure 7—figure supplements 2 and 3 ) . The resulting eyes were completely composed of PTEN mutant tissue already under standard conditions . However , the eyes were smaller as compared to eyes bearing PTEN clones only . We also observed that clones of TOR mutant cells generated in otherwise wild-type eyes were massively underrepresented . The adult eyes were almost completely composed of wild-type twin clone tissue ( Figure 7B ) , suggesting that not only PTEN mutant cells with a high growth potential but also wild-type cells neighboring with cells impaired in TORC1 function attain a growth advantage and proliferate in place of the slower growing cells . Similar observations were made with mutations impinging on IIS activity ( e . g . , PKB; Figure 7—figure supplement 4 ) . Thus , relative differences in IIS activity cause a differential growth behavior . We next tested how the combination of genetically starved neighboring cells ( mutant for TOR ) with real starvation during development impacts on the growth behavior of PTEN clones . Interestingly , when in competition with TOR mutant cells , PTEN clones did overgrow under NR as opposed to normal conditions ( Figure 7—figure supplement 2 ) . The overgrowth was even more pronounced than without TOR mutant neighbors , and it was apparent already early during larval development ( Figure 7—figure supplement 3 ) . SEM analysis revealed that the additional overgrowth was primarily caused by more cells ( Figure 7—figure supplement 5 ) . Finally , we wondered whether systemically dampening IIS would allow the overgrowth of PTEN clones under normal conditions . We ubiquitously expressed Imp-L2 , which encodes a secreted antagonist of Dilp-2 ( Honegger et al . , 2008 ) . In this context , PTEN clones did overgrow , irrespective of whether they were neighboring with TOR mutant cells or wild-type cells ( Figure 7B ) . The overgrowth of the PTEN clones caused a further reduction in body size , as evidenced by decreased shoulder width ( Figure 7B ) . Cells devoid of PTEN also reacted to the decrease in IIS activity ( by slightly reducing cell size ) but the massive increase in cell number caused the total overgrowth ( Figure 7—figure supplement 6 ) . Thus , systemically decreasing IIS enhances the proliferative potential of PTEN mutant cells .
Tumors with high PI3K pathway activity are associated with increased resistance to starvation . Here , we describe how the loss of the tumor suppressor PTEN contributes to early clonal expansion . PTEN mutant cells tolerate and survive starvation in a clonal situation in Drosophila imaginal discs . This response is completely dependent on high PKB and sustained TORC1 activities within the PTEN clones . PTEN mutant cells also acquire a growth advantage under starvation conditions at the expense of wild-type cells in the immediate neighborhood and in the entire organism . It has previously been demonstrated that activation of PI3K ( or of TORC1 ) results in starvation insensitivity in endoreplicative tissues ( ERTs ) ( Britton et al . , 2002 ) . The role of activated PI3K in starvation resistance of Drosophila epithelia is also not unprecedented . Anaplastic Lymphoma Kinase ( ALK ) -dependent activation of PI3K is necessary for bypassing the amino acid requirement in growing neuroblasts ( neural progenitors ) under NR conditions ( brain sparing ) ( Cheng et al . , 2011 ) . However , imaginal tissues are not spared upon starvation , arguing against an important function of ALK in mediating growth of imaginal discs under NR conditions . Mitotically active imaginal disc cells with high PIP3 levels respond differently to starvation as compared to corresponding cells in the ERTs . PTEN mutant imaginal disc cells do react to starvation by reducing their size like control cells do . The cell size reduction is proportional: PTEN mutant cells are still enlarged with respect to control cells under starvation . However , the cell size reduction is compensated for by a massive increase in proliferation , causing a net increase in mutant tissue . In fact , the PTEN mutant tissue is absolutely enlarged under starvation conditions . PTEN mutant cells therefore gain a growth advantage and can take over a complete organ when the surrounding tissue is starved . This advantage is neither due to a prolonged growth period ( PTEN clones are increased in size early after clone induction ) nor to a complete insensitivity to starvation . It rather reflects a change in their mode to respond to a limited access to nutrients: PTEN mutant cells switch from hypertrophic to hyperplastic growth under starvation . This tolerance towards increased proliferation could favor the selection for secondary mutations that enhance proliferation and thus tumor progression . To our surprise , PTEN clones display increased levels of apoptosis , and selective inhibition of apoptosis within the clones results in tremendous hyperplastic overgrowth . The induction of apoptosis in PTEN mutant tissue contrasts the known pro-survival function of PKB ( Scanga et al . , 2000 ) . However , it indicates that the starved PTEN mutant cells , despite their proliferative advantage , exist on the edge of survival and are highly susceptible to apoptosis . Our findings also suggest that acquisition of factors preventing apoptosis would strongly enhance the growth of tumors lacking PTEN . How can PTEN mutant cells grow at the expense of the non-mutant tissue ? Comparing the P-PKB levels in the heterozygous and wild-type tissue revealed a reduction in the tissue surrounding PTEN clones as compared to tissue adjacent to control clones . Furthermore , cells neighboring with PTEN mutant tissue are smaller than those in the discs containing control clones . Thus , PTEN clones impact on IIS activity in the surrounding tissue , thereby reducing its growth potential . This appears to be sufficient to explain the strong reduction of the surrounding tissue , since the process of ousting the neighboring tissue is completed as early as 60 hr after clonal induction , when the disc still contains relatively few cells . This non-autonomous growth reduction is not restricted to the local neighbors in a cell–cell contact-dependent manner , but rather affects all tissues of the organism . The impact of PTEN mutant tissue on its surroundings suggests that the cells compete for common factors , most likely growth factors and nutrients . Under starvation , the circulating levels of growth factors and nutrients are strongly reduced , and thereby can become limiting ( Cheng et al . , 2011 ) . Drosophila insulin-like peptides ( DILPs ) produced by the insulin-producing cells ( IPCs ) in the brain stimulate the growth of imaginal disc cells via the InR . Starvation reduces the secretion of DILPs from the IPCs , and thus reduces growth ( Geminard et al . , 2009 ) . Removing InR function in PTEN clones only mildly affects the overgrowth . Thus , despite lacking the signal input via InR , these cells are able to accumulate PIP3 , probably because of the basal activity of PI3K . Therefore , low levels of circulating DILPs are sufficient to boost PIP3 levels and PKB activity in PTEN mutant cells under starvation . The cationic amino acid transporter Slimfast ( Slif ) has been described as an upstream activator of TORC1 ( Colombani et al . , 2003 ) . We show that , in contrast to inhibiting TORC1 activity , reducing Slif does not block the initial overgrowth of PTEN clones under starvation . However , it causes the overgrown clones to collapse . We speculate that the high metabolic demands of PTEN mutant cells require an efficient amino acid uptake , which is blocked by reducing Slif . Thus , PTEN mutant cells rely on their ability to efficiently compete for nutrients and growth factors under starvation conditions with their direct neighbors and the peripheral tissue . Since the reduction of Slif affects the growth of the wild-type and the PTEN mutant tissue in a differential manner , this amino acid transporter could represent a target for a dosage-dependent drug therapy of tumors with PI3K activation . Our results suggest the following sequence of events when a single cell embedded in a mitotic tissue loses the tumor suppressor PTEN ( Figure 8 ) . The loss of PTEN function triggers high PIP3 levels , thereby activating PKB and TORC1 , thus stimulating cell growth and division . Under starvation conditions , circulating growth factors and nutrients are scarce . In response to these limited resources , both wild-type and PTEN mutant cells adjust their growth with respect to cellular size . In sharp contrast to wild-type clones , PTEN clones retain high PI3K activity and strongly increase their cell number . This hyperplastic clonal overgrowth depends on withdrawing nutrients from the neighborhood and ultimately the entire organism . The active scavenging of resources is dependent on efficient amino acid transporters ( Slif ) , which also ensure that TORC1 remains active . During their entire growth , cells are at the brink of death as evidenced by the high apoptosis rate within the PTEN clones . Thus , enhanced PIP3 levels in cells that lost the tumor suppressor PTEN not only leads to starvation resistance at the cellular level , but also suppresses growth in their direct tissue neighborhood and in peripheral tissues by competing for nutrients and growth factors . Our findings demonstrate how limiting nutrient conditions enhance the proliferative potential of PTEN mutant cells , independent of additional genetic alterations . 10 . 7554/eLife . 00380 . 041Figure 8 . Model of hyperplastic overgrowth of PTEN mutant tissue under starvation . ( A ) Under standard conditions ( normal food ) , the PTEN mutant tissue overgrows in a hypertrophic manner . The tissue is enlarged because of larger PTEN mutant cells . ( B ) Under starvation conditions , the PTEN mutant tissue is metabolically more active and outcompetes the surrounding wild-type tissue , resulting in a hyperplastic overgrowth . The size of the tissue is further increased because of more PTEN mutant cells . The PTEN mutant tissue is susceptible to apoptosis , and it depends on the function of the amino acid transporter Slimfast . Under both conditions , the PTEN mutant tissue exhibits high insulin signaling activity and is dependent on the functions of PKB and TORC1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00380 . 041
1 liter Drosophila medium contains 100 g of fresh yeast , 55 g cornmeal , 10 g wheat flour , 75 g sugar and 8 g bacto agar , here referred to as standard medium ( or 100 g/l yeast food ) . Starvation food ( or 10 g/l yeast food ) was generated by reducing the amount of yeast without altering the other ingredients . All crosses and experiments were performed at 25°C under non-crowding conditions . To generate a viable hypomorphic situation for PTEN , the null allele PTEN117 and the hypomorphic allele PTEN100A were used in heteroallelic combinations ( Oldham et al . , 2002 ) . For clonal analysis , FRT40 PTEN117 and FRT82 Tsc1Q87X ( Oldham et al . , 2000; Tapon et al . , 2001 ) were used . To generate null mutant clones for PTEN on the third chromosome , the genomic rescue construct FRT82 PTENgenomic rescue ubiGFP ( Gao et al . , 2000 ) was used in a null mutant PTEN117 background . Knockdown of PTEN was achieved using the VDRC line 101475 in combination with GMR-Gal4 and DE-Gal4 ( Morrison and Halder , 2010 ) . For genetic interaction studies with the insulin and TOR pathways , the following alleles were used: FRT82 PKB1 and FRT82 PKB3 ( Stocker et al . , 2002 ) , FRT82 InR5545 ( Fernandez et al . , 1995 ) , FRT82 PI3K92E2H1 ( Halfar et al . , 2001 ) , FRT82 FoxO25 ( Junger et al . , 2003 ) , FRT FoxOΔ94 ( Slack et al . , 2011 ) , FRT40 TOR2L19 ( Oldham et al . , 2000 ) , FRT40 TOREP2353 ( Zhang et al . , 2000 ) and FRT82 Rheb2G5 ( Stocker et al . , 2003 ) . The following transgenic fly lines were used: GMR-Gal4 Thor1; EP-dFoxO ( Junger et al . , 2003 ) , UAS-cherry dFoxO , UAS-Tsc1/2 ( Tapon et al . , 2001 ) , UAS-Dilp2 ( Brogiolo et al . , 2001 ) , UAS-Imp-L2 ( Honegger et al . , 2008 ) and UAS-Atg5RNAi ( Scott et al . , 2004 ) . For blocking apoptosis , FRT40 bsk1 ( Riesgo-Escovar et al . , 1996 ) , UAS-p35 ( Hay et al . , 1995 ) and UAS-bskDN ( Adachi-Yamada et al . , 1999 ) were used . To monitor PI3K activity , a tGPH reporter was used ( Britton et al . , 2002 ) . The amino acid transporter Slimfast was silenced using Slifanti ( Colombani et al . , 2003 ) . A full list of alleles and transgenes used is provided in Supplementary file 1 . Flies were crossed in standard rearing vials for 3 days , transferred to laying cages , and allowed to lay eggs on apple agar plates ( 11 hr in most of the experiments , 3 hr for fat body analysis ) . For induction of eyFlp/FRT clones , eggs were distributed to the different food conditions immediately , whereas for induction of hsFlp/FRT clones , eggs were allowed to hatch , and they were heat-shocked before distribution . For quantification of survival , dead embryos were counted 24 hr after seeding to the food , and survival of the pupae was recorded . All phenotypic analyses on adult flies were performed on females ( unless indicated otherwise ) , and measurement of nuclear size in fat bodies was done on female larvae . Size and number of the ommatidia in SEM pictures , size of the eyes , shoulders , wings , and fat body nuclei were measured using Photoshop CS3 . Student’s t-test ( two-tailed ) was used to test for significance in all the quantification experiments . For determination of dry weight , flies were dried at 95°C for 5 min and individually weighed with a Mettler Toledo MX5 microbalance . Mutant clones in eye and wing imaginal discs were generated with y , w , hsFlp; FRT40 or FRT82 flies . Clones were induced during the first instar ( heat shock for 15 min at 37°C , 38 hr after egg deposition [AED] ) , and larvae were dissected in the third instar before wandering , unless otherwise indicated . For positively marked PTEN knockdown clones , Actin Flp-out Gal4 ( Neufeld et al . , 1998 ) and the MARCM ( Lee and Luo , 2001 ) system were used . Clones generated by the Actin Flp-out Gal4 technique were induced during the first larval instar ( heat shock for 10 min at 37°C , 38 hr AED ) , and larvae were dissected in the third instar before wandering . Eye-specific clones were generated using the eyFlp/FRT system . The exact genotypes are indicated in Supplementary file 1 . Larval imaginal discs were fixed in 4% PFA ( 30 min , RT ) , permeabilized in 0 . 3% PBT ( 15 min , RT ) , blocked in 2% NDS in 0 . 3% PBT ( 1 hr , RT ) , incubated with the primary antibodies overnight ( 4°C ) , washed three times in 0 . 3% PBT , and incubated with secondary antibodies ( 1 hr , RT ) . The nuclei were visualized with 1:2000 DAPI in 0 . 3% PBT ( 15 min , RT ) . Antibodies used in this study were: rabbit α-Drosophila phospho-Akt/PKB Ser505 ( 1:300; Cell Signaling ) , rabbit α-cleaved Caspase 3 ( 1:300; Cell Signaling ) , mouse α-β-Galactosidase ( 1:300; Promega ) , Cy3- and Cy5-coupled α-mouse or α-rabbit IgG ( 1:300; Amersham ) . All the dilutions were made in blocking solution . TUNEL staining was carried out according to the manufacturer’s protocol ( ApopTag Red In Situ Apoptosis Detection KitS7165; Millipore ) . Larval fat bodies were fixed in 8% PFA ( 45 min , RT ) and stained with 1:50 Alexa Fluor 488 phalloidin in 0 . 2% PBT ( 90 min , RT; Molecular Probes ) and 1:500 DAPI in 0 . 2% PBT ( 5 min , RT ) . Western blots on L3 eye imaginal discs were performed according to standard protocols . Antibodies were α-Drosophila phospho-PKB Ser 505 ( 1:1000; Cell Signaling ) , α-PKB ( 1:1000; Cell Signaling ) , α-phospho-S6K ( 1:1000; Cell Signaling ) , α-S6K ( 1:2000; our own antibody ) , α-Tubulin ( 1:10 , 000; Sigma ) , HRP-conjugated α-mouse and α-rabbit IgG ( 1:10 , 000; Amersham ) . Histological sections of the adult fly eyes were performed as previously described ( Basler and Hafen , 1988 ) . For the confocal images , a Leica SPE confocal laser scanning microscope was used . A Jeol JSM-6360LV microscope was used for scanning electron microscope pictures . For color pictures of larvae and adults , a KEYENCE VHX1000 digital microscope was used . For the pictures of the wings and the histological sections of the eyes , a Zeiss Axiophot Microscope was used . In all the quantifications , Student’s t-test ( two-tailed ) was used to test for significance . In each experiment , a minimum of nine individuals was measured for each genotype . Significance is indicated in the Figures using the following symbols: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , n . s . , not significant . Error bars represent the standard deviation . All measurement data are provided in the Source Data Files accompanying the Figures including statistical analyses ( Figure 1—source data 1 , Figure 2—source data 1 , Figure 3—source data 1 , Figure 4—source data 1 , Figure 6—source data 1 , Figure 7—source data 1 ) . | Mutations are permanent changes to a cell’s genome . If one or more mutations result in a cell proliferating in an unregulated manner , it is referred to as a cancer cell . The generation of cancer cells is a relatively common occurrence within organisms , but these rogue cells are generally recognized and destroyed by the organism’s immune system . However , when the immune system fails to identify and eliminate cancer cells , they can proliferate to form malignant , life-threatening tumors . Mutations in a gene called PTEN are often found within cells that develop into cancerous tumors . This gene is normally expressed as a protein that is involved in the regulation of cell division , preventing cells from growing and dividing too quickly . However , when the protein PTEN is absent or non-functional , cells experience enhanced growth , proliferation , and survival . Such cells are also thought to be resistant to nutrient restriction , but the mechanism responsible for this resistance is not well understood . Here , Nowak et al . investigate the behavior of cells lacking PTEN in a fly model under a variety of nutritional conditions . When the supply of nutrients is limited , cells lacking PTEN shift resources from cell growth to cell multiplication . This appears to allow PTEN-deficient cells to outcompete neighboring wild-type cells; Nowak et al . suggest these rapidly proliferating cells are capable of effectively hoarding nutrient stores , both in their immediate vicinity and organism-wide . Further studies that focus on changes in gene expression may be able to uncover the mechanism that allows PTEN-deficient cells to proliferate when nutrients are restricted . Moreover , by shedding light on a factor that has an important influence on tumor development , these results may have implications for cancer treatment strategies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2013 | Nutrient restriction enhances the proliferative potential of cells lacking the tumor suppressor PTEN in mitotic tissues |
Bacteria have to avoid recognition by the host immune system in order to establish a successful infection . Peptidoglycan , the principal constituent of virtually all bacterial surfaces , is a specific molecular signature recognized by dedicated host receptors , present in animals and plants , which trigger an immune response . Here we report that autolysins from Gram-positive pathogenic bacteria , enzymes capable of hydrolyzing peptidoglycan , have a major role in concealing this inflammatory molecule from Drosophila peptidoglycan recognition proteins ( PGRPs ) . We show that autolysins trim the outermost peptidoglycan fragments and that in their absence bacterial virulence is impaired , as PGRPs can directly recognize leftover peptidoglycan extending beyond the external layers of bacterial proteins and polysaccharides . The activity of autolysins is not restricted to the producer cells but can also alter the surface of neighboring bacteria , facilitating the survival of the entire population in the infected host .
Peptidoglycan ( PGN ) is a macromolecule composed of long glycan strands , cross-linked by short peptides , which surrounds most bacterial cells and forms a load-bearing mesh that sustains their shape . Enlargement and remodeling of the PGN mesh during bacterial growth and division requires not only the synthesis of new material , catalyzed by penicillin-binding proteins ( PBPs ) , but also PGN hydrolysis to allow insertion of new material , carried out by PGN autolysins ( Vollmer et al . , 2008 ) . Despite its essential role , PGN may also be viewed as an ‘Achilles heel’ of bacteria during infection , as different hosts have specialized receptors that recognize PGN as a pathogen-associated molecular pattern ( PAMP ) and initiate an inflammatory response to eliminate invading bacteria . Examples of these receptors include LysM proteins in plants ( Willmann et al . , 2011 ) , intracellular NOD-like receptors ( NLRs ) and Toll-like receptors ( TLRs ) in mammals ( Chaput and Boneca , 2007 ) , and peptidoglycan recognition proteins ( PGRPs ) in insects and mammals ( Dziarski and Gupta , 2006 ) . PGRPs , originally isolated due to their high affinity to PGN ( Kang et al . , 1998 ) , are used by Drosophila flies to distinguish between Gram-negative and Gram-positive bacteria , directly at the level of PGN detection . This is achieved through specific PGRPs: PGRP-LC specifically recognizes DAP-type PGN , usually found in Gram-negative bacteria and Gram-positive bacilli , and activates the Imd pathway ( Leulier et al . , 2003 ) , while PGRP-SA recognizes lysine-type PGN , which surrounds most Gram-positive bacteria , and activates the Toll pathway ( Lemaitre and Hoffmann , 2007 ) . Activation of either pathway results in a series of multiple defense reactions that include the production and secretion of antimicrobial peptides into the hemolymph of flies ( Lemaitre and Hoffmann , 2007 ) . In mammals , PGRPs can act as antibacterial agents due to their bactericidal and/or bacteriostatic activity , mediated by PGN hydrolytic activity ( e . g . PGLYRP-2 [Dziarski and Gupta , 2006] ) or by the binding of PGRPs to targets on the bacterial cell surface , which causes the activation of specific bacterial two-component systems , resulting in bacteria killing through a mechanism that includes membrane depolarization and production of hydroxyl radicals ( Kashyap et al . , 2011 ) . Bacterial PGN is concealed by an outer membrane in Gram-negative bacteria , or by layers of proteins and glycopolymers in Gram-positive bacteria . It is therefore usually assumed that an infected organism only recognizes PGN in the form of fragments released into the surrounding medium by the activity of different bacterial or host enzymes ( Nigro et al . , 2008 ) . However , it has been recently shown that PGRP-SA can directly bind PGN at the bacterial surface in conditions such as the absence of wall teichoic acids ( WTAs ) ( Atilano et al . , 2011 ) . Therefore , it is possible that bacteria may have developed different strategies to prevent host receptors from binding PGN on the bacterial surface , thus avoiding detection by the host innate immune system . In order to look for molecules with a role in preventing bacterial recognition by the host , we used PGRP-SA , a host receptor circulating in the hemolymph of Drosophila melanogaster , to test if specific proteins involved in the metabolism of bacterial cell wall were required to conceal the PGN at the surface of the Gram-positive pathogenic bacterium Staphylococcus aureus . We have identified the major autolysin Atl as an essential protein for concealing the bacterial PGN at the cell surface from host detection .
In order to identify factors that Gram-positive bacteria use to conceal their PGN present at the bacterial cell surface from host recognition , we have constructed S . aureus null mutants lacking non-essential genes involved in PGN metabolism and determined the ability of a fluorescent derivative of PGRP-SA ( mCherry_PGRP-SA ) to bind to their surfaces ( Figure 1 ) . Specifically , we tested mutants constructed in NCTC8325-4 S . aureus strain , expressing altered levels of autolysins ( ΔarlR ) , lacking the major autolysin ( Δatl ) or factors that modulate autolysin activity ( ΔfmtA ) , producing non-O-acetylated peptidoglycan with increased susceptibility to lysozyme ( Δoat ) , or producing altered WTAs lacking attached β-O-GlcNAc ( ΔtarS ) or D-alanyl residues ( ΔdltA ) . A mutant lacking the major autolysin Atl , a PGN hydrolase , was identified as the most severely impaired in its ability to avoid recognition by mCherry_PGRP-SA . 10 . 7554/eLife . 02277 . 003Figure 1 . Staphylococcus aureus cells lacking the major autolysin Atl are better recognized by mCherry_PGRP-SA . Exponentially growing bacteria from the parental NCTC8325-4 ( NCTC ) , and its mutant strains lacking genes involved in cell wall metabolism ( see main text for details ) were incubated with mCherry_PGRP-SA in 96-well plates . Cells were pelleted by centrifugation and unbound protein was washed with PBS . mCherry_PGRP-SA bound to each bacterial strain was quantified using a fluorescent image analyzer ( n≥ 10 wells for each strain ) . Results are shown as the median with 25% and 75% inter-quartile range . The dashed line represents the median value obtained when bacteria were absent . Statistically significant differences ( p<0 . 001 , indicated by asterisks ) were observed only between mCherry_PGRP-SA binding to the parental strain and mutants NCTCΔtagO and NCTCΔatl . mCherry_PGRP-SA binding to bacterial cells was also imaged by fluorescence microscopy . The top panels show phase-contrast images of bacteria ( white scale bar represents 1 µm ) and the bottom panels show the mCherry_PGRP-SA binding to their surface . Binding of the protein was observed for NCTCΔtagO and NCTCΔatl , with the later exhibiting the highest binding . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 003 PGN modifications , namely the attachment of teichoic acids or the degree of PGN polymerization ( defined as the ratio between the amount of polymerized muropeptides and the amount of monomeric muropeptides present in the PGN macromolecule ) , have been previously associated with changes in the recognition of S . aureus PGN by host receptors ( Filipe et al . , 2005; Atilano et al . , 2011 ) . However , an S . aureus atl null mutant produces a PGN with a muropeptide composition similar to the parental strain ( Figure 2A ) , that is , it shows no increase in the amount of polymerized muropeptides , which were previously reported to be better inducers of an innate immune response than monomeric muropeptides ( Filipe et al . , 2005 ) . Moreover , the S . aureus atl null mutant produces WTAs ( Figure 2B ) , indicating that the mechanism to conceal PGN dependent on the presence of atl is new , and different from that previously reported for a WTA mutant ( Atilano et al . , 2011 ) . In agreement , the simultaneous deletion of tagO and atl had a synergistic effect on exposing the surface PGN to host recognition ( Figure 2C ) . 10 . 7554/eLife . 02277 . 004Figure 2 . Better recognition of Staphylococcus aureus NCTCΔatl by PGRP-SA is not mediated by alterations in peptidoglycan muropeptide composition or lack of wall teichoic acids production . ( A ) Staphylococcus aureus NCTCΔatl mutant has a similar peptidoglycan ( PGN ) muropeptide composition to the parental strain NCTC8325-4 , as seen by HPLC analysis of mutanolysin-digested PGN . Roman numerals I to V indicate muropeptide species form monomers to pentamers , respectively . ( B ) NCTCΔatl mutant produces wall teichoic acids ( WTAs ) , as shown by PAGE analysis of surface WTAs from NCTC8325-4 , NCTCΔatl , and NCTCΔtagO ( which lacks WTAs ) . ( C ) Deletion of both the tagO and atl genes had a synergistic effect on PGN exposure to the host receptor PGRP-SA , indicating independent mechanisms of WTA and Atl in PGN concealment . S . aureus parental strain NCTC8325-4 and mutant strains NCTCΔtagO , NCTCΔatl , and NCTCΔatlΔtagO were incubated with mCherry_PGRP-SA in 96-well plates . The average amount of mCherry_PGRP-SA bound to bacteria in each well was quantified using a fluorescent image analyzer ( n = 10 wells , for each strain ) , and is represented as the median with 25% and 75% inter-quartile range . The dashed line represents the median value obtained with control samples ( no bacteria added ) . Statistically significant differences ( p<0 . 001 ) are indicated by asterisks and were observed between mCherry_PGRP-SA binding to the parental strain and each mutant as well as between mutants . mCherry_PGRP-SA binding to the cells was also confirmed by fluorescence microscopy ( bottom ) . Gray panels show phase-contrast images of bacterial cells ( white scale bar represents 1 μm ) and black panels show mCherry_PGRP-SA binding . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 004 The atl gene encodes a polypeptide that is cleaved into two proteins capable of digesting PGN: an amidase ( AM ) , which cleaves the amide bond between peptides and glycans , and a glucosaminidase ( GL ) , which cleaves glycan strands ( Figure 3A ) . To determine which one of these activities was required for the role of Atl in PGN concealment , we constructed S . aureus strains expressing , from the atl native chromosomal locus , Atl mutants with impaired amidase activity ( AMH265A [Bose et al . , 2012] ) , glucosaminidase activity ( GLE1128A [Bose et al . , 2012] ) , or both ( AtlH265A/E1128A double mutant ) and confirmed the absence of the expected hydrolytic activity in a zymogram ( Figure 3B ) . 10 . 7554/eLife . 02277 . 005Figure 3 . Amidase and glucosaminidase activities limit mCherry_PGRP-SA binding to the bacterial cell surface . ( A ) Representation of the post-translational cleavage ( black arrows ) of the atl encoded protein . The processed amidase ( AM , purple arrow ) releases peptidoglycan ( PGN ) stem peptides by cutting the bond between the stem peptides ( black circles ) and the N-acetylmuramic acid residue ( open square ) . The glucosaminidase ( GL , blue arrow ) releases muropeptides by cutting the glycosidic linkage between N-acetylmuramic acid ( open square ) and N-acetylglucosamine ( open circle ) . For simplicity , only some examples of AM and GL cleavage sites are indicated by arrows . ( B ) Zymogram analysis of the activity of autolysins extracted from different Staphylococcus aureus atl mutant strains expressing inactive amidase ( AMH265A ) , inactive glucosaminidase ( GLE1128A ) , or both ( AMH265A GLE1128A ) . Top: Cells were harvested at mid-exponential phase , and protein extracts were prepared and run in an SDS–PAGE gel containing Micrococcus luteus cell walls . PGN hydrolytic activity is seen as clear bands in the stained gel and showed that S . aureus encoding mutant proteins AM or GM were lacking only the expected activity . Middle: A similar analysis was carried out with purified proteins AM , GL , and their inactive forms , AMH265A and GLE1128A , respectively , showing that mutant proteins were not active . Bottom: SDS–PAGE of the protein samples loaded into the zymograms confirmed that similar amount of proteins were used . ( C ) S . aureus parental and atl mutant strains were incubated for mCherry_PGRP-SA in 96-well plates . mCherry_PGRP-SA bound to each bacterial strain is represented as the median with 25% and 75% inter-quartile range ( n = 50 wells ) . A significant increase of mCherry_PGRP-SA binding , relative to the parental strain , was observed with S . aureus mutants expressing AMH265A and AMH265A GLE1128A ( p<0 . 0001 ) , but not the GLE1128A mutant . mCherry_PGRP-SA binding to S . aureus parental and atl mutant strains was also confirmed by fluorescence microscopy . Gray panels are phase-contrast images of bacterial cells ( white scale bar represents 1 μm ) and black panels show mCherry_PGRP-SA binding . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 005 When the atl mutants were incubated with mCherry-PGRP-SA , both the mutant having only glucosaminidase activity ( AMH265A mutant ) and the mutant having only amidase activity ( GLE1128A mutant ) were still able to avoid strong PGRP-SA binding , indicating that either amidase or glucosaminidase enzymatic activity was sufficient to impair recognition of S . aureus by PGRP-SA ( Figure 3C ) . Only when both activities were absent ( AtlH265A/E1128A double mutant ) was mCherry-PGRP-SA capable of easily recognizing the bacterial cell surface ( Figure 3C , D ) , albeit to lower levels than those observed for the atl null mutant . This difference may be due to residual amidase or glucosaminidase activity in the S . aureus AtlH265A/E1128A double mutant , not detectable in a zymogram . The proposed role for the amidase and glucosaminidase activities in concealing bacteria was confirmed using purified enzymes . Pre-incubation of S . aureus atl null mutant cells with either AM or GL ( approximately 0 . 4 μM ) completely abolished mCherry_PGRP-SA binding to the bacterial surface ( Figure 4A ) . When both enzymes were added simultaneously to the atl null mutant cells , lower concentrations of each enzyme ( <0 . 02 μM ) were sufficient to conceal bacteria from the host immune receptor ( Figure 4B ) . This is in agreement with the synergy in PGRP-SA binding observed for the S . aureus AtlH265A/E1128 double mutant ( Figure 2C ) . 10 . 7554/eLife . 02277 . 006Figure 4 . Combination of purified staphylococcal AM and GL completely abolishes binding of mCherry_PGRP-SA to the surface Staphylococcus aureus atl null mutant cells . ( A ) mCherry_PGRP-SA binding to NCTCΔatl bacteria pre-incubated with purified AM and GL ( 0 . 37 µM and 0 . 44 µM , respectively ) showed that both enzymes can impair binding of mCherry_PGRP-SA to the cell surface . PBS was used as negative control . ( B ) The use of lower concentrations of purified AM and GL in combination , to treat NCTCΔatl Staphylococcus aureus cells ( final concentrations of 7 . 9 nM and 18 . 7 nM , respectively ) was more effective than the single use of AM ( final concentration of 7 . 9 nM ) or GL ( final concentration of 18 . 7 nM ) in preventing binding of mCherry-PGRP-SA to the surface of NCTCΔatl . Gray panels are phase-contrast images of bacterial cells ( white scale bar represents 1 μm ) and black panels show mCherry_PGRP-SA binding . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 006 We then questioned if the mechanism used by autolysins to conceal PGN at the bacterial surface was dependent on their lytic activity . An alternative hypothesis could be that amidases , or other proteins capable of binding PGN , prevented binding of PGRP-SA by competing for the same PGN substrate , given that the structural fold of PGRPs closely resembles the fold of amidases ( Zoll et al . , 2010 ) . To distinguish between the two hypotheses , we first incubated S . aureus atl null mutant cells with purified AM or with GL , and then washed them thoroughly to ensure complete removal of the proteins ( Figure 5 ) . This procedure did not increase recognition by PGRP-SA , showing that the activity of the autolysins , and not their physical presence , was required for protection . 10 . 7554/eLife . 02277 . 007Figure 5 . Activity , and not the presence of the Atl products , is required to avoid bacterial surface recognition by mCherry_PGRP-SA . ( A ) NCTCΔatl cells were incubated with purified staphylococcal AM , GL , or PBS ( negative control ) and were either washed to clear these proteins from the cell surface , or unwashed to keep the proteins attached . Bacterial cells were then labeled with mCherry_PGRP-SA and imaged by fluorescence microscopy . In both cases ( washed and unwashed ) , mCherry_PGRP-SA was unable to bind S . aureus NCTCΔatl cells , showing that the physical presence of the enzyme is not required for this effect . Gray panels are phase-contrast images of bacterial cells ( white scale bar corresponds to 1 µm ) and black panels are fluorescence microscopy images showing binding of mCherry_PGRP-SA to the surface of bacteria . ( B ) To confirm that AM ( left ) and GL ( right ) were absent from the washed samples , a dot-blot assay using anti-His antibody , which recognizes the His-tagged AM and GL enzymes , was performed before the addition of mCherry_PGRP-SA . Washed cells ( lanes C ) showed no presence of protein , whereas in unwashed cells ( lanes B ) the presence of each individual lytic enzyme was detected ( detection limit for AM and GL is lower than 0 . 37 µg/ml , as seen in lanes A , which corresponds to the pure protein loaded onto the membrane at different concentrations ) . A sample of cells to which no protein was added was used as a negative control ( lanes D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 007 The results described above show that amidase or glucosaminidase activities can efficiently protect S . aureus cells from direct recognition by the Drosophila PGN receptor PGRP-SA . Does this protection confer an advantage to S . aureus cells during infection in the Drosophila model ? Indeed , Tabuchi et al . have previously shown that S . aureus atl mutants are impaired in their virulence ( Tabuchi et al . , 2010 ) . We have confirmed that the atl mutant used in this work , when compared to the parental S . aureus strain NCTC8325-4 , was also severely impaired in its ability to kill wild type Drosophila , both the 25174 strain , recently established from wild caught flies ( Mackay et al . , 2012 ) , as well as the yw strain , which is the parental lineage used to generate the PGRP seml mutant ( Michel et al . , 2001; Figure 6A , B ) . Furthermore , the atl mutant triggered a stronger induction of drosomycin expression ( frequently used as a read-out for Toll activity ) when normalized for the same number of infecting bacteria ( Figure 6C , D ) . Importantly , our data showed that survival of Drosophila infected with an S . aureus atl mutant was dependent upon the presence of a functional PGRP-SA ( Figure 6E ) . Furthermore , these mutant bacteria were capable of propagating similarly to the parental S . aureus strain in flies that lacked a functional PGRP-SA ( Figure 6F ) , showing that impaired virulence is not due to a lower proliferation rate of the atl mutant . 10 . 7554/eLife . 02277 . 008Figure 6 . PGRP-SA is required to control infection by a Staphylococcus aureus mutant lacking the major autolysin Atl . ( A ) Estimated survival curves of wild type ( WT ) 25174 flies infected with Atl producer NCTC8325-4 and NCTCΔatl strains were statistically different ( p<0 . 005 ) and are indicated by asterisks . Absence of Atl activity in the S . aureus Δatl mutant resulted in bacteria that have a decreased ability to kill WT flies . Survival of infected flies ( n = 75 ) was monitored at 12 h intervals for 3 days . ( B ) Estimated survival curves of WT yw flies , the parental lineage used to construct the peptidoglycan recognition protein ( PGRP ) seml mutant used in this study , were also produced as described above and gave similar results ( p<0 . 0001 , indicated by asterisks ) . ( C ) The number of bacteria harvested from infected flies at different time points after infection , during survival assays , was determined by plating in S . aureus growth medium and counting CFUs . ( D ) Drosomycin ( Drs ) expression was determined by qPCR at different time points after infection and is shown after normalization for the number of infecting bacteria present in flies ( panel C ) . A stronger induction of drosomycin expression was observed in flies injected with NCTCΔatl bacteria , 12 h after infection . ( E ) PGRP-SA mutant flies succumbed equally well to infection by NCTC832-4 and NCTCΔatl bacteria ( p>0 . 05 ) , showing that in the absence of a functional PGRP-SA , NCTCΔatl bacteria can proliferate and kill the infected host . ( F ) Similar numbers of NCTC832-4 and NCTCΔatl bacteria were present in PGRP-SA mutant flies 17 h after infection ( T17 ) , confirming that the atl mutant proliferated as well as the parental bacterial strain in seml flies . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 008 Two , non-mutually exclusive , hypotheses can explain the decreased virulence of the atl mutant . ( 1 ) S . aureus wild type bacteria produce amidases that cleave the stem peptide moiety of soluble PGN fragments . Given that stem peptides are required for binding of PGRPs to PGN , amidase activity would therefore reduce the inflammatory activity of the soluble PGN . In contrast , the atl mutant would release intact muropeptides , easily detected by PGRP-SA , and therefore would be unable to evade detection by the innate immune system . ( 2 ) Amidases produced by S . aureus wild type bacteria would shave and remove the most accessible PGN fragments at the surface of bacteria , eliminating the binding sites for PGRPs . As the atl mutant does not produce the Atl amidase , it would have extending PGN fragments at its surface , which would be easily detected by PGRP-SA , inducing an immune response . If the first hypothesis is correct , inactivation of the amidase activity , but not of the glucosaminidase ( which cleaves the glycans and therefore releases intact , inflammatory PGN fragments ) , should result in decreased virulence . Alternatively , if the second hypothesis is mainly correct , inactivation of either the amidase or the glucosaminidase activity should result in decreased virulence , as both activities can remove fragments of detectable PGN ( containing peptides ) from the bacterial surface . We therefore tested the virulence of bacteria with either impaired amidase activity ( AMH265A ) or impaired glucosaminidase activity ( GLE1128A ) and observed that both had a reduced ability to kill flies ( Figure 7A ) in accordance with the second hypothesis . Moreover , injection of seml Drosophila with atl mutant bacteria pre-coated with mCherry_PGRP-SA resulted in increased fly survival ( Figure 7B ) , again favoring a role for PGRP-SA in the recognition of PGN directly at the bacterial surface . Nevertheless , the amount of PGRP-SA present at the surface of the injected bacteria may not have been sufficient to result in full complementation of the ability of seml flies to survive infection . This is in accordance with the hypothesis that a continuous and high expression of PGRP-SA during the entire infection process is required to ensure survival of the infected host ( De Gregorio et al . , 2001 ) . 10 . 7554/eLife . 02277 . 009Figure 7 . Lack of either Atl amidase or glucosaminidase activity , which leads to different types of released peptidoglycan fragments , results in decreased virulence of Staphylococcus aureus . ( A ) Estimated survival curves of wild type ( WT ) 25174 flies infected with atl mutant strains impaired in amidase activity ( AMH265A ) , glucosaminidase activity ( GLE1128A ) , or both ( AMH265AGLE1128A ) , were statistically different ( p<0 . 05 ) from survival curves of flies infected with the Atl producer NCTC8325-4 strain , but not with the NCTCΔatl strain . Statistically significant differences ( *p=0 . 01 and **p<0 . 001 ) are indicated by asterisks and were observed between the estimated survival curve of flies infected with the parental bacteria strain NCTC8325-4 and each survival curve of flies infected with the different atl mutant strains . WT 25174 flies infected with the different atl mutant strains succumbed in a similar manner ( p>0 . 05 ) . ( B ) Estimated survival curve of PGRP-SA seml mutant flies infected with NCTCΔatl bacteria pre-coated with mCherry-PGRP-SA showed that direct binding of the peptidoglycan ( PGN ) host receptor to the surface resulted in an increase of resistance of seml mutant flies to bacterial infection . Statistically significant differences ( p<0 . 0001 ) are indicated by asterisks and were observed between the two estimated survival curves . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 009 Secreted autolysins not only bind to bacterial cell surfaces but they can also be found in the growth medium ( Pasztor et al . , 2010 ) . This raises an interesting hypothesis: can secreted bacterial autolysins protect neighboring bacteria ? If this was true , autolysin producers could in theory protect an entire bacterial population from recognition by the host . To test this hypothesis we collected and filtered the supernatant from a culture of S . aureus parental strain NCTC8325-4 ( Atl producer ) and incubated S . aureus atl null mutant cells in this medium for 30 min . The atl-encoded proteins secreted into the supernatant by the parental strain were capable of modifying the surface of the atl null mutant cells , consequently abolishing their recognition by PGRP-SA ( Figure 8A , B ) . This was not observed when supernatant from a culture of S . aureus atl null mutant was used ( Figure 8A , B ) , showing that atl products ( and not other secreted PGN hydrolases ) are responsible for PGN concealment . 10 . 7554/eLife . 02277 . 010Figure 8 . Atl products secreted by Staphylococcus aureus cells protect atl null mutant cells from PGRP-SA recognition and allow S . aureus to establish a successful infection in Drosophila . ( A ) NCTCΔatl Staphylococcus aureus cells were incubated with TSB medium ( control ) or with supernatants ( sterilized by filtration ) from cultures of NCTC8325-4 ( NCTCsup , containing atl encoded products ) or NCTCΔatl ( Δatlsup ) strains . After washing with PBS , cells were mixed with mCherry_PGRP-SA in 96-well plates . Binding of the protein to the cells was determined as described in Figure 1 . mCherry_PGRP-SA binding to NCTCΔatl cells pre-incubated with supernatant from Atl producer strain NCTC8325-4 was 100-fold lower than to the same cells pre-incubated with supernatant from a culture of NCTCΔatl mutant . The dashed line represents the median value obtained with no bacteria . Statistically significant differences ( p<0 . 001 ) are indicated by asterisks . ( B ) Similar results were observed by fluorescence microscopy of NCTCΔatl S . aureus cells incubated with filtered supernatants from NCTC8325-4 or NCTCΔatl cells ( TSB medium was used as negative control ) and subsequently labeled with mCherry_PGRP-SA . Only the supernatant from Atl producer NCTC8325-4 modified the surface of NCTCΔatl cells and limited mCherry_PGRP-SA binding . The top panels are phase-contrast images of bacterial cells ( white scale bar represents 1 µm ) and the bottom panels show the mCherry_PGRP-SA binding to the bacterial surface . ( C ) Estimated survival curves for wild type Drosophila infected with S . aureus NCTC8325-4 and NCTCΔatl strains that were pre-incubated with bacteria-free supernatant from overnight cultures of each of the S . aureus strains . Flies were infected with approximately 100 S . aureus CFUs and fly survival was assessed every 12 h over 3 days . As expected , survival curves for flies infected with NCTC8325-4 bacteria pre-incubated with both supernatants were statistically indistinguishable ( p>0 . 05 ) . Treatment of NCTCΔatl cells with both supernatants resulted in distinct survival curves ( p<0 . 001 , indicated by asterisks ) , showing that NCTCΔatl cells recover the ability to kill flies if pre-incubated with supernatant from a culture of the Atl producer NCTC8325-4 strain . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 010 Importantly , we observed that atl mutants regained the ability to kill flies if pre-treated with a supernatant from a culture of the Atl producer NCTC8325-4 but not if pre-treated with supernatant from a culture of the atl mutant ( Figure 8C ) . Moreover , previously bound PGRP-SA could be removed by autolysin activity present in the medium , as seen by time-lapse microscopy of S . aureus atl null mutant cells initially covered by bound mCherry_PGRP-SA and then incubated with supernatant containing amidase and glucosaminidase enzymes ( Figure 9 ) . 10 . 7554/eLife . 02277 . 011Figure 9 . Time-lapse microscopy showing that atl encoded proteins can mediate release of mCherry_PGRP-SA previously bound to the surface of atl null mutant Staphylococcus aureus cells . NCTCΔatl cells , labeled with mCherry-PGRP-SA , were placed on top of a thin layer of agarose containing filter-sterilized supernatant of cultures of NCTCΔatl mutant ( A , Video 1 ) or NCTC8325-4 parental strain ( B , Video 2 ) , and observed by fluorescence microscopy in a time-lapse experiment . The supernatant from the Atl producer parental strain ( B ) , in contrast to the supernatant from the atl null mutant ( A ) , triggered the release of mCherry-PGRP-SA previously attached to the bacterial cell surface . Gray panels are phase-contrast images of bacterial cells ( white scale bar corresponds to 1 µm ) and black panels are fluorescence microscopy images showing binding of mCherry_PGRP-SA to the surface of bacteria . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 01110 . 7554/eLife . 02277 . 012Video 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 01210 . 7554/eLife . 02277 . 013Video 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 013 Together , these data suggest that the autolysins encoded by atl can protect not only the producer cells but also a population of invading bacteria from detection by the immune system during infection , enabling it to kill the infected host . The S . aureus strain used in this work , NCTC8325-4 , is susceptible to most antibiotics and has limited interest from a clinical point of view . We therefore tested if the role of atl-encoded enzymes in making PGN inaccessible to the host was conserved in other strains , namely in methicillin-resistant S . aureus ( MRSA ) strains . MRSA strains are a leading cause of bacterial infections in hospitals and an important cause of community-acquired ( CA ) bacterial infections in the United States . MW2 is a particularly virulent MRSA strain that caused the earliest reported cases of CA-MRSA infection in the United States ( Chambers and Deleo , 2009 ) . Deletion of the atl gene in MW2 led to higher binding of mCherry_PGRP-SA , that is , to improved recognition of the bacterial surface ( Figure 10A ) . More importantly , MW2Δatl cells were severely impaired in virulence , as demonstrated by the decreased capacity to kill Drosophila ( Figure 10B ) . 10 . 7554/eLife . 02277 . 014Figure 10 . Atl encoded proteins are required to conceal CA-MRSA virulent strains from the Drosophila immune system . ( A ) Staphylococcus aureus CA-MRSA MW2 strain and its atl null mutant , MW2Δatl , were incubated with mCherry_PGRP-SA in 96-well plates . The average amount of mCherry_PGRP-SA bound to bacteria in each well was quantified using a fluorescent image analyzer ( n = 10 wells , for each strain ) , and is represented as the median with 25% and 75% inter-quartile range . The dashed line represents the median value obtained with control sample ( no bacteria added ) . mCherry_PGRP-SA binding to MW2 cells was significantly different ( p<0 . 05 , indicated by asterisks ) from the binding to MW2Δatl . mCherry_PGRP-SA binding to the bacterial cell surface of MW2 and MW2Δatl bacteria was also confirmed by fluorescence microscopy ( bottom ) . Gray panels are phase-contrast images of bacterial cells ( white scale bar corresponds to 1 μm ) and black panels show mCherry_PGRP-SA binding . ( B ) Estimated survival curves for wild type ( WT ) and PGRP-SA deficient ( seml ) flies infected with S . aureus MW2 and MW2Δatl . MW2Δatl is impaired in its ability to kill WT flies , showing that lack of Atl strongly reduces MW2 pathogenicity . PGRP-SA deficient flies are killed in less than 24 h by both S . aureus strains , showing that flies control MW2Δatl infection in a PGRP-SA dependent manner . Statistically significant differences ( p<0 . 0001 ) are indicated by asterisks and were observed between the two estimated survival curves . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 014 The fact that PGN sensing is an evolutionarily conserved recognition mechanism in innate immunity , suggests that the strategy used by S . aureus to conceal PGN from the host could be effective across different animal phyla and even used by other bacteria . We therefore tested if we could observe this phenomenon in another Gram-positive pathogenic bacteria , namely Streptococcus pneumoniae , a frequent cause of pneumonia in developed countries and a major cause of infant mortality by septicemia in developing countries ( Kadioglu et al . , 2008 ) . Lack of LytA , the major amidase of S . pneumoniae , resulted in bacteria that were better recognized by mCherry_PGRP-SA , a phenotype that was reversed when LytA was expressed from a plasmid ( Figure 11A ) . Interestingly , it is possible that autolysins can participate in inter-species protection , as S . pneumoniae LytA , an amidase that releases PGN stem peptides , and Streptomyces globisporus mutanolysin , a commercially available muramidase that releases muropeptides , were capable of protecting S . aureus atl null mutant cells from PGRP-SA recognition ( Figure 11B ) . 10 . 7554/eLife . 02277 . 015Figure 11 . LytA activity can prevent Streptococcus pneumoniae and Staphylococcus aureus recognition by mCherry_PGRP-SA . ( A ) Streptococcus pneumoniae parental strain R36A , a lytA null mutant ( R36AΔlytA ) , and a complemented strain , expressing LytA from a replicative plasmid ( R36AΔlytAplytA ) , were incubated with mCherry_PGRP-SA in 96-well plates . The average amount of mCherry_PGRP-SA bound to bacteria in each well was quantified using a fluorescent image analyzer ( n = 10 wells , for each strain ) , and is represented as the median with 25% and 75% inter-quartile range . The dashed line represents the median value obtained with control sample ( no bacteria added ) . mCherry_PGRP-SA binding to R36AΔlytA was significantly higher ( p<0 . 05 , indicated by asterisks ) than that observed for the parental strain or R36AΔlytAplytA strain , showing that deletion of the lytA gene in S . pneumoniae increases surface recognition by mCherry_PGRP-SA . Binding of mCherry_PGRP-SA to S . pneumoniae cells was also imaged using fluorescence microscopy ( bottom ) . Gray panels are phase-contrast images of bacterial cells ( white scale bar represents 1 µm ) and black panels show the mCherry_PGRP-SA binding to bacteria . ( B ) S . aureus NCTCΔatl cells were incubated with purified pneumococcal amidase ( LytA ) or with the commercially available mutanolysin , a muramidase from Streptomyces globisporus capable of degrading peptidoglycan into its muropeptide components . Bacterial cells were washed to clear these proteins from the cell surface , labeled with mCherry_PGRP-SA , and imaged by fluorescence microscopy , which showed that both pneumococcal LytA amidase and mutanolysin can prevent detection of NCTCΔatl S . aureus cells by mCherry_PGRP-SA . DOI: http://dx . doi . org/10 . 7554/eLife . 02277 . 015 These results suggest a general role for autolysins , such as amidases , in protecting bacteria from host recognition .
We propose that bacteria have evolved various independent mechanisms to prevent direct recognition by PGN host receptors . We have previously shown that the presence of glycopolymers , such as WTAs , at the bacterial surface contributes to S . aureus evasion from the Drosophila innate immune system ( Atilano et al . , 2011 ) . We now show that Atl also protects bacteria from host recognition but by a different mechanism , given that simultaneous deletion of tagO ( essential for teichoic acid biosynthesis ) and atl had a synergistic effect on exposing the surface PGN to host recognition , when compared to individual deletion of each of these genes , as shown in Figure 2C . This new role for Atl is in accordance with previous reports that have shown in the Drosophila infection model that S . aureus atl mutants are impaired in their virulence ( Tabuchi et al . , 2010 ) . Bacterial autolysins , or murein hydrolases , are present in most bacteria and have been recognized as important players in bacterial metabolism . Autolysins are required , among other roles , to cleave the pre-existing PGN structure so that new PGN subunits are incorporated during synthesis , to ensure daughter cell separation after complete bacterial division or to prompt antibiotic-induced lysis of bacterial cells ( Vollmer et al . , 2008; Uehara and Bernhardt , 2011 ) . Autolysins have also been recognized as important virulence factors , as they may act as adhesins ( Hell et al . , 1998 ) or be required for biofilm formation ( Vollmer et al . , 2008; Uehara and Bernhardt , 2011 ) . The new role for Atl uncovered in this work was surprising , since it has been previously suggested that PGN fragments , released by host enzymes capable of degrading PGN or shed by dividing bacteria during the course of infection , may facilitate host recognition in the context of both NOD receptors and insect PGRPs ( Nigro et al . , 2008 ) . However , we have now shown that it is precisely this autolysin-mediated ‘trimming’ of PGN ( which presumably releases PGN fragments ) that renders S . aureus inaccessible to the host . Direct recognition of PGN seems less likely to play a role in mammalian cells , where PGN is recognized by the intracellular pattern-recognition NOD factors , or in the recognition of Gram-negative bacteria , whose PGN is concealed by the outer membrane , by the Drosophila innate immune system . We propose that a major mechanism of Gram-positive bacteria recognition in innate immunity is the direct binding of PGRPs to PGN on whole bacteria , which have dedicated proteins , namely amidases or glucosamidases , to trim exposed fragments of PGN . These fragments may extend beyond the teichoic acids layer and , through their removal , bacteria avoid their own recognition and possibly that of their neighbors , by the host innate immune system . Therefore , targeting bacterial autolysins in order to prevent their activity may constitute a new approach to decrease bacterial virulence and potentiate the effectiveness of the host immune system .
All bacterial strains used in this study are listed in Supplementary file 1 . Escherichia coli strains were grown in Luria–Bertani broth ( LB; Difco , France ) or Luria–Bertani agar ( LA; Difco ) medium at 37°C with aeration . When needed , the antibiotics ampicillin ( Amp; Sigma-Aldrich , Germany ) and erythromycin ( Ery; Sigma-Aldrich ) were added at a final concentration of 100 µg/ml , and kanamycin ( Km; Sigma-Aldrich ) was added at a final concentration of 30 µg/ml . Lactococcus lactis LL108 strain was grown in M17 broth ( Difco ) , supplemented with sucrose ( 0 . 5 M ) and glucose ( 0 . 5% wt/vol ) at 30°C without aeration . Erythromycin was used when required at 100 µg/ml . S . aureus strains were grown at 30°C with aeration in tryptic soy broth ( TSB; Difco ) or on tryptic soy agar ( TSA; Difco ) . Medium was supplemented when required with Ery at 10 µg/ml and/or 5-bromo-4-chloro-3-indolyl β-D-galactopyranoside ( X-Gal; Apollo Scientific , UK ) at 100 µg/ml . S . pneumoniae was grown in C + Y medium at 37°C , without aeration , or on TSA plates supplemented with 5% vol/vol sheep blood ( Probiológica , Portugal ) . Tetracycline ( Sigma-Aldrich ) and Ery were used , when required , at 1 µg/ml and 0 . 25 µg/ml , respectively . Isogenic Drosophila Bloomington #25174 and #6599 ( yw ) were used as wild type flies . PGRP_SAseml flies were used as a PGRP-SA mutant background ( Michel et al . , 2001 ) . Stocks were raised on standard cornmeal-agar medium at 25°C . All plasmids used in this study are listed in Supplementary file 1 and the sequences of the primers used are listed in Supplementary file 2 . The S . aureus oatA and fmtA null mutants were constructed using the integrative vector pORI280 ( Leenhouts et al . , 1996 ) . To delete the oatA gene from the genome of S . aureus NCTC8325-4 we amplified two 1 . 3 Kb DNA fragments from NCTC8325-4 chromosomal DNA , corresponding to the upstream ( primers P1_oatA and P2_oatA ) and downstream ( primers P3_oatA and P4_oatA ) regions of the oatA gene . The upstream fragment was digested with BamHI and NcoI and cloned into pORI280 . The downstream PCR fragment was digested with NcoI and BglII and then cloned into the previous plasmid , which already contained the upstream fragment , producing the plasmid pΔoatA . To delete the fmtA gene , we amplified two DNA fragments of approximately 0 . 9 Kb from S . aureus NCTC8325-4 chromosomal DNA , corresponding to the upstream ( primers P5_fmtA and P6_fmtA ) and downstream ( primers P7_fmtA and P8_fmtA ) regions of the fmtA gene . These fragments were joined by overlap PCR using primers P5_fmtA and P8_fmtA . The resulting PCR product was digested with BamHI and BglII , and cloned in the pORI280 plasmid , producing the plasmid pΔfmtA . Plasmids pΔoatA and pΔfmtA were sequenced and electroporated into the transformable RN4220 strain at 30°C ( using Ery selection ) as previously described ( Veiga and Pinho , 2009 ) . The plasmids were then transduced to NCTC8325-4 using the phage 80α . The resulting strains , carrying the plasmids integrated into the chromosome , were grown in liquid media , without antibiotic and at 37°C , for several generations to allow for loss of the plasmid and , consequently , of the lacZ and erm genes . Different dilutions of the cultures were plated on TSA plates containing X-Gal ( without Ery ) and incubated at 37°C overnight . White colonies were isolated and their sensitivity to erythromycin was confirmed . Absence of the oatA and fmtA genes ( NCTCΔoatA and NCTCΔfmtA ) was shown by PCR and confirmed by sequence analysis of the amplified DNA fragment . All other S . aureus null mutants described were constructed using the thermosensitive vector pMAD ( Arnaud et al . , 2004 ) . In order to generate the arlR null mutant , 0 . 8 Kb PCR fragments of the upstream and downstream regions of the arlR gene were amplified from chromosomal DNA of NCTC8325-4 , using primer pairs P9_arlR/P10_arlR and P11_arlR/P12_arlR . The two PCR products were joined in an overlap PCR reaction using the primers P9_arlR/P12_arlR . The resulting product was digested with NcoI and BglII and cloned into pMAD , giving rise to the pΔarlR plasmid . To delete the atl gene , the pΔatl plasmid was constructed by amplifying 1 Kb fragments from the downstream and upstream regions of the atl gene , using the primer pairs P13_atl/P14_atl and P15_atl/P16_atl . These two fragments were joined by overlap PCR using the primers P13_atl and P16_atl , and the obtained PCR product was digested with BglII and EcoRI and cloned into the pMAD vector , giving rise to the pΔatl plasmid . To delete the tarS gene , we amplified by PCR two DNA fragments of approximately 0 . 6 Kb , corresponding to the upstream ( primers P17_tarS/P18_tarS ) and downstream ( primers P19_tarS/P20_tarS ) regions of the tarS gene . The two fragments were joined by overlap PCR , using primers P17_tarS and P20_tarS . The resulting PCR product was digested with BglII and NcoI and cloned into pMAD vector , producing the plasmid pΔtarS . To delete the dltA gene , two 0 . 55 Kb DNA fragments were amplified by PCR from the genome of S . aureus NCTC8325-4 , corresponding to the upstream ( primers P25_dltA/P26_dltA ) and downstream ( primers P27_dltA/P28_dltA ) regions of the dltA gene . The two fragments were joined by overlap PCR using primers P25_dltA and P28_dltA , and the resulting PCR product was digested with BglII and EcoRI and cloned into the pMAD vector , producing the plasmid pΔdltA . The pΔarlR , pΔatl , pΔtarS , and pΔdltA plasmids were sequenced and electroporated into S . aureus RN4220 strain at 30°C ( using Ery selection ) and then transduced to NCTC8325-4 using phage 80α . Insertion and excision of the plasmids for deletion of the genes ( arlR , atl , tarS , and dltA ) from the NCTC8325-4 chromosome was completed as previously described ( Arnaud et al . , 2004 ) , with the exception of the NCTCΔdltA strain where the last step was performed at 30°C due to the thermosensitive nature of cells lacking the dltA gene . All gene deletions were confirmed by PCR and the resulting strains were named NCTCΔarlR , NCTCΔatl , NCTCΔtarS , and NCTCΔdltA . To make a NCTCΔatlΔtagO double mutant , the pΔtagO plasmid ( Atilano et al . , 2010 ) was transduced into the NCTCΔatl strain using phage 80α , at 30°C due to the thermosensitive nature of cells lacking the tagO gene . The double mutant was identified among the white colonies sensitive to Ery , by PCR and DNA sequencing , as described above . The point mutants in the Atl amidase ( AM ) and Atl glucosaminidase ( GL ) domains were constructed by site-directed mutagenesis . To inactivate the AM domain , the conserved amino acid H265 was exchanged for an alanine ( H265A mutation ) . To inactivate the GL domain , the amino acid residue E1128 was mutated to an alanine ( E1128A mutation ) . To generate the H265A mutation , a 794 bp region upstream from the codon encoding H265 ( primers P29_AtlAM and P30_AtlAM ) and a 1070 bp region downstream from the codon encoding H265 ( primers P31_AtlAM and P32_AtlAM ) were amplified from NCTC8325-4 genomic DNA by PCR . Joining of the up and downstream regions by overlap PCR ( primers P29_AtlAM and P32_AtlAM ) resulted in the amplification of the DNA fragment encoding the mutated Atl amidase domain , which was cloned into the pMAD vector , using BamHI and EcoRI restriction enzymes , to construct plasmid pAMH265A . A similar approach was used to generate the Atl point mutation in the Atl glucosaminidase ( GL ) domain . Briefly , a 659 bp fragment upstream of the codon encoding E1128 ( primers P33_AtlGL and P34_AtlGL ) and a 680 bp fragment downstream of the codon encoding E1128 ( primers P35_AtlGL and P36_AtlGL ) were amplified by PCR . The two fragments were joined by overlap PCR using primers P33_AtlGL and P36_AtlGL , digested and cloned into the EcoRI/BamHI sites of the pMAD vector , yielding plasmid pGLE1128A . After sequencing , plasmids pAMH265A and pGLE1128A were electroporated into RN4220 and then transduced into NCTC8325-4 . Insertion into and excision from the chromosome generated S . aureus point mutants in the amidase and glucosaminidase domains . The strains carrying the Atl mutations were verified by sequencing and were named AMH265A and GLE1128A . A double mutant , named AtlH265A/E1128A , mutated in both the amidase and glucosaminidase domains , was constructed by phage transduction of plasmid pGL1128 into the AMH265A strain , followed by insertion into and excision from the chromosome , as described above . The S . pneumoniae lytA null mutant strain was constructed in the background of R36A strain , using the integrative vector pORI280 . To delete the lytA gene , we amplified two 1 . 0 Kb DNA fragments from R36A chromosomal DNA , corresponding to the upstream ( primers P41_LytA and P42_LytA ) and downstream ( primers P43_LytA and P44_LytA ) regions of the lytA gene . The two fragments were then joined by overlap PCR using primers P41_LytA and P44_LytA . The PCR product was digested with BamHI and EcoRI and cloned into plasmid pORI280 , producing the plasmid pΔlytA . This plasmid was propagated in L . lactis and purified before being transformed into the unencapsulated S . pneumoniae R36A strain . The R36AΔlytA strain was obtained by insertion and excision of the plasmid into the chromosome , as previously described ( Leenhouts et al . , 1996 ) . For the construction of a plasmid expressing LytA ( plytA ) , the lytA gene from S . pneumoniae R36A genome was amplified with primers P45_LytA and P46_LytA , digested with NheI and BglII , and cloned into plasmid pBCSLF001 ( Henriques et al . , 2011 ) . A 3177 bp fragment of the atl gene , missing the pro-peptide sequence , was amplified by PCR using genomic DNA of S . aureus strain COL with primers Pexp1 and Pexp4 , which included BamHI and SalI restriction sites . The DNA fragment was purified , digested with the respective restriction enzymes , and cloned into pET28a ( + ) plasmid ( Novagen , Merck Millipore , UK ) using E . coli DH5α , yielding pET-AMR1R2R3GL . To construct a plasmid expressing the AM recombinant protein , plasmid pET-AMR1R2R3GL was used as the template for site-directed mutagenesis using primers Pexp_stop1 and Pexp_stop2 , to change codon 775 from AAA ( Lys ) to TAA ( stop codon ) resulting in pET-AMR1R2 . To construct a plasmid expressing the GL recombinant protein , primers Pexp2 and Pexp4 with BamHI and SalI restriction sites , were used to amplify a 1446 bp sequence encoding for the GL domain of Atl , using COL genomic DNA as the template . The DNA fragment was purified , digested with the respective restriction enzymes , and cloned into pET28a ( + ) plasmid using E . coli DH5α , yielding pET-R3GL . To construct a plasmid expressing protein AMH265A , pET-AMR1R2 was used as the template for site-directed mutagenesis using primers Pexp_H265A1 and Pexp_H265A2 , to change codon 265 from GTA ( His ) to GCA ( Ala ) . A plasmid expressing protein GLE1128A was constructed similarly , using pET-R3GL as the template for directed mutagenesis with primers Pexp_E1228A1 and Pexp_E1228A2 , changing codon GAA ( Glu ) to GCA ( Ala ) . The amplified plasmids were digested with DpnI prior to transformation into E . coli DH5α , resulting in pET-AMH265A and pET-GLE1228A . The recombinant plasmids ( Supplementary file 1 ) were confirmed by restriction analysis and sequencing and used to transform E . coli BL21 ( DE3 ) strain for protein expression . A functional mCherry-tagged derivative of PGRP-SA protein , mCherry-PGRP-SA , was expressed in E . coli and purified using a cobalt affinity resin ( Clontech ) as previously described ( Atilano et al . , 2011 ) . The AM and GL proteins were expressed as N-terminal 6×His-tag fusion proteins . Expression of the recombinant proteins was performed using an auto-induction based expression method . Cells were grown for 18 h at 37°C with shaking , harvested , and resuspended in 1/10 vol of purification lysis buffer ( 50 mM NaH2PO4 , 10 mM imidazole , 300 mM NaCl , pH 8 . 0 ) containing 10 U/ml of benzonase nuclease ( Novagen , Merck Millipore ) and Complete-Mini Protease Inhibitor Cocktail ( Roche , Switzerland ) . After cell disruption , the lysates were cleared . The recombinant proteins were soluble and therefore present in the supernatant . Protein purification was performed using Ni-NTA agarose columns ( Qiagen , Germany ) under native conditions , according to the manufacturer’s instructions . The recombinant proteins were eluted with purification elution buffer ( 50 mM NaH2PO4 , 250 mM imidazole , 300 mM NaCl , pH 8 . 0 ) . The expression and purification yields were monitored by SDS–PAGE . The most concentrated elution fractions were dialyzed for 16 h in a 3500 MWCO SnakeSkin ( Pierce Biotechnology , Rockford , IL ) at 4°C , against 100 mM Tris , pH 7 . 5 . For the purification of S . pneumoniae LytA amidase , E . coli BL21 ( DE3 ) cells were transformed with the plasmid pGL100 ( Garcia et al . , 1987 ) and incubated overnight at 37°C , with vigorous shaking , in LB supplemented with 100 μg/ml ampicillin and 2% lactose . Cells were harvested by centrifugation , resuspended in 20 mM sodium phosphate buffer , pH 6 . 9 , and broken by sonication . Clarified lysate was applied to DEAE-cellulose resin and incubated at 4°C for 1 h with stirring . Bound protein was washed five times with 20 mM sodium phosphate buffer containing 1 . 5 M NaCl , then eluted in the same buffer containing 2% choline . Protein was dialyzed against 20 mM sodium phosphate buffer , pH 6 . 9 to remove the choline and aliquots were stored at −20°C . To quantify mCherry_PGRP-SA binding to bacterial cells , cultures were grown to mid-exponential phase ( OD600 nm 0 . 5 ) , centrifuged , and resuspended in TSB to an OD of 2 . 5 . Approximately 100 µl of each culture was placed into wells of a 96-well plate and harvested at room temperature ( RT ) for 5 min at 3600×g . Cells were washed and then resuspended in 100 µl of sterile PBS ( pH 6 ) . mCherry_PGRP-SA was added to each well at a final concentration of 60 µg/ml , followed by incubation for 5 min at RT . Cells were then harvested and washed with PBS . The fluorescent signal of the mCherry_PGRP-SA was detected using a 532 nm laser in a Fuji FLA-5100 reader ( FUJIFILM Life Science ) and the fluorescence intensity of each well was determined using ImageJ software ( Abràmoff et al . , 2004 ) . To image the mCherry_PGRP-SA binding to bacterial cell surface by fluorescence microscopy , cells were treated as previously described ( Atilano et al . , 2011 ) , except that mCherry_PGRP-SA was added to a final concentration of ≈300 µg/ml , in a reaction volume of 200 µl . After labeling , bacterial cells were centrifuged , washed twice with 200 µl of PBS , resuspended in 20 µl of PBS , and observed by fluorescence microscopy . Preparation and analysis of crude autolytic enzyme extracts from Staphylococcus aureus is described in detail at Bio-protocol ( Vaz and Filipe , 2015 ) . Crude autolytic enzyme extracts were prepared from S . aureus cultures grown in TSB to OD600 0 . 3 at 30°C . The cells were harvested by centrifugation ( 7000 rpm , 15 min at 4°C ) , and the cell pellet was washed in 15 ml of ice-cold 50 mM Tris–HCl ( pH 7 . 5 ) , 150 mM NaCl . Cells were resuspended in SDS 4% ( 500 ml ) and incubated for 30 min at 25°C with stirring . Zymographic analysis of the proteins’ peptidoglycan hydrolytic activities was performed as described earlier ( Heilmann et al . , 1997 ) . Briefly , crude cell lysates or pure proteins isolated as described above were separated by SDS–PAGE on a polyacrylamide gel ( 10% acrylamide/0 . 2% bisacrylamide ) containing inactivated Micrococcus luteus or S . aureus cells ( recovered at mid-exponential phase , washed with water , and heat inactivated by autoclaving at 121°C for 15 min ) as a substrate in the resolving gel . Equal amounts of crude extracts ( 50 μg ) or purified proteins ( 200 ng ) , quantified in a Nanodrop ND-1000 spectrophotometer ( Thermo Scientific , Wilmington , NC ) , were loaded in the gel . After electrophoresis , gels were washed in deionized H2O for 15 min at RT , and then incubated in renaturation buffer ( 0 . 1% Triton X-100 , 10 mM CaCl2 , 10 mM MgCl2 , 50 mM Tris–HCl , pH 7 . 5 ) at 37°C with gentle agitation . The zymogram was stained in methylene blue solution ( 1% methylene blue in KOH 0 . 01% ) for 3 min and destained in water until bands with peptidoglycan hydrolytic activity were observed as clear zones in the opaque gel . WTAs were extracted by alkaline hydrolysis from overnight cultures , analyzed by native PAGE , and visualized by combined alcian blue/silver staining , as previously described ( Atilano et al . , 2010 ) . Mutanolysin ( Sigma ) -digested peptidoglycan was prepared from NCTC8325-4 and NCTCΔatl S . aureus strains as previously described ( Filipe et al . , 2005 ) . The released muropeptides were reduced with sodium borohydride and separated by HPLC on a C18 column ( ODS-Hypersyl , 5 μm , 4 . 6×250 mm; Thermo Scientific ) at a flow rate of 0 . 5 ml/min for 160 min with a gradient from 5% to 30% ( vol/vol ) methanol in 100 mM NaH2PO4 , pH 2 . 0 . The eluted muropeptides were detected by their UV absorption at 206 nm , and their abundance was estimated by calculating the percentage of the integrated area of each peak relative to the total area of the muropeptide peaks , using Shimadzu LC Solution software . Cell-free supernatants from the S . aureus NCTCΔatl mutant and its parental strain NCTC8325-4 were obtained by harvesting overnight cultures at 10 , 000×g for 10 min at 4°C , and filtering the supernatants through a 0 . 22 µm filter ( Merck Millipore ) . Sterility was confirmed by plating the supernatants on TSA plates at 30°C . To image the effect of the secreted Atl products on the PGRP-SA binding , exponentially growing NCTCΔatl and NCTC8325-4 cells ( 500 µl at OD 0 . 5 ) were incubated with the previously isolated cell-free supernatants ( 500 μl ) for 30 min at 30°C . Cells were then washed twice with cold PBS , incubated with mCherry_PGRP-SA , and imaged by fluorescence microscopy . To quantify mCherry_PGRP-SA binding to the parental and the atl null mutant S . aureus cells pre-treated with secreted Atl products , the cultures were grown to mid-exponential phase ( OD ∼0 . 5 ) , centrifuged , and resuspended in TSB to an OD of ∼2 . 5 . A 100 µl volume of each culture was placed into the wells of a 96-well plate and harvested at RT , 5 min at 3600×g . Cells were washed once with 100 µl of PBS and then resuspended in 100 µl of cell-free supernatants ( prepared as described above ) . After 30 min of incubation at 30°C , the cells were washed twice with 100 µl of PBS . Quantification of mCherry_PGRP-SA binding was done as described above . AM ( 15 . 6 and 390 . 1 nM ) , GL ( 18 . 7 and 467 . 4 nM ) , AM plus GL ( 15 . 6 and 18 . 7 nM , respectively ) , LytA ( 3 . 5 nM ) , and mutanolysin ( 108 . 7 nM ) were added to 200 µl of PBS-washed NCTCΔatl cells collected during exponential growth ( 500 µl at OD 0 . 5 ) . After 30 min incubation at 30°C , cells were washed twice with cold PBS , incubated with mCherry_PGRP-SA as described before , and imaged by fluorescent microscopy . To determine the relative amount of AM and GL still associated with the bacterial cells following the treatment with these two enzymes , cells were treated with AM and GL as described above and washed three times with PBS ( control pellets without the enzymes , without washing , or without cells were used ) . Samples were serially diluted 1:2 to concentrations ranging from 6 µg/ml ( undiluted ) to 0 . 37 µg/ml ( lowest dilution ) , and 2 µl were spotted onto a polyvinylidene difluoride ( PVDF ) membrane ( GE Healthcare Life Sciences ) and immunostained with specific anti-His mouse antibodies . Washed , unwashed , and control bacterial cells were also labeled with mCherry_PGRP-SA as described above and observed by fluorescence microscopy . Bacterial cells samples ( 2 µl ) incubated with fluorescent tagged mCherry_PGRP-SA were placed on a thin layer of 1% agarose ( Bio-Rad ) in PBS . For time-lapse microscopy experiments , 2 µl of NCTCΔatl culture , pre-labeled with mCherry-PGRP-SA , were placed on a thin layer of 0 . 6% agarose containing filtered supernatants of either NCTC8325-4 or NCTCΔatl cells . After 2 . 5 min , phase contrast and fluorescence images were taken every 30 s for 20 . 5 min and a final image was acquired after 50 min of incubation . Imaging was performed using a Leica DM 6000B microscope equipped with an iXonEM + EMCCD camera ( Andor Technologies ) , using Metamorph software ( Meta Imaging series 7 . 5 ) . Images were analyzed using ImageJ software ( Abràmoff et al . , 2004 ) . Overnight cultures of bacteria ( 10 ml ) were washed and resuspended in PBS so that their final OD600 was 0 . 350 . Healthy looking adult female flies ( n = 25 ) , 2–4 days old , were injected in the thorax with 32 nl of a bacterial cell suspension ( approximately 100 CFUs ) or with PBS , using a nanoinjector ( Nanoject II; Drummond Scientific , Broomall , PA ) , and fly survival was determined as previously described ( Atilano et al . , 2011 ) . In the fly survival assay for the pathogenicity rescue of atl null mutants , 100 μl of each bacterial cell suspension were incubated with cell-free supernatants of wild type or Δatl mutants ( prepared as described above ) for 30 min at 30°C . Treated bacterial cells were washed twice , resuspended in an equal volume of PBS and further diluted so that approximately 100 CFUs would be injected per fly . Each experiment was performed in triplicate , on different days . Following injection , flies were kept at 30°C and survival assessed every 12 h over a period of 3 days . Since the trends in survival were the same for each sample of a triplicate ( i . e . , survival curves were positioned similarly , relative to one another ) , the three independent biological repeats were combined ( n = 75 ) and estimates of survival curves were plotted . Fly survival was also tested using PGRP-SA pre-coated S . aureus cells . For that purpose , overnight NCTCΔatl bacterial cells were washed and resuspended in PBS ( 200 μl; OD600 0 . 350 ) and pre-coated with mCherry_PGRP-SA at 300 μg/ml for 5 min at 4°C . Cells were washed once and resuspended in an equal volume of ice-cold PBS . PBS suspensions with non-coated atl cells or containing only mCherry_PGRP-SA ( 300 μg/ml ) were used as experiment controls . Controls and samples were further diluted in PBS ( 1:250 ) and used to infect PGRP-SA deficient flies as described above . The bacterial load per insect during infection was assessed as previously described ( Atilano et al . , 2011 ) . Six female flies were homogenized in 1000 μl TSB with a micropestle device ( VWR , Radnor , PA ) until pieces of tissue were no longer visible . Homogenate samples ( 50 μl ) underwent a 10-fold dilution series from 1× to 1/100 , 000× before being spread onto TSB agar plates and onto mannitol salt agar . The number of CFUs was scored after 24 h at 30°C . Triplicate assays were performed . Total RNA was extracted from homogenized female flies ( n = 6 ) using the Total RNA Purification Plus Kit ( Norgen , Canada ) according to the manufacturers’ instructions . A Maxima First Strand cDNA Synthesis Kit ( Thermo Scientific ) was used to produce cDNA . The total RNA ( 500 ng ) was mixed with 5× reaction mix buffer ( 4 μl ) , Maxima enzyme mix ( 2 μl ) , and RNase-free dH2O to a final volume of 20 μl . cDNA was synthesized in a T100 Thermal Cycler ( Bio-Rad ) and stored at −20°C . Drosomycin ( FlyBase annotation symbol: CG10810 ) and tbp ( FlyBase annotation symbol: CG9874 ) expression levels were measured using primer pairs: Drs ( + ) and Drs ( − ) ; Tbp ( + ) and Tbp ( − ) . The housekeeping gene tbp ( Matta et al . , 2011 ) was used as a control to normalize the expression of the Drs gene . qPCR reactions were performed using the SensiFast SYBR No-ROX Kit ( Bioline , UK ) as outlined in the manufacturer’s instructions , in a Rotor-Gene Q real-time PCR cycler with a 72-well rotor ( Qiagen ) . Negative controls had an equivalent amount of total RNA or no cDNA template . Three biological repeats were performed per time point . Gene expression was calculated on the basis of the comparative threshold cycle ( CT ) value ( Schmittgen and Livak , 2008 ) . Levels of gene expression in all groups were shown as a ratio to the control group value . Data for the mCherry_PGRP-SA binding to bacteria assays ( n = 10–50 ) was non-normal but with equal variance , and therefore a non-parametric Kruskal–Wallis test followed by Dunn’s multiple comparison was applied . Estimated survival curves were constructed from the raw data sets and the log-rank ( Mantel–Cox ) test used to determine statistical significance between the curves . For clarity , 95% confidence intervals have been omitted from the graphs . When the bacterial CFU numbers in fly survival assays did not present a normal distribution ( Lilliefors test , p<0 . 05 ) , log10 transformation was used . Repeated measures two-way ANOVA was used to analyze significant differences over time and between bacterial strains . Bonferroni post-tests were used to locate the time points where mean values were statistically separable between the two bacterial strains . Analysis for qPCR data showed that the data followed a normal distribution ( Lilliefors test , p>0 . 1 ) and had equal variance ( Levene’s test , p>0 . 1 ) . Thus , mean values were plotted with 95% confidence and repeated measures two-way ANOVA was used to look for significant differences between bacterial wild type strain and atl null mutant over time . Bonferroni post-tests were used to locate the time points where mean values were statistically separable between the two bacterial strains . All data were plotted and analyzed using GraphPad Prism 6 ( GraphPad Software ) . | While most bacteria are harmless , some can cause diseases as varied as food poisoning and meningitis , so our immune system has developed various ways of detecting and eliminating bacteria and other pathogens . Receptor proteins belonging to the immune system detect molecules that give away the presence of the bacteria and trigger an immune response targeted at the invading pathogen . Peptidoglycan is one telltale molecule that betrays the presence of bacteria . Peptidoglycan is found in the bacterial cell wall , and for many years it was assumed that the immune system detected stray fragments of peptidoglycan that were accidentally shed by the bacteria . However , it was later shown that the immune system could , under certain conditions , detect peptidoglycan when it is still part of the cell wall . This raised an interesting question: do bacteria use other methods to stop peptidoglycan being detected by the immune system ? Now , Atilano , Pereira et al . have found that enzymes called autolysins can conceal bacteria from the receptor proteins that detect peptidoglycan . These enzymes are needed to break the bonds within the peptidoglycan present in the rigid bacterial cell wall to allow the bacteria to grow and divide . ‘Knocking out’ the genes for autolysins allowed the receptor proteins from the fruit fly , Drosophila , to bind to the bacteria; however , the mutant bacteria were able to evade the immune system after they had been treated with the purified enzymes . Atilano , Pereira et al . suggest that the autolysins trim the exposed ends of the peptidoglycan molecules on the surface of the cell wall , which could otherwise be detected by the host . The experiments also show that bacterial pathogens—including a strain of MRSA—with mutations that knock out autolysin activity trigger a stronger immune response in fruit flies , and are therefore less able to infect this host . Autolysins also help to conceal Streptococcus pneumoniae—a bacterial pathogen that is a common cause of pneumonia and infant deaths in developing countries—from detection by fruit flies . The findings of Atilano , Pereira et al . highlight how bacteria employ a number of ways to evade detection . If similar behavior is observed when bacteria infect humans , autolysins could represent a new drug target for the treatment of bacterial diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"immunology",
"and",
"inflammation"
] | 2014 | Bacterial autolysins trim cell surface peptidoglycan to prevent detection by the Drosophila innate immune system |
Voltage-gated ion channels generate electrical currents that control muscle contraction , encode neuronal information , and trigger hormonal release . Tissue-specific expression of accessory ( β ) subunits causes these channels to generate currents with distinct properties . In the heart , KCNQ1 voltage-gated potassium channels coassemble with KCNE1 β-subunits to generate the IKs current ( Barhanin et al . , 1996; Sanguinetti et al . , 1996 ) , an important current for maintenance of stable heart rhythms . KCNE1 significantly modulates the gating , permeation , and pharmacology of KCNQ1 ( Wrobel et al . , 2012; Sun et al . , 2012; Abbott , 2014 ) . These changes are essential for the physiological role of IKs ( Silva and Rudy , 2005 ) ; however , after 18 years of study , no coherent mechanism explaining how KCNE1 affects KCNQ1 has emerged . Here we provide evidence of such a mechanism , whereby , KCNE1 alters the state-dependent interactions that functionally couple the voltage-sensing domains ( VSDs ) to the pore .
Voltage-gated ion channels sense changes in membrane voltage and respond by opening or closing a pore through which selected ions cross the membrane , generating a transmembrane current . These channels consist of four voltage-sensing domains ( VSDs ) surrounding a central pore . In voltage-gated potassium ( Kv ) channels , this structure results from the tetrameric assembly of Kv-α subunits , each of which contain six transmembrane-spanning segments ( S1–S6 ) . S1–S4 of each subunit forms a voltage-sensing domain ( VSD ) , and S5–S6's from all four subunits form the pore . Sensing of membrane voltage occurs within the VSDs , which contain a mobile S4 segment with several highly conserved basic residues . At depolarized voltages , the forces of the membrane electric field on these positively charged residues promotes the outward displacement of S4 toward its activated state ( Papazian et al . , 1995; Larsson et al . , 1996; Silva et al . , 2009; Wu et al . , 2010a; Delemotte et al . , 2011; Jensen et al . , 2012 ) . The pore contains the ion permeation pathway that can be opened and closed by the reorientation of the intracellular portions of the S6 helices ( Liu et al . , 1997; Webster et al . , 2004; del Camino and Yellen , 2001; Jiang et al . , 2002 ) . Critical to voltage-dependent gating are the interactions between the VSDs and the pore , which couple the activation of the VSD to the opening of the pore , resulting in a voltage-gated conductance ( Chen et al . , 2001; Lu et al . , 2001 , 2002; Long et al . , 2005; Lee et al . , 2009; Zaydman et al . , 2013 ) . In their pioneering work on the action potential of the squid giant axon , Hodgkin and Huxley empirically derived a model for the K+ conductance in which a transmembrane pathway is gated by four voltage-dependent particles ( Hodgkin and Huxley , 1952 ) . The legacy of the Hogdkin and Huxley model can still be found in the current Kv channel models , which assume that ( 1 ) VSD activation and pore opening are two-state ( i . e . , all or none ) processes and ( 2 ) gating and permeation are independent so that the VSD conformation changes the probability of pore opening , but does not affect the properties of the open pore . Several recent studies call into question the assumption that VSD activation and pore opening are two-state processes . Computational and experimental studies have demonstrated that VSD activation actually occurs in a series of stepwise transitions due to salt bridge interactions between the basic residues on S4 and acidic residues on S1 and S2 , which define resting , intermediate , and activated states ( Papazian et al . , 1995; Tiwari-Woodruff et al . , 1997; Wu et al . , 2010a; Delemotte et al . , 2011; Jensen et al . , 2012; Lacroix et al . , 2012 ) . With regards to pore opening , recordings of single channel currents from Kv channels revealed multiple open states discernable by their different conductance levels ( Chapman et al . , 1997 ) , although the identities of these subconductance states remain unclear . In our present study of KCNQ1 ( Kv7 . 1 , KvLQT1 ) channels , we found that both the intermediate and fully-activated states of the VSD yielded robust pore-opening . Remarkably , the intermediate-open and activated-open states had different permeation and pharmacological properties revealing that VSD-pore interactions determine both open probability and open conformation , demonstrating that gating and permeation are not independent . KCNQ1 channels generate currents with very different properties as a result of tissue specific expression of KCNE family accessory subunits ( Abbott , 2014 ) . In the heart , channels formed by KCNQ1 and KCNE1 subunits are responsible for the slow-delayed rectifier potassium current , IKs ( Barhanin et al . , 1996; Sanguinetti et al . , 1996 ) , which plays a critical role in limiting action potential duration when beta-adrenergic tone is elevated . The importance of this response is highlighted by a large set of loss-of-function mutations of KCNQ1 or KCNE1 that have been associated with Long QT Syndrome and result in an elevated risk of fatal arrhythmias during times of stress ( Paavonen et al . , 2001; Schwartz et al . , 2001; Hedley et al . , 2009 ) . Although KCNE1 is a small , single-membrane-spanning peptide , its coassembly dramatically alters every physiologically relevant property of the KCNQ1 channel: voltage-dependence , current kinetics , inactivation , current amplitude , single channel conductance , selectivity , and pharmacology ( Wrobel et al . , 2012; Sun et al . , 2012 ) . The mechanism of how KCNE1 modulates KCNQ1 has been a longstanding topic of debate with several groups arguing that KCNE1 alters VSD activation ( Nakajo and Kubo , 2007; Ruscic et al . , 2013 ) , several other groups arguing that KCNE1 alters pore opening ( Rocheleau and Kobertz , 2008; Osteen et al . , 2010 ) , and several reports claiming that KCNE1 directly contributes to the inner structure of the pore ( Wang et al . , 1996; Tai and Goldstein , 1998 ) . However , these mechanisms are not able to simultaneously account for the effects of KCNE1 on gating and the observed changes in permeation and pharmacology . Here we made three observations regarding the function of homomeric KCNQ1 channels that were not previously reported . First , we found that VSD activation occurs in two resolvable steps through a stable intermediate state . Second , we observed that the intermediate-state of the VSD is sufficient to promote KCNQ1 channel opening , resulting in both intermediate-open and activated-open states . Third , we observed that the intermediate-open and activated-open states have different permeation and pharmacological properties . With these critical observations , we were able to reexamine how KCNE1 affects KCNQ1 . We found that coexpression of KCNE1 prevented the intermediate-open state and changed the properties of the activated-open state . The apparent decoupling of pore opening from the resting to intermediate transition of the VSD suggested that KCNE1 changes how the VSD and pore interact . Consistent with this hypothesis , changing the VSD-pore interactions directly via a single point mutation also prevented the intermediate-open state and modified the properties of the activated-open state . Using a kinetic model , we demonstrated that the effects of KCNE1 on VSD-pore interactions , as suggested by our data , are sufficient to simultaneously explain most of the changes in activation gating without any direct impacts on VSD activation or pore opening . Furthermore , as VSD-pore interactions were found to determine the open-pore properties , the effects of KCNE1 on permeation and pharmacology could also be rationalized . Therefore we conclude that altering VSD-pore interactions is likely the primary mechanism through which KCNE1 modulates KCNQ1 .
Voltage-clamp fluorometry ( VCF ) simultaneously monitors VSD-activation and pore opening ( Mannuzzu et al . , 1996 ) . In our VCF records , a fluorophore attached to the S3–S4 linker of pseudo-WT KCNQ1 channels ( with mutations C214A/G219C/C331A ) reports on conformational changes associated with VSD activation ( Figure 1A–H , green ) , while the ionic currents report the opening of the pore ( Figure 1A–H , black ) ( Osteen et al . , 2010; Zaydman et al . , 2013 ) . We observed multiple components of the fluorescence signals for KCNQ1 ( Figure 1A–D ) and , as recently reported ( Barro-Soria et al . , 2014 ) , for KCNQ1+KCNE1 ( Figure 1E–H ) . Most of the total change in fluorescence intensity was due to a fast component occurring at hyperpolarized voltages ( Fmain ) , but a small additional change was observed due to a slow component occurring at highly depolarized voltages ( Fhigh ) . Fhigh was more prominently observed when KCNQ1 channels were labeled with a different dye ( Figure 1—figure supplement 1A , B ) , or coexpressed with a mutant KCNE1 ( Figure 1—figure supplement 1C , D ) . 10 . 7554/eLife . 03606 . 003Figure 1 . KCNE1 suppresses the intermediate-open state of KCNQ1 . ( A–H ) Fluorescence ( green ) and current ( black ) signals from Xenopus oocytes injected with cRNA encoding pseudo-WT ( C214A/G219C/C331A ) KCNQ1 alone ( KCNQ1 , A–D ) or coinjected with cRNAs encoding pseudo-WT KCNQ1 and KCNE1 ( KCNQ1+KCNE1 , E–H ) . The cells were labeled with Alexa 488 C5-maleimide . ( A and E ) GV and FV relationships ( solid ) with the main and high voltage FV components plotted ( dotted lines ) . ( B and F ) normalized fluorescence and current responses to a 60 mV pulse shown with fits ( thin grey lines ) to a single- or bi-exponential function . Averaged fast ( C and G ) and slow ( D and H ) tau values of fluorescence and current responses to various voltage pulses . ( I ) Intermediate- ( E1-R2 , top ) and activated- ( E1-R4 , bottom ) state homology models of KCNQ1 after 100 ns of MD simulation . Side view of one VSD ( left ) and bottom view of the pore ( right ) . ( J ) Currents from the cells expressing E160R/R231E ( E1R/R2E , top ) or E160R/R237E ( E1R/R4E , bottom ) both alone ( −KCNE1 , middle ) or with KCNE1 ( +KCNE1 , right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 00310 . 7554/eLife . 03606 . 004Figure 1—figure supplement 1 . Improved resolution of Fhigh . ( A and B ) VCF data from cells expressing pseudo-WT KCNQ1 , labeled with Alexa 546 C5-maleimide ( KCNQ1 [Alexa 546] ) . ( C and D ) VCF data from pseudo-WT KCNQ1+R67E/K69E/K70E KCNE1 labeled with Alexa 488 C5-maleimide ( KCNQ1+RKK/EEE [Alexa488] ) . ( A and C ) Normalized current and fluorescence responses to a 60 mV pulse . ( B and D ) GV and FV relationships ( solid ) with the main and high voltage FV components plotted ( dotted lines ) . Current signals ( black ) , fluorescence signals ( color , red = Alexa 546 , green = Alexa 488 ) . The RKK/EEE mutation in KCNE1 causes a leftward shift of Fmain , relative to that of WT KCNE1 , which increases the separation between Fmain and Fhigh . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 00410 . 7554/eLife . 03606 . 005Figure 1—figure supplement 2 . GV/FV relationships are maintained in channels the mutation R243Q in KCNQ1 . GV ( solid black ) and FV ( solid green ) relationships for R243Q/psWT KCNQ1 channels expressed alone ( R243Q ) ( A ) or with R67E/K69E/K70E KCNE1 ( R243Q+RKK/EEE ) ( B ) . Green dotted lines show the main and high voltage FV components . The GV relationship of pseudo-WT channels ( black dashed lines ) is significantly different from that of R243Q; however , the relationship of the GV to different FV components are preserved . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 00510 . 7554/eLife . 03606 . 006Figure 1—figure supplement 3 . MD simulations predict that , unlike the resting-state , both the intermediate- and activated-states of the VSD stabilize pore opening through state-dependent protein and lipid interactions . ( A ) Snapshots of one VSD ( left , side view ) or the pore domain ( right , bottom view ) following 100 ns of MD simulations . In the resting , intermediate or activated VSD , E160 ( E1 , red ) forms a salt bridge with R228 ( R1 ) , R231 ( R2 ) or R237 ( R4 ) , respectively . ( B ) Averaged ( over several trajectories ) pore radius vs position along the axis normal to the membrane ( Z ) . ( C ) Snapshots of the PIP2 intrasubunit-binding site in the three states . In the resting/closed state , PIP2 interacts with positive residues of S4 ( cyan ) . When the VSD is intermediate or activated , PIP2 shifts closer to S6 ( yellow ) and anchors its positive residues ( K354 and K358 ) . ( D ) Probability of salt bridges formation between positive residues of S6 ( K354 and K358 ) and PIP2 . The lipid interacts with S6 only when the VSD is intermediate or activated , not when it is resting . Error bars represent SD . K354 and K358 interactions are not statistically different for the E1-R2 and E1-R4 states . Figure 1—figure supplement 3B represents the averaged pore radius profiles along the axis normal to the membrane ( Z ) . In the activated/open and intermediate states , the minimal radiuses of the pore at this level are 3 . 5 ± 0 . 4 Å and 2 . 8 ± 0 . 6 Å respectively . For comparison , in the Kv1 . 2 open state , the corresponding radius ( pdb 3LUT [Chen et al . , 2010] ) is 4 . 2 Å , in the Kv1 . 2/2 . 1 paddle chimera open state ( pdb 2 R9R [Long et al . , 2007] ) it is 4 . 2 Å also , and in the NavMS open state ( pdb 3ZJZ [Loussouarn et al . , 2003] ) it is 2 . 3 Å . Therefore , the minimal pore radius at the intercellular gate level in the models of the Kv7 . 1 activated and intermediate states corresponds to the open pore . In the resting/closed state , this radius decreases to 1 . 5 ± 0 . 5 Å . This is similar to the closed states of KcsA ( pdb 1K4C [Zhou et al . , 2001] ) , NavAB ( pdb 4EKW [Payandeh et al . , 2012] ) and NavAP ( pdb 4DXW [Zhang et al . , 2012] ) , where these values are 1 . 1 , 1 . 2 and 0 . 9 Å respectively . The activated/open , intermediate and resting/closed states of Kv7 . 1 differ by their properties as evidenced from the reported experimental data . Taking advantage of our simulations , we attempted to investigate whether the interactions between PIP2 and positive residues of the Kv7 . 1 intrasubunit binding site are different . Indeed PIP2 interacts preferably with the VSD ( S4 ) when the channel is resting/closed or with the pore ( S6 ) when the channel is activated/open ( Kasimova et al . , 2014 ) ( Figure 1—figure supplement 3C , top and bottom panels ) . In the intermediate state , the lipid forms salt bridges with both S4 ( R243 ) and S6 ( K354 and K358 ) simultaneously ( Figure 1—figure supplement 3C , middle panel ) . Its equilibrium position is also between these in the activated/open and resting/closed states . Interestingly , the probability of interaction between PIP2 and S6 ( K354 and K358 ) is rather high ( Figure 1—figure supplement 3D ) . The average values are slightly higher for the intermediate than for the activated/open states: 40 and 26% for K354 , 68 and 42% for K358 respectively . However , this difference is statistically insignificant due to the estimated error bars . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 00610 . 7554/eLife . 03606 . 007Figure 1—figure supplement 4 . VSD mutations reveal that KCNE1 suppresses currents from intermediate-open states and increases those from activated-open states . ( A ) Cartoons illustrating the mutational strategy used to arrest the VSD near the intermediate- and activated-states . The E160R ( E1R ) mutation was used to disrupt the native electrostatic interactions between the S2 and S4 segments that stabilize the VSD as it undergoes activation . E1R was paired with R231E ( R2E ) or R237E ( R4E ) to stabilize the putative intermediate- and activated-states , respectively . ( B ) Currents from oocytes expressing WT or mutant KCNQ1 subunits alone ( −KCNE1 , left ) or with KCNE1 ( +KCNE1 , right ) in response various voltage test pulses . ( C and D ) Averaged steady-state current–voltage relationships for cells expressing WT or mutant KCNQ1 subunits alone ( C , −KCNE1 ) or with KCNE1 ( D , +KCNE1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 00710 . 7554/eLife . 03606 . 008Figure 1—figure supplement 5 . Surface membrane expression of E1R/R2E and E1R/R4E . Biotinylation of intact oocytes allowed separation of membrane proteins from the cell lysate using streptavidin beads . The membrane fraction and the cell lysate were subjected to Western Blot using antibodies against KCNQ1 ( Top ) or Gβ , a soluble protein not found in the membrane . ( A ) E1R/R2E and E1R/R4E reached the cell membrane with similar efficiency . ( B ) KCNE1 did not decrease the expression of E1R/R2E to the membrane . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 008 Strikingly , KCNE1 shifted the conductance-voltage ( GV ) relationship so that it correlated with a different component of the fluorescence-voltage ( FV ) relationship . In the absence of KCNE1 , pore opening ( i . e . , GV ) occurred in a similar voltage range as Fmain ( Figure 1A , Figure 1—figure supplement 1B ) . In contrast , in the presence of KCNE1 , pore opening was not observed unless more depolarized voltages were applied , as with Fhigh ( Figure 1E , Figure 1—figure supplement 1D ) . These FV-GV correlations were not likely to be coincidental because they were maintained in the presence of a KCNQ1 mutation , R243Q , that shifted the voltage dependence of channel opening but did not change the correlation of the GV to Fmain or Fhigh in the absence or precense of RKK/EEE KCNE1 , respectively ( Figure 1—figure supplement 2 ) . KCNE1 also altered the time-dependence of pore opening . KCNQ1 current onset had two exponential components following the two timecourses of the fluorescence ( Figure 1B–D ) . With KCNE1 , the channels remained closed during the fast fluorescence increase and , after this initial delay , the channels opened with a single timecourse , similar to the slow fluorescence component ( Figure 1F–H ) . Both the steady-state and kinetic VCF data can be easily explained if one VSD transition ( Fmain ) is sufficient to open KCNQ1 , but an additional transition ( Fhigh ) is required for KCNQ1+KCNE1 . The observation of two fluorescence components ( Figure 1A–H ) is consistent with the suggestion from previous studies ( Silva et al . , 2009; Wu et al . , 2010a ) that , in KCNQ1 , VSD activation occurs in two sequential transitions due to electrostatic interactions between E160 ( E1 ) in S2 and arginine residues in S4 . These interactions stabilize discrete resting , intermediate , and activated states . We built homology models of KCNQ1 with the VSDs in states where E1 forms a salt bridge with R228 ( R1 ) , R231 ( R2 ) , or R237 ( R4 ) ( Figure 1—figure supplement 3 ) , the three S4 arginines that are known to be critical for voltage-sensing in KCNQ1 ( Shamgar et al . , 2008; Wu et al . , 2010b ) . In KCNQ1 a neutral residue ( Q234 ) is located in the canonical third arginine position ( R3 ) of other Kv channels; therefore , we did not model the E1-R3 state . In Molecular Dynamic ( MD ) simulations , we found that the pore was more dilated when the VSDs were in the E1-R2 or E1-R4 states than in the E1-R1 state ( Figure 1—figure supplement 3 ) . These simulations suggest that KCNQ1 channels can open when the VSDs assume intermediate or activated states ( Figure 1I , Figure 1—figure supplement 3 ) . To capture these states experimentally , we engineered pairwise charge reversal mutations to arrest the VSD near the E1-R2 or E1-R4 states ( Figure 1J , Figure 1—figure supplement 4 ) . As shown previously ( Wu et al . , 2010a ) , for KCNQ1 channels , mutating E1 to arginine ( E1R ) caused a severe loss of current that was partially rescued by a charge reversing ( R to E ) mutation at the R2 ( E1R/R2E ) or R4 ( E1R/R4E ) positions ( Figure 1—figure supplement 4 ) . E1R/R2E and E1R/R4E channels were constitutively open ( Figure 1—figure supplement 4C , D ) , suggesting that the VSD was trapped in states where the paired residues interact . When KCNE1 was coexpressed , the E1R/R2E currents were eliminated and the E1R/R4E currents were increased by nearly 10-fold ( Figure 1J , Figure 1—figure supplement 4 ) , suggesting that KCNE1 suppresses the intermediate-open state and potentiates the activated-open state . Importantly , we found that the abundance of E1R/R2E subunits in the cell membrane was similar when expressed alone or coexpressed with KCNE1 , indicating that the inhibition of E1R/R2E currents was due to a functional effect on the intermediate-open state , not an effect on surface expression ( Figure 1—figure supplement 5B ) . Altogether , the results in Figure 1 are consistent with a model in which the VSD undergoes two sequential transitions , resting-to-intermediate and intermediate-to-activated . The first transition is sufficient for KCNQ1 to open , resulting in both intermediate-open and activated-open states . With KCNE1 , the second transition is required for opening because the intermediate-open state is suppressed . In addition to affecting KCNQ1 channel gating , KCNE1 also alters permeation and pharmacology ( Sun et al . , 2012 ) . For example , as previously demonstrated by Pusch et al . ( 2000 ) , KCNQ1 channels had a higher Rb+/K+ permeability ratio than KCNQ1+KCNE1 channels ( Figure 2A , B ) and , as reported by Cohen and colleagues ( Wang et al . , 2000 ) , KCNQ1 channels were more sensitive than KCNQ1+KCNE1 channels to the inhibitor XE991 when short duration pulses ( comparable to the length of the cardiac action potential ) were applied ( Figure 2—figure supplement 1 ) . We used the E1R/R2E and E1R/R4E mutations to examine if VSD conformation affects permeation and pharmacology . Strikingly , E1R/R4E channels had a significantly lower Rb+/K+ permeability ratio and a significantly lower apparent affinity for XE991 compared to E1R/R2E and WT KCNQ1 channels , which were similar ( Figure 2 ) . These results reveal that different VSD conformations yield functionally distinct open states . Furthermore , the similar properties of E1R/R2E and WT KCNQ1 suggest that the intermediate-open state is either the most populated or the most conductive open-state for WT KCNQ1 channels , a notion that is further supported by our observation that E1R/R2E currents are 2–3 times larger than E1R/R4E currents ( Figure 1J , Figure 1—figure supplement 4C ) despite having similar levels of membrane expression ( Figure 1—figure supplement 5A ) . Relative to E1R/R4E alone , coexpression of KCNE1 with E1R/R4E significantly decreased the Rb+/K+ permeability ratio ( Figure 2A , B ) and increased the apparent affinity for XE991 ( Figure 2C , D ) indicating that , in addition to eliminating the intermediate-open state , KCNE1 alters the properties of the activated-open state . Of note , we found that E1R/R4E+KCNE1 and WT KCNQ1+KCNE1 channels displayed similar Rb+/K+ permeability ratios ( Figure 2B ) and XE991 sensitivities ( Figure 2D ) , further supporting our conclusion from Figure 1 that WT KCNQ1+KCNE1 currents are conducted by channels in the activated-open state . 10 . 7554/eLife . 03606 . 009Figure 2 . Permeation and pharmacological properties depend on VSD conformation . Currents from cells expressing WT KCNQ1 alone ( KCNQ1 , black ) , E160R/R231E ( E1R/R2E , red ) , E160R/R237E alone ( E1R/R4E , blue ) , E160R/R237E+KCNE1 ( E1R/R4E+KCNE1 , green ) , or WT KCNQ1+KCNE1 ( KCNQ1+KCNE1 , grey ) . ( A ) Currents from a single cell in external solutions containing 100 mM of Na+ , K+ , or Rb+ . The currents were elicited by first stepping the voltage to +60 mV for 5 s then to −60 mV for 3 s tails . ( B ) Averaged Rb+/K+ permeability ratios calculated by comparing the tail current amplitudes . ( C ) . Currents before ( CTL ) and after ( XE991 ) bath application of 5 μM XE991 in the external solution . ( D ) Fraction of original current inhibited after 2 s of depolarization vs concentration of XE991 applied shown with fits to the hill equation with a hill coefficient of 1 . N . S . = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 00910 . 7554/eLife . 03606 . 010Figure 2—figure supplement 1 . Inhibition of KCNQ1+KCNE1 channels by XE991 develops slowly over time . Currents in response to a +60 mV test pulse from oocytes expressing KCNQ1 ( A ) or KCNQ1+KCNE1 ( B ) , recorded in control ( CTL ) and 5 μM XE991 external solutions . ( C ) Time dependence of XE991 inhibition—the averaged ratio of the current in 5 μM XE991 to that in control solutions is plotted vs depolarization time . ( D ) Averaged fraction inhibited after 200 ( dashed line ) , 500 ( thin line ) or 2000 ( thick line ) ms of depolarization is plotted vs concentration of XE991 for oocytes expressing WT KCNQ1 ( black ) or WT KCNQ1+KCNE1 ( grey ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 010 In our previous studies , we have shown that KCNE1 increases the apparent affinity of KCNQ1 for the membrane lipid phosphatidylinositol 4 , 5-bisphosphate ( PIP2 ) ( Li et al . , 2011 ) and that , in KCNQ1 , PIP2 binding at the VSD-pore interface mediates the VSD-pore interactions that energetically couple VSD activation to pore opening ( Zaydman et al . , 2013 ) . Taken together these findings suggest that KCNE1 affects the VSD-pore interactions . To test this hypothesis , we studied the impact of changing VSD-pore interactions directly via a point mutation of KCNQ1 , F351A . F351 on S6 is highly conserved among Kv channels and participates in interactions with the S4/S5 linker that are known to be critical for VSD-pore interactions and coupling ( Lu et al . , 2001; Tristani-Firouzi et al . , 2002; Long et al . , 2005; Haddad and Blunck , 2011 ) . Remarkably , in VCF experiments , we observed that the F351A GV was correlated with Fhigh instead of Fmain ( Figure 3A , left ) , revealing that , similar to KCNE1 ( Figure 1E ) , F351A suppressed the intermediate-open state . As a result , F351A current onset was delayed and slowed ( Figure 3A , right ) , resembling WT KCNQ1+KCNE1 current onset , as reported previously ( Boulet et al . , 2007 ) . We also observed that , compared to WT KCNQ1 , F351A caused significant changes in the Rb+/K+ permeability ratio ( Figure 3B ) and the timecourse of inhibition by XE991 ( Figure 3C , middle , right ) . Of note , these properties of F351A were not identical to those of WT KCNQ1+KCNE1 . However , such differences are not surprising as prior studies ( Kang et al . , 2008; Xu et al . , 2008; Chung et al . , 2009; Lvov et al . , 2010; Strutz-Seebohm et al . , 2011; Wang et al . , 2011; Chan et al . , 2012; Xu et al . , 2013 ) have located KCNE1 at the VSD-pore interface and have suggested that KCNE1 engages in a very broad and complex set of interactions with KCNQ1 ( Sun et al . , 2012 ) ; therefore , it would be unreasonable to expect that a single point mutation , such as F351A , would alter the VSD-pore interactions in exactly the same way as KCNE1 . Nonetheless , our studies of F351A provide evidence that the VSD-pore interactions determine the permeation and pharmacological properties of the pore as well as which VSD transitions are required for the pore to open . Therefore , altering VSD-pore interactions is a single mechanism that can explain all of these effects of KCNE1 . 10 . 7554/eLife . 03606 . 011Figure 3 . Altering VSD-pore coupling directly , by the mutation F351A , changes the gating permeation , and pharmacology of KCNQ1 channels . ( A ) VCF recordings from oocytes expressing pseudo-WT/F351A ( C214A/G219C/C331A/F351A ) , labeled with Alexa 488 C5-maleimide . Left–GV ( red ) and FV ( solid green ) relationships with the main and high voltage FV components plotted ( dotted green lines ) . Right–normalized fluorescence ( green ) and current ( red ) responses to a 60 mV pulse , the current from a cell expressing pseudo-WT KCNQ1+KCNE1 is shown for comparison ( grey ) . ( B ) Left–currents from a single oocyte expressing F351A in external solutions containing 100 mM of Na+ , K+ , or Rb+ . Right–averaged Rb+/K+ permeability ratios for WT KCNQ1 ( black ) and F351A ( red ) . ( C ) Left–currents from an oocyte expressing F351A in control and 5 μM XE991 external solutions . Middle–time dependence of XE991 inhibition—the averaged ratio of the current in 5 μM XE991 to that in control solutions is plotted vs depolarization time . Right–averaged fraction inhibited after 200 ( dashed line ) , 500 ( thin line ) or 2000 ( thick line ) ms of depolarization is plotted vs concentration of XE991 for oocytes expressing WT KCNQ1 ( black ) , F351A ( red ) , or WT KCNQ1+KCNE1 ( grey ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 011 The observation that KCNE1 suppressed the intermediate-state opening ( Figure 1 , Figure 1—figure supplement 4 ) strongly suggested that KCNE1 affects the interactions between the intermediate-state of the VSD and the pore . Does KCNE1 also affect the interaction between the activated-state of the VSD and the pore to modulate the activated-open state ? In order to detect such effects of KCNE1 , we used the apparent affinity of E1R/R4E for PIP2 as a proxy for the strength of VSD-pore interactions in the activated-open state of KCNQ1 . The rationale behind this approach was that our previous study has demonstrated that the apparent affinity for PIP2 correlates with the strength of VSD-pore interaction ( Zaydman et al . , 2013 ) . In our experiments , we used CiVSP , a voltage-sensitive lipid phosphatase ( Iwasaki et al . , 2008 ) , to cause a rapid decrease in membrane PIP2 upon depolarization ( Falkenburger et al . , 2010 ) and observed the resulting time-dependent decay in ionic currents resulting from net unbinding of PIP2 . We found that the decay of E1R/R4E+CiVSP currents were significantly slower and less severe when KCNE1 was present ( Figure 4A ) , suggesting that KCNE1 increases the affinity of the activated-open state of KCNQ1 for PIP2 . Consistent with this idea , WT KCNQ1+KCNE1 currents , which come from activated-open states exclusively ( Figure 1 ) , were observed to be relatively insensitive to the activity of CiVSP as long as they were maintained in the activated-open state by sustained membrane depolarization ( Figure 4B , C ) . When the channels were permitted to close , by hyperpolarizing the membrane potential , PIP2 unbinding was facilitated as evidenced by decreased current amplitude observed with application of a subsequent depolarizing pulse ( Figure 4B , C ) . Altogether these results reveal that KCNE1 causes the activated-open state to have a high apparent affinity for PIP2 , suggesting that KCNE1 may increase the strength of VSD-pore interactions in the activated-open state . 10 . 7554/eLife . 03606 . 012Figure 4 . KCNE1 increases the apparent affinity of the activated-open state for PIP2 . ( A ) Responses of currents from oocytes expressing E1R/R4E alone ( left ) or E1R/R4E+KCNE1 ( middle ) to rapid depletion of PIP2 by CiVSP ( VSP , blue ) . The membrane voltage was pulsed to +60 mV to activate CiVSP . Currents were normalized to the value 200 ms after depolarization for comparison with currents from oocytes not expressing CiVSP ( black = channel subunits alone , blue = channel subunits + CiVSP ) . Right–averaged current responses for oocytes expressing E1R/R4E+CiVSP ( −KCNE1 ) or E1R/R4E+KCNE1+CiVSP ( +KCNE1 ) . ( B ) CiVSP responses ( VSP , blue ) of currents from oocytes expressing WT KCNQ1 or WT KCNQ1+KCNE1 . Two +80 mV depolarizing pulses were applied , spaced 30 s apart ( note: time scale is broken ) . The currents were normalized to the value 200 ms into the first depolarizing pulse for comparison with currents from an oocyte expressing channel subunits alone ( black ) . ( C ) Left–double pulse protocol in which two 25 s depolarizing pulses were applied . In between the two pulses the membrane potential was set to various voltages for 10 s . After each sweep , the membrane potential was held at −80 mV for 300 s to deactivate CiVSP and allow for PIP2 regeneration by the endogenous lipid kinases . Middle–currents from an oocyte expressing KCNQ1+KCNE1+CiVSP subjected to the voltage protocol shown and normalized to the value at the end of the first pulse ( I1 ) . Right–fraction of the first pulse current available on the second pulse ( I2/I1 ) is plotted vs voltage of intervening 10 s ( blue ) . The voltage-dependence of I2/I1 ( solid blue line ) is similar to that of the GV relationship for WT KCNQ1+KCNE1 ( dotted black line ) suggesting that the open probability during the 10 second interpulse determines the unbinding of PIP2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 012
The basic experimental phenomena revealed in our study are summarized in Figure 5A . We observed that , in KCNQ1 , VSD activation occurred in two steps , with and without KCNE1 ( Figure 1A–H ) . The first step , leading to an intermediate-state , promoted the opening of homomeric KCNQ1 channels causing the GV to overlap with the first component of the FV ( Figure 1A ) . KCNE1 prevented the channel from opening while the VSD is in the intermediate-state and potentiated opening while the VSD is in the activated state resulting in a GV relationship with a similar voltage-dependence as the second component of the FV ( Figure 1E ) . We observed that different states of the VSD yielded functionally different open states ( Figure 2 ) of the pore indicating that the unique sets of interactions with different conformations of the VSD determine the probability of pore opening and selects for different open-pore conformations . In agreement with this interpretation , changing the VSD-pore interactions directly , via the mutation F351A , suppressed the intermediate-open state and changed the properties of the open channel ( Figure 3 ) . The ability of this single point mutation at the S4–S5/S6 interface to change the same variety of properties as KCNE1 suggests that a single mechanism for the effects of KCNE1 could be to alter the VSD-pore interactions . 10 . 7554/eLife . 03606 . 013Figure 5 . Modeling the effects of KCNE1 . ( A ) Cartoon illustrating the observed effects of KCNE1 on KCNQ1 . The VSDs transit between resting ( R , grey patterned ) , intermediate ( I , red patterned ) , and activated ( A , blue patterned ) states . Each VSD conformation has unique interactions ( double arrow ) with the closed ( C ) and open ( O ) conformation of the pore . KCNE1 suppresses the intermediate-open ( OI ) state and modulates the activated-open ( OA ) states . ( B ) Kinetic model of KCNQ1 channel gating where the k parameters are the intrinsic transition rates of the VSD and the pore , and the θ parameters explicitly represent VSD-pore interactions . Fourth power notation ( [ ]4 ) indicates that the model includes four VSDs . ( C and D ) Experimental data and model simulations with ( +KCNE1 ) and without KCNE1 . Steady-state current- ( C , top ) , conductance- ( C , middle ) , and fluorescence- ( C , bottom ) voltage relationships . ( D ) Current ( black ) and fluorescence ( green ) responses at various voltages . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 01310 . 7554/eLife . 03606 . 014Figure 5—figure supplement 1 . Scheme for voltage-depedendent gating . The channel is modeled as a pore coupled to four voltage-sensing domains . VSD activation occurs in two transitions , resting ( R ) to intermediate ( I ) along the horizontal and intermediate to activated ( A ) along the vertical . The pore can be fully closed ( C , back face , grey ) or fully open ( O , front face , black ) , with any combination of VSD-states . The state-notation indicates the conformation of the pore and the combination of the four VSDs ( subscripted ) . For example CRRRR indicates the state where the pore is closed and all four VSDs are resting . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 01410 . 7554/eLife . 03606 . 015Figure 5—figure supplement 2 . Balanced gating model showing transition rates . As described in the text the k parameters represent the intrinsic rates of the two domains and the θ parameters represent the effect of VSD-pore interactions at different states . The closed to open transitions dependent on the intrinsic rate of the pore ( kco ) divided by the closed state interactions with the VSDs . The open to closed transitions depend on the koc divided by the open state interactions the VSDs . n ( X ) indiciated the number of VSDs in state X . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 01510 . 7554/eLife . 03606 . 016Figure 5—figure supplement 3 . Desriptions , values and units for free parameters of gating model . The values represent those used in the kinetic modeling of KCNQ1 gating . ΔKCNE1 column indicates the changes used to simulate the effects of KCNE1 on KCNQ1 channel gating . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 01610 . 7554/eLife . 03606 . 017Figure 5—figure supplement 4 . Descriptions , values , and units for constant parameters used in gating model simulation . DOI: http://dx . doi . org/10 . 7554/eLife . 03606 . 017 Experimentally , two effects of KCNE1 on VSD-pore interactions were observed . ( 1 ) KCNE1 prevented the channel from opening when the VSD was in the intermediate-state suggesting that KCNE1 changes the interactions between the intermediate-state of the VSD and the pore ( Figure 1 ) . ( 2 ) KCNE1 greatly increased the apparent affinity of the activated-open state for PIP2 suggesting that KCNE1 stabilized the activated-open state interactions ( Figure 4 ) . Using kinetic modeling , we sought to test if these experimentally observed effects of KCNE1 on VSD-pore interactions are sufficient to explain the many effects of KCNE1 on KCNQ1 activation gating . Previous Kv channel gating models assume that all four VSDs ( Zagotta et al . , 1994 ) or each VSD ( Horrigan et al . , 1999 ) must fully activate before the pore can open . Therefore , we could not use these established models as they are not able to represent the coupling between three different states of the VSD and the pore . We developed a new gating model that better reflects the understanding that the VSD and pore can fold and function independently ( discussed in our previous review [Zaydman and Cui , 2014] ) , but couple their motions through state-dependent interactions . In our gating model , the VSDs can occupy resting ( R ) , intermediate ( I ) , or activated ( A ) states , and the pore can occupy closed ( C ) or open ( O ) states ( Figure 5B , Figure 5—figure supplement 1 ) . Assuming that the intrinsic activation of each VSD is identical and non-cooperative , as assumed in the ZHA model ( Zagotta et al . , 1994 ) , the different combinations of the states of the four VSDs and the pore give rise to only 30 possible channel states . Two sets of parameters , k's and θ's , determine the transitions among these channel states ( Figure 5B , Figure 5—figure supplement 2 , 3 ) . The ‘k’ parameters represent the intrinsic tendencies of the pore to open and close and a VSD to undergo its transitions , and would be measured directly if the VSD and pore could be decoupled entirely . The k parameters concerning VSD transitions ( kRI , kIR , kIA , kAI ) are assumed to be exponentially dependent on voltage according to Equation 1 . On the other hand , the k parameters concerning pore transitions ( kCO and kOC ) are assumed to be voltage-independent ( i . e . , constant ) . ( 1 ) kij=kijo∗exp ( zij∗F∗V/ ( R∗T ) ) where: kij is the voltage-dependent rate of transition from VSD state i to VSD state j , kij0 is the rate of the ij transition at 0 mV , zij is the valence of the ij transition , F is the faraday constant , V is voltage , R is the universal gas constant , T is the absolute temperature . The ‘θ’ terms explicitly represent the net effect of all VSD-pore interactions within each channel state . For example , θΙC , represents the net stabilization of the intermediate-closed state due to all interactions between the intermediate-state of the VSD and the closed-state of the pore . If θIC is greater than 1 , these interactions will slow transitions away from the intermediate-closed state . The full model is shown in Figure 5—figure supplement 2 and a shorthand version appears in Figure 5B . The main difference between our model and previous allosteric gating models is that the reference state is an imaginary state in which the two domains are completely isolated ( i . e . , decoupled ) from each other , rather than using transitions among the resting states as a reference . The advantage of our approach is that each parameter has an intuitive physical meaning: the k parameters represent the net effect of all interactions within a state of the VSD or the pore , and the θ parameters represent the net effect of all interactions between the VSD and the pore in a specific channel state . In order to parameterize our model , we started with parameter values from two previous KCNQ1 channel gating models ( Silva and Rudy , 2005; Zaydman et al . , 2013 ) and were able to reasonably replicate the experimentally observed KCNQ1 channel gating behavior ( Figure 5C , D ) with some adjustments to the parameter values to account for the differences in the schemes of these models ( Figure 5—figure supplement 3 ) . In order to simulate the effects of KCNE1 on VSD-pore interactions , we made only two changes to the VSD-pore interactions , as suggested by experimental results ( Figure 5—figure supplement 3 ) . ( 1 ) Experimentally , when KCNE1 was coexpressed , the pore remained closed while the VSD was in the intermediate state ( Figure 1E , Figure 1—figure supplement 4 ) . Thus , to model KCNE1 , we strengthened the intermediate-closed state interaction , θIC . ( 2 ) Also , in experiments , KCNE1 increased the activated-open state affinity for PIP2 ( Figure 4 ) . Accordingly , we strengthened the activated-open state interaction , θAO . Making only these two changes mimicked all of the effects of KCNE1 on KCNQ1 activation gating ( Figure 5C , D ) . The GV shifted to more depolarized voltages correlating with Fhigh because the intermediate-closed state interactions became more energetically favorable than the intermediate-open state interactions , preventing pore opening until a more depolarized voltage-range where the intermediate-to-activated transition occurred . In the absence of intermediate-state opening , the resting-to-intermediate transition occurred among closed states and introduced the characteristic delay of current onset of several hundred milliseconds . The maximal current amplitude was increased several fold due to stabilization of the activated-open state . Furthermore , the stabilization of the intermediate-closed state reduced the current near physiological resting voltages and caused a left shift in main component of the FV curve , two phenomena observed experimentally ( Figure 5C ) . The kinetics and steady-state gating behavior predicted by our model were not quantitatively identical to those in experiments; such discrepancies were expected due to several overly simplistic assumptions that we used to limit the number of states in our model . Particularly , we assumed that PIP2 binding was saturated , that all open states had identical conductance , and that inactivated states did not exist . However , previous studies highlight that KCNQ1 is not saturated with PIP2 and that KCNE1 dramatically increases PIP2 binding ( Li et al . , 2011 ) , that KCNE1 increases the apparent single channel conductance ( Sesti and Goldstein , 1998; Yang and Sigworth , 1998 ) , and that KCNE1 prevents the observation of a partially inactivated state ( Pusch et al . , 1998; Tristani-Firouzi and Sanguinetti , 1998 ) , all of which may contribute to gating and macroscopic current amplitude . Revision of our model to include the influence of PIP2 binding and the different properties of intermediate- and activated-open states will require additional studies to better define these properties . Despite these limitations , it is remarkable that we were able to capture all the major effects of KCNE1 on the activation gating of KCNQ1 . This model illustrates that the effects of KCNE1 on VSD-pore interactions are sufficient to explain how KCNE1 affects the activation gating of KCNQ1 without any additional effects on VSD activation or pore opening . Furthermore , as our experimental data demonstrate that VSD-pore interactions determine the permeation and pharmacological properties ( Figures 2 , 3 ) , our proposed mechanism , that KCNE1 affects VSD-pore interactions , provides a relatively complete explanation for how KCNE1 affects KCNQ1 . Central to understanding the modulation of KCNQ1 by KCNE1 is a longstanding controversy regarding which stoichiometries of KCNQ1:KCNE1 may exist in the fully assembled channel . Several groups have argued that association of KCNE1 with KCNQ1 dimers during an early stage of biogenesis leads to a fixed 4:2 KCNQ1:KCNE1 stoichiometry and breaks the fourfold symmetry of the channel ( Wang and Goldstein , 1995; Chen et al . , 2003; Morin and Kobertz , 2008; Plant et al . , 2014 ) . Other groups have argued that various stoichiometries , 4:0–4 KCNQ1:KCNE1 , are possible , depending on the relative levels of subunit expression ( Blumenthal and Kaczmarek , 1994; Cui et al . , 1994; Wang et al . , 1998; Nakajo et al . , 2010; Li et al . , 2011 ) . In the present study , we coinjected oocytes with KCNQ1 and KCNE1 transcripts at a 1:1 ratio because previous work , from our lab ( Li et al . , 2011 ) and others ( Nakajo et al . , 2010 ) , demonstrated that this ratio is sufficient to saturate the functional effects of KCNE1 on KCNQ1 . In our modeling , we assume that coexpression of KCNE1 affects all four subunits identically as if the channel were saturated by KCNE1 , that is , with a 4:4 stoichiometry . This assumption may represent yet another reason why the simulated and experimental gating behavior are not quantitatively identical . Furthermore , it is important to ask , given the ongoing controversy regarding stoichiometry , if the presence of multiple populations of channels with different stoichiometries could complicate our interpretation of our data . Fortunately , the F351A mutation demonstrates that the major finding of our study—the ability to couple pore opening to different transitions within the VSD—is intrinsic to the KCNQ1 subunit and observed even in the absence of the KCNE1 subunit . In the voltage-gated ion channel field , VSD-pore coupling , aka electromechanical coupling , has been a loosely defined term referring to the experimental observation that pore opening is more likely when the VSDs are activated at depolarized voltages . In our model of voltage-dependent gating , coupling is represented in a very different way than in the previously established models . The landmark linear models of Shaker Kv channels ( Schoppa et al . , 1992; Zagotta et al . , 1994 ) , which require that all four VSDs be activated before the pore can open , do an excellent job replicating the gating of Shaker Kv channels , but do not explicitly define coupling . Another landmark model , the Horrigan-Cui-Aldrich ( HCA ) model ( Horrigan et al . , 1999 ) , permits pore-opening without prior activation of all VSDs . In the HCA model , coupling is quantitatively defined by a single term , D , representing how much the closed-open equilibrium of the pore is biased towards open when a single VSD is transitioned from its fully resting to its fully activated state . Our model further develops from these two previous models . Like the shaker model , we assumed two transitions occurring sequentially and independently within each VSD . Similar to the HCA model , we assumed that pore opening can occur with the four VSDs in any combination of states . In our model , we decomposed D into its elementary components , that is , VSD-pore interactions at each channel state . As a result , each parameter ( k , θ ) in our model represents a specific set of interactions that exist at the same time in the physical world . The relative differences in the strengths of all these state-dependent VSD-pore interactions lead to the experimental observation that a change in VSD conformation leads to a change in the probability of pore opening , that is , coupling . It is important to understand that coupling is a result of all of these state-dependent interactions and a perturbation of any of these interactions may alter the coupling . These points are illustrated by our modeling of the KCNE1 effect on KCNQ1 channel gating where strengthening the intermediate-closed interactions caused an apparent decoupling of pore-opening from the resting-to-intermediate transition of the VSD , that is , the open probability was no longer increased by the transition of the VSD to the intermediate-state at intermediate voltages . Alternatively , weakening the intermediate-open state interactions would also decouple opening from the resting-to-intermediate state of the VSD; however , this would not reproduce the leftward shift of the first FV component ( Fmain ) that we observed when KCNE1 was expressed ( Figure 5C ) . The topic of how KCNE1 modulates KCNQ1 has been studied for many years . Early studies by Goldstein and colleagues argued that KCNE1 intercalates deeply into the pore and directly lines the permeation pathway ( Goldstein and Miller , 1991; Wang et al . , 1996; Tai and Goldstein , 1998 ) . Later , the Kass and George labs argued that KCNE1 associates outside the pore and modulates the pore properties through an allosteric mechanism ( Kurokawa et al . , 2001; Tapper and George , 2001 ) . Then , the efforts of several labs have detected probable interactions between the extracellular ( Xu et al . , 2008; Chan et al . , 2012 ) , transmembrane ( Tapper and George , 2001; Chung et al . , 2009; Lvov et al . , 2010 ) , and intracellular ( Haitin et al . , 2009 ) regions of KCNQ1 and KCNE1 . These data have been used to build homology models of the KCNQ1+KCNE1 channel ( Kang et al . , 2008; Lundby et al . , 2010; Xu et al . , 2013 ) , all of which place KCNE1 in a cleft between a VSD and the pore where it participates in a broad set of interactions with KCNQ1 . Even after all of these excellent studies , there was a lack of a biophysical mechanism to explain how these interactions alter the gating , permeation , and pharmacology of KCNQ1 . Prior biophysical studies , which have focused on how KCNE1 slows current onset , have come to conflicting conclusions that KCNE1 either slows VSD activation ( Nakajo and Kubo , 2007; Ruscic et al . , 2013 ) or pore opening ( Rocheleau and Kobertz , 2008; Osteen et al . , 2010 ) . We believe that the interpretations of these data were limited by several missing pieces of information revealed in the present study: ( 1 ) VSD activation occurs through a stable intermediate state , ( 2 ) full activation of the VSD is not necessarily required to open the pore , and ( 3 ) the distinct interactions between different states of the VSD and the pore determine both the probability of opening and the open-state conformation . Also , it is likely that these previous studies did not consider an effect on the state-dependent VSD-pore interactions because such interactions are not explicitly represented in previously established gating models . Recently , the Larsson and Kass labs reported the presence of a second fluorescence transition in KCNQ1+KCNE1 channels , which they attributed to pore opening ( Osteen et al . , 2010 , 2012 ) or a concerted step in which pore opening and further S4 movement occur simultaneously ( Barro-Soria et al . , 2014 ) . From these observations , they concluded that KCNE1 changes the number of subunits that must be activated before pore opening can occur leading to a right shifted GV and a delay in current onset . This model is most similar to ours in that KCNE1 is changing the relationship between VSD activation and pore opening; however , our model is different in several regards . We believe that the second FV component reports on an intrinsic transition of the VSD rather than the pore opening step . This assumption was based on our observation that homomeric KCNQ1 channels also exhibited a second FV component , with similar properties to that of KCNQ1+KCNE1 , which occurred in a range of voltages that was more positive than required for pore opening ( Figure 1A–H , Figure 1—figure supplement 1 ) . Thus , in our model , channel opening requires an additional transition within a single VSD , rather than the activation of additional subunits . A natural consequence of such gating behavior is the existence of both intermediate-open and activated-open states , which could be detected experimentally and are of great functional importance as the different open states have different conductive and pharmacological properties ( Figure 2 ) .
Point mutations were engineered using overlap extension and high-fidelity PCR . Each mutation was verified by DNA sequencing . cRNA was synthesized using the mMessage T7 polymerase kit ( Applied Biosystems ) . Pieces of ovarian lobes were excised from Xenopus laevis by laparotomy . Stage V or VI oocytes from X . laevis were isolated by collagenase ( Sigma Aldrich , St Louis , MO ) digestion . 9 . 2 ng of KCNQ1 cRNA was microinjected with or without 2 . 3 ng of KCNE1 cRNA ( using the Drummond Nanoject , Broomall , PA ) into each oocyte . For CiVSP expression , 2 . 3 ng of CiVSP cRNA was coinjected . Injected cells were incubated at 18°C for up to 7 days before recording in ND96 solution ( 96 mM NaCl , 2 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 5 mM HEPES , pH 7 . 6 ) . In our previous work ( Kasimova et al . , ) , we have built models of the Kv7 . 1 activated/open and resting/closed states using homology modeling ( Eswar et al . , 2007 ) with the Kv1 . 2 crystal structure in its activated/open state ( pdb code 3LUT [Chen et al . , 2010] ) , α , and in its δ conformational state ( Delemotte et al . , 2011 ) as templates . Here , we applied a similar protocol to prepare a model of the Kv7 . 1 intermediate state . Each state is characterized by a unique set of interactions within the VSD . In particular , E160 ( E1 ) forms a salt bridge with R237 ( R4 ) in the activated/open or with R228 ( R1 ) in the resting/closed states ( Figure 1I , S3a ) . We assumed that , when the VSD is intermediate , E1 interacts with the residue located in between of R4 and R1 , namely R231 ( R2 ) . Based on this assumption , the γ conformation of Kv1 . 2 ( Delemotte et al . , 2011 ) with the E1-R2 , was considered as a template for the Kv7 . 1 intermediate state model . This model was further embedded in a palmitoyl-oleyl-phosphatidylcholine ( POPC ) hydrated bilayer and immersed in a 150 mM K+Cl− solution . Due to the importance of PIP2 for Kv7 . 1 function ( Loussouarn et al . , 2003; Eswar et al . , 2007 ) , four molecules of this lipid were placed at the channel's intrasubunit sites located at the interface between the voltage sensor and the pore ( Zaydman et al . , 2013; Kasimova et al . , ) . The MD simulations were performed using NAMD ( Smart et al . , 1996 ) . Langevin dynamics was applied to keep the temperature ( 300 K ) and the pressure ( 1 atm ) constant . The time-step of the simulations was 2 . 0 fs . The equations of motion were integrated using a multiple time-step algorithm . Short- and long-range forces were calculated every 1 and 2 time-steps respectively . Long-range electrostatics was calculated using Particle Mesh Ewald ( PME ) . The cutoff distance of short-range electrostatics was taken to be 11 Å . A switching function was used between 8 and 11 Å to smoothly bring the vdW forces and energies to 0 at 11 Å . During the calculations , chemical bonds between hydrogen and heavy atoms were constrained to their equilibrium values . Periodic boundary conditions were applied . The protein backbone was constrained during 100 ns allowing PIP2 to sample possible interactions with the channel's positive residues . Based on this trajectory , the time evolution of salt bridges formation was monitored . Several residues of Kv7 . 1 interacted with the lipid headgroups temporarily , revealing different configurations of the system where corresponding salt bridges were either formed or broken . In total , for the activated/open , intermediate and resting/closed states we have identified 9 , 8 and 8 the most frequent configurations respectively . These were considered as starting points for the final equilibration step , involving gradual release of the protein backbone and subsequent relaxation of the entire system during 100 ns for each . For all the trajectories , the root mean square deviation ( RMSD ) from the initial structure reached a plateau starting from ∼50 ns . The simulation stretch from 50 to 100 ns was used for further analysis . In order to estimate a degree of the Kv7 . 1 pore dilation at the intercellular gate level , we applied HOLE ( Smart et al . , 1996 ) . 50 conformations of Kv7 . 1 spread equidistantly along the last 50 ns were extracted from each MD trajectory . For these conformations , the pore radius along the axis normal to the membrane ( Z ) was calculated . The obtained profiles were considered to estimate an average profile and error bars ( SD ) for each of the channel states . To analyze the salt bridge formation between PIP2 and Kv7 . 1 , we measured the minimal distance between the nitrogen atoms of arginine and lysine charged groups and the oxygen atoms of the PIP2 phosphates . The salt bridges were assumed formed if the calculated distance was less than 3 . 2 Å . The probabilities of salt bridge formation were simultaneously estimated for four subunits of the channel as a ratio between the number of frames with a formed salt bridge to its total number . The error bars correspond to a standard deviation ( SD ) calculated between values obtained from several MD runs . Relative conductance-voltage ( GV ) relationships were generated by estimating the instantaneous tail current values following test pulses to various voltages and normalizing to the value following the highest voltage test pulse . For calculation of relative fluorescence changes , a baseline fluorescence was extrapolated by fitting a line to the fluorescence at the holding potential during the 2 s preceding application of the voltage pulse . ΔF/F was calculated as ( F ( t ) -baseline ( t ) ) /baseline ( t ) , where F ( t ) is the raw fluorescence intensity at time t and baseline ( t ) is the extrapolated baseline value at time t . Fluorescence voltage-relationships were derived by normalizing the ΔF/F value at the end of a four second test pulse to various voltages to the value of the highest voltage test-pulse . FV and GV curves were fits with one or the sum of two Boltzmann equations in the form 1/ ( 1 + exp ( −z*F* ( V − V1/2 ) /RT ) ) where z is the equivelant valence of the transition , V1/2 is the voltage at which the transition is half maximal , R is the gas constant , T is absolute temperature , F is the Faraday constant and V is the voltage . FV curves were derived from the value at the end of the test pulse , GV curved were derived from estimating the instantaneous tail current amplitude . All averaged data reflects n = 6 or more from at least two batches of oocytes . Pairwise comparisons were achieved using Student's t test , multiple comparisons were performed using an ANOVA with Tukey's Post-Hoc Test . All error bars represent standard error mean . 100 mM Na+ ( 96 mM NaCl , 4 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 5 mM HEPES ) 100 mM K+ ( 100 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 5 mM HEPES ) 100 mM Rb ( 96 mM RbCl , 4 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 5 mM HEPES ) ND96 ( 96 mM NaCl , 4 mM KCl , 1 . 8 mM CaCl2 , 1 mM MgCl2 , 5 mM HEPES ) XE991 from Sigma Aldrich . Alexa fluors from Molecular Probes . | Cells are surrounded by a membrane that prevents charged molecules from flowing directly into or out of the cell . Instead ions move through channel proteins within the cell membrane . Most ion channel proteins are selective and only allow one or a few types of ion to cross . Ion channels can also be ‘gated’ , and have a central pore that can open or close to allow or stop the flow of selected ions . This gating can be affected by the channel sensing changes in conditions , such as changes in the voltage across the cell membrane . Research conducted more than half a century ago—before the discovery of channel proteins—led to a mathematical model of the flow of potassium ions across a membrane in response to changes in voltage . This model made a number of assumptions , many of which are still widely accepted . However , Zaydman et al . have now called into question some of the assumptions of this model . Based on the original model , it has been long assumed that the voltage-sensing domains that open or close the central pore in response to changes in voltage must be fully activated to allow the channel to open . It had also been assumed that the voltage-sensing domains do not affect the flow of ions once the channel is open . Zaydman et al . have now shown that these assumptions are not valid for a specific voltage-gated potassium channel called KCNQ1 . Instead , this ion channel opens when its voltage-sensing domains are either partially or fully activated . Zaydman found that the intermediate-open and activated-open states had different preferences for passing various types of ion; therefore , the gating of the channel and the flow of ions through the open channel are both dependent on the state of the voltage-sensing domains . This is in direct contrast to what had previously been assumed . The original model cannot reproduce the gating of KCNQ1 , nor can any other established model . Therefore , Zaydman et al . devised a new model to understand how the interactions between different states of the voltage-sensing domains and the pore lead to gating . Zaydman et al . then used their model to address how another protein called KCNE1 is able to alter properties of the KCNQ1 channel . KCNE1 is a protein that is expressed in the heart muscle cell and mutations affecting KCNQ1 or KCNE1 have been associated with potentially fatal heart conditions . Based on the assumptions of the original model , it had been difficult to understand how KCNE1 was able to affect different properties of the KCNQ1 channel . Thus , for nearly 20 years it has been debated whether KCNE1 primarily affects the activation of the voltage-sensing domains or the opening of the pore . Zaydman et al . found instead that KCNE1 alters the interactions between the voltage-sensing domains and the pore , which prevented the intermediate-open state and modified the properties of the activated-open state . This mechanism provides one of the most complete explanations for the action of the KCNE1 protein . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Domain–domain interactions determine the gating, permeation, pharmacology, and subunit modulation of the IKs ion channel |
Disassembling microtubules can generate movement independently of motor enzymes , especially at kinetochores where they drive chromosome motility . A popular explanation is the ‘conformational wave’ model , in which protofilaments pull on the kinetochore as they curl outward from a disassembling tip . But whether protofilaments can work efficiently via this spring-like mechanism has been unclear . By modifying a previous assay to use recombinant tubulin and feedback-controlled laser trapping , we directly demonstrate the spring-like elasticity of curling protofilaments . Measuring their mechanical work output suggests they carry ~25% of the energy of GTP hydrolysis as bending strain , enabling them to drive movement with efficiency similar to conventional motors . Surprisingly , a β-tubulin mutant that dramatically slows disassembly has no effect on work output , indicating an uncoupling of disassembly speed from protofilament strain . These results show the wave mechanism can make a major contribution to kinetochore motility and establish a direct approach for measuring tubulin mechano-chemistry .
Microtubules are protein polymers that grow and shorten by addition and loss of αβ-tubulin subunits from their tips ( reviewed in Desai and Mitchison , 1997 ) . In addition to supporting cell structure and serving as tracks over which motor enzymes move , the filaments can act more directly to produce force and movement – that is , to do mechanical work – independently of motor enzymes . Microtubule polymerization can generate pushing forces ( Dogterom and Yurke , 1997; Janson et al . , 2003 ) and depolymerization can generate pulling forces ( Coue et al . , 1991; Koshland et al . , 1988; Lombillo et al . , 1995 ) . An important example of microtubule pulling occurs at kinetochores , where disassembling microtubule tips drive mitotic chromosome movements ( Desai and Mitchison , 1997; Inoué and Salmon , 1995; McIntosh et al . , 2010 ) . Similar depolymerization-driven pulling might occur at other cellular locations as well , for example at the cell cortex , where disassembling tips might generate pulling forces to position the spindle in the cell , ( Laan et al . , 2012; Nguyen-Ngoc et al . , 2007; Carminati and Stearns , 1997; Kozlowski et al . , 2007 ) or at spindle poles , where they might drive poleward microtubule flux ( Waters et al . , 1996 ) . The mechanical work that a disassembling microtubule tip exerts on an isolated kinetochore , or on a collection of kinetochore subcomplexes , can be directly measured in vitro ( Volkov et al . , 2013; Akiyoshi et al . , 2010 ) . But the mechanism underlying this force production is unknown . Two classes of models have been proposed to explain how disassembling microtubules produce force , conformational wave and biased diffusion ( Koshland et al . , 1988; Hill , 1985; Asbury et al . , 2011 ) . The central tenet of the conformational wave model is that individual rows of tubulin subunits , the protofilaments , pull on the kinetochore as they curl outward from a disassembling microtubule tip . Strain energy is trapped after GTP hydrolysis in the microtubule lattice , because intrinsically curved GDP-tubulin subunits are held in a straight ( i . e . , strained ) configuration by their binding to neighboring subunits ( Desai and Mitchison , 1997; Caplow and Shanks , 1996 ) . This stored strain energy is released during tip disassembly , when the protofilaments curl outward from the tip and break apart , forming a conformational wave that propagates down the microtubule ( Mandelkow et al . , 1991; Nogales and Wang , 2006 ) . Kinetochores are proposed to harness this wave to produce useful mechanical work . The central tenet of the alternative biased diffusion model is that the energy of interactions between a kinetochore and a microtubule creates a thermodynamic force that pulls the kinetochore toward the microtubule tip , analogous to the interfacial forces that draw fluid into a capillary ( Hill , 1985; Asbury et al . , 2011 ) . These two models are not mutually exclusive and , in principle , a purely biased diffusion-based mechanism could operate independently of any spring-like action of the protofilaments . The conformational wave mechanism , however , requires curling protofilaments to generate powerful ‘working strokes’ . A seminal study by Grishchuk and co-workers used a laser trap assay to show that disassembling microtubule tips can exert brief pulses of force against an attached bead ( Grishchuk et al . , 2005 ) . Their analysis suggested that the conformational wave might be capable of generating very high forces , up to ~50 pN , but the actual measured forces were much lower ( <0 . 5 pN ) and probably were restricted by interference of the attached beads with the short working strokes of the protofilaments . Displacement amplitudes were not reported . Because of these limitations , the energy carried by the curling protofilaments was not determined . Fundamentally , the work output of the conformational wave mechanism must be limited by the amount of curvature strain energy carried by GDP-protofilaments , which dictates how forcefully they can curl outward from the tip . Convincing measurements of protofilament strain energy should therefore reveal how efficiently they can produce mechanical work via the wave mechanism . Moreover , protofilament strain is fundamental to all current models of microtubule dynamic instability , and it is generally thought to drive rapid disassembly ( Desai and Mitchison , 1997; Nogales and Wang , 2006; VanBuren et al . , 2005; Molodtsov et al . , 2005 ) . Thus , measuring the strain energy in curling protofilaments will also provide insight into the basic mechano-chemistry of tubulin . Based on the pioneering work of Grishchuk et al . ( 2005 ) we have developed a modified ‘wave assay’ that overcomes limitations inherent to their study . Interference from the attached bead was minimized by using recombinant tubulin with an engineered , flexible tether . By applying a feedback-controlled laser trap , nm-scale displacements were measured as functions of force , enabling direct observation of the spring-like elasticity of curling protofilaments and showing that they carry a substantial fraction of the energy of GTP hydrolysis in the form of curvature strain . To probe the relationship between strain energy and disassembly rate , we measured the wave energy of a slow-disassembling tubulin mutant . Surprisingly , a 7-fold decrease in disassembly rate had no effect on conformational wave energy , which reveals that the speed of disassembly can be uncoupled from curvature-derived protofilament strain . We present a simple model to explain how strain energy and disassembly speed can be uncoupled .
The prior laser trap study demonstrated for the first time that disassembling microtubule tips can exert brief pulses of force on microbeads attached to the filaments by strong inert linkers , such as biotin-avidin ( Grishchuk et al . , 2005 ) . However , pulses were detected in fewer than half of the trials , pulse durations varied over 300-fold , and relaxation of the beads into the center of the trap was slower after the trials that failed to produce pulses . These observations suggest that the bead-microtubule attachments , which consisted of multiple biotin-avidin bonds ( approximately 3 to 8 ) , restricted outward curling of the protofilaments . Moreover , because a fixed trap was used without feedback control , pulse amplitudes were probably limited by the maximum distance over which the curling protofilaments could exert force ( i . e . , by their working stroke length ) , rather than by their total capacity for work output . These limitations made it difficult to quantitatively assess the force generating potential of the system . We therefore sought to improve the assay by developing a single molecule tethering scheme and by using a feedback-controlled trap . To begin our modified wave assay , we grew dynamic microtubule extensions from coverslip-anchored seeds . The extensions were assembled from recombinant yeast αβ-tubulin , with a His6 tag engineered onto the C-terminal tail of the β subunit . Microbeads were tethered to the sides of individual , growing filaments via single anti-His antibodies , creating a strong yet flexible tether ( ~36 nm in length; see Materials and methods ) . A bead-microtubule assembly was held in the laser trap ( Figure 1a and b ) and feedback control was initiated to apply a constant tension , which reduced Brownian motion and facilitated detection of microtubule-driven movements . The distal microtubule plus end was then severed with laser scissors to induce disassembly ( Franck et al . , 2010 ) . When the disassembling tip reached the bead , it generated a brief pulse , during which the bead first moved against the force of the laser trap , then relaxed back toward the trap center , and finally detached as the microtubule disassembled past the tether ( Figure 1c-e ) . At low opposing force , a pulse was nearly always observed ( 90% , or 148 of 164 events recorded at <5 pN ) . The pulses were large , often >60 nm ( Figure 1d and e ) , which is more than twice the width of the microtubules . These observations show that disassembling tips can generate pulses of movement more reliably than previously observed . The pulses were also fast , with average risetimes between 0 . 1 and 0 . 3 s ( depending on the level of force; Figures 1d , e and 2a–b ) , which is 5- to 10-fold faster than in the previous recordings . These observations suggest that our modified tethering scheme imposed less restriction on the outward curling of the protofilaments . 10 . 7554/eLife . 28433 . 003Figure 1 . Measuring the tubulin conformational wave with a feedback-controlled laser trap . ( a ) A bead is tethered to the side of a microtubule via a single antibody bound to the C-terminal tail of β-tubulin and placed under tension using the laser trap . The trap is feedback-controlled to keep a fixed separation from the bead ( Δx ) , thereby maintaining a constant level of tension . Microtubule disassembly is induced by cutting the tip with a second laser . ( b ) Video-enhanced differential interference contrast ( VE-DIC ) image of a 900 nm bead tethered to a single microtubule under laser trap tension ( from Video 1 ) . Approximate locations for the coverslip-anchored portion of the microtubule ( white arrow ) , the laser trap center ( red dashes ) , and the plus end tip ( yellow chevron ) are indicated . ( c , d ) Example record showing trap force ( c ) and bead displacement ( d ) versus time . Grey trace shows raw bead-trap separation after converting to force by multiplying by the trap stiffness . Black trace shows same data after smoothing with a 250 ms median filter . When the disassembling tip arrives at the bead , the bead initially moves against the trapping force and then releases as the microtubule disassembles out from underneath it . The pulse amplitude , a , and risetime , t , are indicated . ( e ) Gallery of additional example records , measured at the indicated levels of tension . Data in ( c - e ) were collected using 900 nm beads . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 00310 . 7554/eLife . 28433 . 004Figure 2 . Tubulin waves generate large forces . ( a , b ) Mean pulse risetime versus force ( a ) and distributions of risetime at indicated forces ( b ) for wild-type microtubules . The mean risetime across all forces is depicted by the dashed line in ( a ) . ( c , d ) Mean pulse amplitude versus force ( c ) and distributions of amplitude at indicated forces ( d ) for pulses generated by wild-type yeast microtubules . Total pulse energy , W , is estimated from the area under the line-fit in ( c ) , shaded grey . Error bars show standard errors ( for N = 6 to 87 amplitudes; N = 3 to 78 risetimes ) . All data in ( a - d ) were collected using 900 nm beads . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 00410 . 7554/eLife . 28433 . 005Figure 2—figure supplement 1 . Properties of wild-type tubulin waves measured using different bead sizes . ( a – f ) Mean amplitudes versus force ( a , c , e ) and distributions of amplitude at indicated forces ( b , d , f ) for pulses generated by wild-type yeast microtubules , measured with 320 nm beads ( a , b ) , with 440 nm beads ( c , d ) , and with 900 nm beads ( e , f ) . ( g – l ) Mean pulse risetimes versus force ( g , i , k ) and distributions of risetime at indicated forces ( h , j , l ) for wild-type microtubules measured with 320 nm beads ( g , h ) , with 440 nm beads ( i , j ) , and with 900 nm beads ( k , l ) . Dashed lines in ( a , c , and e ) show fits used to estimate the stall forces , unloaded amplitudes , and total pulse energies ( shaded gray areas ) , which are plotted against bead size in Figure 4a–c . Horizontal dashed lines in ( g , i , and k ) show mean risetimes across all forces , which are plotted against bead size in Figure 4d and e . Error bars show standard errors ( for N = 5 to 87 amplitudes; N = 3 to 78 risetimes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 005 Our modified wave assay enabled us to measure pulse properties as functions of force for the first time . Pulse amplitudes decreased as the force of the laser trap was increased ( Figure 2c and d ) . This behavior demonstrates directly that curling protofilaments exhibit spring-like elasticity . Eventually a ‘stall force’ was reached , at which the pulses were completely suppressed ( Figure 2c ) . Depending on bead size , the stall force ranged from 8 to 16 pN ( Figure 2—figure supplement 1 ) , which is at least 16-fold higher than the maximal force measured in the previous study ( <0 . 5 pN ) . The increased force production may be explained by our use of a force clamp , by our less restrictive tethering scheme , or by a combination of these two factors . It is also formally possible that the force generating capacity of microtubules grown from yeast tubulin ( used here ) is intrinsically higher than that of microtubules grown from bovine brain tubulin ( used in the previous study ) . However , we consider this possibility unlikely because the shapes and lengths of curling protofilaments are very similar in yeast and vertebrate cells ( McIntosh et al . , 2013 ) and because , at the level of tubulin structure , the internal curvature of unpolymerized αβ-tubulin ( i . e . , the rotation required to superimpose α- onto β-tubulin ) is also very similar ( Ayaz et al . , 2014 , 2012 ) . In any case , our results show that protofilaments curling outward from a disassembling microtubule tip behave like springs and can generate forces much higher than previously recorded . Measurements of wave-driven bead movement can potentially be used to estimate the total capacity of the conformational wave for mechanical work output , provided the mechanism underlying movement in the assay is understood . Beads in our assay were linked to the microtubules through the flexible C-terminal tails of β-tubulin . Flexible tethering implies that when a microtubule-attached bead is placed under tension , the tether should become extended and the bead surface should initially rest against the microtubule wall at a secondary contact point ( Figure 3a ) . Starting from this initial condition , we considered two scenarios for how the pulses of bead movement might be generated . In the ‘lateral push’ scenario , the curling protofilaments push laterally against the bead at the secondary contact point , causing the bead to pivot about the base of the tether ( Figure 3b ) . The bead acts as a lever in this case , but because the fulcrum is located at the tether , away from where the curling protofilaments exert their force , the predicted leverage is only modest ( ~2 fold , depending on bead size and tether length ) . In the second scenario , ‘axial pull’ , the microtubule first disassembles past the secondary contact point , allowing the bead to rotate under laser trap tension into an end-on configuration relative to the microtubule tip ( Figure 3c ) . Then the working stroke occurs when curling protofilaments encounter the tether and pull axially on the bead ( Figure 3d ) . There is no leverage in this case . The unamplified trapping force opposes protofilament curling directly . 10 . 7554/eLife . 28433 . 006Figure 3 . Proposed mechanisms underlying conformational wave-driven bead movement in the assay . ( a ) Initially , when a bead is placed under tension it rests against the microtubule wall at a secondary contact point . ( b ) In the lateral push scenario , the curling protofilaments push laterally against the secondary contact point , causing the bead to pivot about the base of the tether . Lateral deflections from the protofilaments , h , produce larger axial displacements of the bead , a . If A is the tether point and B is the point of bead-microtubule contact and C is the bead center , then ABC defines a right triangle and the amount of leverage is given by the ratio of sides BC/AB . The predicted leverage for 900 nm diameter beads attached via 36 nm tethers is a·h−1=2 . 4 . ( c ) In the axial pull scenario , the microtubule first disassembles past the secondary contact point , allowing the bead to rotate under laser trap tension into an end-on configuration relative to the microtubule tip . ( d ) Then the working stroke occurs when curling protofilaments encounter the tether and pull axially on the bead . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 00610 . 7554/eLife . 28433 . 007Figure 3—figure supplement 1 . A rare example record in which the initial pulse , from a stable baseline , was followed by bead relaxation toward the trap center and then by a second pulse ( double arrow ) . Such secondary pulses were seen in only 2% of all recorded events ( 18 of 760 ) . These rare secondary pulses might be generated by axial pulling . However , the lack of any relaxation before the primary pulses indicates that these were not preceded by bead rotation into an end-on configuration , and thus were not generated by axial pulling . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 007 Because of the relatively large bead radius , its rotation into an end-on configuration would produce an obvious relaxation toward the trap center , which in the axial pull scenario must precede the working stroke by ~200 ms ( the time required for tip disassembly to propagate from the secondary contact point to the tether ) . However , we found that the bead position was nearly always stable prior to the pulses ( Figure 1e ) , except in a very small fraction of trials ( ~2% , 18 of 760 ) during which the initial pulse , from a stable baseline , was followed by bead relaxation toward the trap center and then by a second pulse ( Figure 3—figure supplement 1 ) . These rare secondary pulses might be generated by axial pulling . However , the lack of any relaxation before the primary pulses indicates that these were not preceded by rotation into an end-on configuration , and thus were not generated by axial pulling . Thus , it seems that the lateral push mechanism underlies bead movement in most cases . To test more directly whether the lateral push model was operational , we examined how pulse amplitudes varied with laser trap tension and bead size . Altering bead size is predicted to have two consequences . First , larger beads should increase leverage and therefore decrease the amount of laser trap tension required to suppress the pulses . Consistent with this prediction , stall forces decreased from 16 . 2 ± 3 . 0 pN to 8 . 4 ± 0 . 9 pN as bead diameter was increased from 320 to 900 nm ( Figure 4a ) . The second prediction , also a consequence of leverage , is that larger beads should produce larger pulse amplitudes when the opposing tension is low enough to allow unhindered movement . Indeed , the maximum pulse amplitudes , extrapolated to zero tension , increased from 45 . 2 ± 3 . 6 nm to 64 . 3 ± 3 . 6 nm as bead diameter was increased from 320 to 900 nm ( Figure 4b ) . The relationships for stall force-vs-bead diameter and for unloaded amplitude-vs-bead diameter can be predicted quantitatively from simple geometric considerations , given estimates of the height that the curls project from the microtubule surface ( ~20 nm , based on electron micrographs of disassembling tips ) ( Mandelkow et al . , 1991; McIntosh et al . , 2013 , 2008 ) and of the tether length ( ~36 nm; see Materials and methods ) . The predicted curves fit our data well , and they are relatively insensitive to the precise tether length ( Figure 4a and b ) , suggesting that the lateral push model provides a good description of the underlying mechanism . 10 . 7554/eLife . 28433 . 008Figure 4 . Stall forces and pulse amplitudes vary with bead size , but pulse energy is invariant . ( a ) With increasing bead size , the leverage increases and therefore the trapping force required to completely suppress the pulses ( i . e . , the ‘stall force’ ) decreases . ( b ) Unloaded pulse amplitudes ( i . e . , amplitudes extrapolated to zero tension ) increase with bead size , because the amplification ratio increases ( see Figure 3b ) . Dotted curves in ( a ) and ( b ) show predictions assuming a tether length of 36 nm and a curl height , h = 20 nm . Gray shaded regions show predicted ranges for tether lengths ranging from 30 to 42 nm . ( c ) The total pulse energy , W , is independent of bead size . Horizontal dotted line in ( c ) shows global estimate of pulse energy , W = 304 ± 24 pN·nm , from a weighted fit of the wild-type data across all bead sizes . ( d , e ) Mean pulse risetimes as a function of bead size . Wild-type data in ( d ) are replotted in ( e ) with an expanded scale for comparison to the mutant , T238V . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 00810 . 7554/eLife . 28433 . 009Figure 4—figure supplement 1 . Estimation of strain energy per tubulin . ( a ) Given the 23° curvature and 8 nm length of a tubulin dimer , a curl height of h = 20 nm implies that the curled segments are ~4 dimers in length . ( b ) A maximum of ~4 curls could push simultaneously against the bead . ( c ) Thus , the mechanical work output , W = 304 ± 24 pN·nm ( Figure 4c ) , may derive from outward curling of as many as 16 tubulin dimers and the wave carries at least 19 pN·nm of energy per dimer . ( d ) Table of estimates of the mechanical strain energy stored after GTP hydrolysis in the microtubule lattice . Our estimate is based on the total mechanical work output of curling protofilaments , measured directly in the conformational wave assay , as explained above and in the text . The previously published estimates were inferred from thermodynamic considerations , ( Desai and Mitchison , 1997; Caplow and Shanks , 1996; Howard , 2001 ) from fitting of computational models to microtubule dynamic rate data , ( VanBuren et al . , 2005; Molodtsov et al . , 2005 ) from molecular dynamics ( MD ) simulations , ( Kononova et al . , 2014 ) and from measurements of the flexural rigidity of whole microtubules ( Mickey and Howard , 1995; Venier et al . , 1994; Felgner et al . , 1996; Pampaloni et al . , 2006; Schaedel et al . , 2015 ) . * Values shown in red were taken directly from the indicated references and used to calculate the values shown in black using the following relations ( which have been previously explained in detail; see Mickey and Howard , 1995 and VanBuren et al . 2005 ) : Lattice strain W and protofilament flexural rigidity EIp were related by W = ½· EIp·d·R−2 where d = 8 nm represents dimer length and R = 20 nm represents the radius of curvature for a relaxed protofilament . ( R = 20 nm corresponds to 23° per dimer; Mandelkow et al . , 1991 ) . Microtubule and protofilament flexural rigidity , EImt and EIp respectively , were related by a factor of 2140 , the ratio of their second moments ( Mickey and Howard , 1995 ) . Microtubule persistence length was defined as EImt divided by thermal energy , kBT = 4 . 1 pN·nm . † Estimates of stored lattice strain per tubulin dimer can be compared to the total free energy available from hydrolysis of GTP , which under typical cellular conditions is ~87 pN·nm ( =21 kBT=13 kcal·mol−1; Desai and Mitchison , 1997 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 009 Measuring pulse amplitude as a function of trapping force enabled us to calculate the total mechanical work output of the assay , W , which is given by the area under the amplitude-vs-force curve ( e . g . , Figure 2c and Figure 2—figure supplement 1 ) . Whereas changes in bead size altered the amplitude-vs-force curve in predictable ways ( as discussed above ) , the total work output was independent of bead size ( Figure 4c ) . This invariance would not be expected if work output was limited by the attached bead , and therefore it suggests that W indeed measures the intrinsic strain energy carried by the curling protofilaments that pushed against the bead . Based on a global average across all three bead sizes , we estimate W = 304 ± 24 pN·nm ( Figure 4c ) , a value 74-fold greater than thermal energy ( kBT ) . Assuming the lateral push model correctly describes the assay geometry , a maximum of 4 curls could push simultaneously against the beads ( Figure 4—figure supplement 1b ) . Given the 23° curvature and 8 nm length of an individual tubulin dimer , ( Mandelkow et al . , 1991; Amos and Klug , 1974 ) the estimated curl height of h = 20 nm further suggests that the curled segments are ~4 dimers in length ( Figure 4—figure supplement 1a ) . Thus , the total work output , W , may derive from outward curling of as many as 16 tubulin dimers , implying that the wave carries at least 19 pN·nm of energy per dimer ( 4 . 7 kBT , or 2 . 7 kcal mole−1; Figure 4—figure supplement 1c ) . These observations establish that the conformational wave carries considerable strain energy that can be harnessed to perform mechanical work , and they provide a direct estimate of the strain per tubulin subunit . Mechanical strain in the microtubule lattice is commonly assumed to drive the rapid disassembly of microtubules , ( Desai and Mitchison , 1997; Nogales and Wang , 2006; VanBuren et al . , 2005; Molodtsov et al . , 2005 ) but it has not been possible to test directly for a relationship between lattice strain and disassembly speed . To determine whether curvature strain dictates the speed of microtubule disassembly , we used our assay to measure the wave energy for microtubules assembled from a slow-shortening mutant tubulin . A number of tubulin mutations that ‘hyperstabilize’ microtubules in vivo have been described , ( Geyer et al . , 2015; Gupta et al . , 2002; Machin et al . , 1995 ) and some have been shown to slow the disassembly of microtubules grown in vitro from purified mutant tubulin ( Geyer et al . , 2015; Gupta et al . , 2002; Sage et al . , 1995 ) . We focused on a particular mutant in which threonine 238 of β-tubulin is replaced by valine ( T238V; see Figure 5a ) ( Geyer et al . , 2015 ) . The mutated residue is buried inside β-tubulin , where it cannot directly perturb inter-subunit lattice contacts . T238V tubulin forms microtubules that disassemble 7-fold more slowly than wild-type ( Figure 5—figure supplement 1 ) . If the rate of disassembly is determined by lattice strain , then the slow-disassembling mutant microtubules should store less conformational strain energy than their wild-type counterparts . 10 . 7554/eLife . 28433 . 010Figure 5 . Hyperstable mutant microtubules produce slower pulses . ( a ) Superposition of polymerized ( ’straight’ , green ) and unpolymerized ( ’curved’ , blue ) conformations of β-tubulin . Residue T238 is inaccessible to solvent and located on a helix ( H7 ) that undergoes piston-like movement between the straight and curved conformations ( which are represented by PDB entries 3JAT and 1SA0 , respectively ) . GDP nucleotide is shown in red . ( b , c ) Example record showing trap force ( b ) and bead displacement ( c ) versus time for a mutant T238V microtubule . Grey trace shows raw bead-trap separation after converting to force by multiplying by the trap stiffness . Black trace shows same data after smoothing with a 250 ms median filter . The pulse amplitude , a , and risetime , t , are indicated . ( d ) Gallery of additional example records for mutant T238V microtubules , measured at the indicated levels of tension . Data in ( b - d ) were collected using 900 nm beads . Note the different time scales here in comparison to Figure 1c–e . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 01010 . 7554/eLife . 28433 . 011Figure 5—figure supplement 1 . Hyperstable mutant T238V tubulin disassembles more slowly than wild-type . ( a ) Selected images from a movie of an individual yeast microtubule dissembling in vitro , recorded by video-enhanced differential interference contrast ( VE-DIC ) microscopy . The white arrow marks the coverslip-anchored seed . The yellow chevron marks the disassembling plus end . ( b ) Mean disassembly speeds measured in vitro for wild-type and mutant T238V microtubules ( by VE-DIC ) . Error bars show standard deviations ( for N = 16 to 24 microtubules ) . ( c ) Fluorescence images of individual wild-type and mutant T238V microtubules assembling in the presence of 50 nM Bim-GFP , which binds preferentially near the growing ends but also decorates the body of the microtubules at lower intensity . ( d ) Average Bim1-GFP fluorescence intensity versus distance from tip ( for N = 9 microtubules of each type ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 011 Contrary to this prediction , however , the conformational wave energy for T238V microtubules was indistinguishable from that of wild-type microtubules . T238V microtubules in the wave assay produced pulses that were ~5 fold slower than wild-type ( Figures 4e , 5c , d and 6a ) , consistent with their slower disassembly speed . However , the wave amplitude-vs-force relation for T238V microtubules was essentially identical to that of wild-type ( Figure 6c ) . Thus , even though they disassemble at very different rates , wild-type and T238V microtubules must store similar amounts of curvature-derived mechanical strain ( Figure 4c ) . The similar amplitude-vs-force curves further suggest that the mutation does not substantially alter the intrinsic curvature , flexural rigidity , or contour length of protofilament curls , because all of these properties together determine pulse amplitude . 10 . 7554/eLife . 28433 . 012Figure 6 . Hyperstable mutant microtubules produce pulses with identical energy . ( a ) Mean pulse risetime versus force for mutant T238V microtubules . Wild-type data ( from Figure 2a ) is shown for comparison . The mean risetimes across all forces for T238V and wild-type microtubules are depicted by the dashed blue and red lines , respectively . Error bars show standard errors ( for N = 6 to 25 amplitudes; N = 2 to 78 risetimes ) . ( b ) Distributions of risetime at indicated forces for wild-type and T238V microtubules . ( c , d ) Mean amplitude versus force ( c ) and distributions of amplitude at indicated forces ( d ) for pulses generated by mutant T238V microtubules . Wild-type data ( from Figure 2c ) is shown in ( c ) for comparison . Total pulse energy , W = 280 ± 50 pN·nm , estimated from the grey-shaded area under the line-fit , is similar for both types of microtubules . All data in ( a - d ) were measured with 900 nm beads . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 012 How might the rate of microtubule disassembly be uncoupled from mechanical strain in the lattice ? To begin addressing this question , we developed a simple model for the energy landscape of a single GDP-tubulin dimer curling outward from a disassembling tip ( Figure 7 ) . The mechanical strain energy carried by the dimer was modeled as a function of its bend angle by assuming it behaves as a slender elastic rod with a naturally bent shape , with 23° of curvature when fully relaxed , ( Mandelkow et al . , 1991 ) and with ~5 kBT of strain at 0° curvature . ( Its flexural rigidity was chosen such that the fully straightened dimer carries a strain energy similar to our estimated value , 19 pN·nm; see Figure 7—figure supplement 1a . ) The energy of the lateral bonds the dimer forms with its neighbors in the microtubule wall was assumed to follow a simple ( Lennard-Jones ) function of the bend angle ( Figure 7—figure supplement 1b ) . These mechanical strain and lateral bond energies were added together to calculate a total free energy landscape ( Figure 7—figure supplement 1c ) . The predicted landscape implies a curling reaction that proceeds via a high-energy transition state . We envision that the lateral bonds are short-range interactions , such that they break before much curling has developed . With this assumption , the high-energy transition state should closely resemble the initial , straight conformation ( and the curling reaction can be considered ‘Eyring-like’ [Howard , 2001] ) . 10 . 7554/eLife . 28433 . 013Figure 7 . Free energy landscape for a curling αβ-tubulin . ( a ) The model considers a single αβ-tubulin ( highlighted ) as it bends outward from a microtubule . For simplicity , only two protofilaments are depicted . The curling subunit is shown ( arbitrarily ) at the base of a previously formed protofilament curl . ( b ) Hypothetical free energy landscapes for wild-type ( red curve ) and mutant T238V tubulin ( blue curve ) as functions of subunit curvature , φ . Lateral bonding initially holds the tubulin in a straight conformation ( strained , φ = 0° ) . Curling then proceeds via a high-energy transition state ( open circles ) , which is reached without the development of much curvature ( φ ~ 2° ) . Stronger lateral bonding in T238V increases the height of the transition energy barrier , reducing the rate of curling relative to wild-type . Relaxation from the highly strained transition state to the naturally curved ground state ( at φ = 23° , with free energy arbitrarily set to zero ) drives movement in the wave assay . Because T238V and wild-type have similar transition energies , they produce conformational waves with similar energy . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 01310 . 7554/eLife . 28433 . 014Figure 7—figure supplement 1 . Free energy landscape for a single curling αβ-tubulin subunit , calculated by adding independent contributions from mechanical strain and lateral bonding . ( a ) Mechanical strain energy , U , is calculated as a function of bend angle , φ , by assuming the αβ-tubulin subunit behaves like a slender elastic rod with a naturally bent shape , a bend angle of φ = 23° when fully relaxed , and a constant flexural rigidity , chosen such that the fully straightened dimer carries a strain energy similar to our estimated value ( ~19 pN·nm ) . Strain energy functions for wild-type and mutant T238V tubulin are assumed to be identical . Yellow shading in the cartoons indicates mechanical strain . ( b ) The energy of the lateral bond that the subunit makes with its neighbors in the microtubule lattice , G , follows a simple ( Lennard-Jones ) function of bend angle . Mutant T238V tubulin ( blue curve ) is proposed to form stronger lateral bonds than wild-type ( red curve ) . ( c ) The total free energy , T , is the sum of mechanical strain and lateral bond energies ( U + G ) . Stronger lateral bonding by T238V tubulin increases the height of the transition energy barrier , thereby slowing the rate of curling relative to wild-type . Provided the lateral bonds are highly localized , such that they break before much curling has developed , their contribution to the conformational wave energy is small ( irrespective of the exact shapes of the energy functions ) . Thus , changes in curling rate can occur without necessarily changing the conformational wave energy . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 014 According to this model , the slower disassembly of T238V microtubules is explained by an increase in the height of the activation barrier , which could arise either because the energy of the transition state is higher or because the energy of the starting state ( i . e . , when the tubulin is straight and laterally bonded ) is lower . Our data exclude the possibility of a substantially higher transition state energy because this would lead to a higher wave energy for the mutant , which we did not observe . We therefore propose that the mutation specifically strengthens lateral bonds , thereby lowering the energy of the starting state and raising the activation barrier , without altering the intrinsic 23° curvature or the mechanical rigidity of the dimer ( Figure 7b ) . To further explore whether the T238V mutation might strengthen lateral bonds in the microtubule lattice , we took two additional approaches . In one approach , we probed the conformation of tubulin in the lattice using the plus-end-tracking EB1-family protein , Bim1 . Like other EB1 proteins , ( Zanic et al . , 2009; Bieling et al . , 2007 ) Bim1-GFP brightly decorates the growing plus-ends of wild-type yeast microtubules , with a strong preference for the growing ends over the remainder of the filament lattice ( Geyer et al . , 2015 ) . We observed similar bright Bim1-GFP decoration at the growing plus-ends of mutant T238V microtubules as well , but the lattice of the mutant T238V microtubules retained an abnormally high affinity for Bim1-GFP ( Figure 5—figure supplement 1c and d ) . This observation indicates that the lattice conformation of T238V tubulin retains structural characteristics that are normally found only near the ends of growing microtubules ( GTP-cap-like ) , which may be associated with stronger lateral bonding in the lattice . As a second approach for examining the effects of the T238V mutation , we devised a new ‘plucking’ assay to measure the forces required to remove tubulins from growing microtubule ends . We fortuitously found , using the same flexible tethers devised for the wave assay ( i . e . , single anti-His antibodies bound to a His6 tag on the C-terminal tail of β-tubulin ) , that individual beads could be linked to the growing ends ( rather than the sides ) of single , dynamic microtubules . If increasing tension was then applied ( 0 . 25 pN·s−1 ) , the end-bound bead could be detached ( Figure 8a and b ) . End-bound beads were readily detached in this manner , but side-bound beads generally did not detach , even at the maximum laser trap tension ( ~40 pN under the conditions used here ) . Usually , detaching an end-bound bead by force triggered immediate disassembly of the microtubule ( 43 of 57 detachments , ~75% ) , which confirms that tubulin dimers were forcibly removed ( Figure 8c ) . Given the single antibody-based linkages , the number of plucked dimers was probably low , but possibly greater than one or two . The average force required to pluck tubulins from a wild-type microtubule end was 8 . 3 ± 0 . 6 pN ( Figure 8d and f ) . Plucking tubulins from mutant T238V microtubules required considerably more force , 19 . 0 ± 1 . 6 pN on average ( Figure 8e and f ) . This higher plucking force is consistent with stronger lateral bonds , although it could arise from a strengthening of longitudinal bonds , or a strengthening of both kinds of bonds . Whether it would also occur in the context of a disassembling end remains uncertain; nevertheless , the observation shows that mutant T238V tubulin forms relatively stronger tubulin-tubulin bonds compared to wild-type , at least in the context of an assembling end . 10 . 7554/eLife . 28433 . 015Figure 8 . More force is required to pluck hyperstable mutant tubulin subunits from the microtubule end . ( a ) A bead is tethered to the end of a growing microtubule via a single antibody bound to the β-tubulin C-terminus and then tested with a 0 . 25 pN·s−1 force ramp . ( b ) Usually , detaching the bead by force triggers immediate disassembly of the microtubule ( 43 of 57 detachments , ~75% ) , indicating that tubulin dimers were forcibly removed . ( c ) Selected frames from Video 3 , showing an end-tethered bead under tension ( 2 . 2 s ) , detachment of the bead ( 3 . 2 s ) , and tip shortening ( 4 . 3 s ) . White arrows mark the coverslip-anchored segment of the microtubule . Yellow chevrons mark the plus end . ( d , e ) Example records of tensile force versus time for beads tethered to the ends of wild-type and hyperstable mutant T238V microtubules . Arrows mark plucking forces . Gray dots show raw data . Colored traces show same data after smoothing with a 500 ms boxcar average . ( f ) Distributions of plucking force for wild-type and mutant T238V tubulins . Dotted vertical lines indicate averages for wild-type tubulin , 8 . 3 ± 0 . 6 pN ( mean ± SEM; N = 24 ) , and for T238V , 19 . 0 ± 1 . 6 pN ( N = 19 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 01510 . 7554/eLife . 28433 . 016Video 1 . Example of wave assay . A bead tethered to the side of a coverslip-anchored microtubule is initially held under laser trap tension ( here , ~1 pN ) . The distal plus end of the microtubule is severed by laser scissors ( at 0 s ) , triggering disassembly . When the disassembling end reaches the bead , it causes a brief pulse of motion ( 0 . 7 s ) before the bead detaches ( 1 . 0 s ) . After bead detachment , the microtubule continues disassembling while the stage also moves rightward under feedback control . Red dashes mark the approximate location of the center of the laser trap . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 01610 . 7554/eLife . 28433 . 017Video 2 . Second example of wave assay . A bead tethered to the side of a coverslip-anchored microtubule is initially held in the laser trap , at low tension ( <1 pN ) . Feedback control is initiated ( at −5 . 6 s ) to apply higher tension ( 4 pN ) , and then the distal plus end of the microtubule is severed ( 0 s ) . The bead detaches when it is reached by the disassembling end ( 1 s ) . After bead detachment , the microtubule continues disassembling while the stage moves rightward under feedback control . Red dashes mark the approximate location of the center of the laser trap . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 01710 . 7554/eLife . 28433 . 018Video 3 . Example of plucking force assay . A bead linked to the assembling plus end of a coverslip-anchored microtubule is subjected to increasing tension until the bead detaches . After bead detachment , the microtubule plus end disassembles , indicating that tubulin dimers were forcibly removed from the end . Two views of the same movie are shown . At right , the plus end and the coverslip-anchored portion of the microtubule are indicated by red and yellow markers , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 28433 . 018
As with any cantilevered spring , the amount of force that curling protofilaments can produce depends on how they are coupled to the object on which they are pushing ( Molodtsov et al . , 2005; Efremov et al . , 2007 ) . The amount of strain energy they carry is a more fundamental quantity , and therefore less sensitive to geometric details of the coupling . Ultimately , this strain energy determines the maximum force-generating capacity of the conformational wave mechanism . It is also fundamentally important for all current models of microtubule dynamic instability , and numerous previous studies have attempted to estimate its magnitude . Thermodynamic approaches ( Desai and Mitchison , 1997; Caplow and Shanks , 1996; Howard , 2001 ) and analyses based on the bending rigidity of intact microtubules ( Mickey and Howard , 1995 ) have yielded estimates spanning more than an order of magnitude ( Figure 4—figure supplement 1d ) . But these methods can only infer the stored strain indirectly . Our wave assay has provided a more direct approach . To measure the energy carried by the conformational wave we modified an assay pioneered in a previous study , ( Grishchuk et al . , 2005 ) adding a feedback-controlled laser trap and other improvements to prevent the microbeads used in the assay from restricting the curling of the protofilaments . Our results clearly demonstrate the spring-like elasticity of the protofilaments and establish that they carry a very substantial amount of curvature strain ( >74 kBT ) , which can be harnessed to perform mechanical work . Our data further show that movement in the assay is likely driven by a lateral push mechanism , in which the curling protofilaments push laterally on the bead , causing it to pivot around the flexible tether linking it to the microtubule . Based on this arrangement , we estimate that the measured work output derives from outward curling of as many as 16 tubulin dimers , implying an energy per dimer of at least 19 pN·nm ( 4 . 7 kBT , or 2 . 7 kcal mole−1 ) . Ultimately this energy is derived from the chemo-mechanical cycle of tubulin . Soon after a tubulin dimer assembles into a microtubule , GTP is hydrolyzed and a portion of the free energy liberated by this chemical reaction is stored as curvature strain in the microtubule lattice ( Desai and Mitchison , 1997; Nogales and Wang , 2006 ) . Our estimate of this stored strain represents ~22% of the total free energy available from GTP hydrolysis ( see Figure 4—figure supplement 1d ) , indicating that tubulin can convert chemical energy into mechanical work with an efficiency similar to other , more conventional molecular motors , such as kinesin , ( Howard , 1996 ) myosin , ( Rief et al . , 2000 ) and dynein ( Gennerich et al . , 2007 ) . The curvature strain energy carried by curling protofilaments is easily sufficient to make a major contribution to the motility of isolated kinetochores and recombinant kinetochore subcomplexes in vitro ( Volkov et al . , 2013; Akiyoshi et al . , 2010; Asbury et al . , 2006; Powers et al . , 2009; Tien et al . , 2010 ) . The mechanical work harnessed by these reconstituted couplers from a disassembling microtubule tip is the product of the opposing force , F , against which they move , multiplied by the distance moved , δ . The distance moved per released tubulin dimer , δ = 0 . 61 nm , is known from the structure of the microtubule lattice ( i . e . , the 8 nm length of an αβ-tubulin dimer divided by 13 protofilaments ) ( Amos and Klug , 1974 ) . The maximum force against which a yeast kinetochore has been observed to track in vitro with disassembly is F = 8 . 5 ± 1 . 9 pN ( based on a line-fit to tracking speeds measured as a function of tension ) ( Akiyoshi et al . , 2010 ) . Multiplying these gives F·δ = 5 . 2 ± 1 . 2 pN·nm of work per released dimer ( equivalent to 1 . 3 ± 0 . 3 kBT per dimer , or 0 . 74 ± 0 . 17 kcal mole−1 ) , which is lower than our estimate of curvature strain in the protofilaments , by ~3 fold . Tip-couplers made by linking recombinant Dam1 complexes at high density to beads via long tethers can sometimes track against even higher forces , up to F = 30 pN ( Volkov et al . , 2013 ) . This maximum force corresponds to F·δ = 18 pN·nm per released dimer ( 4 . 5 kBT per dimer , or 2 . 6 kcal mole−1 ) , a value nearly equal to our estimated protofilament strain . The most directly comparable in vivo results are the classic microneedle-based measurements of Nicklas , who found that the total force required to stall anaphase chromosome movement in grasshopper spermatocytes was 700 pN ( Nicklas , 1983 , 1988 ) . Assuming this load was shared by 15 kinetochore-attached microtubules leads to an often-cited estimate of F ≈ 50 pN per microtubule ( Nicklas , 1983 , 1988 ) . If all the energy driving chromosome movement was derived from disassembly of these kinetochore-attached microtubule tips , then the work harnessed per released dimer would be F·δ = 30 pN·nm . Our results suggest that the majority of this energy could be derived from protofilament curvature strain . For decades , the conformational wave model has remained a compelling but unproven hypothesis for kinetochore motility . By establishing that protofilaments carry enough curvature strain energy to make a major contribution to kinetochore movement , our findings lend strong support to the conformational wave hypothesis . Our study also reveals how the rate of microtubule disassembly can be altered dramatically by tubulin mutations that do not necessarily affect the amount of lattice strain or the intrinsic protofilament curvature . While the hyperstable mutant T238V tubulin that we examined here did not exhibit low lattice strain , we speculate that such low-strain mutants exist . We envision that the wave assay developed here will be useful for identifying low-strain tubulin mutants and for examining the mechano-chemistry of other recombinant tubulins .
Plasmids to express wild-type yeast αβ-tubulin with a His6 tag fused to the C-terminus of β-tubulin were previously described ( Ayaz et al . , 2014 , 2012; Johnson et al . , 2011 ) . A plasmid to express the T238V mutation of Tub2p ( yeast β-tubulin ) was made by QuikChange mutagenesis ( Stratagene ) , using an expression plasmid for wild-type Tub2 as template and with primers designed according to the manufacturer’s instructions . The integrity of all expression constructs was confirmed by DNA sequencing . Wild-type or mutant yeast αβ-tubulin was purified from inducibly overexpressing strains of S . cerevisiae using nickel affinity and ion exchange chromatography ( Ayaz et al . , 2014 , 2012; Johnson et al . , 2011 ) with the exception that the T238V mutant was eluted from the nickel-affinity column with 200 mM NaCl . T238V nickel elution fractions were treated with Universal Nuclease ( Pierce ) at room temperature for 1 hr prior to ion exchange chromatography . Tubulin samples for the laser trap assays were prepared at UT Southwestern , aliquoted and snap-frozen in storage buffer ( 10 mM PIPES pH 6 . 9 , 1 mM MgCl2 , 1 mM EGTA ) containing 50 μM GTP , shipped on dry ice to the University of Washington , and stored at −80°C . For each experiment , a small channel ~3 mm wide was formed by bonding a plasma-cleaned glass coverslip to a clean glass slide using two parallel strips of double-stick tape . GMPCPP-stabilized , biotinylated seeds were assembled from bovine brain tubulin , anchored to the coverslip surface , and then washed with a solution of 1 mM GTP in BRB80 ( 80 mM PIPES , 120 mM K+ , 1 mM MgCl2 and 1 mM EGTA , pH 6 . 9 ) as previously described ( Akiyoshi et al . , 2010; Franck et al . , 2010; Powers et al . , 2009; Franck et al . , 2007 ) . Alternatively , for some experiments , axonemes purified from sea urchin sperm ( Waterman-Storer , 2001 ) were adsorbed directly to the coverslips and then washed as above . From the coverslip-anchored seeds or axonemes , dynamic microtubule extensions were grown and simultaneously decorated with anti-His-beads ( prepared as described below ) by introducing a suspension of the beads together with ~1 μM of either wild-type or mutant T238V yeast tubulin , freshly thawed , in microtubule growth buffer ( BRB80 supplemented with 1 mM GTP , 5 mM DTT , 25 mM glucose , 200 µg mL−1 glucose oxidase , 35 µg mL−1 catalase ) and then incubating the slide at 30°C for ~20 min . Only the microtubule minus ends were anchored to the glass coverslip – otherwise the filaments were unsupported . To prepare the anti-His-beads , commercially available streptavidin-coated polystyrene microbeads ( Spherotech Inc . , Libertyville , IL ) were functionalized by incubation of ~36 pM beads with 25 pM biotinylated anti-Penta-His antibodies ( #34440 , Qiagen , Valencia , CA ) for 30 min , washed extensively , and stored at 4°C for up to several months . Just prior to each experiment , a small aliquot of the beads was incubated with a mixture of plain and biotinylated BSA ( at 40 and 0 . 4 mg mL−1 , respectively ) for >30 min , diluted into growth buffer with tubulin , and then used as described above . Pre-incubation with BSA was important for preventing non-specific attachment of the sparsely anti-His-decorated beads to the microtubules . Control experiments with beads lacking anti-His antibody confirmed that the attachments were specific after the BSA pre-incubation . To ensure that most beads attached via single antibodies , the ratio of antibodies to beads during bead functionalization was kept very low , ~1:1 , such that the fraction of beads under manual manipulation that would attach to the growing end of a microtubule was less than 50% . Active beads attached readily to growing ends , but not to the sides of microtubules . Their preference for growing ends was expected because the anti-His antibodies on each bead should become quickly and stably occupied by individual , unpolymerized ( and His-tagged ) tubulin dimers upon initial mixing of the beads and tubulin . Thus , bead-microtubule attachments presumably occurred via incorporation of bead-tethered tubulins into growing ends . Laterally attached beads , which are required for the wave assay , were nevertheless found readily after microtubule polymerization had commenced for ~20 min . These lateral attachments presumably arose by polymerization of microtubules past beads that were initially end-attached . Our combination laser trap and laser scissors instrument has been described previously ( Franck et al . , 2010 ) . Briefly , the microscope incorporates two lasers , a 1064 nm wavelength laser for trapping and a 473 nm laser for cutting microtubules . The trapping laser is focused to a diffraction-limited spot in the center of the field of view and the cutting laser is focused to an ellipse several micrometers away , to ensure that it does not interfere with trap operation . Both lasers are controlled independently by manual shutters . Individual microtubules can be severed by brief exposure to the cutting laser ( <1 s ) . To perform the wave assay , a suitable laterally attached bead was first selected and placed under laser trap tension . The bead-microtubule assembly was bent slightly upward , away from the coverslip , to prevent interactions between the bead and the coverslip , or between the disassembling tip and the coverslip , which would have interfered with the movements generated by the conformational wave . The distal , growing end of the microtubule was then severed using laser scissors to induce disassembly . The desired load was maintained by adjusting the position of the specimen stage under feedback control , implemented using custom software written in LabView ( Source code 1 ) . Significant forces could only be applied in the longitudinal direction , along the axis of the microtubule , because of the arrangement of the assay , with the beads tethered to flexible microtubule extensions anchored only by their minus ends to the coverslip . Flexibility of the unsupported microtubule extensions prevented the application of piconewton-scale forces in transverse directions . Bead-trap separation was sampled at 40 kHz while stage position was updated at 50 Hz to maintain the desired load for as long as the bead remained attached to the microtubule . The bead and stage position data were decimated to 200 Hz before storing to disk . Brief recordings ( <20 s ) were sufficient to capture the wave pulses . Up to 40 events could be recorded during a single 1 hr experiment , depending on the scarcity of suitably attached beads . Individual amplitudes and risetimes for all recorded pulses , as well as the means and standard errors for each measurement condition , are given in Supplementary file 1 . The beads in our experiments were tethered to the microtubules through a linkage that consisted of streptavidin on the bead surface , a biotinylated anti-Penta-His antibody ( mouse monoclonal IgG1 ) , and the C-terminal tail of β-tubulin , which is a disordered 30-amino-acid polypeptide segment exposed at the microtubule surface . Approximate lengths for streptavidin , 7 nm , and for the antibody , 18 nm , were estimated from PDB structures 1AVD and 1IGT , respectively . A length of 3 . 6 Å per amino acid was assumed for the C-terminal tail of β-tubulin , based on lengths measured for other mechanically unfolded polypeptides ( e . g . , see Schwaiger et al . , 2002 ) , yielding an estimate for the 30-amino-acid tail of 11 nm ( excluding the His6 tag , which is presumably bound up by the antibody ) . Adding these values for streptavidin , IgG , and the β-tubulin tail yielded a total of 36 nm for the complete tether . We considered deviations from this estimated length to examine how sensitively it would affect the predicted curves for stall force-vs-bead diameter and for unloaded amplitude-vs-bead diameter . The predicted curves were similar and fit our data well for tethers ranging from 30 to 42 nm ( see Figure 4 ) . Fluorescence imaging of growing microtubule tips decorated with Bim1-GFP was performed and analyzed as previously described , ( Geyer et al . , 2015 ) with the use of wild-type or T238V yeast microtubules grown in 1 mM GTP and in the presence of 50 nM Bim1-GFP . Beads for the plucking force assay were prepared and dynamic microtubule extensions grown from coverslip-anchored seeds exactly as described above for the wave assay . Single freely diffusing beads were selected and held near an individual growing microtubule end using the laser trap . Once the bead attached to the microtubule , increasing tension was applied with feedback control to maintain a constant loading rate , 0 . 25 pN s−1 , until the bead detached . Upon bead detachment , microtubule tip state was determined visually , from video-enhanced differential interference contrast ( VE-DIC ) images displayed live during the experiments . Usually , the microtubule began disassembling immediately after detachment of the bead ( 43 of 57 detachments , ~75% ) , which indicates that tubulin dimers were forcibly removed . Only detachments that were followed immediately by microtubule disassembly were included in the analyses of plucking force . The same stocks of antibody-decorated beads were used for the plucking force measurements with wild-type and T238V tubulin . Thus , the different plucking strengths cannot be attributed to different numbers of antibodies on the beads , or different types of antibodies , or different numbers of bonds between the beads and the microtubules . All the individual plucking force values are given in Supplementary file 1 . | Before a cell divides it must separate its chromosomes , the ribbons of DNA that carry its genes . To do this , filaments called microtubules attach by their ends to the chromosomes and then shorten , pulling the chromosomes to opposite sides of the cell . The microtubules are made of thousands of subunits packed together to form miniature tubes , and shorten by losing subunits from their ends . Why don’t the chromosomes simply fall off the ends of these microtubules , which are crumbling under their grip ? How can a crumbling filament exert a pulling force ? The shape of the ends of the microtubules suggests a possible answer . The subunits that make up each microtubule are arranged in rows , called protofilaments , that run along the length of the microtubule . When a microtubule shortens , its protofilaments first curl outward from the end and then crumble apart . If the curling protofilaments are strong enough , they could act like springs , hooking the chromosome and pulling on it as they curl outward . Curling protofilaments can exert some pulling force , but how much force was not known . To investigate , Driver et al . used an instrument called a laser trap , or laser tweezers , to record tiny movements and forces exerted by individual microtubules on microscopic plastic beads . The microtubules came from yeast cells , and had been engineered to carry a tag on their surface that enabled them to attach to the beads in a way that did not interfere with the curling action of the protofilaments . The experiments revealed that curling protofilaments do indeed behave like strong springs , and can make a major contribution to moving chromosomes . Fully understanding how microtubules pull on chromosomes could help to design anti-cancer drugs that prevent cells from dividing . Drugs that target microtubules are already used against certain cancers , but they cause considerable side effects because microtubules are important in many types of cells . However , drugs that specifically prevent curling protofilaments from tugging on chromosomes could potentially treat cancer with fewer side effects . It remains to be seen whether such drugs can be developed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | Direct measurement of conformational strain energy in protofilaments curling outward from disassembling microtubule tips |
GSK-3 is an essential mediator of several signaling pathways that regulate cortical development . We therefore created conditional mouse mutants lacking both GSK-3α and GSK-3β in newly born cortical excitatory neurons . Gsk3-deleted neurons expressing upper layer markers exhibited striking migration failure in all areas of the cortex . Radial migration in hippocampus was similarly affected . In contrast , tangential migration was not grossly impaired after Gsk3 deletion in interneuron precursors . Gsk3-deleted neurons extended axons and developed dendritic arbors . However , the apical dendrite was frequently branched while basal dendrites exhibited abnormal orientation . GSK-3 regulation of migration in neurons was independent of Wnt/β-catenin signaling . Importantly , phosphorylation of the migration mediator , DCX , at ser327 , and phosphorylation of the semaphorin signaling mediator , CRMP-2 , at Thr514 were markedly decreased . Our data demonstrate that GSK-3 signaling is essential for radial migration and dendritic orientation and suggest that GSK-3 mediates these effects by phosphorylating key microtubule regulatory proteins .
Glycogen synthase kinase ( GSK-3 ) α and β are serine/threonine kinases that act as key downstream regulators in multiple signaling pathways , including Wnt/β-catenin , receptor tyrosine kinase ( RTK ) /PI3K , and Sonic hedgehog ( Shh ) ( Kaidanovich-Beilin et al . , 2012 ) . GSK-3s act via mechanisms that include regulation of transcription factors , control of multiple aspects of cellular metabolism , and phosphorylation of cytoskeletal proteins ( Hur and Zhou , 2010; Kaidanovich-Beilin and Woodgett , 2011 ) . Most often , although not invariably , GSK-3s function as negatively acting kinases by inhibiting the functions of substrates at baseline . Inhibition is then relieved via signaling pathways that engage GSK-3 ( Doble and Woodgett , 2003; Kaidanovich-Beilin and Woodgett , 2011 ) . For most GSK-3 substrates , phosphorylation by another kinase near the GSK-3 site ( ‘priming’ ) is required for , or enhances , GSK-3 substrate phosphorylation ( Cohen and Frame , 2001; Doble and Woodgett , 2003; Kaidanovich-Beilin and Woodgett , 2011 ) . Priming kinases for GSK-3 substrates include cyclin dependent kinase-5 ( cdk5 ) , a kinase that is known to regulate important neurodevelopmental events like radial migration ( Tanaka et al . , 2004; Cole et al . , 2006; Li et al . , 2006; Xie et al . , 2006 ) . In the nervous system , GSK-3β has long been thought to be a target of lithium used in treatment of bipolar disorder ( Klein and Melton , 1996; O'Brien et al . , 2004 ) . Some of the GSK-3β effects related to lithium actions are due to regulation of signaling downstream of dopamine receptors ( Beaulieu et al . , 2004 , 2008; Urs et al . , 2012 ) . More recently GSK-3 signaling has been implicated in the pathogenesis of schizophrenia ( Emamian et al . , 2004; Mao et al . , 2009; Emamian , 2012 ) . Disrupted in Schizophrenia-1 ( DISC1 ) , mutated in some familial cases of schizophrenia , is thought to function in part by modulating GSK-3β effects on progenitor proliferation ( Mao et al . , 2009; Singh et al . , 2011 ) . Despite the obvious importance of GSK-3 signaling in pathogenesis and treatment of psychiatric disorders , there are important gaps in information on the role of GSK-3 in the developing brain . It has clearly been established that GSK-3 signaling is a strong regulator of radial progenitor proliferation in the developing cerebral cortex and that these effects are at least partly meditated through β-catenin ( Chenn and Walsh , 2002; Kim et al . , 2009 ) . Additionally , a recent study demonstrated an important role for GSK-3 in regulating INP amplification , an effect associated with GSK-3β binding to the scaffolding protein Axin ( Fang et al . , 2013 ) . Thus , critical roles for GSK-3 signaling in processes that control neuron number in the developing telencephalon have been established . In contrast , functions of GSK-3 in regulating developing cortical neurons are much less clear . Multiple in vitro studies have suggested roles for GSK-3 in regulating neuronal polarity and axon growth and branching ( see Hur and Zhou , 2010 for review ) . These functions are thought to be mediated via GSK-3 phosphorylation of microtubule-associated proteins ( MAPs ) including Collapsin response mediator protein-2 ( CRMP-2 ) ( Yoshimura et al . , 2005 ) , Adenomatous polypsis coli ( APC ) ( Shi et al . , 2004; Zhou et al . , 2004 ) , Tau ( Stoothoff and Johnson , 2005 ) , microtubule-associated protein 1B ( MAP1B ) ( Trivedi et al . , 2005 ) , Doublecortin ( DCX ) ( Bilimoria et al . , 2010 ) , and subsequent regulation of cytoskeletal dynamics . In general , inhibition of GSK-3β via serine 9 ( ser9 ) and GSK-3α via serine 21 ( ser21 ) phosphorylation and subsequent relief of phosphorylation of downstream targets is thought to be required for the formation of the axon and subsequent axonal growth ( Jiang et al . , 2005; Hur and Zhou , 2010 ) . In a similar vein , a study employing in utero electroporation of an activating construct suggested that GSK-3 inhibition was essential for radial-guided cortical neuronal migration downstream of STK11 ( LKB1 ) via a mechanism involving APC ( Asada and Sanada , 2010 ) . A prediction of this work might be that GSK-3 deletion would enhance axon growth and radial migration . However , to date these effects of GSK-3 on neuronal polarity and migration have not been confirmed with mouse genetic studies . Further , mice with point mutation knockins that prevent ser9/21 phosphorylation are viable and have not been reported to show defects in neuronal morphology or migration ( McManus et al . , 2005; Gartner et al . , 2006 ) . Finally , GSK-3 may regulate migration and morphology of cortical neurons via entirely different pathways for example via mediating effects of semaphorin signaling ( Chen et al . , 2008; Renaud et al . , 2008; Nakamura et al . , 2009 ) . We have now assessed GSK-3 functions in newly born cortical neurons using Neurod6 ( Nex ) -Cre ( Goebbels et al . , 2006 ) to mediate recombination of Gsk3 floxed alleles in INPs and newly born excitatory neurons . We demonstrate , surprisingly , that GSK-3 activity is essential for radial neuron migration in all areas of the cortex and in the hippocampus . In contrast , tangential migration is not affected after Gsk3 deletion in cortical interneuron precursors . Remarkably , the migration effects appear to be independent of Wnt/β-catenin signaling that mediates GSK-3 functions in neuronal progenitors . The few upper layer neurons that reached their normal location exhibited strikingly abnormal orientation of basal dendrites . GSK-3 control of migration and morphology is correlated with the regulation of phosphorylation of DCX on ser327 and CRMP-2 on Thr514 . We conclude that GSK-3 is a critical regulator of neuronal migration and morphogenesis and that GSK-3 regulation is mediated by phosphorylation of key cytoskeletal proteins .
To investigate the function of GSK-3 in developing cortical neurons , we generated Gsk3a−/−Gsk3bloxp/loxp:Neurod6-Cre mice ( Gsk3:Neurod6 ) . The Neurod6-Cre line induces recombination in intermediate progenitors of the dorsal telencephalon and early postmitotic neurons beginning at approximately embryonic day 11 ( E11 ) ( Goebbels et al . , 2006 ) . Prior studies using crosses with reporter lines indicate that recombination occurs in virtually all excitatory pyramidal neurons in the dorsal telencephalon ( Goebbels et al . , 2006; Monory et al . , 2006 ) . Western blot analysis of lysates from the whole cortex , shows a 60% decrease in GSK-3β protein at E19 . 5 as compared with heterozygous litter mates ( Figure 1C ) . Remaining GSK-3β protein in the mutants is likely due to lack of recombination in interneurons and developing glia . At E19 , Gsk3:Neurod6 brains are roughly the same size as littermate controls . However , Gsk3:Neurod6 mice die shortly after birth ( P0–P3 ) for reasons that have not yet been determined . 10 . 7554/eLife . 02663 . 003Figure 1 . GSK-3 signaling is essential for proper lamination of the developing cortex . ( A–A' ) Cux-1 staining ( red ) in coronal sections from control and Gsk3:Neurod6 mice at E19 . 5 . Cux-1 neurons are strikingly mislocalized in Gsk3:Neurod6 mutants ( orange arrows ) including a small population of neurons that remain in the ventricular zone ( yellow arrowhead ) . Nuclei were counterstained with DRAQ5 . Scale bar = 500 μm . ( n = 4 ) . ( B–B' ) Cux-1 staining in parasagittal vibratome sections from control and Gsk3:Neurod6 mutants at E18 . 5 . Cux-1 expressing neurons ( arrows ) are mislocalized in Gsk3:Neurod6 mutants and populate the deeper layers of the cortex along the entire rostrol/caudal axis . Scale bar = 200 μm . ( C ) Representative Western blot confirms strongly reduced GSK-3β protein levels in the E19 . 5 Gsk3:Neurod6 cortex compared to heterozygous control ( n = 3 het control , n = 3 CKO ) . Relative Density *p<0 . 05 , unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 00310 . 7554/eLife . 02663 . 004Figure 1—figure supplement 1 . Migration defect apparent by E16 after Gsk3 deletion in cortical excitatory neurons . Coronal cryostat sections at E16 . Layer 6 TBR1 neurons populate appropriate layers in control and Gsk3:Neurod6 mutants . Cux-1 ( red ) expressing neurons , marker for upper layer 2/3 neurons , are mislocalized in Gsk3:Neurod6 mutant coronal sections . L1 staining labels axon tracts . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 004 To explore neuronal functions of GSK-3 , we first assessed cortical lamination at E16 , a time of rapid neuronal migration along radial glial processes . Deep layer neurons appeared to be normally positioned ( Figure 1—figure supplement 1 ) . Thus staining with Tbr1 , a layer 6 marker , revealed a distinctive band in the deeper layers of the cortex in both controls and Gsk3:Neurod6 mutants . In contrast , we noted clear abnormalities in the localization of Cux1 expressing neurons that normally populate Layers 2/3 ( Figure 1—figure supplement 1 ) . A clear band of Cux1 positive neurons has formed in controls by E16 . In contrast Gsk3:Neurod6 mice exhibit a dispersion of Cux1 cells with fewer neurons reaching the outermost layer , even at this early developmental stage . Dramatic mislocalization of layer 2/3 neurons was apparent by E19 . 5 . Coronal sections through developing somatosensory cortex showed a large population of Cux1 expressing neurons essentially 'stuck' in the intermediate zone and throughout the deeper cortical layers ( orange arrows ) ( Figure 1A–A’ ) . Indeed some Cux1-expressing neurons in mutants were observed in the ventricular zone ( yellow arrowhead ) . The migration defect in Gsk3:Neurod6 mutants was striking along the entire rostrol/caudal axis at E18 . 5 , as observed in parasagittal sections ( arrows ) ( Figure 1B–B' ) . The migration defect was particularly prominent anteriorly , a developmental profile corresponding to the neurogenic gradient of the developing cortex ( Caviness et al . , 2009 ) . We also generated Gsk3aloxp/loxp , Gsk3bloxp/loxp: Neurod6-Cre mice ( Gsk3loxp:Neurod6 ) using a Gsk3aloxp/loxp mouse line , which harbors loxp sites flanking exon 2 of Gsk3a . Western blot analysis of lysates from the whole cortex at P0 , verified an 85% decrease in GSK-3α and a 76% decrease in GSK-3β protein when compared with wild-type littermate controls ( Figure 2B ) . At P0 as expected , Gsk3loxp:Neurod6 mice showed the same migration failure of upper layer neurons as Gsk3:Neurod6 ( Figure 2A ) . Inspection and quantification , ( Figure 2A , C ) showed most Cux1+ neurons stuck in deeper layers in mutants , whereas in controls a heavy majority of neurons had already migrated to the most superficial layers . An increase in neurogenesis could in theory account for some of the Cux1+ neurons found in deeper layers , but counts of Cux1+ cells revealed no major difference in numbers between Gsk3 mutants and controls ( control Cux1/total = 32 . 45% ± 3 . 74 , mutant = 34 . 31% ± 0 . 649 , p=0 . 561 , unpaired t-test ) . 10 . 7554/eLife . 02663 . 005Figure 2 . Migration defects in Gsk3-deleted mice are persistent . ( A ) P0:Cux 1 staining ( red ) in coronal sections of Gsk3loxp:Neurod6 mutants and littermate heterozygote controls . Cux1-expressing neurons are localized to layer2/3 in controls ( denoted by yellow dashed lines ) while Cux1-expressing neurons are localized throughout the cortical plate in the mutants . ( n = 5 , scale bar = 100 μm ) . ( B ) Representative Western blot of P0 cortical lysates confirms strongly reduced GSK-3α and GSK-3β protein levels in Gsk3loxp:Neurod6 mutants when compared to Gsk3aloxp/loxpGsk3bloxp/loxp controls . GAPDH was probed as a loading control ( n = 3 control , n = 3 CKO ) . ( C ) P0 quantification of control and Gsk3loxp:Neurod6 Cux1 neurons using 8 bin analysis spanning white matter ( WM ) to the pial surface ( PS ) , ( n = 2 het control , n = 2 CKO ) . ( D–D' ) P7: Gsk3loxp:Neurod6 mutants stained with Cux1 ( red ) show persistent altered lamination with Cux1-expressing neurons spread throughout all layers of the cortex . Littermate controls show normal Cux1 distribution in layer 2/3 . Scale bar = 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 00510 . 7554/eLife . 02663 . 006Figure 2—figure supplement 1 . Gsk3 overexpression enhances radial migration . ( A ) E19 . 5 coronal sections showing normal migration after in utero electroporation of Neurod1-Cre and lox-STOP-lox-Ai9 ( red neurons ) and ( B ) Coronal sections showing abnormal neuronal migration after electroporation of Neurod1-Gsk3b and Neurod1-GFP ( green neurons ) . Enhanced migration is apparent in B . Scale bar = 200 μm . ( C ) E19 . 5 quantification of migration in control and Gsk3 over-expressing neurons using 8 bin analysis as previously described . Significantly more Gsk3 over-expressing neurons were found in the outermost bin 8 and fewer Gsk3 over-expressing neurons populated bin 6 . p-values shown in figure , unpaired t-test . ( control n = 2 mice from two individual litters , 1276 neurons vs Gsk3 overexpression n = 3 mice from two individual litters , 1959 neurons ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 006 A few Cux1 neurons migrated successfully to layer 2/3 ( area denoted by yellow lines ) in Gsk3loxp:Neurod6 mutants ( Figure 2A ) . Whether this subset of normally placed neurons did not undergo recombination at an early enough stage for migration to be regulated could not be determined . However , these properly positioned neurons exhibited abnormalities of dendritic orientation ( see Figure 5 ) . Interestingly , mice with floxed alpha rather than null alpha alleles survived somewhat longer than the Gsk3:Neurod6 mice but died after the second postnatal week ( P15–P17 ) . At P7 , Gsk3loxp:Neurod6 mutants exhibited a striking lamination phenotype involving Cux1 neurons ( Figure 2D–D' ) . This persistent lamination defect demonstrates that Gsk3 deletion results in permanent migration failure and does not simply result in a delay . To further assess functions of GSK-3 in radial migration , we over-expressed Gsk3b in newly born cortical neurons using a pNeurod1 ( Neurod1 ) vector ( Guerrier et al . , 2009 ) . Co-electroporation of Neurod1-Cre and lox-STOP-lox Ai9 in control embryos at E15 ( Figure 2—figure supplement 1A ) was compared to co-electroporation of Neurod1-Gsk3b and Neurod1-Egfp in experimental embryos ( Figure 2—figure supplement 1B ) . Importantly , over-expression of Gsk3b did not inhibit migration as might have been expected from prior studies ( Asada and Sanada , 2010 ) . In fact , Gsk3b over-expression enhanced neuronal migration and resulted in greater than normal numbers of neurons populating the outermost layers of the cortex by E19 . 5 ( Figure 2—figure supplement 1C ) . Importantly , the migration defect in the developing cortex is specific to excitatory pyramidal neurons . In order to assess interneurons , we used the Dlx5/6-Cre ( Stenman et al . , 2003 ) line to generate conditional mice lacking Gsk3 in GABAergic interneurons . A robust decrease of GSK-3β protein ( 84% ) was observed in E18 MGE lysates from Gsk3:DLX5/6-Cre mice when compared to littermate controls ( Figure 3D , E ) . Interneuron migration was monitored using the AI3 reporter line ( Gsk3-Ai3:Dlx ) . Surprisingly , in both controls and Gsk3 mutants , interneurons exhibited robust migration along the two migratory streams ( yellow arrows ) from the medial ganglionic eminence ( MGE ) ( Figure 3A–A' ) . In Gsk3 mutants , as in controls , interneurons entered all areas of the cortical plate by E19 . 5 . Quantification is shown in ( Figure 3—figure supplement 1 ) . These results are not meant to imply that migration of interneurons was normal in every respect as we did not assess migration of specific interneuron subsets . 10 . 7554/eLife . 02663 . 007Figure 3 . GSK-3 signaling is dispensable for tangential migration , but required for radial hippocampal migration . ( A–A' ) E19 . 5 coronal sections showing EYFP-expressing interneurons in heterozygous control and Gsk3:Dlx5/6 mutants crossed with the Ai3 reporter line . Gsk3-deleted interneurons ( green ) enter the cortex in two streams in both controls and mutants ( arrowheads ) . Mutants showed no overt migration defect . Nuclei were counterstained with Hoechst . ( n = 3 ) . ( B–B' ) E19 coronal sections of control and Gsk3:Neurod6 mutants showing CTIP2 ( green ) expressing neurons in the hippocampus . In the Gsk3:Neurod6 mutants , the pyramidal cell layer ( green ) does not extend laterally into a compact CA1 region and remains dispersed ( yellow arrowheads ) . Fimbrial axonal projections appear normal in Gsk3:Neurod6 mutants ( orange arrow ) . Nuclei were counterstained with DRAQ5 . Scale bar = 500 μm . ( n = 3 ) . ( C–C' ) Higher magnification of hippocampal area shown in ( B ) . The Gsk3:Neurod6 mutants show disrupted cytoarchitecture . In the mutants , DRAQ5-labeled cells are mislocalized and diffuse ( arrowheads ) and fail to form clearly defined CA1/CA3 regions of the hippocampus . The Gsk3:Neurod6 mutant mice also lack a clearly defined hippocampal sulcus ( green bars ) and dentate gyrus ( DG ) . ( D ) Representative Western blot of E18 MGE lysates confirm strongly reduced GSK-3β protein after recombination with Dlx5/6-Cre . ( E ) Quantification of protein knockdown in D ( n = 3 WT , n = 3 CKO , unpaired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 00710 . 7554/eLife . 02663 . 008Figure 3—figure supplement 1 . No apparent migration defect in Gsk3:Dlx5/6 mice . P0 quantification of Ai3-positive neurons in control and Gsk3-deleted interneurons using 8 bin analysis spanning white matter to the dorsal stream . p-values reaching significance are shown in figure , unpaired t-test . ( n = 2 controls , 3894 total cux1 neurons , n = 2 cko , 3681 total cux1 neurons ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 008 In order to assess the generality of GSK-3 regulation of radial migration , we assessed migration in developing hippocampus . The hippocampus , like the cortex , is an area where developing neurons migrate along radial glial-like processes ( Nowakowski and Rakic , 1979; Eckenhoff and Rakic , 1984 ) . Neurod6-Cre expression is evident in developing hippocampal neurons as early as E14 ( Goebbels et al . , 2006 ) , allowing us to delete Gsk3 in those cells . Pyramidal neurons generated from the hippocampal primordia undergo migration along radial processes to form CA1/CA3 and the dentate gyrus ( DG ) ( Altman and Bayer , 1990 ) . The transcription factor CTIP2 marks neurons in the developing CA1 region ( Figure 3B ) . Gsk3:Neurod6 mice exhibited a striking hippocampal migration defect . CTIP2 expressing neurons did not migrate properly ( yellow arrows ) and as a consequence CA1 did not fully develop ( Figure 3B–B' ) . As a result , CA1-3 and the DG ( arrowheads ) were disorganized , and the hippocampal sulcus was not well defined ( Figure 3C–C' ) . These defects were striking in rostral areas , as shown , although somewhat less pronounced in caudal sections ( data not shown ) . Interestingly , fimbrial axonal projections formed in Gsk3:Neurod6 mice ( Figure 3B–B' , orange arrows ) demonstrating that even though migration fails , hippocampal neurons were able to polarize and extend appropriately directed axons . To verify that the migration defect was cell autonomous and to visualize morphology of Gsk3-deleted neurons , we introduced Cre and EGFP into a subpopulation of developing neurons in Gsk3a−/−Gsk3bloxp/loxp mice . Neurod1-Cre and lox-STOP-lox lacZ/Egfp ( Z/EG ) plasmids were injected into the ventricles and co-electroporated at E14–15 . 5 . Electroporation at this age targets radial progenitors that generate mainly upper layer neurons . This co-electroporation technique allowed us to visualize individual Gsk3-deleted neurons in an otherwise control background . Labeled neurons were imaged at late embryonic and postnatal stages . Deletion of Gsk3 in individual neurons phenocopies the migration delay seen in the Gsk3:Neurod6 mutants . At E19 most control neurons were located in the upper cortical layers as expected ( Figure 4A ) . In contrast , most Gsk3-deleted neurons had cell somas that were localized to the deeper layers of the cortex and very few were found in upper layers ( Figure 4A' ) . Importantly , Gsk3-deleted neurons were clearly able to project axons ( orange arrows ) suggesting that initial polarization had proceeded in the absence of Gsk3 . Further , most Gsk3 null neurons in the cortical plate elaborated a long leading process directed toward the pial surface ( yellow arrowheads ) ( Figure 4A' ) . Thus at least the initial stages of dendritic arborization also appeared to proceed in the absence of Gsk3 . 10 . 7554/eLife . 02663 . 009Figure 4 . GSK-3 deletion delays the multipolar to bipolar transition . ( A–A' ) Representative E19 coronal sections after in utero electroporation at E14 . 5 with Neurod1-Cre and Z/EG plasmids . Electroporated cells were visualized with anti-EGFP ( green ) , and nuclei were stained with DAPI ( blue ) . Gsk3-deleted neurons remain in the deeper layers of the cortex but elaborate a long pial-directed process ( yellow arrowheads ) . Gsk3-deleted neurons elaborate axons projecting towards the corpus callosum ( orange arrows ) . Scale bar = 200 μm ( n = 5 , two independent litters ) . ( B–B' ) Coronal sections at P10 after E14 . 5 electroporation , as in A . Gsk3-deleted neurons remain in the deeper layers of the cortical plate and fail to reach layer 2/3 ( denoted with yellow bars ) . Scale bar = 200 μm ( n = 3 , 2 independent litters ) . ( C ) Higher magnification of Gsk3-deleted neurons in B' ( box ) . Gsk3-deleted neurons ( green ) in deeper layers co-label with Cux ( red ) ( orange arrows ) . Nuclei were stained with Dapi . ( D ) Quantification of control and Gsk3-deleted neurons in upper ( layer 2–3 ) vs deeper layers of the cortex at P10 . ( n = 3 , 4209 total neurons counted , 2234 control vs 1975 Gsk3 deleted ) **p=0 . 003 , unpaired t-test . ( E ) Gsk3 deletion delays the multipolar to bipolar transition . Still images from time-lapse imaging of slice cultures at 3DIV . pCAG-dsRED or Neurod1-Cre;Z/EG was injected into the ventricles of Gsk3a−/−Gsk3bloxp/loxp embryos and electroporated at E15 . Representative images were taken at time 0 , 6 , and 12 hr . Control dsRed neurons migrate through the cortical plate ( yellow , red , and blue arrows show individual neurons at the different time points ) . ( n = 2 controls ) . Gsk3-deleted neurons fail to migrate through the cortical plate and exhibit persistent multi-polar morphology ( yellow arrowheads ) . ( n = 4 mutants ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 00910 . 7554/eLife . 02663 . 010Figure 4—figure supplement 1 . Gsk3-deleted neurons polarize and are highly dynamic . ( A ) Stage progression analysis of dissociated control and Gsk3-deleted neurons . Stage 1 immature neurons display lamellapodial and filopodial protrusions . Stage 2 neurons have transitioned to form multiple short neurites ( multipolar morphology ) , and stage 3 neurons exhibit a single neurite extending to become an axon ( see Dotti et al . , 1988 ) . There is no significant delay in polarization ( n = 3 , 1341 neurons ) . ( B ) Live imaging of dissociated Gsk3-deleted neurons reveal dynamic neurites . Images were taken every 11 min . Representative images at time 0 ( red ) , 220 min ( blue ) , and 440 min ( green ) . Merged image is pseudo colored . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 010 The GSK-3 migration defect was strikingly persistent into the postnatal period . In mice electroporated at E14–15 . 5 and analyzed at P10 , Gsk3-deleted neurons remained in the deeper layers and subcortical white matter ( Figure 4B–B' ) . Egfp-expressing neurons co-labeled with Cux1 ( red ) demonstrating that they had acquired the proper laminar markers but were unable to attain the proper position ( Figure 4C , arrows ) . Again , long apical processes that reached the pial surface were elaborated by Gsk3-deleted neurons , basal dendrites formed , and axons projecting towards the corpus callosum were evident . Thus , Gsk3 regulates some critical aspect of migration but not the early stages of axon and dendrite formation . In quantifying our results , we found that in control animals the majority of electroporated neurons ( 73 . 7% ) were found in the upper layers of the cortex ( yellow dashed lines ) by P10 ( Figure 4B–B' , D ) . In contrast , Gsk3-deleted neurons remained in the deeper layers of the cortex and only 23% had reached the upper layer 2/3 by P10 ( p=<0 . 005 ) . A number of in vitro studies have suggested that Gsk3 regulates neuronal polarization . It is plausible that some defect or delay in the polarization might account for migration failure . To test this idea , we electroporated Neurod1-Cre and Z/EG into the Gsk3a−/−Gsk3bloxp/loxp cortex , plated cortical cells in dissociated culture and assessed stage progression at 3 days in vitro ( 3DIV ) . We observed no statistically significant difference in the stage progression between control and Gsk3-deleted neurons ( Figure 4—figure supplement 1A ) . Further , Gsk3-deleted neurons that successfully extended an axon remained highly dynamic ( Figure 4—figure supplement 1B ) and extended and retracted neurites . Thus at least some processes that require complex cytoskeletal regulation proceed normally in the Gsk3-deleted neurons . To assess the cell biological mechanisms of GSK-3 regulation of neuronal migration , we co-electroporated Neurod1-Cre;Z/EG or a control pCAG-dsRED construct into the lateral ventricles of control and Gsk3a−/−Gsk3bloxp/loxp mice . This was followed by live imaging of migration ex vivo in a cortical slice preparation ( Hand et al . , 2005 ) at 3DIV . In controls electroporated with pCAG-dsRED , labeled neurons transitioned from multipolar to bipolar morphology and migrated through the cortical plate over a period of 12 hr , as expected ( Figure 4E ) . The progress of individual neurons could readily be tracked and is indicated for three examples by the progress of the colored arrowheads in the three panels . In contrast , most Gsk3-deleted neurons failed to translocate through the intermediate zone and remained in a multipolar state in the outer subventricular zone ( arrowheads ) ( Figure 4E and Videos 1 , 2 ) . Further , during the 12 hr of observation most Gsk3-deleted neurons did not transition to a bipolar morphology , a step thought to be required for radial-guided migration . 10 . 7554/eLife . 02663 . 011Video 1 . Live cell imaging of radially migrating neurons in a cortical slice preparation . Control neurons electroporated at E15 . 5 migrate towards the pial surface after 3 days ex vivo . Neurons are imaged using time-lapse microscopy with images taken every 45 min for a 20-hr session . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 01110 . 7554/eLife . 02663 . 012Video 2 . Gsk3-deleted neurons fail to migrate in cortical slice preparation . Gsk3a−/−Gsk3bloxp/loxp mice electroporated at E15 . 5 with Neurod1-cre and Z/EG have an elongated multipolar stage and do not migrate towards the pial surface . Images were taken every 45 min over a 20-hr imaging session . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 012 That fact that multiple cytoskeletal proteins are GSK-3 substrates might suggest that Gsk3 deletion would have profound effects on dendrite and axonal arborization . To address GSK-3 regulation of cortical neuronal morphology , we deleted Gsk3 using in utero electroporation of Neurod1-Cre as outlined above , and analyzed dendritic morphology at P15 . As demonstrated above , most Gsk3-deleted neurons failed to migrate and populated the deeper layers of the cortex at P15 . Many of these migration arrested neurons exhibited abnormal dendritic arbors ( data not shown ) . Because improper laminar position might affect dendritic arborization , we focused analysis on a small subset of Gsk3-deleted neurons that reached layer 2/3 . Images at E16 ( Figure 1—figure supplement 1 ) and E19 ( Figure 4A' ) show that a few Gsk3-deleted neurons are normally positioned and suggest that these normally positioned neurons did not undergo a substantial delay in migration . All of the neurons elaborated dendritic arbors and extended an axon into the corpus callosum . However , many of these normally positioned Gsk3-deleted neurons exhibited markedly abnormally oriented basal dendrites ( Figure 5A , arrows ) . In many cases , basal dendrites were oriented towards the pial surface rather than towards the deeper cortical layers ( Figure 5B–D ) . Additionally , Gsk3-deleted basal processes grew longer and were more branched than those of control neurons ( Figure 5D and Figure 5—figure supplement 1C , D ) . Many Gsk3-deleted neurons also exhibited striking defects in the apical dendrite ( Figure 5C , C' , E orange arrows and Figure 5—figure supplement 1B ) . Thus , apical dendrites of the Gsk3-deleted neurons , although properly oriented towards the pial surface often extended branches close to the soma that extended apically rather than laterally ( Figure 5E , orange arrows and Figure 5—figure supplement 1B ) . 10 . 7554/eLife . 02663 . 013Figure 5 . GSK-3 signaling is required for proper dendrite orientation . ( A ) Gsk3 deleted neurons at P15 shown after in utero electroporation at E14 . 5 with Neurod1-Cre and Z/EG plasmids . Multiple neurons with obvious abnormalities in dendritic orientation were observed in the upper layers of the cortex ( orange arrows ) . Scale bar = 200 μm . ( B–B' ) Control and Gsk3-deleted neurons in the upper layers at P15 , immunostained with antibodies against eGFP ( black ) using same methods as Figure 4 . Gsk3-deleted neurons have abnormally polarized arbors indicated by orange arrows . Scale bar = 50 μm . ( C–C' ) Neurolucida reconstructions of control and Gsk3-deleted neurons in the upper layers of the cortex . The axon ( red ) projects towards the ventricle in control and Gsk3-deleted neurons . Both apical dendrites ( orange ) and basal dendrites ( blue ) are more branched ( orange arrows ) and basal dendrites ( blue ) are mispolarized ( blue arrowheads ) in Gsk3-deleted neurons . Scale bar = 100 μm . ( D ) Basal dendrite quantification . Dendrogram shows that basal dendrites more frequently project towards the pial surface in Gsk3-deleted neurons when compared to control basal dendrite orientation . ( n = 3 , n = 3 CKO; 15 control and 15 Gsk3-deleted neurons quantified ) . ( E ) Apical dendrite dendrogram indicates polarization and length of processes . Control apical dendrites project pially ( 90° ) . Numerous small apical branches form near the soma and project laterally ( orange arrows ) . Gsk3-deleted neurons also project pially-directed apical dendrites . However , Apical branches have a pially-directed orientation , resulting in abnormal morphology ( orange arrows , also see C' ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 01310 . 7554/eLife . 02663 . 014Figure 5—figure supplement 1 . Quantification of dendritic branching at P15 in control and Gsk3-deleted neurons . ( A ) Representitive images of control and Gsk3-deleted neurons used for Sholl analysis with specific processes pseudocolored for identification . ( B ) Apical dendrite Sholl analysis of Gsk3-deleted neurons shows significantly increased branching close to the soma . p=0 . 034 . Additionally , Gsk3-deleted neurons display a trend toward increased branching in areas further from the soma . ( C ) Basal dendrite Sholl analysis of Gsk3-deleted neurons shows significantly increased branching in basal dendrites in areas furthest away from the soma ( p=0 . 015 ) . ( D ) Basal dendrite sholl analysis of dendritic lengths reveals altered morphology of Gsk3-deleted neurons . Gsk3-deleted neurons have increased lengths of basal dendrites in areas beginning 75 μm away from the soma . p=0 . 023 , p=0 . 0118 , and p=0 . 0202 . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 014 Perhaps surprisingly Gsk3-deleted neurons extended axons into the callosum towards the contralateral cortex ( Figure 4A , A’ orange arrows ) . Axonal arborization in the contralateral cortex had reduced density , an abnormality that is under further investigation ( data not shown ) . Signaling via β-catenin is an obvious candidate to mediate GSK-3 regulation of migration . In the canonical Wnt cascade , Wnt signaling through frizzled receptors leads to dishevelled and GSK-3 sequestration , β-catenin accumulation and enhanced β-catenin/TCF-mediated transcription ( Kaidanovich-Beilin and Woodgett , 2011 ) . In radial progenitors , β-catenin signaling is clearly an important mediator of the effects of GSK-3 deletion on proliferation ( Chenn and Walsh , 2002; Kim et al . , 2009 ) . To determine the role of GSK-3 regulation of β-catenin in developing cortical excitatory neurons , we utilized a β-catenin ( Ctnnb1 ) mouse that harbors loxP sites flanking exon 3 ( Ctnnb1Ex3:Neurod6 ) ( Harada et al . , 1999 ) . Exon 3 encodes the residues that GSK-3 phosphorylates to signal β-catenin degradation; thus deleting exon 3 stabilizes β-catenin . Using a β-catenin antibody directed at the residues encoded by exon3 verified a reduction in this protein fragment at P0 as expected after Neurod6-mediated recombination ( Figure 6C , H ) . Migration of Cux1 neurons was entirely normal in these animals ( Figure 6A , Figure 6—figure supplement 1A ) . Further , in contrast to Gsk3:Neurod6 , Ctnnb1Ex3:Neurod6 mice survive , breed , and have no overt behavioral phenotype . They display a rostral midline defect resulting from lack of the hippocampal commissure ( data not shown ) , as seen in other models using stabilized β-catenin ( Chenn and Walsh , 2003 ) . 10 . 7554/eLife . 02663 . 015Figure 6 . Lamination in other signaling mutants . ( A ) P0 representative coronal sections of control ( heterozygous for floxed allele ) and Ctnnb1Ex3:Neurod6 mutants stained for Cux1 ( red ) and Hoechst ( blue ) . ( n = 5 ) Scale bar = 100 μm . ( B ) E18 Coronal sections of control and Dvl123:Neurod6 showing Cux-1 ( red ) and DRAQ5 ( blue ) staining . Cux-1 neurons reach layer 2/3 in both controls and Dvl123:Neurod6 triple mutants . ( n = 3 ) . ( C ) Western blot verification of protein deletion after recombination of floxed alleles with Neurod6-Cre . Ctnnb1Ex3:NeuroD6 ( n = 3 WT , n = 3 CKO ) , Dvl123:NeuroD6 ( n = 3 WT , n = 3 CKO ) . GAPDH was probed as a loading control . ( D–F ) P0 representative coronal sections of control ( heterozygous for floxed allele ) and indicated mutant lines stained for Cux1 and Hoechst . Scale bars are 100 μm , at least n = 5 per line . ( G ) Western blot verification of protein deletion in mutant lines . GAPDH was probed as a loading control . N's refer to numbers of mutants and paired controls . Stk11:Neurod6 ( n = 2 ) , Cdc42:Neurod6 ( n = 2 ) , Pten:Neurod6 ( n = 3 ) . ( H ) P0 Western Blot quantification of Ctnnb1Ex3:Neurod6 , Dvl123:Neurod6 , Stk11:Neurod6 , Pten:Neurod6 and Cdc42:Neurod6 lines . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 01510 . 7554/eLife . 02663 . 016Figure 6—figure supplement 1 . Quantification of lamination in other signaling mutants . ( A–D ) P0 quantification of Cux1 expressing neurons using 8 Bin quantification in conditional mutants and control shown in Figure 6 . ( A ) Ctnnb1Ex3:Neurod6 , ( B ) Stk11:Neurod6 , ( C ) Cdc42:Neurod6 , ( D ) Pten:Neurod6 . ( n = 2 het control mice , n = 2 CKO mice per line ) . Between 2000 and 3000 Cux1 neurons were scored in each of the control pairs and each of the mutant pairs . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 016 For additional assessment of the role of β-catenin signaling , we performed global analysis of GSK-3 transcriptional targets at E18 in control and Gsk3:Neurod6 cortical lysates using Affymetrix microarray analysis . Consistent with loss of Gsk3 in our conditional mutants , probe level information specific to exon 2 of Gsk3b showed that exon 2 was decreased by an average 2 . 15-fold . However , classic Wnt pathway target genes downstream of β-catenin/ ( TCF ) /LEF-1 transcription factors , including CyclinD1 , Brachyury , Wisp1 , Cdx1 , and Engrailed2 were unchanged ( data made available in GEO , GSE58727 ) . Finally mice in which β-catenin signaling is abrogated after Neurod6-Cre mediated recombination ( Ctnnb1loxp/loxp:Neurod6 ) also show no change in migration of Cux1 neurons ( ES Anton , personal communication , June 2014 ) . To further assess a potential role of Wnt/β-catenin signaling in cortical lamination , we created a conditional dishevelled 2loxp/loxp ( Dvl2 ) mouse ( Ohata et al . , 2014 ) . We then generated a triple mutant by crossing our floxed Dvl2loxp/loxp with existing Dvl1 and Dvl3 nulls and the Neurod6-Cre line ( Dvl123:Neurod6 ) . Deletion of all three Dvls presumably completely abrogates Wnt signaling via the canonical pathway . A robust decrease of DVL2 protein ( 91% ) was observed in P0 cortical lysates from Dvl123:Neurod6 mutants when compared to Cre− Dvl2loxp/loxp controls ( Figure 6C , H ) . These mice are born but die immediately at P0 . Remarkably , lamination in the triple allele mutant Dvl123:Neurod6 appears relatively normal at E18 ( Figure 6B ) . This result further supports the idea that GSK-3 regulation of migration is not mediated by the WNT/β-catenin cascade . Recent work utilized RNAi and GSK-3 S9A mutant constructs to conclude that STK11 ( LKB1 ) inactivation of GSK-3 via ser9 phosphorylation alters neuronal migration ( Asada and Sanada , 2010 ) . However , Gsk3 knock-in S9A/S21A Gsk3 ‘constitutively active’ mice develop normally and migration defects have not been reported ( Jiang et al . , 2005; Gartner et al . , 2006 ) . To further address the issue of GSK-3 inhibition downstream of STK11 , we genetically deleted Stk11 from developing excitatory neurons ( Stk11:Neurod6 ) . Stk11:Neurod6 mutant mice die around P20 . Western blot analysis verified a 66% decrease in STK11 ( LKB1 ) protein in Stk11:Neurod6 mutant mice when compared to heterozygous littermate controls ( Figure 6G , H ) . Perhaps surprisingly , no gross lamination abnormalities were observed ( Figure 6D , Figure 6—figure supplement 1B ) . Thus in our hands , STK11 ( LKB1 ) regulation of GSK-3 activity is not essential for radial-guided neuronal migration in vivo . We also deleted a key regulator of Par6-aPKC , Cdc42 , using Neurod6-Cre . aPKCs are also known to phosphorylate Ser9/Ser21 of GSK-3 ( Etienne-Manneville and Hall , 2003 ) . Cdc42:Neurod6 mutants also die shortly after birth , but again we observed no gross lamination defect in these mutant mice ( Figure 6E , Figure 6—figure supplement 1C ) . Western blot analysis verified an 87% decrease in CDC42 protein levels in mutants when compared to littermate wild type controls at P0 ( Figure 6G , H ) . In receptor tyrosine kinase ( RTK ) cascades , GSK-3 lies downstream of PI3K/ AKT signaling . Signals transduced through this cascade inhibit GSK-3 via phosphorylation on Ser9/Ser21 ( Hur and Zhou , 2010 ) . The phosphatase PTEN suppresses PI3K signaling . To determine if PI3K signaling affects migration , we genetically deleted Pten in neurons using Neurod6-Cre . Western blot analysis at P0 revealed an 88% decrease in PTEN protein in our conditional knockouts when compared to littermate wild-type controls ( Figure 6G , H ) . Though Pten:Neurod6 mutants die around birth , no overt lamination defects were observed ( Figure 6F , Figure 6—figure supplement 1D ) . These findings taken together suggest that regulation of GSK-3 phosphorylation at ser9/ser21 is not important in the control of radial migration . To determine the status of relevant GSK-3 targets , we conducted western blot analysis of GSK-3 substrates that have been implicated in migration regulation . Recently , the key migration regulator doublecortin ( DCX ) was shown to be a GSK-3 substrate ( Bilimoria et al . , 2010 ) . In P0 cortical lysates of Gsk3loxp:Neurod6 mutants , we observed a 61% decrease in phosphorylated DCX on Ser327 ( Figure 7A–A' ) . These results demonstrate that DCX is importantly regulated by GSK-3 in developing cortical neurons . 10 . 7554/eLife . 02663 . 017Figure 7 . Phosphorylation status of GSK-3 substrates . ( A–A' ) Western blots of P0 cortical lysates from Gsk3loxp:Neurod6 mutants and wild-type controls performed in triplicate . Levels of GSK-3 proteins and phospho-target proteins are shown . Strong reductions in phosphorylation of doublecortin on ser327/Thr321 and CRMP-2 on Thr514 are evident . ( A' ) Quantification of relative densities from A . p values shown in figure ( n = 3 , unpaired t-test ) ( B–B' ) Western blots of cortical lysates at P0 showing levels of other GSK-3 targets . No changes were observed in phosphorylation of dynamin , pCREB , or pFAK . No change was observed in cleaved caspase-3 . GAPDH was used as a loading control . ( B' ) Quantification of relative protein densities ( n = 3 , unpaired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02663 . 017 CRMP-2 family proteins have also recently been implicated in control of radial migration ( Ip et al . , 2014 ) . We found an 80% decrease in phosphorylated CRMP-2 on Thr514 in P0 cortical lysates from Gsk3loxp:Neurod6 mutants ( Figure 7A–A' ) . Changes in the phosphorylation status of DCX and CRMP-2 have been implicated in migration control , raising the possibility that reduced functions of these proteins toward microtubules are responsible for the effects of Gsk3 deletion . GSK-3 regulation appeared to be surprisingly specific . Thus , we did not detect changes in pSer722 FAK ( Bianchi et al . , 2005 ) , pSer744 dynamin-1 ( Clayton et al . , 2010 ) , and pSer129 CREB ( Fiol et al . , 1994; Bullock and Habener , 1998 ) ( Figure 7B–B' ) . Presumably other kinases contribute to phosphorylation of the putative GSK-3 sites in vivo . However , we cannot rule out changes that would be masked by the dilution effect of non-recombined cells , changes confined to specific cellular compartments , or changes that might be more apparent later in development . There was no increase in cleaved caspase-3 staining or other evidence of apoptosis in the mutants at P0 ( Figure 7B–B' ) . We did see increases in cleaved caspase-3 staining in cortex in Gsk3loxp:Neurod6 mice starting at later postnatal stages and these changes currently under investigation ( data not shown ) .
In this work , we have documented a cell autonomous requirement for GSK-3 signaling in migration and dendritic orientation of cortical excitatory neurons . GSK-3 activity is critical for the radial migration of later born , Cux1-expressing neurons in all regions of cortex , and for radial migration in the hippocampus . The GSK-3 requirement is specific for radial migration as tangential migration proceeded despite Gsk3 deletion . GSK-3 regulation of migration appears to be independent of Wnt/β-catenin and PI3K signaling . Rather , Gsk3 deletion is associated with striking reductions in phosphorylation of key microtubule regulatory proteins , DCX and CRMP-2 . Our work builds on a growing body of evidence establishing critical requirements for GSK-3 in cortical neuronal development , with GSK-3 signaling having different functions at different developmental stages . In radial progenitors , GSK-3 is a critical mediator of proliferation via regulation of β-catenin ( Kim et al . , 2009 ) . At later stages , recent work has demonstrated that GSK-3 is a key mediator of the amplification of the INP pool via interactions with the scaffolding protein Axin ( Fang et al . , 2013 ) . Axin-GSK-3 binding in the cytoplasm is clearly required for expansion of the INP pool , although the mechanism of GSK-3 action was not specified in this study . Interestingly , β-catenin regulation of transcription in the nucleus was required for differentiation of INPs into neurons . These effects of GSK-3 signaling on radial progenitors and INPs are presumably a strong determinant of final cortical neuronal number in mature animals . We now find that excitatory neuron development is also under important GSK-3 regulation . Thus , Cux1-expressing neurons require GSK-3 signaling for timely multipolar to bipolar transition and migration along radial processes . Deeper layer neurons expressing Tbr1 were not strongly influenced , but we cannot be certain that GSK-3 protein was fully depleted in these earlier born neurons at the time migration was occurring . The behavioral consequences of migration failure and dendritic arbor abnormalities could not be assessed due to early death of the Gsk3-mutant animals . Whether death was due to cortical abnormalities or defects in other cells that underwent recombination with Neurod6-Cre could not be determined . However , even a very mild form of this type of migration defect would presumably be catastrophic for human brain development . In contrast to regulation of progenitor proliferation and neural differentiation , GSK-3 regulation of radial migration appears to be independent of Wnt/β-catenin signaling . Thus , neither the stabilization of β-catenin nor deletion of all Dvls in developing cortical neurons using the Neurod6-Cre driver produced major defects in cortical layering . Interestingly , the migration defect also appears to be independent from a migration defect associated with mutations of the schizophrenia-associated protein , DISC1 . DISC1 regulates progenitor proliferation via interactions with GSK-3 , decreased GSK-3 kinase activity and , ultimately , increased β-catenin signaling ( Mao et al . , 2009 ) . Interestingly , a recent study demonstrated that DISC1 regulates migration independently of GSK-3 via effects on the centrosome ( Ishizuka et al . , 2011 ) . However , our work reported here clearly demonstrates that GSK-3 also has a critical role in the regulation of cortical neuronal migration . Radial migration is an enormously complex process requiring timely progenitor differentiation , dramatic morphological change , intricate mechanisms for cell and nuclear movements , and dynamic sensing of multiple cues that start and stop the process . Not surprisingly , GSK-3 joins dozens of other molecules implicated in the control of radial migration ( see Ayala et al . , 2007 for review ) . Comparisons with the literature suggest that effects of GSK-3 deletion are among the most severe that have yet been observed . Interestingly , GSK-3 importantly regulates multiple proteins implicated in migration and is an important mediator in several of the signaling pathways involved . Thus GSK-3 is known to phosphorylate DCX ( Bilimoria et al . , 2010 ) , FAK ( Bianchi et al . , 2005 ) , dynamin ( Clayton et al . , 2010 ) , neurogenin ( Li et al . , 2012 ) , CRMP-2 ( Uchida et al . , 2005; Yoshimura et al . , 2005 ) , and MAP1B ( Trivedi et al . , 2005 ) . Further , GSK-3 mediates Reelin signaling ( Beffert et al . , 2002 ) , LKB1 effects ( Asada and Sanada , 2010 ) , Cdc42 effects ( Etienne-Manneville and Hall , 2003 ) , integrin signaling ( Guo et al . , 2007 ) , semaphorin signaling ( Eickholt et al . , 2002; Uchida et al . , 2005 ) , and other pathways that have been implicated in control of radial migration . Finally , GSK-3 shares multiple substrates with cdk5 , a kinase that is situated among the most important regulators of neuronal migration and acts as a ‘priming’ kinase for GSK-3 signaling ( Xie et al . , 2006 ) . Clearly GSK-3 effects on neuronal development extend well beyond the phase of migration . For example , although dendritic arbors form , abnormalities in dendritic orientation were striking in Gsk3-deleted neurons . Importantly , interference with semaphorin signaling mediators also produces abnormalities of apical process development and orientation of basal dendrites ( see below ) . In general , GSK-3 acts via two classes of mechanisms: one where GSK-3 activity inhibits substrate function or availability and another where GSK-3 activity is required for substrate function . Therefore , we might expect Gsk3 deletion to enhance processes normally inhibited by GSK-3 activity and to inhibit processes that require GSK-3 activity . Multiple studies have suggested that inhibition of GSK-3 kinase activity is important for morphological functions such as establishment of neuronal polarity and cellular migration . The effects of GSK-3 inhibition are thought to be mediated by dephosphorylation of CRMP-2 , APC and other cytoskeletal mediators with resulting stabilization of microtubules at the tips of axons ( Etienne-Manneville and Hall , 2003; Jiang et al . , 2005; Yoshimura et al . , 2005; Hur and Zhou , 2010 ) . Another recent study employing in utero electroporation suggested that an STK11 ( LKB1 ) /GSK-3 pathway resulting in GSK-3β ser9 phosphorylation and APC localization at the leading edge was important in cortical neuronal migration ( Asada and Sanada , 2010 ) . In the studies outlined above , inhibition of GSK-3β kinase activity is indicated by phosphorylation on Ser9 and inhibition of GSK-3α by phosphorylation on Ser21 . However , a major unresolved paradox is that mice with Gsk3a and Gsk3b point mutation knock-ins that prevent ser9 and ser21 phosphorylation respectively do not exhibit widespread morphological abnormalities ( Gartner et al . , 2006 ) . Similarly , truncation mutants of the Drosophila GSK-3 homologue , Shaggy , that lack inhibitory phosphorylation sites are not associated with morphological abnormalities ( Papadopoulou et al . , 2004 ) . In the context of neuronal migration , although GSK-3 ser9/21 phosphorylation is enhanced downstream of Reelin signaling ( Beffert et al . , 2002; Gonzalez-Billault et al . , 2005 ) , GSK-3β kinase activity towards the important substrate MAP1B is actually increased rather than diminished ( Gonzalez-Billault et al . , 2005 ) . Finally , our findings reported here do not support critical roles in radial migration for STK11 , CDC42 , or PTEN , all of which are known to regulate GSK-3 ser9/21 phosphorylation . These findings , taken together , raise questions about the importance of negative regulation of GSK-3 via Ser9/21 phosphorylation on functions of microtubule binding proteins related to migration in vivo . Our results demonstrate that Gsk3 deletion blocks radial migration and that Gsk3 over-expression may enhance it . Thus our results are more in line with a mechanism in which GSK-3 activity is required for normal substrate function . Several proteins have been shown to require GSK-3 kinase activity for normal function . These include DCX where GSK-3 mediated phosphorylation is thought to be required for DCX's actions in regulating axon branching and possibly in migration ( Bilimoria et al . , 2010 ) . Additional examples possibly relevant to migration are GSK-3 regulation of FAK phosphorylation at ser722 ( Bianchi et al . , 2005 ) and dynamin-1 phosphorylation of ser774 ( Clayton et al . , 2010 ) . GSK-3 phosphorylation regulates functions of these proteins in complex ways , but possibly enhances functions that may be required for neuronal migration . A well-studied example of a pathway that is associated with upregulation of GSK-3 activity is semaphorin signaling . Semaphorin family members have multiple roles as chemorepellants and chemoattractants for many classes of axons and dendrites ( for review see Pasterkamp , 2012 ) . Semaphorin signaling strongly activates GSK-3 ( Eickholt et al . , 2002 ) resulting in phosphorylation of a key MAP , CRMP-2 ( Uchida et al . , 2005 ) , as well as other CRMP family members ( Cole et al . , 2006; Yamashita and Goshima , 2012 ) . Intriguingly , during neuronal polarization semaphorin signaling is associated with suppression of axons , and formation of dendrites ( Shelly et al . , 2011 ) , consistent with the idea that dephosphorylated CRMP-2 may favor axon formation , whereas phosphorylated CRMP-2 may be critical to proper formation of dendritic arbors . Importantly , migration and dendritic abnormalities that we have described here are consistent with cortical neuron phenotypes reported due to manipulation of semaphorin and CRMP signaling . Interfering with semaphorin , NP1 or Plexin via silencing RNAs and gene knockouts has been reported to interfere with radial migration ( Chen et al . , 2008; Renaud et al . , 2008 ) . Application of exogenous semaphorin supports apical process formation ( Polleux et al . , 2000 ) . Further , reduced semaphorin signaling causes bifurcation of CA1 pyramidal dendrites in the developing hippocampus ( Nakamura et al . , 2009 ) . Downstream of semaphorins , interfering with CRMP functions has been reported to affect migration and result in branched apical dendrites and abnormal dendrite orientation ( Yamashita et al . , 2012; Yamashita and Goshima , 2012; Ip et al . , 2012; for a review see , Ip et al . , 2014 ) . Finally , in the context of cortical neuronal migration , phosphorylation of CRMP-2 promotes migration . Thus a GSK-3 phosphomimetic CRMP-2 can rescue migration defects associated with knockdown of the semaphorin signaling mediator , α2-chimaerin ( Ip et al . , 2012 ) . Taken together , these findings suggest that inhibition of CRMP2 phosphorylation at Thr514 in Gsk3-deleted neurons may partially explain the migration and dendritic orientation abnormalities we have observed . In sum , we have demonstrated a key cell autonomous role for GSK-3 signaling in regulating radial migration and dendritic orientation of cortical excitatory neurons . Our Gsk3 deletion results are not in line with what might have been predicted from prior studies that have correlated ser9/21 phosphorylation with relief of negative GSK-3 phosphorylation and negative regulation of cytoskeletal-associated proteins . In particular , we do not find evidence that an STK11 ( LKB1 ) /GSK-3 inhibitory pathway is a key migration regulator . In contrast , our results are consistent with the idea GSK-3 phosphorylation of DCX and CRMP-2 is required for appropriate cytoskeletal regulation and migration . Finally , our data are consistent with the idea that GSK-3 activity is essential for the effects of semaphorin and CRMP-2 signaling on cortical neuronal development in vivo .
Mice were cared for according to animal protocols approved by the Institutional Animal Care and Use Committees of the University of North Carolina at Chapel Hill . GSK-3 lines were generously provided by Jim Woodgett . Gsk3a mice possessing exon 2 deletions and Gsk3a and Gsk3b loxp flanked exon 2 mice have been previously described ( Doble et al . , 2007; MacAulay et al . , 2007; Patel et al . , 2008 ) . Gsk3:Neurod6 mice were generated by mating Gsk3a−/− , Gsk3bloxp/loxp with Neurod6:Cre mice generously provided by Dr KA Nave ( Goebbels et al . , 2006 ) . Gsk3 mutant mice and Neurod6-Cre mice were maintained on mixed genetic backgrounds . For all experiments , four allele mutants were compared with littermate controls . Triple allelic male mutants ( Gsk3a+/-bf/f:Neurod6 ) fail to survive into adulthood with death occurring around P25 for reasons not yet determined . Ctnnb1 exon 3 ( Ctnnb1Ex3 ) floxed mice were previously described ( Harada et al . , 1999 ) and maintained on a mixed C57/Bl6 129 background . Stk11-floxed mice have been previously described ( Bardeesy et al . , 2002 ) and were maintained on a mixed C57/Bl6 129 , CF1 background . Pten-floxed mice ( Groszer et al . , 2001 ) , Cdc42-floxed mice were purchased from Jackson lab and have been previously described ( Chen et al . , 2006 ) . Dvl2 floxed allele was generated in the UNC Neuroscience Center Molecular Neuroscience Core using conventional methodology ( Ohata et al . , 2014 ) . Dvl1 ( Beier et al . , 1992 ) and Dvl3 ( Tsang et al . , 1996 ) mutant mice were purchased from The Jackson Laboratory ( Bar Harbor , Maine ) have been previously described . All Dvl mutant mice were maintained on a mixed background . Results from mutants and control littermates shown in all of the figure panels were based on at least three experiments from independent litters , unless otherwise noted . Mice were anesthetized using 2 , 2 , 2-Tribromoethanol ( 4 mg/10 g mouse ) or isofluorine , and embryos were exposed at E14 . 5 or E15 . 5 . Plasmids were mixed with fast green and microinjected into the lateral ventricle of embryos using a picospritzer . Embryos were electroporated with five 50 ms pulses at 30–35V with a 950 ms interval and returned to the abdominal cavity . Plasmids used were: Neurod1-Cre , lacZ/Egfp , lox-STOP-lox Ai9 , Neurod1-Gsk3b , and Neurod1-EGFP . A CF1 foster dam was used to aid postnatal survival studies . Depending on the experiment , mice were perfused at E19 . 5 , P0 , P10 , or P15 with 4% paraformaldehyde ( Sigma-Aldrich , St . Louis , MO ) in phosphate buffered saline ( PBS ) and post fixed overnight . Cortical progenitor cells were electroporated ex vivo at E15 . 5 as described previously ( Hand et al . , 2005 ) . Briefly , E15 . 5 embryos were decapitated , plasmids were injected into lateral ventricle followed by electroporation with four 30V pulses that were 30–40 ms in duration and separated by a 100-ms interval . Following electroporation , brains were dissected and vibratome sectioned at 250 μm . Slices were transferred to Poly-D-Lysine and laminin coated culture insert ( Millipore , Billerica , MA ) in a FluoroDish ( World Precision Instruments , Sarasota , FL ) , then 2 ml Basal Medium Eagle with FBS , N2 ( Gemini Bio-Poducts , West Sacramento , CA ) , B27 ( Life Technologies , Grand Island , NY ) , penicillin-streptomycin ( Life Technologies ) , and L-glutamine ( Life Technologies ) supplements were added . Slices were cultured for 3 days at 37°C and live imaged using an Olympus FV1000 Confocal microscope with stage incubator . E14 . 5–15 . 5 dorsal cortices were electroporated with Neurod1-Cre; Z/EG , dissected in 4°C Hank's Balanced Salt Solution ( HBSS ) ( Life Technologies ) and dissociated into single cells using Trypsin ( Life Technologies ) according to previously described methods ( Hand et al . , 2005 ) . Neurons were plated on glass bottom dishes ( MatTek ) coated with 0 . 1 mg/ml Poly-D-Lysine ( Sigma ) and 5 μg/ml Laminin ( Sigma ) . Cells were cultured in Neuralbasal-A Medium ( Life Technologies ) , supplemented with 1X B-27 ( Gibco , 17 , 054-044 ) , L-glutamine ( Life Technologies ) , penicillin-streptomycin ( Life Technologies ) , N2 ( Life Technologies ) , and FBS . Neurons were fixed with 4% PFA and stained for stage progression analysis . Mouse cortices were dissected from three control ( Gsk3a−/−Gsk3bloxp/+:Neurod6 ) and mutant ( Gsk3:NeuroD6 ) or three control ( Gsk3aloxp/loxpGsk3bloxp/loxp ) and three mutant ( Gsk3aloxp/loxpGsk3bloxp/loxp:NeuroD6 ) mice from independent litters , collected in RIPA lysis buffer supplemented with protease and phosphatase inhibitors and cleared by centrifugation . Proteins were separated on SDS-PAGE gradient gels , transferred to a PVDF membrane and probed for proteins of interests and secondary HRP-conjugated antibodies were used for detection . Blots were washed and detection was performed with a commercially available ECL kit . ImageJ software ( NIH ) was used for quantification of band intensities , and intensities were normalized with total protein or the load control . Statistical analyses were conductive using Prism ( GraphPad Software Inc . , La Jolla , CA ) . Differences were considered statistically significant at p<0 . 05 and are included in each figure . The following antibodies were used: GSK-3 ( Invitrogen , Carlsbad , CA ) , Actin ( Cell Signaling Technology Inc . , Danvers , MA ) , pFAK ser722 ( Santa Cruz Biotechnology , Inc . , Dallas , TX ) , FAK ( Cell Signaling Technology Inc ) , pCRMP-2 Thr514 ( Cell Signaling Technology Inc ) , CRMP-2 ( Cell Signaling Technology Inc ) , pCREB ser129 ( Santa Cruz Biotechnology , Inc . ) , CREB ( Santa Cruz Biotechnology , Inc . ) , pDCX ser327 , DCX ( Cell Signaling Technology Inc . ) , pDynamin-1 ser744 ( Santa Cruz Biotechnology , Inc . ) , Dynamin-1 ( Santa Cruz Biotechnology , Inc . ) , PTEN ( Cell Signaling Technology Inc . ) , LKB1 ( Upstate ) , β-catenin ( Cell Signaling Technology Inc ) , DVL2 ( Cell Signaling Technology Inc ) , GAPDH ( Cell Signaling Technology Inc ) , Cleaved Caspase-3 ( Cell Signaling Technology Inc ) . Briefly , 100–150 μm free-floating vibratome sections of brains were collected in PBS , blocked with 5% normal serum in PBS with 0 . 1% Triton X-100 and incubated with primary antibodies in blocking solution overnight at 4°C . Cryosectioned tissue samples were immunostained as previously described ( see Newbern et al . , 2011 ) . P15 in utero vibratome sections were blocked with 5% normal serum in PBS with 0/1% Triton X-100 and 2% DMSO , and incubated with primary antibodies for 3 days . Slices were rinsed with PBST for 24 hr and incubated with secondary antibodies for 3 days in PBS with 0/1% Triton X-100 . The following primary antibodies were used in our study: L1 ( Millipore ) , Cux1 ( Santa Cruz Biotechnology , Inc . ) , Neuronal Nuclei ( Millipore ) , GFP ( Aves Labs , Inc . , Tigard , OR ) , CTIP2 ( Abcam , Cambridge , MA ) DRAQ5 ( Fisher ) , TBR1 ( Abcam , Cambridge , MA ) , HOECHST ( Sigma ) , RFP ( Rockland , Gilbertscille , PA ) . After rinsing , sections were then incubated with Alexa-conjugated secondary antibodies ( Invitrogen ) overnight at 4°C , rinsed three times in PBS , and mounted with gel/mount ( Sigma ) . Gsk3 overexpression analysis was conducted with Neurod1-Cre and lox-STOP-lox-Ai9 plasmids as control and Neurod1-Gsk3b and Neurod1-eGFP plasmids for overexpression . Electroporations were performed at E14 . 5 and analyzed at E19 . 5 . For controls and Gsk3 over-expression , multiple comparable cortical sections were analyzed . For analysis of layering , the cortical plate was equally divided into eight bins , and cells in each bin were counted using ImageJ software . Bin 7–8 included layers 1–3 . The percentage of GFP+ neurons in each bin were determined from multiple comparable sections . Postnatal day 10 analysis after in utero electroporation was conducted with Gsk3a+/−Gsk3bloxp/loxp control Gsk3−/−Gsk3bloxp/loxp embryos . EGFP-filled neurons were counted and localized to either upper or deeper layers of cortex . Differences were considered statistically significant at p≤0 . 05 using an unpaired t-test . Lamination quantification of mutant mice at P0 ( Gsk3loxp:Neurod6 , Gsk3:Dlx5/6 , Ctnnb1Ex3:Neurod6 , Dvl123:Neurod6 , Stk11:Neurod6 , Cdc42:Neurod6 , Pten:Neurod6 ) was conducted using heterozygous controls ( Cre+ , loxP/+ ) compared to conditional knockouts . Vibratome sections ( 3–5 per mouse ) were imaged at 40X with a 2 . 4-μm optical slice and Cux-1 cells were counted in an 8 bin analysis . Data were quantified as average percentage of Cux-1 neurons per bin with differences considered statistically significant at p≤0 . 05 using an unpaired t-test . Note that the histograms for controls Figure 2 and Figure 6—figure supplement 1 ( based on a total of 10 control animals ) all look very similar reflecting the fact that almost all Cux1 neurons have reached Layer 2–3 by P0 in control mice . The histogram for the Gsk3 mutants in Figure 2 is drastically different and reflects the distribution of Cux 1 neurons observed by inspection in many additional Gsk3:Neurod6 and Gsk3loxp:Neurod6 mice . In contrast , the histograms showing Cux1 neuron distribution for the other mutants were indistinguishable from controls . Determination of number of Cux1 neurons at P0 in Gsk3loxp:Neurod6 mice with Gsk3aloxp/loxpGsk3bloxp/WT:Neurod6 heterozygous as controls was performed as follows: Three 100-μm thick vibratome sections were imaged and one to two 100-μm cortical columns were obtained per slice for a total of five columns per mouse . Total Cux1 cells and total cell numbers were counted using an 8 bin analysis with ImageJ software . Data were quantified as average percentage of Cux1/total per bin and differences were analyzed using an unpaired t-test . Images were collected on Zeiss LSM 710 and Zeiss LSM 780 confocal microscopes . Z-stack images were collected with 10X or 20X objectives and tiled together to generate high-resolution images of whole brain sections . Binned quantification images were taken using a 20× objective and a 2-μm optical slice . Randomly identified sections of the electroporated area in presumptive somatosensory cortex were imaged for quantification . Live imaging of migrating neurons was acquired on an Olympus FV1000 Confocal microscope with stage incubator . On average , 40 μm Z-stack images were acquired and merged every 30–60 min for 8–20 hr . A 40X objective at a 0 . 8-μm optical slice was used for GSK-3loxp:Nex cux-1 cell counting . P15 in utero experiments used Gsk3a+/+Gsk3bloxp/loxp control Gsk3a−/− Gsk3bloxp/loxp or Gsk3aloxp/loxp Gsk3bloxp/loxp embryos . Images for the dendritic arbor reconstruction were acquired on a Zeiss LSM 7 MP multiphoton system using a W-PlanApochromat 20x/1 . 0 IR-corrected water immersion objective . On average , Z-stacks were acquired from 90 to 100 μm range with 0 . 65 μm steps . Dendritic reconstructions were performed using Neurolucida . A total of 15 control and 15 Gsk3-deleted cells from three mice each were reconstructed for dendritic analysis . Apical and basal dendrite polarity was quantified by generating dendrograms ( Neurolucida ) from reconstructed images . Sholl analysis calculated the number of dendrite intersections and dendritic lengths , Sholl quantification was conducted with an initial 15-μm somal radius and 20 μm concentric radial steps . Dorsal cortices were dissected from three E18 control ( Gsk3a−/−Gsk3bloxp/loxp ) and mutant ( Gsk3;Neurod6 ) embryos derived from two independent litters . RNA was prepared using the MiRNAeasy kit ( Qiagen , Valencia , CA ) and analyzed using the Affymetrix Mouse Gene 2 . 0 St array ( Affymetrix , Santa Clara , CA ) . Following scanning of the array , basic data analysis was carried out using the Partek Genomics Suite Version 6 . 12 . 0712 ( Partek , Inc . , St . Louis , MO ) Transcripts up or down-regulated by 1 . 5-fold were considered interesting candidates . Microarray data have been made available through Gene Expression Omnibus , accession number GSE58727 ( GEO , NCBI ) . | In the brain , one of the most striking features of the cerebral cortex is that its neurons are organized into different layers that are specifically connected to one another and to other regions of the brain . How newly generated neurons find their appropriate layer during the development of the brain is an important question; and , in humans , when this process goes awry , it can often result in seizures and mental retardation . An enzyme called GSK-3 regulates several major signaling pathways important to brain development . The GSK-3 enzyme switches other proteins on or off by adding phosphate groups to them . Morgan-Smith et al . set out to better understand the role of GSK-3 in brain development by deleting the genes for this enzyme specifically in the cerebral cortex of mice . Mice have two genes that encode slightly different forms of the GSK-3 enzyme . Deleting both of these in different groups of neurons during brain development revealed that a major group of neurons need GSK-3 in order to migrate to the correct layer . Specifically , the movement of neurons from where they arise in the central region of the brain to the outermost layer ( a process called radial migration ) was disrupted when the GSK-3 genes were deleted . Morgan-Smith et al . further found that cortical neurons without GSK-3 were unable to develop the shape needed to undertake radial migration because they failed to switch from having many branches to having just two main branches . Additional experiments revealed that these abnormalities did not depend on certain signaling pathways , such as the Wnt-signaling pathway or the PI3K signaling pathway that can control GSK-3 activity . Instead , Morgan-Smith et al . found that two proteins that are normally targeted by the GSK-3 enzyme have fewer phosphate groups than normal in the cortical neurons that did not contain the enzyme: both of these proteins regulate the shape of neurons by interacting with the molecular ‘scaffolding’ within the cell . The GSK-3 enzyme was already known to modify the activities of many other proteins that affect the migration of cells . Thus , the findings of Morgan-Smith et al . suggest that this enzyme may coordinate many of the mechanisms thought to underlie this process during brain development . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2014 | GSK-3 signaling in developing cortical neurons is essential for radial migration and dendritic orientation |
After cardiac ischaemia , a prolonged decrease of coronary microvascular perfusion often occurs even after flow is restored in an upstream artery . This 'no-reflow' phenomenon worsens patient prognosis . In the brain , after stroke , a similar post-ischaemic 'no-reflow' has been attributed to capillary constriction by contractile pericytes . We now show that occlusion of a rat coronary artery , followed by reperfusion , blocks 40% of cardiac capillaries and halves perfused blood volume within the affected region . Capillary blockages colocalised strongly with pericytes , where capillary diameter was reduced by 37% . The pericyte relaxant adenosine increased capillary diameter by 21% at pericyte somata , decreased capillary block by 25% and increased perfusion volume by 57% . Thus , cardiac pericytes constrict coronary capillaries and reduce microvascular blood flow after ischaemia , despite re-opening of the culprit artery . Cardiac pericytes are therefore a novel therapeutic target in ischaemic heart disease .
Coronary heart disease is the leading cause of mortality worldwide , causing 7 . 4 million deaths in 2015 . To treat cardiac ischaemia , primary percutaneous coronary intervention is used to re-open the culprit coronary artery , but this does not guarantee reperfusion of the downstream capillaries supplying the myocardium ( Krug et al . , 1966; Kloner et al . , 1974 ) . A lack of capillary reperfusion - ‘no-reflow’ - affects up to 50% of patients ( Niccoli et al . , 2009 ) , and predicts a raised prevalence of deleterious complications including congestive heart failure , malignant arrhythmias and cardiac death ( Niccoli et al . , 2009; Wu et al . , 1998; Morishima et al . , 2000; Ørn et al . , 2009; Kloner , 2011 ) . No-reflow has been attributed to swollen endothelial cells and/or cardiomyocytes compressing the lumen of capillaries ( Kloner , 2011 ) , or to leukocytes adhering to and plugging capillaries ( Engler et al . , 1983 ) . However , although depleting blood of granulocytes reduces no-reflow ( Engler et al . , 1986 ) , leukocyte adhesion to capillaries in reperfused hearts is modest compared to the adhesion seen in post-capillary venules ( Habazettl et al . , 1999 ) . No treatment exists for coronary no-reflow after decades of investigation , even though it predicts a worse outcome after myocardial ischaemia ( Niccoli et al . , 2009; Wu et al . , 1998; Morishima et al . , 2000; Ørn et al . , 2009; Kloner , 2011 ) . We therefore sought an alternative explanation for coronary capillary narrowing , which might trap leukocytes or red blood cells physically , based on recent advances in our understanding of the similar no-reflow phenomenon which occurs after brain ischaemia ( O'Farrell and Attwell , 2014 ) . In the brain and retina , blood flow is partly regulated by contractile pericytes located on capillaries ( Hirschi and D'Amore , 1996 ) , which respond to vasoactive messengers released from active neurons and astrocytes by altering their tone , thus altering capillary diameter and blood flow ( Peppiatt et al . , 2006; Puro , 2007; Hall et al . , 2014; Biesecker et al . , 2016; Duan et al . , 2016; Kisler et al . , 2017 ) . Furthermore , ischaemia , spinal cord injury and epilepsy lead to a contraction of pericytes , causing constriction of capillaries ( Hauck et al . , 2004; Peppiatt et al . , 2006; Yemisci et al . , 2009; Hall et al . , 2014; Li et al . , 2017; Leal-Camanario et al . , 2017 ) , and for ischaemia this is followed by the pericytes dying in rigor ( Hall et al . , 2014 ) which hinders subsequent capillary dilation . Pericytes also exist in the heart ( Tilton et al . , 1979 ) , where they are the second most numerous cell type after endothelial cells ( Nees et al . , 2012 ) . They are poorly understood ( only 1 in 5000 cardiac papers in PubMed mentions pericytes ) : it is debated whether they are contractile ( Tilton et al . , 1979; Joyce et al . , 1985 ) and attention has focused on other possible functions for them including angiogenesis , immune defence , haemostasis and cardiac regeneration ( Nees et al . , 2013; Chen et al . , 2015 ) . Here we show that ischaemia-induced constriction of coronary capillaries by pericytes contributes to no-reflow after cardiac ischaemia .
The left ventricular coronary capillary bed comprises an array of parallel capillaries linked by occasional connector capillaries running roughly orthogonal to the main direction of flow ( Figure 1A ) . Labelling capillaries with FITC- or Alexa647-isolectin B4 , and pericytes in rat with antibody to NG2 ( Figure 1A ) or PDGFRβ ( Figure 1C ) , or in mouse with transgenic expression of DsRed under the NG2 promoter ( Figure 1C ) , showed that the great majority of coronary capillaries were contacted by pericytes . Pericytes differ from vascular smooth muscle cells in that their somata are spatially separated along the capillary , located either with a bump-on-a-log appearance on the straight parts of capillaries , or at the branch points of capillaries ( Attwell et al . , 2016 ) . More of the parallel capillaries than of the connector capillaries linking adjacent parallel capillaries received pericyte contacts ( 92% vs 66% , p=2 . 2×10−5 , Figure 1B ) . The great majority of pericytes labelled for both NG2 and PRGFRβ ( Figure 1D ) , although a small fraction ( 6 . 4 ± 0 . 2% ) of PDGFRβ-expressing pericytes did not express NG2 ( p=2 . 3×10−7 ) . Overall there is one pericyte every ~60 μm along rat coronary capillaries , independent of location in the left ventricle ( Figure 1F ) although there is variability around this mean distance ( Figure 1A , G ) . Pericytes extend circumferential processes ( Figure 1E , G ) , the contraction of which would directly alter capillary diameter . Contraction of their longitudinal processes could alter the stiffness of the capillary wall and regulate its deformability by passing blood cells . Either of these mechanisms could regulate capillary blood flow . Indeed , varying degrees of contraction of these pericytes is a plausible explanation for the large variance of capillary blood transit time in the coronary circulation ( Rose and Goresky , 1976 ) , which has a significant effect on oxygen extraction by the myocardium ( Ostergaard et al . , 2014 ) . We found that approximately 50% of pericytes are located close to varicosities of tyrosine hydroxylase expressing sympathetic axons ( Figure 1H–J ) , suggesting the possibility of noradrenergic regulation of pericyte tone . Pericytes are conventionally assumed to constrict capillaries using α-smooth muscle actin ( α-SMA: Joyce et al . , 1985; Skalli et al . , 1989 ) , but variability in the labelling observed for α-SMA and data showing expression of other actin isoforms in pericytes has led to a suggestion , for CNS pericytes , that γ-actin might instead be the relevant actin isoform ( DeNofrio et al . , 1989; Grant et al . , 2017 ) . We therefore examined antibody labelling for α-SMA , β-actin and γ-actin ( Figure 1K–M ) . α-SMA labelling occurred in 36 . 4 ± 6 . 7% of 57 pericytes ( Figure 1N ) , and was rarer than labelling for β-actin ( 62 . 3 ± 2 . 8% of 30 pericytes ) or γ-actin ( 76 . 0 ± 3 . 4% of 47 pericytes ) . α-SMA-expressing pericytes were most commonly observed in capillaries closer to the arteriole end of the capillary bed , while β- and γ-actin were seen in pericytes across the capillary bed . To examine the possible role of contraction of pericytes in ischaemic pathology , we occluded the left anterior descending ( LAD ) coronary artery for 45 mins , depriving the anterior wall of the left ventricle and part of the right ventricle of blood ( see Materials and methods ) . We then removed the occlusion to allow reperfusion for 15 mins , so that any rapidly reversible obstruction of blood vessels produced by ischaemia would be removed , leaving only long-lasting vessel obstruction contributing to the no-reflow phenomenon . At the end of this period FITC-albumin in gelatin was perfused to visualise vessels where flow was present , and the tissue was sectioned and labelled with isolectin B4 conjugated to Alexa Fluor 647 to visualise non-perfused vessels . In control ( sham LAD artery occlusion ) hearts , perfusion was visible throughout the whole cross section of the left and right ventricles ( Figure 2A ) . The perfused blood volume per unit area ( assessed from the mean FITC-albumin intensity in regions of interest ) was fairly uniform around the left ventricle ( Figure 2C ) , although somewhat higher in the posterior wall ( regions of interest ( ROIs ) 4–6 of Figure 2C ) . In contrast , occlusion and reperfusion of the LAD artery resulted in microvascular perfusion being greatly reduced after the period of ischaemia in half of the left and right ventricles ( Figure 2B ) , as quantified for the left ventricle in Figure 2C ( reduced by 49% compared to sham-operated animals in ROIs 7–10: p=0 . 0004 ) . Higher magnification images revealed that in control hearts only 3% of capillaries were blocked ( i . e . not perfused by FITC-albumin ) in the left ventricle . In contrast , LAD artery occlusion and reperfusion increased this to 40% in the affected area ( Figure 2D ) . Some capillaries were completely perfused and some were completely unperfused throughout the imaged area , while some capillaries showed an abrupt cessation of perfusion ( Figure 3A–F ) , with a profound decrease of FITC-albumin intensity that occurred over a few microns ( Figure 3G ) . Since leukocytes are both larger and less deformable than erythrocytes ( Schmid-Schönbein et al . , 1981; Downey et al . , 1990; Komatsu et al . , 1990; Doerschuk et al . , 1993 ) , it seemed more likely that leukocytes rather than red blood cells would get stuck at capillary regions of reduced diameter . Surprisingly , labelling with antibody to neutrophil elastase or ICAM-1 revealed no leukocytes lodged at 46 blockage sites examined ( although , as a positive control , they were seen outside vessels , Figure 3K , usually post-capillary venules ) . Similarly , labelling for the erythrocyte protein glycophorin A revealed red blood cells ( Figure 3L ) associated with only a small percentage of blockage sites ( 18% of 44 blockages ) , and even where red blood cells were trapped at capillary constrictions it did not always lead to a block of blood flow ( as shown by FITC-albumin passing the red blood cells ) . However , examining NG2 labelling at the sites of block revealed that many blockages occurred close to pericytes , in some cases with pericyte processes appearing to visibly constrict the capillary at the block location ( Figure 3A–C ) . To assess rigorously whether block occurred disproportionately close to pericytes , we measured the distance of 42 blockages to the nearest pericyte soma . The cumulative probability distribution of this distance showed the median distance to be 3 . 6 μm ( Figure 3H ) . In contrast , when we used the same 42 images to calculate what the average distance of a blockage to the nearest pericyte soma would be if blockages were randomly placed on the blocked capillaries in the images ( see Figure 3—figure supplement 1 ) , the cumulative probability distribution for that distance predicted a median distance of 11 . 6 μm ( Figure 3H; significantly different , p=3 . 9×10−5 ) . Thus , ischaemia-evoked capillary block occurs disproportionately close to pericytes . To verify that pericytes constrict coronary capillaries in ischaemia , we measured the diameter of the FITC-albumin labelled lumen at pericyte somata , and at a distance 10 μm upstream of the soma ( since most circumferential processes of pericytes are located less than 10 μm from the soma ) . In sham-operated animals , the ratio of the diameter at the soma to that upstream was 1 . 058 ± 0 . 015 ( n = 20 , significantly greater than one , p=0 . 001 , Figure 3I ) , consistent with the presence of the pericyte soma inducing growth of the capillary lumen , as found previously for brain pericytes ( Hall et al . , 2014 ) . In contrast , for capillaries after ischaemia , this ratio was 0 . 822 ± 0 . 022 ( n = 60 , significantly less , p=1 . 4×10−8 , Figure 3I ) , implying that constriction occurs near the pericyte soma . The absolute diameter measured at pericyte somata was reduced by 37% after ischaemia and reperfusion ( compared to sham-operated hearts , p=3 . 1×10−6 , Figure 3J ) . Pericytes near blockage sites were tested for labelling of the different actin isoforms mentioned above . All 4 such pericytes tested for α-SMA labelling exhibited labelling ( Figure 3M ) , as did 5 out of 6 tested for β-actin , and all 6 tested for γ-actin . Since only a small fraction of pericytes may need to constrict a capillary to abolish its blood flow , further work is needed to determine which is the main actin isoform responsible for the ischaemia-evoked contraction of pericyte processes and reduction of blood flow . Since cardiac catecholamine transporters reverse and release noradrenaline and adrenaline in ischaemia ( Lameris et al . , 2000 ) , and noradrenaline acts on pericytes in the brain to constrict capillaries ( Peppiatt et al . , 2006 ) , we assessed whether blocking adrenergic α1 receptors reduced the percentage of capillaries blocked after ischaemia . We found no effect of injecting terazosin ( 0 . 5 mg/kg i . v . , 5 mins before occluding the LAD artery , which produced a 20–33 mm Hg decrease of blood pressure ) , with 48 . 8 ± 3 . 7% of capillaries blocked in the presence of terazosin ( in 13 images covering 383 capillaries in 2 hearts; not significantly different from the 44 . 7 ± 4 . 9% seen for 14 images covering 366 capillaries in 2 hearts in interleaved experiments without terazosin , p=0 . 52 ) . We therefore sought an alternative pharmacological approach to reducing no-reflow after ischaemia . Adenosine has been suggested to reduce no-reflow after cardiac ischaemia , although its clinical utility remains uncertain ( Berg and Buhari , 2012; Su et al . , 2015 ) . Adenosine is also thought to relax pericytes ( Matsugi et al . , 1997; Li and Puro , 2001; Gaudin et al . , 2014 ) . We infused adenosine intravenously ( at 0 . 5 mg/kg/min , similar to doses used previously to treat no-reflow in humans and other animals: see Materials and methods ) , from 5 mins before the end of ischaemia until 10 mins of reperfusion had occurred ( see Materials and methods ) , and examined its effect on capillary block . Adenosine significantly reduced ( p=0 . 03 ) the decrease of coronary perfusion seen after ischaemia ( Figure 2C , the mean value in ROIs 7–10 in Figure 2C was increased by 57% compared to ischaemia without adenosine , and was not significantly different from that in control conditions , p=0 . 77 ) . This increase of flow could reflect adenosine acting both on arteriolar smooth muscle and on pericytes . However , adenosine also reduced by one quarter the percentage of capillaries that were blocked after reperfusion , from ~40% to ~30% ( p=0 . 007 , Figure 2D ) . An analysis like that in Figure 3H showed that the remaining sites of block were still significantly associated with pericyte locations ( p=0 . 001 ) . Since the capillary blockages induced by ischaemia in the absence of adenosine are disproportionately associated with pericytes , and since the adenosine was only applied around the period of reperfusion , these data suggest that , at least in part , adenosine reduces no-reflow by reversing the constriction of pericytes that ischaemia induces . To test this hypothesis , we compared the diameter of capillaries at pericyte somata with the diameter 10 μm upstream of the soma and found that adenosine significantly ( p=0 . 0045 ) reduced the constriction evoked by ischaemia at pericyte somata ( Figure 3I ) , implying a specific effect on pericytes ( rather than a general capillary dilation produced by upstream arteriole dilation ) . The absolute capillary diameter at pericyte somata after ischaemia was increased by 21% using adenosine ( p=0 . 025 , Figure 3J ) . Thus , adenosine decreases no-reflow by relaxing pericytes .
Our data suggest a novel therapeutic target for reducing no-reflow after cardiac ischaemia: during ischaemia pericytes constrict and block coronary capillaries , probably because pericyte [Ca2+]i rises when ion pumping stops in ischaemia . Three lines of evidence support this idea . First , sites of capillary blockage are disproportionately associated with pericyte locations ( Figure 3H ) . Second , after ischaemia the diameter of capillaries is reduced specifically at pericyte somata ( Figure 3I , J ) . Third , adenosine , which is thought to relax pericytes , increases the post-ischaemic diameter of capillaries at pericyte locations ( Figure 3I , J ) and reduces the percentage of capillaries that remain blocked when the upstream artery is reperfused after ischaemia ( Figure 2D ) . Prevention or reduction of pericyte constriction should therefore reduce the flow impairment that causes significant functional impairment after removal of a thrombus in a coronary artery . The most plausible explanation of our data is that , like pericytes in the CNS ( Peppiatt et al . , 2006; Hall et al . , 2014 ) , cardiac pericytes extend actomyosin-containing processes around coronary capillaries that , when their [Ca2+]i rises and contraction is activated , reduce capillary diameter ( see Supplementary Movie 1 of Peppiatt et al . , 2006 ) . Indeed , cardiac pericytes express both actin and myosin ( Joyce et al . , 1985; Skalli et al . , 1989 ) . It is likely that pericytes do not need to completely constrict a capillary in order to stop blood flow in that vessel , because a reduction in diameter ( like the 37% in Figure 3J ) may prevent the passage of leukocytes ( Engler et al . , 1983 ) or red blood cells ( Figure 3L ) , causing a blockage that prevents the passage of both plasma and cells . The fact that we observed few cells in the capillary lumen at blockages may result from their displacement during the perfusion with FITC-albumin in gelatin , or may alternatively reflect capillary block occurring when constriction brings together the layers of glycocalyx ( Secomb et al . , 1998 ) on opposite sides of the capillary . It is uncertain whether , over a long period , cardiac pericyte death in rigor occurs and contributes to generating a long lasting reduction of blood flow , as suggested for the CNS ( Hall et al . , 2014 ) . Clearly not all pericytes die during the 45 min period of ischaemia used here ( which in severe brain ischaemia kills ~50% of pericytes: Hall et al . , 2014 ) because , if they did , then adenosine applied around the time of reperfusion would not be able to reduce the amount of capillary constriction and block occurring . However , adenosine only reduces the number of blocked capillaries by 25% ( Figure 3J ) and conceivably it would produce a greater restoration of flow if pericyte death were prevented . It should be recognised that , as with all studies evoking ischaemia by transiently occluding a coronary artery in rodents , our work is only a partial simulation of human cardiac ischaemia . Typically , human myocardial infarct involves the slow build up of a plaque on an artery , which becomes occluded by a thrombus after plaque rupture occurs . This is followed by mechanical removal of the clot ( which may compound microvascular obstruction by embolising plaque and thrombus material ) , leaving an activated oedematous endothelium across which neutrophils invade the tissue ( Niccoli et al . , 2009; 2016 ) . Nevertheless , at short times after ischaemia , we expect the constriction that we observe of downstream capillary pericytes to occur also in human cardiac tissue , prolonging the period of ischaemia . Previously , a combination of myocyte swelling , endothelial cell blebbing and leukocyte plugging have been proposed to block capillaries in ischaemia ( O'Farrell and Attwell , 2014 ) . Our demonstration that ischaemia selectively reduces capillary diameter near pericytes suggests that , at least at early times , the majority of capillary blockages in fact occur near pericytes that have constricted the capillaries . These data imply that pericytes are a novel drug target for treating the coronary no-reflow that often occurs after primary percutaneous coronary intervention . It would be interesting to test whether deletion of pericyte-capillary interactions mediated by the angiopoietin/TIE and PDGF/PDGFRβ pathways ( Ziegler et al . , 2013 ) affect the contractile properties of pericytes and their response to ischaemia . We used adenosine to relax pericytes ( an effect verified from measurements of capillary diameter at pericyte locations in Figure 3I–J ) , but it may also act on arteriolar smooth muscle and lower blood pressure . To reduce no-reflow therapeutically by preventing pericyte constriction of coronary capillaries , it will be necessary to develop agents that act selectively on pericytes . Encouragingly , in the brain it has been shown that signalling pathways regulating pericyte constriction can differ from those regulating arteriole smooth muscle ( Mishra et al . , 2016 ) . Our data also raise the possibility that other organs susceptible to ischaemic no-reflow pathology , such as the kidneys , muscle and skin , may also be damaged by a pericyte-mediated mechanism . Finally , given the proximity of cardiac pericytes to sympathetic terminals ( Figure 1H–J ) , noradrenaline release may also regulate coronary blood flow at the capillary level in physiological conditions .
Adult Sprague-Dawley rats or , to illustrate morphology , NG2-DsRed mice ( Zhu et al . , 2008 ) expressing the fluorescent protein DsRed in pericytes , of either sex , were sacrificed by UK government approved methods ( anaesthetic overdose with 5% isoflurane in 100% O2 , followed by cervical dislocation ) . Hearts were harvested and immersion-fixed in ice cold 4% paraformaldehyde ( PFA ) . Immunohistochemistry was performed on 150 μm thick transverse ventricular slices . Pericytes were labelled by expression of DsRed under control of the NG2 promoter ( in mice ) , or with anti-NG2 ( Merck Millipore AB5320 1:200 or AbCam ab50009 1:200 ) or anti-PDGFRβ ( Santa Cruz Biotechnology sc-432 1:200 ) antibodies ( in rats ) , and the capillary basement membrane was labelled with isolectin B4 congugated to Alexa Fluor 647 ( Molecular Probes , I32450 ) or FITC ( Sigma-Aldrich , UK , L2895 ) as described ( Mishra et al . , 2014 ) . Pericyte association with sympathetic terminals was imaged using antibody to tyrosine hydroxylase ( Merck Millipore AB1542 1:500 ) . Z-stacks for cell counting were acquired using laser-scanning confocal microscopy ( Zeiss LSM 700 ) . Pericyte intersoma distance was calculated between pairs of pericytes on capillaries within the same imaging plane . Antibodies to α-SMA , β-actin and γ1 actin were AbCam ab5694 ( 1:100 ) , Abbiotec 251815 ( 1:100 ) and AbCam ab123034 ( 1:100 ) . Red blood cells were labelled with antibody to glycophorin A ( AbCam ab9520 , 1:2000 ) . Neutrophils were labelled with antibodies to neutrophil elastase ( Santa Cruz , sc9521 ( M-18 ) , 1:50 ) or ICAM1 ( Abcam , Ab171123 , 1:100 ) . Adult male Sprague-Dawley rats ( 200–220 g ) were anaesthetized with pentobarbital sodium ( induction 60 mg/kg i . p . ; maintenance 10–15 mg/kg/h i . v . ) . The right carotid artery and jugular vein were cannulated for measurement of arterial blood pressure and administration of anaesthetic and tested substances , respectively . Stable levels of blood pressure and heart rate were maintained , and adequate anaesthesia was monitored by the absence of a withdrawal response to a paw pinch . The trachea was cannulated , and the animal was mechanically ventilated with room air using a positive pressure ventilator ( tidal volume of 1 ml/100 g of body weight , ventilator frequency ∼60 strokes min−1 ) . PO2 , PCO2 and pH of the arterial blood were measured regularly and , if required , ventilation was adjusted to maintain values within their physiological ranges . Arterial BP and a standard lead II ECG were recorded using a Power1401 interface and Spike2 software ( Cambridge Electronic Design ) . Body temperature was maintained at 37 . 0 ± 0 . 5°C with a servo-controlled heating pad . The heart was exposed via a left thoracotomy and a 4–0 prolene suture was passed around the left anterior descending ( LAD ) coronary artery to induce a temporary occlusion as previously described ( Mastitskaya et al . , 2012; Basalay et al . , 2016 ) . The animals were subjected to 45 min of LAD artery ligation , followed by 15 min of reperfusion . Successful LAD occlusion was confirmed by paling of the myocardial tissue distal to the suture , elevation of the ST-segment in the ECG , and an immediate 15–30 mm Hg fall in the ABP . In rats LAD occlusion deprives of blood not only the anterior wall of the left ventricle but also part of the right ventricle ( Samsamshariat and Movahed , 2005 ) , as seen in Figure 2 . Control ( sham operated ) animals underwent the same procedures , except that after the suture was passed around the LAD coronary artery it was not drawn tight to occlude the vessel . Ischaemia and sham animals were alternately interleaved . In some experiments i . v . adenosine ( Sigma-Aldrich , St . Louis , Missouri ) was administered ( with the aim of relaxing pericytes ) , starting 5 min before the end of ischaemia and continuing until 10 min of reperfusion ( 0 . 5 mg/kg/min in saline , 2 mg/ml , cumulative dose 7 . 5 mg/kg ) , which lowered the blood pressure reversibly by 10–15 mm Hg ( 10% ) . Control animals received vehicle ( saline ) infusion . For comparison , in cats 0 . 5 mg/kg/min is approximately the upper limit ( Portellos et al . , 1995 ) for restricting the adenosine-induced fall of blood pressure to less than 10% , in rats a cumulative dose of 3 mg/kg reduced infarct size by 31% ( Shafy et al . , 2012 ) , and in humans infusing 0 . 07 mg/kg/min for 3 hr during reperfusion ( a cumulative dose of 12 . 6 mg/kg ) reduced infarct size by 57% in the AMISTAD-II trial ( Ross et al . , 2005 ) , although the REFLO-STEMI trial found no beneficial effect of adenosine ( Nazir et al . , 2016 ) . At the end of ischaemia/reperfusion animals were overdosed with pentobarbital sodium and transcardially perfused with saline ( 200 ml ) followed by 4% paraformaldehyde ( PFA , 200 ml ) for fixation and then 20 ml of 5% gelatin ( Sigma-Aldrich , G2625 ) solution containing FITC-albumin conjugate ( Sigma-Aldrich , A9771 ) to enable visualization of the perfused coronary microvasculature . The hearts were then fixed overnight in 4% PFA , and 150 µm transverse sections made for immunofluorescence staining . Pericytes were labelled with anti-NG2 antibodies ( Merck Millipore ) and the capillary basement membrane and pericytes were labelled ( Mishra et al . , 2014 ) with isolectin B4-Alexa Fluor 647 ( Molecular Probes , I32450 ) . The capillary diameters we measure with this protocol are not affected by vessels being compressed by myocyte contraction . The initial perfusion with calcium-free saline stops the heart in diastole . Evidence that the heart is indeed arrested in diastole is provided by the large volume visible within the left ventricle ( Figure 2A , C ) , which matches that seen during diastole in magnetic resonance images of rat heart , which also show that the volume during systole is far smaller ( see Figure 2B of Crowley et al . , 1997 ) . Indeed , it is clear that the observation of occluded capillaries after ischaemia does not reflect compression by arrest in systole , because essentially no capillary block was observed in control hearts that were not made ischaemic ( only 3% of capillaries were blocked , compared to 40% after ischaemia: see Figure 2D ) . Another demonstration that the observation of occluded capillaries does not reflect compression by arrest in systole is provided by the fact that capillary occlusion ( Figure 2D ) and reduction of FITC-albumin labelling ( Figure 2B ) were only seen on the side of the heart where the blood supply was transiently interrupted , and not on the normally perfused side of the heart , and were specifically associated with pericytes ( Figure 3H ) . Furthermore , our mean capillary diameter in control hearts ( 5 . 38 ± 0 . 28 microns , see Figure 3J ) is similar to that estimated for capillaries in diastole in rat hearts ( measured in relaxed hearts as 5 . 3 microns with a suggested correction to 5 microns: see Henquell et al . , 1976 ) and is greater than the diameter estimated for systole ( ~4 microns: Henquell et al . , 1976 ) . Image z-stacks of overall left ventricular myocardial blood volume , the perfusion of capillaries in the area at risk , and individual vessel blockages , were acquired using laser scanning confocal microscopy ( Zeiss LSM 700 ) , and analysed using ImageJ software ( NIH , Bethesda , MD , USA ) . To quantify blood volume across the left ventricular myocardium , ~30 low power z-stacks were taken of an entire transverse section of each heart ( using a 5X air objective ) , maximum intensity projected , and stitched together using the MosaicJ plugin ( Thévenaz and Unser , 2007 ) of ImageJ ( n = 6 hearts for ischaemia , n = 8 for ischaemia +adenosine , n = 5 for control ) . Usually some FITC-albumin labelled blood remained within the lumen of the ventricles ( although this blood often became detached during tissue processing ) . For ease of interpretation of the images in Figure 2 , this labelling ( and also the area outside the imaged heart ) was digitally recoloured black using Photoshop . All quantification of the FITC labelling was carried out on the unedited images . For quantification of the percentage of capillaries that were perfused , three randomly selected regions of the anterior left ventricle wall were imaged in both ischaemia and sham animals , as this region of myocardium consistently included the ischaemic area ( which showed visible pallor and oedema of the myocardium ) . The person quantifying the images was blinded to the condition that the heart was exposed to . 75 stacks ( 160 μm square , and 20 μm deep ) in 12 different animals were taken ( ischaemia n = 4 , ischaemia +adenosine n = 5 , control n = 3 ) . Blockages of flow in the ischaemic area at risk were identified by abrupt terminations in FITC-IB4 signal . For ischaemia 42 blockages were imaged; for ischaemia +adenosine 14 blockages were imaged . Clots ( or red blood cell rouleaux ) were very rarely observed . For low power blood volume analysis , 12 regions of interest ( ROI ) were drawn clockwise around the left ventricle ( when looked at from above ) from the mid-point of the septum ( as in Figure 2A ) , and the mean intensity of FITC-albumin signal was recorded for each ROI , and normalised to the highest intensity measured in any ROI . These data were averaged over hearts and renormalized so that the mean value in positions 1–3 of Figure 2C was 1 . This signal is assumed to be proportional to the volume of blood perfusing the myocardium . To compare perfusion in the ischaemic zone , we averaged the plotted values over ROIs 7–10 for each heart in the different conditions , and compared the mean values averaged over all the hearts studied in each condition . To quantify the percentage of perfused capillaries , we counted the number of filled and unfilled vessels crossing a line drawn through the centre of each image perpendicular to the main capillary axis . For blockages in flow , we measured the distance along the capillary from the termination of the FITC-albumin signal to the mid-point of the nearest visible pericyte soma on the same capillary . The cumulative probability distribution of the blockage-to-nearest-soma distance ( over all blocked capillaries ) was compared with the distribution predicted for random placement of a blockage on each blocked capillary . If pericytes were a uniform 60 μm distance apart ( Figure 1F ) along unbranched capillaries , then the latter cumulative probability distribution for the blockage-to-nearest-soma distance would increase linearly with distance , to reach a value of 1 at a distance of 30 μm ( half the distance between two pericytes ) . In practice , pericytes vary in their separation , vessels branch and , because it was necessary to image at high magnification to reliably locate blockage sites , sometimes the blocked vessel passes out of the image before one of the pericytes adjacent to the block is reached ( making it uncertain whether all positions on the imaged capillary are in fact closest to the imaged pericyte , rather than one outside of the image ) . We calculated the expected cumulative probability distribution of the distance from imaged pericytes to a randomly positioned vessel blockage as follows . For each blocked capillary the probability , p ( x ) . dx , of a location on the capillary ( and hence a potential blockage location ) existing between distances x and x + dx from the pericyte soma ( with x ≥ 0 ) was calculated as a function of distance from the soma ( see Figure 3—figure supplement 1 ) . For example , for a capillary of length L in the image , with a single pericyte soma located a distance A from one end , the probability is p ( x ) . dx=2 . dx/L for x < A , p ( x ) . dx=dx/L for A < x < L-A , and p ( x ) . dx=0 for x > L A , with the total probability ( integrated along the capillary length ) summing to 1 . For more complicated geometries ( e . g . two pericytes on the blocked capillary , or branching capillaries ) the same approach was used to calculate p ( x ) , taking into account all possible positions on the capillaries . The resulting probability distribution was then averaged over all 42 images used for Figure 3H . To generate a predicted cumulative probability distribution with the same resolution as that obtained for the experimentally observed blockages , this smooth distribution was then sampled at 42 equi-probability points as shown in Figure 3H . The resulting distribution shows a deviation from a linear increase to a value of 1 , at 30 μm distance , because of the factors mentioned above . Statistical analysis employed OriginPro . Data normality was assessed with Shapiro-Wilk tests . Comparisons of normally distributed data were made using 2-tailed Student’s t-tests . Equality of variance was assessed with an F test , and heteroscedastic t-tests were used if needed . Data that were not normally distributed were analysed with Mann-Whitney tests . P values were corrected for multiple comparisons using a procedure equivalent to the Holm-Bonferroni method ( for N comparisons , the most significant p value is multiplied by N , the 2nd most significant by N-1 , the 3rd most significant by N-2 , etc . ; corrected p values are significant if they are less than 0 . 05 ) . Cumulative probability distributions were compared using the Kolmogorov-Smirnov test . An estimate of the sample size needed for a typical experiment is as follows: For a control response of 100% , a response standard deviation of 25% , a response after a manipulation ( ischaemia or adenosine ) of 50% ( 50% inhibition ) , a power of 80% and p<0 . 05 , 6 samples are needed ( http://www . biomath . info/power/ttest . htm ) in each of the control and manipulation groups . The exact numbers depend on the effect size for the manipulation and the standard error of the data . | Heart attacks occur when one of the arteries supplying blood to the heart muscle becomes blocked , usually by a blood clot . Doctors unblock the artery and insert an expanding metal cage called a stent to keep it unblocked . This restores blood flow through the artery . Unfortunately , blood flow often does not return to smaller downstream blood vessels called capillaries . This can lead to further damage to the heart . Scientists have not been able to find a way to reliably open up those capillaries after a heart attack because it is not clear exactly what is keeping them closed . Muscle-like cells called pericytes , which wrap around the capillaries , are one possible culprit for the blockages . Pericytes narrow capillaries in the brain after stroke in animal experiments . These cells are also present on heart capillaries , but scientists do not know much about them . Now , O’Farrell , Mastitskaya , Hammond-Haley et al . show that pericytes are partly responsible for limiting blood flow in capillaries after a heart attack in rats . In the experiments , blood flow through an artery feeding the hearts of anaesthetized rats was restricted , simulating a heart attack . After the blood flow was later restored , 40% of the animal’s capillaries remained blocked . Many blockages occurred near pericytes that had narrowed the capillary preventing blood flow . Treating the rats with a drug called adenosine , which relaxes the pericytes , reduced capillary blockages and increased blood flow in the heart . Although adenosine could help to restore blood flow in the capillaries after a heart attack , it may also relax muscles around arteries and lower blood pressure , and so it may not be an ideal treatment . More studies are needed to determine whether drugs that target only the pericytes could complement existing heart attack treatments that unblock the arteries . If these studies are successful , pericyte-targeting drugs might prevent serious complications after a heart attack , including heart failure , heart rhythm abnormalities and future heart attacks . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2017 | Capillary pericytes mediate coronary no-reflow after myocardial ischaemia |
D2 autoreceptors regulate dopamine release throughout the brain . Two isoforms of the D2 receptor , D2S and D2L , are expressed in midbrain dopamine neurons . Differential roles of these isoforms as autoreceptors are poorly understood . By virally expressing the isoforms in dopamine neurons of D2 receptor knockout mice , this study assessed the calcium-dependence and drug-induced plasticity of D2S and D2L receptor-dependent G protein-coupled inwardly rectifying potassium ( GIRK ) currents . The results reveal that D2S , but not D2L receptors , exhibited calcium-dependent desensitization similar to that exhibited by endogenous autoreceptors . Two pathways of calcium signaling that regulated D2 autoreceptor-dependent GIRK signaling were identified , which distinctly affected desensitization and the magnitude of D2S and D2L receptor-dependent GIRK currents . Previous in vivo cocaine exposure removed calcium-dependent D2 autoreceptor desensitization in wild type , but not D2S-only mice . Thus , expression of D2S as the exclusive autoreceptor was insufficient for cocaine-induced plasticity , implying a functional role for the co-expression of D2S and D2L autoreceptors .
Central dopamine transmission coordinates reinforcement learning , including recognition of reward-predictive stimuli and initiation of goal-directed movements . Natural rewards , reward-predictive cues , and drugs of abuse elicit a rapid increase in dopamine release from dopamine axon terminals and somatodendritic sites within the ventral midbrain . Dopamine release is negatively regulated by the activation of dopamine D2 autoreceptors on somatodendritic and axon terminals ( reviewed in Ford , 2014 ) . Loss of D2 autoreceptor-mediated inhibition results in elevated extracellular dopamine and is associated with perseverative drug-seeking , enhanced motivation for food , and novelty-induced hyperactivity ( Marinelli and White , 2000; Marinelli et al . , 2003; Bello et al . , 2011; Anzalone et al . , 2012; Holroyd et al . , 2015 ) . Chronic D2 autoreceptor activation impairs the formation of dopamine- and glutamate-releasing axon terminals ( Fasano et al . , 2010 ) . Thus , D2 autoreceptors regulate structural and functional plasticity of dopamine neurons and are essential in limiting impulsivity and reward-seeking behaviors . A prominent feature of somatodendritic D2 autoreceptors is their activation of G protein-coupled inwardly rectifying potassium ( GIRK ) channels , resulting in inhibition of action potential firing and subsequent dopamine release . During prolonged activation , desensitization of D2 autoreceptors reduces the D2 autoreceptor-dependent GIRK current . A component of desensitization is dependent on intracellular calcium ( Beckstead and Williams , 2007 ) . Single or repeated exposure to drugs of abuse modifies D2 autoreceptor function ( Henry et al . , 1989; Wolf et al . , 1993; Jones et al . , 2000; Marinelli et al . , 2003; Gantz et al . , 2013; Madhavan et al . , 2013 ) , including a loss of the calcium-dependent component of D2 autoreceptor-GIRK desensitization ( Perra et al . , 2011 ) . The mechanism ( s ) that underlie acute desensitization and drug-induced plasticity of D2 autoreceptor-mediated inhibition remain incompletely characterized . There are two splice variants of the D2 receptor , which differ by a 29-amino acid insert in the third intracellular loop in D2-Long ( D2L ) that is absent in D2-Short ( D2S ) . Biased expression of D2S or D2L receptors alters behavioral responses to drugs of abuse ( Smith et al . , 2002; Bulwa et al . , 2011 ) and has been associated with drug addiction in human studies ( Sasabe et al . , 2007; Moyer et al . , 2011 ) . Functionally distinct roles for D2S and D2L receptors have been proposed based on characterization of mice lacking D2L ( Usiello et al . , 2000; Wang et al . , 2000 ) or D2S ( Radl et al . , 2013 ) . Behavioral and biochemical studies have designated D2L as the postsynaptic receptor expressed on non-dopaminergic medium spiny neurons in the basal forebrain and D2S as the autoreceptor ( Khan et al . , 1998; Usiello et al . , 2000; Lindgren et al . , 2003 ) . However , both D2S and D2L receptors are expressed in dopamine neurons and function as somatodendritic autoreceptors ( Khan et al . , 1998; Jomphe et al . , 2006; Jang et al . , 2011; Neve et al . , 2013; Dragicevic et al . , 2014 ) . Biochemical studies indicate that D2S receptors internalize and desensitize more readily than D2L receptors ( Liu et al . , 1992; Itokawa et al . , 1996; Ito et al . , 1999; Morris et al . , 2007; Thibault et al . , 2011 ) , but acute desensitization and drug-induced plasticity of D2S and D2L receptor-dependent GIRK currents have not been characterized . Using virus-mediated expression of the D2 receptor splice variants in D2 receptor knockout mice , this study reveals that D2S but not D2L receptor-dependent GIRK signaling exhibited calcium-dependent desensitization . Manipulations of pathways involved in D2 autoreceptor desensitization had distinct actions on D2S and D2L receptor-dependent GIRK currents . Lastly , a single in vivo cocaine exposure removed the calcium-dependent component of D2 autoreceptor-GIRK desensitization in wild type mice , but not D2S-only mice; thus , the expression of D2S as the exclusive autoreceptor was insufficient for drug-induced plasticity . Taken together , the results of this study imply a physiological role for the co-expression of D2S and D2L autoreceptors .
To examine the ability of D2S and D2L receptors to activate a GIRK conductance , single isoforms were expressed in midbrain dopamine neurons . Drd2−/− mice received bilateral injections of an adeno-associated viral ( AAV ) vector generating either rat D2S or D2L receptor and GFP expression , as previously described ( Neve et al . , 2013 ) . Infected neurons in brain slices containing the substantia nigra pars compacta ( SNc ) were identified by GFP visualization . Whole-cell patch clamp recordings were made from SNc dopamine neurons using an internal solution containing the calcium chelator , BAPTA ( 10 mM ) , as used previously ( Neve et al . , 2013 ) . Application of a saturating concentration of the D2 receptor agonist , quinpirole ( 30 µM ) , produced an outward current that was reversed by application of the D2 receptor antagonist , sulpiride ( 600 nM , Figure 1A , B ) . There was no difference in the peak amplitude of quinpirole-induced currents mediated by D2S and D2L receptors ( Figure 1B ) . In the continued presence of agonist , D2 autoreceptors desensitize resulting in a decline in the agonist-induced outward current ( Beckstead and Williams , 2007; Perra et al . , 2011 ) . The decline in the quinpirole-induced current mediated by D2S and D2L receptors was indistinguishable ( Figure 1A , C ) . 10 . 7554/eLife . 09358 . 003Figure 1 . When virally expressed in midbrain dopamine neurons , D2S and D2L function as autoreceptors . ( A ) Representative traces of whole-cell voltage clamp recordings , using a BAPTA-containing internal solution , of the outward current in D2S and D2L neurons induced by bath application of quinpirole ( 30 μM ) , which was reversed by sulpiride ( 600 nM ) . ( B ) The amplitude of quinpirole-induced currents in D2S and D2L neurons using BAPTA internal did not differ ( n = 12–14 , unpaired t test ) , shown in reference to the amplitude of the quinpirole-induced currents in WT neurons ( white line ) . ( C ) There was no difference between D2S and D2L in the decline of the D2 receptor-dependent current in the continued presence of quinpirole using BAPTA internal ( two-way ANOVA ) . ( D , E ) Representative traces of spontaneous D2-sIPSCs mediated by D2S and D2L receptors , blocked by sulpiride . Inset boxes are shown enlarged in ( E ) . The frequency and amplitude of D2S- and D2L-sIPSCs were not analyzed since these parameters may be influenced by the expression of D2 receptors in presynaptic dopamine neurons , which cannot be confirmed . ( F ) The duration of D2S-sIPSCs and D2L-sIPSCs did not differ ( n = 84–100 sIPSCs , Mann–Whitney U test ) . ns indicates not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 003 In the SNc , dopamine release from neighboring dopamine neurons elicits D2 receptor-mediated inhibitory postsynaptic currents ( IPSCs ) through the activation of GIRK channels ( Beckstead et al . , 2004; Gantz et al . , 2013 ) . D2S and D2L receptors mediate kinetically-identical IPSCs following electrically stimulated dopamine release ( Neve et al . , 2013 ) . Stimulus-independent dopamine release also occurs , resulting in spontaneous D2 receptor-mediated IPSCs ( Gantz et al . , 2013 ) . In slices from mice infected with either D2S or D2L , spontaneous IPSCs were abolished by application of sulpiride ( 600 nM , Figure 1D , E ) . The durations of D2S , D2L , and wild type D2 receptor-mediated spontaneous IPSCs were identical ( Figure 1F , [from Gantz et al . , 2013 , WT: 515 ± 17 ms , n = 76 sIPSCs] ) . Amplitude and frequency of spontaneous IPSCs are affected by the level of D2 receptor expression and dopamine synthesis ( Gantz et al . , 2013 , 2015 ) . Since these parameters may be influenced by variegated viral infection , the amplitude and frequency of D2S- and D2L-sIPSCs were not compared . Taken together , the results confirm previous work indicating that D2S and D2L can serve as autoreceptors at somatodendritic dopamine synapses ( Neve et al . , 2013 ) . Desensitization in the GIRK current induced by D2 receptor agonists is affected by intracellular calcium buffering . Weak calcium buffering with intracellular EGTA ( 0 . 025–0 . 4 mM ) results in greater decline in the GIRK current induced by D2 receptor agonists , without affecting the decline in the GIRK current induced by GABAB receptor agonists ( Beckstead and Williams , 2007; Perra et al . , 2011 ) . These results were confirmed in wild type mice using internal solutions containing either EGTA ( 0 . 1 mM , EGTA internal ) or BAPTA ( 10 mM , BAPTA internal ) . Application of quinpirole ( 10 µM ) or the GABAB agonist , baclofen ( 30 µM ) , resulted in outward currents that declined in the continued presence of agonist ( Figure 2A , E ) . The peak amplitudes of the quinpirole- and baclofen-induced currents were larger when using BAPTA internal than EGTA internal ( Figure 2A , B , E , F ) . The quinpirole-induced current desensitized more quickly with EGTA internal compared with experiments using BAPTA internal ( Figure 2A , C , D ) . This calcium-dependent desensitization was specific to the D2 receptor since the decline in baclofen-induced current was not dependent on the internal solution ( Figure 2E , G , H ) . Thus , as reported previously ( Beckstead and Williams , 2007; Perra et al . , 2011 ) , D2 autoreceptors exhibited a calcium-dependent desensitization that resulted in a larger decline in the D2 autoreceptor-dependent current when intracellular calcium was buffered with a low concentration of EGTA . 10 . 7554/eLife . 09358 . 004Figure 2 . Weak intracellular calcium buffering reveals calcium-dependent desensitization of D2 autoreceptor-dependent GIRK currents in wild type dopamine neurons . ( A ) Representative traces of whole-cell voltage clamp recordings of the outward current induced by bath application of quinpirole ( 10 μM ) that was reversed by sulpiride ( 600 nM ) , using a BAPTA or EGTA-containing internal solution . ( B ) The amplitude of the quinpirole-induced current was larger using BAPTA than EGTA internal ( n = 15 each , unpaired t test ) . ( C , D ) The decline in quinpirole-induced current was faster using EGTA internal compared to BAPTA ( C: two-way ANOVA followed by Bonferroni , D: unpaired t test ) ( E ) Representative traces of whole-cell voltage clamp recordings of the outward current induced by bath application of baclofen ( 30 μM ) which was reversed by CGP-55845 ( 200 nM ) , using BAPTA or EGTA internal . ( F ) The amplitude of the baclofen-induced current was larger using BAPTA than EGTA internal ( n = 14–16 , unpaired t test ) . ( G , H ) There was no difference in the decline in baclofen-induced current recorded with EGTA and BAPTA internals ( G: two-way ANOVA , D: unpaired t test ) . ns indicates not significant , *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 004 EGTA and BAPTA have a similar affinity for calcium but differ in the kinetics of binding . This property is frequently used to characterize the distance between a calcium source and a calcium sensor . Buffering with EGTA allows calcium to diffuse farther ( microdomain ) than BAPTA , which limits calcium spread ( nanodomain ) from a calcium source . However , the concentrations of EGTA and BAPTA used in this study may also result in different levels of resting free calcium ( Adler et al . , 1991 ) . To determine whether the difference in D2 autoreceptor desensitization observed with the two internals was explained by resting free calcium concentration , the level of free calcium in the BAPTA internal was increased to 300 nM by addition of CaCl2 ( 7 . 37 mM ) ( BAPTA+Ca2+ , see ‘Materials and methods’ ) . The peak amplitude and the decline in the quinpirole-induced current recorded with BAPTA+Ca2+ internal were not different from measurements recorded with BAPTA alone ( Figure 3A , B ) . Interestingly , BAPTA+Ca2+ internal decreased the peak amplitude of the baclofen-induced current significantly relative to the amplitude recorded with BAPTA internal ( Figure 3C ) , and was not different from the current measured with EGTA internal ( Figure 3C ) . The decline in the baclofen-induced current was unaffected by BAPTA+Ca2+ ( Figure 3D ) . 10 . 7554/eLife . 09358 . 005Figure 3 . Increasing resting free internal calcium does not enhance desensitization of D2 autoreceptor-dependent GIRK currents . ( A ) The amplitude of the quinpirole ( 10 μM ) -induced current using BAPTA+Ca2+ internal solution was not different from the amplitudes using BAPTA or EGTA internal solutions ( n = 7–15 , ANOVA followed by Bonferroni ) . ( B ) Increasing resting free calcium with BAPTA+Ca2+ had no effect on the decline in quinpirole-induced current ( two-way ANOVA ) . ( C ) Increasing resting free calcium with BAPTA+Ca2+ internal decreased the amplitude of the baclofen ( 30 μM ) -induced current , making it no greater than the amplitude recorded using EGTA internal ( ANOVA followed by Bonferroni ) . ( D ) Increasing resting free calcium with BAPTA+Ca2+ had no effect on the decline in baclofen-induced current ( two-way ANOVA ) . Additional experiments that demonstrate BAPTA+Ca2+ internal increased resting free calcium can be found in Figure 3—figure supplement 1 . ns indicates not significant , **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 00510 . 7554/eLife . 09358 . 006Figure 3—figure supplement 1 . The positive modulator of the SK channel , NS309 , produces an outward current when using the BAPTA+Ca2+ internal solution . ( A ) Representative trace of whole-cell voltage clamp recordings of the outward current induced by bath application of NS309 ( 10 μM ) , which was reversed by apamin ( 200 nM ) , using a BAPTA+Ca2+ internal . ( B ) NS309 produced a current using a BAPTA+Ca2+ , but not BAPTA or EGTA internal ( n = 5–6 , ANOVA followed by Bonferroni ) . ns indicates not significant , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 006 To verify that the resting free calcium was increased using BAPTA+Ca2+ internal , the positive modulator of the small conductance calcium-activated potassium channel ( SK ) , NS309 ( 10 µM ) was applied . Although NS309 did not produce a current using either BAPTA or EGTA internal , it caused an outward current with BAPTA+Ca2+ internal ( Figure 3—figure supplement 1A , B ) . The NS309-induced current was reversed by the SK channel blocker apamin ( 200 nM ) . Thus , the BAPTA+Ca2+ internal increased resting free calcium . Taken together , the results indicate that the resting free calcium had differential actions on D2 and GABAB receptor-dependent GIRK currents . The magnitude of the GABAB receptor-dependent current was sensitive to resting free calcium , but the decline in current was independent of resting free calcium . The decline in D2 autoreceptor-dependent current was dominated by the spatial regulation of intracellular calcium , not resting free calcium . Recordings were made from dopamine neurons that expressed D2S or D2L receptors using an internal solution containing EGTA ( 0 . 1 mM ) . Application of quinpirole ( 30 µM ) produced an outward current that declined and was reversed by sulpiride ( 600 nM ) . In D2S neurons , the decline using EGTA internal was faster than the decline using BAPTA internal ( Figure 4A , D , Figure 4—figure supplement 1A ) . In contrast , in D2L neurons , the decline with EGTA and BAPTA internal was not different ( Figure 4A , D , Figure 4—figure supplement 1B ) . The insensitivity of the decline in D2L neurons to calcium buffering resulted in significantly more desensitization of D2S than D2L with EGTA internal ( Figure 4B ) . The peak amplitude of the quinpirole-induced currents in D2S and D2L neurons was not different ( Figure 4C ) , indicating the difference between D2S and D2L is unlikely to be due to differences in the level of expression of D2 receptors . Application of the GABAB agonist , baclofen ( 30 µM ) , produced an outward current that was reversed by the GABAB antagonist , CGP-55845 ( 200 nM ) . The peak amplitude of the baclofen-induced current was not different among D2S ( EGTA: 259 ± 28 pA; BAPTA: 531 ± 94 pA ) , D2L ( EGTA: 277 ± 29 pA; BAPTA: 520 ± 43 pA ) , D2-KO ( AAV-GFP-only , EGTA: 279 ± 54 pA; BAPTA: 664 ± 80 pA ) , and wild type dopamine neurons ( p > 0 . 05 ) . There was also no change in the decline in the baclofen-induced current in D2S- or D2L-expressing neurons ( Figure 4E ) . 10 . 7554/eLife . 09358 . 007Figure 4 . D2S but not D2L receptor-GIRK currents exhibit calcium-dependent desensitization . ( A ) Representative traces of whole-cell voltage clamp recordings of the outward current in D2S and D2L neurons induced by bath application of quinpirole ( 30 μM ) which was reversed by sulpiride ( 600 nM ) , using EGTA internal , compared with the BAPTA trace shown in Figure 1A ( scaled and peak-aligned ) . ( B , D ) Using EGTA internal , the decline in quinpirole-induced current was greater in D2S than D2L neurons ( B: two-way ANOVA followed by Bonferroni , D: n = 16 each , one-way ANOVA followed by Fisher's LSD ) . ( C ) The amplitude of quinpirole-induced currents in D2S and D2L neurons using EGTA internal did not differ ( n = 16–17 , unpaired t test ) , shown in reference to the amplitude of the quinpirole-induced currents in WT neurons ( white line ) . ( D ) In D2S neurons the decline in quinpirole-induced current was greater using EGTA internal compared to BAPTA , but not in D2L neurons ( n = 12–16 , one-way ANOVA followed by Fisher's LSD ) . The time course of the decline can be found in Figure 4—figure supplement 1 . ( E ) There was no difference in the decline in baclofen-induced current recorded with EGTA or BAPTA internal in either splice variant ( n = 11–19 , one-way ANOVA ) . ( F ) In neurons from transgenic D2S mice , the amplitude of the quinpirole-induced current was larger using BAPTA than EGTA internal ( n = 7–8 , unpaired t test ) . ( G , H ) Representative scaled and peak-aligned traces of whole-cell voltage clamp recordings from neurons from transgenic D2S mice of the outward currents induced by bath application of quinpirole ( 10 μM ) , which were reversed by sulpiride . The decline in quinpirole-induced current was greater using EGTA internal compared to BAPTA ( two-way ANOVA followed by Bonferroni ) . ns indicates not significant , *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 00710 . 7554/eLife . 09358 . 008Figure 4—figure supplement 1 . Time course of desensitization of D2 receptor splice variant-GIRK currents . ( A ) In D2S neurons , the decline in quinpirole ( 30 μM ) -induced current was greater using EGTA internal compared to BAPTA ( n = 10–16 ) . ( B ) In D2L neurons , the decline in quinpirole-induced current using EGTA internal was no different from BAPTA internal ( n = 10–16 ) . Two-way ANOVAs followed by Bonferroni ) . ns indicates not significant , *p < 0 . 05 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 00810 . 7554/eLife . 09358 . 009Figure 4—figure supplement 2 . Expression and labeling of Flag-D2S receptors in dopamine neurons . ( A ) Representative confocal microscopy images of Flag-D2S receptors clustered on the soma , dendrites , and spine-like structures of dopamine neurons , labeled by incubation of live slices in Alexa Fluor-594 conjugated anti-Flag M1 antibody ( red , Flag-D2S ) , then post-fixed and immunostained for tyrosine hydroxylase ( green , TH ) , scale bars: 1 μm ( upper left inset ) and 5 μm . ( B ) Representative two-photon microscopy images of live dopamine neurons , where Flag-D2S receptors were labeled by incubation of live slices in Alexa Fluor-594 conjugated anti-Flag M1 antibody ( Flag-D2S ) , scale bars: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 009 To minimize potential confounds of ectopic D2 receptor expression in non-dopamine neurons in the midbrain , the calcium-sensitivity of D2S receptor-GIRK desensitization was validated using a transgenic D2S mouse line , generated by a cross between TH-hD2S ( Gantz et al . , 2013 ) and D2 receptor knockout mice . In this line , the expression of Flag-tagged human D2S receptors depends on the tyrosine hydroxylase promoter ( Figure 4—figure supplement 2 ) . In slices from these mice , quinpirole ( 10 µM ) produced larger outward currents using BAPTA internal compared to EGTA internal ( Figure 4F ) . The currents were significantly larger than those recorded in wild type dopamine neurons ( Figures 2B , 4F ) , indicating overexpression of D2 receptors . Despite the overexpression , the magnitude of the decline in the quinpirole-induced current was similar to wild type ( Figures 2C , D , 4H ) . Also consistent with wild type , the decline in the quinpirole-induced current using EGTA internal was significantly faster than the decline using BAPTA internal ( Figure 4G , H ) . Taken together , these results indicate that D2S but not D2L receptor-GIRK signaling exhibited calcium-dependent desensitization . Drugs of abuse change the D2 autoreceptor activation of GIRK conductance ( Arora et al . , 2011; Gantz et al . , 2013; Dragicevic et al . , 2014; Sharpe et al . , 2014 ) . One of these changes is a loss of the calcium-dependent component of D2 autoreceptor desensitization after repeated ethanol exposure ( Perra et al . , 2011 ) . Wild type mice were given a single injection of cocaine ( 20 mg/kg , i . p . ) or an equal volume of saline , and brain slices were made 24 hr later . With EGTA internal , the quinpirole-induced current declined significantly less in slices from cocaine-treated mice compared to control mice ( saline-treated and naïve , Figure 7A ) . In fact , the decline was no longer statistically different from that found with BAPTA internal ( Figure 7A ) . There was no difference in the decline in the quinpirole-induced current in slices taken from control or cocaine-treated mice with BAPTA internal ( Figure 7A ) . There was no change in the decline in the baclofen-induced current , using either internal ( Figure 7F ) . Thus , in wild type mice , a single in vivo cocaine exposure resulted in the loss of calcium-dependent D2 autoreceptor desensitization without affecting GABAB receptor desensitization . 10 . 7554/eLife . 09358 . 014Figure 7 . Effects of a single in vivo cocaine exposure on calcium-dependent D2 autoreceptor desensitization . ( A ) In neurons from cocaine-treated wild type mice using EGTA internal , the quinpirole-induced current declined less compared to naïve/saline-treated mice , to a level comparable to the decline recorded with BAPTA internal . Cocaine exposure did not alter the decline in the quinpirole-induced current when measured with BAPTA internal ( n = 11–26 ) . ( B ) In D2L neurons after in vivo cocaine exposure , there was no difference in the decline in quinpirole-induced current recorded with EGTA or BAPTA internal ( n = 6–7 ) . ( C , D ) In neurons from ( C ) AAV-D2S and ( D ) transgenic D2S mice , the decline in quinpirole-induced current was still greater using EGTA internal compared to BAPTA after in vivo cocaine exposure ( C: n = 10 each , D: n = 8–9 ) . ( D , E ) Co-expression of both splice variants by ( D ) viral expression of D2L in transgenic D2S mice and ( E ) infection with a mixture of AAV-D2S and AAV-D2L removed the calcium-dependence of the decline in the quinpirole-induced current ( D: n = 5–8 , E: n = 6–9 ) and there was no change after in vivo cocaine ( E: n = 7–9 ) . ( F ) Previous cocaine exposure had no effect on the decline in baclofen-induced current recorded with EGTA or BAPTA internal in wild type neurons ( n = 13–27 ) . Comparisons were made with one-way ANOVAs followed when p < 0 . 05 by Fisher's LSD . ns indicates not significant , *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09358 . 014 A loss of calcium-dependent D2 autoreceptor desensitization after cocaine exposure in wild type neurons could be due to a change in calcium signaling or a functional increase in the contribution of D2L receptors . To test whether D2L receptors are involved , Drd2−/− mice that had received midbrain injections of AAV-D2S or AAV-D2L were given a single injection of cocaine ( 20 mg/kg , i . p . ) and brain slices were made 24 hr later . In D2L neurons , cocaine exposure did not alter the decline in quinpirole-induced current ( Figure 7B ) . Likewise , cocaine exposure did not alter the calcium-dependent decline in the quinpirole-induced current in D2S neurons . Similar to naïve AAV-D2S mice , the decline of the quinpirole-induced current was greater using EGTA internal than with BAPTA internal ( Figure 7C ) . Thus , unlike what was found in slices from wild type mice , cocaine exposure did not reduce the calcium-dependent decline in the quinpirole-induced current . This result was not dependent on overexpression as it was also observed in D2S neurons in which quinpirole produced outward currents of similar magnitude to wild type neurons ( data not shown ) . This result was also recapitulated in the transgenic D2S mice ( Figure 7D ) , where expression of D2S receptors in the midbrain is restricted to dopamine neurons . Since wild type , but not D2S-only dopamine neurons exhibited a reduction in calcium-dependent desensitization after cocaine exposure , these results suggest that constitutive or viral-mediated expression of D2S as the exclusive autoreceptor was insufficient for cocaine-induced plasticity . To determine if the expression of D2L was sufficient to enable loss of calcium-dependent D2 receptor desensitization of D2S , Drd2−/− mice received bilateral injections of a 1:1 mixture of AAV-D2S and AAV-D2L . In dopamine neurons from mice infected with both splice variants , the amplitude of the quinpirole-induced currents was similar to those measured in neurons expressing D2S- or D2L-only ( EGTA: 333 ± 48 pA , n = 9; BAPTA: 442 ± 86 pA , n = 6 , see Figures 1B , 4C for comparison ) . This suggests a similar level of D2 receptor overexpression as in neurons that express single variants . Surprisingly , the decline in the quinpirole-induced current was similar between EGTA and BAPTA internals ( Figure 7E ) . Therefore , the viral expression of both D2S and D2L receptors did not mimic D2 receptor-dependent GIRK signaling in naïve wild type mice . Moreover , the decline in the quinpirole-induced current did not change after in vivo cocaine exposure ( Figure 7E ) , suggesting that the viral expression of both D2S and D2L receptors precluded cocaine-induced plasticity in D2 receptor-dependent GIRK signaling . To ensure that this result was not due to preferential expression of D2L receptors following injection of the D2S/D2L virus mixture , dopamine neurons in transgenic D2S mice were infected with AAV-D2L . The expression of D2S was confirmed by labeling dopamine neurons with an Alexa Fluor 594-conjugated M1 antibody and imaging on a two-photon microscope ( e . g . , Figure 4—figure supplement 2B ) . Recordings were made from neurons with Flag-D2S staining that were also GFP+ ( AAV-D2L ) . With EGTA internal , the decline in the quinpirole-induced current in transgenic D2S neurons also expressing D2L was equivalent to the decline measured in neurons receiving the D2S/D2L virus mixture , and significantly less than the decline in the quinpirole-induced current in transgenic D2S-only neurons ( Figure 7D ) . As observed in wild type , there was no change in the decline in the baclofen-induced current after cocaine exposure in any of the groups ( p > 0 . 05 for all comparisons , data not shown ) . Taken together , the results indicate that regardless of the presence of D2S , the viral expression of D2L eliminated calcium-dependent D2 receptor desensitization and precluded cocaine-induced plasticity .
Calcium entry promotes desensitization of D2 autoreceptors in wild type dopamine neurons . Buffering intracellular calcium with BAPTA reduces the decline in D2 autoreceptor- , but not GABAB receptor-mediated GIRK currents ( Beckstead and Williams , 2007; Perra et al . , 2011 ) . In this study , the calcium-dependent component of D2 autoreceptor desensitization was observed in wild type and D2S-only expressing dopamine neurons . However , in neurons where D2L receptors were virally expressed , there was no calcium-dependent desensitization . This was observed whether D2L receptors were expressed alone , or in conjunction with transgenic or virally expressed D2S receptors . This study describes two calcium sources that regulate D2 autoreceptor-dependent GIRK currents: intracellular calcium stores and L-type calcium channels . These intracellular pathways did not regulate desensitization of GABAB receptor-dependent GIRK currents . Consistent with a previous report ( Perra et al . , 2011 ) , depleting intracellular calcium stores removed calcium-dependent D2 autoreceptor desensitization in wild type neurons . Depleting intracellular calcium stores also reduced the magnitude of D2S receptor desensitization to a saturating concentration of agonist , without affecting D2L receptor desensitization . Preventing calcium entry from L-type calcium channels also reduced D2S receptor desensitization , without affecting wild type or D2L receptor desensitization . The results demonstrate that the calcium-dependent component of D2S receptor desensitization was readily modifiable . Desensitization of D2 autoreceptors in wild type dopamine neurons was controlled by elevated concentrations of calcium in intracellular microdomains and could not be enhanced by raising the resting free calcium concentration . The lack of a calcium-dependent component in D2L receptor desensitization could be due to localization outside of the calcium microdomains , despite showing similar distribution in the somatodendritic compartment as D2S receptors ( Jomphe et al . , 2006 ) , or to another property of this isoform . Depleting intracellular calcium stores or blocking L-type calcium channels produced robust augmentation in D2L receptor-dependent GIRK currents produced by iontophoretically applied dopamine that was greater than the augmentation of D2S receptor-dependent GIRK currents . The lack of effect of manipulating calcium signaling on D2L receptor desensitization in the presence of a saturating concentration of quinpirole suggests that the enhanced response of D2L to iontophoretically applied dopamine does not reflect removal of tonic desensitization . Nonetheless , the same intracellular pathways interacting with D2S receptors also modified D2L receptor-dependent GIRK currents . It is therefore likely that D2S and D2L receptors are in similar calcium microdomains and the lack of apparent calcium-dependent desensitization upon saturating agonist exposure is a property specific to the D2L isoform . Drugs of abuse cause functional changes to dopamine neuron physiology , including regulation of D2 and GABAB receptor activation of GIRK conductance ( Beckstead et al . , 2009; Arora et al . , 2011; Perra et al . , 2011; Padgett et al . , 2012; Dragicevic et al . , 2014; Sharpe et al . , 2014 ) . Several recent studies reported drug-induced changes in D2 autoreceptor mediated-GIRK signaling that are contingent on the method of recording ( whole-cell vs perforated-patch , Dragicevic et al . , 2014 ) or the calcium buffering capabilities of the whole-cell internal solution ( Perra et al . , 2011; Sharpe et al . , 2014 ) implicating dependence on intracellular calcium . In this study , 24 hr after a single in vivo cocaine exposure , the calcium-dependent component of D2 autoreceptor desensitization was eliminated , similar to the change observed after repeated ethanol exposure ( Perra et al . , 2011 ) . Thus , this study confirms the Perra et al . ( 2011 ) finding and further demonstrates that the plasticity in D2 autoreceptor function did not require repeated drug exposure . Whether this plasticity was due to a change in calcium-dependent pathways or the D2 autoreceptors themselves was previously unresolved . The findings of this study support the latter . Cocaine exposure may change the calcium-dependent pathways examined in this study . Twenty-four h after a single cocaine exposure , metabotropic glutamate receptor 1 signaling is attenuated ( Kramer and Williams , 2015 ) . The activation of metabotropic glutamate receptors decreases D2 autoreceptor-dependent GIRK currents ( Perra et al . , 2011 ) and may desensitize D2 autoreceptors through calcium release from intracellular stores . A change in the contribution of calcium influx via L-type calcium channels to D2 autoreceptor desensitization may also result from cocaine exposure ( Dragicevic et al . , 2014 ) . In this study , depleting intracellular calcium stores or blocking L-type calcium channels readily removed calcium-dependent D2S receptor desensitization . Given these results , it was surprising that cocaine exposure did not alter calcium-dependent D2S receptor desensitization . This result was recapitulated in dopamine neurons from transgenic D2S mice indicating it was not an artifact of virus-mediated expression . Thus , these results suggest that the expression of D2S as the exclusive autoreceptor is insufficient for cocaine-induced plasticity observed in wild type dopamine neurons . In wild type dopamine neurons , it may be that the expression of D2L receptors is involved in cocaine-induced plasticity . Biased expression of D2S and D2L receptors has been associated with drug abuse . The loss of D2L receptors and concomitant overexpression of D2S receptors in D2L-deficient mice is associated with altered drug-taking ( Bulwa et al . , 2011 ) and conditioned place preference ( Smith et al . , 2002 ) . In addition , single nucleotide polymorphisms that result in overexpression of D2S receptors are observed in humans with a history of drug abuse ( Sasabe et al . , 2007; Moyer et al . , 2011 ) . In this study , the viral expression of D2L receptors , alone or with D2S receptors , resulted in a loss of calcium-dependent D2 receptor desensitization . Moreover , it precluded any further cocaine-induced reduction in calcium-dependent D2 receptor desensitization . These results suggest that the overexpression of D2L receptors resembles cocaine-induced plasticity . Transient elevation in extracellular dopamine up-regulates D2L mRNA ( Zhang et al . , 1994; Oomizu et al . , 2003; Giordano et al . , 2006; Wernicke et al . , 2010; but see; Filtz et al . , 1993; Dragicevic et al . , 2014 ) . In addition , D2L receptors are retained in intracellular compartments more so than D2S receptors and exposure to D2 agonists results in the preferential translocation of existing and nascent D2L receptors to the membrane ( Filtz et al . , 1993; Zhang et al . , 1994; Fishburn et al . , 1995; Starr et al . , 1995; Ng et al . , 1997; Prou et al . , 2001 ) . Thus , it may be that exposure to cocaine in wild type mice increases functional D2L receptors , resulting in the loss of calcium-dependent D2 autoreceptor desensitization . Virally expressed D2L receptors may not be subject to the same regulation as endogenously expressed D2L receptors , in such a way that virus-mediated overexpression of D2L mimics and eliminates any requirement for up-regulation of D2L function . The D2S isoform has been thought to be the D2 autoreceptor due to preservation of autoreceptor-mediated behaviors in D2L-deficient mice and more abundant D2S immunolabeling in the SNc ( Khan et al . , 1998; Usiello et al . , 2000 ) . However , immunolabeled D2L receptors are found in SNc dopamine neurons ( Khan et al . , 1998 ) and rodent studies describe dopamine neurons expressing both D2S and D2L mRNA , D2L-only , or D2S-only ( Jang et al . , 2011; Dragicevic et al . , 2014 ) . Both variants are capable of inhibiting action potential firing ( Jomphe et al . , 2006; Jang et al . , 2011; Dragicevic et al . , 2014 ) . In this study , D2S and D2L receptors , when expressed in dopamine neurons , activated a GIRK conductance and were capable of producing IPSCs occurring from spontaneous fusion of dopamine-filled vesicles . Thus , D2S and D2L can serve as autoreceptors at somatodendritic dopamine synapses , as previously demonstrated ( Neve et al . , 2013 ) . Although many of the basic properties of D2S and D2L receptor-dependent currents were similar , there were some differences that suggest both D2S and D2L are autoreceptors in wild type dopamine neurons . The calcium-dependent component of D2 autoreceptor desensitization in wild type neurons was similar to D2S-only neurons . However , results from manipulating calcium signaling in wild type neurons were more consistent with a mix of D2S and D2L receptor expression . Intracellular pathways for calcium signaling that regulated D2S receptor-GIRK desensitization modified the magnitude of D2L receptor-dependent GIRK currents . In wild type dopamine neurons , manipulating these pathways resulted in a decrease in acute desensitization and larger GIRK currents , suggesting that D2S and D2L receptor regulation may operate in parallel in wild type neurons . In addition , cocaine-induced plasticity occurred in wild type , but not D2S-only neurons , indicating a loss of some process in neurons which express D2S as the exclusive autoreceptor that is permissive to cocaine-induced plasticity . However , the viral-mediated co-expression of D2S with D2L receptors also did not resemble wild type , and instead was similar to D2L-only . Although it is not known to what extent developmental compensation , virus-mediated expression , and variegated D2 receptor expression in Drd2−/− mice affected D2 receptor translation or trafficking ( i . e . affecting the ratio of functional D2S and D2L receptors ) , or other regulatory elements of D2 receptor signaling , this result suggests that in wild type dopamine neurons , the functional expression of D2L may be limited . Changes in calcium signaling or exposure to cocaine may bring about an increased contribution of D2L , although this has yet to be directly demonstrated . Taken together , this study suggests that D2S may serve as the predominant autoreceptor under basal conditions , but the functional contribution of D2L autoreceptors may be revealed after drug exposure . This study advances the understanding of D2 autoreceptor regulation . Two pathways for calcium signaling that regulated D2 autoreceptor-dependent GIRK signaling were identified , which distinctly affected D2S and D2L receptors . In addition , distinct action of in vivo cocaine exposure in wild type , D2S , and D2L receptor-GIRK signaling was demonstrated . Since not all dopamine neurons express both D2S and D2L receptors , this study suggests that D2 autoreceptors in a subset of dopamine neurons are regulated differently by calcium and resistant to cocaine-induced plasticity . Given the heterogeneity of dopamine neurons and their projections ( reviewed in Roeper , 2013 ) , a greater understanding of this subset may reveal insights into plasticity in their projection areas .
All studies were conducted in accordance with the Institutional Animal Care and Use Committees at the VA Portland Health Care System ( VAPORHCS ) and Oregon Health & Science University ( OHSU ) . Mice of both sexes were used in this study ( 65–120 days old ) . Wild type ( C57BL/6 ) mice , obtained from The Jackson laboratory ( Sacramento , CA ) , and TH-hD2S mice were bred at OHSU . Drd2−/− mice were bred at the VAPORHCS Veterinary Medical Unit and were maintained on a C57BL/6 background . Mice were housed in standard plastic containers on a 12 hr light/dark cycle . Food and water were available ad libitum , and after stereotaxic injections , diet was supplemented with Diet Gel RE placed on the floor of the cage . ‘Transgenic D2S’ mice were produced by crossing Drd2−/− mice with transgenic TH-hD2S mice , which express Flag-tagged human D2S receptors under the tyrosine hydroxylase promoter ( Gantz et al . , 2013 ) , as shown by immunostaining for the Flag epitope with an Alexa Fluor 594-conjugated M1 antibody and confocal or two-photon microscopy ( Figure 4—supplement 2 ) . Treated animals received one injection of cocaine ( 20 mg/kg , intraperitoneally ) dissolved in saline , or an equal volume of saline , 22–24 hr prior to use . There were no differences found between saline-treated and naïve mice , so data were combined . D2 receptors were ubiquitously expressed in the midbrain using an AAV vector ( AAV9 D2-IRES-GFP; Virovek , Inc . , Hayward , CA ) encoding rat D2S or D2L receptors ( Neve et al . , 2013 ) , or a 1:1 mixture of AAV-D2S and AAV-D2L . Mice were injected when 65–90 days old . Mice were immobilized in a stereotaxic alignment system under an anesthesia cocktail consisting of 7 . 1 mg/kg xylazine , 71 . 4 mg/kg ketamine , and 1 . 4 mg/kg acepromazine ( 10 ml/kg ) . Mice received bilateral injections , each 500 nl volume at a rate of 200 nl/min , with the injection needle left in place for an additional 5 min before it was slowly withdrawn . The coordinates for injections were ( with respect to bregma ) AP −3 . 26 mm , ML ±1 . 2 mm , DV −4 . 0 mm . After injections , mice recovered in individual or group housing for 3–4 weeks to allow for expression . Infected neurons were identified in the slice by visualization of eGFP . Whole-cell voltage clamp recordings ( holding potential −60 mV ) were made as previously described ( Gantz et al . , 2013 ) . Whenever possible , experiments were conducted blinded to treatment or splice variant expression . Mice were deeply anesthetized with isoflurane and killed by decapitation . Brains were removed quickly and placed in ice-cold physiologically equivalent saline solution ( modified Krebs buffer ) containing ( in mM ) 126 NaCl , 2 . 5 KCl , 1 . 2 MgCl2 , 2 . 4 CaCl2 , 1 . 4 NaH2PO4 , 25 NaHCO3 , and 11 D-glucose with MK-801 ( 10 μM ) , and cut horizontally ( 220 μm ) using a vibrating microtome ( Leica ) . Slices recovered at 30°C in vials with 95/5% O2/CO2 saline with MK-801 ( 10 μM ) for at least 30 min prior to recording . Slices were then mounted in a recording chamber and perfused at a rate of ∼3 . 0 ml/min with 33–35 . 5°C modified Krebs buffer . Recordings were made exclusively from neurons in the SNc identified visually by their location lateral to the medial terminal nucleus of the accessory optic . The neurochemical identity of AAV-D2-infected cells was not verified post-recording . Rather , dopamine neurons were identified by location and electrophysiological properties , namely the presence of spontaneous pacemaker firing of broad ( ∼2 ms ) action potentials at 1–5 Hz in cell-attached mode ( Ford et al . , 2006 ) , characteristic passive membrane properties including capacitance and resting conductance ( Gantz et al . , 2011 ) , and a prominent slow hyperpolarization-activated inward ( Ih ) current ( Ford et al . , 2006 ) . These parameters readily distinguished dopamine neurons from GABAergic neurons . The expression of D2 receptors was not used as a physiological criterion for dopamine neuron identity . However , all dopamine neurons from wild type mice identified by location and electrophysiological properties had a D2 receptor-dependent outward current upon quinpirole application . Recordings were obtained with large glass electrodes with a resistance of 1 . 3–1 . 9 MΩ when filled with internal solution containing either , ( in mM ) 115 K-methanesulfonate , 20 NaCl , 1 . 5 MgCl2 , 10 HEPES ( K ) , 2 ATP , 0 . 2 GTP , 10 phosphocreatine , and 10 BAPTA ( K4 ) or 130 K-methanesulfonate , 20 KCl , 1 MgCl2 , 10 HEPES ( K ) , 2 ATP , 0 . 2 GTP , 10 phosphocreatine , and 0 . 1 EGTA; pH 7 . 33–7 . 40 , 275–288 mOsm . The concentration of CaCl2 required to increase resting free calcium in BAPTA internal was determined with use of the EGTAetc program , provided by EW McCleskey . Within 2 min of break-in , membrane capacitance , series resistance , and input resistance were measured with the application of 3 pulses ( ±2 mV for 50 ms ) averaged before computation using the Axograph ( sampled at 50 kHz , filtered at 10 kHz ) . Series resistance was monitored to ensure sufficient and stable electrical access to the inside of the cell throughout the experiment ( <12 MΩ ) . Cells were dialyzed with internal solution for >10 min prior to drug application ( Foehring et al . , 2009 ) . All drugs were applied through bath perfusion , except dopamine , which was applied by iontophoresis . Quinpirole and baclofen were applied with >10 min between the agonist applications with the application order alternated between recordings . The amplitude and the decline in the currents were not affected by the order in which the drugs applied . Slices were exposed to saturating concentrations of each agonist once . Recordings where the peak amplitude of the current was <50 pA were excluded from decline analysis . Dopamine hydrochloride ( 1 M ) was ejected as a cation with a single pulse ( 2–10 ms , >20 nA ) from a thin-walled iontophoretic electrode placed within 10 μm of the soma once every 50 s . Access resistance was assessed during these recordings with a brief ( 200 ms ) step to −70 mV once every 50 s . Data were acquired using AxoGraph software ( AxographX , Berkeley , CA ) and Chart 7 ( AD Instruments , Colorado Springs , CO ) and were post hoc filtered . The amplitude of currents induced by iontophoretic application of dopamine was determined by averaging the current ±20 ms from the greatest upward deflection . For each cell , 6–24 consecutive currents were averaged to determine ‘baseline’ ( preceding drug application ) and post-drug amplitudes . sIPSCs were detected and analyzed as previously described ( Gantz et al . , 2013 ) . Briefly , single peak sIPSCs with amplitudes greater than 2 . 1 times the SD of baseline noise were detected using a semiautomated sliding template detection procedure with AxoGraph X . Each detected event was visually inspected and discarded if the baseline noise was greater than the sIPSC peak ±1 s from the peak . Duration of sIPSCs was determined by measuring the width at 20% of the peak . Brain slices were prepared and allowed to recover , as described for electrophysiology . Slices were incubated in Alexa Fluor 594-conjugated M1 antibody ( 10 µg/ml ) for 40 min at 35°C . Live slices were observed with a custom-built two-photon microscope using ScanImage Software ( Pologruto et al . , 2003 ) . Expression of eGFP was visualized using a CCD camera of epi-fluorescence activation . Slices for laser-scanning confocal microscopy were washed 10 min in modified Krebs buffer before fixation in 4% paraformaldehyde ( 45 min at 24°C ) in phosphate buffered saline + CaCl2 ( 1 mM , PBS+Ca2+ ) . Slices were blocked and permeabilized in PBS+Ca2+ with 0 . 3% Triton-X and 0 . 5% fish skin gelatin for 80 min . Slices were incubated overnight in rabbit anti-tyrosine hydroxylase antibody ( 1:1000 in PBS+Ca2+ + 0 . 05% NaN3 ) . Washed slices were incubated in Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody ( 1:1000 in PBS+Ca2+ + 0 . 05% NaN3 , 2 hr at 24°C ) . Washed slices were mounted with Fluoromount aqueous medium with #1 . 5 glass coverslips . Images were collected on a Zeiss confocal LSM 780 microscope with a 40× water-emersion lens ( 1 . 2 nA ) . All images were processed with Fiji . CGP-55845 was obtained from Tocris Bioscience ( Minneapolis , MN ) . MK-801 and CPA were obtained from Abcam ( Cambridge , MA ) . Cocaine hydrochloride was obtained from National Institute on Drug Abuse-National Institutes of Health ( Bethesda , MD ) . All other drugs were obtained from Sigma–Aldrich ( St . Louis , MO ) . Values are given as means ± SEM and unless otherwise noted n = number of cells . Data sets with n > 10 were tested for normality with a Shapiro–Wilk test . Significant between-group differences were determined in two group comparisons by unpaired two-tailed t tests or Mann–Whitney U tests , and in more than two groups comparisons by one- or two-way ANOVAs . ANOVAs were followed when p < 0 . 05 with uncorrected Fisher's LSD or Bonferroni's multiple comparisons post hoc tests . Significant differences in within-group comparisons were determined by paired two-tailed t tests . Statistical analysis was performed using GraphPad Prism 6 ( GraphPad Software , Inc . , La Jolla , CA ) . | Dopamine is an important component of the brain's reward system and is commonly referred to as a ‘feel-good’ chemical . It is mainly released from neurons in the brain in response to natural rewards , such as food or sex , and following exposure to , or in anticipation of , certain drugs of abuse ( including cocaine ) . Dopamine-releasing neurons also sense dopamine , and just like someone can change the volume of their voice by hearing themselves speak , dopamine neurons regulate how much dopamine is released based on how much dopamine they sense . This feedback system is known as autoinhibition . These neurons sense dopamine when it binds to , and activates , so-called ‘dopamine D2 receptors’ on their cell surface . But not all D2 receptors are alike . Instead there are two variants called D2S and D2L . Previous studies have shown that D2 receptor signaling in dopamine neurons is altered by the concentration of calcium ions inside these cells . Furthermore , exposure to cocaine and other drugs is known to change how these calcium ions regulate D2 receptor signaling . Now , Gantz et al . have used mice that produce only a single variant of the D2 receptor ( either D2S or D2L ) in their dopamine neurons to uncover similarities and differences between the two variants . The experiments show that localized increases in calcium ion concentration make D2S less capable of autoinhibition , like D2 receptors in neurons from wild type mice , without affecting autoinhibition by D2L . In further experiments , some of these mice were given cocaine before D2 receptor signaling was assessed . In dopamine neurons from wild type mice , a single exposure to cocaine eliminates the calcium-dependent regulation; thus , cocaine treatment causes a D2L-like response . In contrast , cocaine treatment did not affect the calcium-dependent regulation when only one variant of the D2 receptor was present . This implies that dopamine neurons must have both D2S and D2L receptors before the drug can induce changes in D2 receptor signaling . These findings also challenge the long-held view that the D2S receptor is the predominant form involved in autoinhibition . The next challenge is to determine how cocaine induces an apparent switch from D2S to D2L and the implications of this switch for the development of cocaine addiction . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Distinct regulation of dopamine D2S and D2L autoreceptor signaling by calcium |
Precise coordination of synaptic connections ensures proper information flow within circuits . The activity of presynaptic organizing molecules signaling to downstream pathways is essential for such coordination , though such entities remain incompletely known . We show that LRP4 , a conserved transmembrane protein known for its postsynaptic roles , functions presynaptically as an organizing molecule . In the Drosophila brain , LRP4 localizes to the nerve terminals at or near active zones . Loss of presynaptic LRP4 reduces excitatory ( not inhibitory ) synapse number , impairs active zone architecture , and abolishes olfactory attraction - the latter of which can be suppressed by reducing presynaptic GABAB receptors . LRP4 overexpression increases synapse number in excitatory and inhibitory neurons , suggesting an instructive role and a common downstream synapse addition pathway . Mechanistically , LRP4 functions via the conserved kinase SRPK79D to ensure normal synapse number and behavior . This highlights a presynaptic function for LRP4 , enabling deeper understanding of how synapse organization is coordinated .
Multiple levels of synaptic organization ensure accurate , controlled information flow through neuronal circuits . Neurons must first make an appropriate number of synaptic connections with their postsynaptic partners . Each of these synaptic connections must have appropriate strength that can be modified by plasticity and homeostasis as a result of experience and activity changes . Further , there must be an appropriate balance between excitatory and inhibitory synapses . Finally , recent work has shown that these connections also occupy precise locations with regards to the three-dimensional structure of the synaptic neuropil . Indeed , circuit models for diverse neuronal ensembles fail to recapitulate functional patterns unless these aspects are accounted for ( Kim et al . , 2014; Vlasits et al . , 2016 ) . The misregulation of any one of these organizational parameters can result in neurodevelopmental disorders and intellectual disabilities like autism ( Mullins et al . , 2016 ) , epilepsy ( Bonansco and Fuenzalida , 2016 ) , and other synaptopathies ( Grant , 2012 ) . Revealing the molecular mechanisms that ensure all of these facets are achieved is a critical step in understanding circuit assembly and function . Synaptic organizers like Neurexins / Neuroligins , Teneurins , protein tyrosine phosphatases ( PTPs ) , leucine rich repeat transmembrane proteins ( LRRTMs ) , and Ephrin / Eph receptors , among others , ensure the proper number , distribution , and function of synaptic connections ( Hruska and Dalva , 2012; Mosca , 2015; Siddiqui and Craig , 2011; Südhof , 2008; Takahashi and Craig , 2013; de Wit and Ghosh , 2016 ) . Loss-of-function mutations in these key synaptogenic molecules have deleterious structural , functional , and organizational consequences for synapses and circuits . At the vertebrate neuromuscular junction , one of these critical organizers is LRP4 . There , it forms a receptor complex with MuSK in muscle fibers to promote clustering of acetylcholine receptors in response to motoneuron-derived agrin ( Zhang et al . , 2008; Kim et al . , 2008; Weatherbee et al . , 2006 ) . Muscle LRP4 can also function as a retrograde signal with an unknown motoneuron receptor to regulate presynaptic differentiation ( Yumoto et al . , 2012 ) . In these roles , the known functions from LRP4 are overwhelmingly postsynaptic . However , a number of lines of evidence suggest a broader role , beyond postsynaptic , for LRP4 . First , motoneuron-derived LRP4 can regulate presynaptic differentiation , demonstrating a role for neuronal LRP4 ( Wu et al . , 2012 ) . Second , in the vertebrate central nervous system ( CNS ) , agrin is not essential for synapse formation ( Daniels , 2012 ) though LRP4 can regulate synaptic plasticity , development , and cognitive function ( Gomez et al . , 2014; Pohlkamp et al . , 2015 ) , through functioning in astrocytes in some cases ( Sun et al . , 2016 ) . In this vein , the Drosophila genome contains an LRP4 homologue , but no clear agrin or MuSK homologues ( Adams et al . , 2000 ) , so any role for LRP4 there must be agrin-independent . Here , we show in the Drosophila CNS that LRP4 is a presynaptic protein that regulates the number , architecture , and function of synapses . LRP4 functions largely through the conserved , presynaptic SR-protein kinase , SRPK79D . LRP4 and SRPK79D interact genetically and epistatically , as SRPK79D overexpression can suppress lrp4-related phenotypes . Unexpectedly , this role for LRP4 occurs preferentially in excitatory neurons , as impairing lrp4 in inhibitory neurons has no effect . As little is known about the presynaptic determinants ( save neurotransmitter-related enzymes and transporters ) of excitatory versus inhibitory synapses , this may suggest a new mode for distinguishing such synapses from the presynaptic side . Thus , LRP4 may represent a conserved synaptic organizer that functions presynaptically , cell autonomously , and independently of agrin to coordinate synapse number and function .
We identified CG8909 as the fly LRP4 homologue ( Figure 1—figure supplements 1 and 2A ) , which is predicted to be a single-pass transmembrane protein whose domain organization resembles that of mammalian LRP4 ( Figure 1A ) . Drosophila LRP4 shares 38% identity with human LRP4 overall , 61% identity within the LDL-repeat containing extracellular portion , and 28% identity in the intracellular tail . Consistent with previous expression data from whole-brain microarrays ( Chintapalli et al . , 2007 ) , we determined that LRP4 was expressed throughout the adult brain using antibodies against the endogenous protein ( Figure 1B–C ) or an lrp4-GAL4 transgene that expresses GAL4 under the lrp4 promoter and visualized with either Syt-HA ( Figure 1D ) or an HA epitope-tagged LRP4 ( Figure 1—figure supplement 2C ) . All methods revealed similar patterns of expression in the antennal lobes ( Figure 1 and Figure 1—figure supplement 2C–E ) , optic lobes , and higher olfactory centers including the mushroom body and the lateral horn ( Figure 1B , D ) . Antibody specificity was validated by the complete loss of signal in a deletion ( see below ) of the lrp4 coding region ( Figure 1C ) . We further investigated LRP4 in the antennal lobe , the first olfactory processing center in the Drosophila CNS , which has emerged as a model circuit for studying sensory processing ( Wilson , 2013 ) and whose synaptic organization was recently mapped at high resolution ( Mosca and Luo , 2014 ) . 10 . 7554/eLife . 27347 . 003Figure 1 . LRP4 is a synaptic protein expressed in excitatory neurons . ( A ) Domain structure of Drosophila LRP4 . Numbers indicate amino acids . EXT , extracellular side . INT , intracellular side . ( B ) Representative confocal image stack of a control Drosophila brain stained with antibodies against endogenous LRP4 ( green ) and Bruchpilot ( inset , magenta ) demonstrating expression throughout the brain . ( C ) Representative confocal image stack of an lrp4dalek null brain stained with antibodies against LRP4 ( green ) and Brp ( inset , magenta ) demonstrating antibody specificity . ( D ) Representative confocal image of a Drosophila brain expressing UAS-Syt-HA via lrp4-GAL4 and stained with antibodies to HA ( D , green ) and N-Cadherin ( inset , magenta ) . The expression pattern resembles that of endogenous LRP4 , supporting the specificity of lrp4-GAL4 . ( E ) Representative single slice within a single antennal lobe glomerulus of a brain processed for expansion microscopy ( proExM ) expressing LRP4-HA and Brp-Short-mStraw in all ORNs via pebbled-GAL4 and stained with antibodies to HA ( E , E” , green ) and mStraw ( E’-E” , magenta ) . LRP4 localizes to synaptic neuropil regions . ( F ) High magnification image of the region bounded by dashed lines in ( E ) and stained as above . Arrows indicate LRP4-HA localization adjacent to / not directly overlapping with Bruchpilot-Short . Arrowheads indicate overlapping LRP4-HA and Brp-Short localization . ( G–K ) Representative high magnification confocal stack images of neuronal cell bodies surrounding the antennal lobe in animals expressing UAS-mCD8-GFP via lrp4-GAL4 and stained for antibodies against GFP ( G-K , green ) and other cell-type markers ( G’-K’ , magenta ) . Merge channels ( G’’–K’’ ) show colocalization of lrp4 with the neuronal marker ELAV ( G’’ ) but not the glial cell marker Repo ( H’’ ) . Neurons positive for lrp4 show colocalization with choline acetyltransferase ( ChAT , I’’ ) , and the vesicular glutamate transporter ( vGlut , J’’ ) , but little to no colocalization with the inhibitory neurotransmitter GABA ( K’’ ) , suggesting that lrp4-positive cells are largely excitatory neurons . The percentage of GFP-positive cells that are ALSO positive for the cell-type specific marker are as follows: Elav = 99 . 50 ± 0 . 19% overlap; Repo = 0 . 38 ± 0 . 18% overlap; ChAT = 59 . 13 ± 2 . 48% overlap; vGlut = 22 . 38 ± 1 . 28% overlap; GABA = 0 . 25 ± 0 . 16% overlap . For all cases , n = 8 animals , ≥ 200 cells per animal . Values = mean ± s . e . m . Scale bars = 50 µm ( B–D ) , 150 μm ( B-D , insets ) , 25 μm ( E–F ) , 10 μm ( G–K ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 00310 . 7554/eLife . 27347 . 004Figure 1—figure supplement 1 . Sequence alignment of Drosophila , mouse , and human LRP4 homologues . Multiple sequence alignment of Drosophila LRP4 ( CG8909; accession AAF48538 . 1 ) , Mus musculus LRP4 ( accession NP_766256 . 3 ) , and Homo sapiens LRP4 ( accession NP_002325 . 2 ) . Red shading = identical residues . Yellow shading = similar residues . The transmembrane domains are underlined . Considerable identity is seen throughout the extracellular side of the protein , and stretches of identity and similarity are also observed in the intracellular side . Asterisks denote a putative internalization signal , NPxY ( Hussain , 2001 ) , which is conserved on the intracellular side of all three species . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 00410 . 7554/eLife . 27347 . 005Figure 1—figure supplement 2 . LRP4 reagents and patterns of LRP4 expression . ( A ) Genomic region of lrp4 . Top bar represents physical position on the X chromosome ( in base pairs ) , and the blue arrow represents the lrp4 genomic region flanked by other genes ( yellow ) . Primer sets are indicated by forward and reverse arrows ( see B ) . The exon structure is displayed with 5’ and 3’ UTRs shaded in gray and coding exons numbered and shaded in beige . The region deleted by the lrp4dalek mutation is indicated in pink . RNAi targets are shown below in orange . The position of the GAL4 in the GMR90B08-GAL4 line is shown below and region of the protein against which antibodies were raised are noted below . ( B ) PCR analysis of genomic DNA from control and lrp4dalek adults . The presence of bands corresponding to Exon 2 and Exon 7–8 in control and heterozygous flies and their absence in lrp4dalek demonstrate loss of the coding region . The presence of a 315 bp band in heterozygous and homozygous lrp4dalek flies ( Flank ) but not in control is a result of non-homologous end joining of the 5’ and 3’ UTRs following deletion of the gene . ( C ) Representative confocal maximum intensity projections of the antennal lobe region of an lrp4-GAL4 animal expressing HA-tagged LRP4 and stained with antibodies to HA ( C , C’’ , green ) and N-Cadherin ( C’-C’’ , magenta ) . LRP4-HA localizes to regions of synaptic neuropil , similar to endogenous staining ( Figure 1 ) . ( D–E ) Representative confocal maximum intensity projections of antennal lobes in animals expressing UAS-FRT-Stop-FRT-mCD8-GFP using lrp4-GAL4 but where FLP expression ( removing the stop codon ) is restricted to either ORNs using eyFLP ( D ) or PNs using GH146-FLP ( E ) and stained with antibodies to GFP ( green ) and N-Cadherin ( magenta ) . Intersectional analysis reveals lrp4 expression in both ORNs as well as PNs . Scale bars = 10 µm ( C ) , 5 μm ( D–E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 00510 . 7554/eLife . 27347 . 006Figure 1—figure supplement 3 . Validation of expansion microscopy in Drosophila . ( A ) A representative Drosophila brain stained with antibodies against endogenous Brp ( green ) and processed for proExM . Organization of the fly brain is maintained as are identifiable landmarks . ( inset ) Unexpanded brain stained with antibodies against Brp , shown at the same scale as the expanded brain . Individual Brp puncta are resolvable in the expanded brain but not in the unexpanded brain . ( B ) A representative antennal lobe in a Drosophila brain expressing Brp-Short-mStraw ( magenta ) in all ORNs using the pebbled-GAL4 driver and stained with antibodies against mStraw ( magenta ) . Following proExM processing , glomerular structure and fine synaptic detail are still present . ( inset ) An equivalent single section from a different brain of the same genotype . In all cases , note the 4–5 fold isotropic expansion of tissue allowing for enhanced resolution while still using confocal microscopy . Scale bars as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 006 LRP4 was enriched in the synaptic neuropil of the antennal lobe ( Figure 1B ) . As this neuropil is made up of processes from multiple classes of olfactory neurons , all of which make presynaptic connections there , we used intersectional strategies with lrp4-GAL4 to identify which neurons expressed lrp4 . These approaches revealed lrp4 expression in both olfactory receptor neurons ( ORNs; Figure 1—figure supplement 2D ) and projection neurons ( PNs; Figure 1—figure supplement 2E ) . Because of the observed neuropil expression of LRP4 ( Figure 1B–C ) , we sought to examine the localization of LRP4 with regards to a known synaptic protein , the active zone scaffolding component Bruchpilot ( Wagh et al . , 2006 ) . However , due to the density of CNS neuropil , colocalization analyses using light level microscopy have inherently low resolution . Therefore , we applied expansion microscopy ( Chen et al . , 2015 ) to the Drosophila CNS to improve the resolution of colocalization analysis . This technique uses isotropic expansion of immunolabeled tissue ( Tillberg et al . , 2016 ) while maintaining the spatial relationship between protein targets and allowing for enhanced resolution with confocal microscopy . Using protein-retention expansion microscopy ( proExM ) , we obtained reliable , ~4 fold isotropic expansion of Drosophila CNS tissue ( Figure 1—figure supplement 3 ) . To specifically examine the relationship between LRP4 and active zones only in ORNs , we expressed HA-tagged LRP4 and Brp-Short-mStraw using the pebbled-GAL4 driver ( Sweeney et al . , 2007 ) . LRP4-HA expressed using lrp4-GAL4 localizes to similar regions as LRP4 antibody staining ( Figure 1B and Figure 1—figure supplement 2C ) , suggesting the fidelity of this transgene . Within individual expanded glomeruli of proExM-treated brains , LRP4 and Brp localized to similar regions ( Figure 1E ) and , when examined at high magnification , LRP4 localized either coincidentally with Brp ( Figure 1F , arrowhead ) or to the space adjacent to active zones ( Figure 1F , arrow ) . This combination of active zone and periactive zone localization is similar to that of known synaptic organizers ( Jepson et al . , 2014; Li et al . , 2007; Mosca et al . , 2012 ) . Thus , LRP4 is a synaptic protein that localizes to nerve terminals . Given widespread expression throughout the brain , we sought to identify the cell types that express LRP4 . To accomplish this , we used lrp4-GAL4 driven mCD8-GFP as this approach , in addition to labeling similar neuropil regions as the antibody , also highlighted the cell bodies of lrp4-positive cells . We co-stained brains for various cellular and neuronal-subtype markers and quantified the overlap between cells positive for lrp4-expression and expression of these various labels . Nearly all lrp4-positive cells observed ( 99 . 5% ) expressed the neuronal marker ELAV ( Robinow and White , 1988 ) ( Figure 1G ) , indicating that these cells were neurons . Few ( 0 . 4% ) expressed the glial marker Repo ( Xiong et al . , 1994 ) ( Figure 1H ) . The majority of lrp4-positive cells ( 59 . 1% ) also expressed choline acetyltransferase ( ChAT; Figure 1I ) , a marker for cholinergic excitatory neurons . We also observed partial overlap between lrp4-positive neurons and vGlut ( 22 . 4%; Figure 1J ) , the vesicular transporter for glutamate . In the fly brain , glutamatergic neurons can be either excitatory or inhibitory ( Liu and Wilson , 2013 ) . Interestingly , there was little overlap ( 0 . 3% ) between lrp4 and GABA , the major inhibitory neurotransmitter in Drosophila ( Figure 1K ) . Thus , LRP4 is expressed at synaptic terminals of a subset of excitatory cholinergic neurons and a subset of glutamatergic neurons that may be excitatory or inhibitory , but is excluded from inhibitory GABAergic neurons . As both the expression and localization of LRP4 were consistent with the protein serving a synaptic role , we sought to determine whether disrupting its function in excitatory neurons would affect synapse number . To image these connections , we expressed fluorescently tagged synaptic markers ( Fouquet et al . , 2009; Leiss et al . , 2009; Mosca and Luo , 2014 ) and used previously established methods to estimate the number of active zones and postsynaptic receptor puncta ( Mosca and Luo , 2014 ) in olfactory neurons in antennal lobe glomeruli ( Figure 2A ) . These methods show stereotyped active zone numbers and densities in ORNs and can reveal the function of synaptic proteins in mediating these aspects ( Mosca and Luo , 2014 ) . Further , measurements from these methods are consistent with our own electron microscopy ( Mosca and Luo , 2014 ) as well as results from ultrastructural reconstructions of all synapses in individual glomeruli ( Tobin et al . , 2017 ) demonstrating their utility . To perturb LRP4 function , we created a null mutation ( lrp4dalek ) using the CRISPR-Cas9 system ( Gratz et al . , 2013 ) that removed the entire coding region ( Figure 1—figure supplement 2A–B ) . lrp4dalek mutants were viable with a slightly reduced body size . 10 . 7554/eLife . 27347 . 007Figure 2 . LRP4 perturbation in excitatory neurons alters synapse number . ( A ) Schematic diagram of the fly brain with major regions labeled and the olfactory regions examined in this study shaded in red ( AL , antennal lobe ) or yellow ( LH , the lateral horn ) . Olfactory receptor neurons ( ORNs , black ) , excitatory projection neurons ( ePNs , red ) , and local interneurons ( LNs , brown ) are indicated . White dashed lines represent a glomerulus . Magnification: the antennal lobe region with the three glomeruli examined here highlighted: DA1 ( green ) , VA1d ( blue ) , and VA1v ( purple ) . ( B–E ) Representative high magnification confocal stack images of VA1v ORN axon terminals in the VA1v glomerulus of males expressing Brp-Short-mStraw and stained with antibodies against mStraw ( red ) and N-Cadherin ( blue ) . Loss of lrp4 ( lrp4dalek ) and RNAi against lrp4 expressed only in ORNs ( ORN lrp4IR-2 ) show fewer Brp-Short-mStraw puncta while LRP4 overexpression in ORNs ( ORN LRP4 OE ) increases the number of Brp-Short-mStraw puncta . ( F–G ) Representative high magnification confocal maximum intensity projections of DA1 and VA1d PN dendrites in males expressing Dα7-EGFP , a tagged acetylcholine receptor subunit . Loss of lrp4 ( lrp4dalek ) also results in fewer Dα7-EGFP puncta . ( H ) Quantification of Brp-Short-mStraw puncta ( red , left axis ) and neurite volume ( black , right axis ) in VA1v ORNs . ( I ) Quantification of Dα7-EGFP puncta ( green , left axis ) and neurite volume ( black , right axis ) . ****p<0 . 0001; ***p<0 . 001; ns , not significant . Statistical comparisons in 2H ( one-way ANOVA with correction for multiple comparisons ) are with control . Statistical comparisons between two samples are done via Student’s t-test . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 00710 . 7554/eLife . 27347 . 008Figure 2—figure supplement 1 . Representative antennal lobe images for genetic lrp4 manipulations . ( A–D ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals in males expressing Brp-Short-mStraw and stained with antibodies against mStraw ( red ) and N-Cadherin ( blue ) . Loss of lrp4 using independent RNAi transgenes expressed only in ORNs all result in fewer Brp-Short-mStraw puncta . ( E–K ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals in males expressing mCD8-GFP and stained with antibodies against GFP ( green ) and N-Cadherin ( blue ) . mCD8-GFP staining ( and thus , neurite volume ) is unchanged by loss of lrp4 via null mutation ( F , lrp4dalek ) or ORN expressed RNAi ( G-J , ORN lrp4IR1-4 ) , or LRP4 overexpression ( K , ORN LRP4 OE ) . ( L–M ) Representative high magnification confocal stack images of DA1 and VA1d PN dendrites expressing membrane-tagged tdTomato ( mtdT ) and stained with antibodies against tdTomato ( red ) . Loss of lrp4 ( M ) does not affect neurite volume as measured by surface rendering of tdTomato staining . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 00810 . 7554/eLife . 27347 . 009Figure 2—figure supplement 2 . lrp4 perturbation in females affects synapse number . ( A–E ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals in females expressing Brp-Short-mStraw and stained with antibodies against mStraw ( red ) and N-Cadherin ( blue ) . Loss of lrp4 using independent RNAi transgenes expressed in ORNs ( b-e , ORN lrp4IR-1-4 ) result in fewer Brp-Short-mStraw puncta . Note that VA1v is sexually dimorphic in size and synapse number ( Mosca and Luo , 2014; Stockinger et al . , 2005 ) hence we examined males and females separately . ( F–J ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals in females expressing mCD8-GFP and stained with antibodies against GFP ( green ) and N-Cadherin ( blue ) . In all cases , mCD8-GFP staining is unaffected by lrp4 loss . ( K–L ) Representative high magnification confocal maximum intensity projection of VA1v ORNs in females expressing Brp-Short-mStraw ( K ) or mCD8-GFP ( L ) while concomitantly overexpressing LRP4 and stained with antibodies against mStraw ( K , red ) or GFP ( L , green ) and N-Cadherin ( blue ) . LRP4 overexpression increases Brp-Short-mStraw puncta number without affecting mCD8-GFP staining . ( M ) Quantification of Brp-Short-puncta ( red , left axis ) and neurite volume ( black , right axis ) in ORNs . In both cases , n ( antennal lobes ) is noted at the bottom of each column . **** p<0 . 0001; ns , not significant . Statistical comparisons ( one-way ANOVA corrected for multiple comparisons ) are with control unless noted . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 00910 . 7554/eLife . 27347 . 010Figure 2—figure supplement 3 . lrp4 RNAi reduces synapse number in multiple glomeruli . ( A–I ) Representative high magnification confocal maximum intensity projections of ORN axon terminals in males expressing Brp-Short-mStraw ( red ) projecting to the DL4 and DM6 ( A–C ) , VA1d ( D–F ) , or DA1 glomeruli ( G–I ) and stained with antibodies to mStraw ( red ) and N-Cadherin ( blue ) . RNAi-mediated knockdown of lrp4 reduces Brp-Short puncta number in multiple glomeruli ( B–C , E–F , H–I ) demonstrating that lrp4 generally affects synapse number in olfactory glomeruli and is not restricted to the VA1v ORNs . ( J ) Quantification of Brp-Short-mStraw puncta expressed as a percentage of control puncta for different glomeruli . For all glomeruli , a similar reduction in puncta number was observed . ****p<0 . 0001 . Statistical comparisons ( one-way ANOVA with correction for multiple comparisons ) are with control . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bars = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 01010 . 7554/eLife . 27347 . 011Figure 2—figure supplement 4 . lrp4 RNAi reduces Syd1 puncta in presynaptic ORN terminals . ( A–C ) Representative high magnification confocal maximum intensity projections of DA1 ORN axon terminals in males expressing DSyd1-GFP and stained with antibodies to GFP ( green ) and N-Cadherin ( blue ) . Presynaptic knockdown of lrp4 ( B , lrp4IR-1; C , lrp4IR-2 ) reduces the number of DSyd1-GFP puncta when compared to Control ( A ) . ( D ) Quantification of DSyd1-GFP puncta . ****p<0 . 0001 . Statistical comparisons ( one-way ANOVA corrected for multiple comparisons ) are with control unless noted . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 011 In ORN axon terminals projecting to the VA1v glomerulus in males ( Figure 2B ) , lrp4dalek mutants ( Figure 2C , H ) showed a 31% reduction in the number of puncta for Brp-Short , an active zone marker , compared to control adults ( Figure 2B , H ) . This phenotype was recapitulated when we expressed any of four independent transgenic RNAi constructs against lrp4 only in ORNs ( Figure 2D , H , and Figure 2—figure supplement 1 ) , demonstrating that LRP4 functions presynaptically in regulating active zone number . These changes were independent of glomerular volume: lrp4 loss-of-function had no effect on neurite volume ( Figure 2H and Figure 2—figure supplement 1 ) . Though the intensity of Brp-Short puncta across some genotypes trended slightly downward , it did not reach statistical significance ( data not shown ) . We also observed that lrp4 disruption ( using lrp4dalek mutants and presynaptic RNAi expression ) caused a quantitatively similar reduction of active zone numbers in VA1v ORN axon terminals in females in this sexually dimorphic glomerulus ( Figure 2—figure supplement 2 ) , and in ORN axon terminals projecting to the VA1d , DA1 , DL4 , and DM6 glomeruli ( Figure 2—figure supplement 3 ) . This suggests that lrp4 phenotypes are not specific to particular glomeruli . Beyond Brp-Short , we observed similar phenotypes with an independent presynaptic marker , DSyd-1 ( Owald et al . , 2012 ) , that is also punctate at ORN terminals ( Mosca and Luo , 2014 ) ( Figure 2—figure supplement 4 ) . We further examined the consequences of lrp4 disruption on the number of Dα7 acetylcholine receptor puncta in PN dendrites postsynaptic to the ORN axon terminals imaged above . Loss of lrp4 decreased Dα7-EGFP puncta numbers by 29% compared to controls ( Figure 2F–G , I ) . This deficit was also independent of neurite volume ( Figure 2I and Figure 2—figure supplement 2 ) , again demonstrating that lrp4 perturbation phenotypes did not result from decreased neuronal projection size . Further , both the presynaptic active zone and postsynaptic acetylcholine receptor phenotypes were quantitatively similar . While likely that the postsynaptic AChR number decreases concomitantly with the presynaptic active zone number ( which is controlled by presynaptic LRP4 ) , we cannot exclude an additional postsynaptic role for LRP4 ( see Discussion ) . However , it is evident that the loss of LRP4 reduces synapse number as assayed both pre- and postsynaptically . The above experiments demonstrated the necessity of presynaptic LRP4 in ensuring the proper number of synaptic connections . However , with known presynaptic organizers like Neurexin , overexpression results in added boutons ( Li et al . , 2007 ) and active zones ( Craig and Kang , 2007 ) . To test for LRP4 sufficiency in synapse addition , we overexpressed HA-tagged LRP4 presynaptically in otherwise wild-type ORNs . LRP4 overexpression increased the number of Brp-Short puncta by 30% ( Figure 2E , H , and Figure 2—figure supplement 2 ) ; this increase was also independent of neurite volume ( Figure 2h and Figure 2—figure supplements 2–3 ) as the glomeruli remained the same size . Thus , there is a direct relationship between presynaptic LRP4 expression and synapse number in excitatory neurons: removing LRP4 reduces , while overexpressing LRP4 increases , synapse number . Though light level analyses accurately report fold-changes in synapse number ( Chen et al . , 2014; Mosca and Luo , 2014 ) , we sought to independently confirm and extend our analyses using electron microscopy . Using transmission electron microscopy ( TEM ) on the fly antennal lobe , we quantified synapse number in putative ORN terminals based on morphology ( Rybak et al . , 2016; Tobin et al . , 2017 ) in both control ( Figure 3A ) and lrp4dalek ( Figure 3B ) adult brains . T-bar profiles were evident in both genotypes , but were reduced in number by 31% in mutant terminals ( Figure 3C ) , which exactly matched the reduction observed by Brp-Short puncta measurements ( Figure 2H ) . Terminal perimeter was slightly but significantly increased in lrp4dalek terminals ( Figure 3D ) , resulting in a 36% reduction in T-bar density when compared to control ( Figure 3E ) . These results are consistent with those observed via confocal microscopy , and demonstrate that LRP4 is necessary for the proper number of synapses in putative ORN terminals of the antennal lobe . 10 . 7554/eLife . 27347 . 012Figure 3 . Loss of LRP4 causes defects in T-bar number and morphology . ( A–B ) Representative transmission electron micrographs of putative ORN terminal in Control ( A ) and lrp4dalek ( B ) adult antennal lobes . Loss of lrp4 results in fewer observed T-bar profiles ( asterisk ) and a larger terminal perimeter . Scale bar = 1 µm . ( C ) Quantification of T-bar profiles per terminal in Control and lrp4dalek terminals . Loss of LRP4 results in a 31% reduction of T-bars . ( D ) Quantification of terminal perimeter in Control and lrp4dalek adults . Mutant terminals have a 13% greater perimeter than control terminals . ( E ) Quantification of the T-bar density per µm of terminal perimeter . Loss of LRP4 causes a 36% reduction in T-bar density when the increased terminal perimeter is accounted for . For ( C–E ) , Control has n = 5 animals , 2688 terminals and lrp4dalek has n = 3 animals , 3123 terminals . The number of terminals measured is listed below the genotype . ****p<0 . 0001 . Statistical comparisons ( two-tailed Student’s t-test ) are done between genotypes . Error bars represent mean ± s . e . m . ( F–H ) Representative transmission electron micrographs of individual T-bar profiles ( asterisk ) in control adults . Single ( F ) , double ( G ) , and triple ( H ) profiles are readily visible . ( I–Q ) Representative transmission electron micrographs of individual T-bar profiles in lrp4dalek adults . As in control flies , single ( I ) , double ( J ) and triple ( K ) T-bar profiles were visible . The majority of T-bars , however , demonstrated morphology defects including those that lacked table tops ( L ) , were detached from the membrane ( M–N ) , were misshapen ( N–P ) , and profiles containing four or more connected T-bars ( Q ) . These all represent morphological defects that are not observed ( or very rarely observed ) in control adults . Scale bar = 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 012 Brp-Short assays alone cannot distinguish between normal and impaired active zones . We therefore examined the ultrastructural morphology of individual active zones to determine if LRP4 had an additional role in the biogenesis of the T-bar itself . In both control ( Figure 3F–H ) and lrp4 ( Figure 3I–K ) terminals , we observed single ( Figure 3F–I ) , double ( Figure 3G , J ) , and triple T-bars ( Figure 3H , K ) suggesting that LRP4 is not absolutely required for T-bar formation and some elements of organization . However , whereas irregular T-bars in control animals were rare ( <5% of total T-bars ) , the majority of T-bars in lrp4 mutants displayed one or more defects ( Figure 3L–Q ) , including immature T-bars that lacked tops ( Figure 3L ) , detached T-bars ( Figure 3M ) , misshapen T-bars of varying configurations and aggregations ( Figure 3N–P ) , and multiple T-bars beyond those observed in control animals ( Figure 3Q ) . Thus , in addition to controlling the number of synapses , LRP4 is also required for individual active zones to assume normal morphology , attach to the membrane , and have proper spacing . Thus , LRP4 has multiple , critical roles in central synapse formation . The preferential expression of lrp4 in excitatory but not inhibitory neurons ( Figure 1 ) suggests that it promotes synapse addition specifically in excitatory neurons . To test this , we used Brp-Short to examine synapse number in GABAergic inhibitory neurons projecting to the antennal lobe using the GAD1-GAL4 driver ( Ng et al . , 2002 ) . Though GAD1-positive neurons project throughout the antennal lobe ( Mehren and Griffith , 2006 ) , we restricted our analyses to the DA1 glomerulus , where we observed reductions in excitatory synapses ( Figure 2—figure supplements 3–4 ) following LRP4 disruption . When LRP4 function was impaired using the lrp4dalek mutant or RNAi in these neurons , synapse number was unaffected ( Figure 4A–B , D ) . Thus , the reduction of synapse number under LRP4 loss-of-function conditions appeared specific for excitatory neurons . 10 . 7554/eLife . 27347 . 013Figure 4 . Effects of LRP4 perturbation on inhibitory neuron synapse formation . ( A–C ) Representative high magnification confocal maximum intensity projections of GAD1-positive inhibitory neurons , which project to the DA1 glomerulus ( dashed line ) , in males expressing Brp-Short-mStraw and stained with antibodies against mStraw ( red ) and N-Cadherin ( blue ) . Due to the proximity of inhibitory neuron cell bodies to the antennal lobe , saturated somatic signal is observed . Loss of lrp4 ( lrp4dalek ) does not affect puncta number but overexpression of LRP4 ( GAD1 LRP4 OE ) increases Brp-Short puncta . ( D ) Quantification of Brp-Short-mStraw puncta ( red , left axis ) and neurite volume ( black , right axis ) in GAD1 neurons . Neither loss of lrp4 nor RNAi against lrp4 expressed in inhibitory neurons affects puncta number or neurite volume . ****p<0 . 0001; ***p<0 . 001; ns , not significant . Statistical comparisons ( one-way ANOVA with correction for multiple comparisons ) are with control . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 013 Interestingly , when LRP4 was overexpressed in inhibitory neurons , we observed a 35% increase in synapse number without an accompanying change in neurite volume , similar to what we observed for excitatory neurons ( Figure 4C–D ) . This suggests that , while inhibitory GABAergic neurons do not normally utilize LRP4 to regulate synapse number , they possess the downstream machinery necessary for LRP4 to function in adding synapses . Thus , when LRP4 is exogenously expressed in these cells , it can co-opt this machinery for synapse addition . As such , excitatory and inhibitory neurons likely use distinct cell surface synaptic organizers ( LRP4 for excitatory neurons ) that converge on common mechanisms for synapse addition . Though we initially restricted our analyses to the antennal lobe , we also observed lrp4 expression throughout the brain , including two higher order olfactory neuropil: the mushroom body and the lateral horn ( Figure 1B–D ) . To determine whether LRP4 could also serve as a synaptic organizer in these brain regions , we examined the effects of lrp4 perturbation on both excitatory and inhibitory synapses in the lateral horn ( LH , Figure 5A ) , a higher order olfactory center involved in innate olfactory behavior ( Heimbeck et al . , 2001 ) . We used Mz19-GAL4 to label projection neurons whose dendrites and cell bodies are restricted to the antennal lobe region , but whose axon terminals make excitatory synapses in the lateral horn ( Berdnik et al . , 2006 ) . To label inhibitory synapses , we used the Mz699-GAL4 driver , which is expressed in inhibitory projection neurons ( iPNs ) whose dendrites project to the antennal lobe and whose axons project to the lateral horn ( Lai et al . , 2008; Liang et al . , 2013 ) . Mz699-GAL4 also labels a small subset of third-order neurons that project dendrites largely void of presynaptic terminals to the ventral lateral horn ( Liang et al . , 2013 ) . Thus , we consider synaptic signal labeled by Mz699-GAL4 as being contributed mostly by iPNs . 10 . 7554/eLife . 27347 . 014Figure 5 . LRP4 perturbations similarly affect higher order olfactory centers . ( A ) Schematic diagram of the fly brain with major regions labeled and the olfactory regions examined in this study shaded in red ( AL , antennal lobe ) or yellow ( LH , the lateral horn ) . Excitatory projection neuron ( ePN , dark red ) and inhibitory projection neuron ( iPN , teal ) axons are indicated . Magnification: the lateral horn region with the regions innervated by excitatory Mz19-positive projection neuron axons ( ePNs , dark red ) and inhibitory Mz699-positive projection neuron axons ( iPNs , teal ) examined here highlighted . ( B–C ) Representative high magnification confocal maximum intensity projections of Mz19-GAL4 positive PN axon terminals in the lateral horn in males expressing Brp-Short-mStraw and mCD8-GFP and stained for antibodies against mStraw ( red ) , GFP ( green ) , and N-Cadherin ( blue ) . Loss of lrp4 ( B , lrp4dalek ) reduces synapse number compared to control ( A ) . ( D–E ) Representative high magnification confocal maximum intensity projections of Mz699-GAL4 positive inhibitory projection neuron ( iPN ) axon terminals in the lateral horn in males expressing Brp-Short-mStraw and mCD8-GFP and stained for antibodies against mStraw ( red ) , GFP ( green ) , and N-Cadherin ( blue ) . Loss of lrp4 ( E , lrp4dalek ) does not affect synapse number compared to control ( D ) . ( F ) Quantification of Brp-Short-mStraw puncta ( red , left axis ) and neurite volume ( black , right axis ) in Mz19-positive excitatory projection neurons . Loss of lrp4 and RNAi against lrp4 expressed in those neurons reduces puncta number but leaves neurite volume unaffected . The similar reduction in puncta number between mutants and PN-specific RNAi reveals the cell autonomous nature of the lrp4 phenotype . ( G ) Quantification of Brp-Short-mStraw puncta ( red , left axis ) and neurite volume ( black , right axis ) in Mz699-positive inhibitory projection neurons . Neither loss of lrp4 nor lrp4 RNAi expressed in those neurons affects puncta number , similar to inhibitory neurons in the antennal lobe . Loss of lrp4 reduces neurite volume by 11% but RNAi does not . Overexpression of LRP4 in these neurons ( LRP4 OE ) results in a 28% increase in the number of Brp-Short puncta . ****p<0 . 0001; **p<0 . 01; ns , not significant . Statistical comparisons ( one way ANOVA with correction for multiple comparisons ) are with control . Error bars represent mean ± s . e . m . n ( lateral horns ) is noted at the bottom of each column . Scale bars = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 01410 . 7554/eLife . 27347 . 015Figure 5—figure supplement 1 . Representative lateral horn images for LRP4 genetic manipulations . ( A–C ) Representative high magnification confocal maximum intensity projections of Mz19-GAL4 positive PN axon terminals in the lateral horn in males expressing Brp-Short-mStraw and mCD8-GFP and stained for antibodies against mStraw ( red ) , GFP ( green ) , and N-Cadherin ( blue ) . Presynaptic knockdown of lrp4 ( B , lrp4IR-1; C , lrp4IR-2 ) results in fewer Brp-Short puncta compared to Control ( A ) but does not affect mCD8-GFP staining . ( D–G ) Representative high magnification confocal stack images of Mz699-GAL4 positive inhibitory projection neuron ( iPN ) axon terminals in the lateral horn in males expressing Brp-Short-mStraw and mCD8-GFP and stained for antibodies against mStraw ( red ) , GFP ( green ) , and N-Cadherin ( blue ) . Presynaptic RNAi against lrp4 ( E , lrp4IR-1; F , lrp4IR-2 ) has no effect on Brp-Short puncta or neurite volume . Overexpression of LRP4 in these neurons ( G , LRP4 OE ) , however , increases synaptic puncta without changing neurite volume . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 015 In lrp4 mutants , the number of excitatory lateral horn synapses was reduced by 40% , consistent with a role for LRP4 in synapse formation ( Figure 5B–C , F ) . PN perturbation of lrp4 using RNAi reduced synapse number similarly to the loss-of-function allele , demonstrating a presynaptic role for lrp4 in these neurons ( Figure 5F and Figure 5—figure supplement 1 ) . These changes were independent of neurite volume , which remained unaffected ( Figure 5F ) . Perturbation of lrp4 in Mz699-positive iPNs , however , had no effect on the number of synapses ( Figure 5D–E , G , and Figure 5—figure supplement 1 ) despite a slight reduction in neurite volume in lrp4dalek mutants ( Figure 5G ) . Despite a lack of a loss-of-function phenotype , we observed an increase in synapse number when we overexpressed LRP4-HA in Mz699-positive neurons ( Figure 5G and Figure 5—figure supplement 1 ) . Thus , the results of lrp4 perturbation on excitatory and inhibitory synapses in the lateral horn resembled those of the antennal lobe , suggesting a general role for LRP4 in promoting excitatory synapse number . Given the role for LRP4 in the specific regulation of excitatory synapse number , we sought to determine whether the consequences of LRP4 disruption were accompanied by functional changes in behavior . We examined fly attraction to the odorant in apple cider vinegar using a modified olfactory trap assay ( Larsson et al . , 2004; Potter et al . , 2010 ) ( Figure 6A ) , an ethologically relevant assay that requires flight and/or climbing to follow odorant information within a larger arena ( Min et al . , 2013 ) . As presynaptic LRP4 regulates ORN synapse number , we used RNAi against lrp4 expressed selectively in all ORNs using pebbled-GAL4 to assess olfactory attraction . Control flies bearing a single copy of pebbled-GAL4 or one of four different lrp4 RNAi transgenes alone exhibited a strong preference for apple cider vinegar ( Figure 6B ) . Flies bearing both transgenes ( and thus , reduced lrp4 expression ) exhibited a near complete abrogation of attractive behavior and were no longer able to distinguish the attractive apple cider vinegar from a water control ( Figure 6B ) . Movement , wall climbing , and flight were still observed in these flies ( data not shown ) , suggesting that this was not due to widespread defects in motion , consistent with our selective perturbation of LRP4 function in ORNs . Thus , presynaptic LRP4 in ORNs is necessary for normal olfactory attraction behavior . 10 . 7554/eLife . 27347 . 016Figure 6 . Loss of presynaptic LRP4 abolishes olfactory attraction behavior . ( A ) Cartoon of the olfactory trap . ( B ) Quantification of preference index [ ( # of flies in odor vial – # of flies in control vial ) / total # of flies] between apple cider vinegar ( odor ) and water ( ctrl ) . Genotypes are indicated below . Control flies with only a GAL4 or UAS-RNAi transgene demonstrate high preference for the attractive odorant in apple cider vinegar . Flies expressing lrp4 RNAi in ORNs have this attraction abrogated . Flies expressing RNAi against GABABR2 in ORNs still display robust attractive behavior while concurrent expression with lrp4 knockdown largely suppresses the loss of attractive behavior . To ensure an equivalent number of transgenes in each genotype , UAS-mCD8-GFP was included ( not listed ) to control for potential transgenic dilution . ****p<0 . 0001; **p<0 . 01; *p<0 . 05; ns , not significant . Statistical comparisons ( one-way ANOVA with correction for multiple comparisons ) are with control unless otherwise noted . Error bars represent mean ± s . e . m . n ( cohorts of 25 flies tested ) is noted at the bottom of each column . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 016 A complete loss of olfactory attraction was unexpected for a manipulation that reduced synapse number by ~30% . One potential explanation is that , while the remaining 70% of synapses were detected by the Brp-Short assay , they were functionally impaired . This would be consistent with the myriad of morphology defects observed in lrp4 mutant T-bars via TEM ( Figure 3I–Q ) . In Drosophila , olfactory information flow is regulated by presynaptic inhibition by local GABAergic interneurons onto excitatory ORNs via the GABAA and GABABR2 receptors ( Olsen and Wilson , 2008; Root et al . , 2008 ) . If the remaining synapses were indeed weakened by the loss of LRP4 , reducing inhibition onto those ORNs might suppress the behavioral phenotype . To test this hypothesis , we inhibited the GABABR2 receptor in ORNs using RNAi , which by itself did not affect the olfactory attraction behavior ( Figure 6B ) . Simultaneous knockdown of GABABR2 and lrp4 , however , markedly suppressed the behavioral phenotype associated with lrp4 knockdown alone ( Figure 6B ) . This manipulation did not suppress the morphological phenotype , however , as the reduction in Brp-Short puncta was still apparent ( 1297 ± 25 . 62 puncta , n = 39 antennal lobes for Or47b-GAL4 > UAS-lrp4IR2 + UAS-mCD8-GFP vs . 1191 ± 48 . 91 puncta , n = 12 antennal lobes for Or47b-GAL4 > UAS-lrp4IR2 + UAS-GABABR2IR , p>0 . 2 ) . These results suggest that olfactory attraction behavior requires a proper level of net excitatory drive in the antennal lobe circuit and that defects caused by weakened excitatory synapses can be compensated for by reducing inhibition . To understand how LRP4 could regulate excitatory synapse number and olfactory behavior , we investigated the mechanism by which it functions . In examining lrp4dalek mutant larvae and larvae where lrp4 was specifically knocked down in all neurons using RNAi , we observed impaired localization of active zone material ( Figure 7A–C ) . Under normal circumstances , the active zone marker Bruchpilot ( Wagh et al . , 2006 ) and the synaptic vesicle marker Synaptotagmin I ( DiAntonio et al . , 1993 ) were barely detectable in larval transverse nerves ( Figure 7A ) , due to their proper trafficking to or maintenance at synaptic sites . However , in lrp4dalek mutants , Bruchpilot improperly accumulated in the transverse nerves ( Figure 7B ) . This kind of accumulation is rarely observed in wild type , but is also most notably associated with loss of SRPK79D ( Figure 7C ) , a conserved serine-arginine protein kinase that localizes to NMJ terminals and negatively regulates premature active zone assembly before Bruchpilot reaches the fly NMJ ( Johnson et al . , 2009; Nieratschker et al . , 2009 ) . In both lrp4 and srpk79D mutants , Brp accumulation was not accompanied by focal accumulations of Synaptotagmin I , indicating that axonal trafficking is not generally impaired ( Figure 7A–C ) ( Gindhart et al . , 1998; Johnson et al . , 2009; Nieratschker et al . , 2009 ) . Because of the similarity in the transverse nerve phenotypes and the role of SRPK79D at peripheral synapses , we hypothesized that LRP4 and SRPK79D could operate together in the CNS to regulate synapse number . 10 . 7554/eLife . 27347 . 017Figure 7 . LRP4 is required for normal synaptic SRPK79D localization in the CNS . ( A–C ) Representative images of larval transverse nerves stained with antibodies to Bruchpilot ( Brp , green ) , Synaptotagmin I ( Syt I , red ) , and HRP ( blue ) . Loss of lrp4 ( B , lrp4dalek ) and srpk79d ( C , srpkatc ) result in improper axonal accumulations of Brp . This is not a general trafficking defect , as Syt I is absent from focal accumulations . ( D ) Representative high magnification confocal slice of VA1v ORNs expressing Brp-Short-mStraw and venus-SRPK79D and stained with antibodies to mStraw ( red ) , GFP ( green ) , and N-Cadherin ( blue ) . SRPK79D largely colocalized with Brp-Short-mStraw but Brp-Short-positive / SRPK79D-negative and Brp-Short-negative / SRPK79D-positive puncta were also observed ( D’’ ) . ( E ) Representative confocal slice within a single antennal lobe glomerulus of a brain expressing venus-SRPK79D and LRP4-HA in all ORNs , processed for proExM , and stained with antibodies to venus ( green ) , HA ( red ) , and N-Cadherin ( blue ) . Distinct regions of overlap between venus-SRPK79D and LRP4-HA ( E” ) are observed , though this represents a subset of venus-SRPK79D localization . ( F–G ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals expressing venus-SRPK79D in control ( F ) and lrp4dalek ( G ) backgrounds and stained with antibodies to GFP ( green ) and N-Cadherin ( blue , inset ) . Loss of lrp4 results in reduced synaptic SRPK79D . ( H ) Quantification of venus-SRPK79D ( green , left axis ) and N-Cadherin fluorescence ( blue , right axis ) . SRPK79D fluorescence is markedly reduced in lrp4dalek animals , but N-Cadherin staining is unaffected , demonstrating specificity . ( I–J ) Representative high magnification single confocal slices of the antennal lobe where all ORNs are expressing venus-SRPK79D and LRP4-HA via the pebbled-GAL4 driver and the brains subsequently processed using proximity ligation assays to determine whether the two proteins were close enough to interact . The brains were stained with antibodies to venus ( green ) and HA ( blue ) and PLA-specific probes ( red ) to detect proximity ligation events . When PLA-specific probes are not added , no signal is observed ( I” ) but when present , positive PLA signal ( J” ) indicates close physical proximity between LRP4-HA and venus-SRPK79D . Positive PLA signal represents a subset of SRPK79D or LRP4 expression , as in ( E ) . **p<0 . 01; ns , not significant . Statistical comparisons ( one-way ANOVA with correction for multiple comparisons ) are with control unless otherwise noted . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bars = 10 µm ( A–D , I–J ) , 25 µm ( E ) , 20 µm ( F–G ) , 33 µm ( F-G insets ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 01710 . 7554/eLife . 27347 . 018Figure 7—figure supplement 1 . Proximity ligation assays reveal LRP4 and SRPK79D interactions . ( A–D ) Representative single confocal slices of antennal lobes expressing venus-SRPK79D alone ( A ) , LRP4-HA alone ( B ) or venus-SRPK79D and LRP4-HA ( C–D ) in all ORNs using the pebbled-GAL4 driver and stained with antibodies against venus ( green ) , HA ( blue ) , and processed using proximity ligation assays ( red ) . When either protein is expressed alone ( a–b ) , no PLA signal is observed ( A” , B” ) . When both are present , however , PLA signal can be observed ( C” , D” ) suggesting that the two are close enough physically to interact . High magnification of a single glomerulus ( D” , the dashed boundary in C ) indicates that the PLA-positive signal represents a subset of LRP4 or SRPK79D localization . Scale bars = 20 µm ( A–B ) , 10 µm ( C ) , 5 µm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 018 As SRPK79D antibodies are not available , we utilized Venus-tagged SRPK79D transgenes to examine CNS localization . When expressed only in VA1v ORNs , venus-SRPK79D localized to axon terminals and overlapped with Brp-Short , demonstrating localization with and adjacent to CNS active zones ( Figure 7D ) . This was reminiscent of LRP4-HA localization in ORNs ( Figure 1F ) so we turned to proExM to more precisely assess the spatial relationship between SRPK79D and LRP4 . In the ORNs of expanded individual glomeruli , LRP4-HA and venus-SRPK79D exhibited coincident and adjacent localization ( Figure 7E ) . However , SRPK79D was expressed more broadly throughout ORNs , suggesting that only a subset of SRPK79D colocalizes with LRP4 . This may indicate both LRP4-dependent and -independent roles for SRPK79D . We also examined this synaptic localization in lrp4dalek mutants: loss of lrp4 reduced synaptic SRPK79D levels by ~50% ( Figure 7F–H ) . This reduction was specific for SRPK79D , as the staining for other markers , like the general neuropil label N-Cadherin , was unaffected ( Figure 7F–H ) . These results demonstrate that LRP4 is necessary for the proper localization and / or expression of SRPK79D and suggest that SRPK79D might act downstream of LRP4 to regulate synapse number . Due to their spatial proximity , we next employed proximity ligation assays ( PLA ) to determine whether LRP4 and SRPK79D are spatially close enough to interact . PLA uses oligonucleotides conjugated to secondary antibodies ( Greenwood et al . , 2015; Söderberg et al . , 2006 ) : if the epitopes are sufficiently close ( 30–40 nm ) , the oligonucleotides can be ligated together and detected using a fluorescent probe . The result can be observed using confocal microscopy and preserve , to a high degree , the spatial localization of the proteins involved . PLA has been used to examine protein-protein interactions at the NMJ ( Wang et al . , 2015 ) but not , to our knowledge , in the CNS . To examine this , we co-expressed venus-SRPK79D and LRP4-HA in all ORNs using pebbled-GAL4 ( Sweeney et al . , 2007 ) , stained both targets with oligonucleotide-conjugated secondary antibodies and performed PLA assays ( Figure 7I–J and Figure 7—figure supplement 1 ) . As expected , both proteins localize to the axon terminals of ORNs . When either is expressed singularly ( Figure 7—figure supplement 1A–B ) or the probes are not added ( Figure 7I ) , no PLA signal is observed . However , in the presence of both transgenes and the appropriate probes ( Figure 7J and Figure 7—figure supplement 1C–D ) , we detected positive signal indicating that the proteins were close enough to interact . The PLA signal represented a subset of LRP4 or SRPK79D staining patterns , suggesting that there are roles independent of the other for each protein . Taken together , this data suggests that LRP4 interacts with SRPK79D to maintain SRPK79D localization at the synapse . The interaction with , and reliance on LRP4 for synaptic SRPK79D localization suggested that the two function together . If so , we would expect that the two would display phenotypic similarity and interact in the same genetic pathway . We observed phenotypic similarity in larval nerves ( Figure 7A–C ) , but we further sought to study this at CNS synapses . To test the interactions between LRP4 and SRPK79D with respect to effects of synapse number , we conducted loss-of-function , genetic interaction , and genetic epistasis experiments between genetic perturbations of both . First , reducing srpk79D function presynaptically using an established RNAi ( Johnson et al . , 2009 ) expressed in VA1v ORNs resulted in a 15% reduction in the number of Brp-Short puncta compared to control ( Figure 8A–B , E ) . Thus , SRPK79D is required for normal CNS synapse number . We further sought to understand if LRP4 and SRPK79D interacted genetically . To examine this , we performed a transheterozygote genetic interaction assay . When single copies of either lrp4 or srpk79D were removed , there was no evident phenotype ( Figure 8—figure supplement 1B–C , E ) . However , when one copy of each was concurrently removed , we observed a significant reduction in Brp-Short puncta ( Figure 8—figure supplement 1D–E ) . This suggests that the two function in the same genetic pathway and may work together to ensure proper synapse number . Given the reduction in synaptic SRPK79D present in lrp4 mutants , we examined whether these reduced SRPK79D levels are the root cause of its synapse reduction . We overexpressed SRPK79D in presynaptic ORNs either in control or lrp4dalek mutant backgrounds . Presynaptic overexpression of SRPK79D in VA1v ORNs partially suppressed the synaptic phenotype associated with the lrp4dalek mutation , resulting in 92% of the normal number of synapses ( Figure 8A , C–E ) , whereas overexpression of SRPK79D in a wild-type background had no effect ( Figure 8E ) . Finally , we sought to determine whether srpk79D was required for the increase in Brp-Short puncta associated with LRP4 overexpression ( Figure 2H ) . When LRP4 was overexpressed concurrently with srpk79D RNAi , the phenotype resembled that of srpk79D RNAi alone ( Figure 8E ) . This suggests that LRP4 requires SRPK79D to mediate its overexpression phenotype , likely by functioning through SRPK79D to increase the number of synapses . Combined , these indicate that LRP4 and SRPK79D closely interact presynaptically in the same genetic pathway to ensure the proper number of excitatory synapses . 10 . 7554/eLife . 27347 . 019Figure 8 . SRPK79D and LRP4 genetically interact to control synapse morphology and function . ( A–D ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals in males expressing Brp-Short-mStraw and stained with antibodies to mStraw ( red ) and N-Cadherin ( blue ) . Presynaptic RNAi against srpk79D ( srpk RNAi ) reduces the number of puncta , but less so than loss of lrp4 ( lrp4dalek ) . Presynaptic overexpression of SRPK79D in an lrp4dalek background ( lrp4dalek + SRPK ) restores puncta number to control levels . ( E ) Quantification of Brp-Short-mStraw puncta . Note that overexpression of SRPK79D in an otherwise wild-type background has no gain-of-function effects on puncta number . Further srpk79D function is needed to enable the LRP4 overexpression-induced increase in synaptic puncta number . n ( antennal lobes ) is noted at the bottom of each column . ( F ) Quantification of preference index in the olfactory trap assay . Flies overexpressing SRPK79D in ORNs show strong attractive behavior , while ORNs expressing RNAi against lrp4 or srpk79D abrogate attraction to apple cider vinegar . This phenotype can be suppressed by concurrent overexpression of SRPK79D . UAS-mCD8-GFP ( not listed ) was used to ensure equivalent numbers of transgenes in each genotype . n ( cohorts of 25 flies tested ) is noted at the bottom of each column . ( G ) A model for LRP4 function at olfactory synapses . At wild-type axon terminals , LRP4 in presynaptic ORNs ( orange ) interacts with a putative postsynaptic partner ( blue ) , resulting in SRPK79D ( beige ) retention at the terminal and a full complement of active zones ( black T ) . Here , the putative ligand is depicted as having a postsynaptic PN source , but alternate sources ( such as glia or local interneurons ) are also possible . In the absence of LRP4 , less synaptic SRPK79D is present and active zone number is reduced . The size of the terminal itself does not change but the synapse number ( i . e . , number of active zones ) within that terminal space is reduced . Further , T-bar defects like a floating T-bar can also be seen . SRPK79D overexpression in an lrp4 mutant restores synaptic SRPK79D and active zone number , despite the absence of LRP4 . Thus , the LRP4 largely functions in synaptic organization through downstream SRPK79D . ****p<0 . 0001; ***p<0 . 001; ns , not significant . Statistical comparisons ( one-way ANOVA with correction for multiple comparisons ) are with control unless otherwise noted . Error bars represent mean ± s . e . m . n ( antennal lobes for E , cohorts of 25 flies tested for F ) is noted at the bottom of each column . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 01910 . 7554/eLife . 27347 . 020Figure 8—figure supplement 1 . lrp4 and srpk79D interact genetically to control Brp-Short puncta number . ( A–D ) Representative high magnification confocal maximum intensity projections of VA1v ORN axon terminals in the VA1v glomerulus of males expressing Brp-Short-mStraw and stained with antibodies against mStraw ( red ) and N-Cadherin ( blue ) . ( E ) Quantification of Brp-Short-mStraw puncta . Loss of one copy of either lrp4 ( B ) or srpk79D ( C ) does not affect puncta number , but concurrent loss of one copy of each gene ( D ) results in fewer puncta . *p<0 . 05 . Statistical comparisons ( one-way ANOVA corrected for multiple comparisons ) are with control unless noted . Error bars represent mean ± s . e . m . n ( antennal lobes ) is noted at the bottom of each column . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 27347 . 020 In light of the synapse number defects , we also examined the functional consequences of srpk79D perturbation on olfactory behavior . Flies expressing srpk79D RNAi in all ORNs demonstrated a nearly complete abrogation of attraction behavior ( Figure 8F ) that was indistinguishable from the lrp4 RNAi phenotype . In light of the suppression of the synapse number phenotype , we also examined whether SRPK79D overexpression could suppress the lrp4 loss-of-function behavioral phenotype . Control flies bearing the pan-ORN pebbled-GAL4 or the SRPK79D overexpression transgene alone exhibited strong attraction towards apple cider vinegar ( Figure 8F ) . Further , SRPK79D overexpression in all ORNs did not affect this robust attraction . Driving both SRPK79D overexpression and lrp4 RNAi in all ORNs , however , resulted in a partial suppression of the behavioral phenotype associated with lrp4 RNAi ( Figure 8F ) . As the synaptic level of SRPK79D is positively regulated by LRP4 and SRPK79D overexpression suppresses the morphological and functional phenotypes associated with lrp4 loss-of-function , SRPK79D is likely a key downstream effector of LRP4 in regulating synapse number and thus , normal olfactory attraction behavior .
Coordination of excitation and inhibition is critical to proper circuit function . Imbalances in excitation and inhibition lead to epileptic states ( Badawy et al . , 2012 ) and social dysfunction ( Yizhar et al . , 2011 ) , and may also underlie many autism spectrum disorders ( Mullins et al . , 2016; Nelson and Valakh , 2015 ) . The mechanisms that maintain this balance are incompletely understood , though likely involve multiple aspects including the number of each type of neuron , their firing rates , release probabilities , synaptic strength , and neurotransmitter receptor sensitivities . Such regulation likely requires distinguishing excitatory from inhibitory neurons at both pre- and postsynaptic levels . Excitatory and inhibitory synapses are identified postsynaptically by distinct neurotransmitter receptor , scaffolding protein , and adhesion molecule repertoires ( Craig and Kang , 2007; Sheng and Kim , 2011; Ziv and Fisher-Lavie , 2014 ) . Postsynaptic factors like Neuroligin 2 ( Graf et al . , 2004 ) , Gephyrin ( Choii and Ko , 2015 ) , and Slitrk3 ( Takahashi et al . , 2012 ) organize inhibitory GABAergic synapses while LRRTMs organize excitatory synapses ( Siddiqui et al . , 2013; de Wit et al . , 2009 , 2013 ) . Thus , postsynaptic regulation can occur by differential modulation of these factors . Little is known , however , about the presynaptic identifiers of excitatory versus inhibitory neurons . Recent work identified Punctin / MADD-4 as a determinant of excitatory versus inhibitory neuromuscular synapses in C . elegans , though as a secreted factor that functions via postsynaptic interaction ( Maro et al . , 2015; Pinan-Lucarré et al . , 2014; Tu et al . , 2015 ) . Further , Glypican4 can localize to excitatory presynaptic terminals and interact with LRRTM4 ( de Wit et al . , 2013 ) but its synaptogenic activity is also provided by astrocytes ( Allen et al . , 2012 ) and thus is not neuronal specific . Proteomic comparisons ( Biesemann et al . , 2014; Boyken et al . , 2013 ) suggest few differences beyond those pertaining to neurotransmitter synthesis enzymes and transporters . But these components may not be sufficient to distinguish presynaptic excitatory from inhibitory neurons . In the Drosophila olfactory system , for example , glutamate can be inhibitory when its postsynaptic partners express glutamate-gated chloride channels ( Liu and Wilson , 2013 ) . This suggests that pre- and postsynaptic regulators may exist to distinguish excitatory and inhibitory synapses , though it is unclear what those presynaptic regulators might be . Our data suggests that LRP4 may be a candidate presynaptic organizer specific for excitatory connections . LRP4 is expressed in a subset of excitatory cholinergic neurons , excluded from inhibitory GABAergic neurons , and expressed in a subset of glutamatergic neurons that may be excitatory or inhibitory ( Figure 1 ) . Though we cannot rule out inhibitory neuron expression in the case of the glutamatergic subset , the phenotypes associated with LRP4 perturbation are consistent with an excitatory neuron-specific role . Thus , LRP4 may not only serve an identifying role at excitatory synapses , but also a functional one . Loss of lrp4 results in fewer excitatory synapses but has no effect on inhibitory synapses . However , both excitatory and inhibitory neurons show increased synapse number with lrp4 overexpression ( Figures 2 and 4–5 ) . This shared competency suggests that both neurons contain machinery that can be engaged downstream of LRP4 ( or the cell surface ) to add synapses . Thus , proteins like LRP4 may represent identifiers of excitatory or inhibitory terminals that function by engaging common mechanisms to add synapses . At the mouse NMJ , LRP4 is the well-established postsynaptic receptor for motoneuron-derived Agrin ( Kim et al . , 2008; Zhang et al . , 2008 , 2011 ) and regulates synapse formation ( Weatherbee et al . , 2006 ) and maintenance ( Barik et al . , 2014 ) . However , additional roles for LRP4 exist at the level of the presynaptic motoneuron . A retrograde signal composed of LRP4 from the postsynaptic muscle interacts with an unknown receptor in the motoneuron ( Yumoto et al . , 2012 ) to regulate presynaptic differentiation . Thus , at the mouse NMJ , postsynaptic LRP4 has both cell-autonomous and non-cell autonomous roles . In addition , presynaptic LRP4 has been implicated to regulate acetylcholine receptor clustering via MMP-mediated proteolytic cleavage ( Wu et al . , 2012 ) . In the mouse CNS , LRP4 regulates synaptic physiology ( Gomez et al . , 2014; Pohlkamp et al . , 2015 ) , learning and memory , fear conditioning , and CA1 spine density ( Gomez et al . , 2014 ) . Though CNS LRP4 most commonly associates with postsynaptic densities ( Tian et al . , 2006 ) , it also fractionates with synaptophysin-positive membranes ( Gomez et al . , 2014 ) . Indeed , the observed CNS phenotypes have not been localized to a particular pool of LRP4 . Our identification of Drosophila LRP4 as a key player in CNS synaptogenesis , however , posits a cell-autonomous presynaptic role . While we cannot rule out an additional , perhaps concurrent , postsynaptic role , our work is the first to demonstrate clear cell-autonomous presynaptic functions for LRP4 . Indeed , LRP4 is expressed in PNs and may localize to PN dendrites within the antennal lobe ( Figure 1—figure supplement 2 ) . In such a case , it could function either presynaptically , at dendrodendritic presynapses ( Rybak et al . , 2016; Tobin et al . , 2017 ) or as a postsynaptic factor . Moreover , as the Drosophila genome lacks clear Agrin and MuSK homologs , this suggests a synaptic function of LRP4 that evolutionarily precedes Agrin and MuSK recruitment to vertebrate NMJ synaptogenesis . It remains open whether this presynaptic function is conserved in the mammalian CNS and , if so , what signal LRP4 receives . In Drosophila , the signal cannot be Agrin and in the mammalian CNS , Agrin is not essential for CNS synapse formation ( Daniels , 2012 ) . Thus , the Agrin-independence of CNS LRP4 may be conserved across systems . Moreover , our finding that LRP4 promotes excitatory , but not inhibitory , synapse formation and function is consistent with reduced excitatory but normal inhibitory input in hippocampal CA1 neurons of lrp4 mutant mice ( Gomez et al . , 2014 ) . Moreover , we find that LRP4 in the Drosophila CNS functions through the SR-protein kinase SRPK79D . Impaired srpk79D function reduces synapse number and overexpression can suppress the functional and morphological defects associated with lrp4 loss ( Figures 7–8 ) . This kinase is evolutionarily conserved ( Johnson et al . , 2009 ) and the three mammalian homologues ( Zhou and Fu , 2013 ) are widely expressed in the mouse brain ( Lein et al . , 2007 ) , including in the hippocampus . From yeast to human , SRPKs regulate spliceosome assembly and gene expression ( Zhou and Fu , 2013 ) but have not been studied in mammalian synapse formation . It will be interesting to test if these kinases also function in the mammalian CNS . Combined , however , these commonalities suggest a basic conservation between invertebrate and vertebrate systems for future study . Recent work implicated LRP4 in both amyotrophic lateral sclerosis ( ALS ) and myasthenia gravis ( MG ) , two debilitating motor disorders with a worldwide prevalence of ~1/5000 . Distinct ALS and MG populations are seropositive for LRP4 autoantibodies ( Tsivgoulis et al . , 2014; Tzartos et al . , 2014 ) and double seronegative for Agrin or MuSK , suggesting that seropositivity is not a byproduct of generalized NMJ breakdown . Further , injection of LRP4 function-blocking antibodies into mice recapitulates MG ( Shen et al . , 2013 ) . Beyond peripheral symptoms , cognitive impairment ( besides that as frontotemporal dementia ) also occurs in a subset of ALS patients ( Ringholz et al . , 2005 ) . Thus , understanding the roles of LRP4 in the peripheral and central nervous systems has marked clinical significance . Our identification of an evolutionarily conserved kinase , SRPK79D , as a downstream target of LRP4 signaling may offer a window into those roles . As SRPK79D overexpression suppresses the behavioral and the synaptic phenotypes of lrp4 loss ( Figure 8 ) , if it functions similarly in the mammalian CNS , SRPKs could be a target for therapeutics . Further investigation of how LRP4 functions in the CNS will provide new insight not only into the cognitive aspects of these debilitating motor disorders , but also into the fundamental aspects of excitatory synapse formation .
The lrp4 mutation was designed following published methods ( Gratz et al . , 2013 ) . Two lrp4-specific chimeric RNAs ( chiRNA ) were cloned into the pU6-BbsI-chiRNA vector as follows - A1 , corresponding to an optimal PAM site 2 bp 5’ of the start ATG ( using primers: 5’ CTTCGGCGAGTTTGTGTACATGTC 3’ and 5’ AAACGACATGTACACAAACTCGCC 3’ with a phosphate at the 5’ end ) and A2 , corresponding to an optimal PAM site 34 bp 3’ of the TAG stop codon ( using primers 5’ CTTCGAATCGGTAAATGGTTTCAG 3’ and 5’ AAACCTGAAACCATTTACCGATTC 3’ ) . Both the A1 and A2 chiRNA plasmids ( 250 ng / μL ) and a pHsp70-Cas9 plasmid ( 500 ng / μL ) were injected into MB03015 embryos ( stock BL23835 ) to produce lrp4 deletions . MB03015 flies bear a Minos-based Mi{ET1} insertion ( Bellen et al . , 2011 ) between exons 5 and 6 of the lrp4 open reading frame; adults with the insertion are marked by expression of a GFP reporter in the eye . Successful events were screened for by the loss of GFP: as the PAM sites were distant from and flanking the insertion , loss of fluorescence likely indicated removal of the intervening sequences ( the lrp4 coding region ) . Five such lines ( representing identical events ) were recovered and homozygous viable stocks established: the allele was named dalek due to the ‘extermination’ of the lrp4 gene , and in homage to the classic villains of ‘Doctor Who’ . Loss of lrp4 was assessed using genomic DNA prepared from control and lrp4dalek adults using the QIAgen DNeasy Blood and Tissue Kit ( QIAgen , Valencia , CA ) . Genomic PCR bands corresponding to exon 2 ( 534 bp using primers 5’ TGTATTCCACGAACCTGGGTATG 3’ and 5’ CAAAATGCAGCGCCCATTGTT 3’ ) and the exon 7–8 junction ( 615 bp using primers 5’ AGTCTTGATGGTAGCAATAGGCAT 3’ and 5’ CTCTGGTAGATTTTGACACTG 3’ ) revealed the absence of both regions in lrp4dalek . The lrp4dalek deletion was further confirmed by the presence of a 315 bp ‘Flank’ band ( with some background bands present only with the lrp4dalek deletion ) representing the connection of sequences from the 5’ and 3’ UTRs ( amplified by primers 5’ AACAGAATCGGAACAGCAGTT 3’ and 5’ GAGCTTTAACAGGACACGTTT 3’ ) not present in control samples ( see Figure 1—figure supplement 2B ) . Finally , antibody staining ( see below ) revealed the elimination of LRP4 signal in the lrp4dalek allele , suggesting the creation of a null allele . An adult Drosophila cDNA library was made according to manufacturer’s protocol using the GeneRacer Kit ( ThermoFisher Scientific , Catalog #L150201 , Waltham , MA ) . From the library , the lrp4 cDNA was amplified using the forward primer 5’ CACCATGTATTTGACAGCCTTT 3’ and the reverse primer 5’ TGTGATAGTCGAGAGCGT 3’ ( without the endogenous Stop codon ) and cloned directly into the pENTR vector using the pENTR/D-TOPO Cloning Kit ( ThermoFisher Scientific , Catalog #K240020 , Waltham , MA ) . Complete cDNA clones were verified by sequencing . UAS-LRP4-HA was made by recombining pENTR-LRP4 with pUAST-attB-Gateway-3xFLAG-3xHA29 via LR clonase . The resultant pUAST-attB-LRP4-3xHA-3xFLAG was transformed into the ΦC31 landing site 86Fb on the 3rd chromosome using standard methods . Custom antibodies were made by Pierce Custom Services ( ThermoFisher , Rockford , IL ) against the C-NKRNSRGSSRSVLTFSNPN peptide corresponding to residues 1921–1939 of the intracellular side of LRP4 . Rat antisera were Ig-purified and then used at a dilution of 1:200 on adult brains . The specificity of the antibody was verified by the absence of signal in the lrp4dalek mutant . The Drosophila melanogaster ( CG8909; accession AAF48538 . 1 ) , Mus musculus ( accession NP_766256 . 3 ) , and Homo sapiens ( accession NP_002325 . 2 ) LRP4 sequences were obtained from NCBI . CLUSTALW alignment was performed using PSI/T-Coffee for transmembrane proteins ( http://tcoffee . crg . cat/apps/tcoffee/do:tmcoffee ) and expressed graphically using ESPript3 . 0 ( http://espript . ibcp . fr/ESPript/ESPript/ ) . All controls , stocks , and crosses were raised at 25°C . Mutants and transgenes were maintained over balancer chromosomes to enable selection in adult or larval stages . The GMR90B08-GAL4 ( Pfeiffer et al . , 2008 ) line was used to examine lrp4 expression ( referred to as lrp4-GAL4 ) . Four UAS-RNAi lines against differing regions of lrp4 were also identified: UAS-lrp4-RNAi 1 ( v29900 , Vienna Drosophila Resource Center ) , UAS-lrp4-RNAi 2 ( v108629 , Vienna Drosophila Resource Center ) , UAS-lrp4-RNAi 3 ( JF01570 , Harvard TRiP Collection ) , UAS-lrp4-RNAi 4 ( JF01632 , Harvard TRiP Collection ) . The following GAL4 lines enabled tissue-specific expression: Or47b-GAL4 ( VA1v ORNs ) ( Vosshall et al . , 2000 ) , Or67d-GAL4 ( DA1 ORNs ) ( Kurtovic et al . , 2007 ) , Or88a-GAL4 ( VA1d ORNs ) ( Vosshall et al . , 2000 ) , AM29-GAL4 ( DL4 and DM6 ORNs ) ( Endo et al . , 2007 ) , Mz19-GAL4 ( DA1 , VA1d , DC3 PNs ) ( Jefferis et al . , 2004 ) , Mz699-GAL4 ( inhibitory projection neurons that project to the lateral horn ) ( Lai et al . , 2008; Liang et al . , 2013 ) , GAD1-GAL4 ( GABAergic inhibitory neurons ) ( Ng et al . , 2002 ) , pebbled-GAL4 ( all ORNs ) ( Sweeney et al . , 2007 ) . The following UAS transgenic lines were used as either reporters or to alter gene function: UAS-Syt-HA ( Robinson et al . , 2002 ) , UAS-Brp-Short-mStraw ( Fouquet et al . , 2009 ) , UAS-DSyd1-GFP ( Owald et al . , 2010 ) , UAS-Dα7-GFP ( Leiss et al . , 2009 ) , UAS-mCD8-GFP ( Lee and Luo , 1999 ) , UAS-3xHA-mtdT ( Potter et al . , 2010 ) , UAS-FRT-Stop-FRT-mCD8-GFP ( Hong et al . , 2009 ) , UAS-Dcr2 ( Dietzl et al . , 2007 ) , UAS-GABABR2-RNAi ( Root et al . , 2008 ) , UAS-srpk79D-RNAi ( Johnson et al . , 2009 ) , UAS-venus-SRPK79D-#28 ( Johnson et al . , 2009 ) , UAS-venus-SRPK79D-#1A ( Johnson et al . , 2009 ) . Intersectional analyses were done using the eyFLP3 . 5 construct ( Chotard et al . , 2005 ) which expresses FLP in ORNs , but not PNs and GH146-FLP ( Hong et al . , 2009 ) , which expresses in 2/3 of all olfactory PNs but not ORNs . The srpk79Datc allele ( Johnson et al . , 2009 ) was used to remove srpk79D function . Adult brains were dissected at 10 days post eclosion as previously described ( Mosca and Luo , 2014; Wu and Luo , 2006 ) . Third instar larvae were dissected as previously described ( Mosca and Schwarz , 2010 ) . The following primary antibodies were used: mouse antibody to Bruchpilot ( 1:40 , DSHB , Catalog #mAbnc82 , Iowa City , IA ) ( Laissue et al . , 1999 ) , rabbit antibody to Synaptotagmin I ( 1:4000 ) ( Mackler et al . , 2002 ) , rat antibody to N-Cadherin ( 1:40 , DSHB , Catalog #mAbDN-EX #8 , Iowa City , IA ) ( Iwai et al . , 1997 ) , rat antibody to HA ( 1:100 , Roche , Catalog #11867423001 , Basel , Switzerland ) , mouse antibody to choline acetyltransferase ( ChAT ) ( 1:100 , DSHB , Catalog #mAbChAT4B1 , Iowa City , IA ) ( Takagawa and Salvaterra , 1996 ) , mouse antibody to ELAV ( DSHB , mAb9F8A9 , 1:100 ) ( O'Neill et al . , 1994 ) , rabbit antibody to GABA ( 1:200 , Sigma-Aldrich , Catalog #A2052 , St . Louis , MO ) , mouse antibody to Repo ( 1:100 , DSHB , Catalog #mAb8D12 , Iowa City , IA ) ( Alfonso and Jones , 2002 ) , rabbit antibody to vGlut ( 1:500 ) ( Daniels et al . , 2008 ) , rabbit antibody to dsRed ( 1:250 , Clontech , Catalog #632496 , Mountain View , CA ) , chicken antibody to GFP ( 1:1000 , Aves Labs , Catalog #GFP-1020 , Tigard , OR ) , Alexa647-conjugated goat antibody to HRP ( 1:100 , Jackson ImmunoResearch , Catalog #123-605-021 , West Grove , PA ) . Alexa488- , Alexa568- , and Alexa647-conjugated secondary antibodies were used at 1:250 ( ThermoFisher Scientific and Jackson ImmunoResearch , Various Catalog #s ) . CF633-conjugated secondary antibodies were used at 1:250 ( Biotium ) . FITC-conjugated secondary antibodies were used at 1:200 ( Jackson ImmunoResearch , Catalog #703-095-155 , West Grove , PA ) . Brains were processed as described and stained using rabbit anti-GFP antibodies at 1:500 ( ThermoFisher Scientific , Catalog #A-11122 , Waltham , MA ) with FITC-conjugated secondary antibodies and mouse anti-HA antibodies at 1:250 ( Sigma-Aldrich , Catalog #A2095 , St . Louis , MO ) with Alexa647-conjugated secondary antibodies , leaving the red channel open . For PLA , we used the DuoLink Mouse Rabbit in situ PLA kit ( Sigma-Aldrich , Catalog #DUO92101 , St . Louis , MO ) . Following the last wash after secondary antibody incubation , the brains were incubated in the anti-mouse and / or anti-rabbit PLA probes at a 1:5 dilution for 2 hr at 37°C . Brains were then washed thrice for 10’ each with Wash Buffer A , and incubated in Ligation solution ( 1:40 ligase in ligation buffer ) for 1 hr at 37°C . Brains were washed in Wash buffer A for three times at 10’ each and then incubated in Amplification solution ( 1:80 dilution of polymerase in Amplification buffer ) for 2 hr at 37°C . Finally , brains were washed three times for 10’ each in Wash Buffer B , and incubated in SlowFade overnight before mounting . Controls without Probes went through the identical process as those with probes , but with water substituted for the probes themselves in the first PLA step . Brains were imaged as described via confocal microscopy . All images were obtained using a Zeiss LSM510 Meta laser-scanning confocal microscope ( Carl Zeiss , Oberkochen , Germany ) using either a 40 × 1 . 4 NA PlanApo or a 63 × 1 . 4 NA PlanApo lens . Images of synaptic puncta ( Brp-Short-mStraw or Dα7-GFP ) and neurite membrane ( mCD8-GFP , 3xHA-mTDT ) were imaged , processed and quantified as previously described ( Mosca and Luo , 2014 ) with the following adjustments: images of synaptic puncta in the lateral horn ( Mz19-GAL4 , Mz699-GAL4 , Figure 5 , Figure 5—figure supplement 1 ) were imaged at 63X , with an optical zoom of 2 . Mz19 and Mz699 images were processed with a spot size of 0 . 6 µm and neurite volume calculated with a smoothing of 0 . 2 µm and a local contrast of 0 . 5 µm . Images were processed and figures prepared using Adobe Photoshop CS4 and Adobe Illustrator CS4 ( Adobe Systems , San Jose , CA ) . For antibody staining comparisons between genotypes , samples were imaged and processed under identical conditions . Fluorescence intensity was measured with ImageJ ( NIH , Bethesda , MD ) . Transmission electron microscopy was performed on 10 day old adult control and lrp4dalek male brains as previously described ( Mosca and Luo , 2014 ) . Putative ORN terminals were identified based on morphology ( Rybak et al . , 2016; Tobin et al . , 2017 ) and quantified as described ( Mosca and Luo , 2014 ) . Terminal perimeter was measured using ImageJ ( NIH , Bethesda , MD ) and used to calculate T-bar density . All quantification was done with the user blind to the genotype . Protein retention expansion microscopy ( Tillberg et al . , 2016 ) was modified for use with Drosophila brain tissue . Fixed and antibody-labeled brains were treated with 100 μg / mL acryloyl-X , SE ( ThermoFisher Scientific , Catalog #A20770 , Waltham , MA ) overnight at room temperature and then embedded in polyelectrolyte gel for two hours at 37°C . Slices containing brains were excised from solidified polyelectrolyte gel and immersed in digestion buffer with 200 μg / mL Proteinase K ( ThermoFisher Scientific , Catalog #AM2546 , Waltham , MA ) overnight at room temperature . Slices achieved maximum expansion after five washes with deionized water . Fully expanded gel slices were anchored to the bottom of a petri dish with 2% low melting point agarose . Confocal microscopy images were obtained on a Leica SP8 with a 25x water immersion objective ( Leica Microsystems , Wetzlar Germany ) . Statistical analysis was completed using Prism 6 . 07 ( GraphPad Software , Inc . , La Jolla , CA ) . For representative datasets , the experimenter was blind to genotype during quantification and data analysis . Significance between two samples was determined using student’s t-test . Significance amongst multiple samples was determined using one-way ANOVA with a Tukey’s post-test to correct for multiple comparisons . Significance between two samples ( for EM ) was determined using a two-tailed student’s t-test . Olfactory trap assays were constructed as described ( Potter et al . , 2010 ) . Flies were raised in a 12/12 light/dark incubator . For each cohort , 25 flies of the appropriate genotype were starved overnight in a 1% agar vial in complete darkness . They were anesthetized briefly on ice and transferred to the olfactory trap , which contained an experimental vial of apple cider vinegar ( ACV: Safeway , Palo Alto , CA ) and a control vial of water . Flies were then left in the trap for 16 hr in complete darkness before being quantified . Preference index was calculated as ( FliesACV – FliesWater ) / FliesTotal . See Supplementary file 1 for a listing of complete genotypes by figure panel . | The connections between nerve cells , called synapses , often malfunction in disease , injury and during aging , and to understand how this happens we first need to know how they work normally . At a synapse , one nerve cell sends a signal to the other . The signal is a chemical substance , which binds to specialized proteins called receptors on the receiving nerve cell . At excitatory synapses , the chemical signal activates the receiver; at inhibitory synapses , it does the opposite . Communication at synapses typically only goes in one direction because the sender and receiver at a synapse are not interchangeable; they contain different molecules that support their distinct roles . To complicate matters , the same molecule may sometimes be present on both sides of a synapse with a different role in each . Moreover , not all synapses exist between two nerve cells; some synapses also form between nerve cells and muscle fibers to control the movement of the muscles . Mosca et al . set out to identify new players involved in forming synapses , and to identify differences in the formation of nerve cell-to-nerve cell versus nerve cell-to-muscle connections . Mosca et al . were interested in particular in a protein called LRP4 . In mammals , LRP4 is largely present on the muscle side of nerve cell-to-muscle synapses , where it acts as a receptor for a chemical signal called Agrin . However , fruit flies — which lack Agrin – also possess the gene for LRP4 , suggesting that it has other roles too . Mosca et al . now show that LRP4 is present in the nerve cell-to-nerve cell synapses found in the fruit fly’s brain . Further experiments reveal that fruit fly LRP4 plays an important role on the sender side of these synapses . Reducing the amount of LRP4 in the fruit fly brain reduces the number of excitatory , but not inhibitory , synapses . This suggests that fruit fly LRP4 may help regulate the formation of excitatory synapses . Understanding how synapses form , and the differences between excitatory and inhibitory connections , could provide new insights into disorders of impaired synapse formation such as schizophrenia . LRP4 has also been implicated in disorders , such as amyotrophic lateral sclerosis ( ALS ) and myasthenia gravis , in which impaired communication between nerves and muscles causes muscles to weaken . Improved understanding of how synapses work may lead to better drugs to treat these disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Presynaptic LRP4 promotes synapse number and function of excitatory CNS neurons |
Cerebellar climbing fiber activity encodes performance errors during many motor learning tasks , but the role of these error signals in learning has been controversial . We compared two motor learning paradigms that elicited equally robust putative error signals in the same climbing fibers: learned increases and decreases in the gain of the vestibulo-ocular reflex ( VOR ) . During VOR-increase training , climbing fiber activity on one trial predicted changes in cerebellar output on the next trial , and optogenetic activation of climbing fibers to mimic their encoding of performance errors was sufficient to implant a motor memory . In contrast , during VOR-decrease training , there was no trial-by-trial correlation between climbing fiber activity and changes in cerebellar output , and climbing fiber activation did not induce VOR-decrease learning . Our data suggest that the ability of climbing fibers to induce plasticity can be dynamically gated in vivo , even under conditions where climbing fibers are robustly activated by performance errors .
The cerebellum is thought to implement a supervised learning algorithm , with the climbing fiber input to the cerebellum providing error signals that induce learning . According to this model , performance errors activate neurons in the inferior olive and their climbing fiber axons , which in turn trigger the induction of plasticity in the cerebellar cortex to produce adaptive changes in behavior ( Marr , 1969; Albus , 1971 ) . Previous work has shown that this process is regulated by a feedback loop from the cerebellar cortex to the inferior olive ( Andersson and Armstrong , 1987; Hesslow and Ivarsson , 1994; Medina et al . , 2002; Rasmussen et al . , 2008; Yang and Lisberger , 2013; reviewed in Apps , 1999; Gibson et al . , 2002 ) . In this study , we describe a second level of regulation in animals undergoing learning: even when climbing fibers are robustly activated by performance errors , the ability of that climbing fiber activity to trigger plasticity can be regulated by the state of the cerebellar circuit . Thus , the cerebellum is not a slave to its climbing fiber ‘teachers’ , but rather plays an active role in determining whether it will adapt in response to the error signals it receives from the climbing fibers . The question of whether error signals carried by climbing fibers are the trigger for learning has been controversial , since the evidence from different studies has been inconsistent . Climbing fibers are activated by performance errors during a wide range of motor learning tasks ( for a review , see Ito , 2001 ) , and several lines of evidence suggest that this provides a potent trigger for cerebellar plasticity . A single spike in a climbing fiber reliably triggers a complex spike in its Purkinje cell targets ( Eccles et al . , 1966 ) . Calcium transients associated with complex spikes can induce synaptic plasticity in the cerebellar cortex in vitro ( for reviews , see Hansel et al . , 2001; Ito , 2001 ) . In vivo , climbing fiber activation can replace the unconditioned stimulus used to induce a form of cerebellum-dependent classical conditioning ( Mauk et al . , 1986; Steinmetz et al . , 1989; Jirenhed et al . , 2007; Rasmussen et al . , 2013 ) or the sensory feedback used to induce oculomotor learning ( Nguyen-Vu et al . , 2013 ) . Moreover , during a smooth pursuit oculomotor learning task , there is a tight , trial-by-trial correlation between the occurrence of individual spikes in a climbing fiber and changes in cerebellar output on the subsequent trial , suggesting that climbing fiber spikes provide a potent and reliable trigger for plasticity ( Medina and Lisberger , 2008; Yang and Lisberger , 2010 , 2013 ) . Despite the considerable evidence that climbing fiber activity provides error signals controlling motor learning , several studies have called this idea into question . Recordings during some cerebellum-dependent tasks have revealed a dissociation between the encoding of errors and the induction of learning over the course of a training session ( Catz et al . , 2005; Ke et al . , 2009; Kitazawa et al . , 1998; Ojakangas and Ebner , 1992 ) . In addition , perturbation of the most extensively studied form of climbing fiber-triggered plasticity , long-term depression of the parallel fiber-to-Purkinje cell synapses ( pf-Pk LTD ) , impairs certain learning tasks while leaving other cerebellum-dependent learning tasks intact ( Aiba et al . , 1994; Boyden et al . , 2006; De Zeeuw et al . , 1998; Feil et al . , 2003; Hansel et al . , 2006; Katoh et al . , 2000; Katoh et al , 2005; Kim and Thompson , 1997; Kishimoto et al . , 2001; Kishimoto et al . , 2001; Koekkoek et al . , 2003; Schonewille et al . , 2011; Shibuki et al . , 1996; Shutoh et al . , 2002 , 2003; van Alphen and De Zeeuw , 2002; Welsh et al . , 2005; Yanagihara and Kondo , 1996 ) . Thus , after decades of research , the conflicting evidence has left unresolved the question of whether error signals in the climbing fibers are the driver of motor learning . A potential reconciliation of the seemingly inconsistent results in the literature is suggested by recent studies in reduced preparations . Climbing fiber spikes were originally thought to convey a reliable , invariant , all-or-none signal for plasticity . However , more recently , it has been shown that a Purkinje cell’s response to its climbing fiber input can be graded , which , in turn , can regulate the efficacy of climbing fiber stimulation to induce plasticity in vitro and in decerebrate preparations ( Weber et al . , 2003; Carey and Regehr , 2009; Maruta et al , 2007; Mathy et al . , 2009; Rasmussen et al . , 2013 ) . However , to date , there has been no evidence about whether the ability of the error signals carried by climbing fibers to induce learning is also regulated in awake , behaving animals . When climbing fibers are activated by performance errors , does this reliably trigger plasticity , or can the impact of these error signals in the climbing fibers be gated ? To address this question , we compared two closely related oculomotor learning paradigms that rely on the same cerebellar microcircuit: learned increases and decreases in the gain of the vestibulo-ocular reflex ( VOR; Figure 1 , Figure 2A , B; for a review , see Boyden et al . , 2004 ) . 10 . 7554/eLife . 02076 . 003Figure 1 . Circuit for VOR motor learning . Vestibular stimuli ( head movements ) drive eye movement responses through VOR interneurons in the vestibular nuclei . Vestibular signals are also conveyed , via parallel fibers and interneurons , to Purkinje cells in the cerebellar floccular complex . Purkinje cells also receive input from climbing fibers that respond to retinal slip , which indicates a performance error--a failure of eye movements to stabilize a visual image on the retina . Climbing fiber activity is hypothesized to drive plasticity in the other inputs to Purkinje cells . Changes in Purkinje cell output can influence eye movements through their inhibitory effect on VOR interneurons in the vestibular nuclei . An extracellular recording from a Purkinje cell can detect complex spikes , which reflect spikes in its single climbing fiber input with a one-to-one correspondence , and simple spikes ( 71 ± 9 sp/s , mean ± SEM ) , which greatly outnumber complex spikes ( 0 . 98 ± 0 . 19 sp/s ) and thus are the major output from the Purkinje cells . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 00310 . 7554/eLife . 02076 . 004Figure 2 . Climbing fibers encode errors during VOR motor learning . ( A and B ) VOR learning is induced by pairing a vestibular stimulus with a moving visual stimulus . During VOR-increase training ( A ) , the visual stimulus ( grey trace ) is paired with a vestibular stimulus in the opposite direction ( red ) . During VOR-decrease training ( B ) , the visual stimulus is paired with a vestibular stimulus moving in the same direction ( blue ) . In panels A–D , upward and downward deflections indicate contraversive and ipsiversive motion , respectively , as defined relative to the side of the brain on which the neural recordings in panels E–H were made: if the climbing fiber was recorded from the left cerebellar flocculus , then leftward head rotations and leftward image motion on the retina were defined as ipsiversive . ( C and D ) During the visual-vestibular training stimuli , there is retinal slip , reflecting the failure of the eye movements to stabilize the visual stimulus , which is a performance error . Retinal slip is plotted in °/s , same scale as A , B . ( E and F ) Responses of the climbing fibers during vestibular stimuli in the direction accompanied by contraversive retinal slip ( the ‘on’ direction ) , which drives an increase in firing in the climbing fibers . During VOR-increase training ( E ) , the probability of a climbing fiber spike was elevated in a window 75–250 ms ( grey box ) after the onset of each ipsiversive vestibular stimulus ( i . e . , climbing fiber activity in the left flocculus increased during head rotation to the left ) . During VOR-decrease training ( F ) , the same climbing fibers increased their firing probability 75–250 ms after the onset of a contraversive vestibular stimulus . All climbing fibers that responded during one training paradigm also responded during the other paradigm , and with a similar response amplitude ( t ( 9 ) = 1 . 77 , p=0 . 11 paired t-test ) . For both training paradigms , climbing fiber activity was suppressed during interleaved vestibular stimuli in the opposite direction from those shown here ( Figure 2—figure supplement 1E , F ) . ( G and H ) The corresponding Purkinje cell simple spike firing rate . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 00410 . 7554/eLife . 02076 . 005Figure 2—figure supplement 1 . Climbing fiber responses during stimuli in the ‘off’ direction . Our VOR training paradigms used vestibular stimuli in alternating directions . Figure 2 shows the results for vestibular stimuli in the direction accompanied by contraversive retinal slip , which drives an increase in firing in the climbing fibers . Here , we show the results for vestibular stimuli in the direction accompanied by ipsiversive retinal slip: contraversive vestibular stimuli during VOR-increase training ( A ) and ipsiversive vestibular stimuli during VOR-decrease training ( B ) . These trials were not included in the analyses in Figures 3–5 because the climbing fiber activity was low during these trials . Data are from the same cells and experiments shown in Figure 2 . ( A and B ) Training stimuli . Vestibular stimulus and visual stimulus velocity are shown in °/s . Ipsiversive ( downward deflections of traces in A–D ) and contraversive ( upward deflections ) are defined relative to the side of the brain on which the climbing fibers in panels E and F were recorded ( defined as in Figure 2 ) . ( C and D ) Retinal slip velocity ( °/s ) during training . ( E and F ) Climbing fibers in the floccular complex respond to ipsiversive retinal slip with a decrease in their firing rate below baseline . Climbing fiber responses are plotted as the probability of a spike in each in 50 ms bin , as in Figure 2E , F ( G and H ) Purkinje cell simple spike firing rate . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 005 The VOR functions to stabilize visual images on the retina by using vestibular signals to elicit eye movements that compensate for head movements . To successfully stabilize images , the gain of the VOR ( amplitude of the eye movement response to a vestibular input ) needs to be well calibrated . This calibration occurs through a form of motor learning that depends on the cerebellar floccular complex ( flocculus and ventral paraflocculus ) ( Robinson , 1976; Ito et al . , 1982; Lisberger et al , 1984; Nagao , 1983; McElligott et al . , 1998; Rambold et al . , 2002 ) . Climbing fibers in this part of the cerebellum are robustly activated by performance errors during both VOR-increase and VOR-decrease learning: if the eye movements driven by the VOR are too small or too big , that performance error results in image motion on the retina ( retinal slip ) , which is encoded by the floccular climbing fibers ( Ghelarducci et al . , 1975; Graf et al . , 1988; Simpson and Alley , 1974; Stone and Lisberger , 1990; Figure 2C–F and Figure 2—figure supplement 1C–F ) . The same climbing fibers encode performance errors during both VOR-increase and VOR-decrease learning . Previous efforts to determine whether these error signals in the climbing fibers are what drive VOR learning have been inconclusive . Over the course of both VOR-increase and VOR-decrease learning , the Purkinje cell simple spike output during the VOR is altered in a manner consistent with the induction of LTD in parallel fiber inputs that were coactive with the climbing fibers during training ( Dufossé et al . , 1978; Hirata and Highstein , 2001; Lisberger et al . , 1994; Miles et al . , 1980; Nagao , 1989; Watanabe , 1984 , 1985 ) . Also , direct , optogenetic activation of the climbing fibers can induce VOR-increase learning , when paired with a vestibular stimulus ( Nguyen-Vu et al . , 2013 ) . On the other hand , a previous attempt to induce VOR-decrease learning with climbing fiber stimulation was unsuccessful ( Nguyen-Vu et al . , 2013 ) . Moreover , studies of multiple lines of mice deficient in pf-Pk LTD have not consistently found impairments of VOR learning , and the reported impairments are more pronounced for VOR-increase than VOR-decrease learning ( Aiba et al . , 1994; Boyden et al . , 2006; De Zeeuw et al . , 1998; Feil et al . , 2003; Hansel et al . , 2006; Kim and Thompson , 1997; Koekkoek et al . , 2003; Schonewille et al . , 2011; van Alphen and De Zeeuw , 2002 ) . These results raised the possibility that the efficacy of the error signals carried by climbing fibers to induce plasticity is reduced during VOR-decrease training . In this study , we provide convergent evidence for this possibility from recording and stimulation experiments . The regulation of climbing fiber efficacy for inducing plasticity represents a new component of the learning algorithm implemented by the cerebellum .
We recorded from Purkinje cells in the floccular complex of two adult rhesus monkeys during VOR-increase and VOR-decrease training . An extracellular recording from a Purkinje cell provides simultaneous access to two distinct physiological signals: complex spikes , which provide a one-to-one readout of spikes in the single climbing fiber innervating the Purkinje cell ( Eccles et al . , 1966 ) ; and simple spikes , which reflect the impact of all excitatory and inhibitory inputs as well as the intrinsic excitability of the Purkinje cell ( Figure 1 ) . The climbing fiber input to Purkinje cells in the flocculus encodes the retinal slip that drives VOR learning ( Simpson and Alley , 1974; Ghelarducci et al . , 1975; Graf et al . , 1988; Stone and Lisberger , 1990 ) . Retinal slip can induce an adaptive increase or decrease in the gain of the VOR , depending on its direction relative to the vestibular stimulus ( Boyden et al . , 2004 ) . To induce a learned increase in VOR gain , visual image motion is paired with a vestibular stimulus in the opposite direction ( Figure 2A ) ; to induce a learned decrease in VOR gain , visual image motion is paired with a vestibular stimulus in the same direction ( Figure 2B; Collewijn and Grootendorst , 1979; Vercher and Gauthier , 1990–1991; Pastor et al . , 1994; Raymond and Lisberger , 1996 ) . Most climbing fibers in the flocculus increase their firing in response to contraversive retinal slip ( image motion away from the side on which the cell is recorded ) ( Simpson and Alley , 1974; Raymond and Lisberger , 1998; Stone and Lisberger , 1990; Figure 2C–F ) . During VOR-increase training , the contraversive retinal slip that activates climbing fibers occurs during ipsiversive vestibular stimuli ( Figure 2A , C , E and Figure 2—figure supplement 1A , C , E ) , whereas during VOR-decrease training , the contraversive retinal slip and associated increase in climbing fiber activity occur during contraversive vestibular stimuli ( Figure 2B , D , F and Figure 2—figure supplement 1B , D , F ) . The same climbing fibers encode retinal slip during both training paradigms , and the amplitude of the response is the same ( Figure 2E , F ) . What distinguishes the two paradigms , and carries information about the required direction of learning , is whether the increased climbing fiber activity occurs during vestibular stimuli in one direction vs the other . During each learning paradigm , the climbing fiber activity coincides with vestibular stimuli in the appropriate direction to potentially induce adaptive changes in the eye movements . Climbing fiber-induced plasticity can reduce Purkinje cell simple spike output , either through long-term depression ( LTD ) at the parallel fiber-Purkinje cell synapses ( Ito et al . , 1982 ) or through potentiation of inhibitory inputs to the Purkinje cell ( Jörntell and Ekerot , 2003; Kano et al . , 1992; Tanaka et al . , 2013 ) . The error signals carried by climbing fibers during VOR-increase and VOR-decrease training should trigger reductions in Purkinje cell output during ipsiversive and contraversive vestibular stimuli , respectively . A reduction in Purkinje cell firing during ipsiversive vestibular stimuli would cause VOR interneurons in the vestibular nuclei to receive inhibition from Purkinje cells that is more out-of-phase with the excitatory input they receive from vestibular afferents ( Figure 1 ) , leading to larger responses in the vestibular nuclei and hence an increase in VOR gain ( Ito , 1982 ) . In contrast , a climbing fiber-triggered reduction in Purkinje cell firing during contraversive vestibular stimuli would cause the inhibition from Purkinje cells to be more in-phase with the excitatory vestibular afferents , resulting in smaller responses of vestibular nuclei neurons and a decrease in VOR gain . Such changes in Purkinje cell responses have been reported by several laboratories after VOR learning ( Dufossé et al . , 1978; Hirata and Highstein , 2001; Lisberger et al . , 1994; Miles et al . , 1980; Nagao , 1989; Watanabe , 1984 , 1985 ) . However , the question of whether the error signals in the climbing fibers are what triggers these changes has been controversial ( Ke et al . , 2009; Nguyen-Vu et al . , 2013; reviewed in Ito , 1982; Lisberger , 1988 ) . One approach we used to address this question was a trial-by-trial analysis of the correlation between activity in the climbing fibers and plasticity of the Purkinje cell responses during training . The activity of an individual climbing fiber encodes retinal slip in a probabilistic manner ( Figure 2E , F ) . Climbing fibers fire at a low rate , and even when retinal slip is present , an individual climbing fiber does not fire on every trial . We harnessed this natural variance to assess whether the activation of the climbing fibers during training is what drives plasticity in the VOR circuit . If a spike in the climbing fiber on one trial induces plasticity in the inputs to its Purkinje cell target , this might be detected as a change in the Purkinje cell’s simple spike response on the next trial , as previously reported during smooth pursuit eye movement learning ( Medina and Lisberger , 2008; Yang and Lisberger , 2010 , 2013 ) . Therefore , we assessed the extent to which the presence or absence of a spike in its climbing fiber input on one trial predicted a change in a Purkinje cell’s simple spike response on the subsequent trial during VOR learning . We identified pairs of consecutive trials in which there was a spike in the climbing fiber on the first trial of the pair , but not the second trial ( CF–No CF pairs ) , and calculated the trial-to-trial change in the simple spike response ( Figure 3A ) . Likewise , we identified pairs of consecutive trials in which there was no spike in the climbing fiber on either trial ( No CF–No CF pairs ) , and calculated the trial-to-trial change in the simple spike response . During VOR-increase training , there was a tight , trial-by-trial correlation between the activity of the climbing fiber input to a given Purkinje cell and the induction of changes in its simple-spike output ( Figure 3B ) . If there was a spike in the climbing fiber input on the first trial of a pair , there was a reduction of about 8 spikes/s in the firing rate of the Purkinje cell on the subsequent trial ( Figure 3B , C; CF–No CF , red ) . On trials with no climbing fiber spike , there was no reduction in Purkinje cell firing rate on the subsequent trial ( Figure 3B , C; No CF–No CF , black ) . The sensory error ( retinal slip speed ) was indistinguishable on trials with a climbing fiber spike vs without a climbing fiber spike , and therefore cannot account for the difference observed on the trial after the climbing fiber spike ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 02076 . 006Figure 3 . Trial-by-trial effects of climbing fiber activity on Purkinje cell output . ( A ) Schematic illustrating the analysis . Isolated Purkinje cells were recorded during the VOR training paradigms illustrated in Figure 2 . Pairs of consecutive trials were identified in which a complex spike , indicative of a spike in the climbing fiber input to the cell ( • , CF ) , occurred 75–250 ms after stimulus onset in the first trial of the pair , but not in the subsequent trial ( left , CF–No CF pair ) . The Purkinje cell’s simple spike firing rate during the first trial was subtracted from the second trial to calculate the change in Purkinje cell response on the trial after the climbing fiber spike ( right ) . This analysis was performed for stimuli in the ‘on’ direction for the climbing fiber: ipsiversive vestibular stimuli during VOR-increase training , and contraversive vestibular stimuli during VOR-decrease training . ( B ) Trial-to-trial changes in Purkinje cell responses during VOR-increase training . If there was a spike in the climbing fiber on the first trial of a pair , there was a decrease in Purkinje cell firing on the subsequent trial ( CF–No CF pairs , red ) . If there was no spike in the climbing fiber on the first trial , there was no detectable change in Purkinje cell firing on the subsequent trial ( No CF–No CF pairs , black ) . Changes were considered significant if they lay outside the 95% confidence interval from a bootstrap distribution of the data ( shown for CF–No CF data , dashed black lines , see ‘Materials and methods’ ) . There were no significant changes in the No CF–No CF trial pairs based on their confidence interval . ( C ) The Purkinje cell firing rate ( baseline subtracted ) during the first ( dashed lines ) and second trials ( solid lines ) of CF–No CF ( red ) and No CF–No CF pairs ( black ) during VOR-increase training . ( D ) Trial-to-trial changes in Purkinje cell responses during VOR-decrease training . There was no reduction of Purkinje cell firing on the trial after a climbing fiber spike ( CF–No CF , blue , dashed black lines are 95% confidence intervals from the bootstrap distribution ) . ( E ) The Purkinje cell firing rate ( baseline subtracted ) during the first ( dashed lines ) and second trials ( solid lines ) of CF – No CF ( blue ) and No CF – No CF pairs ( black ) during VOR-decrease training . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 00610 . 7554/eLife . 02076 . 007Figure 3—figure supplement 1 . Similar retinal slip in trials with vs without a climbing fiber spike . During both VOR-increase ( left ) and VOR-decrease training ( right ) , the retinal slip on the trials in which the climbing fiber spiked ( CF trials; red , blue ) was indistinguishable from the retinal slip on the trials in which there was no climbing fiber spike ( No CF trials; black; F ( 1 , 18 ) = 0 . 06 , p=0 . 80 for VOR-increase and F ( 1 , 18 ) = 0 . 02 , p=0 . 88 for VOR-decrease , ANOVA ) . Thus , there was no difference in the sensory ( visual ) error that could explain the different changes in Purkinje cell output and behavior on trials after CF trials vs No CF trials . Positive and negative values indicate contraversive and ipsiversive retinal slip , respectively , defined relative to the side of the brain on which the cell was recorded . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 007 The tight , trial-by-trial correlation between the occurrence of a climbing fiber spike on one trial , and the change in Purkinje cell simple spike output on the next trial suggests that the activation of the climbing fibers is what is triggering the changes in Purkinje cell output during VOR-increase training . In striking contrast , in the very same set of cells , a trial-by-trial analysis during VOR-decrease learning revealed no correlation between climbing fiber activity and plasticity of Purkinje cell responses ( Figure 3D , E; CF–No CF , blue ) . Unlike VOR-increase training and all previous trial-by-trial analyses of smooth pursuit learning ( Medina and Lisberger , 2008; Yang and Lisberger , 2010 , 2013 ) , there was no reduction in Purkinje cell firing on the trial after a climbing fiber spike during VOR-decrease training ( Figure 3D , E; CF-No CF , blue ) . The Purkinje cells did undergo gradual , adaptive changes in their responses over the course of the full VOR-decrease training session , which could be detected during the same ∼90-s VOR-decrease training sessions used for the trial-by-trial analysis ( Figure 4 ) . Across trials of VOR-decrease training , there was a progressively lower Purkinje cell firing rate during contraversive vestibular stimuli , similar to what has been reported previously using much longer training periods ( Dufossé et al . , 1978; Hirata and Highstein , 2001; Lisberger et al . , 1994; Miles , Braitman , et al . , 1980; Nagao , 1989; Watanabe , 1984 , 1985 ) . This observation , along with the observation that lesions of the floccular complex disrupt both VOR-increase and VOR-decrease learning ( Ito et al . , 1982; Koekkoek et al . , 1997; Lisberger et al . , 1984; Nagao , 1983; Rambold and Churchland , 2002 ) , indicates that the Purkinje cells in our sample participate in VOR-decrease learning as well as VOR-increase learning . However , there was no trial-by-trial correlation between the climbing fiber activity and the plasticity of the Purkinje cell responses to suggest a causal relationship during VOR-decrease training , as there was during VOR-increase learning . Thus , although the performance errors elicited equally robust responses in the climbing fibers during the two learning paradigms , those responses seem to have a different impact in terms of their ability to induce plasticity . 10 . 7554/eLife . 02076 . 008Figure 4 . Changes in Purkinje cell output during the full VOR-increase and VOR-decrease training sessions . ( A ) Responses of example Purkinje cells during VOR-increase ( red ) and VOR-decrease ( blue ) training sessions , demonstrating a gradual change ( reduction ) in the firing rate of the two Purkinje cells over the course of ∼30 individual trials . The baseline-subtracted simple spike firing rate was measured during the first 100 ms of each ipsiversive ( VOR-increase ) or contraversive ( VOR-decrease ) vestibular stimulus , and plotted as a function of trial number . The slope of the linear regression ( colored lines ) provided a measure of the rate of change in the Purkinje cell’s response during the training session . ( B ) Average change in Purkinje cells during training , measured from the slopes of the linear regressions for all training sessions and all cells . There was a consistent reduction in Purkinje cell firing rate during contraversive vestibular stimuli over the course of ∼90 s of VOR-decrease training ( t ( 26 ) = -2 . 51 , p=0 . 019 , one sample t-test ) and a trend for a decrease in firing during ipsiversive vestibular stimuli over the course of VOR-increase training ( t ( 27 ) = −1 . 88 , p=0 . 071 , one sample t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 008 What could be preventing the error signals in the climbing fibers from inducing adaptive changes during VOR-decrease training ? One trivial possibility would be that during contraversive vestibular stimuli there are simply not enough synapses active simultaneously with the climbing fibers to serve as a substrate for associative plasticity . This seems unlikely because there is robust activation of mossy fiber inputs to the cerebellar flocculus during both contraversive and ipsiversive vestibular stimuli ( Noda , 1986 ) , and a substantial fraction of Purkinje cells increase their firing in response to contraversive vestibular stimuli ( Dufossé et al . , 1978; Lisberger and Fuchs , 1978; Miles et al . , 1980; Watanabe , 1984; Nagao , 1989; Pastor et al , 1997; Lisberger et al . , 1994; Raymond and Lisberger , 1997; Hirata and Highstein , 2001; Blazquez et al . , 2003; Ke et al . , 2009 ) . Thus , there should be cerebellar synapses coactive with the climbing fibers during both training paradigms , and therefore some other aspect of the state of the cerebellar circuit is more likely to be regulating the ability of the error signals carried by climbing fibers to induce plasticity . Recent studies have highlighted the duration of the complex spike as a factor that may regulate the induction of plasticity by the climbing fibers . The complex spike consists of a large initial spike followed by a variable number of spikelets . When the number of spikelets , and hence the complex spike duration , is manipulated in vitro or in decerebrate animals ( Carey and Regehr , 2009; Mathy et al . , 2009; Rasmussen et al . , 2013 ) , the probability of climbing fiber-induced plasticity can be altered , with shorter complex spike durations associated with a lower probability of climbing fiber-induced plasticity . We evaluated whether shorter complex spike durations could explain the lack of a trial-to-trial effect of climbing fiber activity during VOR-decrease training , by analyzing the waveform of complex spikes during each training paradigm ( Figure 2E , F ) , in the same analysis windows used for the trial-by-trial analysis . Surprisingly , we found that complex spike waveforms during VOR-decrease training had more spikelets and were of longer duration than those during VOR-increase training ( Figure 5A ) . Thus , the reduced efficacy of climbing fibers to induce plasticity during VOR-decrease learning cannot be explained by a shorter complex spike duration . 10 . 7554/eLife . 02076 . 009Figure 5 . Measures of climbing fiber effects and circuit state during training . ( A ) Complex spike waveforms during VOR-increase ( red ) and VOR-decrease ( blue ) training . Left , example cell: top , mean waveforms; bottom , overlaid individual complex spikes . Note second spikelet during VOR-decrease training . Across all cells , the complex spike waveforms were of longer duration during VOR-decrease training ( center , t ( 9 ) = −2 . 67 , *p<0 . 05 , paired t-test ) and had more spikelets ( right , t ( 9 ) = −2 . 71 , *p<0 . 05 , paired t-test ) compared to VOR-increase training . Horizontal black bars and grey rectangles indicate mean ± SEM ( B ) The length of the pause in simple spike firing following a complex spike was similar during VOR-increase and VOR-decrease training ( NS , t ( 9 ) = 0 . 79 , p=0 . 45 , paired t-test ) . Measurements in panels A and B were made on the same complex spikes used for the trial-by-trial analysis in Figure 3 . ( C ) Mean Purkinje cell simple spike firing ( baseline subtracted , first 250 ms of trials ) during stimuli in the ‘on’ direction for the climbing fibers ( Figure 2E–H ) was higher during VOR-decrease training ( blue ) , than during VOR-increase training ( red; t ( 9 ) = −3 . 39 , **p<0 . 01 , paired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 009 We also considered the possibility that the length of the climbing fiber-triggered pause in simple spike firing may affect the potency of climbing fiber activity to induce learning ( Maiz et al . , 2012 ) . However , there was no significant difference in the duration of the post-complex spike pause during VOR-increase and VOR-decrease training ( Figure 5B ) . It is conceivable that climbing fiber-triggered plasticity would be less effective at reducing Purkinje cell simple spike firing if the firing rate was already very low . Therefore , we compared the average simple spike responses during vestibular stimuli in the direction associated with elevated climbing fiber activity: ipsiversive and contraversive vestibular stimuli during VOR-increase and VOR-decrease training , respectively . For both , firing rate decreased relative to baseline during the training stimulus ( Figure 2G , H ) , but this reduction was smaller during VOR-decrease training than VOR-increase training ( Figure 5C ) . This suggests that there is no floor effect limiting plasticity of the simple spike responses during VOR-decrease training . Nevertheless , the difference in simple spike rates during the two training paradigms suggests different patterns of excitatory and inhibitory input to Purkinje cells , which might regulate climbing fiber-triggered plasticity . The results of our trial-by-trial analysis suggest that error signals carried by climbing fibers contribute to the induction of plasticity during VOR-increase but not during VOR-decrease training . However , such correlational evidence cannot , by itself , establish causality , and the relationship between the short-term effects observed in the trial-by-trial analysis and long-term learning has not been established ( Yang and Lisberger , 2013 ) . Thus , to provide a causal test of the climbing fiber role in VOR learning , and to analyze climbing fiber-induced plasticity over a longer time scale , we used an optogenetic stimulation approach in mice . We tested whether direct , optogenetic activation of climbing fibers could replace the sensory feedback ( visual error signals provided by retinal slip ) that normally drives VOR learning , and induce learning when paired with a vestibular stimulus in the absence of any visual feedback . We recently reported preliminary evidence that climbing fiber stimulation can induce VOR-increase but not VOR-decrease learning ( Nguyen-Vu et al . , 2013 ) . However , the single set of stimulation parameters used in that study was designed to elicit a maximal response , and could potentially have recruited additional plasticity that masked the expression of any climbing fiber contribution to VOR-decrease learning . Moreover , there was no test of whether climbing fiber activation could be initiating changes in the VOR circuit that would support the delayed expression of VOR-decrease learning at later time points beyond the training session , which is plausible given the evidence that different mechanisms can support oculomotor learning over different time scales ( Titley et al . , 2007; Shutoh et al . , 2006; Okamoto et al . , 2011 ) . Here , we address those limitations by testing a broader range of stimulation parameters and by measuring learning 2 hr after training as well as during the 30-min training session ( Figures 6 and 7 ) . 10 . 7554/eLife . 02076 . 010Figure 6 . Optogenetic mimicry of error signals in the climbing fibers induced VOR-increase but not VOR-decrease learning . ( A ) Optogenetic training paradigms . Left , to mimic error signals carried by climbing fibers during visual-vestibular VOR-increase or VOR-decrease training ( see Figure 2A–F ) , we optogenetically activated floccular climbing fibers during the ipsiversive phase ( CF stim + Ipsi Vestibular , red ) or contraversive phase ( CF stim + Contra Vestibular , blue ) of a 1 Hz sinusoidal vestibular stimulus ( black ) . Climbing fibers were activated bilaterally using a single pulse of blue light repeated at 1 s intervals ( cyan ) , see ‘Materials and methods’ ) . Right , before and after each training block , the VOR gain was measured in the absence of climbing fiber stimulation . Representative eye velocity traces from the same mouse pre-training ( grey ) and 2 hr after optogenetic VOR-increase training ( red ) or 2 hr after optogenetic VOR-decrease training ( blue ) . ( B ) Motor learning induced by pairing climbing fiber activation with a vestibular stimulus . Learning was measured as the % change in VOR gain relative to pre-training , and depended on the training condition ( F ( 2 , 17 ) = 3 . 81 , p<0 . 05 , repeated measures two-way ANOVA ) . When climbing fibers were activated during the ipsiversive phase of the vestibular stimulus ( red ) , the VOR gain increased relative to control training with the vestibular stimulus in the absence of climbing fiber stimulation ( Vestibular-only , black ) , and relative to training with climbing fiber activation paired with the contraversive phase of the vestibular stimulus ( blue , *p<0 . 05 , Fisher’s post-hoc test ) . The changes in VOR gain induced by pairing climbing fiber activation with the contraversive vestibular stimulus were not significantly different from the vestibular-only control ( p=0 . 62 , Fisher’s post-hoc test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 01010 . 7554/eLife . 02076 . 011Figure 6—figure supplement 1 . Optogenetic activation of climbing fibers . ( A ) In vivo extracellular optrode recording from a Purkinje cell in the flocculus showing the complex spikes elicited by the optogenetic activation of its climbing fiber input . Climbing fibers were activated by a single pulse of 473 nm light ( cyan , 2 ms duration , 0 . 3 mW/mm2 ) repeated at 1 s intervals , and delivered unilaterally to the cerebellar flocculus . Individual waveforms show the optogenetically elicited complex spikes with the stimulus artifact subtracted . ( B ) Overlay of 73 optogenetically elicited complex spike waveforms from the same Purkinje cell as in A , stimulated at 1 Hz . ( C ) Histogram showing simple spike rate aligned on the time of optogenetically elicited complex spikes ( t = 0 ms ) in the same cell . The pause in simple spike firing is similar to that observed after spontaneous complex spikes ( Goossens et al . , 2001; Sato et al . , 1992 ) . Bin size , 2 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 01110 . 7554/eLife . 02076 . 012Figure 7 . VOR learning induced by a range of climbing fiber stimulation protocols . VOR learning was measured immediately ( A ) and 2 hr after training ( B ) . Climbing fibers were stimulated during the ipsiversive ( red ) or contraversive ( blue ) phase of the vestibular stimulus , to roughly mimic climbing fiber responses during visual-vestibular VOR-increase or VOR-decrease training , respectively . Climbing fibers were stimulated once ( CF 1x ) or three times ( CF 3x ) per cycle of the vestibular stimulus , unilaterally or bilaterally . Each bar represents the mean ± S . E . M change in VOR gain induced by each training paradigm relative to the pre-training baseline . Bars ‘outlined in bold’ are significantly different from zero ( p<0 . 05 ) ; asterisks indicate significant difference from the vestibular-only control ( *p<0 . 05; **p<0 . 01; ***p<0 . 001 ) ; # indicates the learned changes in VOR gain were different when the same climbing fiber stimulation was applied during the contraversive vs ipsiversive phase of the vestibular stimulus ( red vs blue bars , p<0 . 05 ) ; one sample t-test , Wilcoxon signed rank test , or Mann–Whitney test ( see ‘Materials and methods’ ) . Numbers indicate the number of mice for each training condition . Climbing fiber stimulation alone ( light grey ) induced no learning ( p=0 . 12 , Wilcoxon signed rank test for immediately post-training; p=0 . 79 , one sample t-test for 2 hr post-training ) , but had a significant effect when paired with the ipsiversive phase of the vestibular stimulus ( p<0 . 0001 , Kruskal–Wallis test for training condition; post-hoc Dunn’s multiple comparison tests vs vestibular-only , *p<0 . 05 , **p<0 . 01 , ***p<0 . 0001 ) . Optogenetic VOR-decrease training with stimulation of the climbing fibers during the contraversive phase of the vestibular stimulus , to roughly mimic their response during VOR-decrease training ( blue ) did not induce an associative decrease in the VOR below the vestibular-only control ( dark grey ) . Instead , there was a slight increase relative to vestibular-only immediately after training ( F ( 3 , 50 ) = 3 . 00 , p<0 . 05 , one-way ANOVA for training condition; *p<0 . 05 , Dunnett’s multiple comparison test ) , which was not significant 2 hr after training ( p=0 . 49 , Kruskal–Wallis test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02076 . 012 To optogenetically stimulate climbing fibers , virus carrying ChR2 was injected into the inferior olive , and several weeks later , the climbing fiber terminals in the cerebellar flocculus were illuminated with blue light ( Nguyen-Vu et al . , 2013 ) . ChR2 expression in the climbing fibers was verified anatomically , and optrode recordings from Purkinje cells demonstrated that complex spikes could be elicited by brief pulses of light to the flocculus ( Figure 6—figure supplement 1 ) . To roughly mimic the pattern of climbing fiber activation observed during visually driven VOR-increase training ( Figure 2A , C , E; Ke et al . , 2009; Raymond and Lisberger , 1998; Simpson and Alley , 1974; Watanabe , 1984 , 1985 ) , optogenetic climbing fiber activation was paired with ipsiversive vestibular stimuli . This pairing was done in total darkness , that is , in the absence of retinal slip , the sensory error signal that normally drives VOR learning . The vestibular stimulus consisted of 1 Hz sinusoidal head rotation . The climbing fiber stimulation was bilateral , with the climbing fibers in the flocculus on each side activated in alternation , using a single , 15-ms pulse of light centered on peak ipsiversive head velocity: climbing fibers in the right flocculus were stimulated during peak rightward head velocity , and climbing fibers in the left flocculus were stimulated during peak leftward head velocity ( Figure 6A , red ) . As a control , the same vestibular stimulus was delivered in the absence of any climbing fiber stimulation during the 30-min training period . This ‘vestibular-only’ training induced habituation of the VOR , a gradual decrease in VOR gain , as described previously by several groups ( Dow and Anastasio , 1999; Stahl , 2004; Boyden et al . , 2006; Gutierrez-Castellanos et al . , 2013 ) ( Figure 6B , black and Figure 7A , dark grey ) . If the climbing fibers were stimulated during the ipsiversive phase of the same vestibular stimulus , the VOR gain was higher than the vestibular-only control at the end of training ( Figure 6B , red and Figure 7A , pink ) . This effect of climbing fiber stimulation was long-lasting; when retested 2 hr after training , the VOR gain was still higher compared to vestibular-only training ( Figure 6B , red and Figure 7B , pink ) . For both training paradigms , there was an increase in VOR gain between the end of training and the 2 hr retest , presumably reflecting decay of the non-associative habituation ( Figure 6B , black and Figure 7 , dark grey ) . To roughly mimic the pattern of climbing fiber activation during VOR-decrease training ( Figure 2B , D , F ) ( Ke et al . , 2009; Raymond and Lisberger , 1998; Simpson and Alley , 1974; Watanabe , 1984 , 1985 ) , the climbing fibers were optogenetically activated during the contraversive phase of the vestibular stimulus ( Figure 6A , blue ) . This pairing had no detectable effect , immediately or 2 hr post-training . There was no significant reduction of the gain of the VOR below the habituation level observed in response to training with the vestibular stimulus alone ( Figure 6B , blue and Figure 7 , light blue ) . In contrast , when the vestibular stimulus was paired with an appropriate visual stimulus ( see ‘Materials and methods’ ) , there was a bigger decrease in VOR gain than induced by the vestibular stimulus alone , demonstrating that an associative decrease in VOR gain was possible ( Figure 7A , white ) . Therefore , we tested whether stronger climbing fiber stimulation could induce an associative decrease in VOR gain . We stimulated the climbing fibers three times during each cycle of the vestibular stimulus ( 125 ms inter-stimulus interval ) , either unilaterally or bilaterally . These stronger climbing fiber stimulation paradigms induced bigger increases in VOR gain when delivered during the ipsiversive phase of the vestibular stimulus ( Figure 7 , dark pink and red ) . However , when timed to induce VOR-decrease learning , increasing the number of climbing fiber stimuli had no effect ( Figure 7 , blue and dark blue ) . Thus , none of our training paradigms induced a decrease in VOR gain below the vestibular-only control , either immediately or 2 hr after the end of training . It is possible that climbing fiber stimulation could have induced an associative decrease in VOR gain had we been able to more closely mimic the natural climbing fiber responses to a performance error . However , the same climbing fiber stimulation protocols were effective at inducing robust and graded learning when timed relative to the vestibular stimulus to mimic VOR-increase training . Moreover , the lack of VOR-decrease learning in response to optogenetic climbing fiber stimulation was consistent with the trial-by-trial analysis , which found no evidence for a contribution of the natural , visually-elicited error signals in the climbing fibers to the induction of VOR-decrease learning . Thus , the stimulation results , together with the trial-by-trial recordings , suggest that VOR-decrease learning is not induced by the climbing fibers , even though they carry error signals that could potentially guide learning , and the very same climbing fibers contribute to the induction of VOR-increase learning .
The convergent evidence from recording and stimulation experiments , as well as previous , perturbation studies , strengthens the interpretation of the results . A previous stimulation study reported a failure of climbing fiber stimulation to induce VOR-decrease learning ( Nguyen-Vu et al . , 2013 ) , however that negative finding was difficult to interpret , because it might have simply reflected a failure of the stimulation protocol used to adequately mimic the natural patterns of climbing fiber activation present during normal VOR-decrease learning . Therefore in the current paper , we paired stimulation experiments with highly complementary recording experiments , which showed that the natural climbing fiber responses present during visual-vestibular VOR-decrease training are not correlated with the induction of plasticity either . The stimulation experiments can demonstrate causality but are inherently ‘unnatural’ . The recording experiments document what happens under more natural conditions , but cannot establish causality . Together , the convergent evidence from the stimulation and recording experiments are considerably more powerful than either alone . Moreover , the current recording and stimulation results are consistent with a previous perturbation study reporting that mice with impaired long-term depression of the parallel fiber-to-Purkinje cell synapses ( pf-Pk LTD ) are selectively impaired on VOR-increase but not VOR-decrease learning ( Boyden et al . , 2006 ) . That study did not rule out a contribution of the climbing fibers to VOR-decrease learning via a mechanism other than pf-Pk LTD , such as rebound potentiation of inhibitory synapses onto the Purkinje cells ( Kano et al . , 1992; Tanaka et al . , 2013 ) or plasticity in the Purkinje cells’ targets caused by the climbing fiber-triggered pause in Purkinje cell simple spiking ( Maiz et al . , 2012 ) . Thus , the present results extend our previous work by demonstrating a selective contribution , not only of pf-Pk LTD , but of all climbing fiber-dependent plasticity mechanisms to VOR-increase learning but not VOR-decrease learning . It is not known why the mechanisms for VOR-decrease and VOR-increase learning are different , although one can speculate that this would allow learning to appropriately weight the different ‘costs’ of having a VOR gain that is too high vs too low . For example , a VOR gain that is too high may be more likely to create eye movement instabilities . The results for VOR-increase learning provide a positive control indicating that our experimental approaches were able to detect climbing fiber-triggered plasticity , which is critical for interpreting the negative results from VOR-decrease learning . Previous , seemingly inconsistent findings about the role of the climbing fibers in cerebellum-dependent learning have been difficult to reconcile because different labs use different behavioral paradigms and different experimental approaches . In contrast , direct comparison of the VOR-increase and VOR-decrease paradigms provides a demonstration that the very same animals and even the very same cells could toggle between either a tight coupling or no apparent coupling between climbing fiber activity and the induction of plasticity , depending on the behavioral context of the specific oculomotor training paradigm . Thus , direct comparison of the positive and negative results from VOR-increase and VOR-decrease learning in the same animals and the same cells , using the same techniques , suggests that climbing fiber efficacy is actively regulated . Recently , there has been considerable interest in the possibility that the duration of the complex spike may regulate the induction of plasticity by the climbing fibers . In vitro , ethanol and neuromodulators such as norepinephrine can affect the duration of a complex spike and the induction of climbing fiber-dependent plasticity ( Carey and Regehr 2009; He et al . , 2013; Belmeguenai et al . , 2008 ) . However , there was no reduction in complex spike duration during VOR-decrease training that would account for the reduced efficacy of climbing fiber spikes suggested by the trial-by-trial analysis . Another factor that may gate the efficacy of climbing fibers to induce synaptic plasticity is the level of inhibition: coactive inhibitory inputs can impair the induction of LTD at the parallel fiber-to-Purkinje cell synapses ( Ekerot and Kano , 1985 ) , but also may serve as a substrate for climbing fiber-induced plasticity ( Jörntell and Ekerot , 2003; Kano et al . , 1992; Tanaka et al . , 2013 ) . The level of inhibitory input is difficult to assess in vivo; however , we measured the rate of Purkinje cell simple spikes , which provides a readout of the net balance of excitation and inhibition . During both training paradigms , simple spike firing rate decreased relative to the spontaneous , pre-stimulus baseline around the time of climbing fiber activation ( Figure 5C ) . This decrease in simple spike rate was smaller during VOR-decrease than VOR-increase training , indicating a higher ratio of excitatory to inhibitory inputs to the Purkinje cells at the time of climbing fiber activity during VOR-decrease vs VOR-increase training . Future studies will be required to determine whether this difference in E/I ratio could contribute to the differential efficacy of climbing fibers to trigger plasticity during the two training paradigms . The relationship between short-term , trial-by-trial plasticity and longer-term learning is not understood ( Yang and Lisberger , 2013 ) . Our results provide some of the first evidence that the changes observed over different time scales may be mechanistically related , by showing parallels between the contribution , or lack of a contribution , of climbing fiber activity to plasticity on different time scales: a single trial , over the course of a 30-min training session , and 2 hr after the end of training . In the trial-by-trial analysis for VOR-increase training , the change in Purkinje cell activity on trials following a climbing fiber spike was remarkably large , approximately 10% of the average firing rate ( Figure 3B ) . If the 10% change observed in a single trial persisted in its entirety and accumulated across trials , there should be a 10 , 000% change during the ∼1000 such trials that occur during an hour of training with the paradigms used in this study , but the observed change in behavior is typically less than 50% ( Pastor et al , 1992; Raymond and Lisberger , 1996; Boyden and Raymond , 2003; Katoh , 2007; Kimpo and Raymond , 2007 ) . This suggests that , at most , a small fraction ( <1% ) of the changes observed on a single trial could persist over the training session , consistent with a previous report that the single trial changes during smooth pursuit learning decay within a few seconds ( Yang and Lisberger , 2010 ) . Nevertheless , the parallels between the changes observed on a single trial and the changes observed over longer time scales suggest that they could be related . In particular , the direction of the trial-by-trial changes in Purkinje cell responses during VOR-increase learning are consistent with those observed over the course of a brief , 90 s VOR-increase training period ( Figures 3B and 4 , red ) and those observed after several hours or days of VOR-increase training ( Dufossé et al . , 1978; Hirata and Highstein , 2001; Lisberger et al . , 1994; Miles et al . , 1980; Nagao , 1989; Watanabe , 1984 , 1985 ) . Moreover , our results from the trial-by-trial analysis were consistent with the changes we observed over tens of minutes to hours in the stimulation experiments ( Figures 6 and 7 ) , in that both suggested a contribution of the climbing fibers to the induction of VOR-increase but not VOR-decrease learning . Our results provide evidence for a novel component of the neural algorithm for cerebellar learning in vivo—namely , a gating of the ability of error signals in the climbing fibers to induce learning . Since previous work has shown that regulation can also occur at the level of olivary spiking , climbing fiber-triggered plasticity seems to be gated at two different stages—at the level of the olivary neurons responding to a performance error by either spiking or not spiking , and at the level of climbing fiber spikes either inducing or not inducing plasticity . Thus , the cerebellum appears to conditionally heed the instructive signals provided by the climbing fibers . Dynamic gating of the neural error signals controlling the induction of learning may be a general feature of neural learning algorithms , and merits systematic investigation in the cerebellum and other circuits . Defining the factors that influence the receptivity of the cerebellum to neural error signals will be critical for developing more sophisticated theories of learning , and for developing strategies to optimize the recruitment of plasticity to aid learning or recovery from injury .
An optogenetic approach was used to stimulate the climbing fibers ( Boyden et al . , 2005 ) . The neurons in the inferior olive whose axons form the climbing fibers in the left and right cerebellar flocculus are in the right and left dorsal cap of Kooy , respectively , which lie on either side of the midline ( Ruigrok et al . , 1992; Tan et al . , 1995; Sugihara and Shinoda , 2004 ) . Retinal slip in a given direction drives opposite responses ( increases vs decreases in firing ) in dorsal cap neurons on opposite sides of the midline . Therefore , to approximate the natural responses of the climbing fibers to retinal slip requires independent control of olivary neurons on either side of the midline , which would be difficult to accomplish using electrical stimulation . The optogenetic approach overcomes this limitation , because ChR2-expressing climbing fibers can be selectively activated on one side of the brain by shining light on their terminals in the cerebellar flocculus . | The cerebellum ( or ‘little brain’ ) is located underneath the cerebral hemispheres . Despite comprising around 10% of the brain’s volume , the cerebellum contains roughly half of the brain’s neurons . Many of the functions of the cerebellum are related to the control and fine-tuning of movement , and people whose cerebellum has been damaged have problems with balance and coordination , and with learning new motor skills . One of the roles of the cerebellum is to control a reflex known as the vestibulo-ocular reflex , which enables us to keep our gaze fixed on an object as we turn our heads . The cerebellum relays information about head movements to the muscles that control the eyes , instructing the eyes to move in the opposite direction to the head . This keeps the image of the object we are looking at stable on the retina . The vestibulo-ocular reflex is controlled by a circuit that includes Purkinje cells ( which are the main output cells of the cerebellum ) and climbing fibres ( which originate in the brainstem ) . Any failure of the vestibulo-ocular reflex to fully compensate for head movements generates an error signal that activates the climbing fibres . These in turn modify the output of Purkinje cells , leading ultimately to adjustments in eye movements . However , Kimpo et al . have now obtained evidence that Purkinje cells can modulate their response to the instructions they receive from climbing fibres . Monkeys sat in a rotating chair while a visual object they were trained to track with their eyes was moved to induce errors in the vestibulo-ocular reflex . When the object was moved so that a bigger reflexive eye movement was required to stabilize the image , the activation of the climbing fibres in response to the error led to a change in the response of the Purkinje cells , as expected . However , when a smaller reflexive eye movement was needed , the error-driven responses of the climbing fibres did not alter the responses of Purkinje cells . Similar results were obtained using pulses of light to artificially activate climbing fibres and thus simulate error signals . The work of Kimpo et al . indicates that the cerebellum does not blindly follow the instructions it receives from the brainstem , but can instead modulate its responses to incoming information about performance errors . Further work is now required to identify factors that influence the responsiveness of the cerebellum: such information could ultimately be used to improve learning of motor skills and recovery from injury . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Gating of neural error signals during motor learning |
Replication of influenza viral genomic RNA ( vRNA ) is catalyzed by viral RNA-dependent RNA polymerase ( vRdRP ) . Complementary RNA ( cRNA ) is first copied from vRNA , and progeny vRNAs are then amplified from the cRNA . Although vRdRP and viral RNA are minimal requirements , efficient cell-free replication could not be reproduced using only these viral factors . Using a biochemical complementation assay system , we found a novel activity in the nuclear extracts of uninfected cells , designated IREF-2 , that allows robust unprimed vRNA synthesis from a cRNA template . IREF-2 was shown to consist of host-derived proteins , pp32 and APRIL . IREF-2 interacts with a free form of vRdRP and preferentially upregulates vRNA synthesis rather than cRNA synthesis . Knockdown experiments indicated that IREF-2 is involved in in vivo viral replication . On the basis of these results and those of previous studies , a plausible role ( s ) for IREF-2 during the initiation processes of vRNA replication is discussed .
The influenza A virus genome is composed of eight single-stranded viral RNA ( vRNA ) segments of negative polarity that form viral ribonucleoprotein ( vRNP ) complexes with viral RNA-dependent RNA polymerases ( vRdRPs ) and nucleocapsid proteins ( NPs ) . Both transcription and replication of vRNA are catalyzed by vRdRPs consisting of three viral proteins: PB1 , PB2 , and PA . Transcription of the influenza virus genome is initiated in a primer-dependent manner . The cap-1 structure of cellular mRNA was recognized by PB2 ( Guilligay et al . , 2008 ) , and the capped RNA is cleaved 10–15 bases downstream of the 5'-terminus by the endonuclease activity of PA ( Dias et al . , 2009; Yuan et al . , 2009 ) . This cleaved , short RNA with a 5’-cap structure serves as a primer for the initiation of transcription . After elongation of the nascent RNA chain , transcription is terminated by a virus-specific polyadenylation mechanism ( Poon et al . , 1999 ) : vRdRP reaches the poly ( U ) stretch of the vRNA template adjacent to its 5'-end and is thought to slip there repeatedly , leading to the addition of a poly ( A ) tail at the 3'-end of the nascent viral RNA transcript . Replication of the viral genomic RNA takes place in a primer-independent manner and proceeds in two steps: in the first step , vRdRP synthesizes full-length RNA copies of positive polarity , termed complementary RNA ( cRNA ) ; and in the second step , progeny vRNAs are amplified from the cRNA template . Both cRNA and vRNA products from each replication step contain a 5'-triphosphate end group ( Hay et al . , 1982; Young and Content , 1971 ) , indicating that both replications are initiated de novo . vRdRP and the viral RNA template are minimal and essential viral factors for viral RNA synthesis ( Deng et al . , 2005; Lee et al . , 2002 ) . NP is thought to play roles in efficient RNA elongation ( Honda et al . , 1988 ) and regulation of vRdRP activity for replication ( Beaton and Krug , 1986; Newcomb et al . , 2009; Portela and Digard , 2002; Shapiro and Krug , 1988 ) . These observations , which were provided using mainly cell-free systems , concord with the fact that an in vivo viral mini-genome replicon system can be reconstituted by transfection of expression plasmids for PB1 , PB2 , PA , and NP and of a plasmid for vRNA driven by cellular RNA polymerase I ( Pleschka et al . , 1996 ) . These viral factors catalyze cell-free viral RNA synthesis , but at a limited level . Furthermore , cell-free viral replication activity using viral factors could be observed in the presence of a large amount of viral factors ( Vreede and Brownlee , 2007 ) . In early cell-free studies using nuclei or nuclear extracts ( NEs ) prepared from infected cells , a robust level of transcription took place and replication to some extent could be observed ( Beaton and Krug , 1984; Jackson et al . , 1982; Nagata et al . , 1989 ) . Dissection and reconstitution of a cell-free viral RNA synthesis system using vRNP complexes isolated from virions , an exogenously added model RNA template , and uninfected NEs indicated that host-derived factors could be involved in the process of viral RNA synthesis ( Shimizu et al . , 1994 ) . Using this assay system , we identified two stimulatory host factors , designated as RNA polymerase activating factors ( RAF ) -I and -II , which were identified as Hsp90 and UAP56/BAT1 , respectively ( Momose et al . , 2001; 2002 ) . Later , we found a factor , designated as influenza virus replication factor ( IREF ) -1 , which upregulates unprimed cRNA synthesis from vRNA by the vRNP complex through promotion of the promoter clearance step and avoidance of premature abortive RNA synthesis . IREF-1 was shown to be identical to the minichromosome maintenance ( MCM ) complex ( Kawaguchi and Nagata , 2007 ) . Recently , several screening studies including high-throughput screenings have indicated that there may be more candidates for the host factors that affect viral RNA synthesis ( Brass et al . , 2009; Hao et al . , 2008; Karlas et al . , 2010; Konig et al . , 2010; Naito et al . , 2007; Watanabe et al . , 2014 ) . Among these potentially relevant candidates , some were further examined and characterized using in vivo and/or cell-free analyses . Here , we focused our study on identification of a novel host factor ( s ) that facilitates the efficient second step of replication , that is , vRNA synthesis from the cRNA template . We found that a crude fraction of uninfected NE enables vRdRP to synthesize unprimed vRNA from the cRNA template effectively , and we thus designated the novel factor ( s ) responsible for this activity as IREF-2 . Further fractionation and purification identified two host-derived factors , pp32 and APRIL . The target of these factors was shown to be a free form of the vRdRP trimeric complex . IREF-2 was found to preferentially upregulate vRNA synthesis rather than cRNA synthesis . Knockdown ( KD ) of IREF-2 using short interfering dsRNA resulted in decreased levels of viral RNA in the infected cells . However , the primary transcription from the incoming vRNP complexes was not affected . These results suggest that IREF-2 functions as a host factor for vRNA synthesis from cRNA in infected cells .
To establish a cell-free viral RNA synthesis system that mimics the second replication step , that is , unprimed vRNA synthesis from a cRNA template , we used micrococcal nuclease-treated vRNP complexes ( mnRNP ) as an enzyme source and the 53 nucleotide ( nt ) -long RNA harboring both terminal sequences of cRNA derived from segment 8 ( designed ‘c53’ ) as an exogenous cRNA model template ( Shimizu et al . , 1994 ) . This enzyme source was established in an early study ( Seong and Brownlee , 1992 ) and well characterized in subsequent studies ( Galarza et al . , 1996; Seong et al . , 1992 ) . These previous studies showed that mnRNP does not support unprimed vRNA synthesis from the cRNA template . NEs prepared from uninfected cells were fractionated through a phosphocellulose column by stepwise elution with increasing concentrations of KCl . P0 . 05 was a fraction unbound to the column , whereas P0 . 2 , P0 . 5 , and P1 . 0 were fractions eluted using 0 . 2 M , 0 . 5 M , and 1 . 0 M salt solutions , respectively . These fractions were examined individually in terms of their ability to promote cell-free vRNA synthesis reaction in the absence of any primer ( Figure 1A , lanes 3–6 ) . The 53 nt-long RNA product was formed only in the presence of the P0 . 05 fraction ( lane 3 ) . The RNA synthesis was dependent on the mnRNP , c53 , and P0 . 05 fractions ( Figure 1A , lanes 8–10 ) . This novel activity present in the P0 . 05 fraction was designated as IREF-2 . Next , we tried to purify and identify the factor responsible for the IREF-2 activity in the P0 . 05 fraction through fractionation using sequential column chromatography ( Figure 1B ) . To determine fractions containing the IREF-2 activity at each column chromatography step , every fraction was individually assayed for RNA synthesis ability in the presence of mnRNP and the c53 model template , but in the absence of any added primer . On the basis of its chromatographic behavior , IREF-2 appears to be highly acidic . At the final purification step using an anion exchanger ( the Mono-Q column ) , the IREF-2 activity was recovered in fractions eluted with about 500 mM of KCl ( fraction numbers 6–10 in Figure 1C , upper panel ) . The fractions from the Mono-Q column were also analyzed by SDS-PAGE followed by silver staining ( Figure 1C , lower panel ) . Comparison of the polypeptide elution patterns and the level of the IREF-2 activity strongly suggested that two polypeptides with molecular masses of approximately 32 and 31 kDa corresponded to the IREF-2 activity ( Figure 1C , arrows ) . The 32- and 31-kDa polypeptides were designated IREF-2α and β , respectively . 10 . 7554/eLife . 08939 . 003Figure 1 . Purification of influenza virus replication factor-2 ( IREF-2 ) . ( A ) IREF-2 activity in uninfected nuclear extracts ( NEs ) . Biochemical complementation assays using a cell-free vRNA replication system for fractions separated by phosphocellulose column chromatography were performed . The fractions , P0 . 05 ( lane 3 ) , P0 . 2 ( lane 4 ) , P0 . 5 ( lane 5 ) , and P1 . 0 ( lane 6 ) , were individually assayed in the cell-free viral RNA synthesis reaction employing 5 ng PB1-equivalent micrococcal nuclease-treated vRNP ( mnRNP ) as an enzyme source and 10 ng of the complementary RNA ( cRNA ) model template ( c53 ) , as described in the 'Materials and methods' . Dinucleotide ApG , serving as primer for viral RNA synthesis , was added to a final concentration of 0 . 2 mM ( lane 1 ) . To confirm the components required for the reactions , cell-free viral RNA synthesis with mnRNP and c53 in the presence of the P0 . 05 fraction was carried out ( lane 7; identical to the conditions of lane 3 ) . Simultaneously , reactions omitting the cRNA model template ( lane 8 ) , mnRNP ( lane 9 ) , or P0 . 05 fraction ( lane 10; identical to the conditions of lane 2 ) were also carried out . After incubation at 30°C for 2 hr , each reaction product was collected and subjected to 10% Urea-PAGE followed by autoradiography . ( B ) Purification scheme of IREF-2 from uninfected HeLa cell NEs . For details regarding the column chromatography , see 'Materials and methods' . ( C ) Profile of the fractions from Mono-Q column chromatography at the final purification step . Each Mono-Q fraction ( fraction numbers 1–11 ) or input material for the Mono-Q column chromatography ( i . e . , unbound fraction of the Uno-S column chromatography ) was individually added to this cell-free viral RNA synthesis reaction in the absence of any added primers ( upper panel ) . Each Mono-Q fraction was subjected to 11 . 5% SDS-PAGE , and polypeptides were visualized by silver staining ( lower panel ) . The closed arrowhead indicates 53 mer RNA products . The open arrowhead ( 50 mer ) indicates the product possibly generated by internal priming of ApG . The arrows indicate two candidate peptides responsible for IREF-2 activity . The molecular weight ( kDa ) positions are denoted on the left side of the panel . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 003 To confirm the polarity of the IREF-2-dependent product , we carried out RNase T2 protection assays . ApG-primed RNA products using v53 and c53 as templates were also prepared as control materials ( Figure 2A , lanes 1 and 4 ) , in addition to the IREF-2-dependent RNA products ( lane 7 ) . Each radioactively labeled ApG-primed or unprimed 53 nt RNA product was hybridized with excess amounts of either the v53 or the c53 probe and then subjected to digestion with RNase T2 , which preferentially digests ssRNA rather than dsRNA . Expectedly , the v53-directed ApG-primed RNA products were protected from digestion with RNase T2 by hybridization with the v53 probe ( lane 2 ) but not the c53 probe ( lane 3 ) , whereas the c53-directed ApG-primed RNA products were protected by hybridization with the c53 probe ( lane 6 ) but not with the v53 probe ( lane 5 ) , verifying the authenticity of this assay . The IREF-2-dependent RNA product was protected by hybridization with the c53 probe ( lane 9 ) but not with the v53 probe ( lane 8 ) . These results clearly indicate that the IREF-2-dependent RNA product is of negative polarity , that is , vRNA . 10 . 7554/eLife . 08939 . 004Figure 2 . Products of influenza virus replication factor-2 ( IREF-2 ) -dependent unprimed RNA synthesis . ( A ) RNase T2 protection assay . Radioactively labeled vRNA products were synthesized in the cell-free viral RNA synthesis system with micrococcal nuclease-treated vRNP ( mnRNP ) and the v53 model template in the presence of ApG ( lanes 1–3 ) , the c53 model template in the presence of ApG ( lanes 4–6 ) , and c53 in the presence of the IREF-2 fraction and in the absence of ApG ( lanes 7–9 ) . Viral RNA products were hybridized with excess amounts of nonlabeled v53 ( lanes 2 , 5 , and 8 ) or c53 ( lanes 3 , 6 , and 9 ) , which was followed by digestion with RNase T2 . Hybridized and digested RNA samples were extracted , collected , and subjected to 10% Urea-PAGE followed by autoradiography visualization . The closed arrowhead indicates 53-nt-long RNAs . ( B ) Analysis of the 5'-terminal structure of IREF-2-dependent unprimed vRNA products . [α-32P] GTP-labeled IREF-2-dependent unprimed vRNA products were prepared in the cell-free viral RNA synthesis system . The radioactively labeled 53-nt-long vRNA products were isolated from 10% Urea-PAGE , which was followed by excision and elution from the gel . A portion of the isolated 53-nt-long products was treated with alkaline phosphatase ( lanes 2 and 4 ) . Both nontreated and alkaline phosphatase-treated [α-32P] GTP-labeled unprimed vRNA products were digested with RNase T2 ( lanes 1–4 ) or snake venom phosphodiesterase ( lane 5 ) . The digested materials were spotted onto a polyethylenimine ( PEI ) -cellulose thin layer and developed with 1 N acetic acid-4 M LiCl ( 4:1 , v/v ) ( left panel; lanes 1 and 2 ) or 1 . 6 M LiCl ( right panel; lanes 3–6 ) and visualized by autoradiography . For mobility standards , nonradiolabeled AMP , ADP , and ATP were also subjected to thin-layer chromatography and are indicated on the left side of each panel . For a marker of pppAp , [γ-32P] ATP-labeled v53 synthesized using T7 RNA polymerase was also subjected to RNase T2 digestion and then to thin-layer chromatography ( lane 6 ) . The expected nucleotide positions are indicated on the right side of each panel by closed arrowheads . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 004 The other important point to be discussed regarding authentic vRNA replication is whether the vRNA product possesses a triphosphate moiety at its 5'-terminus due to de novo initiation . Thus , we analyzed the structure of the 5'-terminus of the IREF-2-dependent vRNA product ( Figure 2B ) . The vRNA product synthesized in the presence of [α-32P] GTP as a radioactive substrate in the reaction was expected to have 5'-xA[32p]Gp…-3' . If the product is synthesized de novo , ‘x’ must be a triphosphate moiety ( denoted as ppp ) . Upon digestion of the [α-32P] GTP-labeled IREF-2-dependent vRNA product with RNase T2 , which cleaves on the 3' side of a phosphodiester bond , pppA[32p] derived from the 5'-terminus could be detected by thin-layer chromatography separation ( Figure 2B , lanes 1 and 3 ) . This product , putative pppA[32p] , was not detected by treatment with alkaline phosphatase prior to RNase T2 digestion ( lanes 2 and 4 ) or by digestion with snake venom phosphodiesterase ( SV-PDE ) , which cleaves on the 5' side of a phosphodiester bond ( lane 5 ) . These results clearly reveal that a triphosphate moiety is present at the 5'-terminus of the IREF-2-dependent vRNA product , thereby indicating that the synthesis of this vRNA product is initiated de novo . Taken together , these results indicate that the IREF-2-dependent unprimed vRNA product is a bona fide unprimed vRNA replication product from the cRNA template . Next , we attempted to reveal the identity of IREF-2α and β polypeptides . To this end , fraction numbers 6 ( for IREF-2α , 32 kDa ) and 9 ( for IREF-2β , 31 kDa ) , shown in Figure 1C , were individually subjected to separation by SDS-PAGE , and polypeptides corresponding to IREF-2α ( 32 kDa ) and β ( 31 kDa ) , were excised from the gel and then digested with trypsin . The trypsin-digested oligopeptides derived from IREF-2α and β were then analyzed using matrix-assisted laser desorption-ionization time-of-flight mass spectrometry ( MALDI-TOF MS ) . By comparing the molecular masses of the oligopeptides from each IREF-2 protein obtained by MALDI-TOF MS analysis ( see Supplementary file 1 ) with the database , we identified these IREF-2 proteins as follows: IREF-2α was shown to be identical to pp32 ( phosphoprotein with a molecular mass of 32 kDa; accession number NP_006296 ) ( Malek et al . , 1990 ) and also known as leucine-rich acidic nuclear protein , inhibitor of protein phosphatase 2A ( I1pp2A ) , and putative HLA class II-associated protein I ( Matilla and Radrizzani , 2005 ) . IREF-2β was shown to be identical to APRIL ( acidic protein rich in leucines; accession number NP_006392 ) ( Mencinger et al . , 1998 ) , also named proliferation-related acidic leucine-rich protein or silver-stainable protein 29 ( Matilla and Radrizzani , 2005 ) . Both IREF-2α/pp32 and IREF-2β/APRIL exhibit high homology ( 71% sequence identity and 81% sequence similarity ) and belong to the acidic nuclear phosphoprotein 32-kDa family ( ANP32 family; pp32 and APRIL are also named ANP32A and ANP32B , respectively ) . These two proteins are encoded by separate genes and expressed mainly in the nuclei of variety kinds of tissues . Each IREF-2 protein is able to exist stably as a monomer and functions redundantly in multiple biological processes as an I1pp2A ( Li et al . , 1996a ) , a ligand to HuR that stabilizes ARE-containing mRNA ( Brennan et al . , 2000 ) , a component for INHAT activity ( Seo et al . , 2001 ) . The C terminus , comprising one-third of each protein , is extremely acidic , being composed of approximately 70% glutamic acid and aspartic acid residues , while the N-terminal region , comprising two-thirds of the protein , currently termed a leucine-rich-repeat , forms a horseshoe-shaped solenoid protein domain ( Huyton and Wolberger , 2007; Tochio et al . , 2010 ) . By previous physical and regulatory mapping analyses , pp32 ( ANP32A ) was shown to be a host protein related to influenza virus infection ( Shapira et al . , 2009 ) . In addition , a previous proteomic study identified pp32 and APRIL ( ANP32B ) as binding partners to the influenza vRdRP complex ( Bradel-Tretheway et al . , 2011 ) . Recently , a genomewide RNAi screening study identified APRIL as one of nine ‘top hit’ genes affecting the viral RNA synthesis process in infected cells ( Watanabe et al . , 2014 ) . However , functional analyses for IREF-2/ANP32s have not been carried out , and thus their functional importance remains unclear . Mono-Q chromatography could not separate IREF-2α/pp32 and IREF-2β/APRIL completely . Furthermore , several other polypeptides were also observed in these fractions ( Figure 1C , fraction numbers 6–10 ) . Therefore , we needed to confirm that pp32 and APRIL have authentic IREF-2 activity before any further molecular characterization . To this end , we prepared native and recombinant IREF-2 proteins and confirmed the quality of each protein preparation by silver staining and western blot analysis with anti-pp32 or anti-APRIL antibodies ( Figure 3—figure supplement 1 ) . Both native and recombinant IREF-2 proteins supported cell-free unprimed RNA synthesis using mnRNP and the c53 model template in a dose-dependent manner ( Figure 3A , lanes 2–4 and 5–7 , or lanes 8–11 and 12–15 , respectively ) . These results clearly indicate that pp32 and APRIL are authentic proteins responsible for IREF-2 activity , able to function independently of each other . We also tested the IREF-2 activity of an acidic protein , TAF-Iβ/SET ( Nagata et al . , 1995 ) . TAF-Iβ/SET was reported to exhibit similar functions to pp32 as an inhibitor of PP2A phosphatase ( Li et al . , 1996b ) or INHAT ( Seo et al . , 2001 ) . TAF-Iβ/SET did not exhibit any IREF-2 activity at all ( lanes 16–18 ) , suggesting that its acidic property is insufficient for IREF-2 activity . 10 . 7554/eLife . 08939 . 005Figure 3 . Characterization of influenza virus replication factor-2 ( IREF-2 ) activities in the cell-free system . ( A ) Dose response of IREF-2 . Native or recombinant IREF-2 proteins were added to the cell-free viral RNA ( vRNA ) synthesis reaction using micrococcal nuclease-treated vRNP ( mnRNP ) and the c53 model template in the absence of primer . Native pp32 ( lanes 2–4 ) , native APRIL ( lanes 5–7 ) , recombinant pp32 ( lanes 8–11 ) , and recombinant APRIL ( lanes 12–15 ) were used as follows: 5 ng ( lanes 2 , 5 , 8 , and 12 ) , 15 ng ( lane 3 , 6 , 9 , and 13 ) , 50 ng ( lanes 4 , 7 , 10 , and 14 ) and 150 ng equivalent ( lanes 11 and 15 ) of IREF-2 proteins . Recombinant TAF-Iβ/SET protein prepared using an Escherichia . coli expression system ( 33 , 110 , and 330 ng ) was also tested ( lanes 16–18 ) . After incubation at 30°C for 2 hr , the RNA products were collected and analyzed by 10% Urea-PAGE followed by autoradiography . ( B ) Template preference of IREF-2-dependent viral RNA synthesis . Viral RNA replication reactions were performed in the cell-free viral RNA synthesis system using mnRNP and either v53 ( lanes 1–5 ) or c53 ( lanes 6–10 ) as viral model templates for vRNA and complementary RNA ( cRNA ) , respectively . ApG at a final concentration of 0 . 2 mM ( lanes 1 and 6 ) or 30 , 100 , and 300 ng of recombinant pp32 ( lanes 3–5 and 8–10 ) was added to the reaction . ( C ) Effect of IREF-2 on cRNA synthesis from vRNP . Cell-free viral RNA synthesis using 2 ng PB1-equivalent of vRNP as the enzyme source and an endogenous genomic vRNA template were carried out in the presence ( lane 1 ) or absence ( lanes 2–8 ) of 0 . 2 mM ApG . Recombinant pp32 ( lanes 3–5 , 30 , 100 , and 300 ng , respectively ) was added to the reactions . As a positive control , 1 . 5 , 5 , and 15 ng of recombinant IREF-1/MCM were also used ( lanes 6–8 ) . The RNA products were collected and analyzed by 4% Urea-PAGE followed by autoradiography . One-third ( 33% ) of the total products derived from the ApG-primed cRNA synthesis were subjected to Urea-PAGE ( lane 1 ) . ( D ) Effect of IREF-2 on cap-snatching viral transcription . Cell-free viral RNA synthesis reactions were performed using mnRNP as the enzyme source and the exogenous model vRNA template ( v53 ) in the presence of 0 . 2 mM ApG ( lane 1 ) or globin mRNA as the 5’-capped RNA donor ( lanes 2–5 ) . Recombinant pp32 ( lanes 3–5 , 30 , 100 , and 300 ng , respectively ) was added to the reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 00510 . 7554/eLife . 08939 . 006Figure 3—figure supplement 1 . Protein profiles of native and recombinant influenza virus replication factor-2 ( IREF-2s ) . Native or recombinant IREF-2 proteins ( 50 ng ) were subjected to 11 . 5% SDS-PAGE followed by silver staining ( lanes 1–4 ) and western blot analysis with anti-pp32 antibody ( lanes 5–8 ) or anti-APRIL antibody ( lanes 9–12 ) . The Mono-Q fraction 6 ( shown in Figure 1C ) was used as native pp32 ( lanes 1 , 5 , and 9 ) . Mono-Q fractions 9 and 10 ( also shown in Figure 1C ) were further purified with Mono-Q and used as native APRIL ( lanes 2 , 6 , and 10 ) . Recombinant pp32 ( lanes 3 , 7 , and 11 ) and APRIL ( lanes 4 , 8 , and 12 ) were prepared using the Escherichia . coli expression system , as described in 'Materials and methods' . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 006 We therefore next examined the template preference of IREF-2 . Either v53 or c53 was used as the model RNA template for the cell-free reactions ( Figure 3B ) . Notably , the unprimed RNA product was observed from the cRNA template in an IREF-2-dependent manner ( Figure 3B , lanes 8–10 ) , but a significant level of cRNA synthesis was not detected when the vRNA template was used ( lanes 3–5 ) . These results suggest that IREF-2 preferentially regulates vRNA synthesis from the cRNA template , that is , the second step of the replication mechanism . Furthermore , the effect of IREF-2 on cRNA synthesis from the vRNP complex was also examined ( Figure 3C ) . Cell-free viral RNA synthesis using vRNP complexes as the enzyme and endogenous vRNA template source , that is , the vRdRP and genomic vRNA of each segment , was performed in the absence of a primer . As previously observed , replicative cRNA synthesis from the genomic vRNA templates was stimulated by recombinant IREF-1/MCM ( Figure 3C , lanes 6–8 ) , which is known to stimulate promoter clearance during replication ( Kawaguchi and Nagata , 2007 ) . In contrast , no significant change induced by pp32 was observed ( lanes 3–5 ) . This finding indicates that IREF-2 is not involved in cRNA synthesis from endogenous vRNA and that the function of IREF-2 is distinct from that of IREF-1/MCM . This observation also confirms the template polarity preference of IREF-2 ( Figure 3B ) . Next , to address whether IREF-2 affects viral transcription , cap-snatching viral transcription was performed in the cell-free system using 5’-capped mRNA of β-globin ( GmRNA ) as a 5’-cap donor for viral transcription ( Figure 3D ) . By adding GmRNA instead of ApG , the transcriptional product could be detected at a 10–12 nt longer size than that of the vRNA template ( compare lanes 1 and 2 ) , consistent with the findings of a previous report ( Plotch et al . , 1979 ) . After addition of pp32 to this cell-free transcription reaction , no obvious change was observed ( lanes 3–5 ) , suggesting that IREF-2 has neither a positive nor a negative effect on viral mRNA transcription . To determine a regulatory target ( s ) of IREF-2 , we performed interaction assays using IREF-2 and viral factors related to the viral RNA synthesis process . To date , no report has been published showing that pp32 and APRIL bind directly to RNA . In fact , the interaction between IREF-2 proteins and the model RNA template was not observed ( Figure 4—figure supplement 1 ) . Therefore , we performed GST pull-down assays using lysates prepared from mammalian cells ( HEK293T ) expressing GST-tagged IREF-2 ( GST-IREF-2 ) . Forty-eight hours after transfection with GST-derivative expression plasmids , the cells were infected with influenza virus at a multiplicity of infection ( MOI ) of 3 for 6 hr , and the viral factors bound to the GST-protein were then analyzed . Each subunit of the vRdRP complex ( PB1 , PB2 , and PA ) was co-precipitated with both GST-pp32 and GST-APRIL , but NP was not ( Figure 4A , lanes 2 and 3 ) . Instead , only NP was co-precipitated with GST-RAF2p48 , which functions as a chaperone for NP ( Momose et al . , 2001 ) ( lane 4 ) . To further confirm this , lysates prepared from cells exogenously expressing polymerases ( Figure 4B , lanes 1–4 ) or NP ( lanes 5–8 ) and GST-derivative proteins were also examined . Exogenously expressed trimeric complexes of the viral polymerases were co-precipitated with both GST-pp32 and GST-APRIL ( lanes 2 and 3 ) , but NP was not detected at all ( lanes 6 and 7 ) . GST-RAF2p48 interacted with NP , but not with any polymerase subunits ( lanes 4 and 8 ) . Thus , we concluded that the binding target of IREF-2 is the trimeric complex of vRdRP , not NP . Moreover , the fact that the interaction between IREF-2 and NP could not be observed in the infected cells suggested that IREF-2 might interact with a free form of vRdRP , but not with vRdRP , in RNP complexes in infected cells . To confirm this , the amount of viral RNA in the complex of IREF-2 and vRdRP was quantitatively determined . Complexes of GST-pp32 with vRdRP were prepared from infected cells by GST pull-down , as shown in Figure 4A ( lane 2 ) . The RNA pulled down with the complexes was then extracted , and the amount of viral RNA ( here , segment 5 ) was determined by reverse transcription-mediated quantitative PCR ( RT-qPCR ) . We found that 1 pmol of vRdRP complexed with GST-pp32 contains <0 . 5 fmol of segment 5 ( vRNA , cRNA , and viral mRNA ) only , suggesting that IREF-2 tends to interact with a free form of vRdRP in infected cells . 10 . 7554/eLife . 08939 . 007Figure 4 . Interaction of influenza virus replication factor-2 ( IREF-2 ) proteins with free forms of viral polymerase trimeric complexes . ( A ) Interaction between IREF-2 and viral proteins in infected cells . HEK293T cells were transfected with 10 μg of plasmids expressing GSTnls ( lane 1 ) , GST-pp32 ( lane 2 ) , GST-APRIL ( lane 3 ) , and GST-RAF2p48 ( lane 4 ) . At 48 hr post transfection , influenza virus was infected at an multiplicity of infection ( MOI ) of 3 . At 6 hr post infection , the transfected and infected cells were collected , lysed , and subjected to GST pull-down assays , as described in 'Materials and methods' . The pulled-down materials and 3% equivalent of the input samples were subjected to SDS-PAGE followed by western blot analysis with anti-PB1 , -PB2 , -PA , -NP , and -GST ( only the pull-down sample ) antibodies . ( B ) Interaction between IREF-2 and viral proteins in the transfected cells . HEK293T cells were transfected with 5 μg of plasmids expressing GSTnls ( lanes 1 and 5 ) , GST-pp32 ( lanes 2 and 6 ) , GST-APRIL ( lanes 3 and 7 ) , and GST-RAF2p48 ( lanes 4 and 8 ) and also co-transfected with the plasmids for the viral polymerase subunits ( 5 μg of pCAGGS-PB1 , 12 . 5 μg of pCAGGS-PB2 , and 2 . 5 μg of pCAGGS-PA [lanes 1–4] or 10 μg of pCAGGS-NP for NP expression [lanes 5–8] ) . Forty-eight hours post transfection , the cotransfected cells were collected , lysed , and subjected to GST pull-down assays . The input samples ( 3% ) and pulled-down materials were subjected to SDS-PAGE followed by western blot analysis . ( C ) Interaction between IREF-2 and trimeric or binary complexes of vRdRP . HEK293T cells were cotransfected with 5 μg of plasmids expressing GSTnls ( lanes 1 , 4 , and 7 ) , GST-pp32 ( lanes 2 , 5 , and 6 ) , GST-APRIL ( lanes 3 , 6 , and 9 ) , 5 μg of pCAGGS-PB1cFlag ( lanes 1–9 ) , 12 . 5 μg of pCAGGS-PB2 ( lanes 1–6 ) , and 2 . 5 μg of pCAGGS-PA ( lanes 1–3 and 7–9 ) . Forty-eight hours post transfection , the lysates from the cotransfected cells were subjected to GST pull-down assays or immunoprecipitation assays with anti-Flag antibody . The precipitated materials were subjected to SDS-PAGE followed by western blot analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 00710 . 7554/eLife . 08939 . 008Figure 4—figure supplement 1 . Electrophoresis mobility shift assay for influenza virus replication factor-2 ( IREF-2 ) and viral RNA . Radioactively labeled 53-nt-long model vRNA and complementary RNA ( cRNA ) probes ( v53 and c53; 246 . 9 cpm/fmol ) were synthesized by T7 RNA polymerase using [α-32P] GTP and isolated by gel excision . Each 500 pM ( final concentration ) of the labeled viral RNA probes , v53 ( lanes 1–7 ) and c53 ( lanes 8–14 ) was incubated with 10 nM or 50 nM of recombinant NP prepared using the Escherichia . coli expression system ( lanes 2 , 3 , 9 , and 10 ) , recombinant pp32 ( lanes 4 , 5 , 11 , and 12 ) , and recombinant APRIL ( lanes 6 , 7 , 13 , and 14 ) in 50 mM HEPES-NaOH ( pH 7 . 9 ) , 50 mM KCl , 0 . 5 U/μl of RNase inhibitor , and 15% ( v/v ) glycerol at 30°C for 30 min . After incubation , each binding mixture was loaded onto 0 . 6% agarose gel ( buffered with TBE ) and separated by electrophoresis ( 50 V for 3 hr ) . The gel was dried and visualized by autoradiography . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 00810 . 7554/eLife . 08939 . 009Figure 4—figure supplement 2 . Interaction between viral RNA-dependent RNA polymerase ( vRdRP ) complexes and endogenous influenza virus replication factor-2 ( IREF-2 ) . HEK293T cells were transfected with plasmids expressing PB1 ( lane 2 ) or PB1cFlag ( lanes 3–5 ) , PB2 ( lanes 2–4 ) , and PA ( lanes 2 , 3 , and 5 ) , as indicated . At 48 hr post transfection , the cells were harvested , lysed , and subjected to immunoprecipitation assays using anti-Flag antibody , as described in 'Materials and methods' . Input and precipitated materials were analyzed by SDS-PAGE , followed by western blot analysis using the antibodies indicated on the right side of each panel . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 009 We also tried to determine which subunit of vRdRP is the binding target of IREF-2 . Cells expressing GST-tagged derivative proteins and vRdRP trimeric complexes or binary subcomplexes were prepared and subjected to GST pull-down assays ( Figure 4C ) . Here , a Flag epitope tag was attached to the PB1 subunit , and formation of vRdRP subcomplexes , such as a trimeric complex ( lanes 1–3 ) and a binary complex of PB1-PB2 ( lanes 4–6 ) and of PB1-PA ( lanes 7–9 ) , was confirmed by immunoprecipitation with anti-Flag antibodies ( ‘Flag IP’ panels ) . Neither of the binary subcomplexes was pulled down with GST-IREF-2 ( ‘GST pull-down’ panels , lanes 5 , 6 , 8 , and 9 ) , whereas the interaction with IREF-2 could be observed for the vRdRP trimeric complex ( lanes 2 and 3 ) . A similar characteristic interaction property could be observed for each endogenous IREF-2 protein ( Figure 4—figure supplement 2 ) . Consistent with this observation , each IREF-2 protein was previously reported to be associated with the vRdRP trimeric complex , but not with the binary complex of PB1-PA ( Bradel-Tretheway et al . , 2011 ) . And interestingly , the PB1-PB2 subcomplex was also found not to be associated with IREF-2 , as shown here . This result strongly suggests that both the PA and the PB2 subunits in the trimeric complex of vRdRP are requisite for stable interaction with IREF-2 . Using a cell-free viral RNA synthesis system , we demonstrated that IREF-2 enables vRdRP to replicate vRNA from a cRNA template preferentially . Next , to show the function of IREF-2 in vivo , the effect of IREF-2 KD on viral RNA synthesis in infected cells was examined . By transfection with short RNAi duplexes targeting pp32 and APRIL , the expression levels of both proteins were strongly blocked ( Figure 5A ) . Both the control KD and the IREF-2 KD cells were infected with influenza virus , and the intracellular viral RNA level was measured . Upon KD of pp32 or APRIL , the levels of vRNA , cRNA , and viral mRNA decreased to about 50% to –80% of the control level in each infection period ( Figure 5B ) . By silencing both pp32 and APRIL ( ‘double-KD’ cells; see Figure 5A , lane 6 ) , each viral RNA level was additively decreased to about 25% to – 50% of the control level ( Figure 5B ) . These results suggest that IREF-2 , pp32 , and APRIL redundantly play a positive role in viral RNA synthesis in infected cells . Although IREF-2 was shown to preferentially regulate vRNA synthesis from a cRNA template in a cell-free system , the decrease patterns of the vRNA , cRNA , and viral mRNA syntheses were comparable in the IREF-2 KD cells . 10 . 7554/eLife . 08939 . 010Figure 5 . Effect of influenza virus replication factor-2 ( IREF-2 ) knockdown in infected cells . ( A ) Endogenous IREF-2 protein levels in knockdown ( KD ) HeLa cells . The whole-cell lysates of control KD cells ( lane 3 ) , pp32 KD cells ( lane 4 ) , APRIL KD cells ( lane 5 ) , and both pp32 and APRIL KD cells ( termed double-KD cells; lane 6 ) were subjected to SDS-PAGE followed by western blot analysis with antibodies , as indicated on the right side of the panel . To measure the KD levels , 10% and 30% of the lysates of the control KD cells were also subjected to SDS-PAGE followed by western blot analysis with antibodies , respectively ( lanes 1 and 2 ) . ( B ) Viral RNA levels in infected KD HeLa cells . Total RNA was extracted from mock-infected or infected KD cells at an multiplicity of infection ( MOI ) of 1 at 3 , 6 , and 9 hr post infection . vRNA , complementary RNA ( cRNA ) , and viral mRNA derived from segment 5 were quantitatively determined by RT-mediated qPCR ( RT-qPCR ) , as described in 'Materials and methods' . ( C ) Primary transcription level in infected KD HeLa cells . Total RNA was extracted from mock-infected or infected KD cells at an MOI of 10 for 3 hr in the presence or absence of cycloheximide ( CHX ) . Viral mRNA derived from segment 5 was quantitatively determined by RT-qPCR . ( D ) Effect of IREF-2 on progeny virus production . Control KD and IREF-2 double-KD A549 cells were infected with FluV ( WSN/33 strain ) at an MOI of 0 . 001 . The number of infectious progeny viruses produced from control KD and IREF-2 double-KD A549 cells at each time point were plotted ( upper panel ) , and the ratios of progeny viruses produced from IREF-2 double-KD A549 cells compared with those from control KD cells were also represented as a percentage of the control ( lower panel ) . Each quantitative result is presented as the average with the standard deviation from at least three independent experiments . Significance was determined using Student’s t test . n . s . : not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 01010 . 7554/eLife . 08939 . 011Figure 5—figure supplement 1 . Effect of influenza virus replication factor-2 ( IREF-2 ) on viral reporter RNA syntheses from a reconstituted model replicon . HEK293T of control and IREF-2 knockdown ( KD ) cells ( approximately , 3 × 105 cells in a 22-mm-diameter dish ) were transfected with expression plasmids pCAGGS-PB1 , -PB2 , -PA , and -NP ( 15 ng each ) . In addition , 1 . 5 ng of either phPolI-vNS-Luc ( for expression of vNS-Luc; left panel ) or phPolI-cNS-Luc ( for expression of cNS-Luc; right panel ) was co-transfected as a viral genomic source . At 24 hr post co-transfection , viral reporter levels in both the control KD and the IREF-2 KD cells were quantitatively determined by RT-qPCR as described previously ( Kawaguchi et al . , 2011 ) and in 'Materials and methods' . Quantification and standard deviations of the viral reporter RNA levels from IREF-2 KD cells expressed as a percentage of the value from the control KD cells . Each experiment was repeated three times . DOI: http://dx . doi . org/10 . 7554/eLife . 08939 . 011 We tried to address whether the effect of IREF-2 on vRNA and cRNA syntheses could be distinguished by using a mini-genome replicon system with either the vRNA reporter ( vNS-Luc ) or the cRNA reporter ( cNS-Luc ) as the source of the viral genome ( Figure 5—figure supplement 1 ) . In both templates , the levels of three kinds of viral RNAs were shown to be significantly decreased by IREF-2 KD as observed in the infected IREF-2 KD cells ( Figure 5B ) . On the basis of the assumption that IREF-2 preferentially stimulates vRNA synthesis rather than cRNA synthesis , the vRNA level is expected to be more reduced by IREF-2 KD than the cRNA level in these experiments . In the case of vNS-Luc as a viral genomic source , the vRNA level appeared to be slightly more decreased by IREF-2 KD ( 12% of the control ) when compared with the decrease in the cRNA level ( 23% of the control ) . Meanwhile , all viral reporter RNA levels were shown to be decreased almost comparably ( 46%-– 48% reductions ) when cNS-luc was used . From these results , the difference in the reduction level of vRNA and cRNA was insufficient to distinguish the effects of IREF-2 KD on vRNA and cRNA syntheses in these reporter experiments . This could be interpreted as follows: vRNA serves as a template for cRNA synthesis and mRNA transcription , and cRNA serves as a template for vRNA synthesis . Therefore , a defect in a certain specific step of viral RNA synthesis would influence the other steps and would result in declines in the accumulation levels of all viral RNA species , as observed in a previous study ( Maier et al . , 2008 ) . Next , we tried to address whether IREF-2 plays a role in transcription . To examine the effect of IREF-2 KD on primary transcription , we used a translation inhibitor , cycloheximide . In infected cells , cycloheximide prevents the synthesis of viral proteins required for the viral RNA replication process , resulting in the blockage of cRNA and vRNA syntheses . Therefore , only viral mRNA synthesis from incoming vRNP is detectable in the presence of cycloheximide . Control KD and pp32 and APRIL double-KD cells ( identical to lanes 3 and 6 in Figure 5A ) were infected at high multiplicity ( MOI of 10 ) for 3 hr in the presence or absence of cycloheximide . In the presence of cycloheximide , the accumulation level of the primary transcription appeared to be almost the same for the control KD and IREF-2 double-KD cells , while in the absence of cycloheximide , a significant reduction was observed in the IREF-2 double-KD cells ( Figure 5C ) . These results demonstrate that IREF-2 had no effect on the primary transcription , consistent with the result presented in Figure 3D showing that IREF-2 had no effect on the cap-snatching transcription in the cell-free reaction system . Thus , it is quite possible that IREF-2 stimulates viral RNA replication in vivo . Nevertheless the function of IREF-2 in cRNA synthesis and vRNA synthesis was not distinguishable in these experiments . Finally , the effect of IREF-2 KD on virus growth was examined by using human alveolar basal epithelial cells ( A549 cell line ) . Upon silencing both IREF-2 proteins ( IREF-2 double-KD ) , significant decrease in progeny virus production ( approximately , 90% of reduction at each time point ) could be observed ( Figure 5D ) . This seems to be attributable to the deficiency in the viral RNA replication process caused by IREF-2 KD .
In this study , we have identified IREF-2 consisting of pp32 and APRIL as host factors , which upregulate the second step of influenza viral RNA replication , that is , vRNA synthesis from a cRNA template . These host factors are well known as members of the ANP32 family and exhibit high homology with each other . Both are evolutionarily conserved in many organisms , including birds and swine , and are ubiquitously expressed in a variety of tissues , including lung tissue . Recently , an small-interfering RNA ( siRNA ) screening study suggested that APRIL is a positive regulator of viral RNA biogenesis ( Watanabe et al . , 2014 ) . In that report , the interaction between host proteins and viral factors singly expressed in mammalian cells was also examined , but interaction between APRIL with vRdRP subunit was not observed , and the authors speculated that the effect of APRIL on viral RNA synthesis might be indirect . However , we have shown that IREF-2 can interact efficiently with the trimeric complex of PB1 , PB2 , and PA , but not with any subcomplexes ( Figure 4C ) . Therefore , APRIL seems not to be co-precipitated with any singly expressed vRdRP subunit ( Watanabe et al . , 2014 ) . It is noteworthy that interaction between IREF-2 and the vRdRP trimeric complexes was observed in infected cells and cells cotransfected with all three subunit expression vectors , while interaction with the RNP complexes could not be observed in the infected cells ( Figure 4A ) . On the basis of these observations , the interaction between IREF-2 and vRdRP appears to be temporary during the process of cRNA template binding and vRNA synthesis , and once vRdRP starts vRNA synthesis , the associating IREF-2 protein may be released from vRdRP . The exact timing of the release of IREF-2 from the RNP complexes during replication , for example , at the initiation step of vRNA synthesis , during the transition step from initiation to elongation , or during elongation , remains to be solved . Cell-free analyses clearly demonstrated that IREF-2 is preferentially involved in vRNA synthesis from the cRNA template rather than in cRNA synthesis from the vRNA template ( Figures 3B , C ) . The accumulation level of viral RNA in infected IREF-2 KD cells was significantly decreased , but not completely abolished ( Figure 5B ) . This incomplete abolishment could be interpreted as due to the involvement of another unidentified host factor ( s ) in the viral RNA replication process . Recent studies have demonstrated that viral NS2 protein and virus-induced small leader RNA play important roles in the RNA replication process ( Perez et al . , 2010; 2012; Robb et al . , 2009 ) . Taken together , these findings suggest that the negative effect of IREF-2 KD on viral RNA replication might be compensated by other host and/or viral factor ( s ) and/or other regulation mechanism ( s ) and that viral replication at some limited level could still be observed in IREF-2 KD cells . Apparently , the function of IREF-2 is distinct from that of IREF-1/MCM ( Figure 3C ) , which plays a role in promoter clearance during viral RNA synthesis , but how and in which step IREF-2 is involved is unclear . Previously , Deng et al demonstrated that the initiation steps of vRNA synthesis and cRNA synthesis take place in different ways: vRNA synthesis is initiated in an ‘internal initiation and realignment’ manner , whereas cRNA synthesis is initiated at the 3’-terminus of the vRNA template ( Deng et al . , 2006b ) . The reported initiation mechanism of vRNA synthesis in this model describes that dinucleotide pppApG is internally synthesized de novo so as to be complementary to U4C5 ( in which the numbers indicate the nucleotide positions of the cRNA from the 3’-terminus and are denoted as ‘c3’U4C5’ in the discussion below ) . The 3’-terminal sequence of cRNA is composed of 3'-U1C2A3U4C5… -5' . Therefore , this internally synthesized pppApG would be realigned to the 3’-terminal nucleotide positions 1 and 2 ( denoted as ‘c3’U1C2’ ) of the template cRNA , followed by subsequent extension resulting in the full-length vRNA product . On the basis of this model ( Deng et al . , 2006b ) , the differential effect of IREF-2 on vRNA and cRNA synthesis ( Figure 3B ) leads to the hypothetical speculation that IREF-2 functions at an initiation process specific to vRNA synthesis using cRNA as the template . In another previous report by Zhang et al , unprimed vRNA synthesis appears to be initiated de novo at the c3’U4C5 position of the cRNA template but ‘realignment’ does not take place . And vRdRP extends a nascent RNA chain from the internal position , resulting in a short vRNA product lacking the first three nucleotides at the 5'-terminal sequence of the authentic vRNA product ( Zhang et al . , 2010a ) . In their cell-free vRNA synthesis system , such abnormal 3 nt-shorter vRNA was a major product , whereas full-sized vRNA was synthesized only to a limited extent at high concentrations of the UTP substrate . In our study , however , IREF-2-dependent vRNA products were shown to be exclusively full length . Taking these results together , we postulated that IREF-2 regulates the initiation step of vRNA synthesis as follows: ( 1 ) by preventing improper extension prior to pppApG realignment , ( 2 ) by facilitating the conformational change of vRdRP and/or the cRNA template to promote the realignment , and/or ( 3 ) by stabilizing the initiation-intermediate complex when pppApG is already realigned onto c3'U1C2 and ready for proper elongation . To address these possibilities , further functional studies are needed . In the two previous reports by Deng et al . and Zhang et al . , cell-free vRNA synthesis could be reproduced without any exogenously added primer and host proteins , while in our cell-free system , no vRNA synthesis occurred without any added primer and host proteins , as was observed previously ( Galarza et al . , 1996; Seong and Brownlee , 1992; Seong et al . , 1992 ) . They used recombinant vRdRP purified from plasmid-transfected mammalian cells ( Deng et al . , 2006b ) or baculovirus-infected insect cells ( Zhang et al . , 2010a; 2010b ) , while we used mainly mnRNP prepared by treatment of virion-derived vRNP complexes with MNase as the enzyme source . Therefore , the intrinsic nature of vRdRPs seems to differ one from the other , and recombinant vRdRP is possibly competent to initiate spontaneously de novo vRNA synthesis to some extent . Another difference among these enzyme sources is the presence of NP in the mnRNP preparation . At present , several lines of evidence suggest that NP plays a positive role ( s ) in regulation of vRdRP for replication and nascent chain elongation , but in the present study , unprimed vRNA synthesis employing mnRNP was not observed even in the presence of NP ( lanes 2 and 10 in Figure 1A , and elsewhere ) . On the other hand , in previously reported preparations of recombinant vRdRP , some host proteins were possibly contaminated ( Deng et al . , 2006a; Zhang et al . , 2010b ) , and an unexpected contribution ( s ) of such host protein contamination to their cell-free reactions could not be completely excluded . The other point for consideration is that the amount of vRdRP used in the cell-free reactions was different . In their cell-free reaction systems , a relatively large amount ( 100 nM ) of vRdRP was used for the reaction ( Zhang et al . , 2010a ) , while we used a much smaller amount of the enzyme source ( approximately 2–3 nM of vRdRP equivalent mnRNP ) in the reaction . Hence , spontaneous vRNA synthesis in the absence of any primer and host factor might have taken place to some extent in their cell-free system . Taking such differences in the quality and quantity of the enzyme sources into account , it is possible that vRdRP intrinsically possesses de novo initiation activity and that IREF-2 may increase the turnover efficiency so that the apparent concentration of vRdRP seems to be increased . At this point , one of the primary interests to be addressed is whether IREF-2 exhibits similar stimulation activity as observed here in the cell-free system using recombinant vRdRP . In addition , the relationship between IREF-2 and NP has also not been addressed because mnRNP was used mainly as an enzyme source throughout this study . Therefore , NP-free vRdRP ( i . e . , recombinant vRdRP ) seems to be a suitable enzyme source to address whether NP is necessary for the regulatory mechanism by IREF-2 . These differences in the quality and quantity of vRdRP must be taken into account for further study . Lastly , we have shown by cell-free studies that IREF-2 is involved in regulation of unprimed vRNA synthesis but not of cRNA synthesis ( Figures 3B , C ) . However , it is possible that our cell-free system might not be completely appropriate or sufficient for cRNA synthesis from a vRNA template , and thus we do not exclude the possibility that IREF-2 also functions in unprimed cRNA synthesis . If virion-derived vRdRP , including mnRNP , could be set up to be adapted to the transcriptional mode rather than to the ( cRNA ) replicative mode , some process might be required for neutralizing the transcriptional mode of vRdRP or switching the mode from ‘transcriptase’ to ‘replicase’ . Given this possibility , not only IREF-2 but also an additional factor ( s ) might be required for efficient unprimed cRNA synthesis in our cell-free system . To confirm this possibility , we are currently preparing recombinant vRdRP complexes possibly with the intrinsic property of influenza vRdRP as ‘replicase’ in infected cells for use in examining the effect of IREF-2 . In conclusion , we have demonstrated that the novel host-derived factors pp32 and APRIL regulate vRNA synthesis at least from the cRNA template . However , a number of questions about the function of IREF-2 remain unanswered . It would be important and beneficial to use recombinant vRdRP with high-purity and an authentic nature as an enzyme source in further detailed studies . In addition , the crystal structure of influenza vRdRP was recently resolved ( Pflug et al . , 2014; Reich et al . , 2014 ) , and therefore , the structure-based mechanism of IREF-2 will likely be elucidated in future studies .
Cells from the human cervical carcinoma HeLa S3 cell line were cultured in a spinner flask with Eagle’s minimal essential medium for suspension cultures ( S-MEM; Sigma , St . Louis , MO ) containing 10% calf serum . The medium was added to the suspension cell culture to maintain a density of 2–6 × 105 cells/ml . When the volume of the suspension cell culture reached 10 liters , the cells were collected by centrifugation and used for preparation of the NEs . HEK293T and A549 cells were grown at 37°C in Dulbecco’s modified Eagle’s medium ( DMEM; Nissui , Japan ) containing 10% fetal calf serum , and monolayer HeLa cells were cultured in Eagle’s minimal essential medium ( MEM; Sigma ) containing 10% fetal calf serum , termed as MEM ( + ) . Influenza virus A/PR/8/34 ( H1N1 ) was grown in the allantoic sacs of 10-day-old embryonated eggs . The viruses were purified from infected allantoic fluid as described previously ( Kawakami et al . , 1981 ) and used for vRNP preparation . Anti-pp32 ( goat polyclonal; Santa Cruz Biotechnology , Dallas , TX ) , anti-APRIL ( goat polyclonal; Abcam limited , UK ) , and anti-GST ( mouse polyclonal; Nacalai Tesque , Japan ) antibodies were commercially purchased . For detection of viral proteins , rabbit polyclonal anti-PB1 , -PB2 , -PA , and -NP antisera were prepared as previously described ( Momose et al . , 2001; 2002 ) and used for western blot analysis . vRNP complexes were prepared from the purified influenza A/PR/8/34 virus as described previously ( Shimizu et al . , 1994 ) . mnRNP was prepared by incubation of vRNP complexes at 30°C for 2 hr with one unit of micrococcal nuclease ( Roche Molecular Biochemicals , Schweiz ) /μl in the presence of 1 mM CaCl2 . The nuclease reaction was terminated by addition of EGTA to a final concentration of 3 mM . The 53 nt-long viral model RNAs ( v53; 5'-AGUAGAAACAAGGGUGUUUUUUCAUAUCAUUUAAACUUCACCCUGCUUUUGCU-3' and c53; 5'-AGCAAAAGCAGGGUGAAGUUUAAAUGAUAUGAAAAAACACCCUUGUUUCUACU-3' ) were synthesized by transcription with RiboMAX Large Scale RNA Production System-T7 ( Promega , Madison , WI ) using a synthetic DNA template as previously described ( Shimizu et al . , 1994 ) . Cell-free viral RNA synthesis was carried out at 30°C in a final volume of 20 μl or 25 μl in the presence of 50 mM HEPES-NaOH ( pH 7 . 9 ) , 3 mM MgCl2 , 50 mM KCl , 1 mM DTT , 500 μM each of ATP , UTP , and CTP and 25 μM GTP , 5 μCi of [α-32P] GTP ( 3000 Ci/mmol ) , 8 U of RNase inhibitor from human placenta ( Toyobo , Japan ) , 10 ng of a 53-nt-long model viral RNA template of negative or positive polarity ( v53 and c53 , respectively ) and approximately 40 ng NP-equivalent , alternatively 5 ng PB1-equivalent of mnRNP as an enzyme source . After incubation , the reactions were terminated by phenol/chloroform extraction followed by precipitation of the RNA products with ethanol . The precipitated materials were subjected to polyacrylamide gel electrophoresis in the presence of urea ( Urea-PAGE ) and visualized by autoradiography . All procedures for preparation and fractionation of the NEs were carried out at 4°C or on ice . The details for purification is described at Bio-protocol ( Sugiyama and Nagata , 2016 ) . Briefly , uninfected HeLa cell NE was prepared as described previously ( Dignam et al . , 1983 ) . For fractionation of the NE , a buffer ( buffer H ) containing 50 mM HEPES-NaOH ( pH 7 . 9 ) , 20% ( v/v ) glycerol , and 1 mM DTT in the presence of the appropriate concentrations of KCl was used for ion-exchange chromatography for purification of IREF-2 . The purification scheme started with NE containing 15 mg protein . NE was loaded onto a phosphocellulose column ( 10-ml bed volume; Whatman P11 ) equilibrated with buffer H containing 50 mM KCl and successively eluted stepwise with buffer H containing 0 . 2 M , 0 . 5 M , and 1 . 0 M KCl . An unbound fraction ( P0 . 05 ) and fractions eluted with 0 . 2 M ( P0 . 2 ) , 0 . 5 M ( P0 . 5 ) , and 1 . 0 M ( P1 . 0 ) of KCl were examined by the cell-free viral RNA synthesis system . The remaining P0 . 05 fraction was loaded onto an anion exchanger Uno-Q column ( Bio-Rad , Hercules , CA ) equilibrated with buffer H containing 50 mM KCl and eluted stepwise with increasing concentrations of KCl . The IREF-2 activity fraction eluted with buffer H containing 0 . 6 M KCl was diluted with buffer H and again loaded onto the Uno-Q column equilibrated with buffer H containing 0 . 25 M KCl . The IREF-2 activity was eluted with a linear gradient of 0 . 25 M to –0 . 8 M KCl in buffer H . The concentrated IREF-2 fraction was diluted two-fold with an equal volume of a buffer containing 50 mM HEPES-NaOH ( pH 7 . 9 ) , 1 mM DTT , and 2 . 0 M ( NH4 ) 2SO4 and loaded onto a hydrophobic SOURCE-PHE column ( GE Healthcare , Piscataway , NJ ) equilibrated with a buffer containing 50 mM HEPES-NaOH ( pH 7 . 9 ) , 10% ( v/v ) glycerol , 1 mM DTT , and 1 . 0 M ( NH4 ) 2SO4 . The IREF-2 activity was unbound to the hydrophobic column and dialyzed with buffer H containing 30 mM KCl . The dialyzed IREF-2 fraction was loaded onto a cation exchanger ( Uno-S , Bio-Rad ) , and the IREF-2 activity was recovered in the fraction unbound to the column . Finally , the IREF-2 fraction was load onto an anion axcahnger ( Mono-Q column; GE Healthcare ) , and the materials captured by the column were eluted with a linear gradient of 0 . 35–0 . 7 M KCl in buffer H . For the RNase T2 protection assay , one-third of [α-32P] GTP-radiolabeled total RNA products synthesized by the cell-free viral RNA synthesis reactions were hybridized with either a nonradiolabeled v53 or a c53 probe by incubation at 85°C for 10 min , followed by incubation at 60°C for 10 min in a hybridization buffer ( 40 mM PIPES-NaOH [pH 6 . 4] , 1 mM EDTA , 0 . 4 M NaCl , and 80% formamide ) and gradual cooling down to room temperature ( Shimizu et al . , 1994 ) . Each hybridized RNA product was subjected to RNase T2 digestion ( five units in 20 mM NaOAc [pH 5 . 2] ) at 25°C for 1 hr . Digestion was terminated by phenol/chloroform extraction , and the digested materials were collected by ethanol precipitation . The precipitated materials were subjected to 10% Urea-PAGE and visualized by autoradiography . At the same time , the remaining one-third of the total RNA products was also subjected to Urea-PAGE as an undigested sample . For analysis of the 5'’-terminus of the unprimed IREF-2-dependent RNA products , RNase digestion and thin-layer chromatography were performed as described previously ( Kawaguchi and Nagata , 2007 ) . After purification by 10% Urea-PAGE and elution from the gel piece , the purified [α-32P] GTP-radiolabeled RNA products were digested with 15 units of RNase T2 in 50 mM NaOAc ( pH 5 . 0 ) , 100 mM NaCl , and 2 mM EDTA , or 0 . 5 units of SV-PDE in 50 mM Tris-HCl ( pH 9 . 0 ) , 100 mM NaCl , and 14 mM MgCl2 at 37°C for 1 hr . The digested products were spotted onto a PEI-cellulose thin layer ( Merck , Germany ) and developed with 1 . 6 M LiCl or 1 N acetic acid-4 M LiCl ( 4:1 , v/v ) and visualized by autoradiography . In the case of alkaline phosphatase treatment , [α-32P] GTP-labeled product was incubated with 0 . 5 units of bacterial alkaline phosphatase in a buffer recommended by the manufacturer ( Toyobo ) at 37°C for 1 hr before nuclease digestion . For mobility standards , nonradioactive adenosine monophosphate ( AMP ) , adenosine diphosphate ( ADP ) , and adenosine triphosphate ( ATP ) were subjected to thin-layer chromatography . For a marker of adenosine 5'-triphosphate and 3'-monophosphate ( pppAp ) , [γ-32P] ATP-labeled v53 synthesized with T7 RNA polymerase was also subjected to gel purification , nuclease digestion , and thin-layer chromatography as described above . Transfected and infected HEK293T cells were harvested with a rubber policeman and washed with PBS ( - ) . The collected cells were suspended in an ice-cold lysis-binding buffer ( 50 mM Hepes-NaOH [pH 7 . 9] , 100 mM NaCl , 50 mM KCl , 0 . 25% NP-40 , and 1 mM DTT ) and lysed by brief sonication . After centrifugation , the crude lysates were incubated with Glutathione-Sepharose 4B resin ( GE Healthcare ) at 4°C for 1 hr . After incubation , the resins were collected by brief centrifugation and washed three times with the lysis-binding buffer . The resin-bound materials were eluted by boiling in the SDS-PAGE loading buffer and subjected to SDS-PAGE , which was followed by detection with the standard western blot analysis procedure . Total RNA from infected HeLa cells or transfected 293T cells was extracted using Sepasol-RNA I Super G ( Nacalai Tesque ) according to the manufacturer’s instructions . The extracted RNA was treated with DNase I and subjected to RT reaction as follows . For synthesis of cDNA derived from segment 5 viral RNAs , each specific primer was used in the RT reaction , as follows: 5'-GACGATGCAACGGCTGGTCTG-3' ( complementary to the 1122 nt-– 1142 nt region from the 5'-terminus ) for the vRNA of segment 5; 5'-TCATCTTTGTTCCTCAA-3' ( 3' terminal region ) for the cRNA of segment 5; and oligo-dT20 ( 5'-TTTTTTTTTTTTTTTTTTTT-3' ) for the viral mRNA . The RT reactions were performed using the hot-start protocol as described elsewhere ( Kawakami et al . , 2011 ) . Total RNA ( 50–100 ng ) and the aforementioned specific primer ( 10 pmol ) were heated in a volume of 5 . 5 μl at 65°C for 5 min , chilled immediately in an ice-water bath , and then preincubated at 48°C . The RT premixture ( 3 μl First Strand buffer [5× , Invitrogen , Carlsbad , CA]; 0 . 75 μl of 0 . 1 M DTT; 0 . 75 μl of 10 mM dNTP mixture; 0 . 25 μl of Superscript III Reverse Transcriptase [200 U/μl , Invitrogen]; 0 . 25 μl RNase inhibitor; and 4 . 5 μl of saturated trehalose [Sigma] in a total of 9 . 5 μl ) was also preincubated at 48°C and quickly added to the RNA-primer mixture and incubated at 48°C for 1 hr . The RT reaction mixture was diluted ten-fold with H2O and incubated at 95°C to terminate the RT reaction . The diluted cDNA sample ( 3 μl ) was subjected to quantification by real-time PCR using the FastStart SYBR Green Master ( Roche ) and the segment 5-specific primer set ( 5 pmol each/15 μl of the qPCR reaction mix ) , 5’-GACGATGCAACGGCTGGTCTG-3' and 5’-AGCATTGTTCCAACTCCTTT-3' . Real-time PCR was performed using the Thermal Cycler Dice Real Time System TP800 ( Takara , Japan ) . Growth curve analysis was performed by infecting confluent A549 cells ( 96 hr after transfection of siRNA ) in 35-mm dishes with FluV/A/WSN/33 virus at an MOI of 0 . 001 and harvesting the supernatant every 12 hr after infection for 60 hr . The virus titers present in the supernatant were determined by a standard plaque assay using MDCK cells . Full-length cDNA of human IREF-2α/pp32 was amplified using KOD polymerase ( Toyobo ) from a cDNA library prepared from HeLa cell mRNA using the specific primers 5'-CGCGGATCCCATATGGAGATGGGCAGACGGATTCATTTAG-3' and 5'-GCGGCTCGAGACGTCAGTCATCATCTTCTCCCTCATCTTCAGGTTCTCGT-3' , corresponding to the pp32 amino-terminal and carboxyl-terminal regions , respectively . The full-length cDNA of human IREF-2β/APRIL was also amplified by using the specific primers 5'-CGGAATTCATATGGACATGAAGAGGAGGATCCACCTGGAG-3' and 5'-GCGGCTCGAGACGTCAATCATCTTCTCCTTCATCATCTGT-3' , corresponding to the APRIL amino-terminal and carboxyl-terminal regions , respectively . To construct the glutathione S-transferase ( GST ) -tagged pp32 expression vectors for Escherichia . coli cells ( pGEX-2T-pp32 ) , the amplified cDNA fragment of pp32 was digested with BamHI and AatII and then cloned into pGEX-2T ( GE Healthcare ) predigested with the same restriction enzymes . To construct E . coli expression vectors for GST-tagged APRIL ( pGEX-2T-APRIL ) , the amplified cDNA fragment of APRIL was digested with EcoRI and AatII and then cloned into pGEX-2T predigested with the same restriction enzymes . To construct mammalian expression vectors expressing GST harboring the nuclear localization signal ( NLS ) at the C-terminal end , termed pCAGGS-GSTnls , DNA fragments were amplified by PCR using the specific primer set 5'-CCCTCGAGCTCGCGGCCGCCGCCATGGGCTCCCCTATACTAGGTTATTGG-3' ( the sequence responsible for the ‘Kozak translational consensus’ is underlined ) and 5’-AAGATCTATGCATGGTACCGCTAGCGACGTCACCCGGGGTCTTCTACC-3' , and pGEX-2T-SV40 NLS-GFP ( kindly gifted by Dr Yoneda , Osaka University ) as the PCR template . The PCR products corresponding to Kozak-GSTnls were trimmed by digestion with XhoI and BglII and then cloned into the predigested pCAGGS mammalian expression vector , resulting in pCAGGS-GSTnls . To construct mammalian expression vectors expressing GST-tagged IREF-2s , pGEX-2T-pp32 and pGEX-2T-APRIL were digested with SwaI ( for cutting within the GST coding region ) and AatII ( for cutting downstream of the termination codon of both IREF2s ) , and each DNA fragment corresponding to GSTC-term-IREF-2 was subcloned into the pCAGGS-GSTnls predigested with SwaI and AatII , resulting in pCAGST-pp32 and pCAGST-APRIL . To construct pCAGST-RAF2p48 expressing GST-fused RAF2p48/BAT1/UAP56 , a DNA fragment corresponding to the GST-RAF2p48-encoding region derived from pGEX-p48 was subcloned into the pCAGGS vector , resulting in pCAGST-RAF2p48 . The nucleotide sequences of all plasmids constructed for this study were confirmed by DNA sequencing ( ABI prism genetic analyzer; Applied Biosystems , Foster City , CA ) . To express recombinant IREF-2 proteins , E . coli BL21 ( DE3 ) cells harboring the expression vectors ( pGEX-2T-pp32 and pGEX-2T-APRIL ) for GST-tagged IREF-2s were cultivated at 30°C until A590 reached 0 . 5 , and protein production was then induced with 1 mM isopropyl-1-thio-β-D-galactopyranoside for 6 hr . The induced cells were harvested by centrifugation and resuspended in a lysis buffer ( 50 mM Hepes-NaOH [pH 7 . 9] , 150 mM KCl , 0 . 25% NP-40 , and 1 mM DTT ) and sonicated well in an ice-water bath . The insoluble material was removed by centrifugation , and crude extracts were recovered as a supernatant fraction was applied to Glutathione-Sepharose 4B resin that had been equilibrated with the lysis buffer . After capturing of the GST-tagged proteins to the resin , lysis buffer was applied to wash away the unbound materials . The washed resins were equilibrated with a digestion buffer ( 50 mM Hepes-NaOH [pH 7 . 9] , 50 mM KCl , 1 mM CaCl2 , and 1 mM DTT ) and digested with thrombin ( Nacalai Tesque ) at 4°C overnight . The materials with the GST portion removed were collected and further purified through a Mono-Q column with KCl linear gradient elution . The purified proteins were dialyzed with buffer H containing 1 mM DTT , and the concentration of each protein was determined by the Bradford method and comparison of band intensities with a standard protein ( BSA ) after separation by SDS-PAGE . For transfection of HEK293T cells with mammalian expression plasmids , plasmid DNA mixtures were prepared in Opti-MEM ( Thermo Fisher Scientific , Waltham , MA ) , and an appropriate amount of PEI ( Sigma Aldrich , St . Louis , MO ) were mixed into the DNA solution ( 1 . 5 μg PEI per 1 μg DNA ) and incubated at room temperature for 10 min . The DNA/PEI complex was added to monolayer cell cultures ( approximately , 3 × 106 cells in a 100-mm-diameter dish ) . Six hours after transfection , the medium was replaced with fresh DMEM and maintained at 37°C . For transfection of siRNAs , trypsinized cells ( HeLa , HEK293T , and A549 ) were seeded , and pp32 siRNA ( custom-designed Stealth RNAi; Invitrogen ) , APRIL siRNA ( ANP32B-HSS116202; Invitrogen ) , and negative control siRNA ( 12935–200; Invitrogen ) were introduced into the cells with Lipofectamine RNAiMAX ( Invitrogen ) according to the manufacturer's protocol . Twelve to twenty-four hours after the first transfection , the cells were again transfected with the same amount of siRNA used at the first transfection and maintained at 37°C for 4 to 5 days until influenza virus infection . For virus infection , monolayer cultures of human cells ( HEK293T , HeLa , and A549 ) were washed with serum-free MEM , and the cells were infected with influenza A/PR/8 virus allantoic fluid at the desired MOI as described in each figure legend . After virus adsorption at 37°C for 1 hr , the infecting allantoic fluid was removed , and the cells were washed with serum-containing medium and maintained at 37°C in the medium for an appropriate period ( 3–9 hr ) . | The influenza or “flu” virus infects millions of people each year , with young children and elderly individuals most vulnerable to infection . The influenza virus stores its genetic material in the form of segments of single-stranded viral RNA . After the virus infects a cell , it replicates this genetic material in a two-part process . First , an enzyme made by the virus – called RNA polymerase – uses the viral genomic RNA as a template to form a “complementary” RNA strand ( called cRNA ) . This cRNA molecule is then itself used as a template to make more viral genomic RNA strands , which can go on to form new viruses . Exactly how viral genomic RNA is made from cRNA is poorly understood , although previous research had suggested that this process may also involve proteins belonging to the invaded host cell . However , these host proteins had not been identified . By mixing virus particles with extracts from uninfected human cells , Sugiyama et al . have now found that two host proteins called pp32 and APRIL help viral genomic RNA to form from a cRNA template . Both of these proteins directly interact with the viral RNA polymerase . Sugiyama et al . then reduced the amounts of pp32 and APRIL in human cells that were infected with the influenza virus . Much less viral genomic RNA – and so fewer new virus particles – formed in these cells than in normal cells . Further work is now needed to understand how the pp32 and APRIL proteins interact with viral RNA polymerase . This could eventually lead to the development of new treatments for influenza . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | pp32 and APRIL are host cell-derived regulators of influenza virus RNA synthesis from cRNA |
Complexin regulates spontaneous and activates Ca2+-triggered neurotransmitter release , yet the molecular mechanisms are still unclear . Here we performed single molecule fluorescence resonance energy transfer experiments and uncovered two conformations of complexin-1 bound to the ternary SNARE complex . In the cis conformation , complexin-1 induces a conformational change at the membrane-proximal C-terminal end of the ternary SNARE complex that specifically depends on the N-terminal , accessory , and central domains of complexin-1 . The complexin-1 induced conformation of the ternary SNARE complex may be related to a conformation that is juxtaposing the synaptic vesicle and plasma membranes . In the trans conformation , complexin-1 can simultaneously interact with a ternary SNARE complex via the central domain and a binary SNARE complex consisting of syntaxin-1A and SNAP-25A via the accessory domain . The cis conformation may be involved in activation of synchronous neurotransmitter release , whereas both conformations may be involved in regulating spontaneous release .
Ca2+-triggered fusion of synaptic vesicles is orchestrated by synaptic proteins , including neuronal SNAREs ( soluble N-ethylmaleimide-sensitive factor attachment protein receptor ) , the Ca2+-sensor synaptotagmin , complexin , and other components ( Südhof , 2013 ) . This work focuses on the cytoplasmic protein complexin and its interactions with SNAREs . Complexin has multiple functions: it activates Ca2+-triggered synchronous release , regulates spontaneous release , and increases the primed pool of synaptic vesicles ( Mohrmann et al . , 2015; Trimbuch and Rosenmund , 2016 ) . Here we summarize some of the key properties of complexin relevant to this work . Synchronous Ca2+-triggered neurotransmitter release depends critically on complexin ( McMahon et al . , 1995 ) , and this activating role of complexin is conserved across all species and different types of Ca2+-induced exocytosis studied to date ( Reim et al . , 2001; Huntwork and Littleton , 2007; Xue et al . , 2008; Maximov et al . , 2009; Hobson et al . , 2011; Martin et al . , 2011; Kaeser-Woo et al . , 2012; Cao et al . , 2013; Yang et al . , 2013 , 2015 ) . Complexin also regulates spontaneous release , although this effect is less conserved among species and experimental conditions: for example , in Drosophila spontaneous release increases with knockout of complexin ( Xue et al . , 2009; Jorquera et al . , 2012 ) . Likewise , knockdown in cultured cortical neurons increases spontaneous release , although knockout of complexin in mice differently affects spontaneous release depending on the particular neuronal cell type ( Maximov et al . , 2009; Kaeser-Woo et al . , 2012; Yang et al . , 2013; Trimbuch et al . , 2014 ) . We study here the homolog complexin-1 of the mammalian complexin family whose primary sequence is highly conserved ( 96% ) in mouse , rat , and human . Complexin-1 binds with high affinity ( ~10 nM ) to the neuronal ternary SNARE complex consisting of synaptobrevin-2 , syntaxin-1A , and SNAP-25A ( McMahon et al . , 1995; Chen et al . , 2002; Pabst et al . , 2002 ) . Complexin consists of four domains ( Figure 1A ) : the N-terminal domain ( amino acid range 1–27 ) is required for activation of synchronous Ca2+-triggered release , the accessory domain ( amino acid range 28–48 ) is required for regulation of spontaneous release , the central domain ( amino acid range 49–70 ) is required for all functions of complexin-1 , and the C-terminal domain ( amino acid range 71–134 ) is involved in vesicle priming and binds to anionic membranes with curvature sensitivity ( Chen et al . , 2002; Maximov et al . , 2009; Kaeser-Woo et al . , 2012; Snead et al . , 2014 ) . The key roles of the four domains of complexin-1 have been reproduced in a reconstituted single vesicle fusion system with neuronal SNAREs , synaptotagmin-1 , and complexin-1 ( Lai et al . , 2014 ) . By varying the complexin-1 concentration , this in vitro study suggested that the regulatory effect on spontaneous fusion and the activating role on Ca2+-triggered release are governed by distinct molecular mechanisms involving different subsets of the four domains of complexin-1 . Recent in vitro experiments revealed that the accessory domain is entirely dispensable for activation of Ca2+-triggered synaptic vesicle fusion , although it is essential for regulation of spontaneous release ( Lai et al . , 2016 ) . Similarly , genetic experiments in Drosophila also suggest distinct mechanisms for activation of Ca2+-triggered release and regulation of spontaneous release ( Cho et al . , 2014 ) . 10 . 7554/eLife . 16886 . 003Figure 1 . Domain diagram of complexin-1 , mutants , and truncations . ( A ) Domain diagram for full-length wildtype complexin-1 ( Cpx [1–134] ) and complexin-1 mutants ( SC , superclamp; NC , no-clamp; 4M , mutation of the central helix that prevents binding to ternary SNARE complex ) . ( B ) Domain diagrams for truncation mutants of complexin-1 ( Cpx [26–134] , Cpx [48–134] ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 003 The central domain of complexin-1 binds to the groove formed by the synaptobrevin-2 and syntaxin-1A α-helices in the complexin-1 / ternary SNARE supercomplex ( PDB ID 1KIL ) ( Chen et al . , 2002 ) . This interaction is antiparallel , i . e . , the direction of the α-helix of the accessory domain of complexin-1 is antiparallel to the direction of the α-helices in the ternary SNARE complex . Functional studies , along with molecular modeling suggested that complexin-1 may inhibit full ternary SNARE complex formation by preventing the C-terminal part of synaptobrevin-2 from binding to the syntaxin-1A / SNAP-25A components of the ternary SNARE complex ( Giraudo et al . , 2006 , 2009 ) . However , a subsequent crystal structure of the 'superclamp' mutant of complexin-1 ( D27L , E34F , R37A ) ( Figure 1A ) in complex with a partially truncated SNARE complex ( containing a truncated synaptobrevin-2 fragment , amino acid range 25–60 ) , along with isothermal titration calorimetry ( ITC ) and light scattering experiments instead found that complexin-1 bridges two partially truncated SNARE complexes: one partial SNARE complex binds to the central domain of complexin-1 while another partial SNARE complex weakly binds to the accessory domain of the same complexin-1 molecule ( PDB IDs 3RK3 and 3RL0 ) ( Kümmel et al . , 2011; Krishnakumar et al . , 2015 ) . However , this weak interaction of the accessory domain was not observable by NMR ( Trimbuch et al . , 2014 ) , and its relevance has been debated ( Trimbuch et al . , 2014; Krishnakumar et al . , 2015 ) . Regardless of the ongoing studies of the biophysical interactions involving the accessory domain of complexin-1 , this domain is essential for regulation of spontaneous fusion both in neurons and in reconstituted systems . For example , in rescue experiments with complexin knockdown in cultured neurons , the complexin-1 superclamp mutant slightly decreased the mEPSC ( miniature excitatory postsynaptic current ) frequency , while a 'poor-clamp' mutant ( K26E , L41K , E47K ) of complexin increased the mEPSC frequency compared to wildtype control ( Yang et al . , 2010 ) . Similarly , genetic rescue experiments in Drosophila showed that the superclamp mutant decreased spontaneous release , whereas the no-clamp mutant ( L41E , A44E , A30E , A31E ) ( Figure 1A ) slightly increased spontaneous release ( Cho et al . , 2014 ) compared to wildtype control . Moreover , introduction of a helix breaking mutation into the accessory domain of complexin increased spontaneous release in C . elegans ( Radoff et al . , 2014 ) . Different models have been proposed to explain the regulatory function of the accessory domain of complexin on spontaneous release: first , the accessory domain inserts itself into the C-terminal half of a partially assembled SNARE complex and , thus , transiently inhibits full assembly ( Giraudo et al . , 2006 , 2009 ) ; second , the accessory domain inserts itself into another SNARE complex , bridging two SNARE complexes ( Kümmel et al . , 2011; Krishnakumar et al . , 2015 ) ; third , the accessory domain is not involved in protein interactions , but instead acts by electrostatic repulsion with the membrane ( Trimbuch et al . , 2014 ) ; fourth , the accessory domain stabilizes the α-helix of the central domain of complexin ( Radoff et al . , 2014 ) . In order to gain more insights into the molecular mechanisms of complexin-1 , we studied the complexin-1 / ternary SNARE supercomplex using single molecule fluorescence resonance energy transfer ( smFRET ) efficiency experiments . We observed two conformations of complexin-1 when bound to the ternary SNARE complex that we refer to as cis and trans . In the cis conformation , the accessory domain cooperates with the N-terminal domain of complexin-1 to induce a conformational change at the membrane-proximal C-terminal end of the bound ternary SNARE complex . In the trans conformation , complexin-1 bridges two SNARE complexes: a ternary SNARE complex and a binary SNARE complex , consisting of syntaxin-1A and SNAP-25A ( also called t-SNARE complex ) , requiring the presence of both the central and accessory domains .
As a prerequisite to study the interaction between complexin-1 and ternary SNARE complex , we wanted to ensure proper assembly of the ternary SNARE complex . SNARE complexes were sequentially assembled on a microscope slide ( Figure 2A and Materials and methods ) . The surface of the microscope slide was passivated using biotinylated BSA and a phospholipid bilayer to prevent non-specific binding and to ensure that the tethered proteins are isolated single molecules in an environment surrounded by lipids . Single molecules consisting of the cytoplasmic domain of syntaxin-1A were tethered to the deposited bilayer through a biotin-streptavidin linkage , leading to a more uniform distribution of the tethered proteins on the surface compared to when using reconstituted full-length syntaxin-1A ( Choi et al . , 2012 ) . SNAP-25A and the cytoplasmic domain of synaptobrevin-2 were then added sequentially to assemble the ternary SNARE complex . 10 . 7554/eLife . 16886 . 004Figure 2 . The dN-SB method achieves proper assembly of the ternary SNARE complex . ( A ) Left panel: schematic of the dN-SB method for assembly of the ternary SNARE complex . The cytoplasmic domain of syntaxin-1A ( SX , colored red ) was surface-tethered through biotin-streptavidin ( orange dot ) linkage to a passivated microscope slide . Next , SNAP-25A ( S25 , colored green ) was added . Subsequently , the cytoplasmic domain of synaptobrevin-2 ( SB , colored blue ) was added . For the dN-SB method , 10 μM dN-SB fragment was added concurrently with SB to the surface-tethered SX-S25 complex . Unbound proteins were washed away before smFRET measurements . Both SX and SB were labeled with fluorescent dyes ( the SNARE label pair SFC2 is indicated by the yellow dots , SX 249 and SB 82 ) . Right panel: properly assembled ternary SNARE complex is expected to produce high FRET efficiency for the SFC2 label pair . ( B ) Location of three SNARE label pairs ( SFC1 , SFC2 , SFC3 ) in the crystal structure of the ternary SNARE complex ( PDB ID: 1SFC ) , as defined in the legend . Separate experiments were performed for each of the three label pairs . Labeling of the two sites of a particular pair was performed separately with FRET donor and acceptor dyes ( Alexa 555 and Alexa 647 , respectively , Figure 2—source data 1 ) and the ternary SNARE complex was formed using the dN-SB method . The analysis was restricted to cases where FRET was observed , i . e . , complexes that contain one donor and one acceptor dye . ( C , D ) smFRET efficiency histograms for the SNARE label pairs SFC1 , SFC2 , SFC3 for the surface-tethered ternary SNARE complex that was formed in the absence ( C ) and presence ( D ) of the dN-SB fragment . ( E ) Summary bar chart of the histograms shown in panels D , E , illustrating the effect of the dN-SB method in suppressing improper subconfigurations between SX and SB during the assembly of the ternary SNARE complex . “% other subconfigurations ( SX:SB ) ” is calculated as the ratio of the areas under the two Gaussian functions that are fit to the low and high FRET efficiency states in the corresponding smFRET efficiency histograms , respectively . Shown are mean values ± SD for the two subsets of an equal partition of the data that are comprised of the observed FRET efficiency values for all molecules for each different condition ( see data summary table in Figure 2—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 00410 . 7554/eLife . 16886 . 005Figure 2—source data 1 . Data summary table for the results shown in Figure 2E . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 005 In order to assess the proper assembly of ternary SNARE complexes , we designed three FRET label pairs ( referred to as SFC1 , SFC2 , SFC3 ) with donor and acceptor fluorescent dyes attached to synaptobrevin-2 and syntaxin-1A ( Figure 2B ) , respectively . These SNARE label pairs probe the syntaxin-1A / synaptobrevin-2 subconfiguration within the ternary SNARE complex at both ends of the complex . These three label pairs are expected to produce high FRET efficiency for properly assembled ternary SNARE complex based on the crystal structure ( PDB ID 1SFC ) ( Sutton et al . , 1998 ) . As expected , the smFRET histograms of the three label pairs showed the expected high FRET states , but they also revealed low FRET efficiency states ( Figure 2C ) . Nearly 50 percent of the complexes exhibited low FRET efficiency when integrating the corresponding broad peaks ( Figure 2E ) . Such low FRET efficiency states had been observed in previous single molecule studies: they are related to improperly assembled SNARE complexes , including antiparallel syntaxin-1A / synaptobrevin-2 subconfigurations , when mixed in the absence of other factors that assist proper SNARE complex formation ( Weninger et al . , 2003; Lou et al . , 2015; Ryu et al . , 2015 ) . In order to obtain more uniform ternary SNARE complexes , we added the C-terminal fragment of the cytoplasmic domain of synaptobrevin-2 ( referred to as 'dN-SB' , amino acid range 49–96 of synaptobrevin-2 ) during ternary SNARE complex formation ( Figure 2A ) . This method had been previously used to obtain more efficient lipid mixing between SNARE-containing liposomes by incubating it prior to trans SNARE complex formation ( Pobbati et al . , 2006; Hernandez et al . , 2012 ) . In earlier work , a similar peptide ( Vc peptide , amino acid range 57–92 of synaptobrevin-2 ) had also been shown to stimulate lipid mixing ( Melia et al . , 2002 ) . We assembled ternary SNARE complexes in the presence of dN-SB ( Figure 2A , referred to as 'dN-SB method' in the following ) . Remarkably , this method nearly eliminated the low FRET efficiency states and the resulting smFRET histograms consisted mainly of one high FRET efficiency state ( Figure 2D , E ) . Our single molecule experiments now provide an explanation of why C-terminal peptides of synaptobrevin produce more SNARE-dependent vesicle fusion when incubating binary ( syntaxin-1A / SNAP-25A ) SNARE complexes with such peptides prior to trans SNARE complex formation ( Melia et al . , 2002; Pobbati et al . , 2006; Hernandez et al . , 2012 ) : the presence of the dN-SB fragment during ternary SNARE complex assembly prevents improper syntaxin-1A / synaptobrevin-2 subconfigurations which are probably not fusogenic . We used this dN-SB method for the assembly of ternary SNARE complexes in all subsequent smFRET experiments described in this work . The two crystal structures of the complexin-1 / ternary SNARE supercomplex ( Chen et al . , 2002; Kümmel et al . , 2011 ) suggest that the accessory domain projects away from the SNARE complex at an angle such that it does not interact with the same SNARE complex . Ensemble FRET efficiency experiments also indicated that the accessory domain projects away from the SNARE complex ( Krishnakumar et al . , 2011; Kümmel et al . , 2011 ) . However , these previous experiments may not have revealed the full conformational space that is available to the complexin-1 accessory domain due to ensemble averaging . In order to fully explore the conformational space of complexin-1 when bound to SNARE complex , we performed smFRET efficiency experiments with fluorescent labels attached to both complexin-1 and the ternary SNARE complex ( Figure 3A ) . The ternary SNARE complex was assembled as described in Figure 2A using the dN-SB method , starting with the surface-tethered cytoplasmic domain of syntaxin-1A alone , which was labeled with FRET acceptor dye . Unlabeled SNAP-25A and the cytoplasmic domain of synaptobrevin-2 were added sequentially to assemble the ternary SNARE complex . After washing unbound proteins , we added complexin-1 which was labeled with a FRET donor dye within the accessory domain . We observed two populations with low and high FRET efficiency , with their means corresponding to 0 . 19 ± 0 . 12 and 0 . 83 ± 0 . 15 , respectively ( Figure 3B and Table 1 ) . We refer to the high FRET efficiency state as the 'cis' conformation of complexin-1 , whereas we refer to the low FRET efficiency state as the 'trans' conformation of complexin-1 ( Figure 3A ) . 10 . 7554/eLife . 16886 . 006Figure 3 . Cis and trans conformations of the accessory domain of complexin-1 when bound to ternary SNARE complex . ( A ) Schematic of smFRET measurements with fluorescent labels attached to the accessory domain of complexin-1 and to the C-terminal end of the ternary SNARE complex . The cytoplasmic domain of syntaxin-1A ( SX , colored red ) was surface-tethered through biotin-streptavidin ( orange dot ) linkage to a passivated microscope slide . Prior to tethering SX was labeled with FRET acceptor dye ( Alexa 647 ) at residue 249 ( SX 249 , yellow dot ) . Next , SNAP-25A ( S25 , colored green ) was added . Subsequently the cytoplasmic domain of synaptobrevin-2 ( SB , colored blue ) and 10 μM dN-SB fragment were added concurrently to the surface-tethered SX-S25 complex ( i . e . , using the dN-SB method ) . Unbound proteins were then washed away . Full-length wildtype complexin-1 ( Cpx , black ) was labeled with FRET donor dye ( Alexa 555 ) at residue position 24 ( Cpx 24 , yellow dot ) and added to the surface tethered ternary SNARE complex at 0 . 01 μM Cpx concentration . Unbound Cpx was then washed away . ( B ) smFRET efficiency histogram for the label pairs described in panel A using wildtype Cpx . Arrows indicate the calculated FRET efficiencies for two crystal structures ( red: 0 . 13 for PDB IDs 3RK3 , 3RL0; green: 0 . 49 for PDB ID 1KIL ) . ( C ) smFRET efficiency histogram for the label pairs described in panel A using wildtype ( WT ) Cpx and the accessory domain mutants ( SC , superclamp; NC , no-clamp ) of Cpx . ( D ) Summary bar chart of the histograms shown in panel C , illustrating the percentage of cis conformations for Cpx and its mutants . The “% cis conformation” is calculated as the ratio of the areas under the two Gaussian functions that are fit to the high and low FRET efficiency states in the corresponding smFRET efficiency histograms , respectively . Shown are mean values ± SD for the two subsets of an equal partition of the data that are comprised of the observed FRET efficiency values for all molecules for each different condition ( see data summary table in Figure 3—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 00610 . 7554/eLife . 16886 . 007Figure 3—source data 1 . Data summary table for the results shown in Figure 3D . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 00710 . 7554/eLife . 16886 . 008Table 1 . γ-corrected smFRET efficiencies between fluorescent dye labels attached to complexin-1 ( Cpx ) mutants ( WT , wildtype; SC , superclamp; NC , no-clamp ) and ternary SNARE complex obtained from smFRET measurements . The γ-factors for the Cpx mutants ( Cpx WT: 1 . 66 , Cpx SC: 1 . 62 , Cpx NC: 1 . 42 ) were empirically estimated , and the corresponding mean values were used to correct the corresponding smFRET efficiency histogram shown in Figure 3A , B using Equation ( 4 ) as described in the Materials and methods . The two FRET efficiencies shown in the table are the peak positions of two Gaussian functions that were fit to the γ-corrected smFRET efficiency histograms , and the error bounds are standard deviations of the peak positions . For comparison , FRET efficiencies were calculated from the specified crystal structures ( PDB IDs: 3RK3 , 3RL0 , and 1KIL , see Materials and methods for the calculation of FRET efficiencies from crystal structures ) . Cpx was labeled with FRET donor dye ( Alexa 555 ) molecule at residue position 24 and syntaxin-1A was labeled with FRET acceptor dye ( Alexa 647 ) molecule at residue position 249 . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 008γ-corrected smFRET efficiencyFRET efficiency calculated from the specified crystal structuresCpx WT0 . 19 ± 0 . 12 ( trans ) 0 . 83 ± 0 . 15 ( cis ) 0 . 13 ( PDB ID: 3RK3 , 3RL0 ) 0 . 49 ( PDB ID: 1KIL ) Cpx SC0 . 28 ± 0 . 18 ( trans ) 0 . 83 ± 0 . 14 ( cis ) Cpx NC0 . 24 ± 0 . 13 ( trans ) 0 . 81 ± 0 . 12 ( cis ) For comparison , we calculated the corresponding FRET efficiencies from the crystal structures of the complexin-1 / ternary SNARE supercomplex ( Materials and methods ) : the calculated FRET efficiency is 0 . 49 for the supercomplex between complexin-1 and the fully assembled SNARE complex ( PDB ID 1KIL ) and 0 . 13 for the complex between the superclamp mutant of complexin-1 and the partially truncated SNARE complex ( PDB IDs 3RK3 , 3RL0 ) ( Chen et al . , 2002; Kümmel et al . , 2011 ) . Thus , the cis conformation of complexin-1 corresponds to a FRET efficiency that is larger than that derived from the crystal structures , including error estimates ( Table 1 ) . Ensemble averaging in previous ensemble FRET experiments of the complexin-1 / ternary SNARE complex ( Krishnakumar et al . , 2011; Kümmel et al . , 2011 ) would have masked the two distinct conformations of complexin-1 that we uncovered: our smFRET data suggests that there are two distinct conformations ( cis and trans ) of complexin-1 that both exist when bound to ternary SNARE complex . We next tested the effect of the so-called superclamp and no-clamp mutations ( Figure 1A ) on the smFRET efficiency histograms ( Figure 3C ) . The percentage of the cis conformation slightly increased for the superclamp mutant , whereas it substantially decreased for the no-clamp mutant of complexin-1 compared to wildtye complexin-1 ( Figure 3D ) . The large effect of the no-clamp mutant on the population of the high FRET efficiency state suggests that the cis conformation of complexin-1 involves an intimate interaction of the accessory domain of complexin-1 with the ternary SNARE complex . Since the cis conformation of complexin-1 suggests an intimate interaction with the membrane-proximal C-terminal end of the ternary SNARE complex , we tested if this interaction affects the conformation of the ternary SNARE complex itself . We used the same SNARE label pairs as described in Figure 2B to probe the conformation of the ternary SNARE complex . Upon addition of 1 μM full-length wildtype complexin-1 , a low FRET efficiency state emerged for the SFC1 and SFC2 label pairs at the C-terminal end of the ternary SNARE complex , but not for the SFC3 label pair at the N-terminal end ( Figure 4A , B and Table 2 ) with occasional transitions between high and low FRET states of the SNARE complex ( Figure 4—figure supplement 1 ) . The population of the low FRET efficiency state increases as the complexin-1 concentration is increased , reaching saturation around 1 μM within experimental error ( Figure 4C , E ) . At higher concentration there is a higher likelihood of complexin-1 / ternary SNARE supercomplex formation consistent with the dissociation constant of this complex of around 10–100 nM ( Pabst et al . , 2002; Li et al . , 2007 ) , and consequently , also a higher likelihood of bound complexin-1 in the cis conformation . We note that the low FRET efficiency state that is induced by the cis conformation of complexin-1 is different from the low FRET efficiency state that is observed when ternary SNAREs are assembled without the dN-SB method ( compare Figures 2C and 4B ) . 10 . 7554/eLife . 16886 . 009Figure 4 . The cis conformation of complexin-1 induces a conformational change at the membrane-proximal C-terminal end of the ternary SNARE complex . Surface-tethered ternary SNARE complexes were assembled using the dN-SB method as described in Figure 2A . The SNARE complex was labeled with each of the three SNARE label pairs SFC1 , SFC2 , SFC3 ( as defined in Figure 2B ) in separate experiments . Labeling of the two sites of a particular pair was performed separately with FRET donor and acceptor dyes ( Alexa 555 and Alexa 647 , respectively , Figure 4—source data 1 ) and the ternary SNARE complex was formed using dN-SB method . The analysis was restricted to cases where FRET was observed , i . e . , complexes that contain one donor and one acceptor dye . Representative single molecule fluorescence intensity time traces for the SNARE label pair SFC2 are shown in Figure 4—figure supplement 1 . A data summary table for all experiments in this figure is provided in Figure 4—source data 1 . ( A ) smFRET efficiency histograms for SNARE label pairs SFC1 , SFC2 , SFC3 in the absence of full-length wildtype complexin-1 ( Cpx [1–134] ) ( identical to Figure 2D ) . ( B ) smFRET efficiency histograms for SNARE label pairs SFC1 , SFC2 , SFC3 in the presence of full-length wildtype complexin-1 ( Cpx [1–134] ) . 1 μM Cpx [1–134] was then added to form supercomplex with the ternary SNARE complex . ( C ) smFRET efficiency histograms for the SNARE label pair SFC2 in presence of wildtype full-length complexin-1 ( Cpx [1-134] ) at the specified concentrations . ( D ) smFRET efficiency histograms for SNARE label pair SFC2 in the presence of 1 μM full-length wildtype ( WT ) complexin-1 ( Cpx ) and its mutants ( SC , superclamp; NC , no-clamp; and 4M mutation of the central domain that prevents binding to SNARE complex , Figure 1A ) . ( E ) Summary bar graph of the histograms shown in panel D . The “% conformational change” is calculated as the ratio of the areas under the two Gaussian functions that are fit to the low and high FRET efficiency states in the corresponding smFRET efficiency histograms , respectively . Shown are mean values ± SD for the two subsets of an equal partition of the data that are comprised of the observed FRET efficiency values for all molecules for each different condition ( see data summary table in Figure 4—source data 1 ) . ( F ) Summary bar graph of the histograms shown in panel E . The “% conformational change” is calculated as the ratio of the areas under the two Gaussian functions that are fit to the low and high FRET efficiency states in the corresponding smFRET efficiency histograms , respectively . Shown are mean values ± SD for the two subsets of an equal partition of the data that are comprised of the observed FRET efficiency values for all molecules for each different condition ( see data summary table in Figure 4—source data 1 ) . ( G ) smFRET efficiency histograms for SNARE label pairs SFC1 and SFC2 in the presence of 1 μM truncated complexin-1 ( Cpx [26–134] ) fragment ( Figure 1B ) . ( H ) smFRET efficiency histograms for SNARE label pairs SFC1 and SFC2 in the presence of 1 μM truncated complexin-1 ( Cpx [48–134] ) fragment ( Figure 1B ) . ( I ) Summary bar chart of the histograms in panels B , C , H , and I , illustrating the effect of WT complexin-1 and its truncation mutants on the conformation of the SNARE complex for the specified label pairs . Label pair SFC3 was not tested for the truncation mutants of complexin-1 since Cpx has no effect on the N-terminal end of the ternary SNARE complex . The “% conformational change” is calculated as the ratio of the areas under the two Gaussian functions that are fit to the low and high FRET efficiency states in the corresponding smFRET efficiency histograms , respectively . Shown are mean values ± SD for the two subsets of an equal partition of the data that are comprised of the observed FRET efficiency values for all molecules for each different condition ( see data summary table in Figure 4—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 00910 . 7554/eLife . 16886 . 010Figure 4—source data 1 . Data summary table for the results shown in Figure 4A–D , G , H . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 01010 . 7554/eLife . 16886 . 011Figure 4—figure supplement 1 . Representative single molecule fluorescence intensity time traces for the SNARE label pair SFC2 ( Alexa 555 attached to residue 82 of the cytoplasmic domain of synaptobrevin-2 and Alexa 647 attached to residue 249 of the cytoplasmic domain of syntaxin-1A ) in the presence of 1 μM full-length wildtype complexin-1 ( Cpx [1-134] ) . Occasional transitions between low and high FRET efficiency states were observed ( characterized by a simultaneous increase of acceptor fluorescence and decrease of donor fluorescence intensity ) , as well as opposite translations from high FRET to low FRET efficiency states . Acceptor photobleaching is characterized by disappearance of acceptor fluorescence . Donor photobleaching is characterized by disappearance of donor fluorescence . FRET efficiencies were measured during the first 5 s ( corresponding to 50 frames ) , i . e . , well before such photobleaching events occurred . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 01110 . 7554/eLife . 16886 . 012Table 2 . γ-corrected smFRET efficiencies between fluorescent labels attached to the ternary SNARE complex ( label pairs SFC1 and SFC2 as defined in Figure 2B ) obtained from smFRET measurements in the presence and absence of 1 μM full-length wildtype complexin-1 ( Cpx [1-134] ) . Mean γ-factors ( Table 2—source data 1 ) were used to correct the smFRET efficiency histograms shown in Figure 4A–B using Equation ( 4 ) as described in the Materials and methods . The FRET efficiencies shown in the table are the peak positions of one or two Gaussian functions that were fit to the γ-corrected smFRET efficiency histograms , and the error bounds are standard deviations of the peak positions . As explained in the text , the population of the low FRET efficiency state in the presence of Cpx likely corresponds to the cis conformation of bound complexin-1 , whereas the population of the high FRET efficiency state likely corresponds to the trans conformation of bound complexin-1 or SNARE complex alone . For comparison , FRET efficiencies were calculated from the crystal structure of the ternary SNARE complex ( PDB ID: 1SFC , see Materials and methods for the calculation of FRET efficiencies from the crystal structures ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 01210 . 7554/eLife . 16886 . 013Table 2—source data 1 . The means of the empirical γ-factors for SNARE label pairs SFC1 and SFC2 in the presence and absence of 1 μM complexin-1 ( Cpx [1-134] ) for the data shown in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 013γ-corrected smFRET efficiency - Cpxγ-corrected smFRET efficiency +CpxFRET efficiency calculated from the crystal structure ( PDB ID: 1SFC ) SFC10 . 85 ± 0 . 170 . 27 ± 0 . 11 0 . 85 ± 0 . 140 . 98SFC20 . 90 ± 0 . 180 . 31 ± 0 . 13 0 . 86 ± 0 . 120 . 98 To confirm that the change in FRET efficiency is due to a conformational change of the SNARE complex and not caused by dye effects , we followed established procedures in the single molecule field ( McCann et al . , 2012; Akyuz et al . , 2013; Munro et al . , 2014 ) and performed fluorescence anisotropy measurements of the individual fluorophores attached to the SNARE label sites in the presence and absence of full-length wildtype complexin-1 ( Table 3 ) . There was very little or no change in fluorescence anisotropy for all label sites in the presence of complexin-1 compared to SNARE complex without complexin-1 , suggesting that restrictions of the rotational freedoms of the dye molecules by bound complexin-1 are unlikely . Moreover , the sums of the donor and acceptor intensities are similar for the single molecule fluorescence intensity time traces before photobleaching occurs ( see representative time traces in Figure 4—figure supplement 1 ) , suggesting that there is no large protein induced fluorescence enhancement as seen in some systems when a protein binds near a fluorophore at short distance ( Hwang and Myong , 2014 ) . We also ruled out photo-physical effects on quantum yield and detection efficiency as an explanation for the observed appearance of the population of the low FRET efficiency state by empirically determining the γ-factor for individual molecules ( Figure 5 , Table 2—source data 1 and Materials and methods ) . The low FRET efficiency state persisted after application of the γ-correction . Finally , we note that measurements of binding kinetics and equilibrium binding constants for the labeled proteins ( Materials and methods ) agree with literature values of unlabeled proteins ( Pabst et al . , 2002 ) , again suggesting that the fluorophores do not interfere with the binding surfaces . Taken together , all these controls and observations suggest that the population of the low FRET efficiency states observed for the SNARE label pairs SFC1 and SFC2 ( Figure 4B ) represent a distinct conformational state of the ternary SNARE complex that is induced by complexin-1 . 10 . 7554/eLife . 16886 . 014Table 3 . Fluorescence anisotropy measurements of Alexa 555 or Alexa 647 dyes linked to the specified residues in the cytoplasmic domains of syntaxin-1 ( SX ) or synaptobrevin-2 ( SB ) , either in isolation or as part of the ternary SNARE complex , in the presence and absence of 1 μM full-length wildtype complexin-1 ( Cpx [1-134] ) . The numbers after 'SX' and 'SB' are the residue positions of the corresponding labeling sites . The formation of ternary SNARE complex and the presence of complexin-1 did not substantially affect the fluorescence anisotropy . Shown are means ± SD for n = 3 replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 014ConstructsSteady state anisotropy ( r ) Alexa 5550 . 20 ± 0 . 06Alexa 555 ( linked to SB 91 ) 0 . 26 ± 0 . 01Alexa 555 ( linked to SB 91 , in ternary complex ) 0 . 31 ± 0 . 01Alexa 555 ( linked to SB 91 , in ternary complex ) + 1 μM Cpx0 . 32 ± 0 . 03Alexa 555 ( linked to SB 82 ) 0 . 27 ± 0 . 01Alexa 555 ( linked to SB 82 , in ternary complex ) 0 . 27 ± 0 . 02Alexa 555 ( linked to SB 82 , in ternary complex ) + 1 μM Cpx0 . 28 ± 0 . 06Alexa 555 ( linked to SB 82 , in ternary complex ) + 10 μM Cpx0 . 28 ± 0 . 1Alexa 555 ( linked to SB 28 ) 0 . 24 ± 0 . 03Alexa 555 ( linked to SB 28 , in ternary complex ) 0 . 32 ± 0 . 01Alexa 555 ( linked to SB 28 , in ternary complex ) + 1 μM Cpx0 . 32 ± 0 . 004Alexa 6470 . 15 ± 0 . 08Alexa 647 ( linked to SX 259 ) 0 . 24 ± 0 . 09Alexa 647 ( linked to SX 259 , in ternary complex ) 0 . 27 ± 0 . 1Alexa 647 ( linked to SX 259 , in ternary complex ) + 1 μM Cpx0 . 27 ± 0 . 1Alexa 647 ( linked to SX 249 ) 0 . 24 ± 0 . 02Alexa 647 ( linked to SX 249 , in ternary complex ) 0 . 26 ± 0 . 1Alexa 647 ( linked to SX 249 , in ternary complex ) + 1 μM Cpx0 . 27 ± 0 . 08Alexa 647 ( linked to SX 249 , in ternary complex ) + 10 μM Cpx0 . 28 ± 0 . 1Alexa 647 ( linked to SX 193 ) 0 . 27 ± 0 . 1Alexa 647 ( linked to SX 193 , in ternary complex ) 0 . 26 ± 0 . 08Alexa 647 ( linked to SX 193 , in ternary complex ) + 1 μM Cpx0 . 27 ± 0 . 0810 . 7554/eLife . 16886 . 015Figure 5 . Empirical γ-corrected smFRET efficiency histograms . Empirical γ-corrected smFRET efficiency histograms for the SNARE label pairs SFC1 ( A ) , SFC2 ( B ) , in the presence ( red ) and absence ( black ) of 1 μM full-length wildtype complexin-1 ( Cpx [1-134] ) . The FRET efficiency was calculated using Equation ( 4 ) as described in the Materials and methods . γ vs . FRET efficiency for the SFC1 ( C ) and SFC2 SNARE label pairs ( D ) , in the presence ( red ) and absence ( black ) of 1 μM Cpx [1-134] . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 015 We conducted further tests using mutations and truncations of complexin-1 . The superclamp mutant ( Figure 1A ) of the accessory domain of complexin-1 slightly increased the population of the low FRET efficiency state compared to wildtype complexin-1 , while the no-clamp mutant did not induce a conformational change of the ternary SNARE complex ( Figure 4D , F ) . These effects correlate with the FRET efficiency measurements of the complexin-1 / ternary SNARE supercomplex with FRET label pairs attached to complexin-1 and syntaxin-1A ( Figure 3C , D ) : there is a slight increase in the population of the high FRET efficiency state by the superclamp mutant corresponding to an increase in the cis conformation of complexin-1 , whereas the no-clamp mutant decreased the population of the high FRET efficiency state corresponding to a decrease of the cis conformation of complexin-1 . Therefore the low FRET efficiency state in Figure 4B likely corresponds to the cis conformation of complexin-1 in Figure 3B , and the high FRET efficiency state in Figure 4B likely corresponds to the trans conformation of complexin-1 . Taken together , these experiments support the notion that the cis conformation of complexin-1 causes a conformational change at the C-terminal end of the SNARE complex . As control , we also tested the 4M mutant ( Figure 1A ) of complexin-1 that blocks the binding of the central domain of complexin-1 to SNARE complex and abolishes all functions of complexin-1 ( Maximov et al . , 2009 ) . As expected , it had no effect on the ternary SNARE complex . Finally , the conformational change at the C-terminal end of the ternary SNARE complex depended on the inclusion of both the N-terminal and the accessory domains of complexin-1 since truncation of one or both of these domains ( Figure 1B ) did not produce the conformational change ( Figure 4G–I ) . The trans conformation of complexin-1 observed by smFRET ( Figure 3A ) should be capable of bridging two SNARE complexes as suggested previously ( Kümmel et al . , 2011 ) . To further corroborate this notion with different SNARE complexes , we first assembled the supercomplex of complexin-1 bound to ternary SNARE complex by tethering unlabeled ternary SNARE complex to the surface of a microscope slide , and then added FRET acceptor dye labeled complexin-1 ( Figure 6A ) . After removing unbound complexin-1 molecules , FRET donor dye labeled binary ( syntaxin-1A / SNAP-25A ) SNARE complex was added in solution and smFRET events were monitored ( Figure 6A ) . The observed isolated bursts of acceptor fluorescence intensity reflect binding events of binary SNARE complex to tethered complexin-1 / ternary SNARE supercomplex ( Figure 6B ) . From dwell time analysis we obtained rate constants koff = 2 . 16 ± 0 . 02 s-1 and kon = 0 . 34 ± 0 . 07 μM-1s-1 , resulting in KD = 6 . 46 ± 1 . 2 μM ( Figure 6C–E ) . 10 . 7554/eLife . 16886 . 016Figure 6 . The trans conformation of complexin-1 can bridge a ternary and a binary SNARE complex through its central and accessory domains . ( A ) Schematic of the single molecule binding experiment . The cytoplasmic domain of syntaxin-1A ( SX ) was surface-tethered through biotin-streptavidin ( orange dot ) linkage to a passivated microscope slide . Ternary SNARE complex ( consisting of SX , the cytoplasmic domain of synaptobrevin-2 ( SB ) , and SNAP-25A ( S25 ) ) was assembled using the dN-SB method . Full-length wildtype complexin-1 ( Cpx [1-134] ) was labeled with acceptor dye ( Alexa 647 ) at residue position 24 ( yellow dot ) . 50 nM of labeled complexin-1 was then incubated for 5 min to form complex with the surface-tethered ternary SNARE complexes . Unbound Cpx molecules were rinsed away . Next , purified binary SNARE complex ( consisting of S25 and SX that was donor dye ( Alexa 555 ) labeled at residue 249 of SX , yellow dot ) was added at a concentration of 100 nM to the surface-tethered Cpx / ternary SNARE supercomplex . The appearance of acceptor dye fluorescence intensity indicates FRET from the binary SNARE complex and to the surface-tethered Cpx / ternary SNARE supercomplex . A data summary table for all experiments in this figure is provided in Figure 6—source data 1 . ( B ) Representative single molecule fluorescence intensity time traces showing smFRET events between the complexin-1 accessory domain of complexin-1 / ternary SNARE supercomplex and a binary ( syntaxin-1A / SNAP-25A ) SNARE complex . The donor fluorescence intensity is colored green ( scale on the right y-axis ) and the acceptor fluorescence intensity ( scale on the left y-axis ) is colored red . Due to the high concentration of the donor labeled proteins in solution , there is no significant effect on the donor intensity upon FRET with an acceptor dye . The stepwise increase in acceptor fluorescence intensity represents bound states and the gaps in between bound states are unbound states . Tbound and Tunbound represent the dwell time of bound and unbound states , respectively . ( C , D ) Dissociation rates ( C , open circles ) and association rates ( D , open squares ) between binary SNARE complex and surface-tethered Cpx mutants ( WT , wildtype; SC , superclamp mutant; NC , no clamp mutant , 4M , mutation of the central helix that prevents binding to ternary SNARE complex , see Figure 1A ) . Rates are calculated as described in Materials and methods . Error bars are standard deviations calculated from two subsets of an equal partition of the data ( see data summary table in Figure 6—source data 1 ) . ( E ) Apparent dissociation constant KD ( = koff/kon ) of binding between binary SNARE complex and Cpx or its mutants . Error bars are standard deviations calculated from two subsets of an equal partition of the data ( see data summary table in Figure 6—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 01610 . 7554/eLife . 16886 . 017Figure 6—source data 1 . Data summary table for the results shown in Figure 6C–E . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 017 The affinity between the complexin-1 / ternary SNARE supercomplex and the binary ( syntaxin-1A / SNAP-25A ) SNARE complex is roughly three times weaker than that between complexin-1 and binary ( syntaxin-1A / SNAP-25A ) SNARE complex alone ( Figure 6E and Materials and methods ) . We tested this interaction by mutations of complexin-1 . The koff rate of the superclamp mutant of complexin-1 was ~1 . 5-fold slower ( 1 . 36 ± 0 . 06 s-1 ) than that of wildtype complexin-1 , while the kon was ~1 . 5-fold faster ( 0 . 56 ± 0 . 06 μM-1s-1 ) , resulting in KD = 2 . 42 ± 0 . 3 μM ( Figure 6C–E ) . Thus , the complexin-1 superclamp mutant interacts with the binary SNARE complex with three-fold higher affinity than wildtype complexin-1 . In contrast , binary SNARE complex binding events with the no-clamp mutant were very rare within our observation period of 100 s and , consequently , we were unable to determine a dissociation constant for the no-clamp mutant , consistent with the notion that it is the complexin-1 accessory domain that establishes the interaction with the binary SNARE complex in our experiment . As expected , the 4M mutant did not bind at all . In summary , the trans conformation of complexin-1 can bridge two SNARE complexes . At variance with this previous study involving partially truncated ternary SNARE complexes ( Kümmel et al . , 2011 ) , our results now show that complexin-1 can also bridge a ternary SNARE complex and a binary SNARE complex .
Previous single molecule experiments indicated the presence of improperly assembled SNAREs , including antiparallel syntaxin-1A / synaptobrevin-2 subconfigurations , when mixed in the absence of other factors that assist proper SNARE complex formation ( Weninger et al . , 2003; Lou et al . , 2015; Ryu et al . , 2015 ) . In some of these studies , the existence of antiparallel SNARE complex subconfigurations were directly probed by using antiparallel reporting FRET dye pairs ( Weninger et al . , 2003 ) . SNARE complex was assembled and then purified , followed by reconstitution into liposomes , which were used to form supported lipid bilayers containing reconstituted SNARE complexes . Thus , surface adsorption effects could not have contributed to the observed antiparallel subconfigurations since the ternary SNARE complex was assembled in solution before reconstitution and surface tethering . An independent study ( Ryu et al . , 2015 ) also suggested the existence of antiparallel SNARE subconfigurations . Using label pairs attached to synaptobrevin-2 and SNAP-25A , a low smFRET efficiency state was observed , consistent with a subpopulation of antiparallel subconfigurations ( Figure S4A and supplementary material pages 2–3 for the case without αSNAP in Ryu et al . 2015 ) . As previously described ( Weninger et al . , 2003 ) , antiparallel syntaxin-1A / synaptobrevin-2 subconfigurations can be suppressed by extensive purification of the ternary SNARE complex , including an additional urea wash step . However , such a urea purification step would not be possible with SNAREs that are reconstituted in membranes . Our smFRET experiments indicate that the dN-SB method ( Pobbati et al . , 2006 ) also results in the proper parallel syntaxin-1A / synaptobrevin-2 subconfiguration within the ternary SNARE complex ( Figure 2 ) . Our results thus provide for an explanation of the dN-SB method . This method is related to an earlier report using a slightly different synaptobrevin fragment , the Vc peptide ( Melia et al . , 2002 ) . There is independent evidence that improperly assembled SNAREs ( including antiparallel subconfigurations ) may be a feature of SNAREs in other contexts as well: for vacuolar fusion , proof-reading by the HOPS complex is essential for maximal fusogenic trans SNARE complex formation ( Zick and Wickner , 2014 ) . In any case , we used the dN-SB method for all subsequent smFRET studies in this work in order to ensure properly assembled SNARE complex . Previously it had been suggested that the accessory domain of complexin-1 blocks binding of the membrane-proximal C-terminal part of synaptobrevin-2 , preventing full zippering of the SNARE complex ( Giraudo et al . , 2006 , 2009 ) , although subsequent experiments suggested that complexin-1 rather bridges partially assembled SNARE complexes ( Kümmel et al . , 2011 ) . Our smFRET experiments now suggest that the accessory domain is involved in either interaction depending on the particular conformation of complexin-1 ( Figure 3 ) . By monitoring the position of the accessory domain with respect to the ternary SNARE complex , we found two major populations in the smFRET efficiency histograms . We note that ensemble averaging in previous FRET experiments probably masked the existence of the two populations ( Krishnakumar et al . , 2011; Cho et al . , 2014 ) . Our smFRET experiments have now resolved two conformations of complexin-1 that both exist when bound to full-length ternary SNARE complex . We ascribe the low and high FRET efficiency state to the trans and cis conformations of complexin-1 , respectively . The superclamp mutant of complexin-1 slightly increases the population of the high FRET efficiency state , while the no-clamp mutant greatly decreases the population of the high FRET efficiency state ( Figure 3C , D ) , suggesting that the accessory domain is important for the cis conformation of complexin-1 . By positioning FRET label pairs at the C-terminal end of the SNARE complex ( Figure 2B ) , we monitored the effect of the cis conformation of complexin-1 . Remarkably , a substantial fraction of the ternary SNARE complex exhibited a conformational change upon addition of complexin-1 as indicated by the emergence of a low FRET efficiency state ( Figure 4B ) . This complexin-1 induced conformational change of the ternary SNARE complex depends on interactions involving the N-terminal , accessory , and core domains of complexin-1 ( Figure 4I ) , suggesting an intimate interaction between all of these domains of complexin-1 and the ternary SNARE complex . Thus , the low FRET efficiency state likely corresponds to the cis conformation of complexin-1 when bound to the ternary SNARE complex . In support of this notion , mutations of the accessory helix of complexin-1 also had an effect on the conformations of complexin-1 when bound to the ternary SNARE complex . When attaching FRET label pairs on complexin-1 and syntaxin-1A within the complexin-1 / ternary SNARE supercomplex , the superclamp mutant slightly increased the population of the high FRET efficiency state ( i . e . , an increase of the cis conformation ) compared to wildtype complexin-1 ( Figure 3D ) . This effect correlates with a slight increase in the population of the low FRET efficiency state by the superclamp mutant when the FRET label pairs were attached to the C-terminal end of the SNARE complex ( Figure 4F ) . Similarly , the effects for the no-clamp mutant were also correlated compared to wildtype complexin-1: there is a decrease of the cis conformation of complexin-1 ( Figure 3D ) and a decrease in the population of the low FRET efficiency state of the SNARE complex ( Figure 4F ) . Therefore , the cis conformation ( Figure 3A ) of bound complexin-1 likely induces the conformational change of the ternary SNARE complex at the membrane-proximal C-terminal end ( Figure 4B ) . As independent evidence for a conformational change at the C-terminal end of the SNARE complex , 1H-15N TROSY-HSQC NMR spectra with 15N labeled ternary SNARE complex and complexin-1 revealed interactions between the complexin-1 N-terminus with all components ( syntaxin-1A , SNAP-25A , and synaptobrevin-2 ) at the C-terminal end of the ternary SNARE complex ( Figure 5b in Xue et al . , 2010 ) . The interaction between complexin-1 and ternary SNARE complex was also examined by EPR ( Lu et al . , 2010 ) . Most of the spin-labeled residues of the accessory domain of complexin-1 exhibited spectral broadening when quaternary complex was formed with the SNARE complex , likely due to formation of an α-helical conformation of the accessory domain of complexin-1 . However , the authors noted that the spectra were sharper than those for solvent exposed residues on the surface of an α-helix that is part of a globular protein . Therefore , the observed sharpening could be due to motion between several conformations that are sampled by the accessory domain , consistent with the range of conformations that are suggested by our smFRET efficiency histogram ( Figure 3B ) . Spin-labeling of the C-terminal half of the cytoplasmic domain of synaptobrevin-2 indicated little effect on the EPR line widths for most of the labels ( Lu et al . , 2010 ) , suggesting that the conformational change at the C-terminal end of the SNARE complex does not involve a dissociation of a part of synaptobrevin-2 , but rather that synaptobrevin-2 is still interacting with the other components of the complexin-1 / ternary SNARE supercomplex . The N-terminal domain ( residues 1–27 ) of complexin-1 is important for activation of fast synchronous Ca2+-triggered release and it interacts with the C-terminal end of the ternary SNARE complex ( Xue et al . , 2007 , 2010; Maximov et al . , 2009 ) . Since we have shown that the cis conformation of complexin-1 induces a conformational change at the C-terminal end of the SNARE complex through a mechanism that involves both the accessory and the N-terminal domains of complexin-1 , we speculate that the cis conformation of complexin-1 is related to activation of Ca2+-triggered release , although it could also play a role in regulation of spontaneous release ( Giraudo et al . , 2006; 2009 ) ( Figure 7 ) . However , we note that mutations of the accessory domain of complexin-1 do not affect the activating function of complexin-1 compared to wildtype neurons in rescue experiments with complexin-1 knockdown ( Yang et al . , 2010 , 2013 ) . Moreover , recent in vitro experiments revealed that the accessory domain is entirely dispensable for activation of Ca2+-triggered synaptic vesicle fusion ( Lai et al . , 2016 ) . These observations can be reconciled with the findings in this work by postulating that the interaction of the N-terminal domain of complexin-1 occurs for the trans SNARE complex that is juxtaposed between the membranes regardless of the accessory domain . In contrast when starting from fully assembled ternary SNARE complex , both the accessory and N-terminal domains are required to induce the observed conformational change at the C-terminal end of the SNARE complex . We speculate that the complexin-1 induced conformation of the SNARE complex may be related to the conformation of a trans SNARE complex that juxtaposes the synaptic vesicle and plasma membranes ( Figure 7 , right panel ) . 10 . 7554/eLife . 16886 . 018Figure 7 . Models of the cis and trans conformations of complexin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 018 We studied interactions of complexin-1 in the trans conformation by first binding complexin-1 to a surface-tethered ternary SNARE complex via the central domain of complexin-1 . The resulting complexin-1 / ternary SNARE supercomplex should allow the N-terminal , accessory , and C-terminal domains of bound complexin-1 to be accessible for other interactions when it is in the trans conformation . The binary ( syntaxin-1A / SNAP-25A ) SNARE complex bound weakly to the preassembled complexin-1 / ternary SNARE supercomplex via the accessory domain ( Figure 6E ) . Together with previous results ( Kümmel et al . , 2011 ) , one complexin-1 molecule in the trans conformation can bridge a variety of SNARE complexes . The correlations between complexin-1 mutants and this bridging interaction ( Figure 6C–E ) suggests that it is probably specific rather than merely a consequence of the flexible character of the binary SNARE complex that may expose hydrophobic elements in certain configurations ( Weninger et al . , 2008 ) . We speculate that the trans conformation of complexin-1 is related to regulation of spontaneous fusion ( Krishnakumar et al . , 2011; Kümmel et al . , 2011 ) ( Figure 7 , left panel ) . Our results also suggest that the two previously proposed SNARE complex 'clamping' models ( Giraudo et al . , 2006 , 2009; Krishnakumar et al . , 2011; Kümmel et al . , 2011 ) are not mutually exclusive ( see both panels in Figure 7 ) . A conformational change at the membrane-proximal C-terminal end of the ternary SNARE complex has been independently observed upon αSNAP binding to ternary SNARE complex by smFRET experiments with labels attached to synaptobrevin-2 and SNAP-25A ( Ryu et al . , 2015 ) . Similar to the complexin-1 induced conformational change at the C-terminal end of the ternary SNARE complex that we observe , a low FRET population was induced by αSNAP that increased as the αSNAP concentration was increased ( compare Figure S4B in [Ryu et al . , 2015] with Figure 4E ) . In contrast , no conformational change was observed at the N-terminal end of the SNARE complex upon αSNAP binding . Since two quite different factors ( complexin-1 and αSNAP ) can induce a conformational change at the C-terminal end of the SNARE complex , this suggests that the C-terminal end has different conformational properties than the N-terminal end . This finding is also consistent with multistage un-zippering of the SNARE complex by single molecule pulling experiments ( Gao et al . , 2012; Marniemi et al . , 1975 ) . Taken together , the observed plasticity of the membrane-proximal C-terminal end of the ternary SNARE complex likely has a functional role in membrane fusion , and the N-terminal domain of complexin-1 assists this process .
The cytoplasmic domain of rat syntaxin-1A ( amino acid range 1–265 ) , fused with a C-terminal biotinylation sequence ( GLNDIFEAQKIEWHE ) , was cloned into the pTEV5 vector ( Rocco et al . , 2008 ) with a N-terminal TEV cleavable 6x-histidine tag . The constructs for rat SNAP-25A ( amino acid range 1–206 ) , for the cytoplasmic domain of rat synaptobrevin-2 ( amino acid range 1–96 ) , for the dN-SB fragment of synaptobrevin-2 ( amino acid range 49–96 ) , for rat complexin-1 with the N-terminal domain deleted ( amino acid range 26–134 ) , and for rat complexin-1 with both the N-terminal and accessory domains deleted ( amino acid range 48–134 ) were also cloned into the pTEV5 vector that includes a N-terminal TEV cleavable 6 His-tag . Full-length complexin-1 ( residues 1–134 ) was cloned into the pET28a vector ( Novagen , EMD Chemicals , Gibbstown , NJ ) that includes a N-terminal thrombin-cleavable 6x-histidine tag . Using site-directed mutagenesis , all cysteines of the wildtype proteins were mutated to serines . Unique labeling sites were introduced by cysteine mutations using QuikChange Kit ( Agilent , Santa Clara , CA ) at surface exposed positions based on the available crystal structures . We generated constructs for the following labeling sites , one at a time: syntaxin-1A S225C , E234C , S249C , S259C , synaptobrevin-2 S61C , A72C , A82C , K91C , and complexin-1 K26C , E24C . We also generated constructs for the superclamp mutant ( D27L , E34F , R37A ) , no-clamp mutant ( A30E , A31E , L41E , A44E ) , and 4M mutant ( R48A , R59A , K69A , Y70A ) of full-length complexin-1 . All proteins were expressed in E . coli BL21 ( DE3 ) by growing the cells to OD600 of about 0 . 8 at 37°C , then induced with 0 . 5 mM IPTG for 4 hr at 30°C . Syntaxin-1A biotinylation was performed in vivo by co-expression with a BirA gene engineered into pACYC184 ( Avidity , Aurora , CO ) and induced with 0 . 5 mM IPTG at 30°C in the presence of 0 . 1 mM biotin for 4 hr . Cell pellets from 1 liter of culture were suspended in 40 ml of PBS ( 50 mM NaH2PO4 , 300 mM NaCl , 0 . 5 mM DTT , pH 8 . 0 ) buffer and 0 . 5 mM PMSF supplemented with EDTA free Complete Protease Inhibitor Cocktail tablets ( Roche , Basel , Switzerland ) . Cells were lysed by sonication at 1 s on and 2 s off pulse for 2 min at 50% power using Sonicator Ultrasonic Processor XL-2020 ( Misonix , Farmingdale , New York ) on ice water . Inclusion bodies were removed by centrifugation with JA-20 ( Beckman , Coulter , Brea , CA ) rotor at 20 , 000 rpm for 30 min . The supernatant was bound to Nickel-NTA agarose beads ( Qiagen , Hilden , Germany ) for 1 hr rotating at 4°C . The protein was washed extensively with PBS containing 20 mM imidazole while bound to Nickel-NTA agarose beads and eluted with the same buffer with 400 mM imidazole . 100 ug of TEV or thrombin protease was added and dialyzed overnight in 20 mM tris , 50 mM NaCl , 0 . 5 mM EDTA , 1 mM DTT , pH 8 . 0 to cleave off the N-terminal 6x-histidine tags . To separate the protease and the cleaved proteins , samples were purified using 1 ml HiTrap Q ( GE Healthcare Bio-Sciences , Piscataway , NJ ) with a linear gradient of 0 . 05 to 0 . 6 M NaCl in 20 mM tris , 0 . 5 mM TCEP , pH 7 . 5 . The soluble synaptobrevin-2 fragment ( amino acid range 1–96 ) does not bind to ion exchange columns . Therefore , cleaved samples were purified using a Superdex 75 size exclusion column ( GE Healthcare Bio-Sciences , Piscataway , NJ ) in 20 mM tris , 150 mM NaCl , 0 . 5 mM TCEP , pH 7 . 5 . Protein purity was checked using SDS-PAGE ( >95% ) . Mutated proteins with single cysteine sites were labeled with Alexa 555 or 647 maleimide ( Invitrogen , Carlsbad , CA ) in 20 mM tris , 300 mM NaCl , pH 7 . 5 with 0 . 5 mM TCEP overnight at 4°C using a rotating platform . Free dye was removed by Sephadex G50 resin ( GE Healthcare , Piscataway , NJ ) . The surface of the quartz microscope slide was coated with biotinylated BSA and passivated with 50 nm egg phosphatidylcholine liposomes ( Avanti Polar Lipids , Alabaster , AL ) . Streptavidin was added to tether the biotinylated and Alexa 647 labeled cytoplasmic domain of syntaxin-1A to the surface , using conditions that produced a density of about 200–300 syntaxin-1A molecules per 45 x 90 μm2 field of view . Binary SNARE complex ( consisting of syntaxin-1A and SNAP-25 ) was formed by adding 1 μM of unlabeled SNAP-25 to the surface with tethered syntaxin-1A , incubated for 5 min , and then washing out the free SNAP-25A molecules that did not form complex . This method achieves the desired 1:1 stoichiometric ratio since the tethered syntaxin molecules are primary isolated at the low concentration ( 100–200 pM ) and do not form homo-oligomeric species . To form ternary SNARE complex , 10 uM of the unlabeled cytoplasmic domain of synaptobrevin-2 ( amino acid range 1–96 ) was added to the binary ( syntaxin-1A / SNAP-25A ) SNARE complex , incubated for 5 min , and then washed out to remove unbound synaptobrevin-2 . For experiments with labeled synaptobrevin-2 , samples were diluted to 1 nM to form ternary SNARE complex . An additional purification step in the presence of the denaturant urea was used previously in order to suppress improper SNARE subconfigurations that can occur when SNARE components are mixed ( Weninger et al . , 2003 ) . However , the setup used in this work precludes the use of urea as an additional purification step since rinsing 7 . 5 M urea inside the microscope slide would disrupt the lipid bilayer surface along with the biotin-streptavidin linkage for protein tethering . Moreover , even with a denaturant purification step , there can be still up to 20 percent of improper subconfigurations ( Weninger et al . , 2003 ) . Instead we included the 10 μM dN-SB fragment of synaptobrevin-2 ( amino acid range 49–96 ) when ternary SNARE complex is assembled by adding synaptobrevin-2 to binary ( syntaxin-1 / SNAP-25A ) complex . After ternary SNARE complex formation , the unbound dN-SB fragments were removed by rinsing with TBS buffer ( 20 mM tris , 150 mM NaCl , pH 7 . 5 ) . We refer to this method as the dN-SB method and used it in all single molecule experiments ( Figures 2–6 ) in order to suppress improper SNARE subconfigurations . For the single molecule FRET experiments in Figures 2–5 , we used protein free observation buffer ( 1% ( w/v ) glucose , 20 mM tris , 150 mM NaCl , pH 7 . 5 ) in the presence of oxygen scavenger ( 20 units/ml glucose oxidase , 1000 units/ml catalase ) and triplet-state quencher ( 100 uM cyclooctatetraene ) . For the single molecule real-time inter-molecular binding studies shown in Figures 6 and 8 , 100 nM of Alexa 555 labeled complexin-1 was added to the observation buffer ( 1% ( w/v ) glucose , 20 mM tris , 150 mM NaCl , pH 7 . 5 ) in the presence of oxygen scavenger ( 20 units/ml glucose oxidase , 1000 units/ml catalase ) and triplet-state quencher ( 100 uM cyclooctatetraene ) in order to prevent fast photobleaching and blinking of the dye molecules . 10 . 7554/eLife . 16886 . 019Figure 8 . Single molecule detection of complexin-1 interacting with SNAREs . ( A ) Schematic of the smFRET binding assay . The cytoplasmic domain of syntaxin-1A ( SX ) was labeled with acceptor dye ( Alexa 647 ) at residue position 249 ( yellow dot ) and surface-tethered through biotin-streptavidin ( orange dot ) linkage on a passivated microscope slide . Surface-tethered and labeled syntaxin-1A was used in isolation , or as part of a binary SNARE complex with SNAP-25A ( SX-S25 ) , or as part of a ternary SNARE complex with S25 and the cytoplasmic domain of synaptobrevin-2 ( SB ) . Complexin-1 ( Cpx ) was labeled with donor dye ( Alexa 555 ) labeled at residue position 26 ( yellow dot ) and added at a concentration of 100 nM in TBS buffer ( 20 mM Tris , 150 mM NaCl , pH 7 . 5 ) to the sample chamber . smFRET is expected if Cpx interacts with the surface-tethered SNARE complexes . A data summary table for all experiments in this figure is provided in Figure 8—source data 1 . ( B ) Representative single molecule fluorescence intensity time traces showing individual binding events between ternary SNARE complex and Cpx . The donor dye fluorescence intensity ( scale on the right y-axis ) is colored green and the acceptor dye fluorescence intensity ( scale on the left y-axis ) is colored red . Due to the high concentration of the donor labeled proteins in solution , there is no significant effect on the donor intensity upon FRET with an acceptor dye . The stepwise increase in acceptor fluorescence intensity represents a bound state and gaps in between bound states represents unbound states . Tbound and Tunbound represent the dwell times of bound and unbound states , respectively . ( C–E ) Acceptor dye fluorescence intensities of Cpx molecules interacting with surface-tethered SX alone ( C ) , binary SNARE ( SX-S25 ) complex ( D ) , and ternary SNARE complex ( SX-S25-SB ) ( E ) . ( F , G ) Dissociation rates koff ( open circles ) and association rates kon ( open squares ) between Cpx and surface-tethered SX , binary SNARE complex ( SX-S25 ) , and ternary SNARE complex ( SX-S25-SB ) . The insets show single exponential fits to histograms of the unbound and bound dwell times . Rates are calculated as described in Materials and methods . Error bars are standard deviations calculated from two subsets of an equal partition of the data ( see data summary table in Figure 8—source data 1 ) . ( H ) Apparent dissociation constant KD ( = koff/kon ) of Cpx binding to SX alone , binary SNARE complex ( SX-S25 ) , and ternary SNARE complex ( SX-S25-SB ) . The inset shows bar graphs of the number of spots quantified from snapshots of the field of view of the acceptor channel corresponding to FRET events when complexin-1 interacts with the three different surface conditions . Error bars are standard deviations calculated from two subsets of an equal partition of the data ( see data summary table in Figure 8—source data 1 ) . ( I , J ) Dissociation rates koff ( G , open circles ) and association rates kon ( H , open squares ) between Cpx and its mutants ( WT , wildtype; SC , superclamp; NC , no-clamp; 4M , mutation of the central complex domain that prevents SNARE complex binding , see Fig , 1A ) and surface-tethered ternary SNARE complex . Rates are calculated as described in Materials and methods . Error bars are standard deviations calculated from two subsets of an equal partition of the data ( see data summary table in Figure 8—source data 1 ) . ( K ) Apparent dissociation constant KD ( = koff/kon ) for binding between ternary SNARE complex and complexin-1 or mutants . Error bars are standard deviations calculated from two subsets of an equal partition of the data ( see data summary table in Figure 8—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 01910 . 7554/eLife . 16886 . 020Figure 8—source data 1 . Data summary table for the results shown Figure 8F–K . DOI: http://dx . doi . org/10 . 7554/eLife . 16886 . 020 Details of the single molecule fluorescence microscopy setup have been described elsewhere ( Choi et al . , 2012 ) . Briefly , single molecule fluorescence intensities were recorded with a prism-type total internal reflection fluorescence ( TIRF ) microscope using 532 nm laser light ( CrystaLaser , Reno , NV ) excitation and detected by an Andor iXon EMCCD camera ( Andor Technology , South Windsor , CT ) at a frame rate of 10 Hz . The acceptor and donor fluorescence intensities were separated using a 640 nm single-edge dichroic beam-splitter ( Shemrock , Rochester , NY ) producing observation channels for both donor and acceptor fluorescence intensities . Fluorescence intensity movies were recorded for 1000 frames ( 100 s ) until most of the dyes were photobleached with 532 nm laser light illumination . Data analysis was performed with the smCamera software from Taekjip Ha’s group at University of Illinois and analyzed with scripts written for MATLAB ( Mathworks ) . For the smFRET experiments shown in Figures 2–5 , fluorescence intensity histograms were generated by accumulating 50 frames from individual fluorescence intensity time traces ( representative examples are shown in Figure 4—figure supplement 1 ) and converted to FRET efficiencies as described in the next section . For the single molecule binding experiments shown in Figures 6 and 8 , the binding events were characterized using hidden Markov modeling implemented in HaMMy software version 4 . 0 ( McKinney et al . , 2006 ) for a two-state system . High FRET represents the bound state and zero FRET represents the unbound state . The dwell times in the bound and unbound states were plotted in histograms and fitted to an exponential function to extract koff and kon , respectively . kon is the rate of the unbound state divided by the concentration of the donor labeled protein in solution . The dissociation constant , KD , is koff/kon . The FRET efficiency ( E ) between two dye molecules with separation ( r ) is given byE=11+ ( rR0 ) 6 , ( 1 ) where the Förster radius R0 is the distance when the efficiency of energy transfer is 50 percent . For the Alexa 555 and Alexa 647 dye pairs , R0 is given as 5 . 1 nm as provided by the manufacturer of the dyes . Förster theory relates the fluorescence intensities of donor ( ID ) and acceptor ( IA ) dyes to the separation of the two dye molecules as ( Stryer and Haugland , 1967; Stryer , 1978 ) E=IAIA+ID . ( 2 ) In the smFRET experiments shown in Figures 2–4 ( except Figure 5 , see next section ) , FRET efficiency refers toE=IA−β ( ID−αIA ) ( IA−β ( ID−αIA ) + ( ID−αIA ) ) , ( 3 ) where αIA corrects for leakage of acceptor emission into donor channel and βID corrects for leakage of donor emission into acceptor channel ( McCann et al . , 2010 ) . The leakage of donor fluorescence into the acceptor channel was measured to be 1 . 7% and the leakage of acceptor fluorescence into the donor channel was 16 . 5% . In Tables 1 and 2 where we compare the FRET efficiencies to those estimated from known-crystal structures , we additionally applied a γ-correction , which accounts for detection efficiency and quantum yield of the two dye molecules ( Dahan et al . , 1999; Ha et al . , 1999 ) :E=IA−β ( ID−αIA ) ( IA−β ( ID−αIA ) +γ ( ID−αIA ) ) , ( 4 ) where γ = ΔIA/ΔID is the γ-factor . We empirically estimated individual γ-factors from the changes in donor and acceptor fluorescence intensities before and after photobleaching of the acceptor dye molecule ( Ha et al . , 1999; McCann et al . , 2010 ) . The mean of the individual γ-factors was then used in Equation ( 4 ) to calculate the FRET efficiencies in Tables 1 and 2 . We empirically determined the γ-factors for many individual fluorescence traces for the SFC1 and SFC2 label pairs in the presence and absence of 1 μM complexin-1 ( Figure 5C–D ) . The mean γ-corrections are relatively small ( Table 2—source data 1 ) for both dye pairs , arguing against a large influence on the sampling of orientations of the dye molecules by complexin-1 binding . Fluorescence anisotropy experiments ( see below , and Table 3 ) also show that dye orientations are not affected by complexin-1 binding to the ternary SNARE complex . After application of the γ-corrections to the FRET efficiency histograms , the low FRET efficiency population is still observed ( Figure 5A–B ) , suggesting that the induction of the low FRET efficiency population must be due to a genuine conformational change of the ternary SNARE complex that is induced by complexin-1 . Moreover , for both SFC1 and SFC2 label pairs , the individual γ-factors for the low and high FRET populations are similar ( Figure 5C–D ) . In Tables 1 and 2 we estimated the distances for FRET pairs from the known high-resolution structures as previously described ( Choi et al . , 2010; McCann et al . , 2012 ) . Briefly , we used simulated-annealing molecular dynamics simulations starting from the crystal structures and conjugated fluorescent dye molecules to reduced cysteine residues ( Choi et al . , 2010 ) . 100 simulations were performed where all protein atoms were fixed except for the fluorescent dye , which was allowed to freely rotate . From the distance measured between the mean dye position of Cy3 and Cy5 , the FRET efficiency was calculated by using Equation ( 1 ) , where R0 is given as 5 . 1 nm for the Alexa 555 and Alexa 647 dye pairs as provided by the manufacturer of the fluorescent dyes . The steady-state anisotropy was measured relative with a Fluorolog spectrofluorometer ( HORIBA scientific , Edison , NJ ) relative to free Alexa 555 and 647 using an integration time of 2 s . For protein complexes singly labeled with Alexa 555 , we used an excitation wavelength of 532 ± 5 nm and an emission wavelength of 575 ± 5 nm . For protein complexes that were singly labeled with Alexa 647 , we used an excitation wavelength of 651 ± 5 nm and an emission wavelength of 671 ± 5 nm . The absorbance value for both dye labeled samples at the respective excitation wavelengths were adjusted to 0 . 05 AU for all samples to perform ensemble anisotropy measurements . SNARE proteins were individually purified and labeled with donor ( Alexa 555 ) and acceptor ( Alexa 647 ) dye molecule before complexes were formed . SNARE complex was assembled by mixing syntaxin-1 and SNAP-25 aliquots followed by addition synaptobrevin-2 fused to a N-terminal 6x-histidine tag at a ratio of 1:2:5 , respectively , overnight at 4°C . The protein mixture was rebound to Nickel-NTA agarose beads ( Qiagen , Hilden , Germany ) and washed extensively with TBS ( 20 mM Tris , 150 mM NaCl , pH 7 . 5 , 0 . 5 mM TCEP ) . Additional washing with TBS containing 7 . 5 M was used in order to favor the correct assembly and folding of the SNARE complex ( Weninger et al . , 2003 ) . TEV was added to the eluted samples to cleave off the hexa-histidine tag of the synaptobrevin-2 construct for 1 hr at room temperature . The sample was then further purified by a Superdex 75 size exclusion column chromatography ( GE Healthcare Bio-Sciences , Piscataway , NJ ) in TBS in order to remove unbound SNAP-25A and synaptobrevin-2 molcules . In order to test the surface tethering method used throughout this work ( Figures 2–6 ) , we compared complexin-1 binding with previous binding studies ( McMahon et al . , 1995; Pabst et al . , 2002; Bowen et al . , 2005; Li et al . , 2007 , 2011 ) . Syntaxin-1A was labeled with acceptor dye at the C-terminus at residue 249 . The donor dye labeling site of complexin-1 , residue 26 , was chosen based on the crystal structure ( PDB ID 1KIL ) of the complexin-1 / ternary SNARE supercomplex ( Chen et al . , 2002 ) to produce high FRET efficiency when bound to the ternary SNARE complex ( Figure 8A , B ) . Representative traces of real time measurements showed stochastic bursts of acceptor fluorescence upon complexin-1 binding to syntaxin-1A , binary , and ternary SNARE complexes ( Figure 8C–E ) . We used hidden Markov modeling to extract dwell times of the bound and unbound states ( McKinney et al . , 2006 ) . Histograms of individual dwell times were fitted with an exponential function to extract the kinetic rate constants koff and kon ( Figure 8F , G ) . From the measured rate constants , we calculated equilibrium dissociation constants ( KD ) ( Figure 8H ) . For the complexin-1 interaction with the ternary SNARE complex , the koff and kon rate constants were 0 . 12 ± 0 . 03 s-1 and 1 . 62 ± 0 . 1 μM-1s-1 , respectively , resulting in KD = 0 . 07 ± 0 . 01 μM ( Figure 8F–H ) . These values agree reasonably well with previous single molecule measurements with ternary SNARE complex that used reconstituted full length syntaxin-1A ( Li et al . , 2007 ) , validating the surface-tethering method used in this work . For the complexin-1 interaction to the binary SNARE complex the koff and kon rate constants were 1 . 8 ± 0 . 07 s-1 and 0 . 92 ± 0 . 03 μM-1s-1 , respectively , resulting in KD = 2 . 0 ± 0 . 03 μM , in good agreement with previous isothermal titration calorimetry ( ITC ) experiments ( KD = 2 . 4 ± 0 . 2 μM , ref . [Li et al . , 2011] ) . For the complexin-1 interaction to syntaxin-1A alone , the koff and kon rates constants were 2 . 96 ± 0 . 44 s-1 and 0 . 16 ± 0 . 001 μM-1s-1 , respectively , resulting in KD = 18 . 5 ± 2 . 8 μM . In summary , complexin-1 has a weak affinity to syntaxin-1A alone , and binds progressively more strongly from binary to ternary SNARE complex ( Figure 8H ) . We tested if the superclamp and no-clamp mutations affect binding to ternary SNARE complex in our single molecule binding assay ( Figure 8I–K ) . As control we also tested the 4M mutant ( Figure 1A ) . The kinetic rate constants of both the superclamp and no-clamp mutants were similar to that of wildtype complexin-1 , resulting in KD = 0 . 063 ± 0 . 01 μM and 0 . 068 ± 0 . 01 μM , respectively ( Figure 8I–K ) . As expected , the kon rate for the complexin-1 4M mutant was not measurable on the time scale of our experiment , and koff rate was 1 . 34 ± 0 . 01 s-1 , which is more than 10 times weaker than that for wildtype complexin-1 . The lack of an effect of the no-clamp and superclamp mutants on the affinity between complexin-1 and SNARE complex suggests that under the particular conditions of this experiment , the binding is dominated by the interaction between the core domain of complexin-1 and ternary SNARE complex . | Nerve cells communicate via electrical signals that travel at high speeds . However , these signals cannot pass across the gaps – called synapses – that separate one nerve cell from the next . Instead , signals pass between nerve cells via molecules called neurotransmitters that are released from the membrane of the first cell and recognized by receptors in the membrane of the next . Prior to being released , neurotransmitters are packaged inside bubble-like structures called vesicles . The synaptic vesicles must fuse with the cell membrane in order to release their contents into the synaptic cleft . Proteins called SNAREs work together with other proteins to allow this membrane fusion to occur rapidly after the electrical signal arrives . Complexin is a synaptic protein that binds tightly to a complex of SNARE proteins to regulate membrane fusion . This protein activates the quick release of neurotransmitters , which is triggered by an increase in calcium ions as the electrical signal reachess the synapse . Complexin also regulates a different type of neurotransmitter release , which is known as “spontaneous release” . The complexin protein is made up of different regions , each of which is required for one or more of the protein’s activities . However , it is not clear how these regions , or domains , interact with SNAREs and other proteins to enable complexin to perform these roles . Choi et al . have now investigated whether the different activities of mammalian complexin are related to the structure that it adopts when it interacts with the SNARE complex . Complexes of SNARE proteins were assembled with one of the SNARE proteins tethered to a surface for imaging . Next , a light-based imaging technique called single molecule Förster resonance energy transfer ( or FRET ) was used to monitor how complexin interacts with the SNARE complex . This technique allows individual proteins that have been labeled with fluorescent markers to be followed under a microscope and can show how they interact in real-time . Using this approach , Choi et al . showed that complexin could adopt two different shapes or conformations when it binds to the SNARE complex . In one , complexin interacted closely with the SNARE complex so that it made part of the complex change shape . In the other , complexin was able to bridge two SNARE complexes . Complexin can therefore interact with SNARE complexes in different ways by using different regions of the protein . These findings provide insight into how complexin may regulate membrane fusion via the SNARE complex . In the future , single molecule FRET could be used to study other proteins found at synapses and understand the other steps that regulate the release of neurotransmitters . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"neuroscience"
] | 2016 | Complexin induces a conformational change at the membrane-proximal C-terminal end of the SNARE complex |
DNA double strand breaks ( DSBs ) have detrimental effects on cell survival and genomic stability , and are related to cancer and other human diseases . In this study , we identified microtubule-depolymerizing kinesin Kif2C as a protein associated with DSB-mimicking DNA templates and known DSB repair proteins in Xenopus egg extracts and mammalian cells . The recruitment of Kif2C to DNA damage sites was dependent on both PARP and ATM activities . Kif2C knockdown or knockout led to accumulation of endogenous DNA damage , DNA damage hypersensitivity , and reduced DSB repair via both NHEJ and HR . Interestingly , Kif2C depletion , or inhibition of its microtubule depolymerase activity , reduced the mobility of DSBs , impaired the formation of DNA damage foci , and decreased the occurrence of foci fusion and resolution . Taken together , our study established Kif2C as a new player of the DNA damage response , and presented a new mechanism that governs DSB dynamics and repair .
DNA damage is frequently induced by both endogenous metabolic products and exogenous genotoxic agents . Upon DNA damage , the cell promptly activates the cellular DNA damage response ( DDR ) , a surveillance mechanism that leads to DNA repair , cell cycle arrest ( checkpoint ) , and apoptosis ( Li and Zou , 2005; Lou and Chen , 2005; Zhou and Elledge , 2000 ) . Among all types of DNA damage , DNA double strand break ( DSB ) is of great toxicity and deleterious consequences . It is therefore crucial for cells to efficiently repair DSBs , whereas defects in DSB repair have been linked to cancer , immunodeficiency , neurological diseases , and aging ( Jalal et al . , 2011; Liang et al . , 2009; Sancar et al . , 2004 ) . The cell employs two major evolutionarily-conserved mechanisms , non-homologous end joining ( NHEJ ) and homologous recombination ( HR ) to repair DNA DSBs ( Goodarzi and Jeggo , 2013; Sancar et al . , 2004 ) . HR restores the broken DNA strands using an intact strand as template , and is available in S and G2 phases after replication of chromatin DNA ( Jasin and Rothstein , 2013 ) . By comparison , NHEJ directly re-ligates the two broken ends of a DSB , and is accessible throughout the entire interphase ( Davis and Chen , 2013; Lieber , 2010 ) . In addition to these core pathways of DSB repair , the spatiotemporal regulation of DSBs has emerged as a new aspect of DNA repair ( Amitai et al . , 2017; Chuang et al . , 2006; Chung et al . , 2015; Hauer and Gasser , 2017; Krawczyk et al . , 2012; Lemaître and Soutoglou , 2015; Levi et al . , 2005; Lottersberger et al . , 2015; Marcomini et al . , 2018; Marnef and Legube , 2017; Miné-Hattab and Rothstein , 2013; Neumaier et al . , 2012; Schrank et al . , 2018 ) . Potentially , the physical mobility of DSBs mediates the sub-nuclear organization and positioning of DSBs to facilitate DNA repair . However , the precise mechanisms which propel and regulate DSB mobility remain largely obscure . Microtubules ( MTs ) are composed of α/β tubulin dimers , and responsible for a variety of cell movements , including the intracellular transport of various vesicles and organelles , and separation of chromosomes in mitosis ( Dogterom et al . , 2005; Forth and Kapoor , 2017; Maizels and Gerlitz , 2015 ) . For example , cargos , including proteins , nucleic acids and organelles , can be moved along MTs by the action of motor proteins which utilize ATP hydrolysis to produce force and movement ( Dogterom et al . , 2005; Forth and Kapoor , 2017; Maizels and Gerlitz , 2015 ) . A major group of molecular motors involved in intracellular transport are kinesins named Kif ( kinesin superfamily protein ) . There are several dozen Kifs in mammalian cells to constitute at least 14 kinesin families ( Hirokawa et al . , 2009; Lawrence et al . , 2004 ) . Unlike most kinesins , Kif2C , also known as Mitotic Centromere Associated Kinesin or MCAK , and other members of the kinesin-13 family do not utilize their ATPase activities to transport cargos , but rather to depolymerize MTs by disassembling tubulin subunits at polymer ends ( Desai et al . , 1999; Hunter et al . , 2003; Walczak et al . , 2013; Wordeman and Mitchison , 1995 ) . During cell division , Kif2C regulates MT dynamics and ensures the proper attachment of MTs to kinetochores , and thereby directing the positioning and movement of chromosomes ( Ganem et al . , 2005; Kline-Smith et al . , 2004; Manning et al . , 2007 ) . In this study we identify and characterize Kif2C as a new factor involved in DSB repair; Kif2C is required for efficient DSB repair via both HR and NHEJ; and interestingly , Kif2C facilitates DSB mobility and modulates the formation , fusion , and resolution of DNA damage foci .
As described in our previous study ( Zhu et al . , 2017 ) , we utilized DNA DSB-mimicking dA-dT oligonucleotides to isolate potential DNA damage-associated proteins in Xenopus egg extract , a cell-free system well-defined for studying DNA damage repair and signaling ( Guo et al . , 1999; Lupardus et al . , 2007 ) . Along with Ku70 , PARP1 , RPA , and many other factors known to be involved in DSB repair , Kif2C was proteomically identified as a co-precipitated protein of dA-dT . We confirmed , in both Xenopus egg extracts and human cell lysates , that Kif2C bound another , and longer , DSB-mimicking template ( Figure 1A and B ) . We then supplemented in the extract either uncut , circular plasmid DNA , or linearized plasmid DNA with free DSB ends . Interestingly , Kif2C associated specifically with the cut plasmid DNA ( Figure 1C ) , further indicating that Kif2C is a DSB-associated protein . Next , we carried out proteomic analysis to identify proteins that were associated with Kif2C . This effort recovered a number of well-established DNA damage response proteins , including Ku70/Ku80 , a DSB end binding complex , H2AX , a histone variant that is phosphorylated in chromatin regions flanking DSBs , and PARP1 , an early responder of various DNA lesions ( Figure 1D ) . The association of Kif2C with these DNA damage factors was subsequently confirmed using both pull-down and immunoprecipitation ( Figure 1E , Figure 1—figure supplement 1A and B ) . Treatment with DNase did not disrupt the protein association ( Figure 1—figure supplement 1C ) , suggesting that it was not mediated by DNA . It has been revealed that the catalytic function of Kif2C is mediated through a motor domain located in the middle region of the protein ( Ems-McClung et al . , 2007; Maney et al . , 2001 ) . Interestingly , both this middle region and the N-terminus of Kif2C exhibited appreciable levels of associations with DNA repair proteins ( Figure 1—figure supplement 1D and E ) , suggesting the involvement of these motifs in DNA repair . The identification of Kif2C as a potential DSB-associated protein was largely unexpected , given that MT assembly is viewed as a cytoplasmic event , except in mitosis after nuclear envelop breakdown . On the other hand , Kif2C is primarily localized to the nucleus in interphase , but the function of Kif2C in intra-nuclear events is unknown . We showed in HeLa cells that Kif2C was recruited to DNA damage sites induced by laser micro-irradiation ( Figure 2A and Video 1 ) . Kif2C was enriched at laser-irradiated sites within 1 min , indicating it as an early responder to DNA damage ( Figure 2A and B ) . We confirmed subsequently that Kif2C co-localized with γ-H2AX foci induced by ionized radiation ( IR , Figure 2C ) ; Kif2C foci co-localized and co-migrated with 53BP1 foci in cells treated with etoposide ( Figure 2—figure supplement 1 & Video 2 ) . Immunofluorescent staining analysis of endogenous Kif2C revealed consistent pattern of co-localization with IR-induced γ-H2AX foci ( Figure 2—figure supplement 2A and B ) . The fast recruitment of Kif2C to DNA damage sites prompted us to examine its dependence on PARP1-mediated PARylation , which undergoes rapid induction ( <1 min ) and removal ( 5–10 min ) . Interestingly , PARP inhibition disrupted the initial recruitment of Kif2C to laser-induced DNA damage sites; in the presence of a PARP inhibitor ( PARPi ) , Kif2C slowly accumulated at DNA damage sites at about 10 min ( Figure 2D and E ) . By contrast , the sustained , but not the initial , recruitment of Kif2C was dependent on ATM ( Figure 2D and E ) . To reveal additional molecular insights into the DNA damage recruitment of Kif2C , we generated multiple truncated segments of Kif2C , and examined their localization in laser-irradiated cells . Interestingly , the N-terminus of Kif2C exhibits efficient recruitment to DNA damage sites; the middle region of Kif2C containing the catalytic motif was very weakly enriched at DNA damage sites; and the C-terminus of Kif2C did not accumulate at DNA damage sites ( Figure 2F and G ) . Consistent with the strong recruitment of Kif2C N-terminus to DNA damage sites , a Kif2C mutant deleted of the N-terminus was deficient in the DNA damage recruitment ( Figure 2—figure supplement 2C and D ) . To identify minimal elements within the N-terminus that mediate DNA damage recruitment , we generated a series of truncation mutants within the N-terminus ( Figure 2H ) . Interestingly , the efficient recruitment of Kif2C N-terminus depended on both a short five amino acid ( aa 86–90 ) and the neck domain ( Figure 2H ) . The neck domain of Kif2C was shown to play a role in MT depolymerization ( Maney et al . , 2001 ) , hence , our study indicates an additional function of this domain in the DNA damage recruitment of Kif2C . By comparison , the aa 86–90 region lies outside of the minimal functional domain of Kif2C’s MT-depolymerizing activity and is not associated with any known mitotic functions of Kif2C . Interestingly , both of these motifs are important for Kif2C recruitment to DNA damage sites as full-length Kif2C deleted of either one exhibited reduced recruitment to the sites of laser cut ( Figure 2I and J ) . Consistent with the recruitment deficiency , these mutants also exhibited reduced association with DNA repair proteins ( Figure 2—figure supplement 3 ) . As we revealed the recruitment of Kif2C to DNA damage , and the association of Kif2C with DSB templates and repair factors , we set out to investigate the function of Kif2C in the DDR . Interestingly , Kif2C knockdown in HeLa cells led to γ-H2AX induction ( Figure 3A ) . The induction of γ-H2AX was also detected in U2OS cells deleted of Kif2C using CRISPR-Cas9-mediated gene editing ( Figure 3B ) . Moreover , cells depleted of Kif2C exhibited increased foci formation of γ-H2AX and 53BP1 ( Figure 3—figure supplement 1 ) . These lines of evidence suggested that Kif2C plays a role in DNA repair , and its removal caused accumulation of endogenous DNA damage . Consistent with this hypothesis , accumulation of DNA breaks in Kif2C knockout cells was shown using a single cell electrophoresis ( comet ) assay ( Figure 3C ) . As expected , the re-expression of RNAi-resistant Kif2C rescued γ-H2AX induction ( Figure 3D ) . By comparison , a G495A Kif2C mutant defective in ATP hydrolysis and MT depolymerization , as characterized previously ( Wang et al . , 2012b ) , was ineffective in suppressing endogenous DNA damage caused by Kif2C depletion ( Figure 3D and Figure 3—figure supplement 1 ) , suggesting that the ATPase activity of Kif2C is required for its function in DNA repair . Previously reported structural insights into the enzymatic action of Kif2C revealed that tubulin-binding , in addition to ATPase , is required for MT depolymerization ( Ritter et al . , 2016; Wang et al . , 2017 ) . For example , a β5 motif with in the motor domain of Kif2C recognizes the distal end of β-tubulin , and R420S , a specific mutation in this motif disrupted tubulin-binding and MT depolymerization ( Ritter et al . , 2016 ) . Like G495A , R420S mutant failed to rescue the accumulation of γ-H2AX in Kif2C knockout cells ( Figure 3E ) , despite that these mutants were expressed at similar levels as WT ( Figure 3D and E ) , and exhibited nuclear localization and DNA damage recruitment ( Figure 3—figure supplement 2A ) . Kif2C depletion or mutation did not cause significant disruption of cell cycle progression ( Figure 3—figure supplement 2B ) . Interestingly , Kif2C depletion did not additively enhance the induction of γ-H2AX in cells pre-treated with nocodazole , an inhibitor of MT assembly , suggesting that Kif2C functions in DNA repair in the context of MT assembly ( Figure 3—figure supplement 3A ) . Moreover , the Δ86–90 Kif2C mutant deficient in DNA damage recruitment was incapable of suppressing endogenous DNA damage in Kif2C KO cells ( Figure 3F ) , indicating that the DNA damage recruitment of Kif2C is required for the prevention of DSB accumulation . Together , these findings suggested that both the DNA damage recruitment of Kif2C and its catalytic activity are likely involved in the DDR . The Kwok laboratory previously identified DHTP ( ( ( Z ) −2- ( 4- ( ( 5- ( 4-chlorophenyl ) −6- ( isopropoxycarbonyl ) −7-methyl-3-oxo-3 , 5-dihydro-2H-thiazolo[3 , 2-a]pyrimidin-2-ylidene ) methyl ) phenoxy ) acetic acid ) ) as an allosteric inhibitor of Kif2C ( Talje et al . , 2014 ) . Interestingly , DHTP treatment in HeLa cells phenocopied Kif2C depletion and caused γ-H2AX accumulation ( Figure 3G ) . Although Kif2C also plays a role in mitosis ( Manning et al . , 2007 ) , DHTP induced γ-H2AX accumulation efficiently in thymidine-arrested interphase cells ( Figure 3H ) , indicating that mitotic defects were not the primary cause of DNA damage . In line with the involvement of Kif2C in the DDR , Kif2C depletion significantly enhanced the response of HeLa cells to DNA damage treatment , as judged by both reduced cell viability and increased cell death ( Figure 3I and J ) . A similar effect was confirmed also in SCC38 cells ( Figure 3—figure supplement 3B and C ) , or in HeLa cells with DHTP treatment ( Figure 3—figure supplement 3D ) . WT , but not Δ86–90 , Kif2C rescued etoposide sensitivity in Kif2C knockout cells ( Figure 3K ) , confirming the direct involvement of Kif2C in the DDR . To assess further the impact of Kif2C on DNA repair , the kinetics of γ-H2AX post-IR treatment was probed in control and Kif2C depleted cells . Compared to the control HeLa cells , those treated with Kif2C siRNA exhibited more sustained γ-H2AX ( Figure 4A–C ) . A similar effect was observed when comparing Kif2C knockout U2OS cells to control U2OS cells ( Figure 4D–F ) . The DNA repair deficiency caused by Kif2C depletion was also confirmed using single cell electrophoresis ( Figure 4—figure supplement 1 ) . Next , we sought to evaluate the impact of Kif2C on specific DSB repair pathways . The repair activity of NHEJ and HR was measured using an intra-chromosomal , I-SceI-induced NHEJ assay and an intra-chromosomal I-SceI-induced HR reporter system , respectively ( Gunn and Stark , 2012 ) ( Figure 4G and H ) . Interestingly , Kif2C depletion reduced both NHEJ and HR by 3–5 fold ( Figure 4G and H ) . It is very intriguing how Kif2C promotes DSB repair , given that Kif2C is unlikely to function as a core factor for both NHEJ and HR . We speculated that Kif2C might function in regulation of DSB movement and dynamics in light of several existing findings . First , emerging evidence in both yeast and mammalian cells indicated increased chromatin mobility at sites of DNA DSBs ( Chuang et al . , 2006; Krawczyk et al . , 2012; Lemaître and Soutoglou , 2015; Levi et al . , 2005; Lottersberger et al . , 2015; Marnef and Legube , 2017 ) , but the underlying mechanism is largely unclear . Second , MTs are well known to support intracellular trafficking of proteins , chromosomes , and other materials , and kinesins are known to produce mechanical work from ATP hydrolysis ( Dogterom et al . , 2005; Forth and Kapoor , 2017; Maizels and Gerlitz , 2015 ) . Recent studies showed that MT dynamics enhanced the motion of chromatin , especially telomeres , in response to DNA damage ( Lawrimore et al . , 2017; Lottersberger et al . , 2015 ) . Third , we showed that the MT depolymerase activity of Kif2C , mediated by ATP hydrolysis and tubulin-binding , is required for the prevention of γ-H2AX accumulation . To investigate this potential role of Kif2C , we quantified the mobility of etoposide-induced DSBs , as marked by GFP-53BP1 foci . The 3D trajectories of unbiasedly selected foci were tracked to determine the distance traveled by these foci ( Figure 5A ) . We observed that Kif2C depletion , or inhibition of its MT depolymerase activity by DHTP , impaired the mobility of DSBs ( Figure 5B–D ) . This effect of Kif2C suppression was comparable to that of Taxol treatment which inhibits MT dynamics and was previously shown to retard DSB movement ( Figure 5C and D ) ( Lottersberger et al . , 2015 ) . To clarify if Kif2C specifically regulates the mobility of damaged chromatin , we analyzed the movement of Centromere Protein B ( CENP-B ) and Pre-MRNA Processing Factor 6 ( PRPF6 ) , as controls . Both CENP-B and PRPF6 form distinct punctate foci in the nucleus that are not DNA damage-induced , and their motilities can indicate undamaged , general chromatin dynamics , and general intra-nuclear dynamics , respectively . Intra-nuclear CENP-B and PRPF6 foci are relatively less dynamic than 53BP1 foci , and there was no significant difference between WT and Kif2C KO or DHTP treatment ( Figure 5—figure supplement 2A and B ) . On the other hand , Taxol treatment reduced foci dynamics of CENP-B and PRPF6 ( Figure 5—figure supplement 3 ) , suggesting a general effect in nuclear dynamics . These data demonstrated that Kif2C mediates DNA damage mobility in a specific manner without affecting other cellular dynamics in general . Formation of DNA damage foci is a landmark feature of the DDR , but the precise mechanism of this process is still unknown ( Huen and Chen , 2010 ) . Previous studies analyzing these foci as potential repair centers suggested the clustering of multiple DSB ends and the subsequent formation of macro-domains ( Asaithamby and Chen , 2011; Aten et al . , 2004; Aymard et al . , 2017; Neumaier et al . , 2012; Roukos et al . , 2013 ) . In yeast cells , persistent DSBs roam within the nucleus to form these repair centers ( Lisby et al . , 2003; Marnef and Legube , 2017 ) . Mammalian DSBs were shown to travel for a similar distance ( ~2 μM ) as yeast DSBs . However , due to a much larger volume of the mammalian nucleus , mammalian DSBs do not roam within the nucleus , but join each other in close proximity ( Marnef and Legube , 2017; Neumaier et al . , 2012; Roukos et al . , 2013 ) . We analyzed the dynamics of DNA damage foci using high-resolution , live-cell imaging ( Figure 5E ) . Interestingly , Kif2C depletion or pre-treatment with DHTP or Taxol reduced the formation of DNA damage foci ( Figure 5F ) , despite the level of DNA damage being rather elevated under Kif2C suppression ( Figure 5—figure supplement 1B ) . To further assess the impact of Kif2C on the dynamics of DNA damage foci , we first allowed the establishment of DNA damage foci ( Figure 5G ) , and then challenged cells with DHTP or Taxol . Interestingly , we observed that the occurrence of foci fusion events was decreased by Kif2C depletion or inhibition ( Figure 5H , Video 3 ) ; furthermore , foci resolution ( disappearance ) was also markedly suppressed by these treatments ( Figure 5I and J and Video 4 ) . Presumably , foci fusion represents the movement of DSBs to form larger repair centers/foci , and foci disappearance reflects DNA repair or reassembly of foci . Together , we showed that Kif2C mediates the formation and dynamics of DNA damage foci . As we showed that both PARP1 and ATM act upstream to mediate the recruitment of Kif2C to DNA damage sites , we speculated that PARP1 and ATM play a role in regulation of DSB movement and DNA damage foci formation . Indeed , inhibition of PARP and ATM both significantly reduced the mobility of GFP-53BP1 foci ( Figure 6A and C ) . Interestingly , PARP or ATM inhibition only moderately retarded GFP-53BP1 foci movement in Kif2C depleted cells ( Figure 6C ) , indicating that PARP and ATM govern DSB mobility largely , although not exclusively , through Kif2C . On the other hand , these data also indicated that Kif2C depletion did not further suppress the mobility of GFP-53BP1 in cells with PARP or ATM inhibition ( Figure 6—figure supplement 1 ) , suggesting that the function of Kif2C in this process is dependent on both PARP1 and ATM , which act upstream to mediate the DNA damage recruitment of Kif2C .
As a member of the MT depolymerase family , Kif2C was shown to govern several aspects of cell division in mitosis , including spindle assembly , chromosome congression , and kinetochore-MT attachment ( Manning et al . , 2007; Sanhaji et al . , 2011 ) . The role of Kif2C in interphase cells is less characterized , despite that its dominant localization in the nucleus suggests possible functions of Kif2C in intra-nuclear processes . Interestingly , we reported here a direct involvement of Kif2C in DNA repair , as a previously undefined interphase function of Kif2C . First , we showed that Kif2C associated with DSB-mimicking structures in Xenopus egg extracts and human cell lysates . Consistently , Kif2C bound several established DNA repair factors , including PARP1 , H2AX , and Ku70/80 . Second , Kif2C was recruited to DNA damage sites in interphase cells via two distinct mechanisms . The initial recruitment of Kif2C occurred within seconds in a PAR-dependent manner , whereas the sustained localization of Kif2C at DNA damage sites was disrupted by ATM inhibition . Thus , we characterized Kif2C as a downstream factor of the PARP1 and ATM-mediated DNA damage responses . Third , Kif2C was required for efficient DNA DSB repair via both NHEJ and HR; consequently , depletion or inhibition of Kif2C leads to both accumulation of endogenous DSB and DNA damage hypersensitivity . Interestingly , a Kif2C mutant ( Δ86–90 ) specifically deficient in DNA damage recruitment was unable to rescue DSB accumulation or etoposide-sensitivity in Kif2C depleted cells . Furthermore , Kif2C inhibition led to DSB accumulation in cells synchronized at G1/S . Together , our studies revealed a new role of Kif2C in facilitating DNA DSB repair that is distinct from its known functions in mitotic progression . While the core pathways of NHEJ and HR have been well studied , an emerging topic of DNA repair lies in the spatiotemporal dynamics of DSBs ( Hauer and Gasser , 2017; Lemaître and Soutoglou , 2015; Marnef and Legube , 2017; Miné-Hattab and Rothstein , 2013 ) . In particular , clustering of DSB ends into ‘repair centers’ has been observed for longer than a decade; and the increased mobility of DSB ends within the nucleus has been reported in yeast and mammalian cells ( Chuang et al . , 2006; Chung et al . , 2015; Krawczyk et al . , 2012; Lemaître and Soutoglou , 2015; Levi et al . , 2005; Lottersberger et al . , 2015; Marnef and Legube , 2017; Neumaier et al . , 2012 ) . However , mechanistic understandings of these phenomena are largely absent within the context of current DDR regulators . We revealed in this study that a specific kinesin motor protein , Kif2C , directly promotes DSB mobility and mediates the formation and fusion of DNA damage foci . Our findings suggested a model that , upon recruitment to DSBs , Kif2C propels the physical movement of damaged chromatin to promote DNA repair , in a manner that relies on the ATPase and tubulin-binding activities of Kif2C; Kif2C facilitates the formation of DNA damage foci , which potentially involves the mobility and clustering of DSBs , as shown previously ( Asaithamby and Chen , 2011; Aten et al . , 2004; Aymard et al . , 2017; Neumaier et al . , 2012; Roukos et al . , 2013 ) . In addition to foci formation , we observed also the occurrence of DSB foci fusion and resolution , indicating that DSBs may undergo dynamic organization and reorganization during DNA repair . These events are reduced by Kif2C depletion or inhibition , thus implicating a role of Kif2C in these processes . In addition to the underlying mechanism , the functional impact of DSB mobility and foci formation remains to be better clarified . It has been generally hypothesized that this pattern of DSB dynamics facilitates DSB repair , for example by keeping DSB ends in close proximity , and increasing the local concentration of repair proteins ( Lottersberger et al . , 2015; Miné-Hattab and Rothstein , 2012; Miné-Hattab and Rothstein , 2013 ) . Furthermore , mounting evidence suggested that DSB mobility may enable homology search during HR ( Marnef and Legube , 2017; Miné-Hattab and Rothstein , 2012; Schrank et al . , 2018 ) . On the other hand , MT and the linker of the nucleoskeleton and cytoskeleton ( LINC ) -mediated DSB mobility was shown to promote NHEJ of dysfunctional telomeres ( Aymard et al . , 2017; Lottersberger et al . , 2015 ) . By characterizing Kif2C as a specific regulator of DSB mobility , our study provided an opportunity to assess the functional involvement of DSB dynamics in repair . Interestingly , we demonstrated that Kif2C is required for the efficient DSB repair via both HR and NHEJ . Future studies shall be directed to determine more precisely how Kif2C mediates DSB dynamics , and how this process may interact with the core repair machinery of HR and NHEJ . We showed that the initial or sustained recruitment of Kif2C to DNA damage sites is dependent on PARP1 or ATM activation , respectively . Thus , we set out to investigate if either PARP1 or ATM governs DNA damage dynamics via Kif2C . Of note , ATM was shown to govern DSB mobility in previous studies ( Becker et al . , 2014; Dimitrova et al . , 2008 ) . PARP1 is known to play a crucial role in sensing DNA damage , recruiting repair factors , and modulating chromatin structure , but its involvement in DSB movement was not reported . We clarified in our study that PARP1 and ATM inhibition markedly retarded DSB mobility . Inhibition of PARP1 or ATM in Kif2C-depleted cells less significantly affected DSB mobility , validating that Kif2C is a major downstream of PARP1 and ATM in regulation of DSB mobility , but at the same time , suggesting the existence of redundant pathways . Since the first report of nuclear actin in Xenopus , the existence of the actin network in the nucleus , and its function in nuclear architecture and genomic regulation , have been well recognized ( Belin et al . , 2015; Caridi et al . , 2018; Grosse and Vartiainen , 2013; Misu et al . , 2017; Schrank et al . , 2018 ) . By comparison , MT assembly is viewed as a cytoplasmic event , except in mitosis after nuclear envelop breakdown . Thus , the function of Kif2C , a MT depolymerase , in DNA repair is largely unexpected . In particular , our studies using established Kif2C mutants and inhibitor suggested that the ATPase and tubulin-binding activities of Kif2C were indispensable for suppressing DNA damage accumulation . Potentially in line with our findings , previous studies showed that MT poisons caused endogenous DNA damage and reduced DNA repair ( Branham et al . , 2004; Lottersberger et al . , 2015; Poruchynsky et al . , 2015; Rogalska and Marczak , 2015 ) . While cytoplasmic MTs can indirectly influence the DDR , for example , via the nuclear import of repair factors ( Poruchynsky et al . , 2015 ) , or via the LINC complex ( Aymard et al . , 2017; Lottersberger et al . , 2015 ) , our study suggested a rather direct involvement of nuclear MT components in the DDR . This is particularly relevant as many kinesins , as well as low levels of tubulins , are present in the nucleus ( Akoumianaki et al . , 2009; Kırlı et al . , 2015; Kumeta et al . , 2013 ) . Interestingly , Kif4A and γ-tubulin were shown to associate with Rad51 and possibly other repair proteins ( Lesca et al . , 2005; Wu et al . , 2008 ) . Yeast kinesin-14 and nuclear pore proteins mediate the perinuclear tethering of telomeric DSBs in yeast cells ( Chung et al . , 2015 ) . Moreover , recent evidence demonstrated that inhibitors of MT assembly reduced the mobility of DSBs ( Lawrimore et al . , 2017; Lottersberger et al . , 2015 ) , further suggesting a nuclear function of MTs . To account for the potential role of MT dynamics in DNA repair , a provocative possibility is that MT assembly occurs in the nucleus after DNA damage , presumably at a low and transient level . Along this line , a previous study visualized increased tubulin nucleation and MT rearrangement after DNA damage , although it was not defined if this event occurs at least partially in the nucleus ( Porter and Lee , 2001 ) . A study in yeast cells detected the assembly of long and stable MTs in the interphase nuclei when cell enters quiescence ( Laporte et al . , 2013 ) ; a more recent study visualized DNA damage-inducible intra-nuclear microtubule filaments ( DIMs ) in yeast cells using GFP-tagged tubulin ( Oshidari et al . , 2018 ) . However , the formation of detectable DIMs in mammalian cells remains to be demonstrated . On the other hand , as an alternative hypothesis to be considered , MT filament assembly via tubulin nucleation may not occur in the nucleus of damaged mammalian cells , but rather , certain regulators and mechanisms of MT assembly/disassembly are employed by the DDR machinery to govern the dynamic movement and repair of broken DNA ends . In all cases , the characterization of Kif2C as a new DDR factor that mediates DNA damage movement and foci formation sheds new light on the spatiotemporal regulation of DNA damage dynamics . Future studies building on these findings shall further delineate the involvement of MT regulators in the DNA damage response .
Human cervix carcinoma ( HeLa ) and bone osteosarcoma epithelial ( U2OS ) lines , authenticated by ATCC , were maintained in Dulbecco’s modified Eagle medium ( DMEM , Hyclone ) with 10% fetal bovine serum ( FBS , Hyclone ) . Human head and neck squamous cell carcinoma UM-SCC-38 cells were authenticated and maintained as in previous studies ( Brenner et al . , 2010; Wang et al . , 2012a ) . Cell viability and death assays were performed as in our previous study ( Wang et al . , 2014 ) . Briefly , cells were incubated for 1–4 days . The numbers of viable cells were counted using a hemocytometer . To measure cell death , trypan blue staining was performed by mixing 0 . 4% trypan blue in PBS with cell suspension at a 1:10 ratio . Ionized radiation was performed using an X-ray cabinet ( RS-2000 Biological irradiator ) . Transfection of expression vectors was carried out using Lipofectamine 2000 ( Invitrogen ) or TransIT transfection reagents ( from Mirus Bio ) following the protocol recommended by the manufacturer . SiRNA targeting Kif2C ( 5-AUCUGGAGAACCAA GCAU-3’ , Integrated DNA Technologies ) was transfected into cells using Lipofectamine RNAi MAX ( Invitrogen ) . A non-targeting control siRNA was used as a control . Kif2C knockout U2OS cells was generating using the established CRISPR-cas9 gene editing method with following single guide ( sg ) sequence: GCAAGCTGACACAGGTGCTG in lentiCRISPR v2 vector ( a gift from Feng Zhang via Addgene plasmid # 52961 ) ( Sanjana et al . , 2014 ) . Knockout efficiency was assessed by TIDE ( Tracking of Indels by Decomposition ) analysis using the web tool ( https://tide . deskgen . com/ ) and confirmed by western blot . Xenopus Kif2C gene was cloned from a Xenopus oocyte cDNA library , and inserted into a pMBP vector with an N-terminal MBP-tag . Kif2C G495A , R420S , Δ86–90 , Δneck , and siRNA-resistant mutants were generated using site-directed mutagenesis ( Agilent ) following the protocol recommended by the manufacturer . The human Kif2C expression vector was obtained from Addgene ( mEmerald-MCAK-C-7 , a gift from Michael Davidson via Addgene , plasmid # 54161 ) . Biotin-labeled double strand DNA fragment ( dsDNA , 500 bp ) was generated using biotin-11-ddUTP ( Thermo Scientific , #R0081 ) incorporation , and PCR amplification using Taq polymerase and a pMBP vector ( as template ) . Biotin-labeled DNA ( produced as above ) or biotin dA-dT ( 70 mer ) was conjugated on streptavidin magnetic beads ( New England Biolabs , #S1420S ) and incubated in Xenopus egg extracts and HeLa cell lysates . The beads were re-isolated using a magnet , washed five times , and then resolved by SDS-PAGE . Homologous recombination assay was performed in a HeLa-derived cell line stably integrated with a DR-GFP reporter cassette ( a gift from Dr . Jeffrey Parvin at the Ohio State University ) . The reporter consisted of direct repeats of two differentially mutated green fluorescent proteins ( GFP ) , Sce GFP and iGFP . SceGFP contains an I-SceI recognition site and in-frame termination codons . An 812 bp internal GFP fragment ( iGFP ) was used by HR to repair the DSB . Briefly , cells were seeded at 3 × 105 cells per well in a 6-well plate one day before siRNA treatment . After removing the siRNA , the cells were grown for 48 hr in fresh medium and transfected with an expression vector of I-SceI endonuclease ( a gift from Dr . Maria Jasin at Memorial Sloan Kettering Cancer Center ) . In this assay , a full-length GFP is expressed only after DSBs introduced by I-SceI endonuclease are repaired by HR , and the level of full-length GFP ( and control β-actin ) expression was quantified by immunoblotting and NIH ImageJ . The NHEJ assay was performed in U2OS-EJ5 cells ( a gift from Dr . Jeremy Stark at the Beckman Research Institute of the City of Hope ) . Briefly , cells were seeded at 3 × 105 cells per well in a 6-well plate 24 hr before siRNA treatment . After removing the siRNA , the cells were grown for 48 hr in fresh medium and transfected with an expression vector of I-SceI endonuclease . In this assay , GFP is expressed only after DSBs introduced by I-SceI endonuclease are repaired by NHEJ , and the level of GFP expression was quantified by immunoblotting and NIH ImageJ . In both HR and NHEJ assays , approximately 3–10% cells in the control-treated groups exhibited GFP-positive . Sodium dodecyl sulfate-polyacrylamide gel electrophoresis ( SDS-PAGE ) and immunoblotting were carried out as previously described ( Ren et al . , 2017 ) , using the following antibodies: γ-H2AX ( A300-081A-M ) , and Ku80 ( A302-627A-T ) from Bethyl Laboratories ( Montgomery , TX ) ; ATM ( sc-377293 ) , DNA-PKcs ( sc-390849 ) , GFP ( sc-9996 ) , γ-H2AX ( sc-517348 ) , Kif2C ( sc-81305 ) , and Ku70 ( sc-56129 ) , from Santa Cruz Biotechnology ( Dallas , TX ) ; H2B ( ab1790-100 ) and α-tubulin ( ab7291 ) from Abcam ( Cambridge , MA ) ; β-actin ( #4970T ) from Cell Signaling Technology ( Beverly , MA ) ; and Artemis ( GTX100128 ) from Genetex ( Irvine , CA ) . Cells were grown on coverglasses , washed with PBS twice , and fixed with 3% formaldehyde with 0 . 1% Triton X-100 for 30 min . 0 . 05% Saponin containing PBS was used to permeabilize the fixed cells followed by blocking with 5% goat serum for 30 min . Primary antibodies were diluted in blocking buffer and incubated with the cells for 2 hr . The cells were then incubated with Alexa Fluor secondary antibodies ( Invitrogen , 1: 2 , 000 ) for 1 hr at room temperature . The nuclei of cells were stained with 4' , 6-diamidino-2-phenylindole ( DAPI ) , and the stained cells were imaged using a Zeiss Axiovert 200M inverted fluorescence microscope at the UNMC Advanced Microscopy Core Facility . Laser micro-irradiation was performed using 405 nm laser under the Zeiss Axiovert 200M Microscope with Marianas Software ( Intelligent Imaging Innovations , Inc Denver , CO ) . EGFP-53BP1 ( or mApple-53BP1 , or Control foci constructs: EGFP-PRPF6 or Cenp-B-mCherry ) -transfected cells were seeded in ibidi µ-Dish 35 mm Quad dish the day prior to imaging . Formation of 53BP1 foci was induced by the addition of 20 µM etoposide . Other compounds such as taxol or DHTP were added to the cells either prior to ( pre-treatment ) or after ( post-treatment ) etoposide treatment . Image acquisition was carried out using a Zeiss spinning disk confocal microscopy system equipped with a 63 × PlanAprochromat oil objective . After cells expressing those constructs were located and the imaging positions were selected , microscopy recordings were then started ( usually 5 min after the last treatment , for consistency reason ) . Imaging of the control foci , that is CENP-B-mCherry and EGFP-PRPF6 , was done in a similar manner except that their formation does not require etoposide addition . For foci mobility , time-lapse recordings were done every 30 s for 10 min . For foci disappearance or fusion , recordings were done every 3 min for one hours . Z-stack images were acquired at 0 . 5 μm intervals covering a range from 6 to 8 μm . Foci tracking was done using all the acquired stacks for positional information using ImageJ ( NIH ) , and the foci number was quantified using the automatic particle counting option . For image presentation in figure panels , 2D-maximum intensity projection images were generated using the ZEN blue software . Data analysis and graph presentations were performed using Excel ( Microsoft ) and KaleidaGraph ( Synergy ) . Student’s t-test was used for statistical analysis . Mean-square displacement was calculated as previously described ( Lottersberger et al . , 2015 ) using the following equationMSD∆t=1n∑i=1nDi ( ∆t ) 2 WhereDi∆t= ( xti-xtGC-xt-∆ti-xt-∆tGC ) 2+ ( yti-ytGC-yt-∆ti-yt-∆tGC ) 2 For protein association studies , MBP Kif2C WT and mutants were expressed in BL21 bacteria cell , purified on amylose beads , and then incubated in HeLa cell lysates for 1 hr at room temperature . The beads were re-isolated using low speed centrifugation , washed five times , and then resolved by SDS-PAGE . For the plasmid DNA pull-down assay in Figure 1C , pMBP plasmid was either uncut or linearized by EcoRV endonuclease ( New England Biolabs , #R3195 ) . MBP-Kif2C was conjugated on amylose beads and incubated in Xenopus egg extracts supplemented with either uncut or linearized pMBP plasmid for 1 hr at room temperature . The beads were re-isolated using low speed centrifugation , washed five times , and then boiled in distilled water . The samples were used as templates for PCR with Taq Polymerase . Cells were washed with PBS , trypsinized , and plated in 0 . 65% low melting agarose . After solidification , slides were incubated in lysis solution ( 1 M NaCl , 3 . 5 mM N-laurylsarcosine , 50 mM NaOH ) for 2 hr . Slides were then washed , and incubated in alkaline electrophoresis buffer ( 50 mM NaOH , 2 mM EDTA ) for 30 min . After electrophoresis for 10 min at 20 V , slides were stained with propidum iodide ( 25 μg/mL ) . Eggs were rinsed in distilled water and de-jellied with 2% cysteine in 1x XB ( 1 M KCl , 10 mM MgCl2 , 100 mM HEPES pH 7 . 7 , and 500 mM sucrose ) . Eggs were washed in 0 . 2x MMR buffer ( 100 mM NaCl , 2 mM KCl , 1 mM MgCl2 , 2 mM CaCl2 , 0 . 1 mM EDTA , 10 mM HEPES ) , and activated with Ca2+ ionophore . Eggs were then washed and crushed by centrifugation at 10 , 000 g . The cytoplasmic layer was transferred to new tubes , supplemented with an energy mix ( 7 . 5 mM creatine phosphate , 1 mM ATP , 1 MgCl2 ) , and then further separated by centrifugation at 10 , 000 g for 15 min . | DNA can be damaged in many ways , and a double strand break is one of the most dangerous . This occurs when both strands of the double helix snap at the same time , leaving two broken ends . When cells detect this kind of damage , they race to get it fixed as quickly as possible . Fixing these double strand breaks is thought to involve the broken ends being moved to 'repair centers’ in the nucleus of the cell , but it was unclear how the broken ends were moved . One possibility was that the cells transport the broken ends along protein filaments called microtubules . Cells can assemble these track-like filaments on-demand to carry cargo attached to molecular motors called kinesins . However , this type of transport happens outside of the cell’s nucleus , and while there are different kinesin proteins localized inside the nucleus , their roles are largely unknown . In an effort to understand how broken DNA ends are repaired , Zhu , Paydar et al . conducted experiments that simulated double strand breaks and examined the proteins that responded . The first set of experiments involved mixing cut pieces of DNA with extracts taken from frog eggs or human cells . Zhu , Paydar et al . found that one kinesin called Kif2C stuck to the DNA fragments , and attached to many proteins known to play a role in DNA damage repair . Kif2C had previously been shown to help separate the chromosomes during cell division . To find out more about its potential role in DNA repair , Zhu , Paydar et al . then used a laser to create breaks in the DNA of living human cells and tracked Kif2C movement . The kinesin arrived within 60 seconds of the DNA damage and appeared to transport the cut DNA ends to 'repair centers' . Getting rid of Kif2C , or blocking its activity , had dire effects on the cells' abilities to mobilize and repair breaks to its DNA . Without the molecular motor , fewer double strand breaks were repaired , and so DNA damage started to build up . Defects in double strand break repair happen in many human diseases , including cancer . Many cancer treatments damage the DNA of cancer cells , sometimes in combination with drugs that stop cells from building and using their microtubule transport systems . Understanding the new role of Kif2C in DNA damage repair could therefore help optimize these treatment combinations . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2020 | Kinesin Kif2C in regulation of DNA double strand break dynamics and repair |
The visual message conveyed by a retinal ganglion cell ( RGC ) is often summarized by its spatial receptive field , but in principle also depends on the responses of other RGCs and natural image statistics . This possibility was explored by linear reconstruction of natural images from responses of the four numerically-dominant macaque RGC types . Reconstructions were highly consistent across retinas . The optimal reconstruction filter for each RGC – its visual message – reflected natural image statistics , and resembled the receptive field only when nearby , same-type cells were included . ON and OFF cells conveyed largely independent , complementary representations , and parasol and midget cells conveyed distinct features . Correlated activity and nonlinearities had statistically significant but minor effects on reconstruction . Simulated reconstructions , using linear-nonlinear cascade models of RGC light responses that incorporated measured spatial properties and nonlinearities , produced similar results . Spatiotemporal reconstructions exhibited similar spatial properties , suggesting that the results are relevant for natural vision .
The brain uses visual information transmitted by retinal neurons to make inferences about the external world . Traditionally , the visual signal transmitted by an individual retinal ganglion cell ( RGC ) has been summarized by its spatial profile of light sensitivity , or receptive field ( RF ) , measured with stimuli such as spots or bars ( Chichilnisky , 2001; Kuffler , 1953; Lettvin et al . , 1959 ) . Although intuitively appealing , this description may not reveal how the spikes from a RGC contribute to the visual representation in the brain under natural viewing conditions . In particular , because of the strong spatial correlations in natural images ( Ruderman and Bialek , 1994 ) , the response of a single RGC contains information about visual space well beyond its RF . Thus , across the RGC population , the responses of many individual cells could contain information about the same region of visual space , and it is not obvious how the brain could exploit this potentially redundant information ( Puchalla et al . , 2005 ) . Complicating this issue is the fact that there are roughly twenty RGC types , each covering all of visual space with their RFs , and each with different spatial , temporal , and chromatic sensitivity ( Dacey et al . , 2003 ) . Furthermore , RGCs show both stimulus-induced and stimulus-independent correlated activity , within and across cell types ( Greschner et al . , 2011; Mastronarde , 1983 ) , which could substantially influence the encoding of the stimulus ( Meytlis et al . , 2012; Pillow et al . , 2008; Ruda et al . , 2020; Zylberberg et al . , 2016 ) . For these reasons , the visual message transmitted by a RGC to the brain is not fully understood . One way to understand how each RGC contributes to vision is to determine how a natural image can be reconstructed from the light-evoked responses of the entire RGC population . This analysis approach mimics the challenge faced by the brain: using sensory inputs to make inferences about the visual environment ( Bialek et al . , 1991; Rieke et al . , 1997 ) . In the simplest case of linear reconstruction , the visual message of an individual RGC can be summarized by its optimal reconstruction filter , that is its contribution to the reconstructed image . Linear reconstruction has been used to estimate the temporal structure of a spatially uniform stimulus from the responses of salamander RGCs , revealing that reconstruction filters varied widely and depended heavily on the other RGCs included in the reconstruction ( Warland et al . , 1997 ) . However , no spatial information was explored , and only a small number of RGCs of unknown types were examined . A later study linearly reconstructed spatiotemporal natural movies from the activity of neurons in the cat LGN ( Stanley et al . , 1999 ) . However , neurons from many recordings were pooled , without cell type identification or the systematic spatial organization expected from complete populations of multiple cell types . More recently , several studies have used nonlinear and machine learning methods for reconstruction ( Botella-Soler et al . , 2018; Parthasarathy et al . , 2017; Zhang et al . , 2020 ) , although these techniques were not tested in primate , or on large-scale data sets with clear cell type identifications and complete populations of RGCs . Thus , it remains unclear what spatial visual message primate RGCs convey to the brain , in the context of natural scenes and the full neural population . We performed linear reconstruction of flashed natural images from the responses of hundreds of RGCs in macaque retina , using large-scale , multi-electrode recordings . These recordings provided simultaneous access to the visual signals of nearly complete populations of ON and OFF parasol cells , as well as locally complete populations of ON and OFF midget cells , the four numerically dominant RGC types that provide high-resolution visual information to the brain ( Dacey et al . , 2003 ) . Data from 15 recordings produced strikingly similar reconstructions . Examination of reconstruction filters revealed that the visual message of a given RGC depended on the responses of other RGCs , due to the statistics of natural scenes . Reconstruction from complete cell type populations revealed that they conveyed different features of the visual scene , consistent with their distinct light response properties . The spatial information carried by one type was mostly unaffected by the contributions of other types , particularly types with the opposite response polarity ( ON vs . OFF ) . Two simple tests of nonlinear reconstruction revealed only minor improvements over linear reconstruction . Similar visual messages and reconstructions were obtained using linear-nonlinear cascade models of RGC light response incorporating measured spatial properties and response nonlinearity . Finally , full spatiotemporal reconstruction with dynamic scenes revealed similar spatial visual messages , suggesting that these findings may generalize to natural vision .
To understand how the visual message conveyed by a single RGC depends on the signals transmitted by others , reconstruction was performed from a given cell alone or with other cells of the same type . Cells of the same type exhibited similar response properties ( Chichilnisky and Kalmar , 2002 ) , with non-overlapping RFs forming a mosaic tiling visual space ( Figure 2 ) . When a single cell was used for reconstruction , its reconstruction filter ( Figure 3A , top ) was much wider than its spatial RF ( Figure 3A , bottom , measured with white noise; see Materials and methods ) , or the spatially localized filter obtained in the full population reconstruction described above ( Figure 1C ) . The full width at half maximum of the average single-cell reconstruction filter was roughly four times the average RF width ( 3 . 6 +/- 1 . 4 across 15 recordings ) . As additional RGCs of the same type were included in reconstruction , the spatial spread of the primary cell’s reconstruction filter was progressively reduced , leveling off to a value slightly higher than the average RF size when the six nearest neighbors were included ( 1 . 3 +/- 0 . 2 across 15 recordings; average filters shown in Figure 3C , widths shown in Figure 3D ) . Both the spatial spread of the single-cell reconstruction filter and its reduction in the context of the neural population can be understood by examining how the optimal filters ( Equation 1 ) depend on the statistics of the stimulus ( S ) and response ( R ) . The matrix RTR represents correlations in the activity of different RGCs . The matrix RTS represents unnormalized , spike-triggered average ( STA ) images , one for each RGC . These natural image STAs were broad ( Figure 3A , top ) , reflecting the strong spatial correlations present in natural scenes ( Figure 3B ) . For reconstruction from a single cell’s responses , RTR is a scalar , and therefore the single-cell reconstruction filter is directly proportional to the natural image STA . However , in the case of reconstruction from the population , RTR is a matrix that shapes the reconstruction filter based on the activity of other cells . Specifically , each cell’s filter is a linear mixture of its own natural image STA and those of the other cells in the population reconstruction , weighted negatively based on the magnitude of their correlated activity . This mixing resulted in the reduction in the width of the reconstruction filter of a given RGC when nearby cells of the same type were included ( Figure 3C ) . When the complete population of RGCs of the same type was included in the reconstruction , the resulting spatially localized filters were similar to the RFs obtained with white noise stimuli ( ρ=0 . 78 +/- 0 . 10 , n = 997 ON and 1228 OFF parasol cells from 15 recordings ) . However , some natural image spatial structure remained and was consistent across recordings , cells , and cell types . Most strikingly , the reconstruction filters exhibited broad vertical and horizontal structure ( Figure 3C , E ) . This is a known feature of natural scenes ( Girshick et al . , 2011 ) , and is present in the images used here ( Figure 3B ) . In addition , the visual scene was more uniformly covered by the reconstruction filters than by the RFs ( Figure 4A , C ) . Coverage was defined as the proportion of pixels that were within the extent of exactly one cell’s filter . The filter extent was defined by a threshold , set separately for the reconstruction filters and for the RFs to maximize the resulting coverage value . Across both the ON and OFF parasol cells in 12 recordings , the average coverage was 0 . 62 +/- 0 . 06 for the RFs and 0 . 78 +/- 0 . 03 for the reconstruction filters ( Figure 4C; p<0 . 001 ) . By comparison , expanded RFs , scaled around each RGC’s center location to match the average filter width , led to a small reduction in coverage ( 0 . 57 +/- 0 . 06; p<0 . 001 ) due to increased overlap . This indicates that the filters are not simply broader versions of the RF , but rather that they are distorted relative to the RFs to fill gaps in the mosaic . To understand how the differences between reconstruction filters and RFs affected the reconstructed images , reconstruction was performed using the spatial RFs in place of the filters ( each RF independently scaled to minimize MSE , see Materials and methods; Figure 4B ) . This manipulation reduced reconstruction performance by 24% ( Δρ=−0 . 12 +/- 0 . 09 across 4500 images from ON and OFF parasol cells in 15 recordings; p<0 . 001; Figure 4D ) , primarily in the lower spatial frequencies , which also contain most of the power in the original images ( Figure 4E ) . The resulting images were noticeably less smooth in appearance than the optimally reconstructed images , and exhibited structure resembling the RGC mosaic ( Figure 4B ) . Thus , although the reconstruction filters generally resembled the RFs , the additional spatial structure related to natural images and the spatial arrangement of RGCs led to smoother reconstructed images . These features may help explain the high consistency in reconstruction performance across many retinas ( see above; Figure 2 ) . The visual message transmitted by RGCs of a particular type could additionally be affected by the other cell types encoding the same region of visual space ( Warland et al . , 1997 ) . To test this possibility , reconstructions were performed using the responses of a single RGC alone ( the primary cell ) , or in combination with each of the four major cell type populations . For each combination , the reconstruction filters of the primary cells were averaged across all cells of the same type for each recording ( Figure 5A ) . Inclusion of all cells of any one cell type reduced the magnitude of the primary cell’s reconstruction filter ( Figure 5B , left ) . This can be understood by noting that the entries in ( RTR ) -1 , which mix the natural image STAs to produce the reconstruction filters , have the opposite sign of the response correlations . As expected , the correlations were positive for same-polarity cells and negative for opposite-polarity cells ( not shown; Greschner et al . , 2011; Mastronarde , 1983 ) . Therefore , the cell’s reconstruction filter was reduced in magnitude by positively weighted cells of the opposite polarity , and by negatively weighted cells of the same polarity . As discussed previously , for parasol cells , inclusion of the remaining cells of the same type substantially reduced the spatial extent of the primary cell’s filter ( Figure 3 ) . However , this did not occur when cells of other types were included in reconstruction instead ( Figure 5B , right , top two rows ) . Specifically , the inclusion of the midget cells with the same polarity only slightly reduced the spatial extent of the parasol cell’s filter , and inclusion of opposite polarity cells of either type had little effect . This is likely because the other cell types provide roughly uniform coverage , whereas the remaining cells of the same type have a gap in the location of the primary cell , resulting in significant shaping by the immediately neighboring cells . In summary , the spatial structure of the visual message of a single parasol cell is primarily influenced by neighboring cells of the same type and is largely unaffected by cells of other types . The filters for the midget cells were also shaped by the inclusion of the remaining cells of the same type ( Figure 5A , second column ) and were largely unaffected by the inclusion of opposite polarity cells of either type . However , unlike parasol cells , midget cell filters were significantly affected by the inclusion of the same-polarity parasol cells ( Figure 5A , third column ) . This is consistent with known correlations between these cell types ( Greschner et al . , 2011 ) , and the asymmetry may be due to the fact that parasol cells tended to have much stronger responses to the natural images than midget cells . Thus , the interpretation of the visual signal from a midget cell does depend somewhat on the signals sent by the same-polarity parasol cell population . The image features represented by each cell type were revealed by analysis of the reconstructed images . In particular , the separate contributions of ON and OFF cells , and of parasol and midget cells , were investigated . To estimate the contribution of ON and OFF cells , reconstruction was performed with ON or OFF parasol cells alone and in combination ( Figure 6A , B ) . Reconstructions using just OFF parasol cells were slightly more accurate than using just ON cells , but both were less accurate than reconstruction using the two types together ( Figure 6C , both: ρ=0 . 76 +/- 0 . 12 , ON: ρ=0 . 64 +/- 0 . 16 , OFF: ρ=0 . 67 +/- 0 . 14 , across n = 2250 images from 15 recordings; all p<0 . 001 ) . Reconstruction using just ON cells failed to accurately capture intensity variations in dark areas of the image , while reconstruction with just OFF cells failed to capture variations in light areas of the image ( for pixel values above the mean value: ρ=0 . 57 for ON and 0 . 26 for OFF , for pixel values below the mean value: ρ=0 . 31 for ON and 0 . 68 for OFF ) . Only a narrow middle range of pixel intensities were effectively encoded by both types ( Figure 6D ) . This is consistent with known output nonlinearities , which suppress responses to stimuli of the non-preferred contrast , and therefore limit linear reconstruction in that range . Thus , both ON and OFF cells were necessary to reconstruct the full range of image contrasts . Reconstruction using the responses of both cell types seemed to encode darker pixels more accurately than lighter pixels ( Figure 6D , bottom panel , black curve ) , consistent with the reconstruction performance from each type separately . This could reflect the fact that ON cells are less dense ( Chichilnisky and Kalmar , 2002 ) , and/or the fact that the natural image distribution is skewed towards darker pixel values ( Figure 6D , bottom panel , gray distribution ) , potentially placing greater weight on the accurate reconstruction of these values . In addition , ON cells exhibit a more linear contrast-response relationship ( Chichilnisky and Kalmar , 2002 ) , so there is less reconstruction performance difference between preferred and non-preferred contrasts . To estimate the contributions of parasol and midget cells , reconstruction was performed using parasol cells or midget cells or both ( Figure 7A , B ) . As expected , reconstruction using both parasol and midget cells was more accurate than using either alone ( Figure 7C , both: ρ=0 . 81 +/- 0 . 10 , parasol: ρ=0 . 77 +/- 0 . 12 , midget: ρ=0 . 73 +/- 0 . 13 , across n = 1050 images from seven recordings; all p<0 . 001 ) . Images reconstructed from midget cells contained more high-frequency spatial structure , consistent with their higher density ( Figure 7D ) . However , the images reconstructed from parasol cells had 50% higher signal-to-noise ( defined as standard deviation across images/standard deviation across repeats ) , resulting in the slightly higher reconstruction performance from parasol cells . The above analysis obscures the significantly different temporal responses properties of these two cell classes . In particular , parasol cells have more transient responses ( de Monasterio , 1978; De Monasterio and Gouras , 1975; Gouras , 1968 ) , which may allow them to convey information more rapidly than midget cells . To test this possibility , image reconstruction was performed using spikes collected over increasing windows of time after the image onset . The reconstruction performance of parasol cells increased quickly and reached 95% of peak reconstruction performance at 80 +/- 20 ms , while the performance of midget cells increased more slowly , and reached 95% performance at 116 +/- 19 ms ( across seven recordings; Figure 7E ) . This difference indicates that spatiotemporal reconstruction will be necessary to fully reveal the distinct contributions of these two classes ( see Discussion ) . The above results indicate that the visual message of each RGC , and the contributions of each cell type , are shaped by correlated activity . However , these analyses do not distinguish between stimulus-induced ( signal ) correlations , and stimulus-independent ( noise ) correlations that arise from neural circuitry within and across cell types in the primate retina ( Greschner et al . , 2011; Mastronarde , 1983 ) . To test the effect of noise correlations , reconstruction performance was evaluated on repeated presentations of test images . This performance was compared to a control condition in which the responses of each cell were independently shuffled across trials to remove noise correlations while preserving single-cell statistics and signal correlations . The reconstruction filters ( computed from unshuffled training data ) were then used to reconstruct the test images , using either the shuffled or unshuffled responses . In principle , shuffling could result in a net increase or decrease in reconstruction accuracy , due to two opposing factors . Because the reconstruction filters incorporate the correlated activity present in training data ( Equation 1 ) , any deviation from this correlation structure in the test data could reduce performance . On the other hand , if noise correlations produce spatial structure in the reconstructions that obscures the structure of the natural images , their removal could enhance reconstruction performance . The relative influence of these competing effects could also depend on the overall fidelity of the reconstruction . Accordingly , the shuffling manipulation was tested using three response measures . In the first , RGC responses were calculated by counting spikes in the 150 ms window after image onset , as above . In the second , the response was measured at the intrinsic time scale of correlations ( ~10 ms; DeVries , 1999; Mastronarde , 1983; Meister et al . , 1995; Shlens et al . , 2006 ) , by counting spikes in fifteen 10 ms bins , and reconstructing with this multivariate response vector instead of the scalar spike count . In the third , spikes were counted only in the 10 ms bin that had the highest average firing rate ( 50–60 ms after image onset ) . While the third approach did not utilize all the available information in the responses , it was used to mimic low-fidelity or rapid perception scenarios , which would have fewer stimulus-driven spikes available for reconstruction . Reconstruction using the first two response measures had similar unshuffled performance ( ρ=0 . 76 +/- 0 . 12 and 0 . 75 +/- 0 . 12 respectively ) , and low variation across trials ( standard deviation across repeats = 0 . 015 ) . With these measures , shuffling had a very small and detrimental effect on reconstruction ( across 3 recordings with 27 repeats of 150 test images: ( 1 ) Δρ=−0 . 0004 +/- 0 . 0017; |Δρ|=0 . 0012 +/- 0 . 0012; p<0 . 001 , ( 2 ) Δρ=−0 . 0008 +/- 0 . 0019; |Δρ|=0 . 0014 +/- 0 . 0015; p<0 . 001 ) . In each case , the magnitude of the change in correlation represented about 10% of the variation in reconstruction accuracy across trials , which represents roughly how much improvement could be expected ( Figure 8 ) . For comparison , shuffling the responses in each time bin independently across trials ( rather than the responses of each cell independently ) had a much larger effect ( Δρ=−0 . 02 +/- 0 . 01 ) , consistent with previous results ( Botella-Soler et al . , 2018 ) , indicating that the autocorrelation structure across time is more important for reconstruction than the noise correlation structure across cells . Thus , in these conditions , noise correlations had a limited impact on reconstruction , regardless of the time scale of analysis . Reconstruction using the third measure had lower unshuffled performance ( ρ=0 . 64 +/- 0 . 14 ) , and higher variation across trials ( standard deviation across repeats = 0 . 039 ) . In this case , shuffling led to a more consistent , but still small , increase in reconstruction performance ( Δρ=0 . 0071 +/- 0 . 0076; |Δρ|=0 . 0075 +/- 0 . 0072; p<0 . 001 ) . The increase represented a larger fraction of the variation in reconstruction accuracy across trials ( 20%; Figure 8 ) . This suggests that in low-fidelity , high-noise situations , noise correlations in the RGC population can partially obscure the structure of natural images , even if reconstruction is designed to take the correlations into account . Linear reconstruction provides an easily interpretable estimate of the visual message , but it may limit the quality of reconstruction by not extracting all the information available in the neural responses and may also differ greatly from how the brain processes the retinal input . Therefore , two simple extensions of linear reconstruction were tested: transformation of the responses using a scalar nonlinearity , and inclusion of interaction terms between nearby cells . In the first case , the response of each cell was transformed using a scalar nonlinearity , and linear regression ( Equation 1 ) was performed to reconstruct images from the transformed response . The stimulus estimate SNL is given by SNL = f ( R ) ⋅WNL , where WNL is a matrix of reconstruction weights ( refitted using the transformed responses ) , and f ( R ) is the scalar nonlinear transform of the population response vector R . This is equivalent to inverting a linear-nonlinear ( LN ) encoding model of the form R = g ( K⋅S ) , where g is the inverse of f , and K is a different set of weights ( note that in general a nonlinear encoder may not require an equivalent nonlinear decoder for optimum performance; see Rieke et al . , 1997 for a full discussion ) . A common form of the LN encoding model uses an exponential nonlinearity , g = exp ( ) ; therefore , the inverse function f = log ( ) was used for reconstruction , and the response for each cell was defined as the spike count plus 1 . A square root transformation was also tested , and yielded similar results ( not shown ) . The relationship to pixel values was more linear for the transformed RGC responses than for the original responses ( Δlinear fit RMSE = −1 . 9 +/- 1 . 5 across n = 2225 cells from 15 recordings; Figure 9A , B ) , indicating that this inverse function captured at least some of the nonlinearity in retinal signals . The nonlinear transformation slightly increased reconstruction accuracy when using the responses of ON or OFF parasol cells alone ( across 15 recordings with 300 images each: ON parasol: Δρ=0 . 013 +/- 0 . 051 , p<0 . 001; OFF parasol: Δρ=0 . 015 +/- 0 . 035 , p<0 . 001; Figure 9C ) . However , it did not help when using the responses of ON and OFF parasol cells together ( Δρ=−0 . 0017 +/- 0 . 032 , p=0 . 001; Figure 9C ) . This likely reflects the fact that the relationship between the true pixel values and the pixel values reconstructed using the original , untransformed responses was already approximately linear when using both cell types , but not when using just one cell type ( Figure 6 ) . In addition , using the raw responses of both cell types was more effective than using the transformed responses of either type alone ( ON parasol: Δρ=−0 . 09 +/- 0 . 1 , p<0 . 001; OFF parasol: Δρ=−0 . 06 +/- 0 . 1 , p<0 . 001 ) , suggesting that intensity information cannot be directly recovered fully from either ON or OFF cells alone . Nonlinear interactions between the signals from different cells could also potentially increase reconstruction performance . To test this idea , the products of spike counts in pairs of neighboring cells were added as predictors in the linear reconstruction . Neighbors were defined as cells with RF centers that were within 1 . 5 times the median nearest neighbor distance between RF centers of the cells of the same type . For parasol cells , this definition resulted in roughly 6 ON and 6 OFF neighbors per cell , as expected ( see Figure 2 ) . Including these interactions produced a small increase in reconstruction accuracy ( Δρ=0 . 0093 +/- 0 . 023 , across three recordings with 300 test images each; p<0 . 001; regularization did not lead to improved performance ) . The primary contribution was from ON-OFF pairs ( ON-OFF: Δρ = 0 . 0089 +/- 0 . 019 , not significantly different than all pairs , p=0 . 2; ON-ON: Δρ=0 . 0021 +/- 0 . 010 and OFF-OFF: Δρ=0 . 0024 +/- 0 . 013 , both significantly different than all pairs , p<0 . 001; Figure 9D ) . The reconstruction filters associated with these interaction terms typically had an oriented structure orthogonal to the line between the RF centers of the two cells ( Figure 9F , G ) , suggesting that the improvement in reconstruction may come primarily from using the joint activation of partially overlapping ON and OFF cells to capture edges in the visual scene . The above analyses revealed that noise correlations and interactions between cells and cell types had a limited impact on reconstruction performance , suggesting that more complicated features of retinal encoding may not be important for linear reconstruction . To further explore this idea , simple LN models were used to simulate RGC responses across all 15 recordings , and the primary features of reconstructions from recorded and simulated spike trains were compared . The simulated spike count of each RGC in response to a given image was calculated by filtering the image with the spatial RF , and then passing that value through a fitted sigmoidal nonlinearity to obtain a firing rate ( see Materials and methods ) . The noise in the recorded spike counts was sub-Poisson ( not shown; see Uzzell and Chichilnisky , 2004 ) ; therefore , the simulated firing rate was directly compared to the trial-averaged , recorded firing rate . This model captured RGC responses to static images with reasonable accuracy ( correlation between simulated and average recorded spike counts: 0 . 76 +/- 0 . 13 across n = 997 ON parasol cells; 0 . 84 +/- 0 . 09 across n = 1228 OFF parasol cells; see Chichilnisky , 2001 ) . Note that by definition , the model incorporated the measured functional organization of the retina , including retina-specific RF mosaic structure and cell-type specific response properties , both of which are necessary to understand the visual message ( see above ) . Reconstructions with recorded and simulated spike trains revealed broadly similar properties in the filters and reconstructed images . The filters fitted to the recorded and simulated spike trains were similar ( ρ=0 . 84 +/- 0 . 09 across 2225 parasol cells from 15 recordings ) , and shared key features , such as horizontal and vertical structure ( Figure 10A , C ) . The reconstructed images themselves were also similar ( correlation between images reconstructed from simulated and recorded spike counts: 0 . 93 +/- 0 . 04 across n = 2250 images from 15 recordings; Figure 10B , C ) , as was the reconstruction performance ( simulated: ρ=0 . 79 +/- 0 . 11; recorded: ρ=0 . 78 +/- 0 . 11; Δρ=−0 . 003 +/- 0 . 03; across 2250 images from 15 recordings; Figure 10C ) . The simulated spike trains also replicated the structure of nonlinear interactions between cells . This was observed by using the simulated responses of ON and OFF cells and the products of the responses of neighboring cells , as above , to reconstruct natural images . The spatial reconstruction filter corresponding to the interaction term between nearby ON and OFF cells was oriented and qualitatively similar to the interaction filters obtained with real data ( Figure 9E , F ) . However , this was not the case for responses simulated using a linear model without any response rectification ( not shown ) – in this case , the filter corresponding to the interaction term had no clear structure . The model reveals that although the visual messages of RGCs depend on their spatial and cell-type specific organization , as well as the statistics of the stimulus , their essential structure can be understood using simple models of RGC encoding . Furthermore , some degree of nonlinear encoding is necessary to explain the oriented interaction filters observed in the data . In natural vision , a continuous stream of retinal responses is used to make inferences about the dynamic external world . Therefore , the reconstruction approach above – using the accumulated spikes over a fraction of a second to reconstruct a flashed image – could fail to capture important aspects of normal vision . To test whether the above results extend to spatiotemporal reconstructions , a naturalistic movie , consisting of a continuous stream of natural images with simulated eye movements superimposed , was reconstructed from the spike trains of RGCs . The spike trains were binned at the frame rate of the movie ( 120 Hz ) , and linear regression was performed between the frames of the movie and the RGC responses in 15 bins following each frame , resulting in a spatiotemporal reconstruction filter for each RGC . A spatial summary of the filter for each cell was obtained by first calculating the average time course of the strongest pixels , and then projecting each pixel of the full filter against this time course ( examples shown in Figure 11A; see Materials and methods ) . This spatial filter was highly correlated with the spatial reconstruction filters of the same cells obtained in the preceding analysis with flashed images ( ρ=0 . 87 +/- 0 . 07 , n = 351 parasol cells from three recordings; Figure 11B ) . The dynamic filters were approximately space-time separable ( explained variance from first principal component = 0 . 85 +/- 0 . 13 ) . The remaining unexplained variance contained significant apparent structure as well as noise ( not shown ) , which may be important for further understanding spatiotemporal processing in the retina and the underlying mechanisms , but was not explored further ( Benardete and Kaplan , 1997a; Benardete and Kaplan , 1997b; Dawis et al . , 1984; Derrington and Lennie , 1982; Enroth-Cugell et al . , 1983 ) . The large fraction of variance explained by a space-time separable filter suggests that the essential spatial features of the visual message observed in spatial reconstructions largely extend to spatiotemporal vision . In addition , the reconstructed movie frames were similar to reconstructions of static images ( between static reconstruction and average reconstructed frame: ρ=0 . 72 +/- 0 . 19 across 120 images from three recordings , Figure 11C ) .
Linear reconstruction of natural images was used to investigate the spatial information transmitted to the brain by complete populations of primate RGCs . The quality of the reconstructions was consistent across retinas . The optimal interpretation of the spikes produced by a RGC – that is its visual message – depended not only on its encoding properties , but also on the statistics of natural scenes and the spatial arrangement of other RGCs . These factors enabled smoother natural image reconstructions from the RGC population than would be expected from the RFs alone . In addition , the visual representation conveyed by each cell type reflected its distinct encoding properties , and for ON and OFF parasol cells , was largely independent of the contributions of other cell types . Overall , the results were consistent with a simple , linear-nonlinear model of RGC encoding , incorporating the spatial properties , contrast-response properties , and collective functional organization of the four major RGC types . Finally , a limited test of spatiotemporal reconstruction indicated that these results may generalize to natural vision . The results show that the dependence of a given RGC’s visual message on the responses of other RGCs , which was demonstrated previously in the temporal domain using a spatially uniform random flicker stimulus ( Warland et al . , 1997 ) , extends to the spatial domain in natural viewing conditions . For decades , the spatial visual message of a RGC has been estimated using its RF , measured with artificial stimuli . However , due to spatial correlations in natural scenes , the response of a RGC contains information about the stimulus far beyond its RF . In this light , it is at first surprising that the visual message is spatially localized and similar to the classical RF ( Figure 3A , C ) . However , nearby regions of visual space are already ‘covered’ by the neighboring RGCs of the same type , and the redundant information in adjacent cells apparently contributes little to representing the image structure . Even so , the visual messages retain some explicit horizontal and vertical natural scene structure , and collective spatial organization , not present in the RFs . This structure results in smoother reconstructions and more uniform coverage of visual space than the coverage provided by the RF mosaic ( Figure 4 ) . In this sense , the visual message of each RGC differs from its RF , specifically in a way that reflects its coordination with other nearby cells . The significance of natural scene statistics for interpreting the neural code has also been suggested in the visual cortex ( Naselaris et al . , 2009 ) , and can be used as a prior to improve image estimates in multi-step reconstruction methods ( Parthasarathy et al . , 2017 ) . Each of the major RGC types conveyed distinct visual representations , consistent with their encoding properties . For the most part , these were independent of the contributions of the other types , indicating that the major primate RGC types , despite covering the same region of visual space , conveyed different stimulus features . However , this separation was clearer for the ON and OFF types than for the parasol and midget cell classes , because the midget cell filters were influenced by the inclusion of same-polarity parasol cells . Further analysis in the temporal domain ( see Figure 7E ) may be necessary to clarify the separation of these two classes . Both ON and OFF cell types were necessary to reconstruct the full contrast range of the images , because responses from a single cell type resulted in less accurate reconstructions even if they were linearized . It is not clear why the retina separates visual information into separate cell type channels . The roughly linear intensity representation by ON and OFF cell types together ( but not individually ) is consistent with suggestions that encoding by multiple cell types with nonlinear response properties could enable relatively simple linear reconstruction by downstream neurons ( DiCarlo et al . , 2012; Gjorgjieva et al . , 2019 ) . There also may be more complicated interactions between different cell types that another reconstruction method could reveal . As new cell types are identified and characterized ( Puller et al . , 2015; Rhoades et al . , 2019 ) , their contributions to vision may be more fully revealed by these linear and simple nonlinear reconstruction approaches . Overall , the results presented here were consistent with predictions from a simple , independent pseudo-linear model for RGC light responses , despite known nonlinearities and correlations in the retinal circuitry . Specifically , replacing the recorded spike trains with simulated spike trains , generated by LN models fitted to each RGC , resulted in similar reconstruction filters and reconstructed images ( Figure 10 ) . Obviously , the LN model by itself cannot explain the many features of encoding observed here; instead , the specific spatial properties , contrast-response properties , and collective organization of the major RGC types captured in the present measurements are crucial for understanding the structure of the visual message . The similarity of reconstruction from LN models and recorded data is consistent with the limited impact of interaction terms and stimulus-independent ( noise ) correlations , the importance of which has been debated ( Cafaro and Rieke , 2010; Ganmor et al . , 2015; Meytlis et al . , 2012; Nirenberg et al . , 2001; Pillow et al . , 2008; Puchalla et al . , 2005; Ruda et al . , 2020; Zylberberg et al . , 2016 ) . While the impact of noise correlations on reconstruction in the present data was limited by the low total noise in the accumulated spike counts , this may not reflect natural vision , in which perception and action occur too quickly to utilize all the stimulus-driven spikes from each RGC , and sometimes must rely on visual inputs with low light levels or spatial contrast ( Ruda et al . , 2020 ) . A low-fidelity situation was mimicked by reducing the spike integration time window to 10 ms , a manipulation that revealed an increased but still small effect of noise correlations . It is also possible that these results would be affected by removing noise correlations from both the training and testing data , but evaluating this possibility would require longer repeated presentations of training stimuli than were performed here . It is uncertain how close the reconstructions presented here are to the best possible reconstructions given the data , and how much additional information could potentially be extracted from the spike trains . Acuity has been shown to track with midget cell RF size ( Dacey , 1993; Merigan and Katz , 1990; Rossi and Roorda , 2010; Thibos et al . , 1987 ) , indicating that the reconstructions shown in Figure 7 may accurately represent the quality of visual information transmitted to the brain . In addition , it has been suggested that simple decoders may be sufficient , even when the encoding is highly nonlinear ( DiCarlo et al . , 2012; Gjorgjieva et al . , 2019; Naselaris et al . , 2011; Rieke et al . , 1997 ) . However , alternative approaches may be worth exploring , and could extract additional information . For example , different measures of response , such as latency ( Gollisch and Meister , 2008; Gütig et al . , 2013 ) and relative activity ( Portelli et al . , 2016 ) , have been shown to convey more stimulus information than spike counts for non-primates under some conditions . This was not the case in the present data , which may be due to high-maintained firing rates in the mammalian retina ( Troy and Lee , 1994; see Figure 1B ) , which make it difficult to identify the first stimulus-driven spike . In addition , recent studies have indicated that nonlinear and deep learning models could improve reconstruction performance for static images , moving patterns , and naturalistic movies ( Botella-Soler et al . , 2018; Kim et al . , 2020; Parthasarathy et al . , 2017; Zhang et al . , 2020 ) . While these approaches make the visual message more difficult to define , they could be used to extract richer information potentially present in RGC responses . Models that are interpretable while allowing for some nonlinearities could also be used to further investigate the visual message ( Pillow et al . , 2008 ) . Attempting to extract more sophisticated visual information may also reveal additional information conveyed by RGCs , for example , by expanding to more complex , dynamic natural stimuli . Spatiotemporal stimuli , which were only explored here in a limited way , and/or chromatic stimuli , could further illuminate the impact of spike timing , the encoding of dynamic and space-time inseparable features , and the distinct roles of the multiple cell types ( Benardete and Kaplan , 1997a; Benardete and Kaplan , 1997b; Berry et al . , 1997; Dacey et al . , 2003; Dawis et al . , 1984; Derrington and Lennie , 1982; Enroth-Cugell et al . , 1983; Masland , 2012; Uzzell and Chichilnisky , 2004 ) . For example , nonlinear spatial summation and motion encoding have been demonstrated in parasol cells but were not utilized here ( Manookin et al . , 2018; Turner and Rieke , 2016 ) . In addition , pixel-wise mean squared error does not accurately reflect the perceived quality of the visual representation . More sophisticated metrics for optimization and evaluation of reconstruction should be explored ( Wang and Lu , 2002; Wang et al . , 2004 ) . By projecting neural responses into a common stimulus space , reconstruction enabled direct comparison and evaluation of the visual signals transmitted downstream . The large collection of recordings used here revealed a consistent visual representation across retinas , in spite of differences in RF mosaic structure and firing rates that make comparing the neural response itself difficult . The information contained in the retinal signal limits the information available to downstream visual areas , so the results presented here could inform studies of visual processing in the LGN , V1 , and other brain structures . For example , the oriented nature of the interaction term filters supports the hypothesis that orientation selectivity in the cortex results from pairs of nearby ON and OFF RGCs ( Paik and Ringach , 2011; Ringach , 2007 ) . In addition , comparing reconstructions from different visual areas using a standard measurement — the reconstructed image — could help reveal how information about the external world is represented at various stages of the visual system . Using reconstruction to understand the signals transmitted by neurons may be increasingly important in future efforts to read and write neural codes using brain-machine interfaces ( BMIs ) . In the retina , certain types of blindness can be treated with implants that use electrical stimulation to activate the remaining retinal neurons ( Goetz and Palanker , 2016 ) . The visual messages described in the present work could be useful for inferring the perceived visual image evoked by such devices , and thus for selecting optimal electrical stimulation patterns ( Goetz and Palanker , 2016; Golden et al . , 2019; Shah et al . , 2019 ) . Reconstruction can also be used to compare the evoked visual representation with the representation produced by natural neural activity . In addition , the observation that reconstructions from different retinas and from recorded and simulated spikes are similar suggests that perfect replication of the neural code of a particular retina may not be necessary . Outside the visual system , many BMIs rely on reconstruction to read out and interpret neural activity , for example controlling prosthetic limbs using activity recorded in the motor cortex ( Lawhern et al . , 2010; Vargas-Irwin et al . , 2010 ) . While these studies typically focus on performing specific tasks , the present results suggest that examination of the reconstruction filters could reveal contributions of diverse cells and cell types in these modalities .
Noise correlation analysis ( Figure 8 ) was limited to the three recordings with the most repeated presentations of the same set of test images ( 27 repeats each ) . For each of the three scenarios described in Results , reconstruction filters were fitted on a single repeat of training data , and then tested using either shuffled or unshuffled testing data . The testing data was shuffled by randomly permuting each RGC’s responses independently across repeated presentations of the same image . Reconstruction performance on the test data was measured as described earlier . Only the three recordings with the most training data were included ( at least 25 , 000 training images each; the same subset was used for the noise correlation analysis ) , so that despite the increase in parameter count ( from ~200 to ~1000 ) , there were still more than enough samples to calculate the weights , and regularization did not improve cross-validated reconstruction performance . Simple linear-nonlinear encoding models ( Chichilnisky , 2001 ) were used to simulate spike trains for reconstruction , for each RGC independently . For each image , the inner product was first computed between the image and the spatial RF ( see section Spatial receptive field above ) , restricted to a local region ( +/- 440 μm from the RF center , corresponding to either 40 × 40 or 80 × 80 pixels depending on the resolution of the images ) . The resulting value was then passed through a sigmoidal nonlinearity , given by ( 3 ) y= b4+ b1b2+exp ( b3∙x1 ) where the parameters {bi} were fitted by minimizing the mean-squared error between the predicted and measured RGC responses , on the same data set used to fit the reconstruction filters . This model was then used to simulate responses to the images used to obtain the fitting data and the images used to obtain the held-out , repeated test data . Reconstruction filters , reconstructed images , and performance were then calculated from the simulated responses in the same way as described above for the recorded responses . Each frame of the spatiotemporal movie was reconstructed using the RGC spikes recorded during that frame and the following frames . Therefore , each RGC included in the reconstruction was fitted with a full-rank , spatiotemporal reconstruction filter . The spikes were binned at the frame rate of the movie , and a filter length of 15 frames ( 125 ms ) was selected to optimize performance . A spatial summary of the spatiotemporal filter ( Figure 11A , B ) was calculated as described above for spatial RFs . The spacetime separability of the filters was calculated using the explained variance from the first component of a singular value decomposition ( limited to a spatially local region to reduce the effects of the many low-magnitude , noisy pixels outside the primary filter peak ) . Three recordings that contained responses to both static , flashed natural images and dynamic , spatiotemporal natural movies were included . 2400 consecutive movie frames were withheld from fitting for comparison of movie frame and static image reconstructions ( Figure 11C ) . | Vision begins in the retina , the layer of tissue that lines the back of the eye . Light-sensitive cells called rods and cones absorb incoming light and convert it into electrical signals . They pass these signals to neurons called retinal ganglion cells ( RGCs ) , which convert them into electrical signals called spikes . Spikes from RGCs then travel along the optic nerve to the brain . They are the only source of visual information that the brain receives . From this information , the brain constructs our entire visual world . The primate retina contains roughly 20 types of RGCs . Each encodes a different visual feature , such as the presence of bright spots of a certain size , or information about texture and movement . But exactly what input each RGC sends to the brain , and how the brain uses this information , is unclear . Brackbill et al . set out to answer these questions by measuring and analyzing the electrical activity in isolated retinas from macaque monkeys . Studying the macaque retina was important because the primate visual system differs from that of other species in several ways . These include the numbers and types of RGCs present in the retina . These primates are also similar to humans in their high-resolution central vision and trichromatic color vision . Using electrode arrays to monitor hundreds of RGCs at the same time , Brackbill et al . recorded the responses of macaque retinas to real-life images of landscapes , objects , animals or people . Based on these recordings , plus existing knowledge about RGC responses , Brackbill et al . then attempted to reconstruct the original images using just the electrical activity recorded . The resulting reconstructions were similar across all retinas tested . Moreover , they showed a striking resemblance to the original images . These results made it possible to comprehend how the light-response properties of each cell represent visual information that can be used by the brain . Understanding how macaque retinas work in natural conditions is critical to decoding how our own retinas process and convey information . A better knowledge of how the brain uses this input to generate images could ultimately make it possible to design artificial retinas to restore vision in patients with certain forms of blindness . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Reconstruction of natural images from responses of primate retinal ganglion cells |
Motion anticipation allows the visual system to compensate for the slow speed of phototransduction so that a moving object can be accurately located . This correction is already present in the signal that ganglion cells send from the retina but the biophysical mechanisms underlying this computation are not known . Here we demonstrate that motion anticipation is computed autonomously within the dendritic tree of each ganglion cell and relies on feedforward inhibition . The passive and non-linear interaction of excitatory and inhibitory synapses enables the somatic voltage to encode the actual position of a moving object instead of its delayed representation . General rather than specific features of the retinal connectome govern this computation: an excess of inhibitory inputs over excitatory , with both being randomly distributed , allows tracking of all directions of motion , while the average distance between inputs determines the object velocities that can be compensated for .
The brain estimates the location of an object in visual space by reading out which retinal ganglion cells ( RGCs ) respond to it . This ‘retinotopic map’ is preserved through the visual pathway and is used to guide behaviour ( Hubel and Wiesel , 1962; Malpeli and Baker , 1975; Bonin et al . , 2011 ) . There is , however , a problem to overcome when this map is used to estimate the position of a moving object: the slow speed with which photoreceptors convert light into an electrical signal causes RGCs to respond ∼70 ms after an object first appears ( Baylor and Hodgkin , 1974 ) . During this delay , a tennis ball served by a professional player will have travelled ∼3–4 m . The fact that the position of the ball can be estimated precisely enough to meet it with a racquet demonstrates that the visual system is able to overcome the phototransduction delay for moving stimuli . Together , these computations are termed motion anticipation , and they begin in the inner retina ( Berry et al . , 1999 ) : when a stimulus is moving , the peak-firing rate of retinal ganglion cells ( RGCs ) occurs earlier than expected from the delayed response to a flashed stimulus . This correction supports accurate target tracking in salamander ( Leonardo and Meister , 2013 ) . Although the retina is capable of processing the visual input in a variety of ways , we still do not understand how most of these transformations are achieved ( Olveczky et al . , 2003; Münch et al . , 2009; Gollisch and Meister , 2010; Bölinger and Gollisch , 2012 ) . One aspect of motion processing that is understood in detail is the generation of directionally selective responses peculiar to a small subset of RGCs ( Borst and Euler , 2011 ) . Analysis of neuronal connectivity using large-scale electron microscopy has demonstrated that this computation involves asymmetric connections with a specific type of inhibitory interneuron , the starburst amacrine cell ( Briggman et al . , 2011 ) , which has dendrites that are themselves directionally selective ( Hausselt et al . , 2007; Yonehara et al . , 2013 ) . Motion anticipation appears to be a more fundamental retinal computation than directional selectivity because it is observed in the large majority of ganglion cells , of different functional types and across many species ( Berry et al . , 1999; Schwartz et al . , 2007 ) . We do not understand how the retina generates motion anticipation , but the ubiquity of this process across ganglion cell types suggests that it reflects general properties of the inner retinal circuit rather than the specific wiring of subtypes of neuron . A fast decrease in the gain with which signals are transmitted through the retina has been proposed to account for motion anticipation , but the site ( s ) of such control have not been identified ( Berry et al . , 1999; Schwartz et al . , 2007; Chen et al . , 2013 ) . One possibility is that moving stimuli induce a fast decrease in the efficiency of excitatory transmission from bipolar cells to ganglion cells , shortening the time-course of excitation . This idea is attractive because bipolar cells are the only route by which excitatory signals are transmitted to ganglion cells , and these synapses have been identified as a major site of gain-control , with an increase in temporal contrast causing fast depression of vesicle release ( Demb , 2008; Jarsky et al . , 2011; Nikolaev et al . , 2013 ) . A second possibility is that some mechanism intrinsic to ganglion cells alters the time-course of the ganglion cell response ( Chen et al . , 2013 ) . We now need to test these various possibilities experimentally . We have used electrophysiology and modelling to investigate the mechanisms of motion anticipation and find that it is not usually exerted through changes in the excitatory inputs to RGCs , but rather depends on the shunting effect of inhibition that RGCs receive from amacrine cells . The non-linear interaction between excitatory and inhibitory synapses is a result of the passive properties of the dendritic tree and remains intact when active conductances are blocked . Motion anticipation operates across most ganglion cells because it depends on general rather than specific features of the retinal connectome: independent and random distributions of excitatory and inhibitory synapses on RGCs ( Freed and Sterling , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Jakobs et al . , 2008; Xu et al . , 2008; Schwartz et al . , 2012 ) ; an excess of inhibitory inputs ( West , 1976; Koontz and Hendrickson , 1987; Freed and Sterling , 1988; Marshak et al . , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Haverkamp et al . , 1997; Zhu and Gibbins , 1997; Owczarzak and Pourcho , 1999; Marshak et al . , 2002; Jakobs et al . , 2008; Koizumi et al . , 2011 ) , and an average excitatory synaptic spacing along dendrites of ∼5 µm ( Freed and Sterling , 1988; Jakobs et al . , 2008; Xu et al . , 2008; Koizumi et al . , 2011 ) . The excess of inhibitory inputs allows tracking of all directions of motion , while the average distance between excitatory inputs determines the object velocities that can be compensated for . This study demonstrates how a computation fundamental to the retinal circuit can be understood in terms of general properties of the connectome and the cable properties of dendrites .
The basic phenomenon of motion anticipation is demonstrated in Figure 1 , where we recorded extracellular spikes from individual RGCs in retinal flat-mounts from goldfish . The receptive field ( RF ) of each ganglion cell was mapped using bars flashed in a random order across the retina ( Johnston et al . , 2014 ) , and then a bar was flashed onto the centre of the receptive field: the delay to peak firing averaged 62 ± 3 ms , largely reflecting the delay in phototransduction ( Baylor and Hodgkin , 1974 ) ( n = 25; Figure 1A , C ) . But when the same bar ( −100% contrast; 160 µm wide , equivalent to 2 . 4° of visual angle ) was moved across the retina at 500 μm s−1 ( 7 . 5° s−1 ) , the peak spike rate occurred 46 ± 13 ms ( n = 25 ) before the leading edge reached the receptive field centre and then activity decayed before the bar left ( Figure 1B , C ) . The time at which a RGC responded most strongly therefore encoded the anticipated position of the bar in retinotopic space ( Berry et al . , 1999 ) ( Figure 1B , C ) . Motion anticipation was observed across many functional types of ganglion cell , including brisk-transient , brisk-sustained and orientation selective cells . RGCs with larger RFs tended to display greater anticipation , with RF size accounting for 35% of the observed variance in the delays for motion ( Pearson's r = −0 . 593 , n = 25 , Figure 1D ) . The three cells that failed to show any motion anticipation also had the smallest RF size . 10 . 7554/eLife . 06250 . 003Figure 1 . Motion anticipation in the retina is not due to a gain change in bipolar cells . ( A ) An example of two ganglion cells responding to a bar flashed on their receptive field centres for 100 ms ( −100% contrast ) . ( B ) The response of the same cells to a bar of width 160 µm ( 2 . 4º ) moving at 500 µm s−1 ( 7 . 5º s−1 ) . The position of the bar relative to a cell's receptive field is shown above for the different time points indicated by the lettered arrows . ( C ) A comparison of the delay for the maximal response to a flashed stimulus ( 62 ± 2 . 6 ms ) and the time of maximal spiking to a moving stimulus relative to the time at which the stimulus reached the centre of the RF ( −46 ± 12 . 6 ms; n = 25 ganglion cells; p < 0 . 0001 ) . BT = Brisk-transient , BS = Brisk Sustained , OS = Orientation-selective , see ‘Materials and methods’ for cell classification . ( D ) The degree of motion anticipation was correlated with the RF size ( Pearson's r = −0 . 593 , n = 25 ) . ( E ) Schematic of retinal feedback circuits in the inner plexiform layer ( IPL ) , excitation and inhibition are represented by green and red arrows respectively . ( F ) Schematic of feed-forward inhibition in the IPL . ( G ) The dynamics of the EPSC evoked by a −100% contrast bar flashed over the RGCs RF centre for 320 ms . Individual cells were normalised before averaging ( n = 12 , SEM in grey ) . ( H ) Example of the EPSC recorded as a bar moves across the receptive field of an OFF ganglion cell ( average of six presentations ) . The peak EPSC lags behind the receptive field centre by 79 ± 17 μm ( 1 . 2 ± 0 . 3º , n = 7 ) . The purple line indicates the expected linear response obtained from convolution of the receptive field with the EPSC in F . The motion evoked EPSC was not significantly different from the expected linear response , indicating that lateral inhibition is not present ( using the Kolmogorov–Smirnov test , n = 7 ) . ( I ) Orientation selective cells did show a clear indication of lateral inhibition ( n = 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 003 The timing of the peak spike response in RGCs might be brought forward if a moving stimulus caused excitation in the ganglion cell to be truncated soon after it began ( Berry et al . , 1999 ) . Such a rapid decrease in the gain of the excitatory input might be caused by ( i ) depression intrinsic to the bipolar cell synapse , as occurs during contrast adaptation ( Rieke , 2001; Demb , 2008; Nikolaev et al . , 2013 ) and/or ( ii ) feedback inhibition , either reciprocal or lateral , that bipolar cell terminals receive from amacrine cells ( Roska et al . , 2000; Tanaka and Tachibana , 2013 ) ( Figure 1E ) . Alternatively , truncation of the spike response might reflect feedforward inhibition from amacrine cells onto RGCs ( Figure 1F ) . To differentiate between these possibilities , we began by isolating the excitatory postsynaptic current in ganglion cells and asking whether the excitatory input generated by a moving stimulus displayed any degree of motion anticipation . Our standard moving stimulus , a bar 2 . 4° wide moving at 7 . 5° s−1 , spends 320 ms at any one point on the retina . When this bar was presented statically for 320 ms over the RF centre , the bipolar cell input decreased rapidly after a short delay ( Figure 1G , n = 12 ) . This decay reflects a combination of two mechanisms controlling the output from bipolar cells: feedback inhibition from amacrine cells ( Roska et al . , 2000; Tanaka and Tachibana , 2013 ) and depression intrinsic to the synaptic terminal , which reflects depletion of vesicles in a state ready for rapid fusion ( Rieke , 2001; Demb , 2008; Nikolaev et al . , 2013 ) . To test whether these presynaptic mechanisms of gain control could generate motion anticipation in RGCs we measured the time-course of the EPSC in response to the moving bar . In 7 out of 9 RGCs the peak excitatory input was delayed , occurring 158 ± 34 ms after the bar reached the receptive field centre ( Figure 1H , black trace ) , demonstrating that the fast gain reduction in the EPSC was not sufficient to generate motion anticipation . Further , the time-course of the EPSC induced by motion was not significantly different from the linear response predicted by convolving the dynamics of the static EPSC measured in Figure 1F with each ganglion cell's RF ( Kolmogorov–Smirnov test , n = 7 ) , indicating that there was no correction at all for the lag in phototransduction . These results rule out events at the bipolar cell terminal as a mechanism of motion anticipation , be they intrinsic depression , feedback inhibition or lateral inhibition ( Figure 1E ) . Two RGCs that we recorded from provided an interesting exception to this pattern: the motion-induced EPSC was significantly truncated over the latter half of the receptive field compared to the linear prediction ( Figure 1I ) . Notably , these were the only two RGCs out of 25 that we sampled to show orientation-selective responses . However , the large majority of RGCs in goldfish and other species are not orientation-selective ( Levick , 1967 ) , indicating that most signals transmitted by bipolar cells are not corrected for the lag in phototransduction . The result in Figure 1 immediately suggest that the inhibitory input which RGCs receive directly from amacrine cells ( Lukasiewicz and Werblin , 1990; Roska et al . , 2000; Masland , 2012; Cafaro and Rieke , 2013 ) may be responsible for motion anticipation . To test the role of feedforward inhibition we measured the motion response of individual RGCs before and after the selective disruption of their inhibitory inputs . Figure 2A shows a RGC's spiking in response to the standard moving stimulus , recorded in cell-attached mode . As usual , firing occurred just before the leading edge reached the receptive field centre ( average of 18 ± 10 ms , n = 6 ) . We then disrupted inhibition in this single ganglion cell by going into whole-cell mode and dialyzing the cell with an intracellular solution containing 120 mM Cl− ( Figure 2A , grey ) . The advantage of this approach over pharmacology or genetic manipulation is that it acutely disrupts the inhibition impinging on a single RGC while leaving excitatory inputs and the rest of the retinal circuitry intact; a similar approach was used to demonstrate direction selectivity occurs postsynaptically in direction-selective RGCs ( Taylor et al . , 2000 ) . Disrupting inhibitory inputs greatly enhanced the response to a moving bar , indicating that under normal circumstances inactivation of Nav channel does not attenuate the RGC response to motion . Importantly , with inhibition disrupted , the location of the peak firing became delayed occurring 210 ± 26 ms after the stimuli had reached the RF centre ( Figure 2C , p < 0 . 0002 ) . In contrast , for the four cells tested , the delay for a flash was unaffected by disrupting inhibition ( 64 . 4 ± 8 . 7 ms vs 62 . 7 ± 5 . 2 ms , Figure 2D ) . We conclude that feedforward inhibition from amacrine cells to RGCs plays the major role in correcting for the lag in phototransduction allowing the retina to correctly signal the position of a moving object . 10 . 7554/eLife . 06250 . 004Figure 2 . Feed-forward inhibition is necessary for motion anticipation . ( A ) Top: cell-attached recording from a single RGC as a 160 µm bar moves across the retina at 500 µm s−1 ( 7 . 5° s−1 ) . Middle: whole-cell recording in the same cell 15 min after going whole-cell with 120 mM Cl− in the pipette . Bottom: spike-time histograms calculated from 20 repetitions of the stimulus for each condition . ( B ) The normalised spike rates from a plotted as a function of the distance of the bars leading edge from the RF centre , which is shown below in red . ( C ) When inhibition was disrupted by introduction of high Cl− , the peak spike rate shifted from −18 ± 11 ms before the leading edge reached the centre to 209 ± 26 ms after the centre was traversed ( n = 6; p = 0 . 0002 ) . ( D ) The delay in response to a flash was not affected by disruption of inhibition ( 64 . 4 ± 8 . 7 ms vs 62 . 7 ± 5 . 2 ms , n = 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 004 To investigate the biophysical basis of motion anticipation we constructed computational models of three RGCs whose morphologies were recovered with 2-photon microscopy ( Figure 3C ) . Although many neurons contain active dendritic conductances ( Magee and Johnston , 1995; Bischofberger and Jonas , 1997; Hausselt et al . , 2007 ) , including some RGCs ( Oesch et al . , 2005; Sivyer and Williams , 2013 ) , we began by exploring the simpler situation in which excitatory and inhibitory conductances interact with just passive properties , as this is the backbone for electrical signaling in dendrites ( London and Häusser , 2005 ) . The time-course of synaptic conductance changes used in simulations mimicked those measured experimentally ( Figure 3A ) , and the average density of synaptic inputs over the dendritic trees also matched known distributions with around one excitatory synapse per 5 µm of dendrite ( Freed and Sterling , 1988; Jakobs et al . , 2008; Xu et al . , 2008 ) and a ∼2:1 ratio of excitatory to inhibitory synapses , reflecting the ratio of the inhibitory to excitatory conductance measured over the receptive field centre ( Figure 3B , 2 . 1 ± 0 . 23 , n = 10 ) . Electron microscopy has consistently demonstrated that inhibitory synapses outnumber excitatory inputs in RGCs and the 2:1 ratio we used for simulations is relatively conservative; inhibition-to-excitation ( I/E ) ratios up to 5:1 have been observed in some ganglion cell types ( Freed and Sterling , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Owczarzak and Pourcho , 1999 ) . These synapses were distributed randomly over the dendritic tree , with the total numbers of 2960 , 3450 and 3120 in the three reconstructed cells that we used . Full details of the simulations are provided in ‘Materials and methods’ . 10 . 7554/eLife . 06250 . 005Figure 3 . Motion anticipation in model ganglion cells depends on feedforward inhibition . ( A ) The time course of the synaptic inputs used in the model were constructed with piecewise functions fit to the synaptic conductance evoked from a 320 ms −100% contrast step over the RF centre , equivalent to the stimulus that the moving bar generates at any one point on the retina ( average data shown in black and fits in colour; see ‘Materials and methods’ ) . ( B ) The inhibitory-to-excitatory conductance ratio measured over the RF centre in response to −100% contrast bar was 2 . 1 ± 0 . 23 ( n = 10 ) . ( C ) Line drawings of the 3 RGCs used for modeling , with Sholl plots to their right indicating the number of dendritic crossings for spheres of increasing distance from the soma . Excitatory synapses were placed randomly across the dendritic trees to give an average inter-synapse distance of 4 . 7 µm . Inhibitory synapses were also distributed randomly giving inhibitory to excitatory synapse ratios of 2 . 36:1 , 2 . 04:1 and 2 . 41:1 for the three RGCs shown . All synapses had identical weights . ( D ) The output of each model RGC in response to a 160 µm bar moving across its dendritic field at 500 µm s−1 , with ( black ) and without inhibition ( grey ) , the corresponding RF is shown below in red . ( E ) Motion anticipation was robust to changes in the membrane resistance ( Rm ) and axial resistance ( Ri ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 00510 . 7554/eLife . 06250 . 006Figure 3—figure supplement 1 . The voltage-clamped EPSCs ( black ) in response to the same moving stimuli as Figure 3D , plotted relative to the RF ( red ) . Note the peak EPSC occurred with a delay for the three model RGCs , similar to the data in Figure 1H . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 006 When the receptive field of these model neurons was probed by mimicking flashed bars , it reproduced a Gaussian receptive field similar to that measured experimentally ( Figures 1 , 3D ) . Crucially , the EPSPs arriving at the soma displayed motion anticipation when tested with our standard moving bar stimulus ( Figure 3D ) , demonstrating that the passive interactions of excitatory and inhibitory synaptic inputs are sufficient to generate motion anticipation . The model also predicted the effects observed experimentally after disrupting inhibition: when inhibitory synapses were removed , the peak EPSP was delayed , occurring 163 ms after the bar entered the receptive field . These basic features of motion anticipation were observed in all of the modelled RGCs , despite their different branching patterns ( Figure 3C , D ) . Consistent with our observations in Figure 1 , the voltage-clamped excitatory currents in these model ganglion cells failed to display motion anticipation ( Figure 3—figure supplement 1 ) . The integration of voltages in dendrites is dependent on the membrane resistance ( Rm ) and intracellular resistance ( Ri ) ( Rall , 1957 ) , and in the RGCs that we recorded from , Rm varied from 83 MΩ to 580 MΩ , with a median of 218 MΩ ( n = 37 ) . The ability of feedforward inhibition to generate motion anticipation was robust to variations in both Rm and Ri over this range , but was destroyed once Rm fell to 50 MΩ ( Figure 3E ) . These simulations therefore indicate that the passive cable properties of RGCs can account for motion anticipation across different dendritic morphologies and electrical properties of RGCs . Voltage-sensitive channels in dendrites can modulate integration of synaptic inputs; for example , NMDA receptors boost EPSPs that are activated in a temporal order moving towards the soma ( Branco et al . , 2010 ) and a proximal-to-distal gradient of voltage-sensitive Ca2+ channels in the dendrites of starburst amacrine cells acts to boost EPSPs moving from the soma toward distal dendrites ( Hausselt et al . , 2007 ) . To test directly whether active conductances were required to generate motion anticipation , we recorded from ganglion cells whose dendrites were made passive by dialysis of 2 mM MK-801 and 10 mM QX-314-Bromide; together these , substances block NMDA receptors and voltage-sensitive Na+ and Ca2+ channels ( Talbot and Sayer , 1996; Kuzmiski et al . , 2010 ) . In addition to preventing any dendritic boosting of EPSPs , the somatic Nav channels were also blocked allowing us to observe the generator potential responsible for spiking . About 10 min after achieving whole-cell access we verified that the drugs had blocked voltage-gated channels by delivering a depolarizing current injection; by this time spikes were blocked and the resultant depolarization was well fit by a single exponential ( Figure 4A ) . In separate experiments we found that 10 min was sufficient to completely fill the dendritic tree of ganglion cells with Alexa 488 , which has a molecular weight around twice that of MK-801 and QX-314 . We provided our standard moving stimulus to these passive cells and compared the motion-evoked EPSP to the expected linear response , obtained by convolution of the flash response with the measured RF ( Figure 4B ) . All cells had EPSPs that peaked significantly earlier than the expected linear response ( p = 0 . 0002 , n = 7 ) , occurring 47 ± 29 ms before the stimuli reached the RF centre and this was not significantly different to the time of peak spike rate observed for the cells in Figure 1 ( 47 ± 29 ms , n = 7 vs 46 ± 13 ms , n = 25 , p = 0 . 9844 ) . These results indicate that active conductances are not necessary for the computation of motion anticipation in the dendrites of RGCs . 10 . 7554/eLife . 06250 . 007Figure 4 . Motion anticipation is evident in the EPSPs of ganglion cells with passive dendrites . ( A ) Ganglion cells were dialyzed with 2 mM MK801 and 10 mM QX-314-Bromide . The efficacy of these drugs was assessed by attempting to fit a large step depolarization with a single exponential; the depolarization was well fit after 10 min of dialysis . ( B ) The EPSP evoked by a 160 µm ( 2 . 4º ) bar flashed for 320 ms over the RF centre of the same ganglion cell in a with passive dendrites . ( C ) Example of the EPSP recorded as a bar moves across the receptive field of an OFF ganglion cell ( average of 10 presentations ) . The EPSP is plotted as a function of the bar's location within the RF and the purple trace below , shows the expected linear response obtained by convolution of the EPSP from B with the RF . ( D ) The average delay for EPSPs in RGCs with passive dendrites was −47 . 2 ± 29 . 4 ms , whereas the expected linear response was always delayed with an average of 233 . 84 ± 19 . 6 ms ( n = 7 , p = 0 . 0002 ) . The delays for the three models are shown in purple , with the linear response representing the model with only excitation . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 007 The ability of the retina to correct for the delay in phototransduction breaks down at high velocities ( Berry et al . , 1999 ) . We plotted the location of maximal spiking relative to the receptive field centre as a function of velocity for 26 ganglion cells recorded on a multi-electrode array and found that motion anticipation occurred when the bar moved at velocities between 3 . 8° s−1 ( 250 μm s−1 ) and 14 . 7° s−1 ( 1000 μm s−1 ) but not at 27 . 8° s−1 ( 2000 μm s−1 ) or higher ( Figure 5A ) . This behaviour was closely reproduced by the model ( Figure 5A ) , which also predicted breakdown at velocities higher than ∼14 . 7° s−1 ( 1000 μm s−1 , Figure 5B ) . The velocity at which motion anticipation decreased is similar to that observed in salamander for similar sized objects ( Leonardo and Meister , 2013 ) . Passive integration of synaptic inputs is therefore sufficient to provide a quantitative account of several fundamental features of motion anticipation . 10 . 7554/eLife . 06250 . 008Figure 5 . Velocity-dependence of motion anticipation . ( A ) Black traces show the spike histograms from a single cell for a 160 µm bar moving at different velocities plotted against the position of the leading edge relative to the RF centre ( average of 30 presentations ) . Purple traces show the average response of the model ( RGC2 ) to the same stimulus parameters , with an I/E ratio of 2 . 04:1 . Note that the peak EPSP starts to lag behind the receptive field centre at higher velocities . ( B ) The average amount of anticipation plotted against velocity for 26 OFF ganglion cells ( black , ±SD ) . Motion anticipation operated until a velocity of about 1 mm s−1 , as also predicted by the three models ( purple ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 008 Motion anticipation causes a ganglion cell to spike preferentially when an object enters its receptive field , and we have shown that feedforward inhibition is necessary for this phenomenon ( Figures 2–5 ) . But why does inhibition have a stronger effect over the latter half of the receptive field ? To investigate the biophysical mechanisms underlying motion anticipation in more detail , we explored a model RGC with a radially symmetric dendritic tree . Figure 6A plots the membrane potential across a dendrite spanning the receptive field as a bar moves across , and compares this to the voltage experienced by the soma . In the absence of inhibition , the somatic voltage was an accurate reflection of the depolarization seen in the dendrites ( Figure 6A , left ) . However , with inhibition present , this was only the case for the first half of the RF , once the stimulus crossed the RF centre dendritic depolarizations had less influence on the somatic voltage . As a result , the peak excitatory drive occurred close to when the bar traversed the RF centre . 10 . 7554/eLife . 06250 . 009Figure 6 . A biophysical explanation of motion anticipation . ( A ) Heat plots of the voltage across the dendritic tree of a simplified model RGC as a function of time in response to a moving bar . The voltage at the soma is plotted below . Left: with only excitation present the somatic voltage follows the voltage seen in the dendrites . Right: with inhibition present , the somatic voltage only followed dendritic excitation in the first half of the RF , then hyperpolarized as the bar moved across the remainder . Note: the arrow marked a represents the stimulus entering the RF; b is time zero , when the leading edge reaches the RF centre , and c marks the leading edge reaching the distal edge of the dendritic field . ( B ) The positions of inhibitory synapses ( red ) relative to excitatory synapses ( green ) strongly affects the depolarisation observed at the soma . Left: For a stimulus moving across the initial half of the RF , distal inhibition is activated first attenuating depolarisation of the soma only modestly . The red voltage trace is the somatic response with inhibition present and the green trace is with only excitation . Middle: Proximal inhibition in the initial half of the receptive field has little effect on spiking as these synapses are activated later than the distal excitation . Right: For a stimulus moving across the distal half of the RF , proximal inhibition is activated before distal excitation and is very effective at reducing depolarisation of the soma by more distal excitatory synapses . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 009 Why does the latter half of the dendritic field appear to electrically uncouple from the soma ? In a landmark theoretical study Koch et al . laid out the conditions that make inhibitory synaptic inputs most effective at counteracting excitatory inputs: ( a ) inhibitory inputs should be located on the path between the excitatory synapse and the soma ( i . e . , proximally along the dendrite ) , and ( b ) inhibitory inputs should be activated before the distal excitatory synapse ( Koch et al . , 1983; Liu , 2004; Hao et al . , 2009; Pouille et al . , 2013 ) . Satisfying these conditions depends critically on whether the object is moving towards the soma or away from it . This idea is explained further in Figure 6B where we examined the influence of a single inhibitory synapse on the ability of a single excitatory synapse to depolarize the soma . As the moving stimulus enters the receptive field an inhibitory input distal to the soma is activated first ( Figure 6B , left ) , but it fails to attenuate depolarization because it is not on the path between the excitatory synapse and the soma ( condition a ) . If the inhibitory input is located proximal to the excitatory input , it will still be ineffective because it is activated after the excitatory drive has reached the soma ( Figure 6B , middle; condition b ) . Only when the stimulus is traversing the latter half of the RF does proximal inhibition occur just before distal excitation to ‘cut-out’ excitation reaching the soma ( Figure 6B , right ) . In this way , the peak-firing rate is brought ‘forward’ in time to compensate for the delay inherent in phototransduction . The conditions in which proximal inhibition blocks excitation arriving at the soma that were laid out by Koch et al . , 1983 immediately suggest a wiring rule for the generation of motion anticipation: place inhibitory synapses in positions proximal to nearby excitatory synapses . Motion anticipation operates for objects moving across the receptive field in any direction , so this pattern would be expected across the whole dendritic tree . Such an organization might be achieved if developmental processes cause inhibitory synapses to be placed systematically in positions proximal to the nearest excitatory synapse . Such a specific wiring rule does not , however , fit with anatomical studies , which consistently indicate that excitatory and inhibitory inputs are distributed independently and at uniform density along RGC dendrites ( Freed and Sterling , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Owczarzak and Pourcho , 1999; Jakobs et al . , 2008; Xu et al . , 2008; Schwartz et al . , 2012 ) . These two constraints , proximal inhibition across the whole dendritic tree , together with random positioning of synapses , would be satisfied if there are more inhibitory inputs per dendrite than excitatory ones , and this is observed: anatomical studies consistently demonstrate that inhibitory synapses outnumber their excitatory counterparts in the large majority of RGCs ( West , 1976; Koontz and Hendrickson , 1987; Freed and Sterling , 1988; Marshak et al . , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Haverkamp et al . , 1997; Zhu and Gibbins , 1997; Owczarzak and Pourcho , 1999; Marshak et al . , 2002; Jakobs et al . , 2008; Koizumi et al . , 2011 ) . To investigate how the ratio of inhibition to excitation affected the generation of motion anticipation we carried out simulations in which we varied the number of inhibitory inputs in the model cells . When the ratio of inhibition to excitation ( I/E ) was increased , the resultant EPSPs became smaller ( Figure 7A ) and the peak of the EPSP shifted forward in time ( Figure 7B ) . In two of the model ganglion cells , the temporal shift in the time-course of excitation began with I/E > 1 , while in the third neuron it began with I/E > 2 ( Figure 7C ) . These observations confirm that the generation of motion anticipation requires an excess of inhibitory inputs when synaptic sites are positioned randomly and independently over the dendritic tree . 10 . 7554/eLife . 06250 . 010Figure 7 . A greater ratio of inhibition to excitation is important for motion anticipation . ( A ) The output of a model RGC in response to the standard moving stimulus for various ratios of inhibitory to excitatory synapses . The number of excitatory synapses was fixed and the darkest trace represents 0 inhibitory synapses . As expected the amplitude of the response becomes smaller with increased inhibition . ( B ) The amplitudes of the EPSPs in a , normalised to compare the time course of the EPSPs . Note that as inhibition increases the peak of the EPSP moves forward in time , no longer occurring with a delay . ( C ) The amount of motion anticipation plotted as a function of the inhibition to excitation ratio ( I/E ) . Note that motion anticipation emerges as the inhibitory synapses start to outnumber the excitatory . ( D ) The velocity dependence of motion anticipation was influenced by the synaptic density . The density of synapses was varied for RGC2 while keeping the I/E ratio fixed . Densities shown for excitatory ( E ) and inhibitory ( I ) as synapses µm−1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06250 . 010 The ability of the retinal circuit to extrapolate motion breaks down at higher object velocities , measured as greater than 1 mm s−1 in most RGCs in salamander , rabbit and goldfish ( Berry et al . , 1999; Figure 5B ) . What aspect of the circuit determines this fundamental characteristic ? Our simulations demonstrated that a key variable was the average distance between excitatory and inhibitory inputs . Several authors have noted that the distance between excitatory synapses is surprisingly consistent over the dendritic tree of most ganglion cells and across many species , at ∼1 synapse per 5 µm ( Freed and Sterling , 1988; Jakobs et al . , 2008; Xu et al . , 2008; Koizumi et al . , 2011 ) . Together with a 2:1 I/E ratio , this characteristic distance predicted the observed critical velocity of ∼1 mm s−1 in our simulations ( Figure 5B ) . When the synaptic density was reduced to a quarter of its initial value , while maintaining a constant I/E ratio , the critical velocity at which motion anticipation broke down decreased from ≥1 mm s−1 to ∼0 . 5 mm s−1 ( Figure 7D ) . It therefore appears that the density of synaptic inputs impinging on a RGC's dendritic tree determines the critical velocity at which motion extrapolation breaks down .
Our measurements in goldfish indicate that the amount of motion anticipation is correlated with the RF size , with small RF exhibiting the least anticipation ( Figure 1D ) . In primate and monkey the specialized midget ganglion cell pathway has RF smaller than those presented here , consequently they may not exhibit motion anticipation . Indeed , measurements of synapse ratios obtained from electronmicroscopy of parafoveal RGCs in monkey and human indicate that some of these small RGCs have I/E ratios of ∼1:1 ( Kolb and Dekorver , 1991; Calkins et al . , 1994; Calkins and Sterling , 2007 ) , which would further diminish their ability to perform motion anticipation . This observation may make sense in the framework of separate ‘what’ and ‘where’ visual pathways ( Goodale and Milner , 1992 ) ; ganglion cells with the smallest RFs , such as the midget or parvocellular ( P ) pathway , are involved in coding object detail and correspondingly have the highest spatial resolution but slowest conduction velocities ( Gouras , 1969 ) . Whereas RGCs with larger RFs , the magnocellular ( M ) pathway , have high conduction velocities ( Gouras , 1969 ) and are involved in coding the spatial location of an object relative to the organism , a computation that would obviously benefit from motion anticipation . For the RGCs involved in the ‘where’ pathway , three general properties of the retinal connectome that give rise to motion anticipation are observed across a wide variety of species: an excess of inhibitory over excitatory inputs , a constant density of these inputs and their random distribution over the whole dendritic tree ( West , 1976; Koontz and Hendrickson , 1987; Freed and Sterling , 1988; Marshak et al . , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Weber and Stanford , 1994; Haverkamp et al . , 1997; Zhu and Gibbins , 1997; Owczarzak and Pourcho , 1999; Marshak et al . , 2002; Jakobs et al . , 2008; Koizumi et al . , 2011; Schwartz et al . , 2012 ) . The generality of these conditions across species underlines the fundamental importance of correcting for the phototransduction delay so that retinotopic mapping can be conserved when encoding the position of a moving object . This study therefore provides an example of the way in which the general rules of retinal wiring determine a key computation of visual processing . We find that the excitatory signal transmitted to RGCs is delayed relative to the retinotopic position of a moving object , indicating that the temporal correction occurs within the ganglion cell ( Figure 1H ) . A priori reasoning also suggests that motion anticipation should be autonomous to each ganglion cell , rather than occurring earlier in the retinal circuitry . Consider a single bipolar cell situated between , and contacting two ganglion cells ( Asari and Meister , 2012 ) . For a stimulus traversing the retina this bipolar cell will drive one ganglion cell at its latter edge and the second ganglion cell at its initial edge . If the gain of this bipolar cell were reduced for the first ganglion cell it would also retard the second cell's ability to respond to the same stimulus . In effect the neural image on the retina would fade as the stimulus moved across visual space . Instead , motion anticipation is generated by passive interactions within the ganglion cell dendritic tree , and is therefore relatively independent of other computations carried out by the circuitry of the inner retina . It is the non-linear interaction of inhibitory and excitatory synaptic inputs within the dendritic tree that shifts the profile of excitation forward in time . Simulations indicate that the density of synaptic inputs along dendrites determines the object velocities that can be corrected . Dendrites are a fundamental computational unit in the nervous system . In many brain regions they perform non-linear operations on the inputs they receive ( Hausselt et al . , 2007; Branco and Häusser , 2010; Smith et al . , 2013 ) , these can result simply from their passive properties or through active membrane conductances ( London and Häusser , 2005 ) . The modelling we carried out as part of this study illustrates why the classical linear-nonlinear receptive field model incompletely describes the response of a ganglion cell to a moving stimulus: excitatory and inhibitory synaptic inputs interact in a non-linear manner that depends on time and position on the dendritic tree . Koch et al . , 1983 have described the shunting effect of an inhibitory input lying on the path between an excitatory input and the soma as an analogue implementation of a Boolean AND-NOT operation in which one input vetoes another . Here we find that this AND-NOT operation is activated when the object moves away from the soma and receptive field centre , aligning the time of peak excitation closer to the time when the object is in the centre of the receptive field . Dendrites of retinal ganglion cells have also been shown to contribute to other non-linear transformations of the visual input . For example the local integration of excitatory and inhibitory inputs in PV-5 cells endows them with sensitivity to approaching objects ( Münch et al . , 2009 ) , whereas in directionally selective ganglion cells dendritic Nav channels are used to boost their direction sensitivity ( Schachter et al . , 2010; Sivyer and Williams , 2013 ) . We show that motion anticipation emerges from the passive properties of dendrites , but it is possible that subtypes of RGCs could augment anticipation using active conductances . For example , NMDA receptors selectively boost EPSPs that move centripetally along a dendrite ( Branco et al . , 2010 ) , such a mechanism could augment motion anticipation in guinea pig OFF α RGCs where NMDA receptors make a significant contribution to visually evoked spiking but not in OFF ∂ or ON α RGCs where NMDA receptors are less conspicuous ( Manookin et al . , 2010 ) . Additionally dendritic spikes combined with gap junction coupling allow a specific subtype of directionally selective ganglion cell to respond much earlier than expected to motion in its preferred direction ( Trenholm et al . , 2013 ) . A fundamental finding of our study is that motion anticipation in ganglion cells arises when feedforward inhibitory synapses outnumber excitatory inputs ( Figure 7C ) . We modelled each inhibitory synapse as purely feedforward ( Figure 1F ) , a ubiquitous and simple circuit motif that numerous types of amacrine cell provide for example , A2 narrow-field , A8 bistratified , A13 and A22 amacrine cells ( Kolb , 2005 ) . There are , however , at least 22 different types of amacrine cell ( MacNeil and Masland , 1998 ) , but we understand little about the specific response properties of these and even less about how the different types are connected to ganglion cell dendrites . The few notable exceptions highlight the sophistication that can be achieved . For example a combination of intrinsic ( Hausselt et al . , 2007 ) and synaptic ( Lee and Zhou , 2006 ) mechanisms endow individual dendrites of starburst amacrine cells with a directional preference for moving stimuli . Our model demonstrates ( Figure 6B ) that the location of an inhibitory synapse on the dendritic tree of a ganglion cell can have a large effect on the output of that ganglion cell . Indeed the operation of direction-selective ganglion cells results from the particular wiring of their starburst amacrine cell inputs ( Briggman et al . , 2011 ) . It is therefore expected that a proportion of the inhibitory inputs impinging on ganglion cells may already represent complex transformations of the visual signal . A concerted effort to elucidate the functional roles of amacrine cells in the retina is now required . Genetically-encoded indicators can be used to image the synaptic output of neurons in the retina ( Dreosti et al . , 2009; Odermatt et al . , 2012 ) and targeting these reporters to different subtypes of amacrine cells should reveal their functional diversity . High-resolution connectomics ( Briggman et al . , 2011; Helmstaedter et al . , 2013 ) is now uncovering how amacrine cells wire to the dendritic trees of different ganglion cells . The challenge in the future is to marry these two approaches so that the functional landscape can be overlaid on the connectomic map to guide explorers of the retina . The balance between excitation and inhibition profoundly affects the gain and tuning of neural responses ( Isaacson and Scanziani , 2011 ) . For instance , in the transition from the anesthetized to the awake state , increased inhibition within the visual cortex is correlated with pyramidal cell responses that are briefer in time and more narrowly tuned in space ( Haider et al . , 2013 ) . It has long been recognized that non-midget RGCs receive many more inhibitory inputs than excitatory and that these are located randomly over the dendritic tree , and independently of excitatory inputs ( Freed and Sterling , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Weber and Stanford , 1994; Owczarzak and Pourcho , 1999 ) . This characteristic wiring can now be understood in the context of the mechanism generating motion anticipation; a surplus of inhibitory inputs on each dendrite ensures that stimuli entering a receptive field are transmitted more effectively to the soma than stimuli traversing the latter half . Our simulations ( Figure 7 ) and the analysis of Koch et al . , 1983; demonstrate that variations in the distance between inhibitory and excitatory inputs will vary the object velocity that provides the most effective shunting of excitation to the soma because this shunting depends on the time delay between these conductances . The velocity tuning of motion anticipation is therefore expected to vary within individual dendrites . Nonetheless , the average intersynaptic spacing observed over many RGCs is surprisingly constant ( Jakobs et al . , 2008; Koizumi et al . , 2011 ) and reproduces the velocity tuning observed physiologically ( Figure 5B and Figure 7D ) . It appears that the general conditions that generate motion anticipation in the retina , topographically organized neurons receiving feedforward inhibition , also exist in downstream areas of the visual pathway , including the superior colliculus or optic tectum ( Bollmann and Engert , 2009 ) , thalamus ( Blitz and Regehr , 2005 ) and area V1 of the visual cortex ( Miller , 2003; Haider et al . , 2013 ) . A topographic organization of sensory inputs combined with feedforward inhibition onto dendrites may therefore be a general mechanism for correcting time delays in neural signals relative to events in the external world .
All procedures were carried out in accordance with the UK Animals ( Scientific Procedures ) act 1986 . Retinae were removed from goldfish ( 15–20 cm long ) and placed in AMES solution ( Sigma–Aldrich , Gillingham , UK ) diluted to 270 mOsM . Pieces of retina ∼1 cm−2 were mounted photoreceptor side down in a recording chamber and perfused at 2–3 ml min−1 with AMES bubbled with 95% CO2/5% 02 . Retinal ganglion cells were visualized under infrared light using a camera and recordings made with an Axopatch 200B ( Molecular Devices , Sunnyvale , CA ) . Extracellular recordings were made in voltage-clamp mode . For whole-cell voltage-clamp recordings , the intracellular solution contained 104 mM CsMeSO4 , 8 mM CsCl , 5 mM Na2 Phosphocreatine , 4 mM HEPES , 2 mM Mg . ATP , 1 mM Na . GTP , 1 mM EGTA and 2 mM QX-314-Chloride , with this solution the calculated reversal potential for Cl− was −59 mV . Voltage-clamping at −60 mV and 0 mV isolated the EPSC and IPSC respectively . For whole-cell current-clamp measurements with high Cl− the intracellular solution contained 120 mM KCl , 10 mM Na2 Phosphocreatine , 4 mM HEPES , 2 mM Mg . ATP , 1 mM Na . GTP , 0 . 15 mM EGTA . For the passive dendrite recordings in Figure 4 the intracellular solution contained 104 mM KMeSO4 , 8 mM KCl , 10 mM Na2 Phosphocreatine , 10 mM QX-314-Bromide , 4 mM HEPES , 2 mM MK801 , 2 mM Mg . ATP , 1 mM Na . GTP , 0 . 15 mM EGTA . Pipettes had a resistance of 5–6 MΩ . Signals were digitized using an ITC18 A-D converter and acquired on a Mac mini using Neuromatic running in Igor Pro 6 ( Wavemetrics , Lake Oswego , OR ) . Spikes were recorded from RGCs on a 60 channel multi-electrode array ( Multichannel Systems , Reutlingen , Germany ) using perforated electrode arrays . Spikes were sorted as described previously ( Johnston et al . , 2014 ) using Wave_clus ( Quiroga et al . , 2004 ) . Stimuli were repeated 30 times each . Only cells with responses to the standard moving bar stimulus at all velocities were included for further analysis . A 852 × 600 pixel monochromatic OLED micro-display ( eMagin , part number EMA-100100 , Bellevue , WA ) was focused onto the photoreceptor layer of the retina through an oil condenser . Pixels measured 4 × 4 µm on the retinal surface . Visual stimuli were delivered via Matlab ( Mathworks , Natick , MA ) using psychophysics toolbox libraries . Visual stimulation and electrophysiology were synchronized by recording the times of screen refreshes and the timing precision was verified with PMTs . The mean irrandiance was 40 nW mm−2 and our standard stimulus was a bar of −100% contrast measuring 160 µm by 2400 µm on the retina . To relate the retinal images to objects in the real world we measured the distance between the centre of the lens and the retina , for the goldfish used in this study ( ∼150 mm in length ) this was ∼3 . 8 mm . All sizes and speeds have been converted to degrees of visual angle according to the equation: R/n = tanV , where R is the size of the retinal image , n is the distance from the retina to the lens centre and V is visual angle . Therefore our standard stimulus of a 160 µm bar covered ∼2 . 4° of visual angle and moved at ∼7 . 5° s−1 . Initially RGCs were classified as OFF vs ON and brisk-transient vs brisk-sustained by flashing full field stimuli of −100 and 100% contrast for 0 . 5 s each . Brisk transient cells responded strongly to stimulus onset then adapted completely . Brisk-sustained cells responded strongly initially and then adapted slightly over the rest of the stimuli . No ON cells were encountered in our recordings . We also tested for both direction-selectivity and orientation-selectivity by moving a bar across the retina at four different angles . We encountered two orientation-selective cells but no direction-selective cells . The orientation selective cells responded strongly to a moving bar in their preferred orientation and failed to spike to a bar orthogonal to this axis . Similar to ( Johnston et al . , 2014 ) , the RF of each RGC was mapped by flashing a −100% contrast 80 µm ( 1 . 2° ) bar at pseudo-random locations across a single axis of the retina ( the same used for motion stimuli ) . The order was then deshuffled and the total spikes for each bar was counted in a 150 ms window starting at each flash time , a Gaussian was then fit to this data . RGCs were filled with 50 µM alexa 488 by dialysis through the patch pipette then , subsequent to electrophysiological recording , a volume containing the RGC dendrites and soma was acquired with a custom built 2-photon microscope similar to ( Esposti et al . , 2013 ) . The morphologies of each RGC were then digitized using the ‘Simple Neurite Tracer’ plugin for ImageJ; these were then exported as SWC format , down sampled using custom written scripts ( available at http://www . igorexchange . com/project/DendritePruner ) and imported to neuroConstruct . Morphologically realistic models of RGCs with synaptic inputs were constructed in neuroConstruct ( Gleeson et al . , 2007 ) and ran in NEURON ( Hines and Carnevale , 2001 ) . The electrical properties used throughout the manuscript were: intracellular resistance ( Ri ) = 180 Ω . cm , membrane capacitance ( Cm ) = 1 µF cm−2 and membrane resistance ( Rm ) = 20 MΩ cm−2 , however , we found that these parameters were not critical for the appearance of motion anticipation ( Figure 4 ) . The dendrites of the retinal ganglion cells traversed a virtual inner plexiform layer that was populated with excitatory bipolar and inhibitory amacrine synapses using the ‘cubic close packed cell packing adaptor’ in neuroConstruct . Bipolar terminals were packed with an effective radius of 5 . 1 µm and amacrine cells with an effective radius of 4 µm . We set connections between synapses and the dendritic tree of each RGC using morphology based connections . Searches for a connection between each synaptic terminal and a target dendrite were completely random with a distance constraint of 10 µm , 300 attempts were initiated for each synapse . This gave a density of 0 . 182 , 0 . 303 and 0 . 189 excitatory synaptic contacts µm−1 of dendrite for RGC1 , 2 and 3 respectively , which is close to the observed density ( Freed and Sterling , 1988; Jakobs et al . , 2008; Xu et al . , 2008 ) . The density of inhibitory synapses was 0 . 426 , 0 . 606 and 0 . 459 synapses µm−1 of dendrite for each RGC , this higher density of inhibitory contacts is a conservative estimate of the number of inhibitory contacts seen with electron microscopy which can be as high as five times the number of excitatory contacts ( Freed and Sterling , 1988; Hitchcock , 1989; Kolb and Nelson , 1993; Owczarzak and Pourcho , 1999 ) . Each synapse was modeled as a point process , and the time-course of the currents were described by three piecewise functions obtained by fits to the measured synaptic conductances ( Figure 3A ) . The three piecewise functions correspond to the onset of the steady state phase ( a ) , the decay after the stimulus ( b ) and the adaptation to the stimulus over the steady state phase ( c ) . The synaptic conductance was G ( t ) = m × ( a + b − c ) where m is a scaling factor . For the excitatory conductance:a ( t ) ={0 . 9871+e ( 52 . 9+t0−t ) 5ift0<t<t0+td+52 . 90otherwise , b ( t ) ={0 . 18∗e ( t0+td−t ) 95 . 9ift0+td+52 . 9<t<t0+td+4750otherwise , c ( t ) ={0 . 921+ ( −0 . 623∗e ( t0+88−t ) 17 . 3 ) + ( −0 . 274+et0+88−t235 . 6 ) ift0+88<t<t0+td+52 . 90otherwise , and for the inhibitory conductance:a ( t ) ={11+e ( t0+62 . 37−t ) 4 . 5ift0<t<t0+td+62 . 40otherwise , b ( t ) ={0 . 2046∗e ( t0+td−t ) 235 . 6ift0+td+1200<t<t0+td+62 . 40otherwise , c ( t ) ={0 . 7+ ( −0 . 63∗e ( t0+82−t ) 69 ) ift0+82<t<t0+td+62 . 40otherwise , where t0 is the stimulus onset time and td is the duration of the stimulus in ms . The delays inherent to the retinal circuitry , comprising both phototransduction delay and synaptic delay , are accounted for in these functions . Scaling factor m was set to 0 . 00003 ns for all synapses . To facilitate assignment of an activation time ( t0 ) to each synapse , the array of bipolar and amacrine synapses were divided in to strips with a width of 30 µm . The RF of each model ganglion cell was measured by stimulating a single 30 µm strip and plotting the amplitude of the somatic EPSP vs space , similar to the performed physiological measurements . To simulate a bar moving across the retina each strip was given a set of values for t0 that reflected the time required for the leading edge of the bar to traverse the 30 µm strip , each synapse was then randomly assigned a value from this set . For example , for a bar moving at 0 . 5 mm s−1 the t0 values of the first strip would range from 0 ms to 60 ms and for the subsequent strip range from 60 ms to 120 ms . For the simple model used in Figure 6 , the soma had a diameter of 15 µm the primary dendrite extended for 20 µm then branched into eight radially symmetric dendrites of 200 µm length with a diameter of 1 µm . | The retina is a structure at the back of the eye that converts light into nerve impulses , which are then processed in the brain to produce the images that we see . It normally takes about one-tenth of a second for the retina to send a signal to the brain after an object first moves into view . This is about the same time it takes a tennis ball to travel several meters during a tennis match , yet we are still able to see where the moving tennis ball is in real time . This is because a process called ‘motion anticipation’ is able to compensate for the delay in processing the position of a moving object . However , it was not known precisely how motion anticipation occurs . Inside the retina , cells called photoreceptors detect light and ultimately send signals ( via some intermediate cell types ) to nerve cells known as retinal ganglion cells . These signals can either excite a retinal ganglion cell to cause it to send an electrical signal to the brain , or inhibit it , which temporarily prevents electrical activity . Each cell receives signals from several photoreceptors , which each connect to a different site along branch-like structures called dendrites that project out of the retinal ganglion cells . Johnston and Lagnado have now investigated how motion anticipation occurs in the retina by using electrical recordings of the activity in the retinas of goldfish combined with computer simulations of this activity . This revealed inhibitory signals , sent from photoreceptors to retinal ganglion cells via a type of intermediate cell ( called amacrine cells ) , play a key role in motion anticipation . The ability to track motion effectively in all directions requires more inhibitory signals to be sent to the dendrites of a retinal ganglion cell than excitatory signals . These two types of input must also be randomly distributed across the cell . Furthermore , it is the density of these input sites on a dendrite that determines how well the retina can compensate for the motion of a fast-moving object . The building blocks required for motion anticipation in the retina are also found in visual areas higher in the brain . Therefore , further work may reveal that higher visual areas also use this mechanism to predict the future location of moving objects . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | General features of the retinal connectome determine the computation of motion anticipation |
Antimicrobial resistant determinants ( ARDs ) can be transmitted from livestock systems through meat products or environmental effluents . The public health risk posed by these two routes is not well understood , particularly in non-pathogenic bacteria . We collected pooled samples from 8 groups of 1741 commercial cattle as they moved through the process of beef production from feedlot entry through slaughter . We recorded antimicrobial drug exposures and interrogated the resistome at points in production when management procedures could potentially influence ARD abundance and/or transmission . Over 300 unique ARDs were identified . Resistome diversity decreased while cattle were in the feedlot , indicating selective pressure . ARDs were not identified in beef products , suggesting that slaughter interventions may reduce the risk of transmission of ARDs to beef consumers . This report highlights the utility and limitations of metagenomics for assessing public health risks regarding antimicrobial resistance , and demonstrates that environmental pathways may represent a greater risk than the food supply .
Food production and food products are important potential sources of antimicrobial resistant ( AMR ) infections in humans . Beef is a widely consumed protein commodity , and production and consumption is expected to increase in the United States and globally ( Daniel et al . , 2011; OECD/Food and Agriculture Organization of the United Nations , 2015 ) . In North American beef production , several critically important antimicrobial drugs ( AMDs ) such as fluoroquinolones , macrolides and third-generation cephalosporins are used , while others are not , e . g . , carbapenems ( World Health Organization , 2011; Food and Drug Administration , 2003 ) . Use of these AMDs is thought to increase the risk of AMR being transmitted to humans through environmental exposures ( i . e . , air , water and soil ) , occupational exposures ( Levy et al . , 1976; Moon et al . , 2015 ) , as well as through consumption of beef products ( Antibiotic Resistance from the Farm to the Table [Internet] , 2014 ) . While surveillance for foodborne AMR pathogens has been part of North American food safety systems for decades ( National Antimicrobial Resistance Monitoring System – Enteric Bacteria ( NARMS ) , 2011; Canadian Integrated Program for Antimicrobial Resistance Surveillance ( CIPARS ) , 2013 ) , we have yet to fully understand and quantify the public health risk posed by transmission of non-pathogenic bacteria that carry antimicrobial resistance determinants ( ARDs ) . These ARDs could pose a risk to human health if the bacteria carrying them become established within the microbiome of the human host , subsequently enabling horizontal gene transfer to pathogens ( Forsberg et al . , 2012; Rolain , 2013 ) ; or if these ARDs are present in opportunistic pathogens that become established within an immunocompromised individual . Establishment within an individual’s microbiome could occur either through the ingestion of contaminated food products or through exposure to environmental effluents disseminated from beef feedlots ( Antibiotic Resistance from the Farm to the Table [Internet] , 2014 ) , i . e . , a facility where cattle are aggregated , reared in pens ( i . e . , outdoor enclosures ) and fed a high-energy ration before being slaughtered . The rate at which ARDs from beef products or production facilities become established within humans is unknown , largely due to an historical reliance on culture and isolation of pathogens and an inability to access the microbial community and its complete repertoire of ARDs ( i . e . , the resistome ) . Several steps in the beef production system could play crucial roles in the transmission of AMR from beef production to humans via environmental interfaces and beef products . In North America , use of AMDs is much greater in feedlots than any other phase of beef production ( Gow et al . , 2008; Rao et al . , 2010 ) , a fact that has raised concerns that these operations could represent the principal phase of beef production in which AMR is acquired or maintained . Furthermore , feedlots are intricately linked to environmental exposure pathways such as air , manure , soil , and water , enabling indirect human exposure to feedlot effluents ( McEachran et al . , 2015 ) . In North America , abattoirs ( i . e . , slaughterhouses ) are a potential control point for the transmission of AMR , as they employ sequential antibacterial interventions to reduce pathogen contamination in beef products; testing of these interventions has demonstrated that they are effective in reducing not only pathogens , but also total bacterial contamination of beef products ( Bacon et al . , 2000 ) . We hypothesized that the antimicrobial interventions and other procedures used in feedlots and abattoirs would exert a measurable effect on the presence , abundance and composition of ARDs in the bacterial populations of cattle , the feedlot environment and market-ready beef products . Furthermore , we hypothesized that the use of a metagenomics approach would enable us to quantify these changes at an ecological level and therefore better understand the risk to public health , compared to the use of a culture-based approach . In order to understand how feedlots and abattoirs affect the transmission of ARDs , it is imperative to track pens of cattle through the beef production system ( Figure 1 ) , documenting AMD use and antimicrobial interventions and describing resistome changes over time . However , research in this area has been constrained by the challenges of tracking beef products and environmental effluents from individuals or pens of cattle , and collecting detailed records of AMD exposures for the cattle being studied . Specific challenges include lack of unique animal identification , use of non-computerized or hard-to-access AMD treatment records , effluents that are difficult to trace ( e . g . , air and runoff water ) , disassembly of carcasses into hundreds of non-linked parts , and the sheer difficulty of obtaining relevant , representative samples from feedlot steers , which weigh over 450 kg . Because of these complexities , studies in this area have been constrained to descriptions of AMR prevalence in isolated sectors of the beef production process without access to relevant AMD exposure data ( Sheikh et al . , 2012 ) ; or they have relied on the use of AMD and AMR data at very abstract levels such as the nation-state ( Chantziaras et al . , 2013 ) . To our knowledge , no studies have specifically tracked antimicrobial use in cattle while investigating antimicrobial resistance in market-ready products or consumers . This dearth of evidence greatly complicates efforts to develop effective policies related to antimicrobial use in livestock with the goal of protecting public health ( Landers et al . , 2012 ) . The objective of this study was to perform a prospective longitudinal analysis of antimicrobial use and resistance in beef production and to exploit shotgun metagenomics to characterize resistome dynamics in the environment and the products of cohort cattle from feedlot through to market-ready product . 10 . 7554/eLife . 13195 . 003Figure 1 . Overview of sampling design . Cattle in this study were born on ranches and entered the feedlots between 3 and 12 months of age . In the feedlots , we collected pooled fecal ( black pin ) , soil ( red pin ) , and drinking water ( blue pin ) samples from 2 pens of cattle in each of 4 feedlots . These samples were collected once around the time that study cattle arrived in the feedlot ( 'arrival' ) , and then once when the same cattle had reached slaughter weight and were ready to exit the feedlot ( 'exit' ) . Study cattle were then loaded onto transport trucks for shipment to the abattoir . Pooled swabs ( green pin ) from the inside walls of the transport trucks were collected immediately after the cattle had been unloaded at the abattoir ( 'truck' ) . Cattle were then placed into a holding pen outside of the abattoir , where pooled fecal ( black pin ) and drinking water ( blue pin ) samples were collected ( 'holding' ) . Cattle then entered the abattoir , where they were humanely slaughtered and their carcasses disassembled into beef products for retail . At the end of this process , we collected swabs ( yellow pin ) from the conveyor belts used to move carcass parts ( 'conveyor' ) , as well as rinsates ( yellow pin ) of the carcass trimmings used to make ground beef ( 'trimmings' ) . See Figure 1—source data 1 for sampling details , including exact sampling dates for all 8 pens in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 13195 . 00310 . 7554/eLife . 13195 . 004Figure 1—source data 1 . Sample collection details , by location , sample matrix and pen . DOI: http://dx . doi . org/10 . 7554/eLife . 13195 . 004
This study reports the use of shotgun metagenomics in a novel investigation of AMR that tracked specific pens of intensively-managed cattle from feedlot entry through slaughter to market-ready product in a longitudinal fashion . While our results are more directly relevant to large North American feedlot operations , the general approach can easily be extended to other sectors of beef production , other countries , and other livestock production systems . This is particularly important given the increasing availability of next-generation sequencing machines worldwide . However , our study also emphasizes the need to properly contextualize shotgun metagenomics results , including understanding the limitations of sequencing coverage in different matrices . While our results suggest that slaughter-based intervention systems minimize the likelihood of intact ARDs being passed through the food chain , they also highlight the potential risk posed by indirect environmental exposures to the feedlot resistome . This concern is especially salient given evidence in this study that ARDs may be 'shared' between the pens of cattle and feedlots within a geographic region , indicating environmental connectivity that could also extend to human habitats through wastewater run-off , manure application on cropland , and windborne particulate matter . The pattern of resistome change during the feeding period suggests that AMD use practices may be a driving force shaping the feedlot resistome , but more research is warranted . In particular , this study utilized pooled samples to explore resistome dynamics within and between groups of animals; the excrement or meat products from animals treated with a rarely used antimicrobial ( e . g . , a fluoroquinolone or β-lactam ) were less likely to be included in the pooled samples compared to end-products of animals treated with a commonly-used antimicrobial ( e . g . in-feed macrolide ) . Future research should include the sampling of individual treated and untreated animals to better discern the effect of AMD exposure on the resistome . Furthermore , the scientific community urgently needs to develop a better understanding of the risk of different resistomes and resistance genes ( Martínez et al . , 2015 ) and to unify ARD nomenclature so that databases are standardized and analyses are comparable across studies ( Hall and Schwarz , 2015 ) . Finally , this study highlights the utility of an ecological , metagenomics , and systems approach to investigating AMR in food production , and provides unique insights that can be used to better inform agricultural and public health policy .
AMU data were obtained from feedlot owners and/or managers , and were aggregated and analyzed at the pen level because the samples were collected as pooled composites . Due to a lack of published studies on the resistome and/or microbiome of beef feedlots and abattoirs , we were unable to perform a formal sample size estimation . Somewhat related studies have utilized <5 animals and <3 biological replicates per animal ( Wichmann et al . , 2014; Durso et al . , 2011 ) , and therefore we felt that composite samples from 8 pens would provide a more representative sample set than those used in currently published studies . While it is unlikely that pooled samples represent all individual animals within a pen or unit , they do represent a group-level sample that can be used to understand group-level dynamics . This is especially true given research indicating the tendency for cohabiting individuals to share similar microbiomes ( Song et al . , 2013; Yatsunenko et al . , 2012 ) . Because livestock animals are typically managed in groups , pooled samples are likely the appropriate unit of sampling . In addition , recent work has shown that pooled samples do not produce statistically or biologically significant differences in prevalence estimates for resistant generic Escherichia coli compared to individual samples ( Benedict et al . , 2013 ) . In order to balance external representativeness with biological replicates , 4 feedlots and 2 pens per feedlot were selected for sampling . Sample types ( 'matrix' ) were chosen based on relevance to public health risk ( composite feces , soil , water in the feedlots; internal truck walls during transport; feces and water in holding pens; and end of conveyor belts and trimmings in the abattoir ) . Selected sampling time points ( i . e . , arrival , exit , transport/truck , holding , and post-slaughter ) represented points in the process of beef production when ARD abundance or transmission may be influenced by production practices , and interventions can be optimally assessed or implemented . Fecal , soil , swab , and trimming samples were collected using sterile gloves sprayed with alcohol and placed into sterilized Whirl-Pak bags ( Nasco , Fort Atkinson , Wisconsin ) . Water samples were collected into containers that had been submerged in bleach for 5 min , rinsed with sterile water and autoclaved . Pooled fecal samples were collected from feedlot and holding pen floors; pooled soil samples were also collected from feedlot pen floors , but not holding pen floors , which were concrete . Investigators walked through pens on diagonal lines , collecting ~30 g of feces or soil from 12 equally spaced locations . The 12 soil and 12 fecal samples from each pen were then placed in one Whirl-Pak bag each ( Nasco ) and mixed thoroughly . Water trough contents in each pen were thoroughly mixed , and water samples ( 1 L each ) were collected and placed into sterile containers . Truck samples were collected using an EZ Reach polyurethane sponge pre-hydrated with 10 mL Dey/Engle neutralizing broth ( World Bioproducts LLC , Mundelein , Illinois ) , which was used to swab the internal walls of each truck ( sides , door and floor ) 20 times on the front and back of each sponge . For each pen , 3 of 5 trucks were randomly selected for sampling , and 1 sponge was used per truck . After slaughter , carcasses in each pen were grouped and processed by USDA quality grade . Post-slaughter samples were obtained when the USDA grading group with the greatest number of carcasses was being disassembled and processed . In the slaughter room , EZ Reach polyurethane sponges pre-hydrated with 10 mL Dey/Engle neutralizing broth were used to collect swab samples at the end of conveyor belts used to process chuck ( i . e . , the shoulder ) and round ( i . e . , hind leg ) primal cuts , and trimmings . The end of these belts represented the last stage in the slaughter and disassembly process , immediately prior to beef being packaged for retail distribution . Sponges were held on each running belt for one minute per side . Beef trimming samples were collected from the trim conveyor belt immediately prior to spraying of the last antimicrobial solution in the slaughter process . All samples were transferred on ice to the Center for Meat Safety & Quality at Colorado State University . The samples collected in Colorado arrived within one hour of collection , and the samples collected in Texas arrived within 48 hr . Upon arrival , fecal , soil , swab , and trimming samples were immediately frozen at -80°C . Water samples were centrifuged at 15 , 000xg for 20 min at 4°C , and 5 mL of the pellet was collected for DNA extraction . Pre-planned analyses included statistical testing of resistome NMDS ordination results by matrix type , sampling location , pen , feedlot and state; formal statistical comparison of resistome richness and diversity metrics between the same factors; and multivariable modeling of log2-fold change in abundance of ARDs , resistance classes and mechanisms between sampling locations . Additional analyses , including those related to the microbiome and procrustes rotation of NMDS ordination results , were performed post-hoc for purposes of hypothesis generation . All ordinations were pre-planned and were conducted on 2 dimensions with 'vegan’s' metaMDS function ( Oksanen et al . , 2014 ) , using Euclidean distances between Hellinger transformed read counts that had been normalized using CSS , as described above ( Legendre and Gallagher , 2001 ) . The metaMDS function enables the discovery of a stable ordination solution using many random starts . The significance of study variables in explaining ordination variation was tested using permutational multivariate analysis of variance using distance matrices as implemented in 'vegan' ( function 'adonis' ) ( Anderson , 2001 ) . In addition , we included post-hoc statistical tests of NMDS ordinations using the Analysis of Similarity test as implemented in 'vegan' ( function 'anosim' ) , in order to provide a measure of effect with the corresponding R-statistic . Post-hoc procrustes superimposition was performed on results of NMDS ordination of resistome ARD and microbiome species composition , for arrival and exit samples , and the M2 statistic was used to assess correlation of ordinations . A non-column-clustered heatmap ( Figure 2A ) was generated on counts of ARDs that had been normalized using CSS as described above . Rows were clustered using the complete linkage method . Richness was defined as the number of unique features ( ARDs , mechanisms , classes , species or genera ) in a sample , while diversity was calculated using Shannon’s Index . Pre-planned comparisons of richness and diversity between samples were conducted using paired Wilcoxon signed rank test due to the presence of repeated measures when comparing different sampling locations ( e . g . , arrival vs . exit , pre- vs . post-slaughter ) . Microbiome analysis was conducted post-hoc as a means to identify potential shifts in the microbial community as a result of pathogen-reduction interventions during the slaughter process; and to identify the amount of correlation between the microbiome and resistome during the time cattle were in the feedlot . Kraken was used to classify reads phylogenetically , using default settings ( Wood and Salzberg , 2014 ) . A very high number of reads for all samples were assigned to Achromobacter xylosoxidans strain NBRC 15126 , a bacteria that should not be prevalent in these samples . Upon further inspection , this genome had been tagged as 'misassembled' and repressed by NCBI . Therefore , we removed the genome from the kraken database and re-ran the program . The output of kraken was converted into a count matrix with taxa as rows and samples as columns , and the count for each cell representing the number of reads classified to that taxon , by sample . Taxa present in fewer than 10 samples were removed from further analysis to provide robust estimates of changes in abundance ( N = 882 out of 3962 taxa ) . The count for each taxon was normalized within samples using CSS and a percentile of 0 . 5 ( Paulson et al . , 2013; McMurdie and Holmes , 2014 ) , and normalized counts were aggregated to the species and genus levels . In order to obtain a community-level view of the pattern of change from pre- to post-slaughter samples , multivariate , zero-inflated Gaussian mixture models were fit to species and genus-level normalized counts using metagenomeSeq’s 'fitZig' function , with 'useCSSoffset' set to 'FALSE' as aggregation was performed with normalized counts ( Paulson et al . , 2013 ) . All models included pen identification number as a covariable to account for potential clustering of observations . The output of fitZig was then transferred into limma’s 'makeContrasts' and 'eBayes' functions to conduct pairwise comparisons of log2-fold change in abundance between sample groups ( Smyth , 2004 ) , adjusting for multiple comparisons using the Benjamini-Hochberg procedure and using a critical α of 0 . 05 . | When bacteria become resistant to antibiotics , it becomes difficult or impossible to treat infections in both people and animals . Antibiotic resistance is a growing problem , and many fear a “post-antibiotic era” in which common infections become life threatening . In order to slow the spread of antibiotic resistance , it is important to understand how and where this resistance develops . In general , using antibiotics increases the likelihood that bacteria will develop resistance . Therefore , locations where antibiotics are commonly used – such as hospitals , long-term care facilities , livestock facilities ( such as feedlots ) and crop production areas ( such as orchards ) – may help antibiotic resistance to develop and spread . However , it is largely unknown how much each location promotes the emergence of antibiotic-resistant bacteria . Until recently , we have only been able to investigate how resistance develops in bacteria grown in a laboratory; or to look for a handful of specific resistance genes in a sample of bacteria collected from people , animals or the environment . Fortunately , a technology called next-generation sequencing now allows us to look at all the resistance genes within all the bacteria in a sample . This may help us to improve our understanding of how and where resistance develops and spreads . Noyes et al . have now used next-generation sequencing to describe the antibiotic resistance potential ( known as the “resistome” ) found in various types of samples collected from feedlots and slaughterhouses involved in producing beef . This showed that the number of different resistance genes in the samples decreased while cattle were in the feedlot and during the slaughter process . Several groups of resistance genes that were detected when the cattle first arrived in the feedlot were not detected at all at the end of the feedlot period . However , some resistance genes were detected throughout the feedlot period , and these tended to be resistance genes that allow the bacteria to evade the same antibiotics that were used in the cattle . In addition , no resistance genes of any type were detected in the samples collected after the cattle had been slaughtered . As well as providing insights into the resistome of beef production , Noyes et al . ’s study also highlights the fact that we need to develop a deeper understanding of the data that come from next-generation sequencing . This may involve developing new laboratory techniques and creating new methods to analyze such data . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Resistome diversity in cattle and the environment decreases during beef production |
An integrated account of the molecular changes occurring during the process of cellular aging is crucial towards understanding the underlying mechanisms . Here , using novel culturing and computational methods as well as latest analytical techniques , we mapped the proteome and transcriptome during the replicative lifespan of budding yeast . With age , we found primarily proteins involved in protein biogenesis to increase relative to their transcript levels . Exploiting the dynamic nature of our data , we reconstructed high-level directional networks , where we found the same protein biogenesis-related genes to have the strongest ability to predict the behavior of other genes in the system . We identified metabolic shifts and the loss of stoichiometry in protein complexes as being consequences of aging . We propose a model whereby the uncoupling of protein levels of biogenesis-related genes from their transcript levels is causal for the changes occurring in aging yeast . Our model explains why targeting protein synthesis , or repairing the downstream consequences , can serve as interventions in aging .
Aging , the gradual decrease in function occurring at the molecular , cellular , and organismal level , is a main risk factor for cardiovascular disease , neurodegeneration , and cancer ( Niccoli and Partridge , 2012 ) . Understanding its driving force is the required step towards enabling interventions that might delay age-related disorders ( de Magalhães et al . , 2012 ) . While this remains an unsolved problem in biology ( Medawar , 1952; Mccormick and Kennedy , 2012 ) , significant advances in the field have shown the process of aging to be malleable at both the genetic and environmental levels , indicating that it is possible for its causal elements to be dissected . The rate of aging , however , is influenced by diverse factors , including protein translation , protein quality control , mitochondrial dysfunction , and metabolism ( Kennedy and Kaeberlein , 2009; Webb and Brunet , 2014; Lagouge and Larsson , 2013; Barzilai et al . , 2012 ) . The multitude of factors involved indicates that aging is a complex and multifactorial process , where ultimately an integrated and systems-level approach might be necessary to untangle the causal forces . Important insights into the complex process of aging originate from research on the unicellular eukaryote Saccharomyces cerevisiae , which can produce 20–30 daughter cells before its death ( Mortimer and Johnston , 1959 , and see Wasko and Kaeberlein , 2014; Denoth Lippuner et al . , 2014 for recent reviews ) . Significant contributions towards global mapping of the aging process have been demonstrated through transcriptome studies ( Egilmez et al . , 1989; Lin et al . , 2001; Lesur and Campbell , 2004; Koc et al . , 2004; Yiu et al . , 2008 ) and genome-wide single-gene deletion lifespan measurements ( reviewed in Mccormick and Kennedy , 2012 ) . However , a major task remains to comprehensively describe the molecular changes that accompany the aging process . As the exponential increase in daughter cells represents a major challenge in terms of generating sufficient numbers of aged cells , to date no comprehensive description of the changes on both the proteome and transcriptome level has been provided . Assuming that the molecular changes occurring along the replicative lifespan of yeast are , in part , responsible for its decreased viability that occurs over time , we reason that revealing the dynamic and interdependent changes that accompany this process would allow us to distinguish cause from consequence in aging . Here , we developed a novel column-based cultivation method that allowed us to generate large numbers of advanced-age cells in a constant environment . Applying next-generation RNA sequencing and shotgun proteomics , we mapped the molecular phenotypes of aging yeast cells at 12 time points , well into advanced age where the majority of cells had died due to aging . Analysis of these dynamic and comprehensive datasets allowed us to identify a general uncoupling of protein levels from their corresponding messenger RNA ( mRNA ) levels . This uncoupling was most apparent in protein biogenesis-related proteins , which we found over-represented relative to their transcripts . Using computational network-based inference methods , we found that changes in these genes had the strongest ability to predict the behavior of other genes , thereby suggesting their causal role in replicatively aging yeast . On the basis of these analyses , we provide a systems-level model of aging unifying and integrating diverse observations made within the field .
To obtain aged yeast cells , we bound streptavidin-conjugated iron beads to biotinylated cells ( adapted from Smeal et al . , 1996 ) from an exponentially growing culture . This starting cohort of mother cells was put into a column containing stainless steel mesh that was positioned within a magnetic field ( Figure 1A , Figure 1—figure supplement 1 ) . The daughter cells do not inherit the iron beads , as the yeast cell wall remains with the mother during mitosis ( Smeal et al . , 1996 ) . By running a constant flow of medium through the column , we washed away the majority of emerging daughter cells . The flowing medium also provided fresh nutrients and oxygen and ensured constant culture conditions , as confirmed for pH , glucose , and oxygen levels ( Figure 1—figure supplement 2A–C ) . By maintaining multiple columns simultaneously , we could harvest cells from the same starting cohort at different time points and thus at different replicative ages ( Figure 1—figure supplement 2D ) . Because we could retain up to 109 mother cells per column ( Figure 1—figure supplement 3 ) , we could produce sufficient numbers of aged cells for performing parallel proteome and transcriptome analyses . Computer simulations showed that the age distribution broadened over time ( Figure 1—figure supplement 4A , B ) . The broadened age distribution results in a lower resolution making detecting the actual changes occurring at later time points more difficult , and we therefore harvested cells at exponentially increasing time intervals to maximize the differences between time points at later ages . 10 . 7554/eLife . 08527 . 003Figure 1 . Experimental design for analysis of molecular changes during the replicative lifespan of yeast and its validation . ( A ) Schematic overview of the column-based cultivation and data analysis pipeline with 16 parallel columns , where ( zoom in ) mother cells ( M ) containing streptavidin-bound ( green triangles ) iron beads ( black circles ) were captured on the magnetized column and aged under constant environmental conditions , while the daughter cells ( D ) were flushed away . Samples are collected in two replicate campaigns ( R1 , R2 ) at indicated time points in the lifespan . ( B ) Flow cytometry-based assessment of viability of mother ( Avidin-fluorescein isothiocyanate positive [AvF] ) and daughter ( AvF negative ) cells in R1 and R2 , calculated for each time point comparing viable ( propidium iodide [PI] negative ) versus inviable ( PI positive ) cells in harvested samples Mix 1–3 ( see figure 2A for explanation of Mix 1–3 ) . The solid black line represents cell viability in time measured for the same strain in the same media using a microfluidic device ( Lee et al . , 2012; data from Huberts et al . , 2014 , was obtained from the authors ) . ( C ) Cell size is qualitatively assessed with median forward scatter of live mothers ( AvF positive , PI negative ) vs live daughters ( AvF and PI negative ) . Dashed line represents the median forward scatter of young cells that have reached the fully-grown cell size to start their first division . ( D ) Aging was qualitatively assessed throughout the experiment by observing an increase in median WGA intensity over time in a population of primarily mothers ( Mix 2 ) compared to a sample composed primarily of daughters flushed out of the column ( Mix 3 ) . Inset: bright field ( BF ) and fluorescence microscopy image of cell stained with AlexaFluor 633 conjugated wheat germ agglutinin ( WGA ) , which selectively binds chitin in bud scars . Scale bar 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 00310 . 7554/eLife . 08527 . 004Figure 1—source data 1 . Table S1: Materials used for construction of novel column-based cultivation method . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 00410 . 7554/eLife . 08527 . 005Figure 1—figure supplement 1 . Setup of the aging columns . ( A ) Prior to being loaded on the aging column , the yeast cells are labeled with membrane impermeable Sulfo-NHS-LC-Biotin ( step 1 , green triangles ) . The LC-linker in Sulfo-NHS-LC-Biotin has a spacer arm length of 22 . 4 Å . The NHS-ester forms a covalent amide bond with primary amine groups in the lysines and at the N-termini of the yeast cell wall proteins . Streptavidin-coated magnetic beads ( black circles , step 2 ) bind with high affinity to the biotin-labeled cells . ( B ) The side view of one column setup . Medium is pumped with a flow rate of 170 ml/h via air permeable silicone tubing ( 1 ) and a T-connector ( 2 ) into the magnetized column holding the magnetic-bead-coupled yeast cells ( 3 ) . The medium leaves the magnetized column via the U-shaped tubing below the column ( 4 ) , a T-connector ( 5 ) and the outlet tubing ( 6 ) into a waste jar ( 7 ) . The medium level in the column is regulated with the air valve on top of the T-connector ( 2 ) in combination with the backpressure caused by medium in the U-shaped tubing after the column ( 4 ) . To disrupt the steady laminar effluent flow , air was allowed to enter the system via T-connector ( 5 ) . During incubation at the columns , the flow was started and clamp 1 ( C . I ) and clamp 3 ( C . III ) were open , while the air valve was closed ( C . II ) . ( C ) The items used to build the setup are presented in a simplified two-dimensional view and listed in Figure 1–source data 1 Table S1 . ( D ) Three-dimensional view of the magnet’s stand with two magnets present . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 00510 . 7554/eLife . 08527 . 006Figure 1—figure supplement 2 . Cellular aging under constant conditions . The aging columns maintain constant oxygen ( A ) and glucose ( B ) concentrations and pH ( C ) during cultivation . Oxygen concentration was determined using the Optical Oxygen Meter Fibox 3 in both fresh medium and the column effluent ( A ) . Glucose concentration was determined by enzyme-based assay Enzytec fluid D-Glucose ( B ) . The pH of the medium was measured by a conventional pH-meter in fresh medium ( t = 0h ) and in the column effluent after 24 and 48 hr in duplicate ( C ) . ( D ) Distribution of replicative ages of ( n ) cells in samples harvested at different time points as determined by counting bud scars in AlexaFluor 633 WGA-labeled cells . The bud scars were counted double blind from confocal z-stack images . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 00610 . 7554/eLife . 08527 . 007Figure 1—figure supplement 3 . Cell counts per time point . The cell counts present in each mixed-cell sample harvested from each time point of the experiment . These values ( along with fractional compositions present in Figure 2—figure supplement 3B ) were used to calculate the weighted lifespan curve presented in Figure 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 00710 . 7554/eLife . 08527 . 008Figure 1—figure supplement 4 . Simulated yeast aging population dynamics . Due to biological cell-to-cell variation in cell division rates , the age distribution of a starting cohort of cells increases at later time points . This results in an increasing overlap of ages in the mother cell populations harvested at later time points , as modeled for a starting cohort of 1000 cells ( see methods: Harvesting time points ) . The age is indicated as the replication life span ( RLS ) . ( A ) shows the distribution of mother cell ages in samples harvested at indicated equally spaced time points , ( B ) shows the distribution when samples are harvested at exponentially spaced time points , minimizing the overlap of information between neighboring samples . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 00810 . 7554/eLife . 08527 . 009Figure 1—figure supplement 5 . Characterization of mixed-cell samples . ( A ) Cells were stained with fluorescein isothiocyanate conjugated avidin ( AvF ) , which only labels cells coming from the initial biotin-labeled cohort ( see Figure 1—figure supplement 1A ) , and with PI , which is permeable only to dead cells and fluoresces upon intercalation with DNA . ( B ) Analyzing the stained samples on a flow cytometer clearly distinguishes the populations of dead or alive mother cells and dead or alive daughter cells , based on fluorescence emission . Quantification of these populations gives the fractional compositions of each mixed-cell sample ( Mix 1–3 in Figure 2—figure supplement 3 ) collected per time point . SSC-A is the side scatter area , FCS-A is theforward scatter area , FL1-A is the fluorescein emission peak area , FL3-A is the PI fluorescence emission peak area . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 009 To assess whether our column-based cultivation method generated correctly aged cells in a reproducible manner , we developed flow cytometric assays to determine the typical phenotypes of aging cells . Avidin-fluorescein isothiocyanate ( AvF ) binding to the biotin-labeled cells distinguished the starting cohort of mother cells from daughter cells ( Figure 1—figure supplement 5A ) . Dead cells were identified using propidium iodide ( PI ) , which fluoresces upon intercalating with the DNA of membrane-permeable dead cells ( Figure 1—figure supplement 5A ) . These two assays were used to determine the fractions of daughters , mothers , and dead cells in a population ( Figure 1—figure supplement 5B ) . From this data , we derived the viability of the mother cells over time , which we found to be in excellent agreement with the lifespan curve of yeast as observed in a microfluidic device ( Huberts et al . , 2014 ) ( Figure 1B ) . Using the forward scatter of the flow cytometer as a rough proxy for cell size , we could qualitatively observe the cell size increase of live mothers that is known to occur in aging mother cells ( Egilmez et al . , 1990 ) ( Figure 1C ) . Similarly , using fluorophore-conjugated wheat-germ agglutinin , which labels bud scars that appear after every division ( Powell et al . , 2003 ) , we observed an increase of bud scar staining on mother cells in the column , as also visualized by confocal microscopy ( Figure 1D , Figure 1—figure supplement 2D ) . These analyses confirmed known changes that characterize aging yeast: increased cell size and bud scars , and decreased population viability ( Figure 1B–C ) . Next , we developed a combined experimental and mathematical method to determine the molecular phenotype of aging mother cells without contributions from daughter or dead cells . The approach exploits the fact that a system of linear equations can be solved when the number of unknowns equals the number of independent equations . Specifically , while we could determine the number of mothers , daughters , and dead cells in a sample using flow cytometry , the contribution of each type of cells to the measured abundance of a particular protein or transcript was unknown . Therefore , by measuring protein and transcript abundances in three mixed samples with various proportions of mothers , daughters , and dead cells , we could mathematically un-mix the abundances . This resulted in un-mixed data for the aging mother cells . Experiments using samples containing mixed cell populations with known molecular phenotypes validated this mathematical un-mixing method for the RNA sequencing ( RNAseq ) transcriptome , targeted ( selected reaction monitoring ) proteome , and global ( shotgun ) proteome data with a <16% average error ( Figure 2—figure supplement 1 and 2; Supplementary file 1 ) . To use this data un-mixing approach , we harvested three mixed samples for each time point ( Figure 2A , Figure 2—figure supplement 3 ) . One sample was collected from the column effluent ( Mix 3 , mainly daughter cells ) . Harvesting all cells from the column and applying a further enrichment step on a larger magnet produced the two other samples: one sample contained mainly aged mother cells ( Mix 2 , 80–99% mothers ) , while the other contained an intermediate composition compared to Mixes 2 and 3 ( wash fraction , Mix 1 ) . In each of these mixed-cell samples , we determined the fraction of mothers , daughters , and dead cells and generated the mixed-population proteomes and transcriptomes . Then , we mathematically un-mixed the proteomes and transcriptomes to obtain the molecular phenotype of aging mother cells . The data was corrected for sampling artefacts related to bead labeling and cell harvesting ( Figure 2—figure supplement 4; supplemental notes 2 and 3 in Supplementary file 1 ) . Together , through this approach , we obtained pure data for aging mother cells and daughter cells . 10 . 7554/eLife . 08527 . 010Figure 2 . Mathematical un-mixing of proteomes and transcriptomes in mixed-cell populations . For each time point in the aging experiment , three samples ( mixed-cell samples 1 , 2 , 3; originating from different harvesting steps ) composed of different fractions of Mother ( M , green ) , Daughter ( D , blue ) and Dead cells ( De , red ) were harvested and analyzed . On the basis of the compositions of the mixed-cell samples ( wM , D , De ) and the determined proteome or transcriptome data of the mixed-cell samples ( Amix1 , 2 , 3 ) , with the mathematical un-mixing , we obtained un-mixed data ( AM , D , De ) over the time course of 72 hr from two replicates . See Figure 1—figure supplement 5 for details about determining the composition of the mixed-cell samples and Figure 2—figure supplement 3 for the un-mixing method . Data from proteome ( B ) and transcriptome ( C ) replicates highly correlated ( Spearman correlation >0 . 85 ) for M ( circles ) and D cells ( squares ) , indicating high reproducibility of the experimental and data processing pipelines . ( D , E ) Levels of random chosen proteins ( D ) and transcripts ( E ) from both replicate measurements ( gray ) and the fit ( solid line ) are indicated for un-mixed mother data . Raw abundance is a measure of mass spectrometry ( MS ) peak intensities ( proteome ) or fragments per kb of transcript per million mapped ( FPKM ) reads ( transcriptome ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01010 . 7554/eLife . 08527 . 011Figure 2—source data 1 . Table S2: The shotgun proteome data processing . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01110 . 7554/eLife . 08527 . 012Figure 2—source data 2 . Table S3: The transcriptome data processing . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01210 . 7554/eLife . 08527 . 013Figure 2—source data 3 . Table S4: The final shotgun proteome data . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01310 . 7554/eLife . 08527 . 014Figure 2—source data 4 . Table S5: The final transcriptome data . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01410 . 7554/eLife . 08527 . 015Figure 2—figure supplement 1 . Validation of the mathematical un-mixing procedure . ( A ) Schematic representation of samples used for validation of the mathematical un-mixing procedure , taken from fermenter-grown yeast . Log-phase represents mid-exponential growth of the culture ( L ) , deceleration phase represents a decreased growth rate around the diauxic shift ( D ) , and stationary phase is a nutrient deprived culture ( S ) . Each phase of cultivation has a unique transcriptional and proteomic signature . ( B ) The abundance of 207 proteins was measured with targeted selected reaction monitoring ( SRM ) proteomics in the samples L , D and S , and in three mixed-cell samples composed of different ratios of L , D , and S . The protein abundance in the pure samples and the abundance derived after mathematical un-mixing of the data obtained from the mixed cell-sample is shown for 10 representative proteins of the 207 proteins . ( C ) As in B , for all 207 proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01510 . 7554/eLife . 08527 . 016Figure 2—figure supplement 2 . Validation of the mathematical un-mixing procedure , shotgun proteome and RNA sequencing . ( A ) As in Figure 2—figure supplement 1C but now for proteome data obtained by shotgun proteomics . The Pearson correlations are as high as 0 . 989 , 0 . 992 , and 0 . 993 , for log , deceleration , and stationary phase samples , respectively , in the log2 scale ( top panels ) . Bottom panels show the relative errors for all proteins quantified; the abundance of the indicated number of proteins is recovered with less than 20% relative error . ( B ) As in ( A ) but here for the messenger RNA sequencing ( mRNAseq ) transcriptome data showing Pearson correlations of 0 . 945 , 0 . 956 , and 0 . 801 for log , deceleration , and stationary phase samples , respectively , in the log2 scale ( top panels ) . The indicated numbers of transcripts were recovered with less than 20% relative error ( bottom panels ) . All abundances are plotted on a log2 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01610 . 7554/eLife . 08527 . 017Figure 2—figure supplement 3 . Generation and composition of the mixed-cell samples . ( A ) I . A cohort of cells with cell-wall attached beads is maintained in the magnetized column ( magnet 1 ) and harvested at set time points ( column fraction ) when also a fraction with mainly daughter cells is collected ( column effluent , mix 3 ) . II . The harvested column fraction was applied for further enrichment on ‘The Big Easy’ EasySep Magnet ( magnet 2 ) . The bead-labeled aged cells stay in the glass tube , while the non-bead-labeled young cells are removed by pipetting . This wash is repeated three times , resulting in a sample enriched for mothers ( mother enriched; mix 2 ) , and a wash fraction ( wash , mix 1 ) . The fractional population sizes of these three mixes , schematically represented in III , were determined ( See Figure 1—figure supplement 5 ) before storage at −80°C . ( B ) The measured compositions of mother , daughters , and dead cells present in each mixed-cell sample harvested from each time point of the experiment . These fractional compositions were used in the mathematical un-mixing procedure . ( C ) Example of the mathematical un-mixing procedure: Hsp104 protein abundances ( mass spectrometry [MS] peak intensity ) for each time point in each of the mixed-cell samples ( left panel ) and the resulting un-mixed abundances visualized as fold changes on a log2 scale ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01710 . 7554/eLife . 08527 . 018Figure 2—figure supplement 4 . Validation of the bead effect correction . The effect of the beads on the proteome was highly reproducible , regardless of the number of beads per sample , and was unrelated to other biological stimuli applied on the cells . ( A ) Samples generated at different steps during biotinylation , bead labeling , and harvesting were assessed for their similarity to a sample that has undergone all processing steps ( sample D ) , as would the starting ( bead labeled ) sample of the experiment . Using targeted ( SRM ) proteomics focusing on 74 proteins known to be either strongly affected or not affected when comparing a processed to an unprocessed sample , we found that the presence of beads alone within a sample ( sample G ) was enough to match the starting bead-labeled sample ( sample D ) . The process of bead labeling itself ( sample ‘H’ , where bead labeling conditions were mimicked ) yielded proteomes that bore little resemblance to our bead-containing samples . ( B ) Cells and bead counts from flow cytometry . A cohort of 4 . 0×108 cells ( pink bar , left ) was labeled with beads , by adding a known number of beads ( 4 . 8 × 108 beads , pink bar , right ) . The number of beads attached to a biotinylated cell population ( 1 . 2 × 108 ) is the difference between free beads before ( 4 . 8 × 108 beads , pink bar , right ) and after bead labeling ( 3 . 6 × 108 cells , gray bar , right ) . The number of cells with at least one bead was counted after bead labeling and cell enrichment on a magnet ( after bead labeling , 1 . 2 × 108 , gray bar , left ) . The yield was on average 1 . 1 beads/cell . ( C ) The number of free beads and beads attached to the cells was determined for each sample with flow cytometry . The ratio of bead to cells increased maximally two fold in both replicates , most likely due to the detachment of cells from beads while being cultivated in the aging columns . ( D ) To study the effect of a small increase in bead concentration per sample we mixed unprocessed cells with different numbers of beads and performed targeted ( selected reaction monitoring [SRM] ) proteomics , using the same 74 proteins for assessment . The median of the measured peak intensities decreased with an increase of beads per sample , indicating a loss of proteins . ( E ) Nonetheless , we found that varying the amount of beads in the sample in the range relevant to the aging experiment , did not alter the degree to which the sample was changed by the presence of the beads . The Pearson correlation of these samples to the standard ( 1 . 06 beads/cell ) was higher than the correlation between two replicates of the standard . We conclude that the bead effect is highly reproducible , and can be redressed with a correction factor specific to each protein ( See Supplementary file 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01810 . 7554/eLife . 08527 . 019Figure 2—figure supplement 5 . Overview of the experimental pipeline . Detailed view of the experimental pipeline to depict number of samples collected and data processing steps . Up to 16 columns could be run simultaneously ( cartoon of red magnet with column ) and harvested throughout the aging procedure ( cartoon of lifespan curve , fraction surviving at each age ) . Time points were exponentially spaced , and covered by two partially overlapping replicate campaigns ( R1 and R2 , dots showing time points ) , of 14 and 8 time points , respectively . For each time point , either two or three samples were required for mathematical un-mixing of the population , that is , early time points ( blue dots ) , contained mainly live mother and daughter cells , without mortality in the population , and therefore required only two samples for the mathematical un-mixing of two unknowns . While later time points ( red dots ) , contained increasing levels of dead cells , and required three samples for the mathematical un-mixing of three unknowns . Replicate 1 consisted of an unprocessed sample , five time points requiring two samples for un-mixing , and 9 time points requiring three samples for un-mixing , totaling 38 samples . Replicate 2 had in the same way 23 samples , and together the two replicates consisted of 61 samples , which were processed with shotgun proteomics and RNA sequencing ( RNAseq ) transcriptomics . After ‘omics’ data was collected , a bead correction was applied to proteome data coming from samples containing beads ( see methods ) , and quality assessment of sequencing data removed four sets of samples from the transcriptome ( see methods ) . The subsequent 61 proteome samples and 50 transcriptome samples were used for mathematical un-mixing , which resulted in mother-specific data for the proteome ( R1 , 15 time points , and R2 , 9 time points ) and transcriptome ( R1 , 12 time points , R2 , 8 time points ) . Corresponding daughter-specific data also resulted from the un-mixing procedure ( not depicted in schematic ) . Finally , a reference time point was selected ( 7 . 8 hr , see Materials and methods ) and the replicate datasets were merged to produce a single time series , for each of the proteome and transcriptome , spanning 12 time points throughout the replicative lifespan of the cells . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 01910 . 7554/eLife . 08527 . 020Figure 2—figure supplement 6 . Selection of genes with highest similarity between replicates . ( A ) The coefficient of variation was calculated between the replicate datasets for each gene-product profile and a cutoff of 0 . 3 was used to select the most reproducible expression profiles between replicates , consisting of ∼90 . 9% of the proteome , and ∼84 . 4% of the transcriptome datasets . ( B ) Example of a gene profile having a coefficient of variation of 0 . 1 ( top panels ) and coefficient of variation of 0 . 3 , which just failed the cutoff for being included in the dataset ( bottom panels ) . Data shown for both proteome ( left panels ) , and transcriptome ( right panels ) , with each replicate measurement ( gray ) and the fit ( colored line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 020 In two experimental series with overlapping time points , we generated 61 samples for both the proteome and the transcriptome as required for un-mixing . After data processing , we obtained high quality data at 12 unique time points during the lifespan of replicatively aging yeast ( Figure 2—figure supplement 5 ) . We found the replicates to be in excellent agreement ( Spearman correlations >0 . 85 ) ( Figure 2B , C ) . A unified dataset was generated for both the proteome and the transcriptome by fitting the replicate datasets with a polynomial regression ( Figure 2D , E ) , only retaining highly reproducible expression profiles ( ∼85% of genes , Figure 2—figure supplement 6 ) , and resampling the fit at the actual time points of the experiment . This yielded profiles for 1494 proteins and 4904 transcripts from aging mother cells . The raw data ( Janssens et al . , 2015a; Janssens et al . , 2015b ) and the data for each processing step are provided in the supplementary Tables S2 and S3 ( Figure 2—source data 1 and 2 ) . The final datasets for aging mother cells are presented in Table S4 ( proteome ) and Table S5 ( transcriptome ) ( Figure 2—source data 3 and 4 ) . Correlation analyses between the proteomes of young cells and the proteomes of aging mother cells confirmed the expected divergence of the aging cell away from the youthful state ( Figure 3A , Figure 3—figure supplement 1 ) . Daughters from later time points showed a partially aged signature ( Figure 3—figure supplement 2 ) , consistent with the notion that rejuvenation of daughter cells is incomplete later in a mother’s life ( Kennedy et al . , 1994 ) . Furthermore , we found agreement between specific proteome changes detected by us and observations present in literature , including changes related to glycolysis , gluconeogenesis ( Lin et al . , 2001 ) , increased expression levels in energy reserve pathway proteins ( Levy et al . , 2012 ) , increases in stress response protein levels ( Erjavec et al . , 2007; Crane et al . , 2014 ) , and mitochondrial changes ( Hughes and Gottschling , 2012 ) ( Figure 3B , Figure 3—figure supplement 3 ) . Also , we confirmed that changes detected in our population-level study similarly occurred at the single-cell level , which excluded the possibility that our observed changes may reflect a gradual enrichment of a long lived subpopulation . Specifically , we see the levels of the stress-related chaperone Hsp104 and the translation elongation factor Tef1 to increase with age ( Figure 3—figure supplement 4 ) , similar to what was shown using a microfluidic platform tracking single cells ( Zhang et al . , 2012 ) . Also , other single protein changes reported to occur in literature match well ( Koc et al . , 2004; Lee et al . , 2012; Hughes and Gottschling , 2012; Zhang et al . , 2012; Lord et al . , 2015; Denoth-Lippuner et al . , 2014; Eldakak et al . , 2010; Sun et al . , 1994 ) ( Figure 3—figure supplement 4 ) . Together , these observations confirm the validity of our novel experimental design . 10 . 7554/eLife . 08527 . 021Figure 3 . The aging proteome . ( A ) The Spearman correlation at progressive time points compared with the young reference sample for the mother and daughter proteome shows a divergence away from a youthful state for the mother . ( B ) The numbers of proteins changing by at least twofold from the reference ( young ) sample per time point . Blue and red bars and text represent changes that had not occurred previously , either up- or down-regulated , respectively . Gray bars and text are changes that already occurred at a previous time point . Gene functional enrichments per grouped time points were derived from Gene Ontologies ( GO ) and are scaled with significance of enrichment obtained by database for annotation , visualization and integrated discovery ( DAVID ) bioinformatics resource version 6 . 7 ( scaling of text: DAVID enrichment score see Materials and methods and Table S6 ( Figure 3—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02110 . 7554/eLife . 08527 . 022Figure 3—source data 1 . Table S6: Full lists of GO-term enrichment scores for all enrichment analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02210 . 7554/eLife . 08527 . 023Figure 3—figure supplement 1 . The aging transcriptome diverges minimally from a young profile . Similar to Figure 3A , but for transcriptome . The Spearman correlation of the transcriptomes of mother and daughter cells at different time point compared to that of a reference ( young ) time point sample . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02310 . 7554/eLife . 08527 . 024Figure 3—figure supplement 2 . Changes in mother-age dependent daughter profiles . Heat maps ( with row clustering based on Euclidean distance ) showing changes for daughter profiles of each mother-age dependent time point for both proteome ( A ) and transcriptome ( B ) . Gene functional enrichments were determined using database for annotation , visualization and integrated discovery ( DAVID ) version 6 . 7 and summarized into representative terms ( see Materials and methods section for details ) . The enrichment score provided by DAVID for the summarized terms were used as a size-scaling factor for the text , with larger words being more significantly enriched ( scaling of text: DAVID enrichment score ) . Enriched terms are shown next to each respective heat map , for genes changing by at least twofold when comparing the daughter coming from the oldest mother age , to the daughter coming from the youngest mother age . This resulted in 33 genes twofold up-regulated and 40 genes twofold down-regulated in the proteome of 1494 proteins , and 31 genes up-regulated and 190 genes down-regulated in the transcriptome of 4904 transcripts . Fold changes are plotted on a log2 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02410 . 7554/eLife . 08527 . 025Figure 3—figure supplement 3 . Profiles that contribute to the enrichments of proteins changing more than twofold . Proteins contributing to the enrichment score for ‘stress response ( general ) ’ , or ‘glycolysis/gluconeogenesis’ that were increasing more than twofold with age , or proteins contributing to the enrichment score for ‘mitochondria ( general ) ’ and ‘DNA replication’ that were decreasing more than twofold with age were selected for visualization ( from Figure 3B ) . The fold changes are plotted on a log2 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02510 . 7554/eLife . 08527 . 026Figure 3—figure supplement 4 . Single protein profiles matching literature . Assessing the protein dynamics on the single cell level that were reported in the literature to occur in aging yeast shows agreement with our global-scale proteome dataset . Specifically , we see protein levels of the stress-related chaperone Hsp104 and the translation elongation factor Tef1 to increase with aging as was shown using a microfluidic platform tracking single cells [Zhang et al . , 2012] . Using another microfluidic platform and green fluorescent protein ( GFP ) -tagged Vph1 protein as a marker for the vacuole , it was found that the vacuole increased in size more rapidly than the cell itself , suggesting a net increase of Vph1 protein levels to occur in the aging cell [Lee et al . , 2012] . Our data shows Vph1 levels to increase with aging , in line with these observations . Furthermore , our proteome also captures the subtle changes described to occur with the Tpo1 protein and aging , where a computational model based on production and inheritance of the protein throughout aging predicted Tpo1 levels to initially increase and then gradually decrease with age [Eldakak et al . , 2010] . A recent study looking at protein abundances in young and old whole-cell extracts found that levels of the nucleoporins Nup116 and Nsp1 decrease with age , while Nup100 and Nup53 did not change significantly [Lord et al . , 2015] , and for one other nucleoporin , Nup170 , was shown that the levels increase with aging [Denoth-Lippuner et al . , 2014] , which we all also detect in our proteome data ( Figure 7D ) . Three proteins whose overexpression results in extended lifespan in yeast , Ras2 [Sun et al . , 1994] , Mxr1 [Koc et al . , 2004] , and Vma1 [Hughes and Gottschling , 2012] were observed to decrease with age . Literature references are according to main text reference numbering . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 026 To obtain further insights into the global changes in protein expression in mother cells , we plotted our dynamic data as heat map expression profiles . We found that changes started at young age , were gradual , and mostly occurred in one direction ( i . e . up , down ) ( Figure 4A , B ) . Specifically , we found that 64% ( 184/288 total changes ) of the proteins that showed a twofold change by the end of the yeast lifespan also showed a significant change in the same direction at an earlier time point ( Figure 3B ) . These findings suggest that aging is a gradual process occurring from early on . 10 . 7554/eLife . 08527 . 027Figure 4 . Protein profiles in aging yeast . ( A ) Expression profiles for the proteome were clustered using the Ward clustering algorithm and plotted in a dendrogram . Visualization of the most prominent ( red line in dendrogram ) protein fold change profiles ( log2 scale ) occurring with age , showing up-regulated ( cluster 1 ) , down-regulated ( cluster 2 ) and mainly flat ( cluster 3 ) profiles . Gene functional enrichments per grouped time points were summarized into representative terms as in Figure 3B . ( B ) Unidirectional changes occurring with aging are illustrated with a heat map of the fold changes ( log2 scale ) of proteins in the aging mother compared to the young reference sample . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02710 . 7554/eLife . 08527 . 028Figure 4—figure supplement 1 . Comparison of aging proteomes and transcriptomes . ( A ) Heat maps ( with row dendrograms based on Euclidean distance ) of proteome ( top panel ) and transcriptome ( bottom panel ) time series data , plotted as fold changes on a log2 scale . ( B ) The raw abundances ( log2 scale ) for the proteome and transcriptome are plotted against one another for young ( left panel , age 7 . 8 hr ) and old ( right panel age 72 hr ) cells . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02810 . 7554/eLife . 08527 . 029Figure 4—figure supplement 2 . Analysis of twofold changes per time point in the aging transcriptome . ( A ) The numbers of transcripts changing by at least twofold from the reference ( young ) sample per time point . Red and blue bars or text represent changes that had not occurred previously , either up- or down-regulated , respectively . Gray bars or text are changes that already occurred at a previous time point . Gene functional enrichments per grouped time points were derived from Gene Ontologies ( GO ) and are scaled with significance of enrichment obtained by database for annotation , visualization and integrated discovery ( DAVID ) version 6 . 7 ( scaling of text: DAVID enrichment score ) . ( B ) Profiles that contribute to the enrichments of transcript changing more than twofold . Transcripts contributing to the enrichment score for ‘integral to membrane’ , or ‘sporulation’ that increased more than twofold with age , or transcripts contributing to the enrichment score for ‘mitochondria ( respiration ) ’ and ‘mitochondria ( translation ) ’ that decreased more than twofold with age were selected for visualization ( from A ) . The fold changes are plotted on a log2 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 02910 . 7554/eLife . 08527 . 030Figure 4—figure supplement 3 . Analysis of aging changes clustered by expression profile . ( A ) Expression profiles for the transcriptome were clustered using the Ward clustering algorithm and plotted in a dendrogram . Three expression profile groups were selected for characterization ( red vertical line ) . ( B ) The three most prominent profile expression clusters for the transcriptome , showing mainly down-regulated ( cluster 1 and 2 ) and up-regulated ( cluster 3 ) profiles . Gene functional enrichments per grouped time points were summarized into representative terms as in Figure 3B . In one case ( asterix , ‘translation regulation’ ) , the enrichment value was scaled down ( from 10 . 2 ) to the score of the next most enriched term ( 5 . 0 ) , for better legibility of the other terms ( with first three letters kept on the original scale ) . Transcript fold changes are plotted on a log2 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 030 We next investigated whether these changes in the proteome data matched transcriptional changes . Interestingly , the RNAseq data showed similar gradual and unidirectional changes occurring from the beginning on ( Figure 4—figure supplement 1A , 2 , 3 ) . To compare the changes between the proteome and transcriptome , we determined the non-parametric Spearman rank correlation , and found a starting correlation of 0 . 75 , a value in agreement with other single-study comparisons between yeast proteomes and transcriptomes ( Csárdi et al . , 2015 ) . When comparing this correlation in time , however , we found that it declined steadily with age , down to a correlation of 0 . 70 ( Figure 5A ) . This decreasing trend was observed regardless of the statistical method used ( Figure 5—figure supplement 1 ) . Furthermore , this trend is also not an experimental artefact , since samples originating from all time points were treated identically , and both proteome and transcriptome datasets originated from the same biological samples . The decrease in correlation between the proteome and transcriptome means that they do not change synchronously . Indeed , during aging , we found different Gene Ontology ( GO ) terms to describe the changes in the proteins and transcripts that show a larger than two-fold change during aging ( Figure 3B vs . Figure 4—figure supplement 2A ) . These results indicate that , over time , protein levels were increasingly uncoupled from their transcript levels . 10 . 7554/eLife . 08527 . 031Figure 5 . A post-transcriptional overrepresentation in protein biogenesis with aging . ( A ) A progressive uncoupling of the proteome from the transcriptome in time is apparent from the decreasing Spearman correlation between the two . ( B ) Co-expression map showing fold changes ( log2 ) of 72 hr aged samples compared to the young reference , plotting the proteome versus the transcriptome . Quadrants 1 and 3 ( Q1 and Q3 ) represent changes where the protein changes match their transcript changes ( coupled ) , while quadrants 2 ( Q2 ) and Q4 ( Q4 ) reflect opposite changes ( uncoupled ) . Summarizing terms per quadrant are derived from Gene Ontologies ( GO ) as in Figure 3B ( scaling of text: DAVID enrichment score ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03110 . 7554/eLife . 08527 . 032Figure 5—figure supplement 1 . Correlation of proteome versus transcriptome using alternative statistical methods for comparison . Comparison of the proteome versus the transcriptome using the dataset of genes in common between the two . Using Pearson correlation on the raw data , Pearson correlation on log2 transformed data , or Spearman or Kendall correlations on the raw data , show similar results: a decreasing correlation of the proteome and transcriptome with age . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03210 . 7554/eLife . 08527 . 033Figure 5—figure supplement 2 . Co-expression map showing fold changes of 10 . 7 , 22 , 45 . 4 and 72 . 3 hr compared to the young reference , highlighting gene products contributing to gene enrichments . Co-expression map as in Figure 5B , showing fold changes of proteins and transcripts at 10 . 7 , 22 , 45 . 4 , and 72 . 3 hr aged time points compared to the young ( 7 . 8 hr ) reference sample . Genes contributing to enrichment scores of the most enriched processes per quadrant at 72 . 3 hr of aging ( sterol biosynthesis from Q1 , translation regulation from Q2 , cortical actin cytoskeleton from Q3 , and endoplasmic reticulum from Q4 ) are shown highlighted for each time point to illustrate their changes . The fold changes are plotted on a log2 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03310 . 7554/eLife . 08527 . 034Figure 5—figure supplement 3 . Change in posttranscriptional protein overabundance with aging . The fold change in abundance of a protein compared to a reference ( young ) sample , minus the fold change of its transcript , gives a quantity for its relative overabundance . Plotted in time are the summed values for the gene products per quadrant of the co-expression map in Figure 5B ( gray points ) , and for all genes of the entire plot summed ( black points ) . This shows a net increase over time of total relative protein overabundance , and a distinct behavior per quadrant . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 034 To identify the most uncoupled cellular processes , we plotted the fold-changes of transcript and protein expression in old and young cells on a gene product co-expression map ( Figure 5A ) . The transcript and protein levels of genes in quadrants 1 ( Q1 ) and 3 ( Q3 ) were ‘coupled’ , meaning that the changes in protein levels followed the changes in transcript levels . Q1 and Q3 were enriched in gene products related to sterol biosynthesis and cytoskeletal and cell wall processes , possibly related to cell growth . In contrast , the expression of gene products in quadrants 2 ( Q2 ) and 4 ( Q4 ) were ‘uncoupled’ , meaning that the changes in protein levels did not follow the changes in transcript levels . In Q2 , the proteins were over-represented relative to their transcripts , that is , there were more proteins per transcript in older cells than in younger cells . Of all analyzed transcript–protein pairs , 38 . 4% were located in Q2 , suggesting a global tendency towards relative protein overabundance with aging ( Figure 5 ) . In line with this global protein overabundance , Q4 contained fewer genes and less GO-term enrichments . Strikingly , Q2 was strongly enriched in ‘translation regulation’ gene products ( i . e . ribosome and protein biogenesis machinery ) ( Figure 5B ) , and the extent of their overabundance progressively increased as the cells aged ( Figure 5—figure supplement 2 , 3 ) . Next we asked whether this increased level of biogenesis-related proteins , uncoupled from transcriptional regulation , was causal for downstream effects during replicative aging in yeast . Identifying causality on a systems-wide level is difficult , and the key challenge is to separate cause from downstream effects . However , our dynamically resolved , comprehensive data offered the possibility to suggest causal relationships . To elucidate the causal order of changes during aging , we reconstructed a high-level directional network revealing the interdependences of changes in transcript expression ( Figure 6 , Figure 6—figure supplement 1A ) . Therefore , we defined each transcript’s expression profile as a network node , and an edge between each pair of nodes as a partial correlation between the nodes’ expression profiles ( Figure 6—figure supplement 1B and C ) . Next , we determined the directionality of the edges , indicated by arrows . We defined directionality to represent the ability of a transcript’s profile to predict the profile of another transcript . Concretely , when looking at two connected nodes , the node that could be explained by the connected node was considered as the responsive node , while the predicting node was considered to be the causal node ( Opgen-Rhein and Strimmer , 2007 ) ( Figure 6—figure supplement 1D and E ) . This relation defined the directionality of the edge . Any transcript that had no predictive ability and could not be predicted by any other transcript was removed from the network analysis . Following this , the nodes were clustered by maximizing the global modularity of the network ( Csardi and Nepusz , 2006 ) ( Figure 6A ) . Finally , the clusters were ranked based on the ratio of causal ( outward arrows ) to responsive nodes ( inward arrows ) per cluster to determine the higher-level causal relations existing between clusters . A sensitivity analysis was performed to determine the optimal sparsity of the network and the cut-off for the partial correlation among transcript profiles , through which we established that the network was a robust representation of the datasets ( supplemental note 4 in Supplementary file 1 , Table S7 , Figure 6–source data 1 ) . These steps produced a high-level directional network , in which the ranking of the clusters with respective GO enrichments revealed causal relations during aging ( Figure 6B ) . 10 . 7554/eLife . 08527 . 035Figure 6 . Network inference identifies protein biogenesis-related genes as causal force during aging . ( A ) The directed and clustered transcriptome network consists of 3631 edges , connecting 1241 nodes in 8 clusters ( see Figure 6—figure supplement 1 and supplemental note 4 , in Supplementary file 1 , for further details ) . Only actual relations are depicted , the causal direction between two nodes is indicated with an arrow , where the arrowhead points to the responsive node . ( B ) Clusters ranked from more causal to more responsive in the causality network ( from blue to red for clusters 1 through 8 ) . The degree of causality is determined by the ratio of the outgoing over incoming connections per cluster ( from A ) . The blue to red arrows indicate the sum of outgoing arrows between two clusters , where arrow thickness is logarithmically scaled to the number of arrows ( from A ) , that is , the summed predictive power of one cluster over the other . Terms per cluster are derived from Gene Ontologies ( GO ) as in Figure 3B ( scaling of text: database for annotation , visualization and integrated discovery [DAVID] enrichment score ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03510 . 7554/eLife . 08527 . 036Figure 6—source data 1 . Table S7: The direction matrices and the sensitivity analyses for the proteomic and transcriptomic high-level directional networks . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03610 . 7554/eLife . 08527 . 037Figure 6—figure supplement 1 . The transcriptome network . ( A ) Cartoon illustrating the pipeline of the network analysis procedure , to go stepwise from gene expression time series ( i . e . a gene profile ) towards a high-level causal network . First , only nodes that have related gene profiles ( based on partial correlations ) , as distinguished from all indirectly related gene profiles ( based on simple correlations ) , are connected in the network ( see B below ) . Second , the directionality of the arrows between two nodes was found by accounting for the relative reduction in the variability between the nodes . This revealed a causal relationship ( see D below ) . Third , highly interconnected nodes were clustered . Finally , based on the clusters and the average directionality among the clusters , a high-level directional network was generated . For further details regarding these steps , see below and supplemental note 4 . ( B ) A simulated example to highlight the first step in ( A ) , showing that the edge between nodes in the network depends on the partial correlation between the gene profiles . Two transcript profiles ( ‘y’ and ‘z’ ) were based on a computationally generated transcript profile ( ‘x’ ) , forming a small artificial network with edges between the nodes x and y in addition to x and z . While the simple correlations between all profiles are high ( >0 . 995 ) , the partial correlations are only high for x with y and x with z ( grey dashed lines ) . Therefore , actual relations were only found from x to z and x to y ( black edge ) . We can thus retrieve the true network , by making use of the partial correlations . ( C ) The undirected network for the transcriptome data . The edges between the nodes indicate only actual relations ( based on partial correlations ) between transcript profiles . All edges connected without partial correlations or nodes linked to the dataset without a partial correlation are omitted in this network . ( D ) An example to highlight the second step in ( A ) that the directionality between two transcript profiles was found by multiple testing of the standardized partial variances of the nodes . The standardized partial variances are the variances once the effect of the related profiles has been removed by regression analysis . For each of the connected node pairs ( e . g . ‘m’ and ‘n’ ) , the direction goes from the profile with the highest standardized partial variance to the lowest . Basically , for a profile with a lower standardized partial variance , much of its variability is explained by the profiles connected to it , while for a profile with a high standardized partial variance , less of its variability is explained by the profiles associated to it . The latter profile has thus a higher ability to predict the first one than vice versa , and makes a profile with high standardized variance causal over a profile with a low standardized variance . The directionality is indicated as an arrow between the nodes . ( E ) The directed network for the transcriptome data . The arrowhead is pointing to the responsive node . For the clustered directed network see Figure 6A and for the high level directional network see Figure 6B . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03710 . 7554/eLife . 08527 . 038Figure 6—figure supplement 2 . Network cluster gene enrichments in the co-expression map . ( A ) The genes represented in cluster 1 of the transcriptome networks ( blue dots ) were mapped on the co-expression map ( gray dots; Figure 5B ) . The percentage of the genes enriched in each of the four quadrants ( Q1–4 ) is indicated , fold changes are plotted on a log2 scale ( B ) p-values for the enrichment of the genes in each cluster of the network in the four quadrants; transcriptome ( top ) and proteome ( bottom ) . ( C ) The p-value for the enrichment of genes in each cluster in Q1 and Q3 together representing a ‘coupled’ change in protein and transcript levels ( left panel ) , and in quadrant Q2 and Q4 ( uncoupled change ) ( right panel ) . A shift towards an uncoupled phenotype in the ‘later’ network clusters is apparent . The p-values are plotted on a log10 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 038 This high-level directional network of the transcriptome data showed that the very first causal-ranked cluster in the network that we detected was highly enriched for gene products associated with protein biogenesis ( i . e . ribosome biogenesis and transfer RNA [tRNA] processing; Figure 6B ) . These are the same biological processes that had uncoupled transcript and protein levels ( Figure 5B ) ; indeed , genes from this causal cluster were enriched in Q2 of the co-expression map , which showed uncoupled expression ( Figure 6—figure supplement 2A and B ) . These analyses suggest that the uncoupling of protein and transcript levels for ‘biogenesis’-related genes has a central role in the aging process , and may affect the transcript and protein abundances of other genes , as elaborated upon in the discussion . The overabundance of proteins relative to transcripts must have consequences for cellular functioning . Protein overproduction could increase cell size , one of the first hallmarks described in yeast aging ( Egilmez et al . , 1990 ) . Increased cell size could reduce glucose influx rates per cell volume and induce metabolic changes , for example , at low rates of glucose influx , cells switch to respiration ( Huberts et al . , 2012 ) . Indeed , in our transcript-based network analysis ( Figure 6B ) as well as in our proteome dataset ( Figure 3B ) , we found that metabolic signatures related to starvation and oxidative stress were consequences of aging . Furthermore , we hypothesized that if protein levels become globally uncoupled from their transcript levels during aging ( Figure 5 ) , the optimal stoichiometry of proteins in complexes may be perturbed ( Figure 7A ) . Indeed , using curated lists of protein complexes ( Cherry et al . , 2012 ) , we found that an increased deviation from the original stoichiometry occurred with aging ( Figure 7B–D , and Figure 7—figure supplement 1–3 ) . We observed many complexes that were not previously implicated in aging to be age-affected , and we found previously implicated protein complexes such as the vacuolar adenosine triphosphatase ( Hughes and Gottschling , 2012 ) and the nuclear pore complex ( Lord et al . , 2015; Denoth-Lippuner et al . , 2014 ) to lose stoichiometry ( Figure 7C and D and Figure 7—figure supplement 1 , 2 ) . The global stoichiometry loss was greater in aged mothers compared with the daughter population ( Figure 7—figure supplement 3A ) , confirming that this is an aging-related phenotype . Additionally , we found that the stoichiometry loss was greater overall at the proteome level than at the transcriptome level ( Figure 7B ) , supporting the observation that protein levels uncouple from their transcript levels . 10 . 7554/eLife . 08527 . 039Figure 7 . Loss of stoichiometry in protein complexes is a consequence during aging . ( A ) Illustrative representation of loss of stoichiometry within a protein complex during aging . Changing levels of proteins may be coordinated ( left ) or uncoordinated and result in a loss of complex stoichiometry ( right ) . ( B ) Stoichiometry loss ( for a single complex defined as the InterQuartile Range ( IQR ) of the distribution of fold changes of the components ) is plotted for all complexes in proteome and transcriptome datasets as bean plots during aging . Thick horizontal line represents the mean of the distribution of all complexes , thin colored lines the individual complexes’ stoichiometry loss , and the outline the distribution of all complexes . The genes in common between the proteome and transcriptome datasets are used . ( C ) Illustration of the loss of stoichiometry of protein complexes during aging for the proteome ( gray lines ) , with specific examples highlighted ( colored lines ) . ( D ) Illustration of the loss of protein stoichiometry in proteasome ( left panel ) and the vacuolar proton transporting V-type adenosine triphosphatase ( ATPase ) , V1 domain ( right panel ) . The protein abundance changes ( log2 scale ) of the complex’ components are plotted in time . The degree of stoichiometry loss is indicated with a box plot . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 03910 . 7554/eLife . 08527 . 040Figure 7—figure supplement 1 . Proteome data of distribution of changes within complexes in the cell . A curated list of protein complexes derived from the ‘cellular component’ gene ontology was downloaded from yeastgenome . org , and the horizontal box plots show the distribution of fold changes ( log2 scale ) occurring in the complex when comparing proteome data of the old ( 72 hr ) sample to the young reference sample . Box-and-whisker plots are presented as follows: the thick black line within the box is the median of the data , the box extends to the upper and lower quartile of the dataset ( i . e . to include 25% of the data above and below the median , respectively ) , whiskers ( dashed lines ) represent up to 1 . 5 times the upper or lower quartiles and circles represent outliers . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 04010 . 7554/eLife . 08527 . 041Figure 7—figure supplement 2 . Transcriptome data of distribution of changes within complexes in the cell . Same as Figure 7—figure supplement 1 but for the transcriptome data . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 04110 . 7554/eLife . 08527 . 042Figure 7—figure supplement 3 . Loss of stoichiometry occurring in the protein complexes . ( A ) Comparison between mother cells and mother-age dependent daughter cells , loss of stoichiometry within complexes . Bean plots showing the distribution of the loss of stoichiometry for all complexes in the cell ( same as in Figure 7B ) , at each time point throughout aging . Mother and daughter cells plotted side by side , for the proteomes ( left panel ) and transcriptomes ( right panel ) , showing that the mother cells’ proteome undergoes a greater degree of loss of stoichiometry within complexes than do mother-age dependent daughter cells . Stoichiometry loss for a single complex is calculated as the interquartile of the distribution of fold changes within the complex at any given time ( i . e . the ‘box’ in Figure 7—figure supplement 1 and 2 ) . Bean plots are drawn as follows: thick horizontal line represents the mean of the distribution of all complexes , thin colored lines the individual complexes’ stoichiometry loss , and the outline the distribution of all complexes . ( B ) Illustration of the loss of protein stoichiometry in the vacuolar proton transportin V-type adenosine triphosphatase ( ATPase ) , V1 domain . The protein abundance changes ( log2 scale ) of the complex’ components are plotted in time . The degree of stoichiometry loss is indicated with a box plot . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 04210 . 7554/eLife . 08527 . 043Figure 7—figure supplement 4 . The proteome network . ( A ) Undirected , directed , and clustered directed networks for the proteome dataset . The clustered directed network consists of 669 edges , connecting 493 nodes in 5 clusters . ( B ) These interactions are summarized in a causal network: clusters are ranked from more causal to more responsive ( from blue to red for clusters 1 through 5 , placed on a turquoise arrow that depicts ranking ) in the causality network . The degree of causality is determined by the ratio of the causal outgoing over incoming connections per cluster ( from A ) . The blue to red arrows indicate the sum of outgoing arrows between two clusters ( from A ) , that is , the summed predictive power of one cluster over the other . Terms per cluster are derived from Gene Ontologies ( GO ) as in Figure 3B ( scaling of text: DAVID enrichment score ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08527 . 043 Being built of fewer genes ( 1494 proteins versus 4904 transcripts ) , the high-level directional network of the proteome was less revealing than that of the transcriptome ( Figure 7—figure supplement 4 ) . The most causal cluster of the proteome network was enriched for chaperone proteins , reflecting a cellular response to internally changing conditions . Such conditions could include metabolic restructuring in response to an increased cell size or to aggregating proteins that are accumulating due to altered protein complex stoichiometry . Furthermore , we found that the causal clusters of the proteome network tended to be expressed according to their transcriptional message ( i . e . coupled expression; Q1 , Q3 ) , while the responsive clusters represented increasingly uncoupled expression ( Q2 , Q4 ) ( Figure 6—figure supplement 2C ) . This both confirmed the response of the cell to the accumulating changes occurring during aging and indicated that the effects of uncoupled protein expression are progressive over time . We see the clear downstream consequences during aging emerging in the proteome , including metabolic shifts , stoichiometric loss , aggregating proteins , and protein overproduction . All of these point to pathways and processes that may become dysfunctional with aging , any of which may ultimately result in cell death .
Using our newly developed culturing and computational methods and state-of-the-art proteomics and transcriptomics analyses , we generated the first systems-level molecular phenotype of replicatively aging yeast . The comprehensiveness of the data allowed us to discover that protein biogenesis machinery genes , including ribosome , tRNA synthesis , and translation regulation genes , have their protein levels uncoupled from their mRNA levels during aging ( Figure 5B ) . Furthermore , the dynamic nature of the data allowed us to pinpoint the transcripts of these genes as having the strongest ability to predict the behavior of others transcripts during aging ( Figure 6B ) . Finally , we observed metabolic changes , protein stress responses , and changes in the stoichiometry of many protein complexes ( Figure 3B , 4 , 7B ) . Based on these analyses , we propose a model whereby the uncoupling of protein levels of biogenesis-related genes from their transcript levels is causal for the changes occurring in aging yeast . The model proposes that proteins of the translation machinery that are uncoupled from transcript levels accumulate in cells with age ( Figure 5B ) . As the biogenesis genes are themselves involved in translation , their uncoupling might contribute to further uncoupling of the proteome from the transcriptome as a whole . This general uncoupling has degenerative effects ( i . e . cell size increase , protein aggregations and loss of stoichiometry in protein complexes ) , that stimulate transcriptional responses in the cell ( i . e . metabolic changes and activated stress responses ) , which further contributes to changes in the proteome . Although we cannot exclude the possibility of other causes even further upstream , the uncoupling of the protein biogenesis machinery is likely an early driver of replicative aging in yeast . A question remains as to why the biogenesis-related class of proteins we identified as having protein levels uncoupled from their transcript levels become over-represented in replicatively aging yeast in the first place . Ribosome footprinting has shown these proteins to be highly translated ( Ingolia et al . , 2009 ) , and protein turnover experiments have shown them to be highly stable ( Belle et al . , 2006 ) ; thus , it is possible that their overabundance may result from the combination of the dynamics of protein biogenesis , protein turnover , and mRNA stability . Interestingly , the ribosomal proteins themselves showed a low degree of loss of stoichiometry at the protein complex level in our data ( Figure 7C ) , supporting the idea that they are still active and contributing to uncoupling in the cell . In any case , the uncoupling of protein and transcript levels has downstream consequences for the cell that may explain many phenotypes of aging . First , cell size may increase due to protein overproduction and result in metabolic changes . Second , proteins being overproduced at different rates will alter protein complex stoichiometry . Many documented phenotypes of aging may result from this , including the formation of protein aggregates ( Erjavec et al . , 2007 ) , increased reactive oxygen species formation by a dysfunctional mitochondrial transport chain ( Laun et al . , 2001 ) , and loss of gene silencing ( Hu et al . , 2014 ) . The sum of these may ultimately lead to system failure for the organism . Directly targeting certain failing protein complexes or downstream deleterious effects results in replicative lifespan extension , but we suggest that many of these effects will prove to be cell type- and growth condition-specific . Our model predicts that a more robust extension of lifespan may be possible in many organisms by targeting the causal factor in aging , protein biogenesis . Indeed , altering the rates of protein production ( i . e . translation ) or degradation ( i . e . autophagy ) have repeatedly been shown to influence longevity across a wide range of organisms ( see ( Wasko and Kaeberlein , 2014; Johnson et al . , 2013; Cuervo , 2008 ) . The translation activators , target of rapamycin ( TOR ) and S6 kinase , fall into this category , and decreases in their activity result in increased lifespan in yeast ( Fabrizio et al . , 2001; Kaeberlein et al . , 2005 ) , worms ( Vellai et al . , 2003; Pan et al . , 2007 ) , flies ( Kapahi et al . , 2004 ) , and mice ( Lamming et al . , 2012; Selman et al . , 2009 ) , as does calorie restriction and drugs such as rapamycin , which are also modulators of protein biogenesis pathways ( Johnson et al . , 2013 ) . Likewise , deletions in ribosomal subunit components have positive effects on lifespan in both yeast ( Steffen et al . , 2008 ) and worms ( Hansen et al . , 2007 ) . Our model suggests why these interventions and mutations have a lifespan-extending effect in a broad spectrum of organisms , namely because protein biogenesis machinery is itself a driver of aging . | Aging is a complex process , and so many scientists use baker’s yeast as a simpler model to understand it . Although many genes that influence aging have been found , all the generated knowledge is still rather fragmented . It also remains difficult to disentangle cause and consequence . That is to say , sometimes a gene that looks like it might cause aging could simply be a gene that responds to an age related phenomenon . To unravel this puzzle of cause and effect , it is necessary to first get an idea on a system level of everything that changes as an organism ages . Now , Janssens , Meinema et al . have managed to map many of the molecular changes that occur as baker’s yeast ages; this is something that has yet to be achieved for any other organism . The work first involved developing a new way of growing baker’s yeast to keep and generate large cohorts of aging yeast cells in a constant environment . It also required the use of a mathematical ‘un-mixing’ tool to separate the data obtained from the aging cohort from the data from the young offspring that the yeast produce while they age . Janssens , Meinema et al . measured both the majority of the transcriptome and much of the proteome of baker’s yeast throughout its reproductive lifespan . The “transcriptome” refers to the collection of RNA molecules in the cell , which are produced whenever a gene is expressed . The “proteome” refers to all the proteins in the cell , which are translated from the RNA transcripts by the cell’s so-called “translational machinery” . These experiments revealed that this yeast’s proteome reflects its transcriptome less and less as it ages . In particular , this ‘uncoupling’ of the proteome from the transcriptome was seen most strongly for the proteins related to the cell’s translational machinery; these proteins accumulated with age relative to their transcripts . Janssens , Meinema et al . then conducted a computational network-based analysis of the data . This indicated that the uncoupling is the driving force behind the aging process . Many of the other molecular changes that occur with aging were predicted to be consequences of this uncoupling . These findings give a framework for many observations in the existing literature . However , it remains unclear why proteins related to translational machinery are overrepresented in aging yeast in the first place . This question should be explored in future work . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2015 | Protein biogenesis machinery is a driver of replicative aging in yeast |
The importance of cultural processes to behavioural diversity in our closest living relatives is central to revealing the evolutionary origins of human culture . However , the bonobo is often overlooked as a candidate model . Further , a prominent critique to many examples of proposed animal cultures is premature exclusion of environmental confounds known to shape behavioural phenotypes . We addressed these gaps by investigating variation in prey preference between neighbouring bonobo groups that associate and overlap space use . We find group preference for duiker or anomalure hunting otherwise unexplained by variation in spatial usage , seasonality , or hunting party size , composition , and cohesion . Our findings demonstrate that group-specific behaviours emerge independently of the local ecology , indicating that hunting techniques in bonobos may be culturally transmitted . The tolerant intergroup relations of bonobos offer an ideal context to explore drivers of behavioural phenotypes , the essential investigations for phylogenetic constructs of the evolutionary origins of culture .
Humans and other social animals exhibit a diversity of behavioural phenotypes attributed to genetic or social ( i . e . , cultural ) evolutionary processes , and their combination , influenced by the environment ( Allen , 2019; van Schaik et al . , 2003; Whitehead et al . , 2019; Whiten , 2017 ) . While culture is identified as a pivotal selective process in human evolution ( Boyd and Richerson , 1995; Whitehead et al . , 2019 ) , its relative contribution to shaping the behavioural diversity observed in non-human animals , including our closest living relatives , remains debated . For instance , in comparison to the other great ape species , little is known about potential cultural traits in bonobos ( Pan paniscus ) ( Whiten , 2017 ) , thereby limiting phylogenetic comparisons . Culture is defined as group-specific behavioural patterns acquired through social learning ( Laland and Janik , 2006 ) . There is ample evidence that some foraging techniques are socially learned ( e . g . , primates [Whiten and van de Waal , 2018; cetaceans [Mann et al . , 2012; carnivores [Thornton and Raihani , 2008] ) and therefore represent good candidates for cultural traits . However , to distinguish whether social processes contribute to the emergence of behavioural phenotypes , it is essential to quantify ecological variation and account for its influence on behaviour expression , a challenging endeavour in wild settings . Few studies have attempted to limit potential ecological confounders by investigating behavioural diversity between neighbouring groups ( Luncz and Boesch , 2014; Pascual-Garrido , 2019; van de Waal , 2018 ) . Nonetheless , in the absence of between-group range overlap , fine-scale ecological variation specific to the locations where behavioural phenotypes are expressed cannot be excluded . Our closest living relatives , bonobos and chimpanzees , hunt a variety of species across groups and populations ( Gilby et al . , 2015; Hobaiter et al . , 2017; Hohmann and Fruth , 2008; Sakamaki et al . , 2016; Samuni et al . , 2018; Wakefield et al . , 2019 ) . However , it remains unclear whether this diversity is independent of large or even small-scale ecological variation in the distribution of prey species ( Hobaiter et al . , 2017; Sakamaki et al . , 2016 ) . Accounting for potential small-scale local ecological drivers is methodologically challenging in chimpanzees , a territorial species ( Mitani et al . , 2010; Samuni et al . , 2017 ) where each group predominantly occupies unique non-overlapping areas . In contrast , the tolerant intergroup relations of bonobos ( Furuichi , 2020 ) permit a context in which different behaviours are expressed by individuals of different groups in the same place and at the same time . Here , we investigate variation in bonobo predation patterns of two groups ( Ekalakala and Kokoalongo ) at the Kokolopori Bonobo Reserve . The groups share an extensive home range overlap ( 65% kernel overlap; Figure 1 , A , B , C ) and regular gene flow , thereby reducing ecological and genetic influences as an explanatory variable for intergroup differences in behavioural expressions ( van de Waal , 2018 ) . Specifically , we tested whether variation in prey preference between the two bonobo groups is explained by a ) environmental variables , such as area usage and seasonality , and/or b ) social factors , such as the number of potential hunters , individual association pattterns , and group identity .
Between August 2016 and January 2020 , we observed 59 successful captures and consumption of mammals by the bonobos , including anomalure , duiker , and squirrel species ( Table 1; Figure 1—figure supplement 1; Video 1 ) . Starting July 2019 , we also collected data on unsuccessful hunts , and documented 11 hunt attempts on duiker and anomalure ( duiker- NEkalakala = 2 , NKokoalongo = 2; anomalure- NEkalakala = 4 , NKokoalongo = 3 ) . Overall , we observed all Ekalakala and 84% of Kokoalongo adult group members ( 100% if considering only individuals that were present for the entire study period ) participating in hunts . Most anomalure and duiker hunts occurred within overlapping ranging areas ( 94% of anomalure and 83% of duiker hunts ) , compared to only 46% of squirrel hunts ( Figure 1 , A , B , C ) . The groups engaged in frequent and prolonged intergroup associations ( 31% of observation days ) , and nine of the hunts ( five duiker , three anomalure , one squirrel ) occurred during intergroup encounters and at times involved between-group meat sharing . Although 45% of the Kokoalongo duiker hunts occurred during encounters , very few to none ( mean = 1 . 4 ) of the Ekalakala individuals were present during these hunts , and none participated ( Supplementary file 1 ) . Due to the cohesiveness of bonobo groups ( Hohmann and Fruth , 2002 ) , the conspicuous nature of anomalure and duiker hunting ( e . g . , distress calls of duikers ) , and since the acquisition of meat often attracts individuals to hunting areas ( Samuni et al . , 2018 ) , we are confident that we observed most anomalure and duiker feeding events . However , as the hunting and feeding of squirrel is often quiet and solitary and since hunting is frequently detected only post capture , we are likely to have underestimated this type of hunting . Kokoalongo bonobos were more likely to capture duiker ( estimate = 4 . 56 , CI95% = [1 . 93 , 8 . 03]; Figure 1D , Table 2 ) and squirrel species ( estimate = 4 . 99 , CI95% = [2 . 34 , 8 . 21] ) , and were less likely to capture anomalure species in comparison with Ekalakala . The same pattern persisted during intergroup encounters ( once we observed anomalure captured by a Kokoalongo female after a hunt by Ekalakala individuals; Supplementary file 1 ) . We found that prey preferences were independent from potential local spatial and temporal ecological variation . Overall , more than 80% of all hunts occurred in overlapping areas ( 95% kernel ) , and neither utilization differences of specific hunt locations ( reflecting varying opportunities to encounter prey species ) nor potential annual seasonal variation strongly affected phenotypic variation in prey types captured ( Table 2 ) . Variation in prey preference can also arise from between-group difference in sizes of female or male association parties , association tendencies amongst party members , or presence of certain specialized hunters . However , the number of adult females or males present during hunts ( i . e . , available hunters ) and the average dyadic association between them had no strong effect on prey outcome ( Table 2 ) . Further , we observed 17 different individuals ( five males and 12 females ) catching prey , encompassing 72% of Ekalakala and 40% of Kokoalongo group members ( see Supplementary file 1 for the distributions of catchers ) . These percentages are likely an underestimation of the overall number of individuals who captured the prey , as their identity was not recorded for 40% of all hunts . Finally , our results are likely independent from genetic variation , as low genetic differentiation is expected ( Schubert et al . , 2011 ) mainly due to regular gene flow attributed to female migration between Ekalakala and Kokoalongo .
We found that bonobo groups that utilize overlapping home ranges and regularly socialize and forage together show group-specific prey acquisition patterns . These group-specific patterns appear independent of genetic and small-scale ecological variation , seasonality , size of hunting parties , or party cohesiveness . The exclusion of these confounders indicates that other drivers of behavioural variation act as mechanisms in prey selection . Observed differences in prey preferences may arise if different techniques are required to locate and capture them . Duiker and squirrel hunting are either strictly terrestrial ( duiker ) or arboreal ( squirrel ) activities , which appear opportunistic and commonly involved a single individual hunter ( more so for squirrel hunting ) . Conversely , anomalure hunting required the engagement of several group members , during which the bonobos employed both terrestrial and arboreal positions . While at this stage it is unclear if hunting techniques in bonobos require time to acquire or involve social learning processes , specialized hunting techniques may be at the basis of the observed group differences . Prey species preference may additionally reflect differences in prey palatability between groups . Although between-group meat sharing of duiker and anomalure may contradict the idea of group specific meat preference , the costs and benefits associated with hunting relative to begging potentially alter consumption decisions . As hunting behaviour is associated with energetic costs , the benefit of capturing favourable prey may persuade hunt decision making . Conversely , once prey is captured , the costs associated with begging are minimal relative to hunting , thereby largely resetting the cost-to-benefit ratio behind foraging decisions . Thus , while palatability may dictate which prey species to pursue , it is expected to have a lesser impact on begging decisions . The ‘impact hunter’ hypothesis ( Gilby et al . , 2015 ) could offer an alternative explanation for prey preference variation , proposing that certain individuals encourage social hunts by assuming hunt initiation costs . However , as this hypothesis addresses social hunt occurrence , it could explain the prevalence of social hunts like anomalure but cannot explain why duiker and squirrel hunting ( opportunistic and largely solitary ) are nearly absent in Ekalakala . Further , we observed many individuals participating in hunts and capturing prey and prey outcome was independent of the number of male or female hunters . Thus , patterns in our data indicate that we indeed document group , instead of individual , tendencies . In the absence of ecological , genetic , or ingroup social dynamic explanations of prey acquisition , the observed group-specific differences may be cultural . Under this assumption , it is puzzling how such group differences would evolve and persist even when prolonged associations between Ekalakala and Kokoalongo should potentially promote intergroup social learning opportunities . Tolerance , at a degree that facilitates social learning in its various forms , is fundamental in converting innovations into transmitted traditions ( Whiten and van de Waal , 2018 ) . To improve ‘learning’ gains , social learners should be selective in the timing of observations and their choice of ‘models’ from whom to learn ( Boyd and Richerson , 1995 ) . Although the two groups associate for extended periods their intergroup relations are complex and unpredictable , characterized by a mixture of affiliative and agonistic exchanges , frequent fission-fusions and heightened arousal . Unpredictability of intergroup interactions is thus expected to hamper intergroup learning opportunities of certain skills which may require extensive time and effort to acquire ( e . g . , hunting techniques ) . Following group psychology predictions of ingroup bias and favouritism ( Brewer , 1993 ) , outgroup members may as well be less appealing ‘models’ for learning . Together , inconsistent intergroup relations and in-group bias may explain how group-specific prey preferences persist despite numerous intergroup learning opportunities . A by-product of divergent hunting techniques is reduced intergroup competition , which is likely adaptive , especially when groups share ranging zones . Thus , group-specific prey preferences in bonobos may have evolved as a form of microlevel niche differentiation that alleviates feeding competition . Investigating the potential impact of culture on behavioural diversity in non-human animals is challenging due to the difficulties of estimating and accounting for local ecological variation as a driver of behavioural diversity . Challenges may even arise when behavioural variation appears between groups that occupy nearby but non-overlapping ranging areas . Bonobo social groups’ regular overlap in ranging area and tolerant interactions , offer fertile ground in which to explore whether variation in behavioural expressions occurs independently of spatial and temporal use of specific habitat locations . Here , by accounting and largely excluding potential local ecological variation , we provide strong indication for culturally transmitted subsistence hunting techniques in bonobos , informing on the evolution of behavioural diversity .
We investigated behavioural diversity between two fully habituated bonobo groups ( Ekalakala and Kokoalongo , followed since 2007 ) at the Kokolopori Bonobo Reserve , Democratic Republic of Congo ( N 0 . 41716° , E 22 . 97552°; [Surbeck et al . , 2017a] ) . We conducted full day party follows of the bonobo groups ( 1102 and 931 observation days in Ekalakala and Kokoalongo , respectively ) and documented all occurrence hunting behaviour ( here defined as capture of mammalian prey ) . All prey types were captured across most months , and both during the dry ( June-August and December-February ) and wet ( March-May and September-November ) seasons . Hunt participants were almost exclusively adult ( >10 years ) individuals , and both sexes were observed to participate . Adult group sizes fluctuated during the study between 9–11 adult individuals in Ekalakala and 16–24 adult individuals in Kokoalongo due to several deaths and migration events ( Supplementary file 2 ) . We recorded data on party locations at one-minute intervals using a GPS ( Garmin 62 ) . We constructed home range utilization distributions of the bonobo groups using kernel density estimates ( Worton , 1989 ) . The home range ( 95% kernel ) of the two groups between August 2016 and December 2019 was: Ekalakala – 35 km2 , Kokoalongo – 40 km2 , and the overlapping area encompassed 64% and 66% of the home ranges of Ekalakala and Kokoalongo , respectively . Habitat structure and spatial distribution of prey species have been used as explanations for variation in hunting behaviours ( Hobaiter et al . , 2017; Sakamaki et al . , 2016 ) . However , as our data originate from two groups with extensive home range overlap , the explanatory power of these drivers is minimized . Nonetheless , we can evaluate intra-range variation in local ecology by accounting for relative home range usage across the groups . To do so , we assigned each hunt with two kernel usage values , one constituting the kernel usage of the group that hunted ( hunt group ) and the other constituting the kernel usage of the group that did not hunt ( other group ) . We used the values to calculate a score of ‘usage difference’ ( i . e . , other group - hunt group; ranging between −50 and 86; mean ± sd: 20 . 19 ± 26 . 10 ) . Higher scores reflected an area that is more predominantly used by the group that hunted . We recorded the cumulative adult party composition at 30 min intervals and marked individuals observed during the hunt scan as potential hunters . Whenever a party composition scan collected either immediately before or during a hunt included individuals of both groups ( representing between-group spatial proximity ) , that hunt was marked as occurring during an intergroup encounter . This approach categorized two hunts as intergroup hunts although members of only one group were present , but accounts for the likelihood that the other group is nearby . We used these party scans to calculate dyadic association values for each dyad and year , using the following equation: SRI = PAB/ ( PA + PB - PAB ) ( Surbeck et al . , 2017b ) . PA and PB represent the number of scans A or B were present , and PAB represents the number of scans both A and B were present . For every hunt , we then calculated the average dyadic association of the hunting party as a proxy of group social cohesion , which may affect the likelihood to capture prey . We applied a Bayesian Regression model with prey type as a categorical response and logit link function to examine the influence of environmental ( area usage and seasonality ) and social ( group identity , presence of potential hunters , and social cohesion ) factors on prey preference expression . We fitted the model in R ( version 3 . 6 . 1 [R Development Core Team , 2016] ) using the function brm of the R package ‘brms’ ( Bürkner , 2017 ) and weakly informative t-distributed priors ( Lemoine , 2019 ) . As predictors , we included the following environmental factors: a ) ‘usage difference’ score as described above , and b ) a seasonal temporal term , by including the sine and cosine of the Julian dates of the hunts converted into a continuous circular variable ( Stolwijk et al . , 1999 ) . The sine and cosine predictors allow for the modelling of a wave like periodic pattern of peaks and valleys , thereby representing potential seasonal oscillations in hunt dates . Additionally , we included the following social factors: a ) group identity of the individual who caught the prey , b ) female and male party sizes ( mean ± sd: Ekalakala - 7 . 19 ± 1 . 47; Kokoalongo – 7 . 05 ± 3 . 62; encounter - 13 ± 7 . 4 ) , and c ) average dyadic associations of hunt party mean ± sd: Ekalakala - 0 . 51 ± 0 . 09; Kokoalongo – 0 . 34 ± 0 . 13; encounter – 0 . 26 ± 0 . 14 ) . Note , if dietary requirements alone were to dictate hunting patterns , then we would expect a random distribution ( reflecting prey species encounter probabilities ) of the different prey species captured within groups instead of group-specific patterns . We ran 2000 iterations over four MCMC chains , with a ‘warm-up’ period of 1000 iterations per chain leading to 4000 usable posterior samples ( Bürkner , 2017 ) . Visual inspection of all MCMC results revealed satisfactory Rhat values ( <1 . 01; [Gelman et al . , 2013] ) , no divergent transitions after warmup , and stationarity and convergence to a common target , suggesting that our results are stable . We report the estimate ( mean of the posterior distribution ) and the 95% credible intervals ( CI95% ) indicating the strength of the effects . For estimate comparability and to ease model convergence , we standardized all numeric variables to mean = 0 and sd = 1 . Our model did not suffer from issues of collinearity , evaluated using Variance Inflation Factors ( Field et al . , 2012 ) with the R package ‘car’ ( Fox et al . , 2020 ) . The data reported in this paper are available as Source data 1 . | No human culture is quite like the next . Societies around the world show exceptional variety in their social norms , beliefs , customs , language and , of course , food . However , the origins of human culture still remain elusive . Studying humans’ closest living relatives , the great apes , is one way to explore how human culture first appeared . Chimpanzees are often studied for this purpose , but other great apes , such as bonobos , are often overlooked . Yet bonobos are less territorial and more tolerant to others than chimpanzees , with different bonobo groups sharing feeding spots and hunting grounds . These traits actually make bonobos an ideal animal for investigating whether differences in group behaviour , such as feeding habits , are distinct cultural trends or just a result of their surrounding environments . With this in mind , Samuni et al . studied the hunting and feeding patterns of two groups of wild bonobos in the Kokolopori Bonobo Reserve in the Democratic Republic of Congo . The two groups share approximately 65% of their home territory , allowing Samuni et al . to examine whether any differences in hunting preferences persisted when the two groups looked for prey in the same environment . The analysis would reveal whether social factors or environmental conditions influenced the hunting and feeding habits of each group . Samuni et al . found the first bonobo group specialized in hunting duiker , a type of antelope , whereas the second group preferred to hunt tree-gliding rodents . However , the location and timing of the bonobo’s hunts did not determine which types of prey they hunted . Across their territory , and regardless of group size or the dynamics between males and females , the groups continued to hunt their preferred prey . This means ecology alone cannot explain bonobo feeding habits and instead , the findings provide a strong indication for cultural variation between the two groups . Since social learning is a part of cultural development , the next challenge will be to determine if and how these group hunting preferences are learned by young bonobos in their social group . For now , these findings provide a glimpse into the emergence of group culture . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"short",
"report"
] | 2020 | Behavioural diversity of bonobo prey preference as a potential cultural trait |
Synaptic vesicles ( SVs ) fuse at active zones ( AZs ) covered by a protein scaffold , at Drosophila synapses comprised of ELKS family member Bruchpilot ( BRP ) and RIM-binding protein ( RBP ) . We here demonstrate axonal co-transport of BRP and RBP using intravital live imaging , with both proteins co-accumulating in axonal aggregates of several transport mutants . RBP , via its C-terminal Src-homology 3 ( SH3 ) domains , binds Aplip1/JIP1 , a transport adaptor involved in kinesin-dependent SV transport . We show in atomic detail that RBP C-terminal SH3 domains bind a proline-rich ( PxxP ) motif of Aplip1/JIP1 with submicromolar affinity . Pointmutating this PxxP motif provoked formation of ectopic AZ-like structures at axonal membranes . Direct interactions between AZ proteins and transport adaptors seem to provide complex avidity and shield synaptic interaction surfaces of pre-assembled scaffold protein transport complexes , thus , favouring physiological synaptic AZ assembly over premature assembly at axonal membranes .
The primary function of the presynaptic active zone ( AZ ) is to regulate the release of neurotransmitter-filled synaptic vesicles ( SVs ) in response to action potentials entering the synaptic bouton ( Südhof , 2012 ) . Before AZ scaffold components ( e . g . , ELKS family protein Bruchpilot: BRP , Rab3-interacting molecule ( RIM ) -binding protein: RBP ) are integrated into synapses , however , they have to be transported down the often very long axons . AZ scaffold proteins are characterized by strings of interaction motifs ( particularly coiled coil motifs ) contributing to the avidity and tenacity of synaptic scaffolds ( Tsuriel et al . , 2009 ) . Therefore they might be considered as ‘sticky cargos’ whose association status has to be precisely controlled during transport . Long-range axonal transport is conducted along polarised microtubules , using kinesin-family motor proteins for anterograde and dyneins for retrograde transport ( reviewed in Maeder et al . , 2014 ) . Kinesin-1 family motor kinesin heavy chain ( KHC , also known as KIF5; Sato-Yoshitake et al . , 1992; Hurd and Saxton , 1996; Takamori et al . , 2006 ) and Unc-104/Imac/KIF1 ( Hall and Hedgecock , 1991; Pack-Chung et al . , 2007 ) have been implicated in the transport of SVs , in conjunction with regulators of this process , such as Syd-1 ( Hallam et al . , 2002 ) , Syd-2/Liprin-α ( Serra-Pagès et al . , 1998; Zhen and Jin , 1999; Miller et al . , 2005; Stryker and Johnson , 2007; Wagner et al . , 2009 ) , RSY-1 ( Patel and Shen , 2009 ) , or ARL-8 ( Klassen et al . , 2010; Wu et al . , 2013 ) . In Caenorhabditis elegans , SV and AZ scaffold proteins exhibit extensive co-transport and undergo frequent pauses , with immobile phases promoting cargo dissociation and assembly ( Wu et al . , 2013 ) . Long axons , typical for Drosophila or mammals , pose high demands for the ‘processivity’ of axonal AZ scaffold component transport . The molecular mechanisms , which provide this processivity and thus block premature assembly processes remain speculative , but might also be relevant in the context of axonal transport deficits of neurodegenerative scenarios ( Millecamps and Julien , 2013 ) . In addition , we know little concerning the composition of cargos destined for synaptic AZs . The electron-dense AZ cytomatrix ( T-bar ) at the Drosophila neuromuscular junction ( NMJ ) is among others composed of oligomers of BRP and RBP ( Kittel et al . , 2006; Fouquet et al . , 2009; Liu et al . , 2011a; Ehmann et al . , 2014 ) . We report here that BRP and RBP , but no other tested AZ components , are co-transported in discrete transport complexes along the axon . Via a screen for RBP interaction partners , we identified the APP-like protein interacting protein 1 ( Aplip1 ) , an adaptor protein previously implicated in SV transport . Further analysis by X-ray crystallography and calorimetry showed that the second and third Src homology 3 ( SH3 ) domain of RBP bind a specific N-terminal proline-rich ( PxxP ) motif of Aplip1/JIP1 with more than 10-fold higher affinity than RBP binds its synaptic ligands ( Ca2+channels/RIM ) by their cognate PxxP motifs . The integrity of this motif was essential to protect axons from forming ectopic axonal synapses , which were observed in aplip1 mutant axons by electron microscopy ( EM ) and super-resolution light microscopy . In summary , we characterize a mechanism of axonal AZ protein transport through a high affinity interaction between preassembled , stoichiometric scaffold protein complexes and the transport adaptor Aplip1 . This high affinity interaction is needed to allow for effective axonal transport and to protect from premature AZ assembly processes .
Firstly , we chose a previously characterized mutant of a serine–arginine ( SR ) protein kinase at location 79D ( srpk79D ) . The SRPK79D protein is a member of the serine–arginine protein kinase family previously shown to be involved in mRNA splicing and processing ( Wang et al . , 1998 ) . Mutants of srpk79D form dramatic BRP aggregates in the axoplasm , while its endogenous substrates remain elusive ( Johnson et al . , 2009; Nieratschker et al . , 2009 ) . The axonal aggregations here served as a sensitive background to screen for proteins that co-accumulate together with BRP in the axon , and therefore indicate a joint transport mechanism . In order to visualise the aggregates forming within axons of srpk79D mutant larvae , we stained with antibodies ( Abs ) directed against the BRP C- and N-terminus ( Figure 1A , as control ) , and further probed for the presence of additional AZ proteins , such as Liprin-α ( Figure 1B ) and Syd-1 ( Figure 1C ) , which interact with BRP at the AZ ( Owald et al . , 2010 , 2012 ) and the small GTPase Rab3 that was previously shown to regulate the distribution of presynaptic components at AZs ( Figure 1D; Graf et al . , 2009 ) . However , none of these AZ proteins showed co-accumulation with BRP in the aggregates ( B as also described in Johnson et al . , 2009 ) . Staining with anti-RBP Abs ( Liu et al . , 2011a ) , by contrast , revealed strong co-localization of BRP and RBP in the axonal aggregates ( Figure 1E ) . Quantification of BRP and RBP co-localization in two different srpk79D mutant null alleles ( atc from Johnson et al . , 2009; vn from Nieratschker et al . , 2009 ) confirmed the impression that the axonal RBP/BRP signals were of identical size ( Figure 1F; mean area of axonal spots , BRPC-term 0 . 3797 ± 0 . 03694 µm2 in srpk79DATC , 0 . 3259 ± 0 . 02212 µm2 in srpk79Dvn; RBPC-term 0 . 3892 ± 0 . 02097 µm2 in srpk79DATC , 0 . 3696 ± 0 . 01645 µm2 in srpk79Dvn; n = 8 nerves; mean ± SEM ) , and that BRP and RBP nearly always co-localized in these aggregates ( Figure 1G; BRPC-term co-localizing with RBPC-term 93 . 26% ± 2 . 172 in srpk79DATC , 95 . 85% ± 1 . 302 in srpk79Dvn; RBPC-term co-localizing with BRPC-term 95 . 7% ± 0 . 9713 in srpk79DATC , 94 . 24% ± 1 . 162 in srpk79Dvn; n = 8 nerves; mean ± SEM ) . 10 . 7554/eLife . 06935 . 003Figure 1 . Co-accumulation of Bruchpilot ( BRP ) and RIM-binding protein ( RBP ) in srpk79D axonal aggregates . ( A–E , I ) Nerve bundles of segments A1–A3 from third instar larvae of the genotypes indicated labeled with the antibodies ( Abs ) indicated . ( A–E , H ) BRP accumulated in axonal aggregates of srpk79D mutants . ( B–D ) Liprin-α ( B ) , Syd-1 ( C ) , and Rab3 ( D ) , did not co-localize with axonal BRP spots . ( E ) By contrast , RBP invariably co-localized with BRP in these axonal aggregates . ( F ) Quantification of mean area of axonal BRP and RBP spots in wild type ( WT ) and srpk79D mutants . BRPC-term spots: 0 . 3797 ± 0 . 03694 µm2 in srpk79DATC , 0 . 3259 ± 0 . 02212 µm2 in srpk79Dvn , 0 . 06895 ± 0 . 01 µm2 in WT; RBPC-term spots: 0 . 3892 ± 0 . 02097 µm2 in srpk79DATC , 0 . 3696 ± 0 . 01645 µm2 in srpk79Dvn , 0 . 09184 ± 0 . 0133 in WT; n = 8 nerves each; all panels show mean values and errors bars representing SEM; ns , not significant , p > 0 . 05 , Mann–Whitney U test . ( G ) Quantification for BRP co-localization with RBP and vice versa in srpk79D mutants . BRPC-term co-localizing with RBPC-term: 93 . 26% ± 2 . 172 in srpk79DATC , 95 . 85% ± 1 . 302 in srpk79Dvn; RBPC-term co-localizing with BRPC-term: 95 . 7% ± 0 . 9713 in srpk79DATC , 94 . 24% ± 1 . 162 in srpk79Dvn; n = 8 nerves each; all panels show mean values and errors bars representing SEM; ns , not significant , p > 0 . 05 , Mann–Whitney U test . ( H ) Two-colour stimulated emission depletion ( STED ) images of axonal aggregates in srpk79D mutants revealed that RBPC-Term label localized to the inside of the axonal aggregates and was surrounded by BRPC-Term label . ( I ) BRP and RBP also co-localized in axonal spots of WT animals ( arrow heads show co-localization of BRP and RBP in the axon ) . Scale bars: ( A–E , I ) 10 µm; ( H ) 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 003 Thus , RBP was the only AZ protein that robustly co-accumulates with BRP in srpk79D mutant axonal aggregates . To further explore the distribution of BRP and RBP in these aggregates we used stimulated emission depletion ( STED ) light microscopy at a resolution of about 50 nm ( Hell , 2007 ) . Two-colour STED microscopy revealed a tight and stoichiometric association of BRP and RBP in the floating axonal aggregates of srpk79D mutants ( Figure 1H ) , reminiscent of EM images showing T-bar super assemblies in these axons ( Figure 1H; Johnson et al . , 2009; Nieratschker et al . , 2009 ) . In fact , the relative distribution of RBP vs BRPC-term was very reminiscent of the organisation at mature , synaptic AZs ( Liu et al . , 2011a ) . The tight association of BRP and RBP in these ectopic aggregates further suggested a co-transport of both AZ components . Indeed , we could identify axonal BRP spots co-positive for RBP ( Figure 1I , arrows ) in wild type ( WT ) larvae as well . Compared to srpk79D mutant axons , WT BRP/RBP co-positive aggregates were present at a lower frequency and displayed a ∼ four times smaller average diameter in control axons ( Figure 1F; mean area of axonal spots , BRPC-term 0 . 06895 ± 0 . 01 µm2 in WT; RBPC-term spots: 0 . 09184 ± 0 . 0133 in WT; n = 8 nerves; mean ± SEM ) . We observed active anterograde and retrograde transport of the BRP ( GFP-labelled ) /RBP ( cherry-labelled ) co-positive spots when using intravital imaging of axons of intact larvae ( Rasse et al . , 2005 ) ( Figure 2A; Video 1 ) . Thus , as our data strongly suggested that BRP and RBP are co-transported , we searched for adaptor proteins coupling them to axonal motors . 10 . 7554/eLife . 06935 . 004Figure 2 . Live imaging of anterograde co-transport between BRP , RBP and APP-like protein interacting protein 1 ( Aplip1 ) . ( A ) Live imaging in intact third instar larvae showed anterograde co-transport of BRPGFP and RBPcherry . See also , Video 1 . ( B ) Schematic representation of Aplip1 domain structure containing two PxxP motifs , one Src-homology 3 ( SH3 ) domain and one C-terminal phosphotyrosine interaction domain ( PID ) ( FL = full-length ) . Lines represent Aplip1 prey fragments recovered in RBP SH3-II+III bait yeast-two-hybrid ( Y2H ) screen . Arrow indicates one single clone that contained only the first of the two Aplip1-PxxP motifs . ( C , D ) Live imaging in intact third instar larvae showed anterograde co-transport of Aplip1GFP and RBPcherry ( C ) , as well as Aplip1GFP and BRP-shortstraw ( D ) . Scale bars: ( A , C , D ) 10 µm . See also , Videos 2 , 3 . ( E ) Quantification of live imaging of BRP-shortstraw flux ( spots passing through an axonal cross-section per minute ) within the genetic backgrounds indicated . Anterograde and retrograde BRP-shortstraw flux was severely impaired in aplip1ek4 and aplip1null mutant background , which was rescued when a genomic rescue construct for Aplip1 was introduced into the aplip1null mutant background . BRP-shortstraw flux per min , control ( n = 14 nerves ) : anterograde: 5 . 267 ± 0 . 975 , retrograde: 2 . 423 ± 0 . 604 , stationary: 0 . 241 ± 0 . 071; aplip1ek4 ( n = 28 nerves ) : anterograde: 0 . 687 ± 0 . 098 , retrograde: 0 . 284 ± 0 . 125 , stationary: 1 . 023 ± 0 . 145; aplip1null ( n = 11 nerves ) : anterograde: 0 . 808 ± 0 . 051 , retrograde: 0 . 085 ± 0 . 064 , stationary: 0 . 354 ± 0 . 148; aplip1null , gen rescue ( n = 26 nerves ) : anterograde: 3 . 783 ± 0 . 861 , retrograde: 2 . 123 ± 0 . 239 , stationary: 0 . 505 ± 0 . 084 . All panels show mean values and errors bars representing SEM . *p ≤ 0 . 05; **p ≤ 0 . 01; ***p ≤ 0 . 001; ns , not significant , p > 0 . 05 , Mann–Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 00410 . 7554/eLife . 06935 . 005Video 1 . Anterograde co-transport of BRPGFP and RBPcherry . Live imaging in intact third instar larvae showed anterograde co-transport of BRPGFP and RBPcherry . Video was captured at 0 . 6 s per frame and played back at 7× real time . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 005 RBP , via its second and third SH3 domain , is known to bind synaptic ligands such as Ca2+ channels and RIM ( Liu et al . , 2011a ) . Both the SH3 domains and the cognate PxxP motifs of the synaptic ligands are highly conserved between mammals and Drosophila ( Liu et al . , 2011a; Südhof , 2012; Davydova et al . , 2014 ) . However , in order to identify novel RBP interaction partners which might be relevant in the context of axonal transport , we performed a large-scale yeast two-hybrid ( Y2H ) screen using a construct consisting of the second and third SH3 domains of Drosophila RBP as bait ( also shown in Figure 3A ) . As expected , several clones representing RIM and the Ca2+ channel α1-subunit Cacophony ( Cac ) were isolated ( not shown ) . In addition , the screen recovered 14 independent fragments of Aplip1 , including a full length cDNA clone ( Figure 2B ) . Aplip1 is the Drosophila homolog of c-Jun N-terminal kinase ( JNK ) -interacting protein 1 ( JIP1 ) , a scaffolding protein that has been shown to bind kinesin light chain ( KLC; Verhey et al . , 2001 ) , Alzheimer's amyloid precursor protein ( APP; Taru et al . , 2002 ) , JNK pathway kinases ( Horiuchi et al . , 2005 , 2007 ) and the autophagosome adaptor LC3 ( Fu et al . , 2014 ) . If Aplip1 was mediating the axonal transport of RBP , moving spots co-positive for both RBP and Aplip1 should be expected . In fact , we robustly observed co-transport of RBPcherry and Aplip1GFP spots in both anterograde ( Figure 2C , arrowhead; Video 2 ) and retrograde ( not shown ) direction at a frequency consistent with the low frequency of single Aplip1GFP moving particles ( not shown ) . Furthermore , we observed BRP-shortstraw co-transport with Aplip1GFP ( Figure 2D; Video 3 ) , as expected with similarly low frequencies as observed for RBP/Aplip1 co-transport ( not shown ) , further pointing towards a co-transport of RBP and BRP in conjunction with Aplip1 . We used the live imaging assay to investigate BRP transport in different aplip1 mutants to directly address whether removal of Aplip1 affects AZ scaffold protein transport . The aplip1null allele completely and specifically removes the aplip1 gene and was generated by P-element excision ( Klinedinst et al . , 2013 ) . By comparison , the aplip1ek4 allele contains a point mutation in the C-terminal kinesin binding domain of Aplip1 that was shown to almost completely abolish the ability of Aplip1 to bind to KLC ( Horiuchi et al . , 2005 ) . Anterograde and retrograde transport of BRP was drastically reduced compared to controls in both aplip1 mutant alleles ( Figure 2E ) . Through the introduction of a genomic ( gen . ) construct of Aplip1 into the aplip1null mutant background ( aplip1null , gen . rescue ) , however , BRP flux ( spots passing through an axonal cross-section in a given time ) could be restored to WT level ( Figure 2E ) . Quantification showed that retrograde transport in the aplip1null mutant situation was somewhat more affected ( 27× less compared to control ) than anterograde transport ( 7× less ) . Both directions appeared equally affected ( about 8× less compared to controls ) in the kinesin-binding defective aplip1ek4 mutant . It is noteworthy that the transport of SV cargo in the same mutant was reduced equally in both directions , whereas transport of mitochondria is only impaired in the retrograde direction ( Horiuchi et al . , 2005 ) . 10 . 7554/eLife . 06935 . 012Video 2 . Anterograde co-transport of Aplip1GFP and RBPcherry . Live imaging in intact third instar larvae showed anterograde co-transport of Aplip1GFP and RBPcherry . Video was captured at 0 . 6 s per frame and played back at 7× real time . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 01210 . 7554/eLife . 06935 . 013Video 3 . Anterograde co-transport of Aplip1GFP and BRP-shortstraw . Live imaging in intact third instar larvae showed anterograde co-transport of Aplip1GFP and BRP-shortstraw . Video was captured at 0 . 414 s per frame and played back at 5× real time . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 013 As our Y2H screen used the SH3-II and -III domains of RBP as bait ( Figure 3A ) , PxxP motifs are expected to mediate the interaction with Aplip1 . In fact , Aplip1 contains two PxxP motifs which were both present in most of the prey clones recovered in the Y2H screen , except for one single clone that contained only the first more N-terminal motif ( Figure 2B , arrow ) . Using a semi-quantitative liquid Y2H assay and a set of Aplip1 constructs containing only either the first or the second PxxP motif ( Figure 3B ) , we mapped the interaction between RBP and Aplip1 to the first of the two candidate PxxP motifs present in all clones isolated ( Figure 2B ) . The second and third SH3 domain of RBP bound to this motif with comparable strength when measured with a semi-quantitative liquid Y2H assay ( Figure 3C; mean ß-Gal4 units for: Aplip1-PxxP1/RBP SH3-II: 24 . 3 ± 6 . 6; Aplip1-PxxP1/RBPSH3-III: 29 . 1 ± 7 . 4; n = 3 independent experiments; mean ± SEM ) . No binding was observed between the second and third SH3 domains of RBP and Aplip1-PxxP2 ( Figure 3C; mean ß-Gal4 units for: Aplip1-PxxP2/RBP SH3-II: 0 . 2 ± 0 . 0; Aplip1-PxxP2/RBPSH3-III: 0 . 2 ± 0 . 1; n = 3 independent experiments; mean ± SEM ) . When mutating either the PxxP1 motif of Aplip1 ( P156 → A; P159 → A , giving rise to AxxA1 ) or introducing mutations known to interfere with PxxP ligand binding into the individual SH3 domains of RBP ( SH3-II*/SH3-III* ) , the interaction was completely abolished ( Figure 3C; mean ß-Gal4 units for: Aplip1-AxxA1/RBP SH3-II: 0 . 1 ± 0 . 1; Aplip1-AxxA1/RBP SH3-III: 0 . 2 ± 0 . 0; Aplip1-PxxP1/RBPSH3-II*: 0 . 1 ± 0 . 0; Aplip1-PxxP1/RBP SH3-III*: 0 . 1 ± 0 . 0; n = 3 independent experiments; mean ± SEM ) . We performed isothermal titration calorimetry ( ITC ) to measure the thermodynamics of the binding directly and compare Aplip1/RBP binding quantitatively to the established synaptic ligands of RBP . We used four different constructs , comprising either single RBP SH3 domains ( I , II , and III ) or a construct of two RBP SH3 domains ( II+III ) ( see also Figure 3A ) . Whereas we could not detect any binding of the Aplip1 peptides to RBP SH3-I , we could determine KD constants for the single SH3-II , SH3-III and the tandem SH3-II+III ( Figure 3D; Figure 3—figure supplement 1 ) domains of RBP . Both SH3-II and SH3-III single domains showed a binding affinity to Aplip1 peptides several fold stronger compared to either Cac , RIM1 or RIM2 ( Figure 3D; Figure 3—figure supplements 2–4 ) . However , the affinity of the Aplip1 peptides to the SH3-II+III domain was the highest observed which is indicative of co-operativity between both domains in peptide binding that could increase the local concentrations of Aplip1 at RBP binding pockets ( BPs ) . 10 . 7554/eLife . 06935 . 006Figure 3 . Aplip1 binds RBP using a high-affinity PxxP1-SH3 interaction . ( A ) Schematic representation of RBP domain structure containing three SH3 domains ( I–III from the N-terminus ) and three Fibronectin 3 ( FN3 ) domains . The corresponding fragments used in the large scale Y2H screen ( SH3-II+III ) and used as bait ( SH3-II and SH3-III ) in the Y2H assay ( C ) against different Aplip1 prey constructs ( B ) are indicated . Different isothermal titration calorimetry ( ITC ) peptides ( SH3-I , SH3-II , SH3-III and SH3-II+III ) used for ITC measurements ( D ) are also shown . ( B ) Schematic representation of Aplip1 domain structure entailing two PxxP motifs , one SH3 and one C-terminal PID . Different preys ( Aplip1-PxxP1 , -AxxA1 and -PxxP2 ) used in Y2H assay ( C ) are indicated . ( C ) Liquid Y2H assay of individual Aplip1 prey fragments against different RBP baits . Aplip1-PxxP1 interacted with both the single SH3-II and -III domains of RBP . Mutating this first PxxP motif ( Aplip1-AxxA1 ) construct abolished the binding . Aplip1-PxxP2 showed no interaction to RBP SH3 domains . Constructs with point-mutated RBP SH3 domains ( SH3-II* , SH3-III* ) abolished the binding to Aplip1-PxxP1 . ( D ) Peptide sequences used for ITC measurements . Aplip1 showed the strongest interaction with RBP compared with Cacophony ( Cac ) , RIM1 and RIM2 , with the strongest affinity ( lowest KD ) between Aplip1 and the RBP SH3-II+III domain . ( E , F ) Crystal structure of Aplip1-peptide ( E; see also , 3D for peptide sequence ) and of Cac-peptide ( F; see also , Figure 3D for peptide sequence ) bound to RBP SH3-II . The SH3 domain is shown in gray surface representation with ( left ) and without ( right ) the respective protein in cartoon representation . The bound peptides are drawn in stick representation . Hydrogen bonds ≤3 . 3 Å are indicated by red dashes . In the right panel , several peptide SH3-II complexes as observed in the asymmetric unit are superimposed and shown in different colors . See also , Tables 1–4 . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 00610 . 7554/eLife . 06935 . 007Figure 3—figure supplement 1 . ITC measurements for Aplip1 and RBP SH3 domains . Quantification of protein-peptide interactions by ITC . Both the raw data and the data integrated are shown . Data were fitted based on the ‘One Set of Sites’ model . ( A ) Titration of RBP-BP SH3-II and the Aplip1 peptide . ( B ) Titration of RBP-BP SH3-III and the Aplip1 peptide . ( C ) Titration of RBP-BP SH3-II+SH3-III and the Aplip1 peptide . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 00710 . 7554/eLife . 06935 . 008Figure 3—figure supplement 2 . ITC measurements for Cac and RBP SH3 domains . Quantification of protein-peptide interactions by ITC . Both the raw data and the data integrated are shown . Data were fitted based on the ‘One Set of Sites’ model . ( A ) Titration of RBP-BP SH3-II and the Cac peptide . ( B ) Titration of RBP-BP SH3-III and the Cac peptide . ( C ) Titration of RBP-BP SH3-II+SH3-III and the Cac peptide . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 00810 . 7554/eLife . 06935 . 009Figure 3—figure supplement 3 . ITC measurements for RIM1 and RBP SH3 domains . Quantification of protein-peptide interactions by ITC . Both the raw data and the data integrated are shown . Data were fitted based on the ‘One Set of Sites’ model . ( A ) Titration of RBP-BP SH3-II and the RIM1 peptide . ( B ) Titration of RBP-BP SH3-III and the RIM1 peptide . ( C ) Titration of RBP-BP SH3-II+SH3-III and the RIM1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 00910 . 7554/eLife . 06935 . 010Figure 3—figure supplement 4 . ITC measurements for RIM2 and RBP SH3 domains . Quantification of protein-peptide interactions by ITC . Both the raw data and the data integrated are shown . Data were fitted based on the ‘One Set of Sites’ model . ( A ) Titration of RBP-BP SH3-II and the RIM2 peptide . ( B ) Titration of RBP-BP SH3-III and the RIM2 peptide . ( C ) Titration of RBP-BP SH3-II+SH3-III and the RIM2 . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 01010 . 7554/eLife . 06935 . 011Figure 3—figure supplement 5 . Crystal structure of Cac-peptide bound to RBP SH3-III domain . The SH3 domain is shown in gray surface representation , with ( left ) and without ( right ) the respective protein in cartoon representation . The bound peptides are drawn in stick representation . Hydrogen bonds ≤3 . 3 Å are indicated by red dashes . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 011 Finally , in order to get a deeper atomic insight into the structural basis of the binding of RBP towards Aplip1 in comparison to its synaptic ligands , we crystallized the Drosophila RBP SH3-II domain together with both an Aplip1 ( Figure 3E; Tables 1 , 2 , 3 ) and a Cac peptide ( Figure 3F; Tables 1 , 3 , 4 ) , and RBP SH3-III with a Cac peptide ( Figure 3—figure supplement 5; Tables 1 , 3 ) . Drosophila RBP SH3-II and -III share 49 . 2% sequence identity and adopt the canonical fold of SH3 domains ( Figure 3E , F; Figure 3—figure supplement 5 ) . Both domains superimpose with a root mean deviation of 0 . 8 Å for 64 pairs of Cα-atoms . Both peptides sequences harbor the canonical class I interaction motif +xΨPxxP ( + , positively charged; x , any amino acid; Ψ hydrophobic amino acid , see Figure 3D for sequence ) and are bound into the respective SH3 domain in ‘plus’ direction . We observed the classical poly-proline helix that allows for mainly hydrophobic protein-peptide interaction in all three structures . We detected the same hydrogen pattern between the protein side chains and peptide backbone in the structure of SH3-II with Aplip1 and Cac . The major difference is the side chain orientation of R1687 of Cac that π-stacks with its guanidinium function with Y1372 , except for one copy , where it forms a salt-bridge to E1341 . The equivalent residue to R1687 of Cac is R153 of the Aplip1 peptide , which forms , by contrast , a bidentate salt-bridge to D1336 ( Table 3 ) . A second major difference is induced by the two consecutive proline residues in the Cac peptide . Consequently , the peptide has a more polyproline type II conformation that brings T1692 closer to the protein surface and allows P1693 to deeper point in a hydrophobic pocket of the SH3-II domain . Whereas the C-terminal portion of the Aplip1 peptide is folded in a short 310 helix , the N-terminus of the Aplip1 peptide adopts a random coil conformation with hydrophobic interactions to the surface of SH3-II . The Cac-derived peptide bound to SH3-III is fully defined in the electron density . However , the peptide main chain interaction with the SH3 domains is conserved . The side chain orientation of Cac R1687 is again different if bound to SH3-II or SH3-III . In complex with SH3-III , R1687 forms a bidentate hydrogen bond to SH3-III D1463 and E1648 . A π-stacking interaction is not possible since Y1372 of SH3-II is replaced by SH3-III L1499 . The central PxxP motifs of Aplip1 superimpose well in both structures if bound to SH3-II and SH3-III . Towards its C-terminus , the Aplip1-PxxP1 peptide adopts a slightly different random coil conformation compared to the structure when bound to SH3-II caused by two additional hydrogen bonds from T1692 and K1695 to the SH3-II domain ( Table 3 ) . 10 . 7554/eLife . 06935 . 014Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 014Data collection StructureRBP SH3-IIRBP SH3-IIRBP SH3-IIIAplip1CacCac PDB entry4Z884Z894Z8A Space groupC2P21I222 Wavelength ( Å ) 0 . 918410 . 918410 . 91841 Unit cell a; b; c ( Å ) 108 . 3; 62 . 4; 163 . 658 . 3; 122 . 2; 68 . 552 . 1; 54 . 3; 73 . 6 α; β; γ ( ° ) 90 . 0; 90 . 3; 90 . 090 . 0; 113 . 2; 90 . 090 . 0; 90 . 0; 90 . 0 Resolution ( Å ) *50 . 00–2 . 0950 . 00–2 . 6450 . 00–1 . 75 ( 2 . 19–2 . 09 ) ( 2 . 74–2 . 64 ) ( 1 . 86–1 . 75 ) Unique reflections64 , 269 ( 7760 ) 25 , 229 ( 2591 ) 10 , 690 ( 1579 ) Completeness*98 . 9 ( 92 . 4 ) 96 . 9 ( 95 . 0 ) 98 . 7 ( 92 . 6 ) <I/σ ( I ) >*7 . 7 ( 2 . 6 ) 8 . 0 ( 2 . 1 ) 14 . 2 ( 2 . 2 ) Rmeas* , †0 . 127 ( 0 . 533 ) 0 . 157 ( 0 . 726 ) 0 . 127 ( 0 . 663 ) CC1/2*99 . 1 ( 68 . 0 ) 98 . 9 ( 81 . 2 ) 99 . 7 ( 76 . 5 ) Redundancy*3 . 7 ( 3 . 7 ) 3 . 5 ( 3 . 2 ) 5 . 6 ( 3 . 1 ) Refinement Non-hydrogen atoms75646239850 Rwork* , ‡0 . 210 ( 0 . 314 ) 0 . 255 ( 0 . 367 ) 0 . 159 ( 0 . 233 ) Rfree* , §0 . 236 ( 0 . 396 ) 0 . 312 ( 0 . 490 ) 0 . 208 ( 0 . 332 ) Average B-factor ( Å2 ) 40 . 852 . 1018 . 8 No . of complexes24101 Protein residues6484/41 . 0663/51 . 174/17 . 6 Peptide residues861/42 . 792/63 . 615/15 . 9 Buffer molecules11/40 . 21/46 . 3– Water molecules57/29 . 6134/30 . 3110/28 . 6 r . m . s . d . # bond length ( Å ) 0 . 0070 . 0050 . 010 bond angles ( ° ) 1 . 2241 . 1401 . 210 Ramachandran outliers ( % ) 0 . 10 . 560 Ramachandran favoured ( % ) 98 . 498 . 0100*values in parentheses refer to the highest resolution shell . †Rmeas = Σh [n/ ( n − 1 ) ]1/2 Σi|Ih − Ih , i|/ΣhΣiIh , i where Ih is the mean intensity of symmetry-equivalent reflections and n is the redundancy . ‡Rwork = Σh|Fo − Fc|/ΣFo ( working set , no σ cut-off applied ) . §Rfree is the same as Rwork , but calculated on 5% of the data excluded from refinement . #Root-mean-square deviation ( r . m . s . d . ) from target geometries . CC , coiled coil . 10 . 7554/eLife . 06935 . 015Table 2 . Completeness of the model for RBP SH3-II and bound Aplip1 peptideDOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 015RBP SH3-IIRangeAplip1Rangechain A1318–1382chain M153–163chain Bx1318–1382chain N155–159chain Cx1318–1381chain O154–163chain Dx1318–1382chain P153–159chain E1319–1381chain Q151–163chain Fx1318–1380chain R153–159chain Gx1318–1381chain S151–163chain Hx1318–1382chain T152–156chain Ix1318–1382chain U152–163chain Jx1318–1381chain V152–158chain Kx1318–1381chain W152–163chain Lx1318–1381chain X152–158Completeness of the model given for the 12 complexes of RBP SH3-II bound to the Aplip1 peptide 149TRRRRKLPEIPKNKK163 . Superscript ‘x’ indicates additional N-terminal residues of RBP SH3-II originating from the linker region between the protease cleavage site and the N-terminus . 10 . 7554/eLife . 06935 . 016Table 3 . Hydrogen bonding interactionDOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 016Aplip1SH3-IIDistanceArg153NAsp1359OD22 . 4Arg153NH2Asp1336OD13 . 0Arg153NH2Asp1336OD22 . 6Lys154NAsn1334OD12 . 9Lys154OAsn1334ND23 . 0Pro156OAsn1376ND22 . 8CacSH3-IIDistanceGly1686NAsp1359OD22 . 7Arg1687NAsp1359OD22 . 8Arg1688NAsn1334OD13 . 0Arg1688OAsn1334ND22 . 9Pro1690OAsn1376ND22 . 8CacSH3-IIIDistanceArg1687NH1Asp1463OD12 . 9Arg1687NH1Glu1488OE23 . 0Arg1687NH2Glu1488OE23 . 1Arg1688NAsn1461OD12 . 8Arg1688OAsn1461ND23 . 0Pro1690OAsn1376ND22 . 9Thr1692OGAsn1376ND22 . 9Lys1695OTyr1451OH2 . 8Ser1697OGLeu1450O2 . 7Hydrogen bonding interaction of RBP SH3-II with Aplip1 and Cac , as well as RBP SH3-III in complex with Cac . Distance ≤3 . 2 Å are given in Å . 10 . 7554/eLife . 06935 . 017Table 4 . Completeness of the model for RBP SH3-II and bound Cac peptideDOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 017RBP SH3-IIRangeCacRangechain A1318–1381chain a1686–1697chain Bx1318–1381chain b1686–1695chain Cx1318–1382chain c1686–1697chain Dx1318–1381chain d1686–1697chain E1318–1382chain e1685–1694chain Fx1318–1382chain f1685–1693chain Gx1318–1382chain g1686–1693chain H1318–1381chain h1686–1693chain Ix1318–1381chain i1686–1693chain Jx1318–1382chain j1686–1697Completeness of the model given for the six complexes of RBP SH3-II and the bound Cac peptide 1685IGRRLPPTPSKPSTL1699 . Superscript ‘x’ indicates additional N-terminal residues of RBP SH3-II originating from the linker region between the protease cleavage site and the N-terminus . Consistent with the idea that Aplip1 is mediating RBP transport , we found axonal aggregates consisting of both RBP and BRP in the aplip1ek4 , as well as the aplip1null allele ( Figure 4B , C ) . This ectopic RBP/BRP accumulation was rescued after introducing a genomic construct of Aplip1 into the aplip1null mutant background ( aplip1null , gen . rescue; Figure 4D ) . Pan-neuronal expression of an Aplip1 cDNA equally rescued the axonal RBP/BRP accumulations ( Figure 4I , quantification in K , L ) . Importantly , however , the expression of an Aplip-AxxA1 cDNA construct ( integrated at the same chromosomal integration site as the control construct; expression and axonal presence confirmed with a newly generated Aplip1 Ab; not shown ) could no longer rescue the RBP/BRP accumulation phenotype ( Figure 4J , quantification in Figure 4K , L ) . Thus , we conclude that Aplip1 is involved in the transport of RBP/BRP to the AZ , whereby its functionality in this context largely depends on the integrity of its N-terminal PxxP1 motif . 10 . 7554/eLife . 06935 . 018Figure 4 . Aplip1-PXXP1 motif is needed for its function as RBP/BRP transport adaptor . ( A–D ) Nerve bundles of segments A1–A3 from third instar larvae of the genotypes indicated labeled with the Abs indicated . ( E , F ) Quantification of BRP/RBP spot numbers . BRP spots per µm2: WT ( n = 8 nerves ) : 0 . 084 ± 0 . 010; aplip1ek4 ( n = 9 nerves ) : 0 . 205 ± 0 . 025; aplip1null ( n = 8 nerves ) : 0 . 183 ± 0 . 025; aplip1null , gen . rescue ( n = 8 nerves ) : 0 . 034 ± 0 . 007; RBP spots per µm2 , WT ( n = 8 nerves ) : 0 . 074 ± 0 . 007; aplip1ek4 ( n = 9 nerves ) : 0 . 180 ± 0 . 019; aplip1null ( n = 8 nerves ) : 0 . 153 ± 0 . 037; aplip1null , gen . rescue ( n = 8 nerves ) : 0 . 025 ± 0 . 006 . All panels show mean values and errors bars representing SEM . *p ≤ 0 . 05; **p ≤ 0 . 01; ***p ≤ 0 . 001; ns , not significant , p > 0 . 05 , Mann–Whitney U test . ( G–J ) Nerve bundles of segment A1–A3 from third instar larvae of the genotypes indicated labeled with the Abs indicated . BRP and RBP co-localized in control animals and accumulated in a co-localizing fashion in axons of aplip1null mutant animals . Re-expression of an Aplip1-WT cDNA construct in the aplip1null background rescued the phenotype , while re-expression of an AxxA1 construct did not . ( K , L ) Quantification of the number of BRP/RBP spots per µm2 axon . BRP spots per µm2 , control ( n = 12 nerves ) : 0 . 084 ± 0 . 010; aplip1null ( n = 16 nerves ) : 0 . 198 ± 0 . 022; WT rescue ( n = 14 nerves ) : 0 . 078 ± 0 . 009; AxxA1 rescue ( n = 14 nerves ) : 0 . 177 ± 0 . 012; RBP spots per µm2 , control ( n = 12 nerves ) : 0 . 071 ± 0 . 013; aplip1null ( n = 16 nerves ) : 0 . 188 ± 0 . 026; WT rescue ( n = 14 nerves ) : 0 . 039 ± 0 . 004; AxxA1 rescue ( n = 14 nerves ) : 0 . 158 ± 0 . 015 . All panels show mean values and errors bars representing SEM . *p ≤ 0 . 05; **p ≤ 0 . 01; ***p ≤ 0 . 001; ns , not significant , p > 0 . 05 , Mann–Whitney U test . Scale bar: ( A–D , G–J ) 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 018 As indicated above , BRP accumulated in the axons of aplip1 mutants as well . Thus , BRP could be transported through Aplip1 via binding to RBP , other yet undetected co-transported AZ proteins , or BRP could bind Aplip1 independently of RBP . We therefore created aplip1/rbp and aplip1/brp double mutants to investigate the functional relation of RBP and BRP with regard to Aplip1-dependent transport . While removing BRP in srpk79D mutants also abolished the axonal RBP spots ( Figure 5—figure supplement 1D ) , removing BRP in aplip1 mutants had no apparent effect on axonal RBP accumulations ( Figure 5B; control in Figure 5A ) . On the other hand , genetic elimination of RBP did not interfere with the accumulation of BRP in aplip1 mutant axons ( Figure 5E; controls in Figure 5C , D ) . Thus , BRP transport also ‘suffers’ from the absence of the Aplip1 adaptor when RBP is removed in parallel . Hence , Aplip1 promotes BRP transport even in the absence of RBP . To address a putative molecular basis of this relationship , we performed a Y2H assay to test for direct interaction between five different BRP constructs and a full length Aplip1 construct ( see Figure 3B for domain structure ) . Despite these efforts , robust interactions between Aplip1 and BRP fragments could not be detected ( data not shown ) . Nonetheless , both RBP but also BRP were easily detected in anti-GFP immunoprecipitations ( IPs ) from a synaptic membrane preparation ( Figure 5F; Figure 5—figure supplement 2 ) derived from Drosophila head extracts of pan-neuronal driven Aplip1-GFP cDNA construct ( Depner et al . , 2014 ) . Of note , within axons of rbpnull mutant larvae , ectopic BRP accumulations could not be observed ( not shown ) . Thus , we provide evidence for an RBP-independent but Aplip1-dependent transport component for BRP , whose mechanistic details have still to be deciphered . Taken together , our results imply that though BRP and RBP are co-transported in the WT situation , their Aplip1-dependent transport can be genetically uncoupled . 10 . 7554/eLife . 06935 . 019Figure 5 . Aplip1 promotes BRP transport in absence of RBP . ( A–E ) Nerve bundles of segments A1–A3 from third instar larvae of the genotypes indicated labeled with the Abs indicated . ( A ) Removing one copy of BRP in aplip1ek4 mutants had no apparent effect on axonal RBP accumulation . ( B ) RBP still accumulates in brpnull;aplip1ek4 double mutants . ( C , D ) Driver control and removing one copy of RBP in motoneuronal driven Aplip1-RNAi had no apparent effect on axonal BRP accumulation . ( E ) BRP still accumulates in rbpnull , aplip1 double mutants Scale bar: ( A–E ) 10 µm . ( F ) Immunoprecipitation ( IP ) of Aplip1GFP with anti-GFP Ab from Drosophila active zone ( AZ ) protein-enriched fraction was followed by Western blot ( WB ) analysis using anti-BRPLast200 and anti-RBPSH3-II+III . Both BRP and RBP could be detected in Aplip1GFP IPs , but are absent in controls ( plain beads; GFP trapped beads ) . ( For whole WBs , see Figure 5—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 01910 . 7554/eLife . 06935 . 020Figure 5—figure supplement 1 . Accumulation of BRP in srpk79D mutant axons is unaffected by removing RBP . ( A–F ) Nerves of segments A1–A3 from third instar larvae of the genotypes indicated labeled with the Abs indicated . Removing BRP in srpk79D mutants ( D ) also abolished axonal RBP spots , while removing RBP in srpk79D mutants did not affect BRP accumulations ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 02010 . 7554/eLife . 06935 . 021Figure 5—figure supplement 2 . IP of Aplip1GFP with anti-GFP ( Full blot ) . Full blot IP of Aplip1GFP with anti-GFP from Drosophila AZ protein-enriched fraction was followed by WB analysis using anti-BRPLast200 and anti-RBPSH3-II+III Abs . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 021 The BRP flux in axons of aplip1 mutants was severely diminished , but not completely abolished ( Figure 2E ) . At the same time , AZ localization of both BRP and RBP at synaptic terminals of aplip1 mutants was still observed in both aplip1 alleles ( not shown ) , although slightly reduced ( not shown ) . This indicates that alternative transport mechanisms and adaptors exist which operate in parallel to Aplip1 , as the synaptic phenotype is relatively weak . In fact , axonal accumulations of BRP have already been described for Acyl-CoA long-chain Synthetase ( Acsl , Liu et al . , 2011b ) as well as for Unc-51 ( Atg1 ) mutants ( Wairkar et al . , 2009 ) . In our experiments , we found RBP to invariably co-cluster with BRP in the mutants mentioned ( Figure 6B , C; control in Figure 6A ) , and equally in mutants of the Drosophila ß-amyloid protein precursor-like ( Appl; Torroja et al . , 1999a , 1999b; Figure 6D ) and Unc-76 ( Gindhart et al . , 2003; Figure 6E ) . The fact that RBP and BRP tightly co-accumulated in axonal aggregates of all these transport mutants strengthens the probability that BRP is always co-transported with RBP . 10 . 7554/eLife . 06935 . 022Figure 6 . Several known transport adaptor mutants showed axonal BRP and RBP co-accumulations . ( A–E ) Nerve bundles of segment A1–A3 from third instar larvae of the genotypes indicated labeled with the Abs indicated . BRP and RBP accumulated in a co-localizing manner in axons of WT ( A ) , acsl ( B ) , unc-51 ( atg-1; C ) , appl ( D ) and unc-76 ( E ) . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 022 To gain a deeper insight into the substructure of the BRP/RBP accumulations in aplip1 mutant axons , we again used two-colour STED microscopy . In contrast to the srpk79D aggregates , however , STED images of axonal BRB/RBP accumulations were reminiscent of mature synaptic AZs ( Liu et al . 2011a ) , with BRPC-term signal surrounding the RBP signal , which , in turn , is oriented closer towards the axonal plasma membrane ( Figure 7A , arrow head; plasma membrane indicated by dashed line ) . Interestingly , in contrast to the floating T-bar super-aggregates in srpk79D mutants ( Johnson et al . , 2009; Nieratschker et al . , 2009 ) , these axonal BRP spots in aplip1 mutants were positive for Syd-1 ( compare Figures 1C , 7B ) . Intriguingly , floating T-bars have been observed in synaptic boutons in syd-1 mutants ( Owald et al . , 2010 ) . Together , this is suggestive of a role of Syd-1 in the membrane-anchoring of AZ proteins . 10 . 7554/eLife . 06935 . 023Figure 7 . Ectopic AZ scaffold and synaptic vesicle ( SV ) accumulation in aplip1 mutant axons . ( A ) Two-colour STED images of axonal aggregates in aplip1ek4 mutants revealed that the structures observed ( arrow heads ) have identical BRP and RBP arrangement , as recently observed at presynaptic AZs ( Liu et al . , 2011a ) . Right panels display magnifications of single axonal AZ . Dashed lines indicate axonal plasma membrane . ( B ) Two-colour STED images of axonal aggregates in aplip1ek4 mutants revealed that the structures observed ( arrow head ) have identical BRP and Syd-1 arrangement as observed at immature presynaptic AZs ( Owald et al . , 2010 ) . Right panels display magnifications of single axonal AZ . Dashed lines indicate axonal plasma membrane . ( C ) Terminal T-bar ( arrow heads ) surrounded by SVs ( arrows ) taken from electron micrographs of WT third instar larvae after conventional embedding . ( D ) Ectopic axonal T-bar ( arrow heads ) taken from electron micrographs from aplip1ek4 mutant third instar larvae after conventional embedding . SVs accumulate around the ectopic T-bar ( arrows ) . ( E ) Magnification of ( C ) . ( F ) Magnification of ( D ) . ( G–J ) Nerve bundles of segment A1–A3 from third instar larvae of the genotypes indicated labeled with the Abs indicated . Syt-1 accumulates at a subset of axonal BRP aggregations in aplip1null and AxxA1 rescue ( H , J ) larvae , but not in control and WT rescue larvae ( G , I ) . ( K ) Quantification of the number of Syt-1 spots per µm2 axon . control ( n = 12 nerves ) : 0 . 004 ± 0 . 002; aplip1null ( n = 16 nerves ) : 0 . 040 ± 0 . 011; WT rescue ( n = 13 nerves ) : 0 . 014 ± 0 . 007; AxxA1 rescue ( n = 13 nerves ) : 0 . 052 ± 0 . 017 . All panels show mean values and errors bars representing SEM . *p ≤ 0 . 05; **p ≤ 0 . 01; ***p ≤ 0 . 001; ns , not significant , p > 0 . 05 , Mann–Whitney U test . Scale bars: ( A , B ) 500 nm; ( C , D ) 100 nm; ( G , J ) 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 02310 . 7554/eLife . 06935 . 024Figure 7—figure supplement 1 . Ectopic AZ protein accumulations in motoneuronal driven Imac- and KHC-RNAi axons . ( A , B ) Two-colour STED images of axonal aggregates in ctrl and Ok6::UAS-Imac-RNAi revealed that BRP and RBP co-accumulate in both genotypes but , in contrast to aplip1 mutants , show no preference concerning orientation towards the axonal plasma membrane ( arrow heads ) . ( C ) Two-colour STED images of axonal aggregates in Ok6::UAS-KHC-RNAi revealed that the BRP-RBP accumulations observed in this genotype mostly show irregular shapes ( arrow heads ) with diverse orientations in the axon . ( D ) The only ectopic axonal electron dense formation ( arrow head ) found in electron micrographs in Ok6::UAS-KHC-RNAi third instar larvae after conventional embedding . ( E , F ) Magnification of ( D ) . SVs ( arrows ) accumulate around the ectopic electron-dense structure ( arrow head ) but are also accumulating all along the axon . Scale bars: ( A–C ) 1 . 5 µm; ( D ) 200 nm; ( E ) 100 nm; ( F ) 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 06935 . 024 Furthermore , we asked whether BRP/RBP aggregates identified in aplip1 mutants represent ectopic AZs forming at the axonal plasma membrane . In fact , EM analysis easily revealed T-bar structures , typical for synaptic terminals ( Figure 7C , arrow heads , magnification in E ) , at axonal plasma membranes of aplip1 mutants ( Figure 7D , arrow heads , magnification in F ) , but never in controls ( not shown ) . We found these ectopic axonal T-bars surrounded by SV profiles ( Figure 7D , arrows ) , very similar to ‘normally positioned’ T-bars at the presynaptic terminal ( Figure 7C , arrows ) . Consistently , the SV marker Synaptotagmin-1 ( Syt-1 ) was found to be associated with BRP/RBP accumulations in aplip1null mutants ( Figure 7H , quantification in Figure 7K ) . This phenotype could be rescued by the expression of an Aplip1 WT cDNA construct ( Figure 7I , quantification in Figure 7K ) but not by the expression of the Aplip1-AxxA1 construct ( Figure 7J; quantification in Figure 7K ) . Thus , a point-like interaction surface of Aplip1 which binds RBP with high affinity is important to block a whole sequence of assembly events at the axonal plasma membrane , including AZ scaffold ( ‘T-bar’ ) formation and the accumulation of SVs . To further support the importance of adaptor protein—cargo interaction in blocking ectopic AZ assembly we downregulated the expression of motor proteins . This also leads to transport defects and ectopic axonal AZ protein accumulations but in principle leaving the adaptor protein—cargo interaction intact . Interestingly , motoneuronal driven Imac-RNAi led to only few axonal BRP/RBP accumulations although with no preference concerning their direction in relation to the axonal plasma membrane ( Figure 7—figure supplement 1B; arrow heads ) . In contrast motoneuronal driven KHC-RNAi showed prominent axonal aggregates consistent of BRP/RBP but most of the time showing an irregular , elongated shape ( Figure 7—figure supplement 1C; arrow heads ) . As mentioned above , proper T-bars were identified in aplip1 mutant axons with ease . In contrast , systematic EM analysis of khc mutant axons revealed just one electron dense material that showed a T-bar-like appearance ( Figure 7—figure supplement 1D; arrow head , magnifications in E , F ) but never in control ( ctrl ) or motoneuronal driven Imac-RNAi . In summary , we find that the SH3-II and -III interaction surface of RBP serves as a multi-functional platform for differential protein interaction with either other AZ components or the transport adaptor and therefore , motor-cargo linkage . Thus , interaction surfaces of RBP/BRP ‘cargo complexes’ might be shielded and blocked from undergoing premature assembly by interactions with transport adaptors , while genetically induced loss of these adaptors might provoke premature AZ assembly .
Large multi-domain scaffold proteins such as BRP/RBP are ultimately destined to form stable scaffolds , characterized by remarkable tenacity and a low turnover , likely due to stabilization by multiple homo- and heterotypic interactions simultaneously ( Sigrist and Schmitz , 2011 ) . How these large and ‘sticky’ AZ scaffold components engage into axonal transport processes to ensure their ‘safe’ arrival at the synaptic terminal remains to be addressed . We find here that the AZ scaffold protein RBP binds the transport adaptor Aplip1 using a ‘classic’ PxxP/SH3 interaction . Notably , the same RBP SH3 domain ( II and III ) interaction surfaces are used for binding the synaptic AZ ligands of RBP , that is , RIM and the voltage gated Ca2+ channel ( Wang et al . , 2002; Kaeser et al . , 2011; Liu et al . , 2011a; Davydova et al . , 2014 ) , though with clearly lower affinity than for Aplip1 . A point mutation which disrupts the Aplip1-RBP interaction provoked a ‘premature’ capture of RBP and the co-transported BRP at the axonal membrane , thus forming ectopic but , concerning T-bar shape and BRP/RBP arrangement , WT-like AZ scaffolds . The Aplip1 orthologue Jip1 has been shown to homo-dimerize via interaction of its SH3 domain ( Kristensen et al . , 2006 ) . Thus , the multiplicity of interactions , with Aplip1 dimers binding to two SH3 domains of RBP as well as to KLC , might form transport complexes of sufficient avidity to ensure tight adaptor–cargo interaction and prevent premature capture of the scaffold components . Our intravital imaging experiments showed that within axons RBP and BRP are co-transport in shared complexes together with Aplip1 , whereas we , despite efforts , were unable to detect any co-transport of other AZ scaffold components , that is , Syd-1 or Liprin-α with BRP/RBP ( not shown ) . In addition , STED analysis of axonal aggregates in srpk79D mutants showed BRP/RBP in stoichiometric amounts , but also failed to detect other AZ scaffold components . Moreover , BRP and RBP co-aggregated in the axoplasm of several other transport mutants we tested ( acsl , unc-51 , appl , unc-76 ) , consistent with both proteins entering synaptic AZ assembly from a common transport complex . Of note , during AZ assembly at the NMJ , BRP incorporation is invariably delayed compared to the ‘early assembly’ phase which is driven by the accumulation of Syd-1/Liprin-α scaffolds ( Fouquet et al . , 2009; Owald et al . , 2010 , 2012 ) . As the early assembly phase is , per se , still reversible ( Owald et al . , 2010 ) , the transport of ‘stoichiometric RBP/BRP complexes’ delivering building blocks for the ‘mature scaffold’ might drive AZ assembly into a mature , irreversible state ( Owald et al . , 2010 ) , and seems mechanistically distinct from early scaffold assembly mechanisms . Previous work suggested that AZ scaffold components ( Piccolo , Bassoon , Munc-13 and ELKS ) in rodent neurons are transported to assembling synapses as ‘preformed complexes’ , so-called Piccolo-Bassoon-Transport Vesicles ( PTVs; Zhai et al . , 2001; Shapira et al . , 2003; Maas et al . , 2012 ) . The PTVs are thought to be co-transported with SV precursors ( Ahmari et al . , 2000; Tao-Cheng , 2007; Bury and Sabo , 2011 ) anterogradely mediated via a KHC ( KIF5B ) /Syntabuli/Syntaxin-1 complex ( Cai et al . , 2007 ) and retrogradely via a direct interaction between Dynein light chain and Bassoon ( Fejtova et al . , 2009 ) . Since their initial description , however , further investigations of PTVs have been hampered by the apparent relative scarcity of PTVs , and by the lack of genetic or biochemical options for specifically interfering with their transport or final incorporation into AZs . Despite efforts we were not able to detect a direct interaction of Aplip1 and BRP although their common transport can be uncoupled from the presence of RBP . One possible explanation could be a direct interaction of Aplip1 to other AZ proteins that are co-transported together with BRP and RBP . It is interesting that the very C-terminus of BRP is essential for SV clustering around the BRP-based AZ cytomatrix ( Hallerman et al . , 2010 ) . Thus , it is tempting to speculate that adaptor/transport complex binding might block premature AZ protein/SV interactions before AZ assembly , but further analysis will have to await more atomic details as we could gain for the RBP::Aplip1 interaction . The down-regulation of the motor protein KHC also provoked severe axonal co-accumulations of BRP and RBP but per se should leave the adaptor protein-AZ cargo interaction intact . In contrast to aplip1 , the axonal aggregations in khc mutants adapted irregular shapes most of the time , likely not representing T-bar-like structures . Thus , our data suggest a mechanistic difference when comparing the consequences between eliminating adaptor cargo interactions with a direct impairment of motor functions . Still , we cannot exclude that trafficking of AZ complexes naturally antagonizes their ability to assemble into T-bars . The idea that proteins/molecules are held in an inactive state till they reach their final target has been observed in many other cell types . For example , in the context of local translation control , mRNAs are shielded or hidden in messenger ribonucleoprotein particles during transport so that they are withheld from cellular processing events such as translation and degradation . Shielding is thought to operate through proteins that bind to the mRNA and alter its conformation while at the correct time or place the masking protein is influenced by a signal that alleviates its shielding effect ( Spirin , 1996; Johnstone and Lasko , 2001 ) . As another example , hydrolytic enzymes , for example , lysosomes , are transported as proteolytically inactive precursors that become matured by proteolytic processing only within late endosomes or lysosomes ( Ishidoh and Kominami , 2002 ) . Particularly relevant in the context of AZ proteins involved in exocytosis , the Habc domain of Syntaxin-1 folds back on the central helix of the SNARE motif to generate a closed and inactive conformation which might prevent the interaction of Syntaxin-1 with other AZ proteins during diffusion ( Dulubova et al . , 1999; Ribrault et al . , 2011 ) . Previously , genetic analysis of C . elegans axons forming en passant synapses suggested a tight balance between capture and dissociation of protein transport complexes to ensure proper positioning of presynaptic AZs . In this study , overexpression of the kinesin motor Unc-104/KIF1A reduced the capture rate and could suppress the premature axonal accumulations of AZ/SV proteins in mutants of the small , ARF-family G-protein Arl-8 . Interestingly , large axonal accumulations in arl-8 mutants displayed a particularly high capture rate ( Klassen et al . , 2010; Wu et al . , 2013 ) . Similarly , both aplip1 alleles exhibited enlarged axonal BRP/RBP accumulations . Thus , the capture/dissociation balance for AZ components might be shifted towards ‘capture’ in these mutants , consistent with the ectopic axonal T-bar formation . It is tempting to speculate that loss of Aplip1-dependent scaffolding and/or kinesin binding provokes the exposure of critical ‘sticky’ patches of scaffold components such as RBP and BRP . Such opening of interaction surfaces might increase ‘premature’ interactions of cargo proteins actually destined for AZ assembly , thus increase overall size of the cargo complexes by oligomerization between AZ proteins and , finally , promote premature capture and ultimately ectopic AZ-like assembly . On the other hand , the need for the system to unload the AZ cargo at places of physiological assembly ( i . e . , presynaptic AZ ) might pose a limit to the ‘wrapping’ of AZ components and ask for a fine-tuned capture/dissociation balance . Several mechanisms for motor/cargo separation such as ( i ) conformational changes induced by guanosine-5′-triphosphate hydrolysis , ( ii ) posttranslational modification as de/phosphorylation , or ( iii ) acetylation affecting motor-tubulin affinity , have been suggested for cargo unloading ( Hirokawa et al . , 2010 ) . Notably , Aplip1 also functions as a scaffold for JNK pathway kinases , whose activity causes motor-cargo dissociation . JNK probably converges with a mitogen-activated protein kinase ( MAPK ) cascade ( MAPK kinase kinase Wallenda phosphorylating MAPK kinase Hemipterous ) in the phosphorylation of Aplip1 , thereby dissociating Aplip1 from KLC . Thus , JNK signaling , co-ordinated by the Aplip1 scaffold , provides an attractive candidate mechanism for local unloading of SVs ( Horiuchi et al . , 2007 ) and , as shown here , AZ cargo at synaptic boutons . Our study further emphasises the role of the Aplip1 adaptor , whose direct scaffolding role through binding AZ proteins might well be integrated with upstream controls via JNK and MAP kinases . Intravital imaging in combination with genetics of newly assembling NMJ synapses should be ideally suited to further dissect the obviously delicate interplay between local cues mediating capturing and axonal transport with motor-cargo dissociation .
Fly strains were reared under standard laboratory conditions ( Sigrist et al . , 2003 ) on semi-defined medium ( Bloomington recipe ) . For all experiments both male and female larvae were used for analysis . The following genotypes were used: WT: +/+ ( w1118 ) . srpk79D: srpk79Datc/srpk79Datc ( unless otherwise noted ) . srpk79Dvn: srpk79Dvn/srpk79Dvn . srpk79Datc: srpk79Datc/srpk79Datc . brpDf/+;srpk79D: Df ( 2R ) BSC29/+; srpk79Datc/srpk79Datc . brpnull/brpDf; srpk79D: brp69/Df ( 2R ) BSC29; srpk79Datc/srpk79Datc . rbpDf/+;srpk79D: Df ( 3R ) S2 . 01/+; srpk79Datc/srpk79Datc . rbpnull/rbpDf; srpk79D: rbpSTOP1/Df ( 3R ) S201; srpk79Datc/srpk79Datc . aplip1ek4: aplip1ek4/aplip1ek4 . aplip1null: aplip1ex213/aplip1ex213 . aplip1 , gen . rescue: aplip1gen . rescue ( ex213 ) /aplip1gen . rescue ( ex213 ) . Aplip1 cDNA rescue: control: elav/+;;aplip1ex213/+ . aplip1null: elav/+;;aplip1ex213/aplip1ex213 . WT rescue: elav/+;UAS-Aplip1-WT/+;aplip1ex213/aplip1ex213 . AxxA1 rescue: elav/+;UAS-Aplip1-AxxA1/+;aplip1ex213/aplip1ex213 . brpDf/+;aplip1ek4: Df ( 2R ) BSC29/+; aplip1ek4/aplip1ek4 . brpnull/brpDf;aplip1ek4: brp69/Df ( 2R ) BSC29; aplip1ek4/aplip1ek4 . Ok6>+: OK6-Gal4/+ . OK6>Aplip1-RNAi;rbpDf/+: OK6-Gal4/UAS-aplip1-RNAi;Df ( 3R ) S2 . 01/+ . OK6>Aplip1-RNAi;rbpnull/Df: OK6-Gal4/UAS-aplip1-RNAi; rbpSTOP1/Df ( 3R ) S201 . acsl: acsl05847/acsl1 . unc51 ( atg-1 ) : atg1ey07351/Df ( 3L ) BSC10 . appl: applBG0264/appl Df ( 1 ) yT7-518 . unc-76: unc-76G0158/y . Aplip1GFP , BRP-shortstraw: OK6-Gal4/UAS-BRP-shortstraw;UAS-Aplip1GFP/+ . Aplip1GFP , RBPcherry: OK6-Gal4/OK6-Gal4;UAS-Aplip1GFP/UAS-Aplip1GFP were crossed to upstream activator sequence ( UAS ) -RBPcherry/UAS-RBPcherry . BRPGFP , RBPcherry: OK6-Gal4/OK6-Gal4;genomicBRPGFP/genomicBRPGFP were crossed to UAS-RBPcherry/UAS-RBPcherry . Live imaging BRP-shortstraw in aplip1 mutant backgrounds ( Figure 2E ) : ctrl: OK6-Gal4/UAS-BRP-shortstraw . aplip1ek4: OK6-Gal4/UAS-BRP-shortstraw;aplip1ek4/aplip1ek4 . aplip1null: OK6-Gal4/UAS-BRP-shortstraw;aplip1ex213/aplip1ex213 . aplip1gen . rescue: OK6-Gal4/UAS-BRP-shortstraw;aplip1gen . rescue ( ex213 ) /aplip1gen . rescue ( ex213 ) . Ok6/+;UAS-KHC-RNAi . Ok6/+;UAS-Imac-RNAi . Stocks were obtained from: brp69 ( Kittel et al . , 2006 ) , Df ( 3R ) S2 . 01 and rbpSTOP1 ( Liu et al . , 2011a ) , aplip1ex213 and aplip1gen . rescue ( ex213 ) gift from Catherine Collins ( Klinedinst et al . , 2013 ) , srpk79Datc ( Johnson et al . , 2009 ) , srpk79Dvn ( Nieratschker et al . , 2009 ) , UAS-Aplip1GFP ( Horiuchi et al . , 2005 ) , UAS-BRP-shortstraw ( Schmid et al . , 2008 ) and genomic BRPGFP ( Matkovic et al . , 2013 ) . The aplip1ek4 , Df ( 2R ) BSC29 , acsl05847 , acsl1 , atg1ey07351 , applBG0264 , appl Df ( 1 ) yT7-518 , Df ( 3L ) BSC10 , unc-76G0158 lines were provided by the Bloomington Drosophila Stock Center . UAS-Aplip1-RNAi , UAS-Imac-RNAi and UAS-KHC-RNAi from VDRC . RBP cDNA was assembled based on exon annotation sequence of RBP-PF isoform from flybase . cDNA clones , AT04807; RH38268 and a gene synthesis fragment from MWG eurofins GMBH , Germany , containing 1–1131 bp of RBP-PF isoform were used to assemble the cDNA . All the fragments were cloned into a modified pENTR4 cloning vector described in Fouquet et al . ( 2009 ) . The final pENTR4 construct contains 5499 bp RBP cDNA was recombined with pTW-Cherry gateway Drosophila transgenic vector . Transgenic flies were generated at Bestgene Inc . , CA , USA and insertion was confirmed by genotyping . To generate the cDNA of Aplip1 ( with WT or mutated first PXXP motif ) , the full length cDNA clone of Aplip1 was kindly obtained from HYBRIGENICS Services , France and used as a template for cloning full length Aplip1 into pENTR/D-Topo ( Invitrogen , Germany ) using the following primers: Aplip1-FL-FW 5′-CACCATGGCCGACAGCGAATTCGAGGAGTT-3′ Aplip1-FL-REV 5′-TCGGCGCGCCCACCCTTCTACTCAATGTAG-3′ Through Gateway reaction , the construct was shuttled into GAL4/UAS vector and sent for injection at BestGene Inc . , CA , USA . Point mutations were introduced into the constructs via mutated primers with the ‘Quick Change II Site-Directed Mutagenesis Kit’ from Stratagene , CA , USA . This induced a change of the prolines of PxxP1 ( 155-PEIP-160 ) into alanines ( 155-AEIA-160 ) after mutagenesis . Following primers were used: Forward 5′ CGTCGTCGCAAGTTGGCGGAAATAGCGAAAAACAAGAAATCT 3′ Reverse 5′ AGATTTCTTGTTTTTCGCTATTTCCGCCAACTTGCGACGACG 3′ For crystallography constructs comprising either the RBP SH3-II ( residue 1318–1382 ) or SH3-III ( residue 1441–1507 ) domain of RBP were amplified by PCR and cloned into the pGEX-6P1 vector using EcoRI and XhoI restriction sites . The following primers were used: SH3II_for 5′-CAGAATTCCGCTATTTTGTGGCCATGTTC-3′ SH3II_rev 5′-TACTCGAGTCACTCCACCTCGGAGACCAT-3′ SH3III_for 5′-CAGAATTCAACATGCCCGTGAAGCGAATG-3′ SH3III_rev 5′-TACTCGAGTCAGTCCGCCAGGAAGTTAGA-3′ The resulting constructs comprise an N-terminal GST-tag that is followed by a PreScission cleavage site and the respective SH3 domain . Correctness of the DNA sequences was confirmed by DNA sequencing . The Yeast two-Hybrid screen for RBP interaction partners was carried out in collaboration with HYBRIGENICS Services , France using the LexA system ( pB27 with bait; pP6 vector with prey ) against the HYBRIGENICS Drosophila melanogaster head ( adult ) library . The vector maps of the bait and prey vectors are confidential ( protected under material transfer agreement ) . The plasmids ( pP6 and pB27 ) encode tryptophan ( Trp ) and leucine ( Leu ) biosynthesis genes , and were successfully double transformed into the TATA strain lacking genes for synthesis of Leu and Trp which can be followed by positive growth in LT media . Reporter genes for the protein–protein interaction are HIS3 , which can be later detected by growth on plates lacking histidin , as well as lacZ which allows the detection of interaction in a more quantitative fashion with a β-galactosidase assay . To transform the yeast cells with the pP6 and pB27 vector respectively the LiAc/single strand DNA/PEG technique was used ( Gietz and Schiestl , 2007 ) . The RBP constructs for Y2H were cloned into pB27 bait vector . The RBP cDNA clone AT04807 ( Drosophila Genomics Resource Centre , IN , USA ) was used as a template for PCR reaction . For amplification the following primers were used: 5′-CAGAATTCGGTCAACCGGGACAACCGGGG-3′ 5′-TAACTAGTTCAGTCGGGCGCGTCCGCCAGGA-3′ The assay was carried out as described in JH Miller ‘Experiments in Molecular Genetics’ 1972 Cold Spring Harbor Laboratories pages 352–355 . The RBP constructs for Y2H were cloned into pB27 bait vector . The RBP cDNA clone AT04807 ( Drosophila Genomics Resource Centre ) was used as a template for PCR reaction . For amplification the following primers were used: IP of elav-Gal4/+;UAS-Aplip1GFP/+ was performed as described in Depner et al . ( 2014 ) . In brief , the experiment was performed as following , 500 µl adult fly heads were mechanically homogenized in 500 µl lysis buffer ( 50 mM Tris pH 8 . 0 , 150 mM KCl , 1 mM MgCl2 , 1 mM EGTA , 10% glycerol containing protease inhibitor cocktail [Roche , Germany] ) . 0 . 4% Sodium deoxycholate was added , and the lysate was incubated on ice for 30 min . The lysate was diluted 1:1 with sodiumdeocycholat-free lysis buffer , then 1% Triton X-100 was added and lysate was kept on the wheel at 4°C for 30 min . After centrifugation for 15 min at 16 , 000×g , the supernatant was used in IPs with GFP-Trap-A beads and blocked agarose beads as binding control ( Chromotek , Germany ) . After incubation overnight at 4°C , beads were washed in buffer without detergent and glycerol . Proteins were eluted from the beads with SDS sample buffer . Afterward , the SDS-PAGE samples were subjected to Western blot ( WB ) . The gel electrophoresis for both SDS-PAGE and Tris-acetate gels was conducted according to the standard protocols ( Laemmli , 1970; Schägger , 2006 ) . Colloidal Coomassie blue stain was used to detect proteins based on manufacture protocol ( Carl‐Roth , Germany and Invitrogen ) . For BRP , RBP and Aplip1 , standard SDS-PAGE gels ( 6–12% ) were used to separate the target protein . Following the separation by gel electrophoresis , the proteins were transferred into a nitrocellulose membrane by wet transfer procedure using cold transfer buffer ( 25 mM Tris , pH 8 . 0 , 150 mM glycine , 20% methanol ) . For visualization of proteins , the membrane was stained using Ponceau-S staining solution ( Sigma–Aldrich , MO , USA ) . 5% milk powder in phosphate buffered saline ( PBS ) was used for blocking of the membrane . Following the blocking , the membrane was incubated with the primary Abs guinea pig BRPLast200 ( 1:5000 , Ullrich et al . , in submission ) and rabbit RBPSH3-II+III ( 1:1000 , Depner et al . , 2014 ) at 4°C for overnight . After several washing steps , the membrane was incubated with horseradish peroxidase ( HRP ) conjugated secondary Abs ( Dianova , Germany ) . For detection , an enhanced chemoluminescence substrate ( GE Healthcare , United Kingdom ) was used and the X-ray film ( GE Healthcare ) development was carried manually . Larval filets were dissected and stained as described previously ( Owald et al . , 2010 ) . The following primary Abs were used: rabbit BRPN-term ( 1:500; Qin et al . , 2005 ) ; rabbit Liprin-α ( 1:500; Fouquet et al . , 2009 ) ; rabbit Syd-1 ( 1:500; Owald et al . , 2010 ) ; rabbit Rab3 ( 1:500; Graf et al . , 2009 ) ; rabbit RBPC-term , rabbit RBPSH3-II+III ( 1:500; Depner et al . , 2014 ) ; rabbit Syt1-CL1 ( 1:1000; gift from N Reist [Mackler et al . , 2002] , Colorado State University , CO , USA ) ; mouse GFP ( 3E6 ) ( 1:500 , Life Technologies , Germany ) , mouse Nc82 = anti-BRPC-term ( 1:100 , Developmental Studies Hybridoma Bank , University of Iowa , Iowa City , IA , USA ) . Except for staining against CacGFP , where larvae were fixed for 5 min with ice-cold methanol , all fixations were performed for 10 min with 4% paraformaldehyde in 0 . 1 mM PBS . Secondary Abs for standard immunostainings were used in the following concentrations: goat anti-HRP-Cy5 ( 1:250 , Jackson ImmunoResearch , PA , USA ) ; goat anti-rabbit Cy3 ( 1:500 , Life Technologies ) ; goat anti-mouse Alexa-Fluor-488 ( 1:500 , Life Technologies ) . Larvae were mounted in vectashield ( Vector labs , United Kingdom ) . Secondary Abs for STED were used in the following concentrations: For Figures 1H , 7A: goat anti-mouse Atto594 ( 1:250 ) ; goat anti-rabbit Atto594 ( 1:250 ) ; goat anti-mouse Atto647N ( 1:100 ) , goat anti-rabbit Atto647N ( 1:100 ) ( ATTO-TEC , Germany ) . For Figure 7B: goat anti-mouse Atto590 ( 1:100 ) ; goat anti-rabbit star635 1:100 ( Atto590 [ATTO-TEC] and star635 [Abberior , Germany] ) coupled to respective IgGs ( Dianova , Germany ) . For Figure 7—figure supplement 1A–C: goat anti-mouse Alexa-Fluor-488 ( 1:500 , Life Technologies ) and goat anti-rabbit Alexa-Fluor-532 ( 1:500 , Life Technologies ) was used . For STED imaging larvae were mounted in Mowiol ( Max-Planck Institut for Biophysical Chemistry , Group of Stefan Hell ) or Prolong Gold antifade reagent ( Life Technologies; Figure 7—figure supplement 1A–C ) . Confocal microscopy was performed with a Leica TCS SP5 ( all except for Figure 4G–J and Figure 7G–J ) or a Leica SP8 ( Figure 4G–J and Figure 7G–J ) confocal microscope ( Leica Microsystems , Germany ) . STED microscopy was performed with a custom-built STED-microscope ( see below ) . Images of fixed and live samples were acquired at room temperature . Confocal imaging of axons was done using a z step of 0 . 25 μm . The following objective was used: 63× 1 . 4 NA oil immersion for NMJ confocal imaging . All confocal images were acquired using the LCS AF software ( Leica , Germany ) . Images from fixed samples were taken from third instar larval nerve bundles ( segments A1–A3 ) . Images for figures were processed with ImageJ software to enhance brightness using the brightness/contrast function . If necessary images were smoothened ( 0 . 5–1 pixel Sigma radius ) using the Gaussian blur function . Quantifications of axonal spot number and size were performed following an adjusted manual ( Andlauer and Sigrist , 2012 ) , briefly as follows . The signal of a HRP-Cy5 Ab was used as template for a mask , restricting the quantified area to the shape of the axon/nerve bundles . The original confocal stacks were converted to maximal projections and after background subtraction , a mask of the axonal area was created by applying a certain threshold to remove the irrelevant lower intensity pixels . The segmentation of single spots was done semi-automatically via the command ‘Find Maxima’ and by hand with the pencil tool and a line thickness of 1 pixel . To remove high frequency noise a Gaussian blur filter ( 0 . 5 pixel Sigma radius ) was applied . The processed picture was then transformed into a binary mask using the same lower threshold value as in the first step . This binary mask was then projected onto the original unmodified image using the ‘min’ operation from the ImageJ image calculator . The axonal spots of the resulting images were counted with the help of the ‘analyze particle’ function with a lower threshold set to 1 . The spot density was obtained by normalizing the total number of analyzed particles to the axonal area measured via HRP . Colocalization of RBP/BRP spots ( Figure 1G ) was counted manually . Data were analyzed using the Mann–Whitney U test for linear independent data groups . Means are annotated ±SEM . Asterisks are used to denote significance: *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001; n . s . ( not significant ) , p > 0 . 05 . For Figures 1H , 7A two-colour STED images were recorded with a custom-built STED microscope which combines two pairs of excitation and STED laser beams all derived from a single supercontinuum laser source ( Bückers et al . , 2011 ) . For Figure 7B STED microscopy was performed as previously described in Li et al . ( 2014 ) . Here , two-colour STED images were recorded on a custom-built STED-microscope ( Göttfert et al . , 2013 ) , which combines two pairs of excitation laser beams of 595 nm and 640 nm wavelength with one STED fiber laser beam at 775 nm . All STED images were acquired using Imspector Software ( Max Planck Innovation GmbH ) . STED images were processed using a linear deconvolution function integrated into Imspector Software ( Max Planck Innovation GmbH , Germany ) . Regularization parameters ranged from 1e−09 to 1e−10 . The point spread function ( PSF ) for deconvolution was generated by using a 2D Lorentz function with its half-width and half-length fitted to the half-width and half-length of each individual image . For Figure 7—figure supplement 1 , STED microscopy was performed with a Leica TCS SP5 time gated STED microscope equipped with a 100× 1 . 4 NA objective using the LCS AF software ( Leica ) for image acquisition . Alexa-Fluor-488 and Alexa-Fluor-532 were excited using a pulsed white light laser at 488 and 545 nm , respectively . STED was achieved with a continous STED laser at 592 nm . In gSTED mode time gated detection started at 1 . 2 ns–6 ns for Alexa488 while for Alexa532 gating time was set to 2 . 3 ns–6 ns . Raw gSTED images were deconvolved using the built-in algorithm of the LAS AF software ( Signal intensity; regularisation parameter 0 . 05 ) . The PSF was generated using a 2D Lorentz function with the full-width half maximum set to 60 nm . Images for figures were processed with ImageJ software to remove obvious background , enhance brightness/contrast and smoothened ( 1 pixel Sigma radius ) using the Gaussian blur function . Live imaging was performed as previously described ( Füger et al . , 2007 ) . Briefly , third instar larvae were put into a live imaging chamber and anaesthetized with 10–20 short pulses of a desflurane-air mixture until the heartbeat completely stopped . For assessing axonal transport , axons immediately after exiting the ventral nerve cord were imaged for 10 min using timelapse confocal microscopy . The flux was determined by manually counting the number of moving spots ( unidirectional for >3 frames ) passing a virtual line in the middle of the nerve bundle . Mean flux was calculated by pooling results from at least three independent larvae and at least six nerves . If little or no flux was observed , additional nerves were imaged to avoid any bias from selecting specific nerves . ITC experiments were performed at 25°C on an iTC200 microcalorimeter ( Malvern Instruments Ltd . , United Kingdom ) . The same peptides were employed as used for the co-crystallization experiments ( see below ) . Lyophilized peptides were resuspended in the final buffer of the proteins ( 10 mM Tris-HCl pH 7 . 4 , 100 mM NaCl ) . RBP SH3-II and SH3-III were both provided at a concentration of 150 µM , RBP SH3-II+III was provided at 78 µM . The proteins were titrated with 16 injections of 2 . 5 µl of either Aplip1 , Cac , RIM1 or RIM2 peptide at a concentration of 2 mM with 2-min intervals . The released heat was obtained by integrating the calorimetric output curves . Binding parameters were calculated using the Origin5 software using the ‘One Set of Sites’ curve fitting model provided by the software . Protein expression was conducted using chemically competent Escherichia coli BL21-CodonPlus-RIL cells . The cells were grown in autoinduction ZY-medium ( Studier , 2005 ) with ampicillin and chloramphenicol for 4 hr at 37°C . Afterwards , the temperature was decreased to 18°C , and the cells were grown overnight . The cells were harvested by centrifugation at 8 , 000×g for 6 min . The cell pellet was resuspended in resuspension buffer ( 40 mM Tris/HCl pH 7 . 5 at RT , 250 mM NaCl , 1 mM DTT , 10 mg/l lysozyme and 5 mg/l DNase I ) and subsequently lysed by sonification for 20 min . The lysate was centrifuged at 56 , 000×g for 45 min to pellet the cell debris . The supernatant was applied for affinity chromatography using 10 ml glutathione sepharose 4B ( GE Healthcare ) . Hereafter , two washing steps were performed using 80 ml washing buffer ( 20 mM Tris/HCl pH 7 . 5 at RT , 250 mM NaCl , 1 mM DTT ) for each step . The GST-tag of the respective SH3 domain was cleaved off on the beads using PreScission protease ( 1 mg/ml ) . Therefore 40 ml washing buffer with PreScission protease in a molar ratio of 1:30 to the maximum loading capacity of the glutathione sepharose were incubated with the beads at 4°C while gently rotating overnight . The PreScission-cleaved constructs were purified using a Superdex 75 26/60 column ( GE Healthcare ) . The protein containing fractions were pooled and concentrated using a 3 kDa molecular weight cut-off concentrator ( Millipore , Germany ) . Protein concentrations were determined by UV-absorption . For crystallization experiment the RBP SH3-II was concentrated to 56 mg/ml and the RBP SH3-III to 62 mg/ml . The same peptides as for ITC measurements were used and synthesized at the Leibniz Institute for Molecular Pharmacology with N-terminal acetylation and C-terminal amidation . The unsolubilized peptides were mixed in a fivefold molar excess with the protein solution and incubated for 2 hr on ice . Insoluble peptide was removed by centrifugation ( 16 , 000×g for 1 min ) prior to crystallization experiments . All crystallization experiments were carried out at 291 K in a sitting drop setup . Crystals of RBP SH3-II bound to the Aplip1 peptide were obtained over a reservoir solution composed of 2 . 2–2 . 6 M ammonium sulfate , 0 . 1 M bicine with final pH 9 . For cryoprotection , the crystals were transferred to a reservoir solution supplemented with 25% ( vol/vol ) glycerol . Crystals of RBP SH3-II bound to Cac were obtained over a reservoir solution of 0 . 2 M Ca ( Ac ) 2 , 0 . 1 M MES pH 6 . 0 , and 20% ( wt/vol ) polyethylenglycol ( PEG ) 8000 . For cryoprotection , the crystals were transferred to a reservoir solution supplemented with 15% ( vol/vol ) PEG 400 . Crystals of RBP SH3-III bound to the Cac peptide appeared over a reservoir solution of 0 . 2 M Li2SO4 , 0 . 1 M MES pH 6 . 5 , and 30% ( vol/vol ) PEG 400 . After cryoprotection the crystals were flash-cooled in liquid nitrogen . Synchrotron diffraction data were collected at the beamline 14 . 2 of the MX Joint Berlin laboratory at BESSY ( Berlin , Germany ) . X-ray data collection was performed at 100 K . Diffraction data were processed with the XDS package ( Kabsch , 2010 ) . The diffraction data of RBP SH3-II/Aplip1-PxxP1 were initially indexed in P622 . Cumulative intensity distribution analysis as well as calculation of the moment of the observed intensity/amplitude distribution performed with PHENIX . XTRIAGE and POINTLESS ( Evans , 2011 ) indicated an unusual intensity distribution , likely caused by twinning . For determination of the correct space group , the diffraction data were processed in P1 . Subsequently , the structure was solved by molecular replacement with the program PHASER ( McCoy et al . , 2007 ) . We used the NMR structure of the SH3-II domain of human RBP ( PDB entry 2CSQ ) as search model and could locate 24 copies of the SH3 domain . Next the diffraction data and the coordinates of our molecular replacement were analysed by the program ZANUDA ( Lebedev and Isupov , 2014 ) revealing that sixfold is in fact broken and C2 is the true symmetry , with sixfold twinning with the six twin operators: h , h , l; h , −k , −l; 1/2h − 3/2k , −1/2h − 1/2k , −k; −1/2h + 3/2k , 1/2h + 1/2k , −l; −1/2h − 3/2k , −1/2h + 1/2k , −l and 1/2h + 3/2k , 1/2h − 1/2k , −l . In total we could locate in the asymmetric unit 12 copies of RBP SH3-II bound to Aplip1-PxxP1 . The crystals of RBP SH3-II and SH3-III bound to the Cac peptide have P21 and I222 symmetry , respectively . Analyses of the diffraction data of the complex of RBP SH3-II and Cac revealed one pseudo-merohedral twin operator ( h , −k , −h − l ) , that was later included in the refinement protocol . The structures of RBP SH3-II and SH3-III each bound to the Cac derived peptide were solved by molecular replacement with our previously determined structure of RBP SH3-II . The asymmetric unit of RBP SH3-II bound to Cac contains 10 complexes and of RBP SH3-III bound to Cac one complex , respectively . The refined molecular replacement solution clearly revealed the presence of the bound Aplip1-PxxP1 peptide in 2mFo − DFc and mFo − DFc electron density maps . For refinement , a set of 4 . 7% of Rfree reflections was generated in P622 and then expanded to C2 to insure equal distribution of the Rfree reflections in all six twin domains . For calculation of the free R-factor of the other two data sets , a randomly generated set of 5% of the reflections from the diffraction data set was used and excluded from the refinement . The structure was manually built in COOT ( Emsley et al . , 2010 ) and refined in REFMAC 5 . 8 . 0073 ( Murshudov et al . , 2011 ) with intensity based twin refinement . In final stages TLS refinement was applied with every protein and peptide chain as single TLS group . The structures with bound Cac peptide were refined with PHENIX . REFINE ( Adams et al . , 2010; Afonine et al . , 2012 ) . Water molecules were picked with COOT and manually inspected . All structures were evaluated with MOLPROBITY ( Chen et al . , 2010 ) and PROCHECK ( Laskowski et al . , 1993 ) . Figures were drawn with PYMOL ( DeLano , 2002 ) . Conventional embedding was performed as described previously ( Fouquet et al . , 2009 ) . Data were analyzed using the Mann–Whitney rank sum test for linear independent data groups ( Prism; GraphPad Software , Inc . ) . Means are annotated ± SEM . Asterisks are used to denote significance ( *p < 0 . 05; **p < 0 . 01; ***p < 0 . 005; not significant , p > 0 . 05 ) . | To pass on information , the neurons that make up the nervous system connect at structures known as synapses . Chemical messengers called neurotransmitters are released from one neuron , and travel across the synapse to trigger a response in the neighbouring cell . The formation of new synapses plays an important role in learning and memory , but many aspects of this process are not well understood . In a specific region of the synapse called the active zone , a scaffold of proteins helps to release the neurotransmitters . These proteins are made in the cell body of the neuron , and are then transported to the end of the long , thin axons that protrude from the cell body . This presents a challenge for the cell , because the components of the active zone scaffold must be correctly targeted to the synapse at the end of the axon , ensuring the active zone scaffold assembles only at its proper location . Siebert , Böhme et al . studied how some of the proteins that are found in the active zone scaffold of the fruit fly Drosophila are transported along axons . Labelling the proteins with fluorescent markers allowed their movement to be examined under a microscope in living Drosophila larvae . The results showed that two of the proteins—known as BRP and RBP—are transported along the axons together . Further investigation revealed that a transport adaptor protein called Aplip1 , which binds to RBP , is required for this movement . Siebert , Böhme et al . established the structure of the part of RBP where this interaction occurs , and found that mutating this region causes premature active zone scaffold assembly in the axonal part of the neuron . The interaction between RBP and Aplip1 is very strong , and this helps to prevent the scaffold assembling before it has reached the correct part of the neuron . Exactly how the transport adaptor and active zone protein are separated once they reach their final destination ( the synapse ) remains to be discovered . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | A high affinity RIM-binding protein/Aplip1 interaction prevents the formation of ectopic axonal active zones |
Inflammation modifies risk and/or severity of a variety of brain diseases through still elusive molecular mechanisms . Here we show that hyperactivation of the interleukin 1 pathway , through either ablation of the interleukin 1 receptor 8 ( IL-1R8 , also known as SIGIRR or Tir8 ) or activation of IL-1R , leads to up-regulation of the mTOR pathway and increased levels of the epigenetic regulator MeCP2 , bringing to disruption of dendritic spine morphology , synaptic plasticity and plasticity-related gene expression . Genetic correction of MeCP2 levels in IL-1R8 KO neurons rescues the synaptic defects . Pharmacological inhibition of IL-1R activation by Anakinra corrects transcriptional changes , restores MeCP2 levels and spine plasticity and ameliorates cognitive defects in IL-1R8 KO mice . By linking for the first time neuronal MeCP2 , a key player in brain development , to immune activation and demonstrating that synaptic defects can be pharmacologically reversed , these data open the possibility for novel treatments of neurological diseases through the immune system modulation .
Neurological disorders represent an enormous source of burden to the individual and to society , with many patients failing to respond to available medication . Growing evidence on genetic components of neurological diseases have been collected during recent years; notably , these genes overwhelmingly point to disorders of synaptic transmission , which led to the coinage of the term ‘synaptopathy’ to indicate a brain disease originating from a dysfunction of the synapse ( Grant , 2012; Grabrucker , 2014 ) . Disruption of synapse function may also be caused by environmental stimuli , with inflammatory cytokines affecting synaptic transmission and modifying the risk and severity of a variety of brain diseases , including autism spectrum disorders , schizophrenia and cognitive disabilities ( Hagberg et al . , 2012; Chugh et al . , 2013; Steinman , 2013 ) . Among the cytokines known to affect synaptic function , the proinflammatory cytokine IL-1β plays a critical role . IL-1β was found to impair brain-derived neurotrophic factor ( BDNF ) -induced expression of molecules critical for activity-dependent synaptic plasticity , including cAMP response element binding protein ( CREB ) , Arc , and cofilin , thus reducing actin polymerization and impairing spine morphogenesis ( Tong et al . , 2012 ) . In addition , IL-1β controls different neuronal functions , including excitability and transmitter release , via multiple biochemical pathways ( Weber et al . , 2010 ) . However , no evidence has been reported yet that IL-1β may affect synapse development and function acting on molecular pathways known to be at the root of synaptopathies . Genetic mouse models of immune deregulation may serve as reliable and reproducible systems for examination of the effects of inflammation on synapse structure and function and elucidation of the molecular processes involved . IL-1R8 , also known as single Ig IL-1-Related Receptor ( SIGIRR ) or TIR8 , belongs to the toll-like receptors ( TLRs ) and interleukin-1R-like receptors ( ILRs ) , a family of conserved proteins involved in immunity and inflammation ( Riva et al . , 2012; Garlanda et al . , 2013a ) . TLRs are receptors able to recognize specific pathogen-associated patterns ( PAMPs ) and necrotic cell-derived danger signals ( DAMPs ) and act as sensors for microorganisms and tissue damage , whereas the IL-1R subfamily includes components of signalling receptor complexes as well as molecules with regulatory function . IL-1R8 dampens the activation of the TLRs and IL-1R signalling pathways by intracellularly interfering with the association of adaptor molecules to the receptor complexes , including NF-κB and JNK ( Riva et al . , 2012; Garlanda et al . , 2013b ) . As a consequence , IL-1R8-deficient mice display exaggerated symptoms of inflammatory conditions ( Garlanda et al . , 2007; Gulen et al . , 2010; Drexler et al . , 2010 ) and demonstrate pronounced susceptibility to the inflammatory challenge posed by microbial LPS ( Garlanda et al . , 2004 ) . IL-1R8 is also present in the brain ( Costelloe et al . , 2008; Polentarutti et al . , 2003 ) . Genetic deficiency for IL-1R8 is associated with inflammatory changes in the brain , including increased levels of LPS-induced tumor necrosis factor α ( TNFα ) and IL-6 in microglia , higher expression of TLR4 , and NF-κB activation ( Watson et al . , 2010 ) . Reduced IL-1R8 expression has been described in patients affected by psoriatic arthritis ( Batliwalla et al . , 2005 ) while SIGIRR variants ( characterized by defective SIGIRR function ) have been found in humans in association with necrotizing enterocolitis ( Sampath et al . , 2015 ) and with systemic lupus erythematosus ( SLE; [Zhu et al . , 2014] ) , all pathologies being characterized by cognitive defects and neurodevelopmental impairment ( Husted et al . , 2013; Rees et al . , 2007; Calderón et al . , 2014; Muscal et al . , 2010 ) . Notably , IL-1R8-deficient mice are impaired in novel object recognition , spatial reference memory and long-term potentiation ( LTP ) , defects that occur in the absence of any external inflammatory stimuli ( Costello et al . , 2011 ) . However , the molecular mechanisms by which IL-1R8 deficiency results in brain defects are still completely unknown . We used IL-1R8 deficient mice to investigate whether genetic hyperactivation of the IL-1R pathway affects synapse function impinging on molecular players involved in synaptopathies . We demonstrate that the activity of the IL-1R pathway directly affects , in neurons , the levels of expression of the methyl-CpG-binding protein 2 ( MeCP2 ) , a synaptopathy protein involved in neurological diseases -Rett syndrome and MeCP2 duplication syndrome- characterized by defective plasticity , impaired cognition and intellectual disability . We also show that pharmacological inhibition of IL-1R activity normalizes MeCP2 expression and cognitive deficits in IL1R8-deficient mice .
To investigate the impact of IL-1R8 deficiency on synapse structure and function , we examined the morphology and plasticity of dendritic spines in primary cultures established from embryonic IL-1R8 KO or WT mice hippocampi , transfected with GFP at DIV 12 . Compared to their WT counterpart , IL-1R8 KO neurons displayed an increased number of immature , thin spines and a decreased number of mature , mushroom spines ( Figure 1A–C ) along with a significant reduction of spine width ( Figure 1D ) . Also PSD-95 puncta density ( Figure 1E ) and size ( Figure 1F ) were significantly lower in IL-1R8 KO neurons . Consistently , the levels of the postsynaptic protein PSD-95 , evaluated by western blotting of culture homogenates were significantly reduced ( Figure 1G and H ) . In line with a synaptic defect , patch clamp recording of IL-1R8 deficient cultures revealed that the frequency , but not the amplitude , of miniature excitatory postsynaptic currents ( mEPSCs ) was significantly reduced ( Figure 1I–K ) . Spine defects ( Figure 2A and B ) and reduction in synaptic markers ( Figure 2C and D ) were also detected in CA1 pyramidal neurons of IL1R8 KO mice with respect to age-matched WT controls . 10 . 7554/eLife . 21735 . 003Figure 1 . IL-1R8 silencing affects spine morphology and function . ( A ) PSD-95 immunocytochemical staining of GFP-transfected , 16 DIV hippocampal cultures from WT or IL-1R8 KO mice . Scale bar 5 mm . ( B-F ) Quantitative analysis of the following parameters: thin and mushroom spine density ( B and C ) ; spine width ( D ) ; PSD-95 puncta density ( E ) and mean size of PSD-95 puncta ( F ) . Number of analyzed neurons: B-D: 32 ( WT ) , 44 ( IL-1R8 KO ) ; E-F: 32 ( WT ) , 29 ( IL-1R8 KO ) ; Student t test . ( G , H ) Western blotting analysis of PSD-95 levels in primary hippocampal cultures , 3 independent experiments , Mann Whitney test . ( I ) Representative mEPSC traces recorded from WT and IL-1R8 KO neurons . ( J ) mEPSC frequency quantitation ( WT: n = 12; IL-1R8 KO: n = 18; Mann Whitney Test ) . ( K ) Cumulative distributions and bar graph of mEPSC amplitude ( WT: 22 , 82 ± 2 , n = 12; IL-1R8 KO: 22 , 11 ± 1 , 5 , n = 18; Mann Whitney test . * indicates significance compared to WT , # indicates , significance compared to IL-1R8 KO . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 00310 . 7554/eLife . 21735 . 004Figure 2 . IL1R8 deficient mice show altered spines and synapses in hippocampal sections . ( A ) Representative images of secondary branches of apical dendrites of WT and IL-1R8 KO mice ( 3 months old ) stained by the Golgi-Cox method and relative quantitation ( B ) . A significant reduction of spine density in IL-1R8 KO mice was evident with respect to WT mice ( number of spines per micron: WT = 1 , 02 ± 0 , 03; number of mice analyzed: 3 , number of examined dendrites: 75; IL-1R8 KO = 0 , 80 ± 0 , 02; number of mice analyzed: 3 , number of examined dendrites: 84; Mann-Whitney test ) . Scale bar , 5 µm . ( C ) Representative fields of the CA1 hippocampal region ( stratum radiatum ) of a WT and IL-1R8 KO mouse brain ( 1 month old mice ) , stained for the vesicular glutamate transporter , vGLUT1 . Scale bar , 15 µm . ( D ) A significant reduction in vGLUT1 area was found in the stratum radiatum of CA1 field of IL-1R8 KO mice ( total area of vGLUT1 positive puncta WT = 0 , 4644 ± 0 , 03420; number of examined fields: 35; IL-1R8 KO = 0 , 2910 ± 0 , 01984; number of examined fields: 47; Mann-Whitney test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 004 To investigate whether IL-1R8 KO neurons are able to undergo synaptic potentiation , hippocampal cultures were subjected to an established chemical LTP ( c-LTP ) protocol based on the culture exposure to 100 μM glycine in KRH devoid of Mg , followed by a washout and recovery in neuronal medium for at least 60 min ( Menna et al . , 2013 ) . Under these conditions , a significant increase in the density of both PSD-95-positive puncta and mushroom spines occurred in WT but not in IL-1R8 KO neurons ( Figure 3A–C ) . Similarly , no increase in mEPSC amplitude and frequency was recorded over time in IL-1R8 KO neurons ( Figure 3D–F ) , univocally indicating that hippocampal IL-1R8 KO neurons are unable to undergo synaptic plasticity . These data show the occurrence of synaptic structural and functional defects in primary cultures from IL-1R8 KO mice . 10 . 7554/eLife . 21735 . 005Figure 3 . IL-1R8 KO neurons do not undergo LTP . ( A ) PSD-95 immunocytochemical staining of GFP-transfected , DIV 16 hippocampal cultures from WT or IL-1R8 KO mice , subjected or not to the LTP protocol . Scale bar 5 μm . ( B and C ) Quantitative analysis of PSD-95 and mushroom spine density of neurons treated as above . Number of analyzed neurons , B: 15 ( WT , no LTP ) , 13 ( WT , + LTP ) , 28 ( IL-1R8 KO , no LTP ) , 18 ( IL-1R8 KO , + LTP ) ; C: 16 ( WT , no LTP ) , 33 ( WT , + LTP ) , 24 ( IL-1R8 KO , no LTP ) , 34 ( IL-1R8 KO , + LTP ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test ) . ( D ) Representative traces of mEPSCs recorded from neurons of WT or IL-1R8 KO mice before and 60 min after LTP induction . ( E and F ) Averaged mEPSC frequency and amplitude of WT and IL-1R8 KO neurons over different recording time points after LTP induction . Normalized mEPSC frequency: WT 0 min: 0 , 99 ± 0 , 06 , n = 11; 20 min 1 , 84 ± 0 , 3 , n = 12; 60 min 1 , 81 ± 0 , 14 , n = 9; IL-1R8 KO 0 min 1 , 0 ± 0 , 07 , n = 10; 20 min 1 , 11 ± 0 , 15 , n = 11; 60 min 0 , 94 ± 0 , 07 , n = 5 . Normalized mEPSC amplitude: WT 0 min: 0 , 99 ± 0 , 03 , n = 11; 20 min 1 , 29 ± 0 , 14 , n = 12; 60 min 1 , 36 ± 0 , 14 , n = 9; IL-1R8 KO 0 min 0 , 96 ± 0 , 05 , n = 10; 20 min 0 , 91 ± 0 , 07 , n = 11; 60 min 0 , 99 ± 0 , 1 , n = 5 . Mann Whitney test . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 005 To determine whether the synaptic defects of IL-1R8 KO neurons are attributable to IL-1 receptor ( IL-1R ) or TLR pathway , both negatively regulated by IL-1R8 , we analyzed spine density and electrophysiological properties in neurons from mice deficient for both IL-1R8 and IL-1R ( IL-1R8 KO IL-1R KO ) . Double IL-1R8 KO IL-1R KO neurons displayed an increase in mushroom spine density ( number of spines/micron , mean ± SEM , WT: 0 , 1581 ± 0 , 01508 , n = 32; IL-1R8 KO IL-1R KO: 0 , 2585 ± 0 , 0154 , n = 41 , Student t test , p<0 , 0001 ) , accompanied by enhanced mEPSCs frequency and amplitude ( mEPSC frequency , mean ± SEM , WT: 1317 ± 0 , 1714 , n = 12; IL-1R8 KO IL-1R KO: 2432 ± 0 , 3187 , n = 17 , Student t test , p<0 , 05; mEPSC amplitude , mean + SEM , WT: 22 , 82 ± 2468 , n = 12; IL-1R8 KO IL-1R KO: 29 , 10 ± 2112 , n = 17 , Student t test , ns , p=0 , 0644 ) . These data suggest a role for IL-1R signaling in controlling synaptic structure and function . To determine if reducing IL-1R activity could restore synaptic plasticity and spatial learning , hippocampal neurons from IL-1R8 KO mice exposed overnight to IL1Ra ( Anakinra ) , a naturally-occurring IL-1 receptor antagonist ( Dinarello , 2009 ) , displayed increased density of both mushroom spines ( Figure 4A and B ) and PSD-95 puncta ( Figure 4A and C ) . Furthermore , IL1Ra-treated neurons from IL-1R8 KO mice recovered their ability to undergo both structural and functional LTP , as indicated by the increased density of spines and PSD-95 ( Figure 4A–C ) and mEPSC frequency and amplitude ( Figure 4D–F ) following the application of the c-LTP protocol . Therefore , the IL-1R8 KO synaptic phenotype results from hyperactivation of the IL-1R pathway as a result of IL-1R8 silencing . Consistently , pharmacological activation of IL-1R pathway in WT neurons through overnight treatment with IL-1β ( 40 ng/ml for 14 hr ) resulted in an increased number of immature thin spines and a decreased number of mature mushroom-type spines ( Figure 4G–I ) and PSD-95 puncta ( Figure 4G and J ) , accompanied by inability to undergo LTP ( Figure 4G–J ) . 10 . 7554/eLife . 21735 . 006Figure 4 . Inhibition of IL-1R signalling restores LTP in IL-1R8 KO neurons . ( A ) PSD-95 immunocytochemical staining of 16 DIV hippocampal cultures from GFP transfected WT or IL-1R8 KO mice , treated or not , at DIV 15 with IL1Ra ( 20 ng/ml ) overnight ( 14 hr ) . Scale bar 5 μm . ( B ) Quantitative analysis of mushroom spine density in control or upon LTP protocol application . Similar results were obtained with IL1Ra at 100 ng/ml . Number of analyzed neurons: 14 ( WT , no LTP ) , 16 ( WT , + LTP ) , 10 ( WT , IL1Ra , no LTP ) , 10 ( WT , IL1Ra , + LTP ) , 26 ( IL-1R8 KO , no LTP ) , 29 ( IL-1R8 KO , + LTP ) , 23 ( IL-1R8 KO , IL1Ra , no LTP ) , 32 ( IL-1R8 KO , IL1Ra , + LTP ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test ) . ( C ) Quantitative analysis of PSD-95 immunoreactivity . Number of analyzed neurons: 9 ( WT , no LTP ) , 26 ( WT , + LTP ) , 27 ( WT , IL1Ra , no LTP ) , 24 ( WT , IL1Ra , + LTP ) , 24 ( IL-1R8 KO , no LTP ) , 62 ( IL-1R8 KO , + LTP ) , 13 ( IL-1R8 KO , IL1Ra , no LTP ) , 12 ( IL-1R8 KO , IL1Ra , + LTP ) . One-way ANOVA analysis of variance followed by post hoc Tukey test . ( D ) Representative traces of WT and IL-1R8 KO neurons treated with vehicle or IL1Ra ( 100 ng/ml ) . ( E and F ) Quantitation of mEPSC frequency and amplitude recorded 60 min after LTP protocol in WT or IL-1R8 KO neurons , treated as above . Analysis of normalized mEPSC frequency and amplitude reveals that only WT neurons and IL1-Ra-treated IL-1R8 KO neurons undergo LTP . Number of recorded neurons: 6 ( WT , no LTP ) , 6 ( WT , + LTP ) , 8 ( WT , IL1Ra , no LTP ) , 8 ( WT , IL1Ra , + LTP ) , 9 ( IL-1R8 KO , no LTP ) , 6 ( IL-1R8 KO , + LTP ) , 8 ( IL-1R8 KO , IL1Ra , no LTP ) , 12 ( IL-1R8 KO , IL1Ra , + LTP ) , Mann-Whitney test . ( G ) Immunocytochemical staining for PSD-95 in GFP-transfected WT neurons treated or not with IL-1β ( 40 ng/ml , overnight ) and subjected or not to LTP stimulation . Scale bar 5 μm . ( H and I ) Quantitative analysis of mushroom and thin spine density . Number of analyzed neurons: 14 ( WT , no LTP ) , 16 ( WT , + LTP ) , 28 ( WT , IL-1β , no LTP ) , 27 ( WT , IL-1β , + LTP ) , one-way ANOVA analysis of variance followed by post hoc Tukey test . ( J ) Quantitative analysis of PSD-95 density . Number of analyzed neurons: 9 ( WT , no LTP ) , 13 ( WT , + LTP ) , 15 ( WT , IL-1β , no LTP ) , 15 ( WT , IL-1β , + LTP ) , one-way ANOVA analysis of variance followed by post hoc Tukey test . Data indicate that application of IL-1β prevents synaptic potentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 006 Of note , exposure of WT neurons to IL-1Ra prevented LTP , as assessed by confocal analysis ( Figure 4A–C ) or electrophysiological recording ( Figure 4D–F ) , indicating that IL-1R acts positively in supporting LTP , even when IL-1R8 expression is not perturbed . In line with these observations , we found that neurons genetically devoid of IL-1 receptor ( IL-1R KO ) were unable to undergo plasticity phenomena ( mushroom spine density: WT , no LTP: 0 , 13 ± 0009 , n = 19; WT , + LTP: 0 , 36 ± 0 , 03 , n = 18; Student t test , p<00001 . IL-1R KO , no LTP: 0 , 41 ± 0 , 03; IL-1R KO , + LTP: 0 , 49 ± 0 , 03 n = 18; Student t test , p ns = 0 , 1024 ) . Therefore , in line with literature data ( Costello et al . , 2011; Schneider et al . , 1998; Coogan and O'Connor , 1999; Avital et al . , 2003 ) , either pharmacological or genetic silencing of IL-1R is per se sufficient to alter dendritic spine morphology and plasticity , indicating that physiological levels of IL-1R activation are required for correct long-term potentiation . To further dissect changes in WT , IL-1R8 KO and WT mice treated with IL-1β , we conducted transcriptomic analysis on cortical tissues . RNA-seq data revealed that genetic ablation of IL-1R8 and pharmacological activation of IL-1R both lead to altered transcription of a shared subset of genes ( Figure 5A and B ) . Treatment of WT mice with IL-1β led to transcriptional alterations in 1084 genes ( 60 . 6% downregulated and 39 . 4% upregulated , Supplementary file 1 ) . IL-1R8 KO mice showed alterations in expression of 639 genes compared to WT ( 49 . 1% downregulated and 50 . 9% upregulated , Supplementary file 2 ) . Comparing the two sets of genes , we observed a highly significant overlap of 193 genes whose expression was altered by either IL-1β treatment and IL-1R8 deficiency ( Fischer’s exact test; P value = 2 . 2e-16; Pearson correlation coefficient = 0 . 95; 61 . 7% downregulated and 38 . 3% upregulated , Supplementary file 3 ) . Gene Ontology ( GO ) enrichment analysis revealed that the deregulated genes are clustered in specific categories referring to biological processes such as hippocampal development and synaptic transmission , among others ( Supplementary file 4 ) . Moreover , consistent with our results in neuronal cultures , transcriptomic analysis of IL-1R8 KO mice treated with Anakinra revealed that this pharmacological treatment largely reversed the transcriptional alterations observed in IL-1R8 KO mice ( 83 . 1%; False Discovery Rate/FDR <0 . 1; 80 . 8% , FDR < 0 . 05 ) , including most of the shared changed genes between IL-1R8 KO mice and in IL1β stimulated mice ( 89 . 0%; FDR < 0 . 1; 88 . 7% , FDR < 0 . 05 ) ( Figure 5C and D , Figure 5—figure supplement 1 and Supplementary file 3 ) . Notably , the reversion affected important synaptic genes downregulated in both IL-1R8 KO and IL-1β-treated mice , such as Mdga1 ( Figure 5—figure supplement 1 ) , Cnih2 and Sez6 ( Supplementary file 3 ) . These data indicate that acute IL-1β treatment and IL-1R8 deficiency trigger overlapping gene programs which are reversed , in the case of IL-1R8 KO mice , by acute exposure to Anakinra . 10 . 7554/eLife . 21735 . 007Figure 5 . Transcriptomic analysis of cortices from WT mice treated with IL-1β and IL-1R8 KO mice reveals common genes with altered regulation , and reversal of altered expression upon treatment of IL-1R8 KO mice with the IL-1β antagonist Anakinra . ( A ) Venn diagram showing significant overlap in the number of Differentially Expressed ( DE ) genes between conditions ( P value = 2 . 2e-16 , Fischer’s exact test ) . IL1β labels DE genes after IL1β administration and IL-1R8 KO show DE genes in IL-1R8 KO mice . ( B ) Heatmap showing a hierarchical clustering of the genes ( rows ) based on fold changes of expression in each sample versus the average level in the WT condition . Color sidebar for the samples is indicative of the condition: WT ( black ) , treatment with IL-1β ( green ) , or IL-1R8 KO ( orange ) . The inset key shows the color scale of the fold change matrix ( log2 values ) , from blue ( downregulated genes ) to red ( upregulated genes ) , and white for non-regulated genes . ( C ) Line chart showing fold change ( log2 values ) against condition ( IL1β , IL-1R8 KO , IL-1R8 KO + Anakinra ) for DE genes in IL-1R8 KO mice ( FDR < 0 . 1 ) and also upon IL1β administration to WT mice ( FDR < 0 . 1 , 193 genes ) . Genes that were reversed upon treatment of IL-1R8 KO mice with Anakinra are shown in red ( 96 upregulated genes ) or blue ( 60 downregulated genes ) . Genes that were not reversed by Anakinra are shown in light grey ( 37 genes ) . ( D ) Bar graph showing the absolute number of genes that were either reversed or non-affected by treatment of IL-1R8 KO mice with Anakinra . Percentage of reversed genes over the total of DE genes in each gene list is also shown . DE gene lists are as follows: IL-1R8 KO FDR < 0 . 1 ( 639 genes ) ; IL-1R8 KO FDR < 0 . 05 ( 264 genes ) ; IL-1R8 KO ∩ IL1β FDR < 0 . 1 ( 193 genes ) ; IL-1R8 KO ∩ IL1β FDR < 0 . 05 ( 71 genes ) . Reversed genes are defined as not found to be DE using the indicated significance threshold in each group , in the comparison of IL-1R8 KO + Anakinra versus WT mice , and those that were differentially expressed in this comparison but in the opposite direction found in IL1β and IL-1R8 KO conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 00710 . 7554/eLife . 21735 . 008Figure 5—figure supplement 1 . RNA-seq profiles in the cortex of WT , WT treated with IL-1β , IL-1R8 KO mice , and IL-1R8 KO mice treated with Anakinra . RNA-seq profiles at the Mdga1 locus presenting downregulation of transcript levels in WT mice in response to IL-1β treatment . Transcript levels for Mdga1 are also reduced in the cortex of IL-1R8 KO mice . Administration of Anakinra to IL-1R8 KO mice restored Mdga1 transcript levels to that of WT ones . Note the very high reproducibility of the profiles between samples . Mdga1 is involved in maintaining the proper balance between excitatory and inhibitory synapses ( Pettem et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 008 We next identified the downstream signaling pathway through which hyperactivation of the IL-1R pathway in IL-1R8 KO neurons influences spine morphology and function . The signaling pathways downstream of IL-1R8 have been described in non-neuronal cells and found to include i ) the inhibition of NF-κB and JNK activation , dependent on IL-Rs or TLRs family member activation and ii ) the IL-1-dependent activation of the Akt-mTOR pathway ( reviewed in Riva et al . [2012] ) . Based on the known involvement of the phosphatidylinositol 3-kinase ( PI3K ) /Akt/mTOR pathway in the induction and maintenance of LTP in different brain regions ( Opazo et al . , 2003; Lee et al . , 2011 ) , we focused on the possibility that the lack of IL-1R8 could result in the hyperactivation of this pathway . This possibility would be in line with the demonstrated cognitive and plasticity deficits occurring in genetic mouse models characterized by increased mTOR signaling ( reviewed in Hoeffer and Klann [2010] ) . Overnight treatment of 15 DIV IL-1R8 KO neurons with 20 nM rapamycin ( Vézina et al . , 1975 ) , which inhibits both mTORC1 and mTORC2 , was sufficient to reverse the defective spine phenotype , with IL-1R8 KO treated neurons displaying an increase in density of both mushroom spines ( Figure 6A and B ) and PSD-95 puncta ( Figure 6A and C ) relative to untreated IL-1R8 KO neurons . Similarly , overnight exposure to 30 μM LY294002 , a specific inhibitor of PI3K , or 20 nM wortmannin , a PI3-K/Akt signal transduction inhibitor , rescued mushroom spine and PSD-95 densities in IL-1R8 KO neurons ( Figure 6 ) . None of the blockers produced damage to the neurons , at least for the time of incubation used . These data demonstrate that hyperactivation of IL-1R in IL-1R8 KO neurons impairs spine morphogenesis and plasticity through the PI3K/AKT/mTOR pathway . Again , WT neurons exposed to the different inhibitors displayed an increase in spine and postsynaptic marker density ( Figure 6A–C ) . These data indicate that the endogenous activation of mTOR/PI3K/Akt pathway , like already shown for IL-1R activation ( Figure 4A–C ) , is required for correct spine morphogenesis . 10 . 7554/eLife . 21735 . 009Figure 6 . Inhibition of mTOR signalling restores LTP in IL-1R8 KO neurons . ( A ) PSD-95 immunocytochemical staining of GFP-transfected 16 DIV hippocampal cultures from WT or IL-1R8 KO mice . At DIV 15 neurons were treated with rapamycin ( Rapa , 20 nM ) , LY294002 ( LY , 30 μM ) or wortmannin ( wort , 20 nM ) overnight ( 14 hr ) . Scale bar , 5 μm . ( B ) Quantitative analysis of mushroom spine density . Number of analyzed neurons: 36 ( WT ) , 17 ( WT , Rapa ) , 10 ( WT , LY ) , 5 ( WT , Wort ) , 42 ( IL-1R8 KO ) , 35 ( IL-1R8 KO , Rapa ) , 28 ( IL-1R8 KO , LY ) , 30 ( IL-1R8 KO , Wort ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . ( C ) Quantitative analysis of PSD-95 puncta density . Number of analyzed neurons: 17 ( WT ) , 7 ( WT , Rapa ) , 9 ( WT , LY ) , 11 ( WT , Wort ) , 9 ( IL-1R8 KO ) , 15 ( IL-1R8 KO , Rapa ) , 15 ( IL-1R8 KO , LY ) , 18 ( IL-1R8 KO , Wort ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . Data indicate that inhibitors of the mTOR pathway restore synaptic potentiation . * indicates significance compared to WT , # indicates significance compared to IL-1R8 KO . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 009 It is known that mTOR is at the cross road of plasticity , memory and disease processes ( Hoeffer and Klann , 2010 ) . Reduced AKT/mTOR signaling and protein synthesis dysregulation has been described in the brain of the MeCP2 KO Rett syndrome animal model , thus indicating that MeCP2 is an upstream regulator of the Akt/mTOR pathway ( Ricciardi et al . , 2011 ) . Western blotting and immunofluorescence analysis of MeCP2 in IL-1R8 KO hippocampal cultures revealed higher levels of the protein with respect to controls ( Figure 7A and B ) , in the absence of differences in the transcript levels ( normalized counts WT = 4857 ± 97 . 46 , WT+IL1β = 4936 ± 281 . 3 , IL-1R8 KO = 4864 ± 201 . 7 , IL-1R8 KO+Anakinra = 3599 ± 232 . 6 , mean ± SEM , p>0 , 05 , one-way ANOVA , Tukey’s multiple comparisons test ) . However , relative to previous reports ( Ricciardi et al . , 2011 ) , we found that IL-1R activation and mTOR pathway can be also an upstream regulator of MeCP2 . Indeed , overnight treatment of IL-1R8 KO neurons with either IL1Ra ( 20 or 100 ng/ml ) or rapamycin ( 20 nM ) resulted in the normalization of MeCP2 levels , as assessed by Western Blotting ( Figure 7A and B ) or immunofluorescence ( Figure 7C and D ) , thus indicating that MeCP2 is increased in IL-1R8 KO neurons as a consequence of the hyperactivation of IL-1R and downstream to the mTOR pathway . Finally MeCP2 levels were increased by immunofluorescence ( Figure 7E and F ) and by western blotting ( Figure 7G and H ) also in cultured WT neurons exposed to IL-1β ( 40 ng/ml for 14 hr ) . As a further support to this view , a reduction of MeCP2 intensity was detected in WT neurons exposed to either IL-1Ra or rapamycin and examined by immunofluorescence ( Figure 7C and D ) . Consistently , MeCP2 levels , analyzed by Western Blotting , were significantly lower in both hippocampus and cortex of adult IL-1R KO mice ( 7 months old ) ( normalized integrated density of MeCP2 levels in hippocampus , WT: 1 ± 0 , 0535 ( n = 5 ) ; IL-1R KO: 0795 ± 0 , 0121 ( n = 5 ) , Student t test , p<0 , 01 . Normalized integrated density of MeCP2 levels in cortex , WT: 1 ± 0 , 0586 ( n = 5 ) ; IL-1R KO: 0 , 7792 ± 0 , 0473 ( n = 5 ) , Student t test , p<0005 ) . 10 . 7554/eLife . 21735 . 010Figure 7 . IL-1R8 KO neurons display higher MeCP2 levels that are responsible for LTP defects in IL-1R8 KO neurons . ( A ) MeCP2 expression analyzed by western blotting in cultured neurons from WT or IL-1R8 KO neurons treated or not with IL-1Ra ( 20 ng/ml ) , IL-1Ra ( 100 ng/ml ) or Rapa ( 20 nM ) . A higher MeCP2 expression is detectable in IL-1R8 KO neurons , which is reduced by IL-1Ra or Rapa . ( B ) Quantitative analysis of MeCP2 expression . Number of replicates: 3 ( WT ) , 3 ( IL-1R8 KO ) , 3 ( IL-1R8 KO , IL1Ra 20 ) , 3 ( IL-1R8 KO , IL1Ra 100 ) , 3 ( IL-1R8 KO , Rapa ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . ( C ) MeCP2 and MAP2 immunocytochemical staining of 16 DIV hippocampal WT or IL-1R8 KO neurons treated overnight ( 14 hr ) with vehicle , IL1Ra ( 20 ng/ml ) or Rapa ( 20 nM ) . Scale bar , 40 μm ( low magnification image ) , 20 μm ( insert ) . ( D ) Quantitative analysis of MeCP2 immunoreactivity in neurons treated as above . Number of analyzed neurons: 88 ( WT ) , 60 ( WT , IL1Ra ) , 55 ( WT , Rapa ) , 52 ( IL-1R8 KO ) , 66 ( IL-1R8 KO , IL1Ra ) , 81 ( IL-1R8 KO , Rapa ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . ( E ) Immunocytochemical staining for MeCP2 and β3-tubulin of 16 DIV WT hippocampal neurons , exposed or not to IL1β ( 40 ng/ml ) overnight . Scale bar , 40 μm . ( F ) Quantitative analysis of MeCP2 immunoreactivity reveals higher MeCP2 levels in IL1β treated neurons . Number of analyzed neurons: 288 ( WT ) , 227 ( WT , IL1β ) ; Student t test . ( G and H ) Representative western blotting ( G ) and quantitative analysis ( H ) of MeCP2 expression in WT neurons treated with IL-1β ( 40 ng/ml ) . Number of replicates , WT: n = 5; WT + IL1β 40: n = 5 independent experiments . Mann Whitney test . * indicates significance compared to WT , # indicates significance compared to IL-1R8 KO . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 010 MeCP2 , a synaptic factor that controls spine morphogenesis and plasticity ( Chao et al . , 2007; Nelson et al . , 2006; Zoghbi , 2003; Asaka et al . , 2006; Guy et al . , 2007; Moretti et al . , 2006 ) , needs to be tightly regulated in the human brain . Indeed , besides the well-known pathological traits - intellectual disability and delayed development - caused by MeCP2 duplication ( Ramocki et al . , 2010 ) or MeCP2 loss of function ( Chahrour and Zoghbi , 2007 ) , even mild differences in MeCP2 expression turned out to profoundly impact human behavior and brain function ( Tantra et al . , 2014 ) . To investigate whether the morphological and functional defects observed in the IL-1R8 KO neurons could directly result from increased MeCP2 levels , we silenced MeCP2 in neuronal cultures from IL-1R8 KO mice through the use of a well characterized shRNA construct ( Sh MeCP , kind gift of Dr . M . Greenberg , Harvard Medical School ) , which has been widely used in literature to reduce MeCP2 expression ( Zhou et al . , 2006; Blackman et al . , 2012; Gangisetty et al . , 2014; Kishi et al . , 2016; Bedogni et al . , 2016 ) . DIV 12 neurons were transfected with shCTRL or shMeCP2 , and examined for their ability to undergo LTP four days later . We first quantitated the extent of MeCP2 reduction in treated cultures and found that the use of shRNA construct allowed to achieve a protein reduction ranging from 59% to 67% . However , given the higher MeCP2 expression in IL-1R8 KO neurons , the levels of MeCP2 in shRNA–treated IL-1R8 KO neurons ended up to be comparable to those in untreated WT neurons ( Figure 8A and B ) . While WT neurons with silenced MeCP2 showed abnormal spine morphology , with a decreased density of mushroom spines ( Figure 8C and D ) and PSD95 puncta ( Figure 8C and E ) , IL-1R8 KO neurons with silenced MeCP2 reverted the morphological phenotype . Even more interestingly , MeCP2 silenced neurons rescued their ability to undergo LTP ( Figure 8F–H ) . Therefore , recovery of MeCP2 to control levels is sufficient to acutely rescue the synaptic defects observed in IL-1R8 KO neurons . 10 . 7554/eLife . 21735 . 011Figure 8 . Higher MeCP2 levels are responsible for LTP defects in IL-1R8 KO neurons . ( A ) MeCP2 and MAP2 immunocytochemical staining and ( B ) quantitative analysis of MeCP2 immunoreactivity in 16 DIV WT and IL-1R8 KO neurons transfected with SH CTRL or SH MeCP2 . Scale bar 40 μm . Number of analyzed neurons: 23 ( WT , SH CTRL ) , 28 ( WT , SH MeCP2 ) , 46 ( IL-1R8 KO , SH CTRL ) , 40 ( IL-1R8 KO , SH MeCP2 ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . * indicates significance compared to WT + SH CTRL , # indicates significance compared to IL-1R8 KO + SH CTRL . ( C ) PSD-95 immunocytochemical staining of SH CTRL- or SH MeCP2-transfected , DIV 16 hippocampal cultures from WT or IL-1R8 KO mice . Neurons were transfected at DIV 12 and the LTP protocol was applied at DIV 16 . Scale bar , 5 μm . ( D ) Quantitative analysis of mushroom spine density in WT and IL-1R8 KO neurons treated as above . Number of analyzed neurons: 30 ( WT , SH CTRL , no LTP ) , 29 ( WT , SH CTRL , + LTP ) , 30 ( WT , SH MeCP2 , no LTP ) , 29 ( WT , SH MeCP2 , + LTP ) , 59 ( IL-1R8 KO , SH CTRL , no LPT ) , 60 ( IL-1R8 KO , SH CTRL , + LTP ) , 54 ( IL-1R8 KO , SH MeCP2 , no LTP ) , 58 ( IL-1R8 KO , SH MeCP2 , + LTP ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . ( E ) Quantitative analysis of PSD-95 density . Number of analyzed neurons: 17 ( WT , SH CTRL , no LTP ) , 15 ( WT , SH CTRL , + LTP ) , 16 ( WT , SH MeCP2 , no LTP ) , 15 ( WT , SH MeCP2 , + LTP ) , 15 ( IL-1R8 KO , SH CTRL , no LTP ) , 15 ( IL-1R8 KO , SH CTRL , + LTP ) , 15 ( IL-1R8 KO , SH MeCP2 , no LTP ) , 17 ( IL-1R8 KO , SH MeCP2 , + LTP ) ; one-way ANOVA analysis of variance followed by post hoc Tukey test . Data indicate that reduction of MeCP2 expression restores synaptic potentiation . ( F ) Representative traces of mEPSC recorded from IL-1R8 KO neurons transfected with SH CTRL or SHMeCP2 before and after LTP induction . ( G and H ) Quantitation of mEPSC frequency and amplitude of neurons treated as above . Analysis of normalized mEPSC frequency and amplitude reveals that only neurons transfected with SH MeCP2 undergo LTP . Number of recorded neurons: 14 ( IL-1R8 KO , SH CTRL , no LTP ) , 7 ( IL-1R8 KO , SH CTRL , + LTP ) , 19 ( IL-1R8 KO , SH MeCP2 , no LTP ) , 7 ( IL-1R8 KO , SH MeCP2 , + LTP ) . Mann Whitney test . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 011 The increased MeCP2 expression detected in primary IL-1R8 KO neurons was also confirmed in vivo . Immunohistochemical analysis of hippocampus ( CA1 region ) and cortex revealed a significantly higher expression of MeCP2 in IL-1R8 KO neurons compared to WT , as indicated by the quantitation of the MeCP2 integrated density value per neuron and the cumulative distribution of neuronal integrated density values ( Figure 9A–E ) . Consistently , western blot analysis of either hippocampus or cortex revealed higher levels of MeCP2 in the brains of IL-1R8 KO mice ( Figure 9F–H ) . Notably , administration of 30 mg/kg Anakinra to IL-1R8 KO mice for three consecutive days significantly reduced MeCP2 protein ( Figure 9F–H ) . These data indicate that higher levels of MeCP2 are detectable in the brain of IL-1R8 KO mice and that the acute inhibition of IL-1R by Anakinra reduces hippocampal and cortical MeCP2 levels . 10 . 7554/eLife . 21735 . 012Figure 9 . Higher MeCP2 levels in the brain of IL-1R8 KO mice . ( A ) Representative images of brain sections ( CA1 hippocampus , upper panels and cortex , lower panels ) of WT and IL-1R8 KO mice ( 1 month old ) stained for MeCP2 and NeuN , as indicated . ( B-E ) Graphs show quantitation of MeCP2 mean integrated density values and the cumulative distributions of neuronal MeCP2 integrated density values in CA1 hippocampal neurons ( B and C ) and cortical neurons ( D and E ) . Number of analyzed mice and neurons: WT: n = 3 mice , hippocampal neurons = 572 , cortical neurons = 481 . IL-1R8 KO: n = 3 mice , hippocampal neurons = 559 , cortical neurons = 471 . Statistical comparison: Mann-Whitney Test for B and D; KolmogorovSmirnov Comparison ( http://www . physics . csbsju . edu/stats/KS-test . html ) , D values are: 0 , 2084 with a corresponding p<0 . 0001 ( panel C ) and 0 , 1396 with a corresponding p=0 . 0005 ( panel E ) . ( F-H ) Western blotting ( F ) and quantitative analysis ( G and H ) of MeCP2 expression in cortices and hippocampi of 3 months old WT and IL-1R8 KO mice or in IL-1R8 KO mice treated with Anakinra ( 30 mg/kg ) for 3 days . Number of analyzed mice: 6 WT , 5 IL-1R8 KO and 5 IL-1R8 KO + Anakinra ( G ) ; 6 WT , 6 IL-1R8 KO and 4 IL-1R8 KO + Anakinra ( H ) . Statistical test: one-way ANOVA analysis of variance followed by post hoc Tukey test . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 012 We finally investigated whether behavioral deficits consequent to the genetic lack of IL-1R8 could be rescued in the adult mice by the pharmacological inhibition of IL-1R . To this aim IL-1R8 KO mice were analyzed for novel place recognition test three days after the treatment with the anti-inflammatory compound glycyrrhizic acid ( 50 mg/kg ) , which binds to high-mobility group box 1 ( HMGB1 ) protein and inhibits IL-1 activity ( Sakamoto et al . , 2001; Mollica et al . , 2007 ) or with 30 mg/kg Anakinra . Both treatments ameliorated the spatial memory impairment of IL-1R8 KO mice ( Figure 10A ) , with 50% IL-1R8 KO mice displaying a significantly improved discrimination index after Anakinra treatment ( p=0 , 0168 paired Student’s t-test ) . IL-1R8 KO mice were also tested for the Morris water maze and the rewarded T maze . In line with the concept that alternation , either rewarded or spontaneous , detects hippocampal dysfunction even better than the Morris water maze ( reviewed in Deacon and Rawlins , 2006 ) , the rewarded T-maze ( Figure 10C ) , but not the Morris water maze ( Figure 10B ) , revealed an impairment of spatial memory in IL-1R8 KO mice . Treatment with Anakinra ameliorated the performance ( Figure 10C ) . Consistent with these data , IL-1R8 KO mice performed a significantly higher number of errors during the acquisition phase with respect to WT and Anakinra-treated IL-1R8 KO mice ( Figure 10D ) . Interestingly , also WT mice treated with Anakinra displayed impaired spatial memory ( Figure 10C ) . The lack of increased number of errors in WT mice treated with Anakinra ( Figure 10D ) is possibly attributable to the enhancement of freezing behavior . These data indicate that Anakinra ameliorates the spatial memory of IL-1R8 KO mice and confirm that blockade of IL-1R originates learning defects in WT mice . 10 . 7554/eLife . 21735 . 013Figure 10 . Anakinra rescues behavioral defects in IL-1R8 KO mice . ( A ) Analysis of novel-place object recognition task in WT and IL-1R8 KO mice shows a defect in spatial learning in IL-1R8 KO mice ( one-way ANOVA followed by Bonferroni multiple comparison test ) . 3 days i . p . treatment with 50 mg/kg glycyrrhizic acid or 30 mg/kg Anakinra ameliorates IL-1R8 KO mice performance . Number of analyzed mice: 15 ( WT ) , 6 ( WT+Anakinra ) , 18 ( IL1-R8 KO ) , 3 ( IL1-R8 KO+glycyrrhizic acid ) and 10 ( IL1-R8 KO+Anakinra ) . ( B ) IL-1R8 deficiency or treatment with Anakinra did not affect the learning performance in the Water Maze navigation task . Number of analyzed mice: 5 ( WT ) , 5 ( WT+Anakinra ) , 5 ( IL1-R8 KO ) , and 5 ( IL1-R8 KO+Anakinra ) . ( C ) IL1-R8 KO mice displayed impaired spatial memory in the rewarded T-maze task , as indicated by the low percentage of animals reaching the criterion . Treatment with Anakinra ameliorated the performance of IL1-R8 KO mice , while worsening the learning ability of WT . Number of analyzed mice: 5 ( WT ) , 3 ( WT+Anakinra ) , 6 ( IL1-R8 KO ) and 5 ( IL1-R8 KO+Anakinra ) . ( D ) IL1-R8 KO mice showed a significant increase in the total number of errors during the acquisition phase as compared to the WT , which was significantly reduced by treatment with Anakinra . One-way ANOVA followed by Tukey’s multiple comparison test . Number of analyzed mice: 6 ( WT ) , 5 ( WT+Anakinra ) , 5 ( IL1-R8 KO ) and 5 ( IL1-R8 KO+Anakinra ) . All the mice were 3–6 months old . DOI: http://dx . doi . org/10 . 7554/eLife . 21735 . 013
Finely tuned expression of immune molecules has been found to modulate CNS function ( Deverman and Patterson , 2009 ) , both throughout normal development and in pathological conditions . During development , immune molecules such as CXCR4 , interferon γ , IL-1β , IL-6 , IL-9 , IL-10 and transforming growth factor β affect neurogenesis , neuronal migration , axon guidance , synapse formation , activity-dependent refinement of circuits and synaptic plasticity ( Deverman and Patterson , 2009; Zhao and Schwartz , 1998 ) . In line with this concept , mice genetically devoid of IL-1R8 , a receptor which dampens the activation of the TLRs and IL-1R signalling pathways , are neurologically impaired , showing defects in novel place recognition task , spatial reference memory and LTP , defects that occur in the absence of any external inflammatory stimuli ( Costello et al . , 2011 ) . However , the molecular processes at the basis of these defects are still elusive . Our study provides two major breakthroughs in the field of the cross talk between nervous and immune systems: first , it demonstrates that hyperactivation of the IL1R pathway results in the overexpression of MeCP2 , a synaptic factor that controls spine morphogenesis , synaptic transmission and plasticity , which is responsible for syndromes associated with intellectual disability such as Rett syndrome and MeCP2 duplication syndrome; second it provides the proof-of-concept that the synaptic and behavioral defects consequent to genetic hyperactivation of the IL-1R pathway can be rescued in the adult by pharmacological treatment with IL-1Ra . Of note , the acute application of IL-1β to mature neurons recapitulates similar synaptic defects and alterations of MeCP2 expression . Consistently , transcriptomic analysis of cortices from WT mice ( treated or not with IL-1β ) and from IL-1R8 KO animals ( treated or not treated with Anakinra ) show that a common set of genes is transcriptionally altered in WT mice treated with IL-1β and in IL-1R8 KO animals , including genes controlling synaptic function and spine morphogenesis . The expression of this set of genes is restored to normal levels by treatment with Anakinra in IL-1R8 KO mice . These findings have important repercussions in the clinic , as they could open the road to the use of anti-inflammatory drugs as therapeutic treatments for neuropsychiatric illnesses such as schizophrenia , autism and mental retardation , for all of which an immune component has been demonstrated ( Khandaker et al . , 2015; Theoharides et al . , 2015; Young et al . , 2016 ) . Our study identifies the crucial involvement of the AKT/mTOR pathway in IL-1R8 KO neurons ( Figure 10H ) . IL-1R8 has been previously reported to regulate mTOR kinase activity in Th17 cells , playing a nonredundant role in controlling mTOR-dependent differentiation , proliferation and cytokine production ( Gulen et al . , 2010 ) . The demonstration that the PI3K/Akt/mTOR pathway is at the root of synaptic defects in IL-1R8 KO neurons is in line with the established involvement of this pathway in dendritic tree development and spine formation in neurons ( Sarbassov et al . , 2006 ) , as well as actin cytoskeleton dynamics and synaptic plasticity ( Jaworski and Sheng , 2006 ) . Remarkably , the mTOR pathway is closely linked to ILR/TLR signaling since PI3 kinase is required for activation of NF-κB by IL-1 ( Gulen et al . , 2010 ) . Furthermore , the signaling module containing the MyD88 adaptor protein , together with phosphorylated IRAK and TRAF6 , which is downstream to both IL-1R and TLRs activation , is essential for PI3K recruitment and Akt activation , making mTOR one of the downstream targets of ILR/TLRs signaling ( Gulen et al . , 2010 ) . Consistent with the involvement of mTOR in the IL-1R8 KO neuron defects , treatment with the PI3 kinase inhibitors , LY294002 and wortmannin restored the neuronal ability to sustain LTP . Moreover , specific inhibition of mTOR by the drug rapamycin rescues synaptic morphology in IL-1R8 KO neurons . These results indicate that the PI3K-regulated pathway is involved in the downstream signaling that connects excessive IL-1R activation to the morphological and functional impairments of dendritic spines observed in IL-1R8 KO neurons . More importantly , our study shows for the first time that activation of IL-1R in neurons increases the levels of MeCP2 , both in vitro and in vivo . Fine-tuning of MeCP2 expression is required for proper synapse function ( reviewed in Na et al . [2013] ) . Indeed , lack of expression of MeCP2 results in deceleration of body and head growth rate , problems in motor and speech capabilities , irregularities in motor activity and difficulties in breathing , and also in cognitive defects characteristic of an autism-spectrum disorder ( reviewed in Percy and Lane [2005] ) . Mouse models of MeCP2 duplication display a neurological phenotype , stereotyped and repetitive movements , epilepsy , spasticity , hypoactivity , early death ( Collins et al . , 2004 ) and defects in dendritic arborization and spine morphology ( Jiang et al . , 2013 ) . In humans , sporadic mutations in the gene coding for MeCP2 results in Rett syndrome ( Amir et al . , 1999 ) , while a double dosage of MeCP2 causes a severe developmental delay and mental retardation ( Lubs et al . , 1999 ) . Thus , MeCP2 levels must be finely tuned as even mild over-expression of this factor can have a robust effect . We first demonstrate here that immune activation controls the expression of MeCP2 in neurons . The possibility that MeCP2 phosphorylation , by which MeCP2 also modulates gene expression levels , is altered in these models is an open issue . MeCP2 can act as a transcriptional activator ( when interacting with CREB ) but primarily functions as a transcriptional repressor associated with mSin3A and HDACs repressor complexes ( Chahrour et al . , 2008; Jones et al . , 1998; Nan et al . , 1998 ) . Consistently , we observed a prevalence of downregulation over upregulation in the expression of the deregulated genes in WT mice treated with IL-1β and in IL-1R8 KO animals , suggesting that the repressor function of MeCP2 might dominate in response to IL-1R activity . Indeed , gene ontology enrichment analysis for the deregulated genes highlighted negative regulation of transcription from RNA polymerase II promoter and negative regulation of gene expression among the most affected biological processes in IL-1R8 KO animals ( Supplementary file 4 ) . The regulation of MeCP2 protein levels by IL-1R8 appears to result from enhanced activity of IL-1R . IL-1β is a fundamental factor of inflammatory responses in the brain . It is expressed at baseline levels in the healthy brain , and its expression increases following peripheral infection , surgery , or brain injury , as well as in neurodegenerative diseases ( Rothwell and Luheshi , 2000 ) . Increases in IL-1β levels lead to cognitive decline , in particular in hippocampal-dependent tasks ( Rachal Pugh et al . , 2001 ) . IL-1β overexpression for two weeks in an inducible transgenic mouse was found to cause impairment in long-term contextual and spatial memory , without effects on short-term and non-hippocampal memory ( Hein et al . , 2010 ) . An increase in IL-1β levels due to Escherichia coli infection leads to defects in contextual fear conditioning , with loss of memory prevented by IL1Ra ( Barrientos et al . , 2009; Frank et al . , 2010 ) . Remarkably , we show that neuron treatment with IL-1β not only leads to an increase in immature spines unable to undergo LTP , but also augments MeCP2 levels . In this context , as with E . coli infection , IL1Ra/Anakinra restores synaptic function . Our demonstration that enhanced IL-1R activity , due to either lack of IL-1R8 or excessive IL-1β signaling , impacts synapse function through regulating the expression of MeCP2 expands in a critical way previous research on the role of MeCP2 in cognition ( Na et al . , 2013 ) : we show here that MeCP2 is a fundamental node linking inflammation with synaptic damage and we demonstrate that inhibition of IL-1R pathway by IL-1Ra restores synaptic structure and function in vitro . We also demonstrate that the acute treatment with Anakinra , a pharmacological agent currently in use to treat chronic inflammation , restores synaptic plasticity in vivo . Although we do not have the proof that this occurs exclusively through MeCP2 reduction , our data indicate that immune drugs may be efficacious for treating neurological deficits associated with immune pathologies with a genetic basis and justify further research into anti-inflammatory treatment for selected brain pathologies . Indeed , recurrent infections have been found to occur in 70% of individuals affected by MeCP2 duplication syndrome , which lead to further deterioration of the general and neurological status , being even fatal in some patients ( van Esch et al . , 2012 ) . Also , in patients affected by cryopyrin-associated periodic syndrome ( CAPS ) , a group of rare autoinflammatory diseases with genetic basis , the levels of IL-1β are fivefold higher than in healthy individuals , leading to persistent unregulated systemic inflammation ( Janssen et al . , 2004; Lachmann et al . , 2009 ) . CAPS is characterized by recurrent bouts of fever with malaise and chills , urticarial neutrophilic , eye redness due to conjunctivitis , arthralgia and myalgia with intense fatigue . Children affected by CAPS are believed to show symptoms of intellectual disability . Mental and hearing defects are reversed following treatment with Anakinra ( Goldbach-Mansky , 2011; Goldbach-Mansky et al . , 2006; Lepore et al . , 2010; Miyamae et al . , 2010; Neven et al . , 2010 ) and with specific neutralization of IL-1β with canakinumab . Our study opens the possibility that further cognitive deterioration may result from the enhanced inflammation and hyperactivation of the IL-1 signaling pathway , which might further increase MeCP2 levels in a harmful positive feedback loop . The challenging possibility that treatment with Anakinra may help interrupting this spiral and ameliorating the cognitive deficits in affected individuals is worth to be tested .
IL-1R8 KO mice ( Garlanda et al . , 2004 ) and double IL-1R8 KO IL-1R KO mice ( Véliz Rodriguez et al . , 2012 ) were obtained from Istituto Clinico Humanitas IRCCS , Milan , Italy . Primary hippocampal cultures were performed from E17 embryos . Tissues for WB analysis were taken from 3-month-old male animals . All the experimental procedures followed the guidelines established by the European Legislation ( Directive 2010/63/EU ) , and the Italian Legislation ( L . D . no 26/2014 ) . Mice ( 3 months old ) were deeply anesthetized with chloral hydrate ( 4%; 1 ml/100 g body weight , i . p . ) and perfused intracardially with 0 . 9% saline solution . The brains were removed and stained by modified Golgi-Cox method as described in ( Menna et al . , 2013 ) with slight modifications . Coronal sections of 100 µm thickness from the dorsal hippocampus were obtained using a vibratome ( VT1000S , Leica , Wetzlar , Germany ) . These sections were collected free floating in 6% sucrose solution and processed with ammonium hydroxide for 15 min , followed by 15 min in Kodak Film Fixer , and finally were rinsed with distilled water , placed on coverslips , dehydrated and mounted with a xylene-based medium . Spine density was counted on the secondary branches of apical dendrites of pyramidal neurons located in the CA1 subfield of the dorsal hippocampus . At least 30 neurons per animal were evaluated . Mice ( 1 month old ) were deeply anesthetized with chloral hydrate ( 4%; 1 ml/100 g body weight , i . p . ) and perfused intracardially with 4% paraformaldehyde . Immunofluorescent staining was carried out on free-floating sections as described in ( Menna et al . , 2013 ) . Free-floating sections at the level of dorsal hippocampus were processed with the specific antibodies as indicated , followed by incubation with the secondary antibodies , counterstained with DAPI and mounted in Fluorsave ( Calbiochem , San Diego , CA , USA ) . Primary antibodies: anti-vGLUT-1 ( guinea pig polyclonal antibody , 1:1000; No . 135 304 Synaptic System ) , anti-MeCP2 ( rabbit polyclonal , 1:200; M9317 Sigma ) , anti-NeuN ( mouse monoclonal , 1:500; MAB377 Millipore ) . Sections were examined by means of a Zeiss LSM 510 META confocal microscope ( Leica Microsystems , Germany ) . Images were acquired in the stratum pyramidale or stratum radiatum of the CA1 subfield of the hippocampus ( as indicated ) using the x40 oil immersion lens with an additional electronic zoom factor of up to 3 ( voxel sizes of 0 . 10 × 0 . 10×1 μm ) maintaining the parameters of acquisition ( laser power , pinhole , gain , offset ) constant among groups . Whole cell voltage-clamp recordings were performed on wild type and transgenic embryonic hippocampal neurons maintained in culture for 13–15 DIV . During recordings cells were bathed in a standard external solution containing ( in mM ) : 125 NaCl , 5 KCl , 1 . 2 MgSO4 , 1 . 2 KH2PO4 , 2 CaCl2 , 6 glucose , and 25 HEPES-NaOH , pH 7 . 4 . Recording pipettes were fabricated from borosilicate glass capillary using an horizontal puller ( Sutter Instruments ) inducing tip resistances of 3–5 MΩ and filled with a standard intracellular solution containing ( in mM ) : 130 Cs-gluconate , 8 CsCl , 2 NaCl , 4 EGTA , 10 HEPES- NaOH , 2 MgCl2 , 4 MgATP , and 0 . 3 Tris-GTP . For miniature AMPA-EPSC recordings tetrodotoxin 1µM , Bicuculline 20 μM and AP5 50 μM ( Tocris ) were added to standard extracellular solution to block the spontaneous action potentials propagation , GABA-A and NMDA receptors , respectively . Recordings were performed at room temperature in voltage clamp mode at holding potential of −70 mV using a Multiclamp 700B amplifier ( Molecular Devices ) and pClamp-10 software ( Axon Instruments , Foster City , CA ) . Series resistance ranged from 10 to 20 MΩ and was monitored for consistency during recordings . Cells in culture with leak currents > 100 pA were excluded from the analysis . Signals were amplified , sampled at 10 kHz , filtered to 2 or 3 KHz , and analyzed using pClamp 10 data acquisition and analysis program . Electrophysiological mEPSC recordings of neurons WT , IL-1R8 KO and IL-1R8 KO IL-1R KO were always performed in the same experimental sessions . Chemical Long Term Potentiation ( LTP ) was performed as in ( Menna et al . , 2013 ) . Induction was performed stimulating synaptic NMDA receptors via glycine . For glycine-induced LTP experiments , hippocampal neurons were transfected at 10DIV with cDNA encoding for EGFP by using Lipofectamine 2000 . After 6 days , cells were perfused with a solution containing ( in mM ) 125 NaCl , 5 KCl , 1 . 2 KH2PO4 , 2 CaCl2 , 1MgCl2 , 6 glucose , and 25 HEPES-NaOH , TTX 0 . 001 , Strychnine 0 . 001 and bicuculline methiodide 0 . 02 ( pH 7 . 4 , KRH ) for 10 min then a solution devoid of Mg2+ and containing glycine ( 100 µM ) was applied for 3 min followed by a wash and recovery in neuronal medium for at least 60 min . After 60 minutes cells were immediately fixed and stained . For patch clamp electrophysiology , the patch pipette electrode contained the following solution ( in mM ) : 130 CsGluconate , 8 CsCl , 2 NaCl , 10 HEPES , 4 EGTA , 4 MgATP and 0 . 3 Tris-GTP . Samples containing 25 mg protein were resolved in 12% sodium dodecyl sulphate-polyacrylamide gels under reducing conditions . After transfer onto polyvinylidene diflouride membranes for 2 hr at 250 mA at 41C , blots were blocked for 1 hr at room temperature in a 5% MILK solution in phosphate-buffered saline ( PBS ) pH 7 . 4 and then incubated with PSD95 ( 1:10000; monoclonal; UC Davis/NIH NeuroMab Facility , CA ) , MeCP2 ( 1:1000; polyclonal; Cell Signaling ) , GAPDH ( 1:4000; polyclonal; Synaptic System , Goettingen , Germany ) at 4°C overnight in PBS 0 . 5% Tween-20 ( PBS-T ) . Subsequently , membranes were washed and incubated for 1 hr at room temperature in PBS-T with the secondary antibodies . Western blotting was performed by means of Chemi-Doc system + Image Lab software ( Bio-Rad ) . Photographic development was by chemiluminescence ( ECL , Amersham Bioscience or Immobilon substrate , Millipore ) . Western blot bands were quantified by the ImageJ program ( rsb . info . nih . gov/ij ) . Total RNA from cortex from individual mice was isolated using TRI ( #T9424; Sigma , Inc ) and resuspended in 100 ul ddH2O treated with diethyl pyrocarbonate ( DEPC ) . Samples were then incubated with DNase I ( #79254; Qiagen , Inc ) for 10 min at room temperature and precipitated with RNA grade potassium acetyate ( Ambion , #AM9610 ) . RNA pellets were finally resuspended in 20 ul ddH2O DEPC . Stranded mRNA-Seq multiplexed libraries were prepared from total RNA from mouse cortex following manufacturer’s instructions ( Illumina , Inc ) . To reduce biological variability , each library was performed with total RNA from three independent mice . A total of 3 libraries ( 9 mice in total ) were performed per condition ( 12 libraries in total ) . Conditions were wild-type mice ( C57bl/6j mice ) , wild-type mice treated with IL1β 8 μg/kg o . n . ( 14 hr ) , IL-1R8 KO mice and IL-1R8 KO mice treated with Anakinra ( 30 mg/kg , i . p . administration ) for three consecutive days . Sequencing was performed in a HiSeq 2500 apparatus in paired-end configuration ( 2 × 125 bp ) . To increase sequencing depth , samples were sequenced in two different lanes . All the libraries were loaded in each of the two lanes . Quality control of the raw data was performed with FastQC ( http://www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) . Libraries were trimmed for adapter removal using Trimmomatic ( Bolger et al . , 2014 ) and mapped to reference genome ( Ensembl GRCm38 ) using TopHat2 ( Kim et al . , 2013 ) and Bowtie2 ( Langmead et al . , 2009 ) . Library sizes of primary mapped reads were between 70 and 96 million reads . Samtools was used to manipulate BAM files ( Li et al . , 2009 ) . For calling of differentially expressed genes ( DEG ) , mapped reads were counted with HTSeq v0 . 6 . 1 ( Anders et al . , 2015 ) and count tables were analysed using DeSeq2 v1 . 10 . 1 R-package ( Love et al . , 2014 ) with a design of one factor with four levels ( ‘wild-type’ , ‘wild-type + IL1β” , ‘IL1-R8 KO’ , ‘IL1-R8 KO + Anakinra’ ) and differences between groups were tested using contrasts for wild-type + IL1β versus wild-type; IL-1R8 KO versus wild-type; IL-1R8 KO + Anakinra versus wild-type . For consideration of differentially regulated genes between conditions , we used adjusted p-value<0 . 1 or adjusted p-value<0 . 05 as indicated in figure legends . Functional annotation and category and pathway analysis were carried out using WEB-based Gene SeT AnaLysis Toolkit ( WebGestalt ) ( Zhang et al . , 2005 ) . All expression data are made publicly available in the GEO Series GSE80446 . Data are presented as mean±standard error ( SE ) from the indicated number of experiments . Statistical analysis was performed using PRISM 6 software ( GraphPad , Software Inc . , San Diego , CA , USA ) . After testing whether data were normally distributed or not , the appropriate statistical test has been used . The Kolmogorov –Smirnov test was used to determine significance in cumulative distributions of mEPSC amplitudes and integrated density values . In particular , Mann-Whitney t test was used to determine significance in an average of mEPSC frequency . To determine significant differences in spine number we used Mann Whitney test or one-way ANOVA followed by specific multiple comparison post hoc tests ( as indicated ) . The differences were considered to be significant if p≤0 . 05 and are indicated by ( * ) or ( # ) ; those at p≤0 . 01 are indicated by double ( * ) or ( # ) ; those at p≤0 . 005 are indicated by triple ( * ) or ( # ) ; those at p≤0 . 0001 by four ( * ) or ( # ) . | Errors that occur while the brain is developing can lead to conditions such as autism and schizophrenia . They can also lead to rare disorders like Rett syndrome and MeCP2 duplication syndromes , which are characterized by severe cognitive and physical disabilities . Many people with these neurodevelopmental disorders have mutations in genes that encode proteins found at synapses , which are the junctions between neurons where the cells exchange information with one another . However , not everyone with these mutations develops a neurodevelopmental disorder , which indicates that other , non-genetic factors also play a part . One of the main non-genetic factors that can influence the risk and severity of neurodevelopmental disorders is inflammation of the brain . Inflammation is a normal part of the body’s immune response to threats such as invading microorganisms or tissue damage . However , abnormal activation of the immune system in early life can trigger excessive inflammation . This increases the risk of a neurodevelopmental disorder , but it is not clear exactly how it does so . Tomasoni et al . set out to test whether the missing link between inflammation and neurodevelopmental disorders might be damage to synapses . The experiments revealed that genetically modified mice with inflammation of the brain have abnormal synapses and are unable to learn properly . These mutant mice also have excessive levels of a protein that influences how synapses function called MeCP2 , which is missing in the brains of people with Rett syndrome and abnormally increased in brains of patients affected by MeCP2 Duplication Syndrome . This is thus the first evidence that directly links inflammation of the brain to a synapse protein implicated in a disorder of brain development . Tomasoni et al . also found that a drug called anakinra – which is used to treat an inflammatory disease called rheumatoid arthritis – reduced levels of MeCP2 in the mutant mice and improved their performance in cognitive tasks . Together , these results raise the possibility that anti-inflammatory medications may be beneficial in the treatment of neurodevelopment disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Lack of IL-1R8 in neurons causes hyperactivation of IL-1 receptor pathway and induces MECP2-dependent synaptic defects |
We aimed to study time trends and levels of mean total cholesterol and lipid fractions , and dyslipidaemias prevalence in Latin America and the Caribbean ( LAC ) . Systematic-review and meta-analysis of population-based studies in which lipid ( total cholesterol [TC; 86 studies; 168 , 553 people] , HDL-Cholesterol [HDL-C; 84 studies; 121 , 282 people] , LDL-Cholesterol [LDL-C; 61 studies; 86 , 854 people] , and triglycerides [TG; 84 studies; 121 , 009 people] ) levels and prevalences were laboratory-based . We used Scopus , LILACS , Embase , Medline and Global Health; studies were from 1964 to 2016 . Pooled means and prevalences were estimated for lipid biomarkers from ≥2005 . The pooled means ( mg/dl ) were 193 for TC , 120 for LDL-C , 47 for HDL-C , and 139 for TG; no strong trends . The pooled prevalence estimates were 21% for high TC , 20% for high LDL-C , 48% for low HDL-C , and 21% for high TG; no strong trends . These results may help strengthen programs for dyslipidaemias prevention/management in LAC .
There is a growing body of evidence about levels , patterns and trends of body mass index , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2019; NCD Risk Factor Collaboration ( NCD-RisC ) , 2017a ) diabetes , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2016 ) blood pressure and hypertension , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2017b; Geldsetzer et al . , 2019 ) yet much less has been reported about dyslipidaemias and cholesterol ( Farzadfar et al . , 2011; NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Unlike Latin America and the Caribbean ( LAC ) , other world regions have multi-country studies or systematic reviews that have informed public health officers and practitioners about the burden of unhealthy lipid profiles ( Noubiap et al . , 2018 ) . Moreover , available evidence already suggests there are non-trivial differences in lipid levels with other regions that deserve further scrutiny ( Farzadfar et al . , 2011; Ponte-Negretti et al . , 2017a; Ponte-Negretti et al . , 2017b ) . These facts show that regional evidence on lipid profiles and trends is limited in LAC , hampering the formulation of health policies and practice guidelines to prevent , treat and control dyslipidaemias with a regional focus . This dearth of evidence has relevant implications for public health , clinical medicine and research in LAC . It is unknown whether surveillance systems are urgently needed to monitor dyslipidaemias , because the current cholesterol levels and whether they have increased or decreased have not been quantified in LAC . In this line , if public health authorities should secure access to lipid-lowering medications in the current struggle to extend universal health coverage , ( Atun et al . , 2015 ) is also unknown because we have not quantified , which dyslipidaemia is the most prevalent in LAC . Finally , research cannot efficiently advance if LAC does not know what evidence is already available; thereby , resources can be targeted to where information is scarce or null . Therefore , we aimed to provide robust evidence about trends of mean levels of total cholesterol and lipid fractions , as well as trends of dyslipidaemias prevalence in LAC . This evidence will guide policies and interventions so that they can focus on the most pressing issues . Also , public health officers can use this information as a starting point for disease surveillance and to monitor progress of interventions or to track targets .
The search yielded 6699 titles and abstracts; of these , 1123 were studied in detail and finally 197 studies met the inclusion criteria ( Figure 1—figure supplement 1 ) . Brazil , with 61 studies , and Chile with 21 studies , contributed with the greatest number of studies to the systematic review ( Figure 1 - Figure 1—figure supplement 1 and Supplementary file 1 , table 1 ) . There were more studies conducted since 2010 ( Figure 1—figure supplement 1 ) . Across studies , the mean proportion of men in the study population was 43% , and the mean age was 48 years ( Supplementary file 1 , table 1 ) . Evidence from 86 studies ( 168 , 553 individuals ) informed the overall estimates on mean total cholesterol . The random-effects meta-analysis revealed a pooled mean total cholesterol of 193 mg/dl since 2005 ( Table 1 ) . During the last years , there seemed to be a negative yet weak correlation with time , signalling a small decrease in mean total cholesterol ( Figure 2 ) . National studies tended to report lower mean levels than community and sub-national studies ( Figure 2 ) . Southern and Tropical Latin America appeared to have higher levels than the other sub-regions ( Figure 2 ) . The total cholesterol prevalence estimates were informed by 68 studies ( 129 , 123 individuals ) overall . The pooled prevalence since 2005 was 21% for total cholesterol ≥240 mg/dl and 34% for total cholesterol ≥200 mg/dl ( Table 1 ) . There was a positive trend with time , signalling an increase yet weak evidence supported this observation ( Figure 3 ) . National studies were evenly distributed; Southern and Tropical Latin America seemed to have higher estimates ( Figure 3 ) . The overall sample for LDL-Cholesterol was 61 studies ( 86 , 854 subjects ) . Since 2005 , the pooled mean was 120 mg/dl ( Table 1 ) . There was a non-significant decreasing trend ( Figure 2 ) . National studies seemed to report lower means , and there was not a clear geographic distribution ( Figure 2 ) . Overall , LDL-Cholesterol prevalence estimates were informed by 29 studies ( 42 , 900 individuals ) . The pooled prevalence of high LDL-cholesterol since 2005 was 21% for LDL-Cholesterol ≥160 mg/dl and 40% for LDL-Cholesterol ≥130 mg/dl ( Table 1 ) , and such estimates have slightly increased ( Figure 3 ) . National studies were evenly distributed along the other studies ( Figure 3 ) . Southern and Tropical Latin America appeared to have higher estimates ( Figure 3 ) . The HDL-Cholesterol mean estimates benefited from 84 studies ( 121 , 282 subjects ) . The pooled mean since 2005 was 47 mg/dl ( Table 1 ) . The time trend of mean HDL-Cholesterol was negative yet non-significant ( Figure 2 ) . National studies were evenly distributed; Southern and Tropical Latin America seemed to show higher means ( Figure 2 ) . The HDL-Cholesterol prevalence estimates were based on 34 studies ( 55 , 164 individuals ) overall . The pooled prevalence since 2005 was 48% for HDL-Cholesterol ≤40 mg/dl in men and ≤50 mg/dl in women ( Table 1 ) . The prevalence of low HDL-Cholesterol had a negative trend , yet not strong evidence supported this finding ( Figure 3 ) . National studies were evenly distributed; Central Latin America seemed to have higher rates of low HDL-Cholesterol ( Figure 3 ) . There were 84 studies ( 121 , 009 people ) included in the mean triglycerides analysis . The pooled mean was 139 mg/dl since 2005 ( Table 1 ) . The mean levels of triglycerides have slightly increased ( Figure 2 ) . Estimates from national studies did not show a strong pattern ( Figure 2 ) . Estimates from Andean and Central Latin America appeared to be higher than those from Southern and Tropical Latin America ( Figure 2 ) . Data from 70 studies ( 109 , 935 people ) informed the triglycerides prevalence estimates overall . The pooled prevalence of high triglycerides was 21% for triglycerides ≥ 200 mg/dl and 43% for triglycerides ≥ 150 mg/dl ( Table 1 ) . The prevalence of high triglycerides has increased ( Figure 3 ) , yet weak evidence supported this finding . National studies did not show any patterns ( Figure 3 ) . As it was the case in mean triglycerides , Andean and Central Latin America seemed to have higher rates of high triglycerides ( Figure 3 ) . Given the overall selection criteria ( population-based studies with blood samples to measure lipid biomarkers ) , studies had moderate risk of bias . For details about the assessment tool and our grading rationale , please refer to Supplementary file 1 pp . 08–09 .
We summarized trends in total cholesterol , HDL-Cholesterol , LDL-Cholesterol and triglycerides; in addition , we also reported on trends of dyslipidaemias with clinical relevance: high total cholesterol , high LDL-Cholesterol , low HDL-Cholesterol and high triglycerides . This work , along with other global estimates , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) can be used for surveillance of lipid levels in LAC . We have pooled recent mean and prevalence estimates , which can serve as a starting point to monitor changes during the next years . This work can also inform policies and interventions so that they can target the dyslipidaemia with the highest prevalence . Moreover , this work can inform regional practice guidelines to include local evidence and address regional needs . Although there has been a marginal decrease in the mean of the four lipid biomarkers over time , the results do not support there have been a substantial change . A substantial change was not observed for prevalence estimates either . These findings are consistent with those from a recent global analysis , in which they found little change in total cholesterol and non-HDL Cholesterol in LAC , a decrease in several areas of North America , Europe , and Oceania , while an increase in East and South East Asia ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . These remarks may suggest that there have been few policies or interventions to improve these lipid profiles in LAC; alternatively , this could suggest that available interventions were not effective . In either case , the results may show a natural progression or variation , rather than the influence of any interventions to improve lipid profiles in LAC . Nonetheless , a potential explanation could be that effective policies were in place and these prevented an increase . A pending task in LAC is a comprehensive and quantitative evaluation of policies to decrease the burden of cardio-metabolic risk factors . Overall , the global work by Taddei and colleagues suggested that lipid levels have decreased in several countries in North America , Europe and Oceania , yet lipid levels have increased in East and South East Asia; as we have discussed before , their findings for LAC agrees with ours showing little change over time ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . These patterns largely mirrors trends in cardiovascular disease mortality: death rates per 100 , 000 people in Eastern and Western Europe have decreased , death rates have increased in East and South Asia , while changes are modest in LAC ( IHME , GHDx , Viz Hub , 2020 ) . Lipid biomarkers are key cardio-metabolic risk factors , thus successful improvement in these at the patient and population level could bring gains in terms of cardiovascular outcomes reduction . Hypercholesterolemia awareness and control may be low in LAC , as it has been exemplified in some cities ( Silva et al . , 2010; Hernández-Alcaraz et al . , 2020; Lotufo et al . , 2016 ) ; we do not have any strong evidence to assume this profile has improved since then . Because awareness , treatment and control for hypertension are still insufficient , ( Geldsetzer et al . , 2019 ) despite the fact that antihypertension drugs may have better availability than lipid-lowering medications , we hypothesize poor treatment rates for dyslipidaemias in LAC . Therefore , this potential poor awareness , treatment and control rates for dyslipidaemias may have translated into the unremarkable trends herein reported for LAC . A recent global analysis also found that lipid levels have changed little in LAC , while in many high-income countries these levels have improved ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) ; this could be a consequence of better awareness and access to treatment in the latter countries . In comparison to other world regions , a recent work located High−income English−speaking countries , Europe and High−income Asia−Pacific with larger mean levels of total cholesterol and HDL-Cholesterol than LAC ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Diet and physical activity profiles , along with unequal access to primary prevention strategies , could be behind this difference . For non-HDL-Cholesterol , their findings revealed fewer regions above LAC , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) which could indicate low prescription of lipid-lowering drugs ( e . g . , statins ) in LAC . Time trends reported by the NCD-RisC largely agrees with our results , depicting LAC sub-regions with a marginal decrease since 1980 with regards to mean total cholesterol , HDL-cholesterol and non-HDL-Cholesterol ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Other world regions have experienced a marked decrease or increase; this may support our hypothesis that little attention have been paid to lipid levels in LAC , resulting in unremarkable time changes . The meta-analysis showed that the most frequent dyslipidaemia in LAC since 2005 was low HDL-Cholesterol . This could have relevant implications for clinical practice guidelines . Major international guidelines ( Grundy et al . , 2019 ) as well as guidelines from LAC ( refer to Supplementary file 1 , Table 4 for a summary of guidelines from selected countries in LAC ) have solid algorithms and recommendations on when to start pharmacological treatment . Nonetheless , statins would reduce LDL-Cholesterol with little impact on HDL-Cholesterol . Consequently , based on our results , it would be advisable to strengthen clinical guidelines with further evidence , recommendations and algorithms to improve HDL-Cholesterol; these should include strong primary prevention strategies ( e . g . , life-styles modification ) . This does not imply that lowering LDL-Cholesterol should not be important in LAC; on the other hand , this suggests that raising HDL-Cholesterol should also be addressed by guidelines and practitioners . Our findings suggest that low HDL Cholesterol is the most common dyslipidaemia trait in LAC since 2005 . Reasons behind this finding can relate to other cardio-metabolic risk factors . While raised body mass index ( e . g . , obesity ) and diabetes have a negative correlation with HDL-Cholesterol , exercise has a positive effect on this lipid fraction ( Rashid and Genest , 2007 ) . The proportion of the population with obesity and diabetes has raised substantially throughout LAC , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2019; NCD Risk Factor Collaboration ( NCD-RisC ) , 2017a; NCD Risk Factor Collaboration ( NCD-RisC ) , 2016; NCD Risk Factor Collaboration ( NCD-RisC ) , 2020b ) which also happens to be the region with one of the largest frequencies of physical inactivity ( Guthold et al . , 2018 ) . Even if clinical guidelines incorporate thorough recommendations to improve HDL-Cholesterol , this may not be achieved without policies or population-based interventions addressing the underlying ( concomitant ) cardio-metabolic risk factors . To successfully increase HDL-Cholesterol in LAC , reducing obesity and diabetes , while providing opportunities to do physical activity , are also needed . This is a comprehensive systematic review conducted in five major search engines . However , limitations should be acknowledged . First , we did not search in grey literature sources; although in these sources we could have found some reports based on national surveys ( e . g . , WHO STEPS results ) , we argue that the results are sufficiently robust to have been largely influenced or driven by research published in grey literature . Second , although we selected only population-based studies to report on the scenario at the general population level , we did not apply other criteria to avoid potential health or clinical bias . For example , we did include a report if it had excluded people on lipid lowering medications . These studies could provide slightly higher estimates , than if they had included both people with and without lipid-lowering drugs . To the best of our knowledge , there has not been a systematic assessment of lipid-lowering medication uptake across LAC , yet we argue that these drugs are still not widely prescribed . Because there is evidence suggesting limitations in accessing hypertension and diabetes medication , ( Attaei et al . , 2017; Chow et al . , 2018 ) we believe that these restrictions would be even greater for lipid-lowering drugs . In this context , we argue that this limitation may have had little impact on our estimates . Third , there were some years for which we did not retrieve any data . This limitation would not have affected the most recent trends and meta-analysis . In this line , we also did not present results for every country in LAC , limiting the extrapolation of our findings to these nations . Future studies with other analytical approaches could improve these limitations and provide information for all years and countries , or even for sub-regions within LAC ( Farzadfar et al . , 2011; NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Fourth , although we only studied adult populations , the age range of the study participants may have not been the same across all selected studies . This could have biased our estimates if , the mean levels or prevalence estimates of the herein studied lipid biomarkers were significantly larger in some age ranges . Extracting published data to verify this hypothesis would considerably reduce the number of observations because not all studies reported their findings by ( consistent ) age groups . This would also be the case for sex , as studies would not always stratify findings by sex . Fifth , extraction of other characteristics of the studies could have provided relevant information to interpret the results and to draw the scenario of lipid research in LAC; among others , relevant features could include whether point-of-care devices were used or blood samples were analysed in laboratories , and whether these followed an international standard . Sixth , LDL-cholesterol could be measured directly or estimated ( e . g . , Friedewald formula ) , though this information was not extracted to further characterize the results . The Friedewald formula is frequently used , and it may underestimate the real value ( Meeusen et al . , 2014 ) . If so , our estimates for LDL-cholesterol need to be interpreted cautiously; further research is needed to understand the magnitude of this potential underestimation and , if needed , to develop a more accurate formula for populations in LAC . In this line , a recent global work reported slightly larger mean levels of non-HDL Cholesterol in comparison to our LDL-cholesterol levels ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Although these are not identical metrics , the agreement between these is typically good . Speculatively , we could hypothesize that some of surveys herein summarized used the Friedewald formula , and the LDL-cholesterol levels were underestimated . This could potentially explain the different results ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Some studies only reported the mean or the prevalence estimate . Although the prevalence estimates are relevant , from a public health perspective the mean and the shape of the distribution are also important and should be reported whenever possible . This calls for authors and reporting guidelines to provide both metrics . Similarly , studies did not report their results by sex or ( consistent ) age groups , precluding us to make estimates by gender and age . A global endeavour reported on levels of total cholesterol for each country in the world until 2010 ( Farzadfar et al . , 2011; these work has been recently updated ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . The evidence from LAC was limited in comparison to our work , which also expands the evidence to include prevalence estimates . Beyond the CARMELA study which comprised seven cities in LAC , ( Pramparo et al . , 2011 ) and the CESCAS study which included cities in three countries , ( Rubinstein et al . , 2011 ) there is a dearth of multi-country studies addressing lipid profiles and other cardio-metabolic risk factors in LAC . These studies and other local research suggested that low HDL-Cholesterol was the most common dyslipidaemia . Our work confirms this observation and strengthens the evidence so that it can inform policies , interventions and guidelines . Recently , an international work updated the 2010 global total cholesterol estimates ( Farzadfar et al . , 2011 ) and provided results for HDL-Cholesterol and non-HDL-Cholesterol ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) . Our work complements this evidence by providing mean levels for other lipid fractions and prevalence estimates of clinically relevant dyslipidaemias in LAC . Their total cholesterol mean estimates for LAC are largely consistent with our findings , and so are the mean levels for HDL-Cholesterol; however , their non-HDL-Cholesterol means are larger than our LDL-cholesterol means . The reasons could be different methodology and analytical approach ( please , refer to the limitations section ) . Our findings , as those by the NCD-RisC , ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2020a ) suggested that mean levels of lipid biomarkers are not the same across LAC countries . Although LAC hosts mostly middle-income countries , there are large within countries inequalities; this is also seen in the Caribbean , where some islands may be high-income countries , but inequalities still exist . Different levels of poverty , access to healthy foods , opportunities for physical activity , and a still fragile primary health system , may explain the differences between sub-regions and countries in LAC . A seminal work in Africa followed a similar methodology and reported , for the general population , a prevalence of 23% for high total cholesterol ( our estimates were seven percentage points higher ) ; 41% for low HDL-Cholesterol ( our estimates were seven percentage points higher ) ; 25% for elevated LDL-Cholesterol ( our estimates were 15 percentage points higher ) ; for triglycerides their prevalence was 16% ( our results were 23 percentage points higher ) ( Noubiap et al . , 2018 ) . They also reported estimates for other populations ( e . g . , people with diabetes ) ( Noubiap et al . , 2018 ) . The higher -worse- estimates we reported for LAC could suggest LAC is ahead in the epidemiological and nutritional transition , in comparison with Africa . An interesting finding , which deserves in-depth scrutiny , is that we found a marginal decrease in mean total cholesterol , yet also a marginal increase in the prevalence of high total cholesterol . We propose two hypotheses . First , aging of the population . Older people may have larger prevalence of dyslipidaemias . As populations are aging and living longer , we could expect larger prevalence estimates , while mean levels get ‘diluted’ or do not necessarily change . Second , and closely related , is that we do not know the drivers of these changes , i . e . , we still need to answer whether the mean or the tails of the distribution are changing and driving the trends . For blood pressure , it has been suggested that the mean is the main driver of trends in raised blood pressure prevalence ( NCD Risk Factor Collaboration ( NCD-RisC ) , 2018 ) ; whether this is the case for lipid biomarkers and dyslipidaemias is still unknown . A third option could be the uptake of lipids-lowering drugs . As more people take these drugs , the population mean would decrease while the prevalence would not change ( or even increase ) . However , as we have argued before , lipids-lowering medication coverage may still be limited in LAC . Improving prevention , care and management for diabetes and hypertension is a clear priority globally and in LAC ( González-Villalpando et al . , 1999; Hall Martínez et al . , 2005; Hernández-Hernández et al . , 2017 ) . Nonetheless , lipid profiles are relevant for public health and clinical practice as well ( Grundy et al . , 2019; NICE , 2014 ) . In fact , they are a major risk factor for cardiovascular events including ischemic heart disease and stroke ( Lewington et al . , 2007; Di Angelantonio et al . , 2009 ) . Also , lipid biomarkers are predictors in several cardiovascular risk scores used to guide treatment allocation for primary cardiovascular prevention ( Goff et al . , 2014; WHO CVD Risk Chart Working Group , 2019; Hajifathalian et al . , 2015 ) . This work provides timely regional evidence to start a research and policy agenda to improve lipid profiles in LAC . Our work has compiled the largest number of data sources across years and countries in LAC . Consequently , it is uniquely positioned to inform local and regional authorities about recent trends in lipids distribution and prevalence estimates . The results could have multiple pragmatic applications . First , they could be used as a baseline upon which build surveillance systems to monitor future trends of lipid biomarkers . Second , our comprehensive search strongly suggests that there is a lack of evidence from several countries , particularly in Central America and the Caribbean . Local and regional authorities should conduct epidemiological studies or population-wide surveys ( e . g . STEPS approach ) ; alternatively , when these have already been conducted , data could be open access for research purposes . Ideally , where data are available , these should meet the FAIR acronym: findable , accessible , interoperable , and reusable . Unfortunately , the first two elements of the acronym are perhaps the least frequent , yet the most important for scientific use of available data . Third , our results could inform prevention strategies and policies . Because we have reported on different lipid fractions , medication ( e . g . , statins ) could be prioritized where LDL-cholesterol is higher , while diet or healthy lifestyles could be a priority where HDL seems to be the most important issue . Levels and prevalence estimates of unhealthy total cholesterol , LDL-Cholesterol , HDL-Cholesterol and triglycerides seemed not to have substantially changed over the last years in LAC . Since 2005 across LAC , the most common dyslipidaemia was low HDL-Cholesterol . These results should inform policies so that they can start or strengthen strategies to improve lipid profiles , thus reducing the burden of cardiovascular events which are strongly associated with unhealthy lipid levels .
This is a systematic review of the literature ( PROSPERO CRD42019120491; PRISMA Checklist available in Supplementary file 1 ) . We aimed to identify trends in total cholesterol and cholesterol fractions in LAC general population; also , to ascertain which dyslipidaemia ( e . g . low HDL-cholesterol ) is the most prevalent in LAC . Research reports were analysed if they targeted adult men and women of the general population . We focused on LAC populations , thus studies with LAC populations in countries outside the LAC region , and studies with only foreign populations in LAC nations , were excluded . Population-based studies were defined as those which followed a random sampling of the general population . Conversely , studies addressing specific populations ( e . g . , shanty towns ) , those with patients ( e . g . , stroke survivors ) , or people with risk factors ( e . g . , smokers ) , were excluded . The outcomes of interest were lipid biomarkers levels and dyslipidaemia prevalence . We focused on clinically and public health relevant lipid biomarkers: total cholesterol , HDL-Cholesterol , LDL-cholesterol and triglycerides . Only studies in which lipid biomarkers were measured with valid methods ( e . g . laboratory or point-of-care devices ) were included; that is , studies which results relied only on self-reported information were excluded . The search was conducted on December 21st , 2018 . We used Scopus , LILACS , Embase , Medline and Global Health; the last three through Ovid . In all of these , the search was conducted without time or language restriction . The search terms are available in Supplementary file 1 pp . 06–07 . Results from each search engine were downloaded and saved in EndNote where duplicates were dropped . A second search for duplicates was conducted with Rayyan , an online tool for systematic reviews ( Ouzzani et al . , 2016 ) . Titles and abstracts were independently reviewed by two researchers ( RMC-L and NP-B; CA-R and CJB-M; LA-F and DS-V ) , and discrepancies were solved by consensus or a third party ( AB-O ) . After this screening phase , selected reports were downloaded and independently studied in detail by two researchers ( RMC-L and NP-B; CA-R and CJB-M; LA-F and DS-V ) ; discrepancies were solved by consensus or by a third party ( AB-O ) . Finally , selected studies were scrutinized again to check for data duplication , i . e . different reports that used the same data ( e . g . , a national survey ) . In this case , the paper which presented more information ( e . g . , all four lipid biomarkers ) , or the one with the largest sample size , was included in the systematic review and meta-analysis . In other words , we aimed to include each study or survey once . The unit of analysis is a study . An extraction form was developed by the authors and tested with a random sample of selected studies; the form was not modified after data collation started . This form included study’s characteristics: mean age , proportion of men , year of data collection , and if it was a nationally representative sample . The extraction form also collated the mean and prevalence estimate as well as a dispersion measurement ( e . g . standard deviation or confidence interval ) of the available lipid biomarkers . We used the risk of bias tool developed by Hoy and colleagues ( Hoy et al . , 2012 ) . Notably , this tool was also used by a systematic review on a similar topic ( Noubiap et al . , 2018 ) . These criteria were implemented in an Excel spreadsheet and evaluated independently by two reviewers ( RMC-L and NP-B; CA-R and CJB-M; LA-F and DS-V ) ; discrepancies were solved by consensus or a third party ( AB-O ) . We present both a narrative and quantitative summary . The narrative summary described the study’s characteristics , while the quantitative summary explored the trends of the lipid biomarkers means as well as prevalence estimates . In addition , following a random-effects meta-analysis and using data from 2005 onwards , we computed the pooled mean and the pooled prevalence estimate for each lipid biomarker and dyslipidaemia trait . We only used the most recent data ( i . e . , from 2005 ) to report on the current -or most recent- levels in LAC , rather than summarizing all available information with no clear time frame . Because the selected studies had different sample size and scope ( e . g . , national surveys versus community studies ) , we conducted the random-effects meta-analysis . Unlike a fixed-effect meta-analysis , in a random-effects meta-analysis large studies would not drive or bias the pooled estimates . This is a systematic review of published scientific evidence and open information . Human subjects did not participate in this work directly and there was no intervention . Approval from an IRB/ethics committee was not requested . The funder had no role in the research question , data collation , analysis or reporting of the results . All the authors collectively are responsible for data accuracy and they all have approved the submitted work . | Cholesterol and triglycerides are fatty substances found in the blood . They are crucial components of cell membranes and important for a variety of processes in the body . But , too much , or too little blood fat can damage the blood vessels . For example , high levels of fat in the blood can clog arteries , which can increase the chances of heart disease , heart attacks and strokes . Fat starts to build up if ‘bad’ fats , such as triglycerides and LDL cholesterol , are too high . But it can also happen if levels of 'good' fats , like HDL cholesterol , are too low . The causes of , and treatments for , these different types of dyslipidaemia ( or fat levels outside normal ranges ) are not the same . So , to plan interventions effectively , public health authorities need to know which type of blood fat imbalance is most common in the local population , and whether this has changed over time . In many parts of the world , this kind of information is available , but in Latin America and the Caribbean the data is incomplete . To address this , Carrillo-Larco et al . reviewed around 200 previous studies from across Latin America and the Caribbean . This revealed that , since 2005 , low HDL cholesterol has been the most common type of dyslipidaemia in this region , followed by elevated triglycerides , and third , high LDL cholesterol . These patterns have changed little over the years . In many parts of the world , public health guidelines for dyslipidaemia focus on treatment specifically for high LDL cholesterol . But this new data suggests that guidelines should also include recommendations for HDL cholesterol , in particular in Latin America and the Caribbean . And , with a clearer understanding of the current pattern of blood fat imbalances in this region , researchers now have a baseline against which to measure the success of any new health policies . In the future , a multi-country study to measure blood fats in the general population could provide even more detail . But , until then , this work provides a starting point for customised health interventions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"medicine"
] | 2020 | A systematic review of population-based studies on lipid profiles in Latin America and the Caribbean |
In fibroblasts , large Ca transients activate massive endocytosis ( MEND ) that involves membrane protein palmitoylation subsequent to mitochondrial permeability transition pore ( PTP ) openings . Here , we characterize this pathway in cardiac muscle . Myocytes with increased expression of the acyl transferase , DHHC5 , have decreased Na/K pump activity . In DHHC5-deficient myocytes , Na/K pump activity and surface area/volume ratios are increased , the palmitoylated regulatory protein , phospholemman ( PLM ) , and the cardiac Na/Ca exchanger ( NCX1 ) show greater surface membrane localization , and MEND is inhibited in four protocols . Both electrical and optical methods demonstrate that PTP-dependent MEND occurs during reoxygenation of anoxic hearts . Post-anoxia MEND is ablated in DHHC5-deficient hearts , inhibited by cyclosporine A ( CsA ) and adenosine , promoted by staurosporine ( STS ) , reduced in hearts lacking PLM , and correlates with impaired post-anoxia contractile function . Thus , the MEND pathway appears to be deleterious in severe oxidative stress but may constitutively contribute to cardiac sarcolemma turnover in dependence on metabolic stress .
Acute ischemic events are now a leading cause of death in the entire world ( World Health Organization , 2013 ) . Ironically , much of the damage from ischemic events can occur during reoxygenation of tissue when mitochondria begin to generate reactive oxygen species ( ROS ) at increased rates ( Jennings , 2013 ) . One of the hallmarks of reperfusion injury is a pronounced swelling of mitochondria , and the extensive opening of mitochondrial permeability transition pores ( PTPs ) that underlies this swelling ( Haworth and Hunter , 1979 ) is a nearly decisive step on the path to cell demise ( Halestrap et al . , 2004 ) . Associated with this progression , mitochondria release cytochrome c to the cytoplasm and thereby activate apoptosis programs that can definitively establish cell fate ( Gottlieb , 2011 ) . Given that both outer and inner mitochondrial membranes become permeable during reperfusion injury , mitochondria with certainty release a large number of metabolites and small proteins . Therefore , the question arises whether , besides cytochrome c , other molecules released from mitochondria might serve signaling functions . Our analysis of Ca-activated endocytosis in BHK fibroblasts ( Hilgemann et al . , 2013 ) suggests that mitochondrial release of coenzyme A ( CoA ) may be important . This hypothesis arises from two facts . First , mitochondria accumulate CoA to high concentrations with respect to the cytoplasm ( Leonardi et al . , 2005 ) . Second , many key cytoplasmic enzymes are modulated by binding long-chain acyl CoA ( Faergeman and Knudsen , 1997 ) , a CoA metabolite that will be generated immediately if CoA is released ( Idell-Wenger et al . , 1978 ) . These include glycogen synthase ( Wititsuwannakul and Kim , 1977 ) , glucose-6-phosphate dehydrogenase ( Kawaguchi and Bloch , 1974 ) , and perhaps AMP kinase ( Faergeman and Knudsen , 1997 ) , as well as KATP potassium channels ( Shumilina et al . , 2006 ) . Our study in BHK cells ( Hilgemann et al . , 2013 ) suggests that release of CoA from mitochondria may promote the palmitoylation of surface membrane proteins , subsequent to generation of acyl CoA . Further , our study suggests that extensive palmitoylation promotes internalization of membrane as a form of ‘lipid raft’-dependent endocytosis ( Doherty and McMahon , 2009 ) . Beyond the function of PTPs at the threshold of cell death , it is now established that transient mitochondrial PTP openings play important physiological functions by causing transient mitochondrial depolarizations that release matrix Ca to the cytoplasm accompanied by the generation of superoxide flashes ( Brenner and Moulin , 2012; Zhang et al . , 2013; Zhou and O’Rourke , 2012 ) . In this light , a possibility emerges that cytoplasmic free CoA could physiologically play important second messenger functions in dependence on its release from mitochondria . On the one hand , mitochondrial depolarization will favor reverse CoA transport to the cytoplasm ( Tahiliani , 1991 ) , and on the other hand transient PTP openings may allow enough CoA to escape from the matrix space , down its more than 50-to-1 gradient , to significantly increase the low micromolar concentration of free cytoplasmic CoA ( Idell-Wenger et al . , 1978 ) . We describe here experiments that address these possibilities in cardiac myocytes and in intact cardiac tissue . First , we show that the function and abundance of cardiac Na/K pumps and Na/Ca exchangers is affected by the expression of the plasmalemma-directed acyl transferase , DHHC5 , then we show that massive endocytic ( MEND ) responses in isolated myocytes depend on acyl CoA and DHHC5 activity , and that MEND responses occur when PTP openings occur during the reoxygenation of anoxic cardiac muscle . Finally , we show that two cardiac membrane proteins are indeed increasingly palmitoylated during reoxygentation , that internalized PLM ( phospholemman ) is more heavily palmitoylated than PLM on the cell surface , and that DHHC5 activity significantly impacts cardiac contractile recovery after anoxia .
Figure 1 demonstrates that expression of the DHHC5 acyl transferase strongly influences Na/K pump activity in cultured cardiac myocytes , namely human fibroblast-derived cardiac myocytes ( iCell Cardiomyocytes , [Ma et al . , 2011] ) . These myocytes are transfected easily to overexpress and knockdown regulatory proteins , they highly express cardiac-specific proteins , such as PLM and the Na/Ca exchanger , NCX1 , and ion transport activities are equivalent to those of adult myocytes ( Fine et al . , 2013 ) . Measuring maximal Na/K pump currents exactly as described in our study ( Fine et al . , 2013 ) , the first two bar graphs of Figure 1 show that overexpression of DHHC5 , together with green fluorescent GFP to identify transfected myocytes , decreases Na/K pump currents by 55% . This decrease becomes still somewhat larger with dual DHHC2/DHHC5 expression . Conversely , knockdown of DHHC5 by siRNA causes a 38% increase of Na/K pump currents , compared to myocytes transfected with scrambled siRNA ( ‘siRNA control’ ) , and dual DHHC5/DHHC2 knockdown causes a 90% increase . Although these changes may arise from either a change of pump activity or a change of pump density in the sarcolemma , data presented next indicate that changes of transporter localization play a large role . 10 . 7554/eLife . 01295 . 003Figure 1 . Na/K pump current densities are inversely dependent on DHHC5 expression in human iCell Cardiomyocytes . From left to right , Na/K pump current densities in iCell Cardiomyocytes transfected with GFP to identify transfected cells , DHHC5 , DHHC5 and DHHC2 , transfected with scrambled siRNA , siRNA for DHHC5 , and siRNA for DHHC5 and DHHC2 . Na/K pump currents are decreased by 55% and 61% with DHHC5 and DHHC5/DHHC2 overexpression , respectively . Current densities are increased by 38% with DHHC5 knockdown and by 90% with DHHC5/DHHC2 knockdown , n >6 for all results . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 003 We next describe myocytes and cardiac tissue from mice that are homozygous for a hypomorphic allele of DHHC5 ( gene-trapped , GT ) with cardiac DHHC5 expression decreased by >80% ( Li et al . , 2011 ) . DHHC5-GT mice have significantly reduced growth rates . Between 4 and 5 weeks , the average weights of DHHC5-GT mice ( 14 ± 0 . 8 g ) were 24% less than control litter mates ( 19 ± 0 . 9 g ) . Between 12 and 14 weeks , the average weights of DHHC5-GT mice ( 21 ± 0 . 2 g ) were 13% less than control litter mates ( 24 ± 0 . 2 g ) . The dimensions of isolated myocytes from young DHHC5-GT animals were also significantly smaller than those from matched WT animals . Therefore , we employed animals 12 to 14 weeks of age for the studies to be described . We used patch clamp with square wave voltage pulses to measure the myocyte surface area as cell electrical capacitance ( Cm ) ( Lariccia et al . , 2011 ) , assuming that 1 pF constitutes 100 μm2 of membrane . Cell area ( i . e . , Cm ) /volume ratios of myocytes were then determined as described in Figure 2A . The volume of each myocyte was estimated from a micrograph of the myocyte still resting on the surface of the chamber employed . From the micrograph , the two-dimensional area of the myocyte was determined in square microns using Image-J software ( http://rsbweb . nih . gov/ij/ ) . The volume was then estimated by assuming an average 5-to-1 ratio of myocyte width to thickness ( volume ( μm3 ) = myocyte area2/myocyte length/5 ) and converted to pl . As shown in bar graphs in Figure 2B , C , the average surface areas of DHHC5-GT myocytes were modestly but significantly increased in comparison to WT myocytes from matched animals . The average volumes of DHHC5-GT myocytes were insignificantly decreased in myocytes from 12 to 14 week old animals . As shown in Figure 1D , the average myocyte area/volume ratio , calculated in pF/pl , was highly significantly increased by 24% . Figure 2E shows that comparable results were obtained via NIC recordings in right ventricles from DHHC5-GT and litter mate mice . The capacitive NIC signal ( i . e . , sarcolemma area in a hemisphere below the 0 . 6 mm recording electrode ) was increased significantly by 21% , on average , in DHHC5-GT ventricles ( n = 5 ) vs WT ventricles ( n = 5 ) . 10 . 7554/eLife . 01295 . 004Figure 2 . DHHC5-GT cardiac myocytes . ( A ) Representative myocyte micrograph used to determine myocyte area/volume ratios . 25 × LWD lens . ( B ) Myocyte surface areas ( Cm ) are significantly increased in DHHC5-GT myocytes vs matched WT myocytes . ( C ) Myocyte volumes are not significantly different in DHHC5-GT myocytes from 4 to 6 week old mice . ( D ) Myocyte surface area/volume ratios are increased by 27% in DHHC5-GT myocytes . ( E ) Cardiac tissue Cm , monitored via NIC recording , is increased by 21% in right ventricular strips from DHHC5-GT mice ( n = 5 ) vs control litter mates ( n = 5 ) . ( F ) Na/K pump current densities are increased by 32% in DHHC5-GT myocytes . ( G ) An average 17% increase of NCX1 current density is not significant . ( H ) Amine PEGylation assay to determine the surface membrane fractions of PLM and NCX1 in WT and DHHC5-GT hearts . Western blot density profiles are shown for three hearts , one control heart that was not PEGylated ( gray ) , one WT heart ( black ) and one DHHC5-GT heart ( red ) . PLM , actin and NCX1 are blotted , and the PEGylated PLM and NCX1 densities ( i . e . , protein resident in the cell surface ) are shifted 5 kD from control densities . ( I ) Fractions of PLM and NCX1 that can be PEGylated , and therefore reside in the sarcolemma , are increased by 27 and 23% , respectively , in DHHC5-GT hearts . For both WT and DHHC5-GT hearts , n = 3 . For all data from myocytes , n >40 using myocytes from three animals . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 004 Next we show in Figure 2F , G that the average current density of Na/K pumps is significantly increased by 37% in DHHC5-GT myocytes , while a 17% increase of Na/Ca exchange ( NCX1 ) current density was not significant . Finally , we tested directly whether pumps and exchangers are on average more retained in the sarcolemma vs internal membranes in DHHC5-deficient vs WT hearts . To do so , we employed a PEGylation assay described in ‘Materials and methods’ to determine the fractions of NCX1 and the regulatory Na/K pump subunit , PLM , resident in the cell surface . Both of these proteins are advantageous for amine PEGylation because they have extracellular N-termini ( Geering , 2006; Ren and Philipson , 2013 ) . Using a 5 kD NHS-PEGylation reagent to label the outer sarcolemma of intact , perfused hearts , Figure 2H shows typical Western blot density profiles from this assay for PLM and NCX1 . Using a least squares Gaussian fit to determine relative protein densities , 34 ± 3% of PLM and 31 ± 2% of NCX1 bands were shifted 5 kD to higher molecular weights in hearts from WT animals . As shown in Figure 3B , the calculated surface fractions of both membrane proteins were significantly increased in DHHC5-GT hearts vs WT hearts , namely by 27 ± 5 and 23 ± 2% , respectively . 10 . 7554/eLife . 01295 . 005Figure 3 . MEND evoked by three different stimuli is inhibited in DHHC5-deficient myocytes . ( A ) KSP MEND . Cytoplasmic application of KSP solution via the patch pipette causes MEND responses in WT myocytes that take place with rough time constants of 2 min and that amount to 24% of the sarcolemma , on average . The KSP MEND responses are decreased by 74% on average in myocytes from DHHC5-GT animals . ( B ) H2O2 MEND . The application and removal of H2O2 ( 80 μM ) results in 38% MEND responses , on average , with MEND occurring only after the oxidative stress is relieved . These MEND responses require that the cytoplasmic solution contains acyl CoA ( mCoA , 15 μM ) . They are more reliable when glutathione ( 3 mM ) is included in the cytoplasmic solution , presumably because final steps of endocytosis require a reducing environment . ( C ) GPCR MEND . MEND occurs slowly over 30 min in the presence of phenylephrine ( 30 μM ) when the cytoplasm contains mCoA ( 20 μM ) . These endocytic responses , while slow , amount on average to >40% of the sarcolemma , and they are inhibited by 67% in myocytes from DHHC5-GT animals . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 005 In BHK cells , large MEND responses can be evoked by rapidly perfusing the cytoplasm of cells via the patch clamp pipette with solutions containing metabolites that promote mitochondria to open PTPs ( ‘KSP solution’ , containing potassium [100 mM] , succinate [5 mM] , and inorganic phosphate [Pi , 1 mM] ) . As shown in Figure 3A , WT myocytes generate 33% MEND responses upon pipette perfusion of KSP solution with rough time constants of 2 . 5 min , and these responses are decreased significantly by 75% in DHHC5-GT myocytes . When the cytoplasmic solution employed in BHK cells contains myristoyl CoA ( mCoA ) , application of H2O2 ( hydrogen peroxide ) ( 80 μM ) for a few minutes primes cells to undergo MEND when the oxidative stress is removed . As shown in Figure 3B , transient H2O2 application ( 80 μM ) also causes large MEND responses in cardiac myocytes ( ∼35% ) when mCoA ( 15 μM ) is included in the pipette . These responses were also strongly reduced in DHHC5-GT myocytes ( p<0 . 01 ) . In BHK cells , activation of PKCs with diacylglycerol surrogates induces MEND responses that constitute 20–40% of the plasmalemma when mCoA is present in cytoplasmic solutions ( Hilgemann et al . , 2013 ) . The equivalent experiments caused much slower responses in myocytes , as well as when phospholipase C-coupled GPCRs ( G-protein coupled receptor ) were activated ( e . g . , by adenosine or phenylephrine ) in the presence of mCoA . Figure 3C illustrates MEND responses that occur when the alpha receptor agonist , phenylephrine ( 50 μM ) ( Puceat et al . , 1994 ) , is used in the presence of mCoA ( 20 μM ) . MEND did not occur unless both phenylephrine and mCoA were present ( n = 10 ) . While slower than responses to KSP solution and transient oxidative stress , sarcolemma area decreases on average by about 40% over 30 min in WTl myocytes , but only 14% in DHHC5-GT myocytes ( p<0 . 05 ) . In murine myocytes , Ca influx can activate MEND responses that occur much more rapidly than those just described ( Lariccia et al . , 2011 ) . When activated by reverse Na/Ca exchange , MEND stops abruptly when Ca influx is terminated ( Lariccia et al . , 2011 ) , suggesting that Ca is acting via a rather direct mechanism that is likely different from MEND responses just described . Nevertheless , Figure 4A demonstrates that Ca-activated MEND is highly significantly reduced in DHHC5-GT myocytes vs WT myocytes . Figure 4A illustrates Ca-activated MEND in a WT myocyte , the top trace being Cm ( i . e . , membrane area ) and the bottom trace being membrane current . When Ca influx is activated , Cm begins to decline after a 3 to 4 s delay and then declines further by 28% with a rough time constant of 8 s . As documented in Figure 2G , exchange currents were on average larger in DHHC5-GT myocytes . However , as shown in Figure 4B , MEND responses were decreased from 27% on average in WT myocytes to only 3% in DHHC5-GT myocytes . Figure 4B shows further that MEND responses in WT myocytes were decreased by 70% when myocytes were preincubated with the PTP/cyclophillin D-specific cyclosporine , NIM811 ( N-methyl-4-isoleucine cyclosporine ) ( 3 μM ) ( Waldmeier et al . , 2002 ) , for 1 hr . However , inclusion of a high CoA concentration ( 4 mM ) in the cytoplasmic solution to acutely inhibit DHHC acyl-transferease activity ( Hilgemann et al . , 2013 ) did not significantly inhibit Ca-activated MEND in WT myocytes . We conclude therefore that palmitoylation reactions do not occur during the protocol of Figure 4A , although this form of MEND nevertheless depends on the palmitoylation state of the sarcolemma . 10 . 7554/eLife . 01295 . 006Figure 4 . DHHC5 facilitates Ca-activated MEND in cardiac myocytes . ( A ) Typical endocytic response caused by Ca influx ( i . e . , reverse Na/Ca exchange current ) in a WT myocyte . Top record , Cm ( i . e . , membrane area ) ; bottom record , membrane current . MEND occurs during Ca influx with a time constant of ∼8 s . ( B ) Ca-activated MEND in WT myocytes amounts to 27% of the sarcolemma on average . MEND is reduced by >80% in myocytes from DHHC5-GT animals and reduced 70% by pretreatment of cells with NIM811 ( 3 μM ) for 1 hr . However , MEND is not reduced by inclusion of a high CoA concentration ( 4 mM ) to acutely block acyl transferace activity . ( C ) Typical MEND response in a cardiac myocyte over-expressing NCX1 by 10-fold . Top trace , Cm ( i . e . , membrane area ) ; bottom trace , membrane current . In these myocytes with four to sixfold larger Na/Ca exchange currents , Ca influx causes MEND that amounts to more than 50% of the sarcolemma in less than 2 s . ( D ) Large Ca transients overcome the dependence of MEND on palmitoylation . Ca-activated MEND in NCX1-overexpressing myocytes is unaffected by 1 hr treatment with NIM811 ( 3 μM ) or by a high cytoplasmic CoA concentration ( 4 mM ) to acutely inhibit palmitoylation . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 006 Figure 4C solidifies this interpretation using myocytes that overexpress Na/Ca exchangers ( NCX1 ) by 10-fold . Both the time course of MEND and the decay of NCX1 current during Ca influx are accelerated nearly 10-fold ( τ <1 s ) , and myocytes internalize more than 50% of their sarcolemma within 1–2 s . Using the myosin II inhibitor blebbistatin ( Farman et al . , 2008 ) ( 10 μM in all solutions ) to inhibit contraction , myocytes did not contract or show obvious morphological changes during these profound responses . As shown in Figure 4D , neither pretreatment with NIM811 for 1 hr ( 3 μM ) nor inclusion of a high CoA concentration ( 4 mM ) in the cytoplasmic solution significantly affected these responses . Clearly , Ca can promote MEND in myocytes by a mechanism that is promoted by , but does not require , membrane protein palmitoylation . Reperfusion of ischemic cardiac tissue is one of the best defined circumstances in which PTP openings occur profusely and correlate with a significant amount of reoxygenation injury after ischemic or anoxic episodes ( Halestrap , 2009 ) . Importantly , much of this damage can be prevented by chemical preconditioning of cardiac tissue with hormones , such as adenosine , that activate PKCε and likely inhibit PTP openings ( Baines et al . , 2003 ) . Given evidence that significant endocytosis can occur in cardiac ischemia/reperfusion via unconventional mechanisms ( Pierre et al . , 2010 ) , we tested next whether MEND indeed occurs during reoxygenation of anoxic cardiac tissue . Figure 5A illustrates NIC recording in rapidly superfused right ventricular strips ( 0 . 5–0 . 7 mm thick; 36°C ) . Using 15 kHz sinusoidal voltage oscillations , the conductance signal component ( IG ) decreases and the capacitive signal component ( ICAP ) increases as the perturbing electrode ( a 0 . 6 mm Ag/AgCl pellet , recessed 0 . 4 mm in a glass capillary ) is brought close to and touches the muscle surface . The conductance signal ( IGX ) then reflects conductance of the extracellular space , which constitutes about 30% of the heart volume . Capacitive signals reflect the relative amount of surface membrane in a hemisphere of tissue below the electrode . The capacitive signal showed little change over 30 min after oxygen-saturated solution was switched to oxygen-free , glucose-free solution with 5 mM deoxyglucose to promote nucleotide depletion . However , when oxygen-saturated , glucose-containing solution was reintroduced , NIC signals decreased over 25 min by 21% on average . 10 . 7554/eLife . 01295 . 007Figure 5 . Electrical recordings of MEND during reoxygenation of anoxic cardiac muscle . ( A ) Noninvasive Cm ( NIC ) recording in superfused right ventricular strips . The capacitive signal ( ICAP ) , reflecting sarcolemmal area in a hemisphere of tissue beneath the electrode , is stable during 30 min periods of anoxia , but the NIC signal decreases on average by 21% over 25 min upon reoxygenation of the tissue . ( B ) Loss of sarcolemma during reoxygenation for 25 min after a 30 min period of anoxia . The first bar graph quantifies the composite response for control ( WT ) muscles . Subsequent bars document from left to right that MEND is strongly inhibited by perfusing hearts for 15 min with FA free albumin ( 60 μM ) before isolating muscle strips , is strongly enhanced by a low concentration ( 0 . 1 μM ) of staurosporine , is potently inhibited by cyclosporine ( CyS ) ( 1 μM ) , is strongly reduced in muscles from DHHC5-deficient mice , is strongly reduced by the ‘preconditioning’ hormone , adenosine ( 100 μM ) , and is strongly reduced in cardiac muscle lacking PLM . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 007 Composite results from similar recordings , shown in Figure 5B , support the interpretation that MEND is occurring in intact cardiac muscle by mechanisms that are related to those described in BHK cells ( Hilgemann et al . , 2013 ) . First , the cardiac MEND responses were strongly decreased when muscles were taken from hearts that had been FA ( fatty acid ) -depleted by 15 min arterial perfusion of FA-free albumin ( 60 μM ) prior to the experiments . Second , a low concentration of staurosporine ( STS; 0 . 1 μM ) , which promotes MEND responses in BHK cells ( Hilgemann et al . , 2013 ) , increased sarcolemma loss to 35% . Third , cyclosporine A ( CsA; 1 μM ) strongly decreased sarcolemma loss . Fourth , the preconditioning hormone , adenosine ( 0 . 1 mM ) ( Ytrehus et al . , 1994 ) , decreased sarcolemma loss by 52% . And fifth , MEND was nearly ablated in muscles from DHHC5-GT animals . As described in the accompanying article ( Hilgemann et al . , 2013 ) , MEND appears to be ‘cargo-dependent’ , being promoted by overexpression of the palmitoylated Na/K subunit , PLM ( Tulloch et al . , 2011 ) , in T-Rex293 cells . Therefore , we examined whether the occurrence of MEND would be altered in cardiac muscle from PLM-deficient mice . As shown in the last bar graph of Figure 5B , reoxygenation-induced MEND was reduced by 58% in right ventricles from PLM knockout mice ( Bell et al . , 2008 ) . As an independent verification of the occurrence of MEND , we describe in Figure 6 the fluid-phase uptake of FITC-labeled dextran ( 4000 MW ) by myocytes in arterially perfused hearts . Hearts were perfused retrograde with either FA free- or 2:1 myristate-loaded albumin ( 60 μM ) during the same anoxia-reoxygenation protocol . During reoxygenation , hearts were perfused with FITC-dextran ( 0 . 5 mM ) for 25 min followed by washout at room temperature for 10 min . Subsequently , epicardial myocytes were imaged several cell layers below the outer cardiac surface; the outer most myocytes were excluded because they are atypical , being enriched in cardiac stem cells ( Schlueter and Brand , 2012 ) . Hearts perfused with FA—free albumin reveal diffuse staining of a limited fraction of myocytes . This reflects nonspecific dextran uptake as a result of reperfusion stress . Hearts perfused with FA reveal , in addition to diffuse staining of a fraction of myocytes , bright punctate staining that is indicative of dextran uptake into large endosomes and vacuoles . Punctate staining becomes very profuse in hearts perfused with STS ( 0 . 1 μM ) and is absent when CsA ( 3 μM ) is perfused with albumin and FA . Statistical analysis , performed as described in Methods , is provided in Figure 6B , together with the additional result that dextran uptake was negligible in the same protocol in ventricles from DHHC5-GT mice . 10 . 7554/eLife . 01295 . 008Figure 6 . Optical recordings of MEND during reoxygenation of anoxic cardiac muscle . ( A ) Micrographs of FITC-dextran uptake in arterially perfused mouse hearts after 30 min of anoxia , followed by 25 min reoxygenation with perfusion of FITC-dextran ( 0 . 5 mM ) and a 10 min washout at 25°C . Confocal images are from myocytes that are a few cell layers below the outer left ventricular cardiac surface . Scale bar , 30 μm . ( B ) Composite results for hearts with control perfusate ( i . e . , 0 . 1 mM albumin with 0 . 1 mM FA ) , with perfusate containing 0 . 1 mM FA-free albumin , with control perfusate containing cyclosporine ( CsA , 2 μM ) , with control perfusate containing staurosporine ( STS , 0 . 1 μM ) , and DHHC5-GT hearts with normal perfusate . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 008 To test whether palmitoylation of membrane-associated proteins is indeed enhanced during reoxygenation of anoxic cardiac tissue , we employed a resin assisted capture assay for acylated proteins ( acyl-RAC ) ( Forrester et al . , 2010 ) to determine the palmitoylation status of proteins that might be important in MEND . In addition to PLM , we examined flotillin-2 because flotillins may be involved in the ordering of proteins into Lo domains ( Langhorst et al . , 2005 ) . In this assay , free cysteines are initially blocked , followed by deacylation of palmitoylated residues with hydroxylamine ( NH2OH ) , cysteine-specific pull-down , and quantification of captured , deacylated proteins via Western blotting . The presence of a single acylation site is ‘positive’ and multiple vs single acylations are not distinguished . Proteins that are stably palmitoylated , such as caveolins ( Parat and Fox , 2001 ) , can be used as loading controls in the assay ( Tulloch et al . , 2011 ) . Usually , we analyzed caveolin-3 ( CAV3 ) content after stripping secondary antibodies employed in initial blots , but results were very similar when analyzed in relation to Western blot loading controls of initial lysates . Figure 7A illustrates Western blots for a pair of hearts . One was subjected to the anoxia/reoxygenation protocol of Figure 5A , and the second heart was perfused with oxygen-containing perfusate for an equivalent time ( 1 hr at 1 . 5 ml/min; 37°C ) . For PLM blotting , we employed an N-terminal ( extracellular ) antibody ( see ‘Materials and methods’ ) to avoid effects of cytoplasmic ( C-terminal ) phosphorylation or palmitoylation on antibody binding . As shown in Figure 7B , anoxia/reoxygenation caused a 71% increase of palmitoylated PLM ( n = 5; p<0 . 01 ) and a 77% increase of palmitoylated flotillin-2 ( n = 5; p<0 . 01 ) relative to caveolin-3 . Consistent with the idea that DHHC5 constitutively affects palmitoylation of these proteins , Figure 7C shows that palmitoylated PLM and flotillin-2 were decreased by 27% and by 51% , respectively , in DHHC5-GT hearts that were perfused briefly to remove blood and immediately frozen . 10 . 7554/eLife . 01295 . 009Figure 7 . Palmitoylation is increased during reoxygenation-induced cardiac MEND . Palmitoylation of PLM and flottilin-2 during reoxygenation of anoxic cardiac tissue . ( A ) Experiments were performed in a pairwise fashion . One heart was quick-frozen after perfusion with oxygen for 45 min ( not treated , NT ) and one after being subjected to the anoxia/reoxygenation protocol , followed by biochemical analysis of palmitoylation . Left lanes show loading controls from initial lysates; right four lanes show samples after deacylation and precipitation with cysteine-reactive beads . No protein is detected without deacylation . Caveolin-3 was blotted on the same gel after stripping secondary antibodies . ( B ) Five paired data sets from experiments as in ‘A’ , with palmitoylation calculated relative to palmitoylated caveolin-3 densities . ( C ) Constitutive palmitoylation of PLM and flotillin-2 in DHHC5-GT hearts , not subjected to anoxia/reoxygenation , is decreased 27 and 51% , respectively . Palmitoylation is not clearly changed in hearts subjected to 30 min anoxia without reoxygenation ( ‘−O2’ ) . ( D ) Combined measurements of the surface membrane fraction and the palmitoylated fraction of PLM . Hearts #1 and #4 were not PEGylated; hearts #2 and #3 were PEGylated with PEGylated PLM fractions amounting to 40 and 46% , respectively , in the initial lysate . In the acyl-RAC pull-down of palmitoylated PLM , the surface membrane ( PEGylated ) fractions of PLM amount to only 17 and 23% in hearts #2 and #3 , respectively . Thus , surface membrane PLM is substantially less palmitoylated than internalized PLM . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 009 We next combined the acyl-RAC assay with the amine PEGylation assay , illustrated in Figure 2H , to determine the fraction of PLM at the cell surface and then additionally determine the fraction of PLM that is palmitoylated at the surface membrane vs internal membrane pools . Using the PEGylation protocol described in ‘Materials and methods’ , Figure 7D shows the PEGylation and acyl-RAC results for four hearts , two of which were PEGylated and two of which were not . The blots shown include a loading control and the acyl-RAC result for each heart with loading amounts adjusted to generate similar densities on the acyl-RAC blots . Densities of the loading controls from PEGylated hearts ( second lanes from the left and from the right ) indicate that 40 and 46% of PLM was PEGylated , respectively , and is therefore sarcolemmal . In the acyl-RAC pull-downs from the same hearts , however , only 17 and 23% of PLM is PEGylated . Therewith , a key tenet of our working hypothesis is verified . Although palmitoylation can target proteins to the surface membrane ( Greaves et al . , 2009 ) , enhanced palmitoylation may favor the removal of palmitoylated proteins from the surface membrane . As described in Figure 5B , MEND is decreased in intact cardiac tissue by treatments that protect the heart from reperfusion damage , namely CsA and adenosine ( Halestrap et al . , 2004 ) . Furthermore , we find that MEND is promoted by an agent that prevents ischemic preconditioning and promotes PTP openings , STS ( Ytrehus et al . , 1994 ) . These results therefore suggest that the occurrence of MEND might negatively impact the recovery of the heart from anoxia-reoxygenation episodes . To test if this is indeed the case , we analyzed contractile function of right ventricular strips paced at 0 . 25 Hz and subjected to 30 min of anoxia in the absence of glucose , followed by reoxygenation and administration of glucose ( 15 mM ) . As shown in the examples in Figure 8A and in the composite data in Figure 8B , contraction became negligible within 8 min of anoxia without glucose . During reoxygenation for 25 min , contractile function of WT ventricles recovered only little , 12% on average . However , contractile function of right ventricles from DHHC5-GT animals recovered substantially , 60% on average , equivalent to some of the most effective ‘preconditioning’ protocols ( Halestrap et al . , 2004 ) . Thus , as already suggested by others ( Belliard et al . , 2012 ) , internalization of sarcolemma may contribute to the acute failure of cardiac function during ischemia/reperfusion episodes . 10 . 7554/eLife . 01295 . 010Figure 8 . MEND correlates with impaired functional recovery of cardiac muscle from anoxia . Contractile function of mouse right ventricular strips , paced at 0 . 25 Hz , during anoxia for 30 min without glucose , followed by reoxygenation with glucose ( 15 mM ) for 25 min . ( A ) Representative force envelopes of right ventricular strips . ( B ) Composite results for right ventricular strips from WT and DHHC5-GT mice . For both sets , n = 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 010
From all molecular players in cardiac excitation-contraction coupling , small changes of Na/K pump activity cause the largest changes of myocyte Ca transients and therefore cardiac contractility ( Hilgemann , 2004 ) . In this light , Na/K pump activity changes caused by modifying DHHC5 and DHHC2 expression ( Figures 1 and 2 ) appear relevant to the physiological , long-term control of basal cardiac contractility . It is certain that palmitoylation of the regulatory pump subunit , PLM , can cause pump inhibition ( Tulloch et al . , 2011 ) . We have shown here , however , that the surface membrane fraction of PLM is increased by approximately the same extent as Na/K pump activity ( Figure 2 ) in DHHC5-deficient myocytes . Furthermore , myocyte surface area/volume ratios are significantly increased . Analysis of ion channel and excitation-contraction coupling function are now required to define more clearly the roles of DHHC5 and DHHC2 in myocytes . However , the present results strongly suggest that DHHC5 activity modulates Na/K pump densities in cardiac myocytes by modifying pump internalization rates . In this connection , the fate of membrane internalized by DHHC5-dependent endocytosis remains to be determined . One possibility is that the pathway delivers surface membrane to autophagosomes ( Moreau and Rubinsztein , 2012 ) . That MEND responses are reduced in PLM-deficient cardiac tissue ( Figure 5B ) correlates well with the fact that MEND is enhanced when PLM is overexpressed in a model cell ( Hilgemann et al . , 2013 ) . Therewith , a significant mechanistic involvement of Na/K pumps in the progression of MEND has been established . Clearly , proteomic analyses of palmitoylation by mass spectroscopy will be essential to evaluate the overall cellular scope of the MEND response . It is certain that mechanisms besides palmitoylation are sufficient to cause MEND in BHK cells . For example , extracellular lysolipids and sphingomyelinase C activities can cause MEND responses that are equivalent in magnitude to Ca-activated MEND ( Lariccia et al . , 2011 ) . In our experience , these different mechanisms act synergistically , and the generation of submicoscopic membrane phase transitions appears to be the convergent effect that precedes endocytosis ( Hilgemann and Fine , 2011 ) . In the present study as well , it is likely that some MEND responses reflect synergistic effects of multiple mechanisms . As shown in Figures 3 and 4 , the occurrence of MEND in four different protocols is strongly inhibited in DHHC5-deficient myoctytes and therefore depends on palmitoylation . MEND occurs over one to several minutes in response to PTP-promoting metabolites , over several minutes in response to transient H2O2-induced oxidative stress , over 30 min in response to a GPCR agonist , phenylephrine , which activates GPCRs and subsequently PKCs , and over a few seconds when Ca transients are activated . In contrast to the delayed MEND responses that occur in BHK cells ( Hilgemann et al . , 2013 ) , Ca transients cause immediate MEND responses in myocytes . These responses , which may represent myocyte wounding responses , are blunted in DHHC5-deficient myocytes when they are activated by moderate Ca transients . However , the explosive responses that occur when Ca transients are large ( Figure 4 ) are unaffected by acute inhibition of DHHC5 activity . Thus , the palmitoylation state of myocytes promotes Ca-activated MEND , but Ca can cause MEND by a mechanism unrelated to DHHC activity . In contrast , myocyte MEND responses to transient oxidative stress and to phenylephrine require an immediate presence of cytoplasmic acyl Co A and therefore are likely to involve palmitoylation as they develop . In the case of MEND induced by KSP solution , we cannot discount at this time that mitochondrial Ca release is playing some role . In cardiac muscle more than 95% of total CoA metabolites reside in the mitochondrial matrix ( Idell-Wenger et al . , 1978 ) , and the cytoplasmic free CoA concentration is about 7 μM ( Idell-Wenger et al . , 1978 ) . Given that mitochondria make up about 30% of myocyte volume ( Zhou et al . , 2011 ) , the physical basis for our working hypothesis is well established in myocytes . Figures 5 and 6 document by independent methods that sarcolemma in intact hearts is significantly internalized upon reoxygenation after an anoxic episode . Internalization is dependent on fatty acids and DHHC5 activity . It is prevented by agents that precondition hearts against reoxygenation damage , namely adenosine and CsA , and it is promoted by low concentrations of staurosporine that promote reoxygenation damage ( Schneider , 2005 ) . The results together suggest that MEND responses occurring during reoxygenation may be relevant to human ischemia/reperfusion events . As expected if MEND contributes significantly to myocyte damage during reoxygenation , contractile function of ventricular muscle is preserved by 60% in DHHC5-GT right ventricles after 30 min anoxia without glucose and 25 min reoxygenation with glucose ( Figure 8 ) . This degree of functional preservation is as good as standard ‘preconditioning’ protocols ( Cave et al . , 1994; Schneider et al . , 2001 ) . Whether the occurrence of MEND is in fact detrimental to functional recovery is uncertain . Loss of substantial amounts of sarcolemma may certainly impact cardiac excitation-contraction coupling . From a different perspective , MEND itself may not be the culprit . MEND appears to rely on the development of submicroscopic membrane phase transitions . Independent of endocytosis , therefore , the occurrence of membrane phase transitions may promote membrane defects and membrane shedding that mediates release of cytoplasmic enzymes and metabolites to the extracellular space . The important conclusion for reperfusion injury is that effective inhibition of palmitoylation may significantly inhibit this part of reperfusion damage , and these same considerations apply to the function of vascular cells and fibroblasts during reperfusion . In conclusion , this article provides evidence that a signaling pathway leading from mitochondrial PTP openings to DHHC5 acyl transferases and endocytosis ( Hilgemann et al . , 2013 ) is activated during the reoxygenation of anoxic cardiac tissue . Multiple lines of evidence support that idea that endocytosis during reoxygentation requires increased membrane protein palmitoylation , and inhibition of the pathway appears to protect muscle from reoxygenation injury . The phenotypes of myocytes that are deficient in DHHC5 acyl activity suggest that the pathway contributes significantly to constitutive sarcolemma turnover and will mediate enhanced sarcolemma turnover in response to transient metabolic stress .
Patch clamp ( Yaradanakul et al . , 2008 ) and myocyte preparation were as described ( Lariccia et al . , 2011 ) . Software employed is available on-line ( https://sites . google . com/site/capmeter/ ) . The UT Southwestern Medical Center Animal Care and Use Committee approved all animal studies . Effects of voltage pulses ( 0 . 1–0 . 5 kHz ) to determine Cm are removed digitally from current records . For myocyte patch clamp and pipette perfusion experiments , highly polished pipette tips with diameters of >6 μm were employed . iCell myocytes ( Ma et al . , 2011 ) were obtained from CellularDynamics ( Madison , WI ) and cultured according to CellularDynamics instructions . NIC recording was performed using a conventional patch clamp circuit and phase-lock amplifier . Right ventricular strips were dissected at 10°C in a relaxing solution that contained 20 mM MgCl2 to prevent electrical activity and contraction . Muscles were mounted horizontally in a chamber maintained at 37°C and superfused with physiological salt solution , described below , at a velocity of 1–2 cm/s . The capacitive component of NIC signals increased with increasing oscillation frequency up to a maximum at ∼23 kHz . The optimal frequency for recording was therefore about 15 kHz . The phase angle was adjusted by up to 2° to insure that the capacitive signal was zero upon touching the probe to a soft rubber septum in the recording chamber . As described in Figure 9 , capacitive signals in NIC recording can be strongly decreased by extracting lipids from right ventricular strips and are very insensitive to sarcolemmal conductance changes of very large magnitudes . One major experimental requirement for reliable NIC recording is that the probe remains mechanically stable with respect to the tissue and that conductance of the tissue ( i . e . , of the extracellular space ) does not change . Conductance signal changes were negligible during NIC recordings presented in Figure 5 . 10 . 7554/eLife . 01295 . 011Figure 9 . NIC recording monitors sarcolemmal area and is insensitive to sarcolemmal conductance changes . ( A ) As described previously ( Lariccia et al . , 2011 ) , ( 2-hydoxypropyl ) -β-cyclodextrin ( HPCD ) extracts cholesterol from fibroblasts without causing Cm changes . However , methyl-β-cylodextrin ( BMCD ) causes large decreases of Cm that presumably reflect extraction of both phospholipids and cholesterol . Similarly , NIC signals from right ventricular strips are unaffected by application of HPCD ( 8 mM ) but are strongly decreased by application of BMCD ( 8 mM ) . ( B ) Muscle incubated with ATP and 0 Ca with 1 mM EGTA to avoid contraction when the sarcolemma develops ruptures . A rather high concentration of sodium dodecylsulfate ( SDS , 2 . 5 mM ) does not significantly decrease NIC signals . Although this concentration of SDS generates sarcolemmal leaks with great certainty , it does not effectively extract the sarcolemma from the intact muscle preparation . By contrast , Triton X100 at 3% very effectively decreases NIC signals from the same muscle , as expected for an effective membrane extraction . These results verify that NIC signals are not affected by increases of sarcolemmal conductance until cellular conductance becomes significant with respect to that of the extracellular space . DOI: http://dx . doi . org/10 . 7554/eLife . 01295 . 011 Standard MEND Solutions minimize all currents other than NCX1 current . Extracellular solution contained in mM: 120 n-methyl-d-glucamine ( NMG ) , 4 MgCl2 ± 2 CaCl2 , 0 . 5 EGTA , 20 TEA-OH , 10 HEPES , pH 7 . 0 with aspartate . Cytoplasmic solution contained in mM: 75 NMG , 20 TEA-OH , 15 HEPES , 40 NaOH , 0 . 5 MgCl2 , 0 . 8 EGTA , 0 . 25 CaCl2 , 1 Pi set to pH 7 . 0 with aspartate . Unless stated otherwise , 8 mM MgATP , 2 mM TrisATP , and 0 . 2 mM GTP were employed in cytoplasmic solutions with a free Mg of 0 . 5 mM . KSP cytoplasmic solution contained 110 KOH , 40 NaOH , 10 histidine , 1 . 0 EGTA , 0 . 2 CaCl2 , nucleotides as just given , and pH 7 . 0 with aspartate . Bath solution in NIC recording and FITC-dextan uptake experiments contained in mM: 120 NaCl , 5 KCl , 0 . 5 NaHPO4 , 0 . 5 MgCl2 , 1 . 5 CaCl2 , 15 histidine , and 15 glucose . During anoxia , 5 mM deoxyglucose was substituted for glucose , 35 mM KCl was added to stop spontaneous activity , 50 μM FA-free albumin with 50 μM myristate was added , and solution was degassed by stirring under vacuum . Unless specified otherwise , reagents were from Sigma-Aldrich ( St . Louis , MO ) . N-terminal ( ‘N3’ ) PLM antibody , a gift of Dr Will Fuller ( Dundee ) , was raised in rabbit against the peptide , EAPQEPDPFTYDYHT , coupled to KLH by Moravian Biotechnology ( Brno , Czech Republic ) and affinity purified against the same peptide coupled to NHS-Sepharose . Unless stated otherwise , error bars represent standard error of six and usually eight or more observations . Significance was accessed by Students t test or , in rare cases of unequal variance , by the Mann-Whitney Rank Sum test . In figures , ‘*’ denotes p<0 . 05 , ‘**’ denotes p<0 . 01 , and ‘***’ denotes p<0 . 001 . After completion of experimental protocols , hearts were perfused for 10 min with 10 mM NEM , flash-frozen in liquid nitrogen-cooled aluminum clamps , and stored at −80°C . Tissue was powdered and vortexed into lysis buffer ( 50 mM NaCl , 50 mM Hepes , 0 . 1% SDS , 1% NP-40 , pH 7 . 4 ) with protease inhibitor cocktail ( Roche ) , placed on ice for 15 min , and centrifuged at 16 , 000 × g for 10 min . Supernatants were then collected and protein determined ( Micro BCA kit , Pierce ) . Acyl-RAC assays were performed as described ( Forrester et al . , 2010 ) with modifications as follows . Duplicate sets of protein lysate were diluted to 2 mg/ml in 250 μl blocking buffer ( 100 mM Hepes , 1 mM EDTA , 2 . 5% SDS , 0 . 2% methyl methanethiosulfonate [MMTS] , pH 7 . 5 ) and incubated at 40°C for 20 min with constant shaking . MMTS was removed by acetone precipitation at −30°C . Pellets were extensively washed with 70% ice-cold acetone and resuspended in 240 μl binding buffer ( 100 mM Hepes , 1 mM EDTA , 1% SDS , pH 7 . 5 ) . The duplicate samples were separated , one was treated with 100 μl of thiopropyl Sepharose beads ( GE Life Sciences ) and 250 mM hydroxylamine , ( HA ) pH 7 . 5 . The other was treated with 250 mM NaCl as negative control . Approximately 40 μl of each sample was retained as ‘total input’ . Samples were rotated at 23°C for 2 . 5 hr , beads were pelleted , and supernatants were retained as ‘unbound’ sample . Beads were extensively washed in binding buffer and acylated protein was eluted at 23°C in 50 μl binding buffer with 100 mM DTT for 20 min . Supernatant was removed , mixed with Laemmli loading buffer , heated at 37°C for 30 min , and separated on SDS-PAGE . Hearts were perfused for 2 min to remove blood , followed by perfusion with 10 mM PEG succinimidyl ester ( MW 5000; Nanocs , PG1-SC-5k ) dissolved in 130 mM NaHCO3 with 4 mM MgCl2 at pH 8 . 7 for 15 min at 22°C and further incubation of hearts in the same solution at 5°C without perfusion . Thereafter , NHS was quenched by renewed perfusion of the bicarbonate perfusate diluted 1:3 with isotonic ( 300 mM ) glycine solution for 15 min . Hearts are then frozen , powdered , and homogenized into RIPA . Western blotting subsequently resolves the PEGylated protein fraction , that is the putative cell surface fraction , as the fraction of protein shifted to a higher molecular weight . To test whether the NHS reagent significantly crossed the surface membrane ( i . e . , whether the sarcolemma became significantly permeable during the procedure ) , we routinely blotted actin , and PEGylation of actin was not detectable in the hearts employed in the results presented in Figures 2 and 7 . PLM-deficient mice were bred from knockout mice provided by Amy L Tucker ( Jia et al . , 2005 ) ( U Virginia , Charlottesville ) . DHHC5-GT and DHHC-WT cardiac myocytes were isolated from littermates of heterozygous crosses at F2 or littermates of F2 × F2 crosses of homozygous WT or DHHC5-GT mice ( Li et al . , 2011 ) . Confocal imaging of hearts was with a TE2000-U Nikon microscope ( 60 × oil immersion , 1 . 45-NA objective; RC-26 recording chamber; Warner Instruments; 40-mW 163-CO2 laser , S Newport Corporation ) at 488 nm using 3% power . Full-frame live recording resolution was 512 × 512 with exposure <1 s ( pinhole , 150 µm ) . Higher resolution captures were 1024 × 1024 with exposures <4 s . Bleaching of fluorophores was negligible . The aorta of isolated hearts was cannulated and retrograde perfusion was implemented with standard oxygenated NIC bath solution at 37°C at a rate of 1 . 5 ml/min with 50 µM albumin in FA-free experiments or 50 µM albumin plus 100 µM myristate for experiments containing FA . Experiments with 0 . 1 µM STS or 5 µM CsA contained FA , albumin , and drug throughout . After 20 min control perfusion , anoxic solution was perfused 30 min , and then oxygenated solution was perfused with 100 mg/5 ml of 4000 MW FITC-dextran ( Sigma ) with flow reduced to 0 . 5 ml/min . After 25 min , dye-free oxygenated solution was perfused at 1 . 5 ml/min for 5 min . Then , hearts were transferred to the microscope and perfused at 23°C . During imaging the heart was repositioned several times to gather images from a wide distribution of myocytes . Averages of at least 16 individual myocytes were calculated per frame and at least 10 regions were observed per heart , yielding >160 observations . Fluorescent punctae were counted within each myocyte , and the occurrence of 20 punctae per cell was taken as threshold for the occurrence of MEND . | Many people who survive a stroke or heart attack experience substantial tissue damage when the blood supply is restored . Much of this damage can be caused by the mitochondria inside the cells releasing a protein called cytochrome c that can cause cells to die in a process called apoptosis . The cytochrome c is released as the outer membrane of the mitochondria becomes permeable and pores called permeability transition pores open up in the inner membrane . Now Lin et al . explore if additional molecules released from the mitochondria might also initiate important cellular responses during the reoxygenation of oxygen-deprived tissue . Lin and co-workers recently showed that the mitochondria of some cells can release a small enzyme cofactor , coenzyme A , which then promotes a cellular response called massive endocytosis . This process can cause up to 70% of the cell surface membrane to be absorbed into the interior of the cell in the form of membrane vesicles . Most forms of endocytosis involve a much smaller fraction of the cell membrane and employ a set of well-known endocytic proteins that are not involved in massive endocytosis . Now , Lin et al . investigate the role of massive endocytosis in cardiac muscle . Electrical and optical measurements reveal that massive endocytosis occurs as cardiac cells that have been deprived of oxygen are reoxygenated . Lin et al . also find that an enzyme called DHHC5 must be present to allow endocytosis to take place during reoxygenation . DHHC5 is an enzyme that catalyzes a process called acylation – the transfer of acyl groups to proteins at the cell surface . Moreover , the deletion of DHHC5 has a beneficial impact on the performance of cardiac muscle after oxygen deprivation , which implies that molecules that inhibit protein acylation might protect the heart from damage during reoxygenation . Together , these results establish new pathological and physiological roles for the acylation , which is one of the most common biochemical modifications made to membrane proteins after they are synthesized . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | Massive palmitoylation-dependent endocytosis during reoxygenation of anoxic cardiac muscle |
The hippocampus is linked with both sleep and memory , but there is debate about whether a salient aspect of sleep – dreaming – requires its input . To address this question , we investigated if human patients with focal bilateral hippocampal damage and amnesia engaged in dreaming . We employed a provoked awakening protocol where participants were woken up at various points throughout the night , including during non-rapid eye movement and rapid eye movement sleep , to report their thoughts in that moment . Despite being roused a similar number of times , dream frequency was reduced in the patients compared to control participants , and the few dreams they reported were less episodic-like in nature and lacked content . These results suggest that hippocampal integrity may be necessary for typical dreaming to occur , and aligns dreaming with other hippocampal-dependent processes such as episodic memory that are central to supporting our mental life .
Dreaming has intrigued humans for thousands of years , being variously interpreted as having premonitory , religious , psychoanalytic , or mnemonic significance . Defined as an internally-generated subjective mental experience during the sleep state ( Cipolli et al . , 2017 ) , dreaming can be present at initial sleep onset ( hypnagogic sleep; Horikawa et al . , 2013; Stickgold et al . , 2000 ) , during non-rapid eye movement ( NREM ) sleep ( Antrobus et al . , 1995; Foulkes , 1962; Nielsen , 2000; Siclari et al . , 2017; Wamsley , 2013; Wamsley et al . , 2007 ) , and rapid eye movement ( REM ) sleep ( Hobson et al . , 2000; Maquet et al . , 2000; Maquet et al . , 2005 ) . Although dreams are not a precise replay of our memories ( Fosse et al . , 2003; Stickgold et al . , 2001a ) , it has been proposed that dreaming may be associated with memory consolidation processes ( Payne , 2010; Wamsley , 2014; Wamsley et al . , 2010; Wamsley and Stickgold , 2019 ) . Indeed , in rodents and humans , patterns of brain activity exhibited during a learning experience were found to be subsequently expressed during sleep ( Peigneux et al . , 2004; Wilson and McNaughton , 1994 ) . Bilateral damage to a brain structure called the hippocampus is known to adversely affect memory processing ( Miller et al . , 2017; Miller et al . , 2020; Scoville and Milner , 1957; Spiers et al . , 2001 ) and daydreaming ( McCormick et al . , 2018 ) . While sleep dreaming has been examined previously in patients with hippocampal lesions , results are mixed , with some studies reporting repetitious and stereotyped dreams ( Torda , 1969a; Torda , 1969b; Stickgold et al . , 2000 ) , whereas others claim that hippocampal damage has no effect on dreaming ( Solms , 1997; Solms , 2013 ) . Possible reasons for these disparate findings include differences in the sleep stages sampled , the inclusion in several studies of patients with psychiatric conditions and below-average IQ and , in other instances , patients were interviewed at a point temporally remote from any dreaming that might have occurred , presenting a challenge for these patients who usually suffer from amnesia . It remains uncertain , therefore , whether hippocampal integrity is necessary for dreaming to occur . If , as we predicted , bilateral hippocampal damage degrades dreaming , this would reinforce the link between dreaming and hippocampal-dependent processes such as memory , potentially moving us closer to an understanding of why we dream . By contrast , if hippocampal-damaged patients who have a reduced ability to experience complex , imagery-rich , spatio-temporal mental events during wake are nevertheless able to have such experiences during sleep , this would need to be explicitly accounted for within theories of hippocampal function . Here , we sought to mitigate the issues affecting previous studies while investigating the frequency , features and content of dreaming in rare patients with selective bilateral hippocampal damage and matched control participants . To do this we employed a provoked awakening protocol ( Nguyen et al . , 2013; Siclari et al . , 2013; Wamsley et al . , 2016 ) . This widely-used approach involved waking participants up at various times during their night’s sleep to report their thoughts in that moment . In this way we examined sleep mentation in a direct and immediate manner .
We assessed four patients ( all right-handed males; mean age 58 . 25 years , SD ±20 . 82 ) with selective bilateral hippocampal lesions and a specific episodic memory deficit ( Table 1 , Material and methods , Supplementary file 1 , Spanò et al . , 2020 ) . Patients were matched to ten healthy control participants ( all right-handed males; mean age 59 . 2 years , SD ±15 . 89 ) based on a number of demographic factors including age , gender , body mass index , non-verbal IQ , and a range of sleep quality measures ( Table 1; Material and methods , Supplementary file 1 , Supplementary file 2 ) . We conducted in-home sleep recording using portable polysomnography ( PSG ) on two consecutive nights – one habituation night to allow for familiarization with the PSG equipment and one experimental night for the collection of dream reports . The purpose of the PSG recording was to ensure that we awakened participants during both NREM and REM sleep , and in a similar manner for the patient and control groups . During the dream sampling night , participants were woken up after a period of 3 min from the onset of either NREM or REM sleep at various times throughout the night ( Materials and methods , Figure 1A ) . Table 2 shows the group summary data and the results of the between-group statistical analyses for each of the measures that are described below . While , for the sake of economy , we present the findings in terms of these group comparisons , given the small sample of these rare patients , caution should be exercised in interpreting the results . We , therefore , include the individual patient data in Tables 1 and 2 permitting the patients to be considered also as a series of case studies . The number of awakenings was not different between patients and control participants ( Figure 1B ) . Furthermore , there were no significant group differences in the proportion of awakenings from NREM and REM sleep ( Figure 1C ) . After an awakening , participants were instructed via a two-way intercom to describe everything that was going through their mind before they were woken up in that moment . They were occasionally probed ( e . g . Can you tell me more about that ? ) to obtain further information ( Materials and methods ) . The amount of probing did not differ between groups . Dream reports were subsequently transcribed for further analyses ( Materials and methods; Figure 2 ) . Although the two groups were woken up a similar number of times , they differed in terms of dream frequency , with patients reporting significantly fewer dreams compared to control participants ( Figure 1D ) . There were no group differences in terms of the proportion of dream recall during NREM and REM sleep . Of note , one patient did not report any dreams at all , but instead typically stated ‘I can’t picture it’ ( Figure 2 ) . Perhaps the patients reported fewer dreams than control participants simply because they forgot any dreams they may have had . Three different types of awakening were evident: when participants reported a dream , when they stated they did not dream at all ( no dream ) , and when they dreamt but could not recall the content ( this is known as a blank dream ) . Although there was a significant difference between patients and controls for dreams ( see above ) and for no dreams , they did not differ in terms of the proportion of blank dreams , which was low for both groups . This suggests that patients could distinguish between situations when they did not dream and those when they dreamt but could not remember . However , no validated objective measure of dreaming exists , and this should be borne in mind when interpreting participants' subjective reports . It is notable that in previous studies , these particular patients could retain information over several minutes , including reporting on their daydreaming ( e . g . McCormick et al . , 2016; McCormick et al . , 2018 ) , which speaks against a rapid decay of sleep mentation as an explanation for their reduced dream frequency . The patients had so few dream reports that comparisons with the control participants should be treated with caution . Nevertheless , we wondered whether any differences in features and content were evident between the groups for the few dreams the patients reported . Because one patient had no dreams whatsoever , he was not included in these analyses . We first examined the number of informative words ( see Materials and methods; Stickgold et al . , 2001b ) used in the dream narratives and , while the patients used fewer such words , overall there was no significant difference between the groups . Nevertheless , adjudicating between the possibility of patients having a generic problem with expressing themselves verbally versus merely having little to describe because the few dreams they had were so impoverished , is challenging . The same issue pertains for assessments during wake . A number of studies addressed this concern during tasks involving the imagination of scenes or future scenarios , counterfactual thinking , describing the present and pictures of scenes . Bilateral hippocampal damage does not affect narrative construction or verbal descriptive ability ( e . g . Race et al . , 2011; Race et al . , 2013; Mullally et al . , 2012; Mullally and Maguire , 2014; Miller et al . , 2020 ) . Considering specifically the patients in the current study , they too had no difficulty performing verbal fluency tests or other tasks where total word count was measured ( Supplementary file 1 ) . Given these general and specific findings , it is unlikely that the patients’ performance was driven by an underlying expressive verbal problem . Next we assessed overall qualitative features of dream reports using experimenter ratings ( Materials and methods; De Gennaro et al . , 2011; Oudiette et al . , 2012 ) . There were no significant group differences in terms of general complexity , vividness , bizarreness , emotional valence ( there were no nightmares ) , and proportion of self-references . In order to probe the dream reports in more depth , we used two scoring methods that are often employed for examining complex mental events ( Materials and methods ) . The first involved a scoring regime typically used for autobiographical memories , the Autobiographical Interview ( Levine et al . , 2002 ) . This allowed us to measure the amount of episodic ( internal ) and non-episodic ( external ) details . The patients included significantly fewer internal details in their dream narratives relative to controls , whereas there was no significant group difference in terms of external details . A second method , usually employed for scoring the content of imagined scenes ( Hassabis et al . , 2007 ) , showed that the dream reports of patients were significantly less rich in content compared to those of the control participants ( Figure 1E ) .
By studying rare patients with selective bilateral hippocampal damage we found that dream frequency was reduced compared to control participants , and the few dreams they had were less episodic-like in nature and lacked content . This accords with previous studies that reported stereotyped dreaming in patients with brain damage that extended beyond the hippocampi ( Torda , 1969a; Torda , 1969b; Stickgold et al . , 2000 ) , and echoes the effects of hippocampal damage on imagination ( Hassabis et al . , 2007 ) , episodic memory ( Miller et al . , 2017; Miller et al . , 2020; Spiers et al . , 2001 ) and daydreaming ( McCormick et al . , 2018 ) . Given our patients’ circumscribed lesions , these results suggest that hippocampal integrity may be necessary for typical dreaming to occur . There are several possible explanations for degraded dreaming in the patients , which we consider in turn . Despite the tendency to confine dreaming to the sleep state , studies have shown that there is often continuity between waking life experiences and dreaming ( Andrillon et al . , 2015; Fosse et al . , 2003; Horikawa et al . , 2013; Schredl and Hofmann , 2003 ) . Given that the patients had difficulty retaining information over longer time scales while awake , then perhaps during subsequent sleep there was little material to process . Hence it could be that the capacity to dream was intact , but underused , in the patients . Alternatively , the core capacity for dreaming might have been compromised , with this in turn affecting memory processing . Dreaming has been linked with the consolidation of information into long-term memory ( Payne , 2010; Wamsley , 2013; Wamsley , 2014; Wamsley et al . , 2010; Wamsley and Stickgold , 2019 ) , but it is unknown whether dreaming plays a functional role in this process ( Wamsley , 2014 ) . If it does , then patients with hippocampal damage may lack this dream-related mechanism for facilitating memory processing , and this could contribute to their amnesia . Another aspect of sleep that has been associated with episodic memory consolidation is slow-wave sleep ( SWS ) , a stage within NREM sleep ( Rasch and Born , 2013 ) . We have previously shown that the hippocampal-damaged patients tested here had significantly reduced SWS and slow wave activity , whereas the time spent in other sleep stages was comparable to that of control participants ( Spanò et al . , 2020 ) . It is unlikely that the patients’ degraded dreaming was caused solely by decreased SWS . This is because we sampled dreaming across both NREM and REM sleep , and of the dreams sampled during NREM sleep in the control participants , just 5% ( SD 11 . 3 ) were during SWS . Overall , however , reduced slow wave activity along with sub-optimal dreaming – if dreaming plays a functional role – may constitute a twofold blow that adversely affected the proper functioning of episodic memory in these patients . A different perspective on the current results involves looking beyond memory . The hippocampus has been implicated in a range of other cognitive functions including thinking about the future ( Addis et al . , 2007; Hassabis et al . , 2007; Kurczek et al . , 2015 ) , spatial navigation ( Maguire et al . , 2006; O'Keefe and Nadel , 1978 ) , daydreaming ( Karapanagiotidis et al . , 2017; McCormick et al . , 2018 ) , aspects of decision making ( Mullally and Maguire , 2014; McCormick et al . , 2016 ) , and visual perception ( Lee et al . , 2005; McCormick et al . , 2017; Mullally et al . , 2012 ) . Patients with bilateral hippocampal damage are also impaired at mentally constructing scene imagery ( Hassabis et al . , 2007 ) , and scene imagery features prominently across cognition ( Clark et al . , 2019; Clark et al . , 2020 ) . Consequently , it has been proposed that one role of the hippocampus could be to facilitate the scene construction process ( Hassabis and Maguire , 2007; Maguire and Mullally , 2013 ) . This may explain why dreaming is degraded in the context of hippocampal pathology , as it too typically involves ( and possibly requires ) scene imagery . In summary , while the functional role of dreaming is as yet unknown , we conclude from our results that hippocampal integrity may be a prerequisite for typical dreaming to occur , and that dreaming seems to align with an array of other cognitive functions that are hippocampal-dependent and which play crucial roles in supporting our everyday mental life .
For all patients , hippocampal lesions resulted from leucine-rich glycine-inactivate-1 antibody-complex limbic encephalitis ( LGI1-antibody-complex LE; Miller et al . , 2017; Miller et al . , 2020 ) . This study was conducted a median of 9 . 5 years after hippocampal damage occurred ( mean 9 years ± SD 2 . 45 ) . Patients ( HPC ) and the dream control participants ( CTL ) were closely matched on a number of demographic factors: gender ( all males ) , age ( Table 1; MWU = 19 . 00 , p=0 . 89 , Cohen’s d = 0 . 08 ) , body mass index ( HPC mean 27 . 68 ± 2 . 51; CTL 25 . 79 ± 2 . 41; MWU = 14 . 00 , p=0 . 40 , Cohen’s d = 0 . 47 ) and general cognitive ability assessed with the Matrix Reasoning subtest of the Wechsler Abbreviated Scale of Intelligence ( WASI; Wechsler , 1999; Table 1; MWU = 7 . 00 , p=0 . 06 , Cohen’s d = 1 . 13 ) . Participants had no history of psychiatric disorders ( e . g . , depression , anxiety ) . Each participant gave written informed consent for participation in the study , for data analysis and for publication of the study results . ‘Materials and methods’ were approved by the University College London Research Ethics Committee . The patients entered the sleep study having already been characterized , relative to matched healthy control participants , in terms of their lesion selectivity and neuropsychological profile as part of previous research studies . Full details of that characterization process are available in McCormick et al . , 2016 , McCormick et al . , 2017 , McCormick et al . , 2018 and Spanò et al . , 2020 . In summary , manual ( blinded ) segmentation of the hippocampi from T2-weighted high resolution structural MRI scans ( 0 . 5 × 0 . 5 × 0 . 5 mm voxels ) showed that our patients had substantial volume loss relative to controls in the left ( MWU = 2 . 00 , p=0 . 009 , Cohen’s d = 1 . 83 ) and right ( MWU = 3 . 00 , p=0 . 013 , Cohen’s d = 1 . 67 ) hippocampus ( Table 1 ) . Expert neuroradiological examination confirmed there was no damage outside of the hippocampi . In addition , automated whole brain voxel-based morphometry showed there were no volume differences between patients and controls anywhere else in the brain . Supplementary file 1 provides the neuropsychological profile ( summary data and statistical analyses ) of the patients across a range of cognitive tests , and indicates the selective nature of their memory loss . Since all patients included in the current study had suffered from LGI1-antibody-complex LE , our findings might potentially not generalize to other forms of hippocampal amnesia . However , it is important to note that other aetiologies that lead to hippocampal-mediated amnesia such as viral encephalitis , hypoxic brain injury secondary to drug overdose , or toxic shock syndrome are associated with circumscribed hippocampal lesions , but frequently also involve anatomical damage elsewhere ( Heinz and Rollnik , 2015; Raschilas et al . , 2002 ) . In addition , these aetiologies lead to co-morbidities and broader cognitive impairment ( Heinz and Rollnik , 2015; Hokkanen and Launes , 2007; McGrath et al . , 1997; Peskine et al . , 2010; Thakur et al . , 2013; Rosene et al . , 1982 ) , which were absent from the clinical and neuropsychological profile of the patients reported here . Therefore , the selection of such a rare group of patients with circumscribed hippocampal lesions allowed us to pinpoint the direct role of the hippocampus in dreaming without the interference of potential confounds associated with heterogeneity in aetiology . Other features associated with LGI1-antibody-complex LE in its initial presentation – such as focal seizures and hyponatraemia related to hypothalamic damage – are also unlikely to explain the effects we observed . Our patients were seizure-free when they were discharged after initial admission , they were not prescribed antiepileptic medication , and none of the patients had seizure recurrence following initial treatment . Thus , unlike in temporal lobe epilepsy , which is associated with ongoing seizures and hippocampal sclerosis ( Kapur and Prevett , 2003 ) , our patients enabled us to study effects on dreaming that were not coincidental with , and sequelae of , seizure activity . Moreover , patients were not undergoing treatment for hyponatraemia , which is consistent with published evidence that persistent hyponatraemia is not a characteristic feature of LGI1-antibody-complex LE ( Bastiaansen et al . , 2017 ) . Therefore , the findings in the current study are unlikely to stem from the above-mentioned potential issues . As dreaming might be influenced by sleep quality ( Schredl , 2009 ) , we confirmed that patients did not differ significantly from control participants on subjective measures of the general quality and pattern of sleep , as well as on objective measures of sleep-related breathing disorders , and sleep-wake patterns across one week ( see Supplementary file 2 ) . Specifically , participants completed standardized questionnaires assessing habitual sleep habits over the last month ( The Pittsburgh Sleep Quality Index; Buysse et al . , 1989 ) , level of daytime sleepiness ( The Epworth Sleepiness Scale; Johns , 1991 ) , and chronotype – whether someone is a ‘morning’ or an ‘evening’ type of person ( The Morningness-Eveningness Questionnaire; Horne and Ostberg , 1976 ) . Moreover , we assessed the severity of obstructive sleep apnoea with the WatchPAT-200 ( Itamar Medical Ltd . , Caesarea , Israel ) , a wrist-worn device that measures the Peripheral Arterial Tone ( PAT ) signal by means of a plethysmographic based finger-mounted probe . Signals were automatically analysed with the zzzPAT software ( version 4 . 4 . 64 . p , Itamar Medical Ltd . , Caesarea , Israel ) to identify respiratory events and sleep states . The outcome measure employed in this study was the PAT apnoea-hypopnea index ( AHI ) , which provides the number of apnoea and hypopnea events per hour during the night . In order to assess sleep-wake patterns , participants wore an Actiwatch 2 ( Phillips Respironics Mini-Mitter ) for seven consecutive days and nights on their non-dominant wrist . Light and activity data were collected in 30 s epochs and analyzed using the Philips Actiware 6 . 0 . 2 software package ( Respironics Actiware 6 . 0 . 2 ) . Data were scored based on available guidelines ( Chow et al . , 2016 ) , with a medium sensitivity ( 40 activity cpm ) , with sleep onset occurring after an immobility period of 10 min , and rise time following an increase in activity level and in light level above 1 . 0 μW/cm2 . Variables of interest were sleep efficiency ( in percent ) , total sleep time ( in minutes ) , sleep fragmentation index ( an index of restlessness ) , night-to-night variability for sleep duration ( Lemola et al . , 2013 ) , average bedtime and mean sleep midpoint ( clock time halfway between bedtime and rise time ) . Participants slept in the same room on both nights . Two sleep researchers were located in a separate , nearby , room one of whom performed sleep staging in real time during online visualization of noise-reduced EEG recordings . An independent , registered sleep technologist , blind to participant group membership and the study aims , later off-line scored the PSG recordings to verify the sleep staging , in line with the revised American Academy of Sleep Medicine manual ( Berry et al . , 2015 ) . For all participants , awakenings occurred after sleep onset , throughout the night . We aimed to assess dreams from both NREM and REM sleep , and we therefore staggered awakenings at intervals that allowed for entry into these stages , or approximately between 30–90 min intervals . For example , a participant may have been woken at 30 min , returned to sleep and was woken again at 90 min . A maximum of 10 awakenings were scheduled per night . Awakenings were not collected at pre-ordained points during the night ( for example , after the third REM period for all participants ) , rather they were based on participants’ specific sleep architecture in order to maximize the number of reports collected . Every participant had a different sleep onset and duration , and so the awakenings were not scheduled at precisely the same clock time across participants . However , as shown in Supplementary file 2 , the sleep quality of the patients and the controls was well-matched , and this included total sleep time , bedtime and midpoint of the night . Once a decision to awaken a participant was made , after a 3 min period without stage shift of either NREM or REM sleep , the other sleep researcher played a non-stressful 500 Hz neutral tone via a two-way , Bluetooth intercom system equipped with a camera for continuous visual monitoring . After the tone was played , if the participant did not wake up , his name was spoken . This two-step awakening procedure was repeated up to five times , if required ( Dumel et al . , 2015 ) . After awakening , participants were instructed by intercom to tell the experimenter everything that was going through their mind before they were woken up . What they said was recorded and subsequently transcribed . Participants were occasionally probed ( e . g . Can you tell me more about that ? ) to obtain further information . Probing followed a structured protocol . This involved first asking participants to freely describe what was in their minds immediately after awakening . Whenever a participant’s response was not clear or only covered parts of the dream , the experimenter asked general follow-up questions , which could echo information already provided ( e . g . ‘…It was a conversation happening in the locker room’ . Can you give us any more specifics about the conversations or anything else that you recall ? ) . Crucially , this probing never involved leading the participant , as can be observed in the examples provided in Figure 2 . This approach is very similar to that of well-established tasks that assess autobiographical memory recall ( Levine et al . , 2002 ) and scene imagination ability ( Hassabis et al . , 2007 ) during wake , where probing in this manner is widely accepted . Transcriptions of dream reports were analysed by a researcher who was blind to participant group membership . Double scoring was performed on 20% of the data by a second researcher . We assessed across-experimenter agreement with inter-class correlation coefficients , with a two-way random effects model looking for absolute agreement , which indicated excellent agreement between the experimenters' scoring ( range: 0 . 9–1 . 0 ) . A dream was defined as any report that included at least one person , one place or one event ( Foulkes and Rechtschaffen , 1964 ) . Dream frequency was calculated as the total number of dreams divided by the total number of awakenings . Word count included words that provided information about the dream ( i . e . , informative words ) , and excluded repetitions , hesitations and fillers , secondary elaborations and metacognitive statements ( Stickgold et al . , 2001b ) . Dream complexity was experimenter scored on a 6-point scale using an adaptation of the Orlinsky score ( Oudiette et al . , 2012 ) ; excluding the no dream and blank dream options , which we recorded separately ) . This ranged from a participant remembering a specific topic , but in isolation , for example a fragmentary action , scene , object , word , or idea unrelated to anything else , to a participant remembering an extremely long and detailed dream sequence of five or more stages . Vividness referred to the clarity and detail of a dream , and was experimenter scored using the 6-point scale of De Gennaro et al . , 2011 , ranging from no image at all , to perfectly clear and as vivid as normal vision . Bizarreness/implausibility of dreams was experimenter scored using the De Gennaro et al . , 2011 6-point scale . This was based on the presence of bizarre elements ( impossible characters or actions ) and/or improbable plot ( discontinuity or unusual settings ) . The emotional valence of a dream was scored using a 5-point Likert scale ranging from very negative to very positive , with three indicating a neutral tone . For self-references , one point was awarded per dream report if a participant reported he was the agent of an action , thought or feeling ( e . g . ‘I was driving my car down a nearby road with a friend . . . ” ) . We used two other scoring methods that are often employed for examining complex mental events in order to probe the dream reports further . The first was the Autobiographical Interview which identifies internal and external details , a distinction that can be conceptualized as the difference between episodic and non-episodic/semantic information , respectively ( Levine et al . , 2002 ) . Internal details refer to the main event described by the participant and comprises the subcomponents: Event ( happenings , individuals present , weather conditions , physical/emotional actions , reactions in others ) ; Time ( time of the day , year , season , month , day of the week , hours ) ; Place ( location of the event ) ; Perceptual ( auditory , olfactory , tactile , visual and visual details , body position , duration of time ) ; Thought/emotion ( emotional states , thoughts , implications ) . The internal details score is the sum of these subcomponents . Anything tangential to the main dream event was scored as external details , including Event ( details from other events outside of the dream ) ; Semantic ( general , ongoing , extended knowledge/event/state of being ) ; Repetition; Other ( metacognitive statements , editorializing ) . The external details score is the sum of these subcomponents . A second method we employed , the Scene Construction Test ( Hassabis et al . , 2007 ) , is usually used for scoring the content of imagined scenes . Here we focused on the content score , which comprises four subcomponents: Entities Present ( objects or people ) ; Spatial References ( places or spatial relationships between entities ) ; Sensory Descriptions ( details that describe an entity ) ; Thought/emotion/action ( thoughts , emotional states , action descriptions ) . The content score is the sum of these subcomponents . All participants underwent PSG in their homes using a Brain Products system ( GmbH , Gilching , Germany ) . The purpose of the PSG recording was to ensure that we awakened participants during both NREM and REM sleep , and in a similar manner for the patient and control groups . Two trained sleep researchers arrived at a participant’s home approximately three hours before the usual bedtime to set up for the PSG . Equipment was then removed by a researcher the following morning upon awakening . PSG was recorded using a 24-electrode cap ( EasyCap; based on the international 10–20 system ) including the following EEG channels: Fp1 , Fp2 , F3 , F4 , C3 , C4 , P3 , P4 , O1 , O2 , F7 , F8 , T7 , T8 , P7 , P8 , Fz , Cz , Pz , Oz , FT9 , FT10 referenced to average mastoids ( M1 and M2 ) ( sampling rate = 500 Hz ) . This montage also included two bipolar electrooculogram channels ( EOG ) , two electromyogram channels ( EMG ) and two electrocardiogram channels ( ECG ) . Sleep staging was performed based on EOG , EMG and the following derivations: F3/M2 , F4/M1 , C3/M2 , C4/M1 , O1/M2 , O2/M1 . All statistical analyses were performed with SPSS 25 . 0 ( IBM Corporation ) . Given that the data did not meet the assumptions of normality and homogeneity necessary for parametric statistics , between-group analyses were performed using non-parametric Mann-Whitney U tests . We also calculated the effect sizes using non-parametric Cohen’s d for all tests performed . In all analyses , the significance level was set at 0 . 05 . | Dreaming has intrigued humans for thousands of years , but why we dream still remains somewhat of a mystery . Although dreams are not a precise replay of our memories , one idea is that dreaming helps people process past experiences as they sleep . If this is true , then part of the brain called the hippocampus that is important for memory should also be necessary for dreaming . Damage to the hippocampus can cause a condition called amnesia that prevents people from forming new memories and remembering past experiences . However , studies examining dreaming in people with amnesia have produced mixed results: some found that damage to the hippocampus had no effect on dreams , while others found it caused people to have repetitive dreams that lacked detail . One reason for these inconsistencies is that some studies asked participants about their dreams the next morning by which time most people , particularly those with amnesia , have forgotten if they dreamed . To overcome this limitation , Spanò et al . asked participants about their dreams immediately after being woken up at various points during the night . The experiment was carried out with four people who had damage to both the left and right hippocampus and ten healthy volunteers . Spanò et al . found that the people with hippocampal damage reported fewer dreams and the dreams they had were much less detailed . These findings suggest that a healthy hippocampus is necessary for both memory and dreaming , reinforcing the link between the two . Hippocampal damage is associated with a number of diseases , including dementia . If these diseases cause patients to dream less , this may worsen the memory difficulties associated with these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2020 | Dreaming with hippocampal damage |
The centromere is a specialized chromatin region marked by the histone H3 variant CENP-A . Although active centromeric transcription has been documented for over a decade , the role of centromeric transcription or transcripts has been elusive . Here , we report that centromeric α-satellite transcription is dependent on RNA Polymerase II and occurs at late mitosis into early G1 , concurrent with the timing of new CENP-A assembly . Inhibition of RNA Polymerase II-dependent transcription abrogates the recruitment of CENP-A and its chaperone HJURP to native human centromeres . Biochemical characterization of CENP-A associated RNAs reveals a 1 . 3 kb molecule that originates from centromeres , which physically interacts with the soluble pre-assembly HJURP/CENP-A complex in vivo , and whose down-regulation leads to the loss of CENP-A and HJURP at centromeres . This study describes a novel function for human centromeric long non-coding RNAs in the recruitment of HJURP and CENP-A , implicating RNA-based chaperone targeting in histone variant assembly .
Specialized chromatin domains called centromeres play an essential role in chromosome segregation , serving as a platform for kinetochore complex formation , which in turn binds spindle microtubules at mitosis ( Verdaasdonk and Bloom , 2011 ) . Although centromeric DNA sequence is not uniform across species , centromere function is conserved ( Sullivan , 2009 ) . In humans , AT-rich 171 bp α-satellite repeats lacking any known genes are the primary DNA component of the centromere ( Waye and Willard , 1987 ) . Centromeres are characterized by the presence of the centromeric histone H3 variant ( CENH3/CENP-A in human ) . Centromeric chromatin has long been considered heterochromatic , despite exhibiting a bivalent organization with heterochromatin-like post-translational modifications ( PTMs ) , such as H3 and H4 hypo-acetylation , and transcription-coupled PTMs , including dimethylation on H3 lysine 4 ( H3K4me2 ) ( Sullivan and Karpen , 2004; Heintzman et al . , 2007; Zhou et al . , 2011 ) . The function of such bivalent modifications has remained mysterious . Indeed , despite the common assumption that centromeres are largely dormant , a number of recent studies have pointed to the importance of transcription at centromeres in multiple organisms , which appears to be essential for the maintenance of centromere integrity ( Hall et al . , 2012 ) . Chemical inhibition of either RNA Polymerase I ( RNAPI ) , or RNA Polymerase II ( RNAPII ) , results in loss of the inner kinetochore protein CENP-C , and in chromosome mis-segregation ( Wong et al . , 2007; Chan et al . , 2012 ) . Centromeric RNA components also seem to contribute to the structural integrity of the mitotic centromere ( Wong et al . , 2007 ) . However , the exact timing of centromeric transcription , the polymerase involved , the identity of centromeric RNAs and their precise role in maintaining native centromere integrity in human cells has been elusive . In this study , we report that centromeric RNAs play a critical role in the maintenance of the human centromere in vivo . Using chromatin immunoprecipitation ( IP ) , and immunofluorescence ( IF ) on chromatin fibers , we find that RNAPII , in conjunction with TATA-box binding protein ( TBP ) localizes to , and actively transcribes native human centromeres from late mitosis to early G1 ( eG1 ) . Biochemical purification and sequencing of the RNA associated with human centromeric chromatin at eG1 reveals a 1 . 3 kb long transcript . This RNA physically interacts with CENP-A and its chaperone HJURP ( Holliday junction recognition protein ) in the pre-assembly soluble state in vivo . Targeted sequence-specific knockdown of the transcript results in the formation of multipolar spindles and lagging chromosomes in subsequent mitoses , leading to chromosome instability . IF analysis of centromeric chromatin fibers demonstrates that these cellular and nuclear phenotypes arise specifically from the abrogation of CENP-A and HJURP localization to the centromere . Together , our data describe a direct physical role for a centromeric long non-coding RNA ( lncRNA ) in HJURP targeting , subsequent CENP-A loading , and the maintenance of centromere integrity . Our study supports the possibility that an lncRNA-based mechanism is involved in targeting CENP-A and its chaperone HJURP to the centromere .
Centromeric transcription has been previously described in human cells , and RNAPII has been implicated in this process ( Saffery et al . , 2003; Wong et al . , 2007; Bergmann et al . , 2011; Chan et al . , 2012 ) . To investigate the timing of centromeric transcription , we used synchronized HeLa cells at G2 , eG1 , and G1/S to track the activated form of RNAPII ( i . e . , serine two phosphorylated , RNAPIIS2P ) on centromeric chromatin fibers throughout the cell cycle by IF ( Figure 1—figure supplement 1 ) . RNAPIIS2P co-localizes with the inner kinetochore protein CENP-B and centromeric α-satellite DNA specifically at eG1 ( Figure 1A , Figure 1—figure supplement 2A ) . We also noted that TBP , a partner of RNAPII normally involved in transcription initiation ( Vannini and Cramer , 2012 ) , is localized on eG1 CENP-A-rich fibers ( Figure 1A ) . These data suggest that centromeres are actively transcribed by RNAPII machinery at eG1 . We next sought to establish whether there was a physical interaction between CENP-A chromatin and RNAPII . In order to achieve this , we extracted chromatin from non-synchronized cells after a short MNase digestion , to obtain long chromatin arrays that are rich in tri- , tetra- , and penta-nucleosomes ( Figure 1—figure supplement 2B ) . From this input chromatin , centromeric chromatin was immunoprecipitated with specific antibodies against either CENP-A , or the inner kinetochore protein CENP-C or no antibody ( mock IP ) . The mock IP control shows no enrichment of any of the centromeric proteins tested ( Figure 1—figure supplement 2C ) . As expected , Western blots revealed reciprocal co-purification of CENP-A and CENP-C ( Figure 1B , left and middle panels ) . RNAPII and its partner TBP also co-purified with CENP-A and CENP-C ( Figure 1B , left and middle panels ) . To further establish an interaction between RNAPII and centromeric proteins , we performed the reciprocal experiment , precipitating RNAPIIS2P from solubilized chromatin , and testing for centromeric partners . While Western blots revealed little or no interaction of RNAPII with CENP-C , a robust and reproducible binding of RNAPII to TBP and CENP-A was observed ( Figure 1B , right panel ) . Thus , active RNAPII machinery is physically associated with CENP-A chromatin at eG1 . Recent studies have indicated that RNAPII transcribes centromeres at mitosis ( Chan et al . , 2012 ) . However , our RNAPII localization data above showed RNAPII enrichment occurs primarily at eG1 ( Figure 1A ) . To examine the consequence of eG1 RNAPII localization at centromeres , HeLa cells were synchronized at G2 , mitosis and eG1 , and briefly treated ( 2 hr ) with drugs to specifically block either RNAPI ( actinomycin D ) or RNAPII ( α-amanitin ) activity ( Figure 1—figure supplement 1 ) ( Bensaude , 2011 ) . After RNA extraction and retro-transcription , we determined the expression of control genes and centromeric α-satellite repeats by semi-quantitative PCR . As expected , actinomycin D and α-amanitin inhibited transcription of target genes of RNAPI ( e . g . , 18S rRNA ) or RNAPII ( e . g . , GAPDH ) , respectively ( Figure 2 , left and middle graphs ) . Compared to non-treated conditions , actinomycin D treatment or α-amanitin treatment in G2 phase had no impact on centromeric α-satellite expression ( Figure 2 , right graph ) . Consistent with a previous study ( Chan et al . , 2012 ) , RNAPII inhibition in mitotic cells revealed a decrease ( 15 . 6% ) in centromeric transcripts ( Figure 2 , right graph ) . However , when RNAPII was blocked in eG1 , a larger reduction ( 35 . 1% ) was observed ( Figure 2 , right graph ) . These results suggest that RNAPII transcribes centromeres not solely at mitosis , but also throughout eG1 . The synchrony of centromeric transcription and CENP-A recruitment onto centromeres at late mitosis-eG1 led us to examine whether active transcription is required for CENP-A loading . To test this hypothesis , we briefly treated eG1-synchronized cells with α-amanitin ( 2 hr ) to block RNAPII activity as above , and quantified potential changes in intensity for CENP-A or CENP-B IF signal using ImageJ . Consistent with its role as a constitutive centromeric DNA-binding protein ( Verdaasdonk and Bloom , 2011 ) , CENP-B staining intensity was heterogeneous ( Figure 3B , Figure 3—figure supplement 1 ) , but identical in both non-treated ( NT ) and α-amanitin-treated cells ( Figure 3A , B; Supplementary file 1 ) , demonstrating that its localization is independent of centromeric transcription . Whereas punctate CENP-A spots can be seen under both conditions , when RNAPII was blocked , the intensity of the CENP-A signal was decreased by ∼50% in cells ( Figure 3A , B; Supplementary file 1 ) . To ensure that this decrease was not due to reduced expression of CENP-A or its chaperone HJURP , total levels of both proteins were quantified by Western blot . No noticeable changes in protein levels were detected upon α-amanitin treatment ( Figure 3—figure supplement 1B ) . These data indicate that equal amounts of CENP-A and HJURP were available at eG1 , but potentially unable to load at the centromere . The decrease of CENP-A signal at centromeres during eG1 after RNAPII inhibition might be due to either the loss of pre-existing CENP-A , or a defect in targeting of newly synthesized CENP-A by its chaperone HJURP . To discriminate between these two hypotheses , we analyzed the localization of CENP-A , CENP-B , RNAPIIS2P , and HJURP on chromatin fibers , with or without α-amanitin treatment at eG1 . Consistent with the known effect of α-amanitin on RNAPII ( i . e . , blocking RNAPII elongation without inducing the release of RNAPII ) ( Nguyen et al . , 1996; Bensaude , 2011 ) , inhibition of transcription did not affect RNAPIIS2P localization onto the centromeric fiber ( Figure 3C , first top panel ) . Similarly , CENP-B localization onto centromeric chromatin fibers remained unaffected as well ( Figure 3C , first panel ) . Consistent with the whole cell data presented above ( Figure 3A , B ) , after RNAPII inhibition , not only were CENP-A signals reduced on chromatin fibers ( Figure 3C , second panel ) , HJURP was almost completely lost ( Figure 3C , second panel ) . The loss of HJURP ( Figure 3C ) combined with the ∼50% decrease of CENP-A signal after α-amanitin treatment ( Figure 3B ) , suggests that RNAPII-dependent transcription is required for the targeting of HJURP , and for the subsequent loading of new CENP-A to the centromeric chromatin fiber at eG1 . Previous data have documented the existence of ncRNA at centromeres in multiple species ( Hall et al . , 2012 ) . In humans , no genes have been annotated within native centromeres , suggesting a transcription event at α-satellite DNA repeats most likely leads to the synthesis of ncRNAs . In order to characterize potential centromeric transcripts , we sought to purify them biochemically . Total RNAs were extracted from cells , DNase I treated to remove genomic contamination , separated on denaturing gels , transferred to Northern blots , and subjected to hybridization with radiolabelled centromeric α-satellite probes , in order to reveal potential complementary transcripts . Northern blots revealed a unique centromeric RNA species migrating at approximately ∼1 . 3 kb ( Figure 4A , Figure 4—figure supplement 1A ) . Control experiments were performed to exclude the possibility of trace genomic DNA contamination contributing to the 1 . 3 kb band . Treatment of RNA samples with RNase A ( Figure 4—figure supplement 1B ) , or purification of RNA from cells treated with α-amanitin ( Figure 4—figure supplement 1C ) , both demonstrated the absence of the 1 . 3 kb band on Northern blots , supporting the interpretation that the 1 . 3 kb band derives solely from an RNA species . The inhibition of transcription was accompanied by the loss of CENP-A and HJURP at the centromere during eG1 ( Figure 3C ) , and our results above ( Figure 4A ) supported the possibility of a unique RNA species present at centromeres in eG1 . A logical prediction arising from these data is that centromeric transcripts might physically associate with the soluble pre-assembly HJURP/CENP-A complex in vivo . Indeed , computational RNA-binding prediction algorithms revealed potential RNA binding residues in both HJURP and CENP-A ( Figure 4—figure supplement 2; Wang et al . , 2010 ) . Thus , to further test this hypothesis , we probed for physical interactions between CENP-A and its chaperone HJURP with centromeric α-satellite transcripts . After a brief MNase digestion of eG1-synchronized cells , we immunoprecipitated CENP-A and HJURP from both , the soluble fraction ( composed of free histones and nuclear factors , SF ) , and the chromatin fraction ( composed of chromatin and associated complexes , CF ) ( Experimental Scheme , Figure 4—figure supplement 3 ) . CENP-A and HJURP complexes were immunoprecipitated from SF and CF . Mock IPs pulled down neither CENP-A nor HJURP ( Figure 4—figure supplement 4A ) . Consistent with HJURP chaperoning CENP-A at eG1 ( Dunleavy et al . , 2009; Foltz et al . , 2009; Shuaib et al . , 2010 ) , these proteins co-purified from both SF and CF ( Figure 4—figure supplement 4A ) . From these IPs , RNAs were purified , electrophoresed , transferred to Northern blots , and subsequently hybridized to the same radiolabelled centromeric α-satellite probes as above ( Figure 4A ) . These Northern blots revealed no RNA signal in the mock IP ( Figure 4—figure supplement 4B ) . In contrast , the 1 . 3 kb RNA is physically associated with CENP-A in both SF and CF ( Figure 4B ) , and interacts with HJURP only in the SF ( Figure 4B ) . These data provide evidence that the 1 . 3 kb centromeric RNA physically associates with the soluble HJURP/CENP-A pre-assembly complex at eG1 . We next sought to purify , clone using a conventional TOPO T/A cloning strategy and sequence CENP-A-associated RNA ( Experimental Scheme , Figure 4—figure supplement 3 ) . This sequencing approach was moderately successful , yielding one sequence of ~675 nucleotides ( cenRNA#1 , Figure 5—figure supplement 1 ) . This RNA sequence is unique and contains four semi-regular spaced 28 bp repeats with a weak homology ( ~52% ) to the canonical CENP-B box ( Supplementary file 2 ) , but does not map to the currently annotated human genome sequence , to any other organisms , or to plasmids . Over the course of the subsequent two years after publication , we made additional attempts to map centromeric RNAs , turning to a high-throughput approach coupled to CENP-A and HJURP RIP-Seq . These data yielded ~435 centromeric-mapping RNA sequences ( Quénet et al . , 2016 ) , which , by RNA FISH localize to centromeres ( unpublished ) . However , the initial cloned CenRNA#1 sequence did not map to this database of centromeric RNAs . This led us to re-sequence 6 clones of CenRNA#1 , which led to the rescue of 12 N’s and 27 base calls changed in the 675bp sequence ( Figure 1 ) . We then performed an expanded sequence search across multiple sequence databases beyond the publicly available NCBI catalog , including the HeLa genome ( after obtaining permission from the HeLa DGAP working group ) , as well as an industrial database for cloning vectors Supplementary files 2 , 3 , 4 . Much to our dismay , we discovered that cenRNA#1 arose largely from fusions of adapters used in the RNA cloning approach and described in Pfeffer et al . ( Figure 2; Pfeffer et al . , 2005 ) . Based on this evidence , we no longer believe that the original sequence provided for putative cenRNA#1 represents a human centromeric transcript . In our initial version , we seek to assess the functional role of cenRNA#1 by an shRNA strategy to down-regulate specifically its expression ( Figure 5—figure supplement 1 ) . The two shRNAs were generated from the putative sequence of cenRNA#1 , against the 28bp repeat element sequence . Cells were transfected with control scrambled ( shRNAscram ) or shRNAcenRNA#1 constructs , and selected with puromycin Cells transfected with shRNAscram displayed no changes in morphology and density , whereas cells treated with shRNAcenRNA#1 displayed significant loss of cell density ( down by ~70% , relative to control ) , and presented aberrant morphology ( Figure 5—figure supplement 4A and 2B ) . We decided to readdress this experiment . Based on the re-sequencing results , three base calls within each shRNA were changed , which removes their uniqueness within cenRNA#1 and changes the percent identity to 26/29 bases ( Supplementary file 6 ) . Neither shRNA had a significant match to the human genome or known transcripts . However , the best hit is to a lncRNA on chromosome 3 ( 15/29 bases ) , which is proximal to the centromere , but not in the pericentromere ( Supplementary file 7 ) . This proximity may potentially explain the positive IF/FISH signal observed in our initial manuscript of CENP-A with cenRNA#1DNA probe ( Figure 5 ) . An independent reproduction of the down-regulation of cenRNA#1 by shRNA approach yielded the same chromosome defect as before ( M . Bui , data not shown ) . To re-test whether some fraction of cenRNA#1 matching sequence plays a role in chromosomal integrity , we next designed a locked nucleic acid antisense oligonucleotide ( LNA ASO ) and analyzed mitosis integrity . Indeed , LNA ASO targeting cenRNA#1 led to modestly increased rates of lagging chromosomes and cells with multi-polar spindles , when compared to either un-transfected , mock transfected , or scrambled transfected cells ( 28% and 13% compared to 10% and 3% for scrambled control; ( Figure 5—figure supplement 5 ) . This result may suggest a potential function for some fraction of this non-centromeric sequence on chromosome segregation and mitotic integrity , but which is not connected to the main findings of our original manuscript . We were curious whether it was the act of centromeric transcription alone , or the product of transcription ( i . e . , centromeric RNAs ) , that was necessary for HJURP and CENP-A targeting to the centromere at eG1 . To distinguish between these hypotheses , we further examined functional consequences arising from the targeted loss of total centromeric α-satellite RNA , without inhibiting RNAPII transcription . Using the centromeric α-satellite consensus sequence ( Waye and Willard , 1987 ) , we designed two shRNAs targeting α-satellite sequences ( shRNAsat1 and shRNAsat2 ) to destroy centromeric transcripts ( Figure 6—figure supplement 1A ) . At 6 days post-transfection with the control scrambled ( shRNAscram ) or shRNAsat constructs and puromycin selection , the expression level of centromeric α-satellite transcript was analyzed by qtPCR . Compared to control shRNAscram , cells transfected with shRNAsat constructs displayed a significant decrease ( ∼70% ) of the centromeric α-satellite transcript ( Figure 6— figure supplement 1B ) , confirming targeted destruction was accomplished . Evaluation of cell morphology by phase contrast microscopy revealed that shRNAsat-transfected cells were less dense ( down by ∼70% , relative to control ) and exhibited abnormal morphology ( Figure 6A ) . Phenotypes included a large and flat cytoplasm , cellular protrusions , and multinucleate cells ( Figure 6A ) . To better elucidate cell defects , we stained with β-actin , which revealed cells with several nuclei and atypical shape ( Figure 6B ) . Reduced cell density and morphological abnormalities could result from cells exiting the replicative cell cycle and undergoing senescence ( Kuilman et al . , 2010 ) . We performed β-galactosidase staining to test for senescence ( Bandyopadhyay et al . , 2005 ) . Senescent BJ cells were used as positive control , and displayed the expected blue color after the β-galactosidase assay ( Figure 6—figure supplement 2 ) . However , no significant increase in senescence was seen in either shRNAscram or shRNAsat-transfected cells ( Figure 6—figure supplement 2 ) . The kinds of cellular morphological changes observed above ( Figure 6A , B ) have previously been linked to defects in cell division , specifically in mitosis ( Carone et al . , 2013 ) . To test this alternative possibility , shRNA-transfected cells were synchronized at mitosis , and at 6 days post-transfection , stained for markers of mitotic spindles ( α-tubulin ) and centromeres ( CENP-B ) , respectively . shRNAscram-transfected cells displayed normal mitotic structures ( Figure 6C ) . In contrast , almost half ( 42 . 2% ) of shRNAsat-transfected cells presented abnormal mitoses , with multipolar spindles and lagging chromosomes ( Figure 6C ) . A mechanistic explanation for the observed mitotic aberrances in the centromeric transcript depleted cells ( Figure 6C ) could be loss of centromere integrity , driven by deficient targeting of HJURP/CENP-A complexes to centromeres . We were curious whether the loss of centromeric transcripts directly abrogated CENP-A and HJURP localization at centromere . Because the loss of centromeric transcripts resulted in reduced cell density and mitotic defects ( Figure 6 ) , there were insufficient cells for biochemical experiments . Thus , we turned to chromatin fibers to investigate this issue . After synchronization at eG1 , chromatin fibers were isolated from shRNAscram cells , or the few remaining of shRNAsat-transfected cells , and stained for RNAPII and centromeric proteins . In both , shRNAscram- or shRNAsat-transfected cells , RNAPIIS2P remains associated with centromeric chromatin fibers ( Figure 7 , first top panel ) , confirming that RNAPII localization and transcription of centromeres are independent of centromeric transcripts . Additionally , CENP-B and CENP-C localization was also unaffected at centromeres ( Figure 7 , second and third panels ) . In contrast , on centromeric fibers from shRNAsat-transfected cells , CENP-A and HJURP were barely detectable ( Figure 7 , fourth and fifth panels ) . These data suggest that at eG1 , CENP-A targeting through its chaperone HJURP is dependent not just on active transcription itself , nor on processes that facilitate centromeric transcription ( Barnhart et al . , 2011; Wang et al . , 2014 ) , but specifically requires the presence of centromeric transcripts . The long-term loss of these centromeric transcripts leads to mitotic defects ( Figure 6 ) , which are physically underpinned by the loss of HJURP recruitment and CENP-A loading ( Figure 7 ) to native centromeres .
Active transcription is thought to be essential for centromere structure and function ( Saffery et al . , 2003; Wong et al . , 2007; Bergmann et al . , 2011; Chan et al . , 2012 ) . In this study , we investigated the mechanistic contribution of transcription , and centromeric transcripts , to centromeric integrity . We show that RNAPII and TBP are loaded onto and transcribe human centromeric chromatin at eG1 ( Figure 1 ) . This cell cycle regulated centromeric transcription is required for the synthesis of centromeric RNAs ( Figure 2 ) . Biochemical purification and analysis reveal a 1 . 3 kb transcript which is physically associated with CENP-A and HJURP in the soluble pre-assembly state ( Figure 4 ) . Targeted destruction of this centromeric RNA leads to the loss of centromere integrity and subsequent mitotic and cellular defects ( Figure 6 ) , which are mechanistically underpinned by the loss of HJURP and CENP-A recruitment to centromeres at eG1 ( Figure 7 ) . Altogether , these data reveal a hitherto unsuspected function for lncRNAs in RNA-dependent chaperone targeting to centromeres in human cells ( Figure 7—figure supplement 1 ) . Several specific questions arise from our observations . First , no active genes have ever been described in human centromeres , making the identification of RNAPII ( Figure 1; Saffery et al . , 2003 ) and TBP at the centromere surprising . Because the repetitive nature of centromeric α-satellite DNA has thus far disallowed complete sequencing , successful annotation of transcriptional motifs that may exist in human centromeres remains to be accomplished . Our data show that centromeric transcription is an event integral to the epigenetic maintenance of centromere integrity , and discovering precisely where such motifs lie within active centromeres is an exciting avenue of research . Second , our functional characterization of a 1 . 3 kb centromeric lncRNA deriving from centromeric transcription reveals its interaction with CENP-A and HJURP at eG1 . Specific depletion of centromeric α-satellite transcripts affects the recruitment of both CENP-A and HJURP proteins , directly implicating centromeric RNA in CENP-A and HJURP targeting onto centromeres at eG1 . We note that previous data have shown that HJURP loading is also dependent on its interaction with the Mis18 complex , in a CDK1-dependent manner ( Barnhart et al . , 2011; Moree et al . , 2011; Müller et al . , 2014; Wang et al . , 2014 ) . However , mutations of CDK1-phosphorylated sites in HJURP only partially abrogate its recruitment in vivo , highlighting the existence of other hitherto unknown CENP-A loading factors ( Wang et al . , 2014 ) . We speculate centromeric lncRNAs are , in fact , the missing factor . Thus , of immediate interest is the elucidation of the structure of the 1 . 3 kb centromeric RNA with its cognate binding domains in CENP-A and HJURP . Third , the exact molecular process involved in targeting lncRNA-nucleoprotein complexes to centromeres is an unexpected and novel avenue to pursue . For example , whereas it is well known that Xist RNA binds its cognate DNA locus only in cis ( Plath et al . , 2002 ) , it is unknown if centromeric transcripts can bind in cis solely to the centromere of origin , or in trans , across all centromeres . An attractive possibility is centromeric RNAs originate from multiple centromeres and serve a redundant function to ensure accurate targeting of CENP-A/HJURP to homologous centromeres . Fourth , our study has potential evolutionary implications . Prior studies have described RNA originating from centromeres in multiple species ( Bouzinba-Segard et al . , 2006; Carone et al . , 2009 ) . In mouse cells , a 120 nucleotide minor satellite RNA is associated with centromeres ( Bouzinba-Segard et al . , 2006 ) , and in tammar wallaby , centromeric transcription results in the production of ∼40 nucleotides crasiRNAs ( centromere repeats-associated short interacting RNAs ) ( Carone et al . , 2009; Lindsay et al . , 2012; Carone et al . , 2013 ) . A logical explanation for the difference in size of ncRNAs generated in different organisms may be the divergent nature of the centromeric DNA sequences across species , which in turn may lead to divergence in the type of centromeric RNAs produced . However , despite this difference , over-expression or down-regulation of mouse minor satellite RNA , or crasiRNAs in tammar wallaby , or the 1 . 3 kb centromeric human RNA identified in our study , leads to similar cellular and mitotic defects . Our data reveal that such RNAs generated from human centromeric transcription bind HJURP and CENP-A in the soluble form and that mitotic loss seen in cells depleted of these lncRNAs is specifically linked to abrogation of HJURP-mediated targeting of CENP-A . Thus , our data suggest an evolutionarily conserved basis for the phenomena of centromeric transcription seen in other organisms . We speculate that accurate CENP-A targeting onto active centromeres probably requires a dual-lock system , coupling chromatin-bound centromeric factors ( such as Mis18 ) , which facilitate cell-cycle regulated centromeric transcription , which in turn results in the production of a lncRNA/CENP-A/chaperone complex that can effectively target CENP-A back to pre-existing active centromeric sites ( Figure 7—figure supplement 1 ) . It is noteworthy that transcription-coupled , chaperone-mediated histone variant assembly governs much of chromatin biology . Our report potentially reveals an RNA-based mechanism by which specialized histone-variant driven chromatin structures might be maintained in vivo .
LNA ASOs were designed and purchased by QIAGEN ( previously Exiqon ) . Supplementary file 9 lists sequences of these LNA ASO sequence . Transfection of HeLa cells with LNA ASO were conducted as in ( Bui et al . , 2017 ) . Briefly , cells were seeded 24 hours before transfection to allow no more than 75% confluency , and transfected using Lonza’s Amaxa Cell Line Nucleofector Kit R ( Cat #VCA-1001 ) with Amaxa Biosystems Nucleofector II electroporation system using program O-005 . After transfection , cells were grown on coverslips with fresh DMEM medium . The day following LNA ASO transfection , cells were synchronized with a double thymidine block and released for 10 . 5 hours on the third day , to enrich for anaphase-cytokinesis staged cells . IF experiments were performed as described previously ( Bui et al . , 2012 ) . Cells were fixed with 4% paraformaldehyde in 1X PBS ( #14190-144; Gibco by Life Technologies ) for 15 min , permeabilized with 0 . 5% Triton X-100 in 1X PBS for 10 min , and blocked with 3% bovine serum albumin ( BSA , #BP9706-100; Fisher Scientific ) in 1X PBS . Coverslips were immuno-stained for CENP-C and α-tubulin for one hour each ( Supplementary file 8 ) . After three washes in 1X PBS , 0 . 1% Tween ( #P7949-500ML; Sigma-Aldrich ) , cells were incubated with secondary antibody ( goat anti-guinea pig or anti-mouse IgG ( H+L ) secondary antibodies , Alexa Fluor568 and Alexa Fluor488 conjugates ( Thermo Fisher Scientific ) ) in 1X PBS for 1 hour at RT in the dark . Finally , cells were washed three times for 5 min at RT . Coverslips were mounted with anti-fade mounting medium Prolong Gold with DAPI . IF slides were observed with a DeltaVision Elite RT microscopy imaging system ( GE Healthcare ) controlling an interline charge-coupled device camera ( Coolsnap ) mounted on an inverted microscope ( IX-70; Olympus ) . Images were captured by using a 60X objective at 0 . 2μm z-sections and analyzed with Image J ( 1 . 50e; Java 1 . 6 . 0_20 ) . Antibodies are commercially available , except the custom CENP-A antibody ( available upon request ) used for CENP-A detection on Western blot . Supplementary file 8 lists all antibodies used for each experiment . HeLa cells were grown at 37°C in a humidified atmosphere containing 5% CO2 , in Dulbecco's modified Eagle's medium high in glucose and L-glutamine ( #11965; Gibco , Grand Island , NY ) supplemented with 10% Fetal Bovine Serum ( #26140 – 079; Gibco ) and 1X Pen/Strep solution ( #10378 – 016; Gibco ) . All synchronizations were done by double thymidine block ( 0 . 5 mM , #T9250; Sigma-Aldrich , Saint Louis , MO ) . After a first block of 19 hr , cells were released for 9 hr , followed by a second thymidine block of 16 hr . Cells then were released for the appropriate time ( 9 hr for G2 , 10 hr for mitosis , and 11 hr for eG1 , Figure 1—figure supplement 1 ) . Synchronization was assessed by flow cytometry . Cells were stained with propidium iodide ( #P817045 , Invitrogen , Grand Island , NY ) and analyzed on a FACScalibur ( Becton Dickinson , San Jose , CA ) . Synchronized cells were treated with either 0 . 2 μg/ml of actinomycin D ( #A2263 , Sigma-Aldrich ) or 2 μg/ml of α-amanitin ( #A1410; Sigma-Aldrich ) to analyze the effect of RNAPI and RNAPII inhibition , respectively , on centromere transcription . RNAs were extracted by Trizol reagent ( #15596 – 026; Ambion , Grand Island , NY ) according to manufacturer protocol . Briefly , cells were resuspended in Trizol , and following 5 min incubation at room temperature ( RT ) , 200 μl of chloroform ( #BP1145-1 , Fisher Scientific , Pittsburgh , PA ) was added . After centrifugation at 12 , 000 rpm for 15 min at 4°C , the clear phase was mixed with 500 μl of isopropanol ( #534021 , Sigma-Aldrich ) and centrifuged . The pellet was washed with 75% ethanol ( diluted from 100% ethanol , #61509 – 0010 , Acros Organics , Pittsburgh , PA ) and resuspended in water complemented with DNase I buffer , DNase I ( #M0303; New England Biolabs NEB , Ipswich , MA ) , and RNase inhibitor ( #M0314 , NEB ) to avoid genomic DNA contamination . After incubation for 30 min at 37°C , the DNase I activity was inhibited by addition of 5 mM EDTA ( #351-027-721 , Quality Biological , Gaithersburg , MD ) and incubation at 65°C for 10 min . RNAs were purified a second time by phenol:chloroform:isoamyl alcohol ( 25:24:1; #AC327115000 ) method and ethanol precipitated . RNAs were conserved at −80°C until further analysis . After quantification by UV-spectroscopy ( 230 , 260 , and 280 nm ) and verification of RNA quality on 1 . 5% agarose gel , equivalent concentrations of RNA were subjected to retro-transcription , using the SuperScript III First-Strand Synthesis System with random hexamer primers ( #18080 – 051 ) , and amplified with Takara PCR kit ( #RR001B; Clontech Laboratories Inc . , Mountain View , CA ) . Control reactions without the reverse transcriptase or complementary DNA were performed to rule out DNA contamination and non-specific amplification , respectively . Primer sequences are included in Supplementary file 3 ( Dunham et al . , 1992 ) . PCR conditions were defined for each analyzed sequence . The setup for GAPDH and centromeric α-satellite were 3 min 94°C; [10 s 98°C , 30 s 57°C , 30 s 72°C] 30 cycles; 5 min 72°C . The conditions for 18S rRNA were 3 min 94°C; [10 s 98°C , 30 s 52°C , 30 s 72°C] 30 cycles; 5 min 72°C . Finally , the PCR status for cenRNA#1 were 3 min 94°C; [10 s 98°C , 30 s 57°C , 30 s 72°C] 45 cycles; 5 min 72°C . Cells were grown on poly-D-Lysine-treated coverslips in six-well plate and synchronized by double thymidine block . After two washes with cold 1× PBS , they were prefixed for 30 s with cold 4% paraformaldehyde ( PFA , #15714 s; Electron Microscopy Sciences , Hatfield , PA ) in PEM ( 80 mM K-PIPES pH6 . 8 , 5 mM EGTA pH7 . 0 , 2 mM MgCl2 ) . Following three washes with cold PEM , soluble proteins were extracted for 5 min on ice with 0 . 5% Triton X-100 ( #327372500 , Acros Organics ) in CSK ( 10 mM PIPES pH6 . 8 , 100 mM NaCl , 300 mM sucrose , 1 mM EGTA , 3 mM MgCl2 ) . Few drops of 4% PFA in PEM were added for 5 min . Slides were then incubated with fresh 4% PFA in PEM for 40 min on ice . After three washes with PEM , cells were permeabilized with 0 . 5% Triton X-100 in PEM for 30 min at RT , washed again three times , and blocked in 1×— TBS , 3% Bovine Serum Albumin , 5% normal goat serum ( #005-000-121; Jackson ImmunoResearch , West Grove , PA ) for 1 hr at RT . Finally cells were incubated with the primary antibody diluted in 1×— TBS , 1% Bovine Serum Albumin , 5% normal goat serum over-night ( O/N ) at 4°C in a humidified chamber . Slides were washed three times for 5 min at RT with 1×— TBS , 0 . 1% Tween 20 ( #P7949 , Sigma-Aldrich ) , and incubated with secondary antibody for 1 hr at RT . After washing , the same protocol was repeated for co-IF , and then cells were stained with DAPI ( 4′ , 6-diamidino-2-phenylindole , #D9542; Sigma-Aldrich ) in 1×— TBS and mounted with mowiol solution ( Amé et al . , 2009 ) . For β-actin IF , a classic protocol was used . Briefly , cells were fixed with 2% PFA , 1× PBS for 10 min on ice . After three washes with 1× PBS , 0 . 1% Triton X-100 , 1% Bovine Serum Albumin , cells were incubated with the primary antibody diluted in 1× PBS , 0 . 1% Triton X-100 , 1% Bovine Serum Albumin , 5% normal goat serum O/N at 4°C in a humidified chamber . Slides were washed three times for 5 min at RT and incubated with secondary antibody for 1 hr at RT . After washing , cells were stained with DAPI in 1× PBS and mounted with mowiol solution . Chromatin fiber protocol was adapted from Sullivan ( 2010 ) . Trypsinized HeLa cells were incubated in hypotonic buffer ( 75 mM KCl ) for 10 min at RT , before cytospining for 10 min at 400 rpm . Slides were immersed in freshly prepared fiber lysis buffer ( 2 . 5 mM Tris HCl pH7 . 5 , 0 . 5 M NaCl , 1% Triton X-100 , 0 . 4 M urea ) for 15 min at RT , then in fixation buffer ( 4% formaldehyde ( #F8775; Sigma-Aldrich ) , 1X PBS , final pH 7 . 4 ) for 10 min at RT , and finally in permeabilization buffer ( 1X PBS; 0 . 1% Triton X-100 ) for 7 min at RT . After blocking ( 1X PBS , 0 . 5% Bovine Serum Albumin , 0 . 01% Triton X-100 ) , chromatin fibers were stained O/N at 4°C in a humidified chamber with primary antibody diluted in blocking solution complemented with 1% normal goat serum . Slides were washed three times for 5 min at RT with 1X PBS , 0 . 05% Tween 20 , before incubation with secondary antibody for 1 hr at RT . After washing , the protocol was repeated for co-IF , and the fibers were then stained with DAPI in 1× PBS and mounted with mowiol solution . When FISH was performed , antibody protein complexes were crosslinked ( 8% formaldehyde diluted in distilled water ) for 10 min at RT , denaturated in 70% formamide ( #F47670; Sigma-Aldrich ) , 2X SSC buffer ( #46 – 020 CM , Corning , Manassas , VA ) for 8 min at 78°C , and then incubated with denaturated probed ( tagged with biotin-16-dUTP [#11093070910 , Roche , Indianapolis , IN] or CyTM5 dUTP [#PA55022 , GE Healthcare , Pittsburgh , PA] by nick translation method ) O/N at 37°C in a humidified chamber . Slides were washed 5 min at 45°C three times with 50% formamide , 2X SSC solution , and four times with 2X SSC , 0 . 05% Tween 20 solution . Slides were blocked in 4X SSC , 0 . 1% Tween 20 , 3% Bovine Serum Albumin for 30 min at RT . Following the incubation with the secondary antibody at 37°C for 1 hr , slides were washed four times for 5 min each at 45°C with 4X SSC , 0 . 1% Tween 20 , stained with DAPI in 2X SSC , and mounted coverslip with mowiol solution . All samples were observed with a DeltaVision RT system ( Applied Precision , Issaquah , WA ) controlling an interline charge-coupled device camera ( Coolsnap , Roper Scientific , Martinsried , Germany ) mounted on an inverted microscope ( IX-70; Olympus , Center Valley , PA ) . Images were captured by using a 60×— at 0 . 2 μm z sections for cell and 100×— objective at 0 . 1 μm z sections for chromatin fiber , deconvolved , and projected by using softWoRx ( Applied Precision ) . Three independent experiments were performed and in each , 5–10 chromatin fibers or 30–50 cells were analyzed per slide . To quantify IF signals , the acquisition of pictures for all samples of an experiment was performed with the same time of exposure during the same day to avoid variability from the instrument . Using ImageJ ( ImageJ 1 . 43U ) , signal intensity of each CENP-A or CENP-B spot inside of the nucleus ( as defined by the DAPI staining ) was extracted . The background level of the nucleus was subtracted from the average value of the spot intensity per cell . For each experiment , the average value of the spot intensity per cell and the ratio signal intensity in treated condition vs signal intensity in non-treated condition was measured . The mean and standard deviation values of three experiments are presented in the Supplementary file 1 . Five F175 flasks of HeLa cells ( 70–80% of confluence ) were used for IP . Cells were trypsinized ( #25300; Gibco ) and washed three times with cold 1× PBS , 0 . 1% Tween 20 coupled with 5 min centrifugation at 800 rpm at 4°C . Nuclei were isolated in TM2 buffer ( 20 mM Tris HCl pH8 , 2 mM MgCl2 , 0 . 5 mM PMSF ) complemented with 0 . 5% NP40 substitute ( #74385 , Sigma-Aldrich ) , and washed once with TM2 buffer . Chromatin was digested 2 min at 37°C with 0 . 2 unit/ml of MNase ( #N3755; Sigma-Aldrich ) in 0 . 1 M TE buffer ( 0 . 1 M NaCl , 10 mM Tris HCl pH8 , 0 . 2 mM EGTA ) complemented with 2 mM CaCl2 . The reaction was stopped by addition of 10 mM EGTA and transferred to ice . After centrifugation for 5 min at 800 rpm at 4°C , the nuclear pellet was resuspended in 1 ml low-salt buffer ( 0 . 5× PBS , 5 mM EGTA , 0 . 5 mM PMSF , protease inhibitor cocktail [#05892953001; Roche] ) , and the chromatin was extracted O/N at 4°C in an end-over-end rotator . An aliquot of the supernatant obtained after centrifugation for 5 min at 8000 rpm at 4°C was saved as input ( 1 . 5% ) . At 4°C , sample was precleared with protein A/G Plus agarose beads ( #sc-2003; Santa Cruz Biotechnology , Dallas , TX ) for 30 min at 4°C in an end-over-end rotator , incubated with the primary antibody for 4 hr , followed by IP with protein A/G Plus agarose beads for 2 hr . After centrifugation , an aliquot was saved as unbound ( UB , 1 . 5% ) , and the bead-associated IP was washed three times with low-salt buffer , and stored at −20°C in Laemmli buffer ( 30 μl ) for Western blot analysis . A general scheme is presented on Figure 4—figure supplement 3 . Five F175 flasks of eG1-synchronized HeLa cells ( 70–80% of confluency ) were used for IP . After trypsinization , cells were washed two times with cold 1× PBS , 0 . 1% Tween 20 , fixed 10 min at RT with 1% formaldehyde , quenched by addition of 125 mM glycine , and washed twice with cold 1X PBS , 0 . 1% Tween 20 . Samples were treated as described above ( i . e . , Chromatin extraction and immunoprecipitation ) in presence of 10 mM Ribonucleoside Vanadyl Complex ( RVC , #1402; NEB ) . After centrifugation 5 min at 800 rpm at 4°C following the MNase digestion , the supernatant was saved and named soluble fraction , whereas the nuclear pellet was suspended in 1 ml low-salt buffer ( 0 . 5× PBS , 5 mM EGTA , 0 . 5 mM PMSF , protease inhibitor cocktail , 10 mM RVC ) and chromatin was extracted O/N at 4°C . The supernatant obtained after centrifugation 5 min at 8000 rpm at 4°C was named chromatin fraction . IPs were performed as described previously with both fractions ( i . e . , Chromatin extraction and immunoprecipitation ) . After washes , bead-associated IPs were divided in two equal samples for protein analysis by Western blot and for RNA study . For RNA study , RNA protein complexes were eluted from protein A/G Plus agarose beads by incubation in an end-over-end rotator for 15 min at RT with 250 μl elution buffer ( 1% SDS; 0 . 1 M NaHCO3 ) . The supernatant was saved and the elution step was repeated once more . All samples ( input , unbound and IP ) were denatured at 65°C for 2 hr with 200 mM NaCl . Proteins were digested with 20 μg of proteinase K ( #AM2548; Ambion ) in the presence of 40 mM Tris HCl pH6 . 5 , 10 mM EDTA at 42°C for 45 min . To avoid genomic DNA contamination , samples were treated with DNase I for 30 min at 37°C . The reaction was stopped by addition of 5 mM EDTA , and RNAs were purified by phenol:chloroform:isoamyl alcohol method and ethanol precipitation . Samples were stored at 80°C until further use in retro-transcription PCR and Northern blot . Samples in Laemmli buffer were denatured 5 min at 95°C , plunged on ice for 2 min , loaded into a 4 20% SDS-PAGE ( #456 – 1093; Biorad , Hercules , CA ) for separation in 1×— Tris Glycine SDS Running Buffer ( #161 – 0732; Biorad ) , and transferred to Whatman nitrocellulose membrane ( #10439396; Sigma-Aldrich ) in Tris Glycine transfer buffer ( #351-087-131; Quality Biological ) diluted in 20% ethanol . Membrane was blocked in Odyssey blocking buffer ( #927 – 40000; Li-Cor , Lincoln , NE ) diluted in 1× PBS ( 1:1 ) at RT for 1 hr , and incubated with primary antibody diluted in blocking buffer complemented with 0 . 1% Tween 20 O/N at 4°C . After three washes in 1× PBS , 0 . 1% Tween 20 , the membrane was incubated with the secondary antibody conjugated to IRDye680 ( #926 – 68072 and #926 – 68073; Li-Cor ) diluted in blocking buffer complemented with 0 . 1% Tween 20% and 0 . 05% SDS for 1 hr at RT . The membrane was washed in the same conditions than previously and proteins were detected by scan on Odyssey CLx scanner ( Li-Cor ) . eG1-synchronized HeLa cells treated with α-amanitin ( as described in Cell culture and RNA Polymerase inhibition ) were resuspended in lysis buffer ( 20 mM Tris HCl pH7 . 5 , 400 mM NaCl , 2 mM dithiothreitol , 1% Nonidet P40 substitute , 0 . 5 mM PMSF , protease inhibitor cocktail ) . After three cycles of freezing and thawing , extract was centrifuged for 20 min at 12 , 000 rpm at 4°C . The cleared suspension was quantified by UV spectroscopy , and 50 μg of proteins were resuspended in Laemmli buffer . After separation into a 4 20% SDS-PAGE , transfer to a nitrocellulose membrane and incubation with primary and secondary antibodies ( as described in Western Blot ) , proteins were detected by scan on Odyssey CLx scanner and quantified with ImageStudioLite software ( Li-Cor ) . For each experiment , the ratio of signal intensity in treated condition vs signal intensity in non-treated condition was measured . The means and standard deviations of three experiments are presented in the figure . Northern blotting protocol was adapted from Summer et al . ( 2009 ) and http://archive . bio . ed . ac . uk/ribosys/protocols/website_Northern_blotting . pdf . 5 μg of Trizol extracted RNAs or 1 μg of immunoprecipitated RNAs was separated on 4% urea PAGE against 0 . 5×— TBE buffer at 25 W for 90 min . RNAs were transferred to Amersham Hybond-NX membrane ( #RPN203T; GE Healthcare ) for 2 hr at 65 V , UV-cross-linked , blocked for 1 hr in SES buffer ( 0 . 5 M Na3PO4 pH7 . 2 , 7% [wt/vol] SDS , 1 mM EDTA ) , and hybridized O/N at 37°C with radiolabelled α-satellite probes ( end-labeling method using primer extension-system AMV reverse transcriptase kit , #E3030; Promega , Madison , WI ) diluted in SES buffer . Membrane was washed in 6×— SSPE ( 1 . 08 M NaCl , 0 . 06 M NaH2PO4 , 20 mM EDTA , pH adjusted to 7 . 4 ) two times for 15 min at 37°C and two times for 30 min at 42°C . Blot was exposed to P32-sensitive film ( Hyblot CL film , #E3012; Denville , Saint Laurent , Canada ) at −80°C to reveal the potential interaction for a short ( <24 hr ) or long ( >24 hr ) period of time . The sequences of the radiolabeled probes are indicated in Supplementary file 10 . Computational prediction of RNA binding residues was performed with BindN+ program ( http://bioinfo . ggc . org/bindn+/ ) with a specificity equal to 85% ( Wang et al . , 2010 ) . We used human CENP-A ( P49450 ) , H3 . 1 ( P68431 ) and HJURP ( Q8NCD3 ) , and Scm3 ( Q12334 ) protein sequences from uniprot database . pGFP-V-RS plasmids expressing shRNA sequence were purchased from Origine ( Rockville , MD ) . Two controls were used for each experiment: the empty vector ( #TR30007 ) and the vector expressing scrambled sequence cassette ( #TR30013; shRNAscram 5′-GCACTACCAGAGCTAACTCAGATAGTACT-3′ ) . Two shRNA sequence cassettes were designed from the centromeric α-satellite consensus sequence ( Waye and Willard , 1987 ) : shRNAsat1 5′-TGTGTGCATTCAACTCACAGAGTTG-3′ and shRNAsat2 5′-CAACTCACAGAGTTGAACCTTCCTT-3′ ( Figure 6—figure supplement 1 ) ; and from the cenRNA#1 sequence ( Figure 5—figure supplement 1 ) : shRNAcenRNA#1 5′-TGCTAGACAGCCAATGCAATTCCTCATTA-3′ . Cells were transfected with Escort II Transfection Reagent ( #L6037; Sigma-Aldrich ) following manufacturer instruction . 48 hr after transfection , the medium was replaced every 2 days with fresh medium complemented with 0 . 5 μg/ml puromycin ( #A1113802; Gibco ) to select transfected cells . At day 6 , cells were either treated for IF or RNA extraction . To detect α-satellite expression level in shRNA-transfected cells , RNAs were extracted , quantified by UV-spectroscopy , and equal quantities were retro-transcribed using Superscript III First-Strand Synthesis kit as described above ( i . e . , RNA extraction , retro-transcription , and Polymerase Chain Reaction ) . To perform qtPCR , complementary DNAs ( cDNAs ) samples were prepared using the iQ SYBR Green supermix ( #170–8880; Biorad ) following manufacturer's protocol . Control reactions without the cDNA were performed to rule out non-specific amplification . The qtPCR was run on Step one plus Real time PCR system ( Applied Biosystem , Grand Island , NY ) . Primer sequences are available on Supplementary file 3 . The comparative cycle threshold ( CT ) method was used to analyze the expression level of α-satellite transcripts . CT values were normalized against the average CT value of the housekeeping gene β-actin . The ΔΔCT values were determined from the scrambled shRNA samples . Relative fold differences ( 2− ΔΔCT ) are indicated on figure . Senescent cells were detected using the protocol developed by Itahana et al . ( 2007 ) . Briefly , 6 days post transfection with shRNA , HeLa cells grown in a six-well plate were washed two times in 1× PBS , fixed 5 min at RT with 3 . 7% formaldehyde in 1X PBS , and washed twice with 1X PBS . Cells were stained with the X-gal staining solution ( 1 mg/ml of X-gal [#B9146; Sigma-Aldrich] , 40 mM citric acid/sodium phosphate buffer pH 6 . 0 , 5 mM potassium ferricyanide [#702587; Sigma-Aldrich] , 5 mM potassium ferrocyanide [#P3289; Sigma-Aldrich] , 150 mM NaCl , 2 mM MgCl2 ) O/N at 37°C . After rinsing , cells were observed under a light microscope for blue color , indicator of senescence . Standard deviation was determined for all quantification measures . To test the significance of these measures , a two-tailed , paired Student's t test was performed . For all tests α was assumed to be 0 . 05 . The p-value is indicated on the figures or tables each time it was evaluated . | Before a cell divides , it copies its chromosomes . Initially , the two copies of each chromosome remain linked via their centromeres . These regions also serve as the attachment sites for the proteins that pull these two copies apart , and eventually segregate the chromosomes equally between the two newly formed cells . Chromosome segregation is the main function of centromeres; and in most organisms , the DNA in these regions is highly repetitive and is not thought to encode any proteins . However , it has been observed that cells need enzymes called RNA polymeraseswhich transcribe stretches of DNA into RNA moleculesto be able to separate the copies of their chromosomes correctly . This suggests that RNAs transcribed from centromeres might be required for cell division , but the identity and function of these RNAs remained elusive . Quénet and Dalal have now discovered that an RNA polymerase localizes to the DNA in human centromeres and produces RNA molecules during the early stages of the cell cycle . Two proteins–one called CENP-A and another that functions as its chaperone–that normally bind to the centromere and determine its structure were found less often in this region of the chromosome if the activity of the RNA polymerase was inhibited . Qunet and Dalal identified a specific RNA molecule that is transcribed from the centromeric DNA , which directly binds to the CENP-A protein and its chaperone before CENP-A is assembled onto the centromeric DNA . Reducing the levels of this RNA within the cells made them unable to separate their chromosomes correctly during cell divisions . Qunet and Dalal also demonstrated that this centromeric RNA is needed to specifically target both the CENP-A protein , via its chaperone , to the centromere . The findings of Qunet and Dalal demonstrate that RNAs produced from a specific part of the chromosome can help target DNA-binding proteins back to that region's DNA sequence . Following on from this work , the next challenge will be to determine if other RNA molecules are used for the same purpose in humans and other species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | A long non-coding RNA is required for targeting centromeric protein A to the human centromere |
Recent studies indicate that soluble oligomers drive pathogenesis in several neurodegenerative proteinopathies , including Alzheimer and Parkinson disease . Curiously , the same conformational antibody recognizes different disease-related oligomers , despite the variations in clinical presentation and brain regions affected , suggesting that the oligomer structure might be responsible for toxicity . We investigated whether polyglutamine-expanded ATAXIN-1 , the protein that underlies spinocerebellar ataxia type 1 , forms toxic oligomers and , if so , what underlies their toxicity . We found that mutant ATXN1 does form oligomers and that oligomer levels correlate with disease progression in the Atxn1154Q/+ mice . Moreover , oligomeric toxicity , stabilization and seeding require interaction with Capicua , which is expressed at greater ratios with respect to ATXN1 in the cerebellum than in less vulnerable brain regions . Thus , specific interactors , not merely oligomeric structure , drive pathogenesis and contribute to regional vulnerability . Identifying interactors that stabilize toxic oligomeric complexes could answer longstanding questions about the pathogenesis of other proteinopathies .
The family of neurodegenerative diseases known as proteinopathies are characterized neuropathologically by deposits of insoluble proteins . In Alzheimer disease ( AD ) , for example , there are extracellular depositions of amyloid-β ( Aβ ) plaques and intracellular accumulations of neurofibrillary tangles of tau ( Glenner and Wong , 1984; Grundke-Iqbal et al . , 1986 ) . Similarly , in Parkinson disease ( PD ) , α-synuclein forms cytoplasmic inclusions known as Lewy bodies ( Lashuel et al . , 2013 ) . Although AD , PD , and other proteinopathies such as Huntington disease ( HD ) and the prion disorders are associated with different proteins possessing various structures , functions , and affecting different brain regions , they all involve the accumulation of β-sheet-rich entities , which suggests some commonality of pathogenic mechanism ( Glabe and Kayed , 2006; Knowles et al . , 2014 ) . Like HD , Spinocerebellar ataxia type 1 ( SCA1 ) is a polyglutamine disease that belongs to this larger class of proteinopathies . SCA1 is caused by the expansion of a CAG repeat that encodes for glutamine ( Q ) in ATAXIN-1 ( ATXN1 ) , a protein that is expressed throughout the brain . The Atxn1 knockin mouse model , which bears a 154 CAG repeat knocked into the murine Atxn1 locus , faithfully reproduces the SCA1 phenotype: progressive motor incoordination due to cerebellar degeneration , cognitive deficits , premature death , and degradation-resistant deposits ( nuclear inclusions , or NIs ) that contain mutant ATXN1 ( Watase et al . , 2002 ) . As with the stable fibrillar deposits first observed in AD over a hundred years ago , the prominence of these NIs led initially to the postulate that this material is the causative agent of disease ( Chiti and Dobson , 2006 ) . Yet the NIs develop primarily in neurons that escape degeneration , not in the cerebellar Purkinje cells ( PCs ) , which are the first to succumb to SCA1 pathology ( Watase et al . , 2002 ) . This curious observation led to the proposal that the ATXN1-containing NIs are not themselves toxic but rather might serve a protective role by sequestering the mutant protein ( Cummings et al . , 1998 , 1999 ) . Recent findings suggest a refinement to this hypothesis: it may be that the primary drivers of toxicity are metastable non-fibrillar species known as soluble oligomers ( Glabe , 2008; Benilova et al . , 2012; Krishnan et al . , 2012 ) . Although toxic oligomers have been identified in HD models and their modulation relates to beneficial outcomes ( Legleiter et al . , 2010; Sontag et al . , 2012 ) their specific role in disease progression in vivo remains unstudied . Furthermore , there are not studies regarding the role of binding partners of the disease-related proteins in the oligomerization process . The inverse correlation between NIs and neuronal integrity in SCA1 , however , lends appeal to the hypothesis that soluble oligomers , rather than fibrils per se , drive neurodegeneration in SCA1 . In this study we sought to determine if and how oligomeric forms of polyQ ATXN1 might contribute to the SCA1 disease state . We report the discovery of polyQ ATXN1 oligomers in the Atxn1 knockin mouse and demonstrate that these oligomers do indeed correlate with disease pathogenesis and motor dysfunction . We also show that polyQ ATXN1 oligomers seed the formation of new oligomers and demonstrate that Capicua ( CIC ) , a key native binding partner of ATXN1 , plays a pivotal role in the stabilization and regional toxicity of these oligomeric species .
In the absence of high-resolution structural data for oligomers , conformation-dependent antibodies can be used to distinguish between different types of amyloid structures by recognizing epitopes that are associated with specific aggregation states , independent of their amino acid sequences ( Kayed et al . , 2003 , 2010 ) . We used the conformational monoclonal anti-oligomer antibody F11G3 to detect ATXN1 oligomers in the Atxn1154Q/+ knockin mouse model . This antibody has been extensively characterized and compared to other anti-oligomer antibodies previously developed using similar methods ( Guerrero-Munoz et al . , 2014a , 2014b ) . Oligomers were apparent in cerebellar extracts of Atxn1154Q/+ but not in wild-type or Atxn1−/− mice ( Figure 1A ) . To confirm the anti-oligomeric nature of F11G3 , we pre-incubated the antibody with different types of oligomers prior to performing IF in brain sections from Atxn1154Q/+ mice . The results verified that F11G3 is indeed highly specific to an oligomeric conformation rather than an amino acidic sequence ( Figure 1—figure supplement 1 ) . Immunofluorescence ( IF ) against both ATXN1 and oligomers revealed substantial co-localization in the cerebellum ( Figure 1B ) . Immunoprecipitation of oligomers from the Atxn1154Q/+ cerebellum confirmed that these metastable entities are formed by ATXN1 ( Figure 1C ) . Atomic force microscopy ( AFM ) images show that these oligomers have an average height of 6 . 8 +/− 3 . 4 nm ( Figure 1D ) . 10 . 7554/eLife . 07558 . 003Figure 1 . ATXN1 oligomers are located in areas prone to SCA1 degeneration . ( A ) Western blot analysis of soluble fractions from cerebella shows the existence of amyloid oligomers exclusively in Atxn1154Q/+ mice model but in neither wild-type nor Atxn1−/− controls . ( B ) Immunofluoresence studies of Atxn1154Q/+ brain sections using anti-Ataxin-1 ( green ) and F11G3 ( red ) confirms the ATXN1 identity of the amyloid oligomers . Scale bar = 15 μm . ( C ) Western blot using anti-ATXN1 antibody showed that the isolated oligomers ( IP with F11G3 ) , were indeed ATXN1 oligomers . These ATXN1 oligomers were IP’d exclusively from Atxn1154Q/+ mouse cerebellum and not from Atxn1−/− controls . ( D ) AFM analysis showing brain-derived ATXN1 oligomers IP from Atxn1154Q/+ mouse using F11G3 . Scale bar 200 nm . ( E ) Double staining using the Purkinje cell ( PC ) marker , calbindin ( green ) and F11G3 ( red ) revealed that ATXN1 oligomers accumulate in PC dendrites before dendritic degeneration is observed ( top panel ) . With progression of dendritic degeneration , ATXN1 oligomers tend to accumulate in both the cytoplasm and nucleus ( middle and bottom panel ) . Scale bar 30 μm . No degeneration is represented by ( − ) , a low level of degeneration is represented by ( + ) , and advanced degeneration by ( ++ ) . Mice analyzed were 28 weeks old . ( F ) Histological staining for ATXN1 ( top panel ) and oligomers ( F11G3 , bottom panel ) in cerebellum and cortex of Atxn1154Q/+ ( left panels ) and WT mice ( right panels ) . Scale bar 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 00310 . 7554/eLife . 07558 . 004Figure 1—figure supplement 1 . ATXN1 oligomers in SCA1 mouse . Whenever the antibody F11G3 was preincubated with the antigen Aβ or IAPP oligomers , no immunolabeling was observed in Atxn1154Q/+ sections that were previously positive for oligomers . Double immunofluorescence between anti-ATXN1 antibody ( green ) and F11G3 ( red ) preincubated with recombinant oligomers show the positive signal only for ATXN1 . Double immunofluorescence between anti-ATXN1 antibody ( green ) and F11G3 ( red ) preincubated with recombinant monomers show positive signal for both antibodies . Scale bar 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 00410 . 7554/eLife . 07558 . 005Figure 1—figure supplement 2 . ATXN1 oligomers levels are higher in the cerebellum than in the cortex . Western blots using F11G3 confirmed higher levels of ATXN1 oligomers in the cerebellum ( Cer ) than in the cortex ( Cx ) in the Atxn1154Q/+ mouse . ATXN1 was detected using 11750 antibody; arrowhead corresponded to 154Q ATXN1 band . 2Q ATXN band ( * ) demonstrated equal amount of loading between the Cer and Cx sample . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 005 If oligomeric ATXN1 does indeed drive pathology , it should be most abundant in cerebellar Purkinje cells , the cells most vulnerable to SCA1 . Staining of PCs at various stages of neurodegeneration in the Atxn1154Q/+ mice revealed high levels of oligomers in the neurites prior to their degeneration; in later stages , oligomers accumulated in the soma ( Figure 1E ) . When we explored the regional localization of oligomeric forms of ATXN1 , we found them associated with neurons that are prone to degeneration ( e . g . , PCs ) but not those that are spared by the disease ( e . g . , cortical neurons; Figure 1F and Figure 1—figure supplement 2 ) . The presence of ATXN1 oligomers in more susceptible neurons suggests that these soluble aggregates are important in the neuropathology of SCA1 . We took advantage of the fact that the knockin mouse model reproduces the progressive phenotype of SCA1 ( Zoghbi and Orr , 2009 ) to investigate the relationship between oligomer levels and degree of motor impairment in the Atxn1154Q/+ mice . Motor incoordination in the rotarod assay becomes apparent in these mice as early as 5 weeks of age and worsens throughout their life span ( Watase et al . , 2002 ) . Accordingly , we tested the levels of ATXN1 oligomers in these mice at an early stage of disease ( 8 weeks of age ) , a middle stage ( 18 weeks ) , and a late stage ( 28 weeks ) . There was considerable variability among animals of the same age in both motor performance and oligomer abundance , but the levels of different molecular weight oligomers in the cerebellum clearly increased as the mice aged ( Figure 2A ) . The differences between individual mice , however , allowed us to search for relationships between motor impairment and different ATXN1 oligomers . We tested each mouse on the rotarod and observed a consistently strong inverse correlation between latency to fall and the level of ATXN1 oligomers ( Figure 2B ) , with oligomers over 245 KDa showing the greatest inverse correlation ( R = 0 . 89 ) . Importantly , no significant changes in monomeric ATXN1 levels ( 2Q or 154Q ) were observed in these mice ( Figure 2—figure supplement 1 ) . Thus , oligomer levels correlated with disease progression as assessed by motor incoordination . 10 . 7554/eLife . 07558 . 006Figure 2 . ATXN1 oligomers form high molecular weight complexes that correlate with motor impairment . ( A ) Western blot of cerebellar lysates of 8- , 18- and 28-week old Atxn1154Q/+ mice using F11G3 . Right panels indicate cropped regions of the blot with exposure adjusted to permit comparative quantification among mice of different ages . ( B ) Correlation plots between the different molecular weight oligomers ( Figure 1A ) and the latency to fall in the rotarod assay in Atxn1154Q/+ mice . Each black dot corresponds to one Atxn1154Q/+ mouse from 1A . ( C ) Size exclusion chromatography ( SEC ) from Atxn1154Q/+ mouse cerebellum probed for oligomers ( F11G3 , left panel ) and ATXN1 ( 11750 , right panel ) . Note that top panels reveal a higher exposure than the bottom panels . In corresponded to Input before fractionation . ( D ) Representative atomic force microscopy ( AFM ) pictures from fractions 7 , 8 , 11 and 12 ( Scale bar 140 nm ) . ( E ) MTT ( 3- ( 4 , 5-Dimethylthiazol-2-yl ) -2 , 5-Diphenyltetrazolium Bromide ) assay performed on cerebellar granule neurons . Cells grown 7 days in vitro ( DIV ) were treated with different fractions ( labeled below the x-axis ) and viability was measured 18 hr following treatment ( black bars ) . Gray bars denote viability following incubation of fraction 11 previously co-incubated with the indicated antibodies . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 00610 . 7554/eLife . 07558 . 007Figure 2—figure supplement 1 . Correlation between the levels of 154Q or 2Q ATXN1 and the latency to fall from the rotarod . ( A ) A slight decrease in the level of 154Q ATXN1 was observed with age in Atxn1154Q/+ samples , but no changes were observed in the levels of 2Q ATXN1 . The antibody 11750 was used to detect ATXN1 in western blot . ( B ) No correlation was observed between the levels of 154Q or 2Q ATXN1 and the latency to fall from the rotarod . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 00710 . 7554/eLife . 07558 . 008Figure 2—figure supplement 2 . ATXN1 oligomers characterization by hydrophobic interaction column . Fraction 11 of the SEC from Atxn1154Q/+ mouse cerebellum was passed through a hydrophobic interaction column ( HIC ) . ATXN1 oligomer complexes were detected in fractions with high levels of exposed hydrophobic domains . F11 corresponded to Fraction 11 input; FT corresponded to the HIC flow through; ( − ) and ( + ) indicates lower and higher levels of hydrophobic fractions , respectively . Oligomers were detected with F11G3 and ATXN1 was detected with 11750 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 00810 . 7554/eLife . 07558 . 009Figure 2—figure supplement 3 . ATXN1 oligomers from Atxn1154Q/+ are highly stable . Dot-blot probed with F11G3 of SEC fractions from Atxn1154Q/+ cerebellum incubated at 37°C for 12 , 24 and 48 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 00910 . 7554/eLife . 07558 . 010Figure 2—figure supplement 4 . ATXN1 oligomeric complex induce cellular toxicity . Cellular staining of CGPC cells after fraction 11 treatment using Tuj1 antibody: the ATXN1 oligomer complex induces degeneration in cells . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 010 To determine the native size of soluble ATXN1 oligomers under non-denaturating conditions to insure that they were not artificially generated during SDS-PAGE , we performed non-denaturating size-exclusion chromatography ( SEC ) of cerebellar samples and subjected these to SDS-Tris-Glycine gels ( Figure 2C ) . Evaluation of individual fractions by western blot with anti-ATXN1 antibody revealed WT and polyQ monomeric ATXN1 in different fractions , as previously shown ( Figure 2C , right panel ) ( Lam et al . , 2006 ) . When we analyzed the SEC fractions with the anti-oligomer antibody , we observed oligomeric species mainly between fractions 10 and 12 ( ranging from 500 to 1000 KDa in size; Figure 2C , left panel ) . Anti-ATXN1 antibody confirmed that the oligomers in these fractions do in fact contain ATXN1 ( Figure 2C , right panel ) ; the range of sizes of these oligomers is thus rather broad . Fraction 11 ( ∼667 KDa ) contained a peak in the signal of oligomers >245 KDa in weight , as well as the presence of lower molecular weight oligomeric species ( ∼180 , ∼135 , ∼110 and ∼98 KDa ) . AFM revealed that those fractions detected by the anti-oligomer antibody ( particularly fractions 11 and 12 ) were indeed oligomeric , whereas the higher molecular weight fractions were fibrillar in nature ( fraction 7 , ∼4000KDa; Figure 2D ) . A hydrophobic interaction column , which separates proteins according to hydrophobic domain exposure , showed that fraction 11 is highly hydrophobic , as is characteristic of many oligomers ( Campioni et al . , 2010 ) ( Figure 2—figure supplement 2 ) . To establish how stable these oligomers are , SEC fractions were incubated at 37°C for up to 48 hr followed by immunoblot analysis using F11G3 . F11G3 showed a strong affinity for fractions 11 and 12 even after 48 hr of incubation , suggesting that these oligomers are highly stable ( Figure 2—figure supplement 3 ) . To determine whether the ATXN1 oligomeric complex can induce cellular toxicity , we treated cultured cerebellar granular neurons ( CGNs ) with SEC fractions from Atxn1154Q/+ cerebellar samples . Cell viability measurement revealed fraction 11 to be highly toxic ( Figure 2E ) . This toxicity was blocked when fraction 11 was pre-incubated with the anti-oligomer antibody F11G3 or anti-ATXN1 antibody . The cellular damage was thus produced by an ATXN1-containing oligomeric complex present in this SEC fraction . IF staining of CGNs with the neuronal marker Tuj1 revealed that the ATXN1 oligomer complex from fraction 11-induced neurodegeneration , as evidenced by neuritic swelling and beading ( Figure 2—figure supplement 4 ) . Neuropathological analysis of the Atxn1154Q/+ mice revealed that ATXN1 oligomers are not evenly distributed throughout the cerebellum but instead are restricted to focal sub-populations of PCs . It is notable that this focal distribution coincided with cellular toxicity: cells with no detectable oligomers looked healthy , displaying many calbindin-positive PCs with extensive projections , while areas with abundant oligomers showed more diffuse calbindin staining and short dendrites ( Figure 3A ) . Given that , in AD and PD , proteins such as Aβ , tau or α-synuclein appear to ‘seed’ the accumulation of other neurotoxic proteins and promote neurodegeneration ( Clavaguera et al . , 2009; Langer et al . , 2011; Lasagna-Reeves et al . , 2012; Rey et al . , 2013 ) , we sought to determine whether a similar mechanism operates in SCA1 . 10 . 7554/eLife . 07558 . 011Figure 3 . ATXN1 oligomer complexes from Atxn1154Q/+ mouse cerebellum induce the formation of new ATXN1 oligomers . ( A ) Histological analysis in Atxn1154Q/+ brain sections reveals the regional presence of oligomers in PCs stained with calbindin ( top panel ) and oligomers stained with F11G3 ( bottom panel ) in degenerating PCs ( bottom right panel ) but not surviving PCs ( bottom left panel ) . ( B ) Cell-based seeding assay: mRFP-ATXN1 ( 82Q ) cells were incubated with the indicated fractions ( Frac 7–14 ) and the oligomeric and fibrillar inclusions were quantified . *p < 0 . 05 . **p < 0 . 01 . ( C ) Immunostaining for oligomers ( F11G3 , green ) of mRFP-ATXN1 ( 82Q ) cells following treatment with fraction 11 . Arrowheads denote internalized oligomers . Scale bar 15 μm . ( D ) Immunostaining for ATXN1 ( 11750 , green signal ) of mRFP-ATXN1 ( 82Q ) after treatment with fraction 11 from Atxn1154Q/+ cerebellum ( top row ) demonstrated that exogenous ATXN1 ( green signal ) is internalized into the cells . Cells treated with fractions 11 from wild-type mouse cerebellum ( bottom row ) did not show internalized ATXN1 entities . Scale bar 15 μm . ( E ) Quantification of the percentage of mRFP-ATXN1 ( 82Q ) cells with oligomeric ( blue bars ) or fibrillar ( red bars ) inclusions following treatment of fraction 11 pre-incubated with the indicated antibodies *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 01110 . 7554/eLife . 07558 . 012Figure 3—figure supplement 1 . ATXN1 oligomers are detected in mRFP-ATXN1 ( 82Q ) cells . Immunostaining for oligomers ( F11G3 , green ) in a stable cell line expressing mRFP-ATXN1 ( 82Q ) ( top panels ) or a similar stable cell line expressing 82Q ATXN1 with the S776A mutation ( bottom panels ) . The S776A mutation prevents formation of Nis and oligomers . Scale bar 15 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 01210 . 7554/eLife . 07558 . 013Figure 3—figure supplement 2 . Anti-oligomer antibody F11G3 does not get internalized into the cells . ( A ) Daoy mRFP-ATXN1 ( 82Q ) stable cell lines were treated with SEC fraction 11 from Atxn1154Q/+ cerebellum pre-incubated with F11G3 for 10 hr . After fixation cells were stained with a secondary anti-mouse IgM alexa 488 ( green signal ) . No intracellular signal was observed in any cell . Low extracellular signal was observed ( arrowhead ) . Scale bar 15 μm . ( B ) Cell media was analyzed by western blot using secondary anti-mouse IgM . Heavy and light chains from IgM were observed in treated samples but not in untreated control . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 013 To evaluate the potential seeding activity of ATXN1 oligomer complexes , we used a previously described cellular model that expresses ATXN1 ( 82Q ) fused to monomeric red fluorescent protein ( mRFP ) ( Park et al . , 2013 ) . These cells form nuclear inclusions , but their capacity to form oligomers had never been evaluated . We first established that these cells do form oligomers ( Figure 3—figure supplement 1 ) . As a negative control , we used cells that express mRFP-ATXN1 ( 82Q , S776A ) , a mutant form of ATXN1 that cannot be phosphorylated at Ser776 and is not toxic ( Emamian et al . , 2003; Jorgensen et al . , 2009; Park et al . , 2013 ) . The anti-oligomer antibody detected the smaller inclusion bodies ( 350–900 nm ) but did not recognize the larger inclusions ( >900 nm ) , suggesting that these larger inclusions are more complex and composed of higher order aggregates , likely to be fibrillar in nature . We next tested whether ATXN1 oligomer complexes from Atxn1154Q/+ mouse cerebellum seed the formation of new oligomers in this cell model . We added cerebellar fractions ( see Figure 2C ) to the cells and quantified the percentage of mRFP-ATXN1 ( 82Q ) cells with oligomeric or fibrillar inclusions , based on size ( 350–900 nm or >900 nm , respectively ) and whether or not the inclusions were detected by F11G3 . The cells exposed to either fraction 11 or 12 developed the largest proportion of oligomeric inclusions ( 55% and 34% , respectively ) in comparison with cells incubated with other fractions or the control group ( 18% ) ( Figure 3B ) . None of the treated groups showed an increase in fibrillar inclusions . This suggests that ATXN1 oligomeric complexes seed the formation of new oligomers but do not promote the formation of ATXN1 fibrillar material . No fibrillar or oligomeric inclusions were observed in any of the mRFP-ATXN1 ( 82Q , S776A ) cells , regardless of the treatment ( data not shown ) . ATXN1 oligomeric complexes may thus act as a seed , but the cell needs to express an entity that is prone to aggregate . To confirm that the seeding effect was in fact produced by ATXN1 oligomers , Daoy mRFP-ATXN1 ( 82Q ) cells treated with Fraction 11 were immunostained with either anti-oligomer antibody ( Figure 3C ) or an anti-ATXN1 antibody ( Figure 3D ) . The anti-oligomer staining ( green fluorescence ) revealed several oligomeric complexes that lacked an mRFP signal , suggesting that exogenously added oligomers were internalized by the cell ( Figure 3C ) . ATXN1 staining revealed that cells also internalized exogenously added ATXN1 species ( Figure 3D ) . We next treated Daoy mRFP-ATXN1 ( 82Q ) cells with fraction 11 pre-incubated with the anti-oligomer antibody , the anti-ATXN1 antibody , or a control antibody . Cells treated with fraction 11 pre-incubated with the anti-oligomer or the anti-ATXN1 antibody blocked oligomer formation ( Figure 3E ) . It is therefore the ATXN1 oligomers present in fraction 11 that are responsible for inducing the formation of new oligomeric inclusions in mRFP-ATXN1 ( 82Q ) cells . To determine the mechanism through which the anti-oligomer antibody blocks the seeding effect of ATXN1 oligomers , we treated Daoy mRFP-ATXN1 ( 82Q ) cells with fraction 11 pre-incubated with the anti-oligomer antibody and performed IF using only a secondary antibody labeled with alexa 488 . Fluorescent images show that the anti-oligomer antibody was not internalized by the cell but bound to the ATXN1 oligomers in the extracellular space . The antibody thus appears to block the internalization of exogenous ATXN1 oligomers rather than interacting directly with intracellular oligomeric species . The presence of anti-oligomer antibody was confirmed by western blot analysis of media from cells treated with only secondary antibody ( Figure 3—figure supplement 2 ) . In SCA1 , both wild-type and expanded ATXN1 share many binding partners ( Lim et al . , 2006 ) . Among these , the transcriptional repressor CIC is of particular interest , because it is the only known binding partner whose protein levels are significantly reduced in Atxn1−/− mice ( Lee et al . , 2011 ) ; its stability is thus dependent on being complexed with ATXN1 ( Lam et al . , 2006 ) . Moreover , our lab has shown that polyQ ATXN1 exerts toxicity through its native CIC-containing complex rather than through aberrant interactions with novel proteins ( Lam et al . , 2006 ) , and that reducing CIC levels by half rescues many SCA1-like phenotypes in Atxn1154Q/+ mice ( Fryer et al . , 2011 ) . When we performed SEC on cerebellar extracts from Atxn1154Q/+ mice , we detected CIC in the same fractions enriched with ATXN1 oligomer complexes . These fractions lost their seeding capacity when pre-incubated with an anti-CIC antibody ( Figure 4—figure supplement 1 ) . Based on these observations , we investigated the role of CIC in the formation and stability of ATXN1 oligomer complexes . We generated Atxn1154Q/+;CicL+/- mice that have a ∼50% reduction of both isoforms of CIC ( Cic-L and Cic-S ) ( Fryer et al . , 2011 ) and observed a significantly lower number of PCs positive for oligomers in these mice ( Figure 4A , B ) . This result was confirmed by western blot analyses on whole cerebellar extracts from these animals ( Figure 4C , D ) . 10 . 7554/eLife . 07558 . 014Figure 4 . Reduction of CIC levels decreases the levels of ATXN1 oligomer complexes in vivo . ( A–B ) Pathological analysis showed fewer PCs from Atxn1154Q/+;CicL+/- mice loaded with oligomers than PCs from Atxn1154Q/+ mice . PCs were stained with calbindin antibody and oligomers with F11G3 . *p < 0 . 05 . ( C–D ) Western blot revealed lower levels of oligomers in the cerebellum of Atxn1154Q/+;CicL+/- mice than in Atxn1154Q/+ mice . Oligomers were measured using F11G3 . **p < 0 . 01 . ( E–F ) Size Exclusion Chromatography ( SEC ) showed that ATXN1 complexes in cerebellar samples from Atxn1154Q/+;CicL+/- mice shift to a higher molecular weight fraction than those in Atxn1154Q/+ cerebella . ( G ) Cell-based seeding assay demonstrated that the absence of one Cic allele in Atxn1154Q/+;CicL+/- mice reduced the formation of new ATXN1 oligomers in Daoy mRFP-ATXN1 ( 82Q ) . Fraction 11 and 12 ATXN1 oligomer complexes from control Atxn1154Q/+ mouse cerebella seeded the formation of new mRFP-ATXN1 ( 82Q ) . *p < 0 . 05 , **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 01410 . 7554/eLife . 07558 . 015Figure 4—figure supplement 1 . CIC is part of the ATXN1 oligomeric complex . ( A ) Size exclusion chromatography ( SEC ) on Atxn1154Q/+ mouse cerebellar soluble fraction revealed that CIC is part of a higher molecular weight oligomeric complex detected with F11G3 ( fractions 10 and 11 ) . ( B ) Pre-incubation of fraction 11 from Atxn1154Q/+ cerebellum with anti-CIC antibody blocked the internalization and seeding capacity of oligomer complexes . Pre-incubation of fraction 11 with an anti-ATXN1-like antibody ( ATXN1-like is a paralog of ATXN1 that lacks the polyQ tract ) failed to block the seeding capacity of ATXN1 oligomer complexes . Ctl = control antibody . *p < 0 . 05 , **p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 01510 . 7554/eLife . 07558 . 016Figure 4—figure supplement 2 . Higher molecular weight complexes in Atxn1154Q/+;CicL+/- do not present an oligomeric conformation . SEC revealed no shift of ATXN1 oligomeric complexes in Atxn1154Q/+;CicL+/- mouse cerebellum compared with Atxn1154Q/+ mice . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 01610 . 7554/eLife . 07558 . 017Figure 4—figure supplement 3 . High molecular weight complexes characterization by filter retardation assay ( FRA ) . ATXN1 insoluble aggregates were detected on Atxn1154Q/+ mouse cerebellar soluble fractions 7 and 8 . In Atxn1154Q/+;CicL+/- mouse cerebellum ATXN1 insoluble aggregates were observed in fraction 7 , 8 and 9 . Fibrillar oligomers detected by OC , are observed in fraction 7 and 8 from Atxn1154Q/+ mouse and fractions 7 , 8 and 9 from Atxn1154Q/+;CicL+/- mouse . No oligomers detected by F11G3 were observed in any of the insoluble fractions . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 017 To our surprise , SEC on cerebellar extracts from Atxn1154Q/+;CicL+/- and Atxn1154Q/+ mice revealed that decreasing the levels of CIC induced a shift of the ATXN1 complexes to a higher molecular weight fraction detected by ATXN1 antibody ( Figure 4E , F ) . Furthermore , these higher molecular weight complexes are not oligomeric: anti-oligomer antibody did not reveal any shift to the higher molecular weight fractions upon loss of Cic ( Figure 4—figure supplement 2 ) . To confirm that the high molecular weight complex present only in Atxn1154Q/+;CicL+/- mice adopted a more organized fibrillar aggregate conformation , we performed Filter Retardation Assay ( FRA ) using the SEC fractions from Atxn1154Q/+;CicL+/- and Atxn1154Q/+ mice . Probing the FRA membrane with anti-ATXN1 antibody revealed the presence of ATXN1 SDS-insoluble aggregates in fractions 7 and 8 in Atxn1154Q/+ mice . In the case of the Atxn1154Q/+;CicL+/- mice , ATXN1 SDS-insoluble aggregates were also present in fraction 9 ( Figure 4—figure supplement 3 ) , confirming that in the context of reduced CIC levels , ATXN1 tends to form more organized fibrillar aggregates . We also probed FRA membranes with OC , a conformational antibody that specifically recognizes fibrillar oligomers and fibrils ( Kayed et al . , 2007 ) , and found fractions 7 and 8 in Atxn1154Q/+ mice , and fractions 7 , 8 , and 9 in Atxn1154Q/+;CicL+/- mice , positive for OC ( Figure 4—figure supplement 3 ) . ATXN1 thus tends to form fibrillar oligomers , amorphous aggregates and/or fibrils when CIC levels are reduced . We detected no oligomers using F11G3 in the FRA membranes; ATXN1 oligomers detected by F11G3 are therefore soluble in 2% SDS ( Figure 4—figure supplement 3 ) and able to escape detection by FRA . To determine whether the ATXN1 complex maintains its seeding capacity in the absence of one Cic allele , we performed the cell-based seeding assay using mRFP-ATXN1 ( 82Q ) cells treated with SEC fractions from either Atxn1154Q/+ or Atxn1154Q/+;CicL+/- cerebellum and quantified the percentage of cells containing fibrillar or oligomeric inclusions . Fractions 11 and 12 from the Atxn1154Q/+;CicL+/- mice were less proficient in inducing the formation of new oligomeric inclusions than those from the Atxn1154Q/+ mice ( Figure 4G ) . To determine whether CIC plays a key role in ATXN1 oligomerization , we co-transfected Hela cells with a consistent amount of ATXN1 ( 82Q ) and varying quantities of CIC . As a negative control , we utilized CIC with the W37A mutation which hinders its interaction with ATXN1 ( Kim et al . , 2013 ) . Increasing amounts of CIC led to increasing levels of oligomers as measured with F11G3 ( Figure 5A , B ) , but increasing amounts of mutant CIC/W37A produced no changes in the levels of ATXN1 oligomers in the cells . To determine whether polyQ-length plays a role in CIC's stabilization of ATXN1 oligomers , we repeated the same co-transfection experiments in HeLa cells as described for ATXN1 ( 82Q ) , this time utilizing nonpathogenic wild-type ATXN1 ( 30Q ) . When cells were co-transfected with ATXN1 ( 30Q ) and increasing amounts of CIC , oligomers were almost undetectable by western blot using F11G3 ( Figure 5C ) . As expected , increasing amounts of mutant CIC/W37A produced no oligomeric signal . To compare the relative amounts of oligomers in the ATXN1 ( 82Q ) and ATXN1 ( 30Q ) cells , we performed direct ELISA using F11G3 and normalized the values by the total amount of ATXN1 measured using anti-ATXN1 antibody . The ELISA quantification confirmed that increasing amounts of CIC corresponded to increasing levels of oligomers in ATXN1 ( 82Q ) cells , whereas no changes in oligomer levels were observed when cells were treated with any amount of mutant CIC/W37A . Cells transfected with nonpathogenic ATXN1 ( 30Q ) failed to produce oligomers , whether they were co-transfected with CIC , CIC/W37A , or not co-transfected at all ( Figure 5D ) . 10 . 7554/eLife . 07558 . 018Figure 5 . ATXN1 requires CIC binding to maintain oligomer conformation . ( A ) Western blot of Hela cells co-transfected with ATXN1 ( 82Q ) and increasing concentrations ( from 0 . 5 to 3 μg ) of wild type CIC ( left panel ) or increasing concentrations of mutant CIC/W37A ( right panel ) . Membranes were probed with F11G3 to determine the levels of oligomers . * indicated change in exposure of the membrane . ( B ) Quantification of oligomers from Western blot analysis ( A ) , demonstrated that increased oligomer levels correlated with increases of transfected wild-type CIC . Mutant CIC/W37A produced no effect on the amount of oligomers detected by Western blot . n = 4 . *p < 0 . 05 , **p < 0 . 01 and ***p < 0 . 001 . ( C ) Western blot of Hela cells co-transfected with ATXN1 ( 30Q ) and increasing concentrations ( from 0 . 5 to 3 μg ) of wild-type CIC ( left panel ) or increasing concentrations of mutant CIC/W37A ( right panel ) . Membranes were probed with F11G3 to determine the levels of oligomers . Almost no oligomers were detected in any of the measured conditions . * indicated change in exposure of the membrane . ( D ) ELISA using F11G3 shows that cells transfected with ATXN1 ( 82Q ) have high levels of oligomers , but cells transfected with ATXN1 ( 30Q ) had almost no detectable oligomers . Increasing amounts of oligomers were observed when cells transfected with ATXN1 ( 82Q ) were co-transfected with increasing amounts of wild type CIC but not mutant CIC/W37A . n = 4 . *p < 0 . 05 , **p < 0 . 01 and ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 01810 . 7554/eLife . 07558 . 019Figure 5—figure supplement 1 . CIC facilitates ATXN1 oligomerization . ( A–B ) Daoy mRFP-ATXN1 ( 82Q ) stable cell lines transfected with the N-terminal fragment of CIC showed a higher number of oligomeric inclusions than cells transfected with the N-terminal fragment of mutant CIC W37A or control ( non-transfected ) cells . *p < 0 . 05 . Scale bar 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 019 We also transiently transfected mRFP-ATXN1 ( 82Q ) stable cells with the N-terminal region of CIC , which contains the ATXN1 binding region ( Lam et al . , 2006 ) . As a negative control , we transfected N-terminal CIC with the W37A mutation . Cells transfected with N-terminal wild-type CIC developed many more oligomeric inclusions than those transfected with mutant CIC or controls ( Figure 5—figure supplement 1 ) . This effect was specific to oligomeric inclusions . We next studied the effect of CIC on the aggregation kinetics of ATXN1 in vitro . Because full-length ATXN1 is insoluble and difficult work with in vitro ( de Chiara et al . , 2005; Lam et al . , 2006; Goldschmidt et al . , 2010 ) , we used ATXN1's AXH domain , which is the region that interacts with CIC and is also responsible for a large part of the protein's toxicity in mice and fruit flies ( Tsuda et al . , 2005 ) and its tendency to aggregate ( Figure 6—figure supplement 1 ) . When the AXH peptide alone was stirred at 37°C , soluble oligomers formed in as little as 12 hr but were no longer detectable after 48 hr . When AXH was mixed with CIC in a 1:1 molar ratio , however , oligomeric species were more abundant and persisted much longer , being detectable for up to 72 hr ( Figure 6A ) . When we blocked AXH-CIC binding ( using CIC W37A ) , AXH oligomer kinetics resembled those of AXH alone ( Figure 6A ) . 10 . 7554/eLife . 07558 . 020Figure 6 . CIC stabilizes ATXN1 oligomers . ( A ) ELISA using anti-prefibrillar oligomer antibody A-11 shows that the N-terminal fragment of wild-type CIC stabilized AXH fragments in the oligomeric conformation , unlike the N-terminal fragment of mutant CIC W37A or AXH incubated alone . Wild-type and mutant CIC alone did not oligomerize at any time point . ( B ) ELISA using anti-fibrillar oligomer antibody OC shows that only wild type , not mutant , CIC partially inhibits the formation of fibrillar oligomers . ( C ) Thioflavin T assay demonstrated that wild-type but not mutant CIC blocks the formation of AXH fibrils . *p < 0 . 05 , **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 02010 . 7554/eLife . 07558 . 021Figure 6—figure supplement 1 . The AXH domain of ATXN1 has a high energy for aggregation . 3D profile method for predicting fibrillizing segments . The top and bottom panels show the predicted energy for fibrillation of every six-residue segment of AXH and α-Synuclein , respectively . Red histogram bars represent hexapeptides with energy below −23 kcal∕mol , which are predicted to form fibrils . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 021 To better characterize the effect of CIC on the aggregation kinetics of ATXN1 , we utilized the conformational antibody OC that specifically detects fibrillar oligomers and fibrils ( Kayed et al . , 2007 ) . It has been shown that these fibrillar oligomers are more mature oligomers , more compact and less toxic than the soluble ones detected with A-11 and F11G3 , known as prefibrillar oligomers ( Krishnan et al . , 2012; Guerrero-Munoz et al . , 2014b ) . Despite the structural similarity of fibrillar oligomers and fibrils , the former are not positive for Thioflavine ( Glabe , 2008 ) . As expected , AXH either alone or with mutant CIC W37A formed significantly more fibrillar oligomers at 24 hr than it did with wild-type CIC ( Figure 6B ) . This result suggests that CIC binding to ATXN1 promotes the formation of and stabilizes soluble prefibrillar oligomers and tends to divert ATXN1 from forming more mature and less toxic aggregates such as fibrillar oligomers and fibrils . To solidify this interpretation , we performed a Thioflavin T ( Thio-T ) assay to determine the effect of CIC on the formation of AXH fibrils . Fluorescence measurements of Thio-T showed that wild-type CIC diminished the formation of AXH fibrils ( Figure 6C ) . The results obtained from our aggregation kinetics assay suggested that CIC not only promotes AXH oligomerization but also stabilizes the oligomers , retarding the kinetics of AXH aggregation to fibrils . Having demonstrated that increasing levels of CIC stabilize ATXN1 polyQ oligomers in cells ( Figure 5 ) , we sought to determine whether the particular susceptibility of the cerebellum to ATXN1 toxicity could be attributable , at least in part , to CIC levels in the tissue . We measured the levels of ATXN1 and CIC in the cerebellum and cortex of wild type mice and found that , under these physiological conditions , CIC levels are higher in the cerebellum than in the cortex , but ATXN1 levels are lower ( Figure 7 ) . The amount of CIC available relative to ATXN1 is thus considerably higher in the cerebellum than in the cortex . This result suggests that in brain regions containing high levels of CIC in relation to ATXN1 ( such as the cerebellum ) , CIC stabilizes ATXN1 in its oligomeric conformation , but where the ratio of CIC to ATXN1 is lower , ATXN1 is not stabilized in its oligomeric confirmation and continues its process of forming non-toxic NIs . 10 . 7554/eLife . 07558 . 022Figure 7 . Comparison of CIC and ATXN1 protein levels in cerebellum and cortex . CIC levels are higher in the cerebellum than in the cortex , while ATXN1 levels are lower in the cerebellum than in the cortex , making the amount of CIC per ATXN1 much higher in the cerebellum . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 022
An overwhelming number of studies have shown that structurally related oligomers are toxic ( Chiti and Dobson , 2006 ) and that the soluble rather than the fibrillar mutant protein is associated with a toxic gain of function ( Saunders and Bottomley , 2009; Zoghbi and Orr , 2009 ) . Oligomers have been specifically associated with Alzheimer disease , Parkinson disease , and frontotemporal dementia , and it has been proposed that the disease pathology could be caused solely by soluble toxic entities of Aβ , α-synuclein or tau , respectively ( Lesne et al . , 2006; Berger et al . , 2007; Shankar et al . , 2008 ) . Yet the molecular mechanism ( s ) of this toxicity have remained poorly understood . Not since Prusiner proposed the existence of a ‘protein X’ that might help form the toxic species has attention been paid to a possible requirement for interaction with another protein partner to form toxic oligomers ( Telling et al . , 1995 ) . Here we have demonstrated not only that polyQ ATXN1 adopts an oligomeric conformation and that this conformation is pathogenic in cellular and animal models of SCA1 , but that the native binding partner of ATXN1 , CIC , is essential for the stabilization of oligomeric polyQ ATXN1 . We believe the data provide compelling reason to investigate the role of native partners involved in the toxicity of protein oligomers in other neurodegenerative proteinopathies . Our study also provide an answer to the longstanding conundrum of differential regional vulnerability , i . e . , why a ubiquitiously expressed protein should prove toxic to only certain brain regions and leave others relatively unharmed . It has been particularly perplexing that regions that develop insoluble nuclear inclusions , which were long thought to be toxic , tend to escape SCA1 pathology ( Watase et al . , 2002 ) . We had previously shown that polyQ ATXN1 exerts toxicity through its native complexes containing CIC rather than through aberrant interactions with novel proteins ( Lam et al . , 2006 ) and that reducing CIC levels by 50% rescues the phenotype of a mouse model for SCA1 , improving motor functions and increasing survival ( Fryer et al . , 2011 ) . The data from the current study provide insight into how CIC modulates pathogenesis and why it is particularly toxic in the cerebellum: CIC stabilizes toxic oligomers and exists in a particularly high ratio with respect to ATXN1 in cerebellar neurons . In the SCA1 knockin mice , the abundance of these oligomers correlated directly with behavioral deficits; without CIC , in vitro and in vivo , ATXN1 oligomers continue on the aggregation pathway and co-assemble into less-toxic fibrillar oligomers , amorphous aggregates and/or fibrils ( Figure 8 ) . SCA1 is therefore triggered not just by a conformational change in ATXN1 but by interactions with a specific partner in a specific brain region that brings about oligomerization and stabilization . The partner not only has to be present , but its levels have to be high enough to stabilize a considerable proportion of ATXN1 oligomers in order to drive pathogenesis . These observations do much to explain why pathology is relatively localized in SCA1 despite the ubiquitous expression of ATXN1 . It seems likely that the physiological cellular environment will prove to be the determining factor in promoting and stabilizing the respective pathological proteins in their toxic conformations in other polyglutamine diseases and proteinopathies . 10 . 7554/eLife . 07558 . 023Figure 8 . ATAXIN-1 oligomer complexes with seeding capability correlate with disease progression in spinocerebellar ataxia type 1: a model . Under normal conditions ( A , left panel ) , wild type ATXN1 forms a transcriptional repressor complex with CIC ( 1 ) and binds to DNA ( 2 ) . Under pathological conditions ( B , right panel ) , polyQ-expanded ATXN1 accumulates and can directly form nuclear inclusions ( 1 ) or it can adopt an oligomeric conformation ( 2 ) . These oligomers form nuclear inclusions ( 3 ) or form a stable oligomeric complex with CIC ( 4 ) that could: act as a dysfunctional transcriptional repressor complex ( 5 ) or be released into the extracellular space and be internalized by neighboring neurons , seeding the formation of new oligomers ( 6 ) . The toxic effect of oligomeric complexes might be mitigated by reducing the total levels of polyQ ATXN1 or hindering the propagation of these toxic complexes ( e . g . , by immunotherapy ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07558 . 023 Amyloid oligomers are characterized by hydrophobic domains; ATXN1 dimers form a heterotetramer with CIC dimers by hydrophobic interactions ( Bemporad and Chiti , 2012; Kim et al . , 2013 ) . In light of this , we propose that CIC interacts with oligomeric polyglutamine-expanded ATXN1 dimers through the exposed hydrophobic domains , which in turn stabilizes the complexes . In other words , CIC—a native ATXN1 interactor—does not discern between a nonpathogenic wild type ATXN1 dimer and a pathologically expanded ATXN1 dimer , and interacts with either; it simply binds to ATXN1 to form a functional transcriptional repressor complex or a dysfunctional oligomeric complex , depending on the polyglutamine tract . The length of the polyQ tract is essential to the accumulation , aggregation and oligomerization of ATXN1 , as has been postulated for other polyglutamine diseases ( Legleiter et al . , 2010 ) . Given that CIC interacts with ATXN1 regardless of polyQ expansion ( Lam et al . , 2006 ) and that CIC stabilizes AXH oligomers , we believe that CIC stabilizes ATXN1 oligomers independent of the polyQ tract length . Nevertheless , a polyglutamine expansion is necessary to initiate the oligomerization process under physiological conditions , in the context of the full-length protein . To our knowledge , this is the first evidence that a native interactor of a toxic protein can stabilize its oligomeric form to promote its bioavailability and downstream toxicity . Furthermore , ATXN1 oligomeric complexes were able to penetrate cells and seed the formation of new ATXN1 oligomers without forming fibrillar inclusions . This could be explained by the addition of monomers to the ends of oligomers , which is kinetically more favorable than the assembly of oligomers into fibrils via sheet stacking interactions ( Wu et al . , 2010 ) . Our findings for SCA1 are likely to be applicable to other members of the larger family of neurodegenerative proteinopathies . It should be worthwhile , for example , to explore whether native interactors of tau and α-synuclein might promote or hinder their formation of toxic oligomers . Without ruling out the role of non-native interactors in protein aggregation , we propose that neuropathological features of a specific disease are determined by both transient and stable native protein interactions and that these interactions in turn affect the amyloidogenic properties of the disease-associated protein . Thus , blocking certain interactions that mediate the formation or stabilization of oligomeric toxic entities , or modulating the downstream effects of such interactions , could be an attractive treatment option for this class of diseases .
All mouse procedures were approved by the Institutional Animal Care and Use Committee for Baylor College of Medicine and Affiliates . Atxn1154Q/+ , and Atxn1−/− mice have been previously described ( Lorenzetti et al . , 2000; Watase et al . , 2002 ) and were backcrossed to C57BL/6 for more than ten generations . The Atxn1154Q/+;CicL+/- mouse was previously generated and characterized ( Fryer et al . , 2011 ) . Briefly , male Atxn1154Q/+ animals were crossed with female CicL+/− to obtain the following genotypes: wild type , CicL+/− , Atxn1154Q/+ , and Atxn1154Q/+;CicL+/- . Twenty-four weeks-old ( unless stated otherwise ) mouse cerebella were dissected and then lysed in 0 . 5% Triton buffer ( 0 . 5% triton X-100 , 50 mM Tris pH 8 , 75 mM NaCl ) supplemented with protease and phosphatase inhibitors ( Sigma , St . Louis , MO ) . The protein lysate was then incubated on ice for 20 min and centrifuged at 13 , 200 r . p . m . for 10 min at 4°C , and the supernatants were portioned into aliquots , snap-frozen , and stored at −80°C until used . Size exclusion chromatography ( SEC ) was performed as previously described ( Park et al . , 2013 ) . Briefly , 700 μl of triton soluble sample was applied to a Superose 6 GL 300 column ( GE Healthcare and Life Science , United Kingdom ) and run at 0 . 3 ml/min using a 0 . 1% triton column buffer . Fractions were collected every 1 ml . A stable Daoy mRFP-ATXN1 ( 82Q ) cell line was generated as previously described ( Park et al . , 2013 ) . Cells were plate in 24 well plates ( 2*104 cells/ml ) . 24 hr later , cells were treated with SEC fractions ( 1 . 5 μg of total protein ) from mouse cerebellum . After 10 hr of treatment , cells were fixed with methanol at −80°C for 45 min . RFP-positive inclusions ranging from 350 to 900 nm were considered oligomeric; inclusions larger than 900 nm and not detected by F11G3 were considered fibrillar . One hundred cells were quantified per group in triplicates . Analyses were manually performed with Image J . For blocking assays , each fraction was mixed with each antibody ( 2 . 5 mg/ml ) in a ratio 1:1 ( vol/vol ) for 1 hr and then added to the cells . Motor coordination was assessed on the Rotarod assay as previously described ( Park et al . , 2013 ) , with four trials for 4 days using 8- , 18- and 28-week-old mice for the correlation measurements . The tester was blinded to animal genotype and treatment . Triton-soluble fractions of brain extracts were run on Tris-Glycine 5% gels , in nonreducing conditions to avoid altering the oligomeric conformation ( NuPage sample buffer without β-mercaptoethanol ) and subsequently transferred onto nitrocellulose . After blocking overnight at 4°C with 10% nonfat dried milk , membranes were probed for 1 hr at room temperature with anti-oligomer antibody F11G3 , provided by Dr Rakez Kayed ( 1:1000 ) , anti-ATXN1 antibody 11750 ( 1:3000 ) , anti-ATXN1 antibody 11NQ , or anti-CIC antibody ( 1:4000 ) diluted in 5% nonfat dried milk . 11750 and 11NQ immunoreactivity was detected with horseradish peroxidase conjugated anti-rabbit IgG ( 1:8000; Jackson ImmunoResearch Laboratories , West Grove , PA ) ; anti-mouse IgM ( 1:6000; Jackson ImmunoResearch Laboratories ) was used for F11G3 and anti-guinea pig IgG ( 1:8000; Jackson ImmunoResearch Laboratories ) . For signal detection , ECL Plus ( Amersham-Pharmacia , Piscataway , NJ , USA ) was used . Paraffin sections were deparaffinized , rehydrated , and washed in 0 . 01 M PBS 3 times for 5 min each time . After blocking in normal goat serum for 1 hr , sections were incubated overnight with rabbit anti-ATXN1 antibody 11750 ( 1:700 ) . The next day , the sections were washed in PBS 3 times for 10 min each time and then incubated with goat anti-rabbit IgG Alexa Fluor 568 ( 1:700; Invitrogen , Grand Island , NY ) for 1 hr . The sections were then washed 3 times for 10 min each time in PBS before incubation overnight with mouse anti-oligomers F11G3 ( 1:300 ) . The next day , the sections were washed in PBS 3 times for 10 min each before incubation with goat anti-IgM Alexa Fluor 488 ( 1:700; Invitrogen ) for 1 hr . Sections were washed and mounted in Vectashield mounting medium with DAPI ( Vector Laboratories , Burlingame , CA ) . The sections were examined using a Zeiss LSM 710 confocal microscope . Immunohistochemistry was performed on paraffin-embedded sections . In brief , sections ( 5 μm ) were deparaffinized and rehydrated . Primary antibodies were detected with biotinylated goat anti-mouse IgG ( 1:2000; Jackson ImmunoResearch Laboratories ) , biotinylated goat anti-mouse IgM ( 1:1500 ) , or biotinylated goat anti-rabbit IgG ( 1:1800 ) ( all from Jackson ImmunoResearch Laboratories ) and visualized using an ABC reagent kit ( Vector Laboratories ) , according to the manufacturer’s recommendations . Bright-field images were acquired using a Carl Zeiss Axio Imager M2 microscope , equipped with an Axio Cam MRc5 color camera ( Carl Zeiss , Germany ) . Sections were counterstained with hematoxylin ( Vector Laboratories ) for nuclear staining . The following antibodies were used for immunostaining: rabbit anti-ATXN1 antibody 11750 ( 1:700 ) , rabbit anti-oligomer antibody A-11 ( 1:600 ) , mouse anti-oligomer antibody F11G3 ( 1:100 ) , and mouse anti-calbindin antibody ( 1:450 ) . Cerebellar granular precursor cells were obtained from E16 wild-type C57BL6 mice . Cells were maintained in DMEM ( Life Technologies , Inc . , Invitrogen , Carlsbad , CA , USA ) supplemented with 10% FBS , glutamine ( 4 mM ) , penicillin ( 200 U/ml ) , streptomycin ( 200 μg/ml ) , and sodium pyruvate ( 1 μM ) . Cells were maintained at 37°C in 5% CO2 . Cells ( ∼10 , 000/well ) were plated in 96-well plates ( Corning Glassworks , Corning , NY ) and grown overnight . SEC fractions ( 0 . 3 μg of total protein ) from mouse cerebellum were added to the cells and incubated for 24 hr . Cell viability was assessed spectrophotometrically using a 3-[4 , 5-dimethylthiazol-2-yl]-2 , 5-diphenyltetrazolium bromide ( MTT ) -based assay according to the manufacturer's specifications ( Sigma–Aldrich , St . Louis , MO , USA ) . CGP or stable Daoy mRFP-ATXN1 ( 82Q ) cells on the coverslip were fixed and permeabilized after their specific treatment . The samples were blocked for 1 hr in 5% goat serum and incubated overnight at 4°C with anti-Nuj1 for CPG cells ( 1:150 , Covance , Cranford , NJ ) . For the Daoy mRFP-ATXN1 ( 82Q ) cells , samples were incubated overnight with anti-oligomer F11G3 ( 1:100 ) , anti-ATXN1 11 , 750 ( 1:500 ) or anti-Myc ( 1:500 , GenScript ) . In all cases cells were then washed and incubated with an Alexa 488-conjugated goat anti-mouse or anti-rabbit antibody ( 1:700 , Invitrogen ) for 1 hr at room temperature . DAPI was used for nuclear staining ( 1:4000 , Invitrogen ) . Cells were then washed for 30 min and cover slipped . For ELISA , plates were coated with 10 μl of sample using 0 . 1 M sodium bicarbonate ( pH 9 . 6 ) as a coating buffer , followed by incubation for 1 hr at 37°C , washing 3 times with Tris-buffered saline with very low ( 0 . 01% ) Tween 20 ( TBS-T ) , and then blocking for 1 hr at 37°C with 10% BSA . The plates were then washed three times with TBS-T; F11G3 ( 1:500 ) , A-11 ( 1:1000 ) , 11 , 750 ( 1:2000 ) , OC ( 1:1000 ) or Tubulin ( 1:2000 ) antibodies ( diluted in 5% nonfat milk in TBS-T ) were added and allowed to react for 1 hr at 37°C . The plates were then washed 3 times with TBS-T , and 100 μl of horseradish peroxidase-conjugated anti-mouse IgM , anti-mouse IgG or anti-rabbit IgG ( diluted 1:10 , 000 in 5% nonfat milk in TBS-T; Promega , Madison , WI , USA ) were added , followed by incubation for 1 hr at 37°C . Finally , plates were washed 3 times with TBS-T and developed with 3 , 3 , 5 , 5-tetramethylbenzidine ( TMB-1 component substrate ) from KPL ( Gaithersburg , MD , USA ) . The reaction was stopped with 100 μl of 1 M HCl , and samples were read at 450 nm . The AXH domain was expressed and purified as previously described ( Kim et al . , 2013 ) . Briefly , the AXH domain ( amino acids 563–689 ) from human ATXN1 was cloned into a pET28a vector , generating an N-terminal His tag . The AXH domain was expressed in Escherichia coli BL21 ( DE3 ) and purified with an Ni-affinity column . The N-terminal His tag was cleaved by TEV protease ( Life Technologies ) and further purified with a Hitrap Q ( GE Healthcare ) anion exchange chromatography and with Superdex 75 ( GE Healthcare ) size exclusion chromatography ( SEC ) . The CIC N-terminal fragment ( amino acids 1–117 ) , wild-type or mutant ( W37A ) , was cloned into a pGEX-4T-1 vector , generating a GST tag . The N-terminal CIC was expressed in Escherichia coli BL21 ( DE3 ) and purified in a GST-affinity column . After AXH purification , the buffer was exchanged by dialysis to 20 mM Tris–HCl ( pH 7 . 0 ) , with 20 mM NaCl , 2 mM β-mercaphtoethanol and 0 . 01% ( wt/vol ) NaN3 and then concentrated to a final concentration of 0 . 3 mg/ml . The AXH solution was stirred at 37°C for 7 days . Aliquots were collected at different time points and stored at −80°C . To determine the effect of CIC in AXH oligomer formation , both peptides were mixed in a 1:1 molar ratio and stirred at 37°C for 7 days . Aliquots were also collected at different time points and stored at −80°C . All experiments were performed in triplicate . For AXH oligomers quantification , ELISA was performed as described above . Briefly , 10 μl of sample was coated overnight . The anti-oligomer antibody A-11 ( 1:1000 ) was utilized to detect oligomers and the anti-ATXN1 antibody 11750 ( 1:500 ) was used to quantify total AXH . Thioflavin T ( Sigma ) binding was measured using a POLARstar OMEGA plate reader ( BMG Labtechnologies , Melbourne , VIC , Australia ) with 440-10 nm/520 nm excitation/emission filters set . 1 μl of 0 . 3 mg/ml of protein sample were mixed with 250 μl of 5 μM ThT , 50 mM glycine buffer ( pH 8 . 5 ) . Fluorescence intensity values of samples were obtained by subtraction of blank . One ml of the sample was diluted 20 times in 1 . 7 M Ammonium Sulfate buffer and then loaded on to the Hydrophobic interaction column column ( 1 ml Phenyl Sepharose Fast Flow , low substitution , from GE ) at 1 ml/min . Fractions were collected every 0 . 5 ml . An AKTA UPC 10 FPLC system from GE at 4° was used for all chromatography steps . To determine the percentage of PCs with ATXN1 oligomers , 5 μm brain sections were stained with anti-calbindin antibody as described above to quantify the total number of PCs in the cerebellum . The number of PCs positive for calbindin was considered as 100% . The adjacent sections were immunostained for oligomers using F11G3 or A-11 . For the quantification of nuclear inclusions , we stained ATXN1 using 11750 antibody and performed nuclear staining with hematoxylin . We considered the total amount of nucleus in the cortex as 100% of the nucleus . Data were analyzed using post-hoc test . The antibody F11G3 was preincubated for 40 min with Aβ oligomers , IAPP oligomer , Aβ monomers or IAPP monomers . After the incubation period , this antibody was used to perform double IF experiments with 11 , 750 in Atxn1154Q/+ mouse cerebellar sections . SEC fractions from Atxn1154Q/+ mouse cerebella were incubated up to 48 hr at 37°C . The presence of oligomers in each fraction at each time point was determined by dot-blot using F11G3 ( 1:1000 ) . A total of 50 μg of SEC fractions protein in 200 ml of 2% SDS was boiled for 5 min and run through a dot blot apparatus under a vacuum onto a cellulose acetate membrane . Membrane was then washed 3x with 0 . 1% SDS and then blocked in 5% milk and subject to western blot analysis using F11G3 ( 1:1000 ) , OC ( 1:1000 ) and 11750 ( 1:8000 ) . Mammalian expression vectors Myc-CiC/WT , Myc-CiC/W37A , GST-ATXN1 ( 82Q ) and GST-ATXN1 ( 30Q ) have been described previously ( Lim et al . , 2008; Kim et al . , 2013 ) . Hela cells were cultured in DMEM medium with 10% fetal bovine serum and were plated in six-well plates 1 day before transfection . On the day of transfection , corresponding DNA constructs were transfected with Lipofectamine 2000 ( Invitrogen ) . In all cases 100 ng of GST-ATXN1 ( 30Q ) or GST-ATXN1 ( 82Q ) were used for transfection . For the CIC construct we transfected 0 . 5 , 1 , 2 or 3 μg . Two days later , cells were lysed for 20 min at 4°C in lysis buffer ( 150 mM NaCl , 50 mM Tris–HCl at pH 7 . 5 , 1 mM EDTA , 0 . 1% Triton X-100 , 0 . 1% Tween-20 , complete protease inhibitor cocktail [Roche , Switzerland] , PhosSTOP phosphatase inhibitor cocktail [Roche] ) . After centrifugation , soluble fraction was used for western blot and ELISA measurements . Lysis was performed on ice for 20 min with brief vortexing using in 0 . 5% Triton buffer ( 0 . 5% triton X-100 , 50 mM Tris pH 8 , 75 mM Nacl ) supplemented with protease and phosphatase inhibitors ( Sigma ) . Cell debris were removed by centrifugation ( 20 min , 15 , 000 r . p . m , 4°C ) and pre-cleared with un-conjugated beads . In parallel , 2 μg of antibody was conjugated to Dynabeads for 1 hr at 4°C with rocking . Lysate was then added to the conjugated beads overnight at 4°C . Beads were then washed 5 × 500 μl of lysis buffer before being eluted using sample buffer and boiling for 10 min . Data are represented as mean ± SEM . Different conditions were compared using one- or two-way ANOVA followed by the indicated post hoc test to compare controls with treatment groups and/or genotypes or by two-tailed Student's test , as appropriate . | Spinocerebellar ataxia type 1 ( SCA1 ) is a progressive neurodegenerative disease in which damage to the brain regions that control movement results in the gradual loss of coordination and motor skills . The disease is a caused by a mutation in the gene that codes for a protein called ATAXIN-1 . In healthy individuals this protein contains up to 39 copies of an amino acid called glutamine . However , the mutant gene can encode for 40 or more copies of glutamine , which results in a longer-than-usual ATAXIN-1 protein with toxic properties . Within the brain , some of the toxic ATAXIN-1 proteins form insoluble deposits , while the rest remain soluble . At first it was assumed that the insoluble deposits were responsible for the neurodegeneration seen in SCA1 . However , closer examination revealed that these deposits form mainly in brain regions that do not degenerate , which suggests that they might instead have a protective role . This is consistent with evidence from research into other brain disorders , including Alzheimer's disease , which suggests that the soluble form of the toxic proteins might be causing these diseases . Lasagna-Reeves et al . now provide the first direct evidence that the soluble form of the toxic ATAXIN-1 proteins are indeed harmful in a mouse model of SCA1 . Experiments reveal that these soluble proteins accumulate in the brain regions that undergo degeneration in SCA1 , such as the cerebellum , but not in those regions that remain intact . Moreover , the motor skills and coordination of the mice get worse as the level of soluble toxic ATAXIN-1 increases . Lasagna-Reeves et al . go on to show that a protein called capicua stabilizes the toxic ATAXIN-1 proteins , which keeps them from forming insoluble deposits . Since capicua is particularly abundant in the cerebellum , this could explain the high levels of toxic ATAXIN-1 in this region and why it is vulnerable to degeneration . Future experiments are needed to investigate whether proteins equivalent to capicua might play a similar role in stabilizing toxic proteins in Alzheimer's , Parkinson's and Huntington's diseases , and whether preventing this stabilization could have therapeutic potential . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | A native interactor scaffolds and stabilizes toxic ATAXIN-1 oligomers in SCA1 |
DNA methylation is extensively remodeled during mammalian gametogenesis and embryogenesis . Most transposons become hypomethylated , raising the question of their regulation in the absence of DNA methylation . To reproduce a rapid and extensive demethylation , we subjected mouse ES cells to chemically defined hypomethylating culture conditions . Surprisingly , we observed two phases of transposon regulation . After an initial burst of de-repression , various transposon families were efficiently re-silenced . This was accompanied by a reconfiguration of the repressive chromatin landscape: while H3K9me3 was stable , H3K9me2 globally disappeared and H3K27me3 accumulated at transposons . Interestingly , we observed that H3K9me3 and H3K27me3 occupy different transposon families or different territories within the same family , defining three functional categories of adaptive chromatin responses to DNA methylation loss . Our work highlights that H3K9me3 and , most importantly , polycomb-mediated H3K27me3 chromatin pathways can secure the control of a large spectrum of transposons in periods of intense DNA methylation change , ensuring longstanding genome stability .
Millions of transposable elements reside in mammalian genomes , far surpassing in number the approximately 25000 protein-coding genes ( Lander et al . , 2001 ) . Most of these elements are retrotransposons , which utilize an RNA intermediate to duplicate and mobilize . Through their activity or their mere presence , transposons can be both beneficial for the evolution of the host genome and deleterious for its integrity . They can modify gene functions through insertional mutagenesis , influence gene transcriptional outputs by acting as promoters or enhancers or induce chromosomal rearrangements through non-allelic recombination ( Goodier and Kazazian , 2008 ) . Accordingly , erratic transposon-related events have been linked to congenital diseases , cancer and infertility ( Kaer and Speek , 2013 ) . Successive waves of transposon expansion and decline have shaped mammalian genomes over evolution , leading to a current occupancy rate of approximately half of the genomic space . Reflecting their various evolutionary origin and multiplication success , resident elements are greatly diverse in structures , numbers and functional properties , which define discrete families of transposons . Long Terminal Repeat ( LTR ) sequences characterize endogenous retroviruses ( ERVs , 12% of the mouse genome ) , which can be further subdivided into three families ( ERV1 , ERVK and ERVL ) , according to the infectious retroviruses they derive from ( Stocking and Kozak , 2008 ) . Non-LTR elements comprise Long and Short INterspersed Elements ( LINEs and SINEs , 20% and 8% of the genome , respectively ) , and also consist of specific sub-families ( Babushok et al . , 2007 ) . The majority of transposons have accumulated nullifying mutations and truncations , but around 1–2% of LINEs and ERVs have intact sequences that embed the protein coding information necessary for their mobilization . Notably , ERVK elements show the greatest level of activity , which causes at least 10% of spontaneous mutations in laboratory mice ( Maksakova et al . , 2006 ) . To minimize their impact on genome fitness , multiple layers of control antagonize transposons at different steps of their life cycle ( Zamudio and Bourc’his , 2010 ) . Notably , restraining mechanisms can differ between cell types . In somatic cells and in the male differentiating germline , DNA methylation is the main transcriptional suppressor of LTR and non-LTR transposons . In these contexts , transposable elements are densely methylated ( Rollins et al . , 2006; Smith et al . , 2012 ) and DNA hypomethylation leads to their de-repression ( Bourc’his and Bestor , 2004; Walsh et al . , 1998 ) . In contrast , the early germline and the early embryo manage to globally control their transposon burden without DNA methylation . These cells naturally undergo genome-wide loss of DNA methylation , likely as part of the acquisition of a pluripotent , flexible state ( Seisenberger et al . , 2013 ) . Moreover , genetic studies have demonstrated that mouse embryonic stem ( ES ) cells can use DNA methylation-independent mechanisms to silence transposons: knocking-out the three active DNA methyltransferases ( Dnmt-tKO ) does not yield significant de-repression of transposons , except Intracisternal A Particle ( IAP ) elements ( Karimi et al . , 2011b; Matsui et al . , 2010 ) In fact , transposon control in ES cells seems to rely primarily on post-translational histone methylation , notably at lysine 9 of histone H3 ( H3K9 ) . H3K9 dimethylation ( H3K9me2 ) , which is deposited by the EHMT2/G9a and EHMT1/GLP lysine methyltransferases , directly and specifically represses class L ERVs ( Maksakova et al . , 2013 ) . H3K9 trimethylation ( H3K9me3 ) can be catalyzed by the SETDB1 ( also known as ESET ) or the SUV39H enzymes . The SUV39H system targets H3K9me3 at evolutionary young LTR and non-LTR transposons , but Suv39h mutant ES cells principally up-regulate LINE1 elements ( Bulut-Karslioglu et al . , 2014 ) . In parallel , SETDB1 , together with its associated co-repressor , the Krüppel-associated box domain ( KRAB ) -Associated Protein 1 ( TRIM28 , also known as KAP1 ) , mainly control H3K9me3-dependent suppression of ERVK transposons- a family to which IAP elements belong ( Karimi et al . , 2011b; Matsui et al . , 2010; Rowe et al . , 2010 ) . TRIM28 is recruited to specific genomic sites via direct interactions with KRAB-zinc finger proteins ( Friedman et al . , 1996 ) , which are a large family of DNA binding factors that co-evolved with ERVs ( Emerson and Thomas , 2009 ) . Therefore , different H3K9 methylation-based mechanisms are utilized to silence different transposons families in ES cells . In contrast , the repressive spectrum of polycomb-mediated H3 lysine 27 trimethylation ( H3K27me3 ) is limited: only Murine Leukemia Virus ( MuLV ) elements are reactivated upon H3K27me3 deficiency ( Leeb et al . , 2010 ) . However , the prevailing view that H3K9 methylation acts as the main transposon controller in ES cells may be biased by two confounding factors . First , conclusions are based on analyses of chromatin modifier mutants , which still harbor high DNA methylation levels . Second , proper transposon repression in Dnmt-tKO ES cells may reflect a long-term adaptation to a DNA methylation-free state rather than a lack of significant role of DNA methylation per se . In fact , how the ES cell genome transitions from a DNA methylation-dependent to -independent mode of transposon control has never been investigated . To study the dynamics of transposon regulation upon DNA methylation loss , we modulated the ES cell methylome by using interconvertible culture systems , which do not modify pluripotency potential . ES cells grown in standard serum-based conditions have heavily methylated genomes ( ~75% of CpG methylation ) ( Stadler et al . , 2011 ) , which is linked to the expression of de novo DNA methyltransferases . ES cells grown in presence of two small kinase inhibitors ( 2i ) down-regulate these enzymes , and have reduced DNA methylation levels ( Leitch et al . , 2013; Ying et al . , 2008 ) . Upon transfer from serum to 2i medium , demethylation occurs with a slow kinetics: several weeks are required to reach 20–30% of CpG methylation . Notably , imprinted genes , major satellite repeats and IAP elements maintain persistent DNA methylation after 2i adaptation ( Ficz et al . , 2013; Habibi et al . , 2013 ) . Addition of vitamin C ( vitC ) can also lower the ES cell methylome . This compound promotes active demethylation by stimulating the TET ( Ten Eleven Translocation ) enzymes , which oxidize 5-methylcytosines to 5-hydroxymethylcytosines that are potential intermediates towards unmethylated cytosines ( Blaschke et al . , 2013 ) . Here , by switching ES cells directly from a serum-based to a 2i+vitC medium , we were able to induce rapid and extensive demethylation genome-wide , mimicking a situation occurring in the early embryo . By combining DNA methylation , chromatin and transcriptional profiling of transposons along with genetic analyses , we found that DNA methylation represses multiple families of transposons in ES cells , but an epigenetic switch towards histone-based control is progressively implemented as DNA methylation disappears . Importantly , we reveal for the first time the specific and overlapping roles of H3K9 and H3K27 trimethylation in controlling distinct transposon families upon DNA demethylation . These findings have important implications for understanding the molecular underpinning of transposon control in the pluripotent cells of the early mammalian embryo .
Dnmt-tKO ES cells are completely devoid of DNA methylation , yet expression levels of most transposable elements remain globally similar to wild-type ( WT ) ES cells ( Karimi et al . , 2011b; Tsumura et al . , 2006 ) . This may indicate the implementation of alternative mechanisms that compensate for DNA methylation-based repression . To analyze dynamic adaptation , we utilized a culture-based system that results in rapid DNA methylation loss: converting ES cells from serum-based to 2i+vitamin C ( 2i+vitC ) culture conditions . To overcome confounding genetic effects , we used the J1 ES cell line , from which Dnmt-tKO mutants were originally derived . Quantification using the methyl-CpG sensitive restriction enzyme-based LUminometric Methylation Assay ( LUMA ) ( Karimi et al . , 2011a ) revealed that CpG methylation linearly decreased from 77% to 13% in six days , and reached a minimal level of 6% after 14 days of conversion ( Figure 1A ) . In comparison , cells grown in serum+vitC or in 2i-only maintained relatively high CpG methylation content after the same treatment duration , with an average of 56% and 22% , respectively ( Figure 1—figure supplement 1A ) , in agreement with a previous study ( Habibi et al . , 2013 ) . This suggests that such a rapid and extensive loss of genomic methylation can only be attained through the synergistic action of 2i-dependent passive demethylation and vitC-dependent active demethylation . 10 . 7554/eLife . 11418 . 003Figure 1 . Kinetics and extent of DNA methylation loss in ES cells upon serum to 2i+vitC conversion ( A ) Time course of global CpG methylation loss measured by LUMA over 14 days ( D0 to D14 ) of conversion from serum to 2i+vitC . Data represent mean and Standard Error of the Mean ( SEM ) between two biological replicates . ( B ) Sequence-specific CpG methylation level measured by bisulfite pyrosequencing . Data represent mean ± SEM between two biological replicates . ( C ) Tukey boxplot representation of genome-wide CpG methylation content as measured by WGBS in different culture conditions . Datasets of J1 Serum , E14 Serum and E14 2i were obtained from previous studies ( Habibi et al . , 2013; Seisenberger et al . , 2012 ) ( D ) CpG methylation distribution over different genomic compartments by WGBS . ( E ) Heatmap and hierarchical clustering of average CpG methylation over 69 transposon families as measured by WGBS . ( F ) Left panel: Tukey boxplot representation of CpG methylation content in Residually Methylated Regions ( RMRs ) ( n = 4 , 100 ) compared to the whole genome in various culture conditions . Right panel: pie chart distribution of 2i+vitC RMRs in different genomic compartments ( left ) and among repeats ( right ) . ( G ) Example of WGBS profile of a genomic region containing two 2i+vitC RMRs mapping to an IAPEY and a L1 elements . Bars represent the methylation percentage of individual CpG sites , between 0 ( unmethylated ) and 100% ( fully methylated ) . Location of LINE and LTR transposons ( RepeatMasker ) are displayed below; the RMRs are highlighted in red . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 00310 . 7554/eLife . 11418 . 004Figure 1—figure supplement 1 . DNA methylation is almost completely erased in 2i+vitC medium . ( A ) Global CpG methylation level in J1 ES cells as measured by LUMA . Mean ± SEM between two biological replicates . ( B ) Smoothed scatter plots of DNA methylation levels at individual CpG between two different conditions as measured by WGBS . Histograms: distribution of CpG methylation levels for all CpG . Pearson correlation between culture conditions . ( C ) Typical example of methylation profiles in serum , 2i-only and 2i+vitC . Each dot represents the average methylation level over 10–15 kb windows . ( D ) Average methylation level in Imprinted Control Regions ( ICRs ) as measured by WGBS . ( E ) Representative genomic region containing the H19 ICR . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 004 To monitor the demethylation dynamics of specific genomic sequences , we performed quantitative bisulfite-pyrosequencing ( Figure 1B ) . All analyzed sequences reached very low levels of CpG methylation upon 2+vitC switch , although at various rates . Young LINE1 transposons ( L1-A and L1-T ) mirrored the dynamics of the genome average , while the CpG-rich promoter of the germline-specific gene Dazl was a fast 'loser' , comparatively . Consistent with their intrinsic ability to maintain high levels of DNA methylation in various contexts of global DNA hypomethylation ( Ficz et al . , 2013; Seisenberger et al . , 2013 ) , the demethylation rate of IAP transposons and the Imprinting Control Region ( ICR ) of the H19-Igf2 locus was slower than the rest of genome . Nevertheless , the combination of 2i and vitC eventually overcame chromatin environments that confer protection of these sequences from DNA demethylation . To determine the extent of DNA demethylation globally in 2i+vitC culture conditions , we carried out whole-genome bisulfite sequencing ( WGBS ) at the conversion end-point . Quality control indicated high genomic coverage , with approximately 55% of CpGs covered at least five times ( Supplementary file 1 ) . Available methylome maps indicate that 71% and 30% of CpG sites are methylated in serum and in 2i-only conditions , respectively ( Habibi et al . , 2013; Seisenberger et al . , 2012; Figure 1C , Figure 1—figure supplement 1B and Supplementary file 1 ) ; in contrast , ES cells grown in 2i+vitC were almost completely unmethylated , with an average CpG methylation of 4 . 6% , which is fully consistent with the LUMA quantification ( Figure 1C ) . Low methylation levels were homogeneously found throughout all genomic compartments , including single-copy genic regions and repeated sequences ( Figure 1D and Figure 1—figure supplement 1C ) . In particular , all transposable element families ( ERVs , LINEs and SINEs ) were affected by 2i+vitC-induced demethylation ( Figure 1E ) . In an attempt to identify individual genomic regions with significant DNA methylation traces ( Song et al . , 2013 ) , we uncovered 4 , 100 Residually Methylated Regions ( RMRs ) ( Figure 1F , G ) , which exhibited an average of 26% of CpG methylation after long-term 2i+vitC conversion . These were also regions prone to high DNA methylation retention in 2i-only conditions ( 65% of CpG methylation ) . We estimated that nearly 75% of the RMRs overlapped with repeated sequences , among which half belonged to the ERVK class . This confirmed the specific ability of these elements , which includes IAPs , to resist genome-wide erasure of DNA methylation . One quarter of the repeat-associated RMRs overlapped with LINEs , however specifically localized around 5’ UTR regions; in contrast , ERVK-associated RMRs encompassed the entire length of these elements ( Figure 1G ) . Notably , Imprinting Control Regions ( ICRs ) , which are usually protected against DNA methylation erasure in 2i conditions , were devoid of any residual DNA methylation in 2i+vitC ( Figure 1—figure supplements 1D , E ) . Our analyses show that only scarce genomic regions retain DNA methylation in 2i+VitC , and even those regions are lowly methylated when compared to other culture systems . Global CpG methylation levels , less than 5% , are unprecedented in male WT cells , both in culture and in vivo ( Seisenberger et al . , 2013 ) . This experimental system provides a valuable means to study the dynamic adaptation of the genome to a loss of DNA methylation . Using the serum to 2i+vitC medium conversion system , we investigated how transposable elements transcriptionally respond to an acute loss of DNA methylation . Through a time-course RT-qPCR analysis of steady-state levels of three classes of retrotranscripts ( LINE1 , IAPEz and MERVL ) , we observed a two-phase pattern: 1 ) an initial up-regulation , which culminates at day 6 ( D6 ) of 2i+vitC conversion , when genomic methylation reaches a low plateau , then 2 ) re-silencing in the absence of DNA methylation ( Figure 2A ) . This was confirmed by amplification-free Nanostring nCounter quantification and further extended to VL30 elements ( Figure 2—figure supplement 1A ) . To rule out background-specific effects , we exposed serum-cultured E14 ES cells to 2i+vitC ( Figure 2—figure supplement 1B ) . Despite some differences in the magnitude of transposon de-repression observed between the J1 and E14 cell lines , the same biphasic pattern of regulation was reproduced . By contrast , the quantity of transposon transcripts remained constant during the conversion from serum to 2i-only or from serum to serum+vitC ( Figure 2—figure supplement 1A ) , which further underscores the synergistic effect of 2i and vitC in releasing DNA methylation-based repression of transposons . Importantly , the transposon transcription burst did not occur upon conversion of Dnmt-tKO cells ( Figure 2—figure supplement 1C ) . Rapid transition from a methylated to an unmethylated state seems to provide a window for transposon reactivation; this is in agreement with the hypothesis that Dnmt-tKO ES cells have likely acquired long-term compensatory mechanisms preventing this relaxation . 10 . 7554/eLife . 11418 . 005Figure 2 . Two phases of transposon regulation upon serum to 2i+vitc conversion . ( A ) Dynamic expression of LINE1 , IAPEz and MERVL families upon conversion from serum to 2i+vitC as measured by RT-qPCR . Values were normalized to Gapdh and Rplp0 and are expressed as the fold change to D0 . Data represent mean ± SEM from five biological replicates . *p<0 . 05 , **p<0 . 01 and ***p<0 . 001 ( Student’s t-test ) . ( B ) Evolution of LINE1-ORF1 protein levels at different time points during medium conversion . ( C ) Distribution of LINE1-ORF1 and IAP-gag protein levels after ImageJ quantification of immunofluorescence intensity in individual cells . Between 1000 and 5000 cells were analyzed per sample . ***p<0 . 001 ( Wilcoxon rank-sum test ) ( D ) Volcano plot representation of up- and down-regulated transposons as measured by RNA-seq between D0 and D6 ( left ) , D6 and D13 ( middle ) , and D0 and D13 ( right ) . Red dots indicate significantly misregulated repeats between two conditions ( fold change >2 and p-value<0 . 05 ) . RNA-seq mapping allowed multiple hits onto the genome . ( E ) Heatmap representation and hierarchical clustering of expression changes for 69 transposon families at D0 , D6 and D13 . Bold names: transposons of specific interest; grey names: transposons that are not significantly up- or down-regulated between any time points . Colors represent on a log2-scale the differential expression between a given time point and the average of the three time points . ( F ) Expression of individual elements from different transposon families at D0 , D6 and D13 in Count per Millions ( CPM ) . Each dot represents a single element . RNA-seq mapping allowed only unique hits in the reference genome; only elements with a minimum of 10 reads in at least one of the sample were conserved . The black bar represents the median of the distribution . Analyzed numbers of distinct transposon copies per family appear into brackets . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 00510 . 7554/eLife . 11418 . 006Figure 2—figure supplement 1 . Two phases of transposon regulation correlate with rapid DNA methylation loss . ( A ) Expression of different transposons in J1 ES cells as measured by Nanostring during the conversion from serum to 2i+vitC ( grey ) , 2i-only ( light blue ) and serum+vitC ( dark blue ) . Data are expressed as fold change to D0 and represent mean ± SEM between eight ( for 2i+vitC ) or two ( for 2i- and vitC-only ) biological replicates . ( B ) Expression of transposons in E14 ES cells . Data are expressed as fold change to D0 and represent mean ± SEM between two biological replicates . p<0 . 05 , **p<0 . 01 and ***p<0 . 001 ( Student t-test ) . ( C ) Expression of transposons in Dnmt-tKO and J1 ES cells measured . Data are expressed as fold change to J1 D0 and represent mean ± SEM between three biological replicates ( for Dnmt-tKO ) . *p<0 . 05 , **p<0 . 01 and ***p<0 . 001 ( unequal variances t-test between WT and Dnmt-tKO at a given day ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 00610 . 7554/eLife . 11418 . 007Figure 2—figure supplement 2 . Cellular heterogeneity of transposon expression . ( A ) Left panel: Immunofluorescence staining for LINE1-ORF1 and NANOG proteins in serum- and 2i+vitC-grown ES cells . Right: panel: Density of LINE1-ORF1 protein levels after ImageJ quantification of immunofluorescence intensity in individual cells . Between 2000 and 5000 cells were analyzed per time point: cells have globally a higher level of LINE1-ORF1 at D6 . ( B ) Same as ( A ) , with staining for IAP-gag and NANOG proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 00710 . 7554/eLife . 11418 . 008Figure 2—figure supplement 3 . Cell proliferation , cell death , and genome integrity . ( A ) Mitotic division rate of J1 ES cells during the conversion from serum to 2i+vitC . Mean ± SEM between three biological replicates . ( B ) Quantification of proliferation marker expression by Nanostring . Data are expressed as fold change to D0 and represent mean ± SEM between seven biological replicates . ( C ) Percentage of cells immunostained against the apoptosis marker PARP and the mitotic marker H3S10 phosphorylation . Between 5000 and 10000 cells were counted at each time point . When available , data represent mean ± SEM between two biological replicates . ( D ) Absolute copy number of L1-ORF2 and IAPΔ1 fragments assayed by qPCR on genomic DNA . Values are expressed as copies per genome , representing the mean ± SEM between two biological replicates . ( E ) Chromosome numbers counted in 20 metaphase spreads in serum and 2i+vitC medium . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 00810 . 7554/eLife . 11418 . 009Figure 2—figure supplement 4 . Genome-wide characterization of transposon relaxation . ( A ) Reconstruction principles of RepeatMasker annotation . ( B ) Pearson correlation and hierarchical clustering between RNA-seq experiments . ( C ) Number of individual LINEs and ERVs with detectable expression at D0 , D6 and D13 of 2i+vitC conversion . Elements with a minimum of 10 uniquely mappable reads were considered . ( D ) Expression of individual elements from different transposon families at D0 , D6 and D13 in Count per Millions ( CPM ) . Each dot represents a single element . RNA-seq mapping allowed only unique hits in the reference genome and only elements that had a minimum of 10 reads in at least one of the sample were conserved . The black bar represents the median of the distribution . Analyzed numbers of distinct transposon copies per family appear into brackets . ( E ) Representative genomic regions comprising LINE1 and ERV repeats . RNA-seq coverage for the two biological replicates is represented in blue and orange , and the overlap in grey . Data represent normalized read density . ( F ) Expression of individual transposable elements and their localization on chromosomes 1 , 5 , 15 and 19 at D0 , D6 and D13 of conversion . Each bar represents the expression of a single element in Count per Millions ( CPM ) . Elements with a minimum of 10 uniquely mappable reads in at least one sample were considered . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 009 We further found that the burst of transposon expression also occurs at the protein level: both LINE1-ORF1 and IAP-gag proteins presented a peak of expression at D6 , which we detected by western blotting ( Figure 2B ) and by quantification of immunofluorescence signals ( Figure 2C and Figure 2—figure supplement 2A , B ) . While IAP-gag staining was uniform among cells at a given time point , LINE1 protein intensity showed great inter-cellular variability , ranging from intense to no signal . Importantly , the level of LINE1 heterogeneity was present throughout the conversion process from serum to 2i+vitC conditions , including at D6 . In an attempt to link this heterogeneity with fluctuating levels of pluripotency , we performed co-staining against NANOG ( Figure 2—figure supplement 2A ) . We could not detect any correlations suggesting that LINE1 heterogeneous expression is not linked to various degrees of pluripotency . Additionally , co-staining with phosphorylated-H2AX did not reveal a correlation between the level of transposon expression and DNA damage ( data not shown ) . We wanted to rule out that the repression phase we observed was not simply a reflection of positive selection of a subset of cells that maintained transposon repression throughout the medium conversion . We found cell proliferation to remain globally constant over the 14-day period of media conversion , as measured by division rate ( Figure 2—figure supplement 3A ) , transcriptional level of different proliferation markers ( Figure 2—figure supplement 3B ) or percentage of histone H3 Serine 10 phosphorylation-positive cells ( Figure 2—figure supplement 3C ) . Similarly , we did not observe increased cell death/apoptosis at any days during the conversion ( Figure 2—figure supplement 3C ) . Finally , despite the transient release of transposon silencing at D6 , we failed to detect transposon multiplication or transposon-induced chromosome rearrangements: genomic copy numbers of LINE1 and IAPEz elements as well as karyotypes were globally similar between cells before ( D0 ) and after the transposon burst ( D14 ) ( Figure 2—figure supplement 3D , E ) . In sum , 2i+vitC induces a transient up-regulation of transposon transcription and translation , but cellular viability and genome integrity remain largely intact . To gain a qualitative and quantitative view of the transcriptional dynamics of transposons upon acute loss of DNA methylation , we performed paired-end RNA-seq at D0 , D6 and D13 of serum to 2i+vitC conversion , in biological replicates . Typically , to map transposons , the choice is either to allow multiple hits at the expense of specificity , or to consider unique reads only and lose substantial information . Here , we combined the two methods , in order to provide in-depth characterization of the dynamics of transposon regulation at the familial level , while bringing insights into intra-familial heterogeneity . We further improved transposon mapping by correcting the RepeatMasker annotation ( Figure 2—figure supplement 4A ) , which tends to overestimate the number of transposon entities by counting a unique element as several individual fragments . This is systematic for ERVs , which are split into internal and LTR sequences , but can also concern any type of transposons with small internal deletions or insertions . Using bioinformatic resources allowing the assembly of different fragments of an element ( Bailly-Bechet et al . , 2014 ) , our reconstructed version gave a census of 588 , 739 LINEs and 497 , 706 ERVs , while the original annotation roughly doubles these numbers , with 989 , 411 LINEs and 969 , 096 ERVs . Finally , we assigned an integrity score to each element ( 1 being the maximum ) , taking into account deletions , insertions and the divergence rate from the consensus sequence . Using a cutoff of 0 . 8 , we predicted a number of 37 , 194 relatively intact LINEs ( 6 . 3% of total LINE elements ) and 15 , 604 ERVs ( 3 . 1% ) in the mouse reference genome . Quality control of our RNA-seq datasets indicated high genomic coverage ( Supplementary file 1 ) and consistency between replicates ( Figure 2—figure supplement 4B ) . Notably , by excluding transposon-derived reads mapping to RefSeq exons , only autonomously transcribed transposons were considered for this analysis . By allowing multiple hits and by weighting each read by its hit number , a total of 58 transposon families were found differentially expressed between at least two of the time points of medium conversion ( Figure 2D , E ) . Volcano plots show that almost all families underwent significant up-regulation from D0 to D6 ( Figure 2D , left panel ) , ranging from modest ( LINEs ) to robust ( MMERGLN ) fold changes . Silencing restoration also occurred globally between D6 and D13 , except for IAPEy or B1 elements , which remained at constant levels ( Figure 2D , middle panel ) . Comparison of transposon expression levels between the two end-points ( D0 and D13 ) indicated skewing in both directions ( Figure 2D , right panel ) . Some families , such as MERVL , SINEs B2 or any LINE1 types , were more strongly repressed after the 2i+vitC conversion at D13 than initially at D0 in serum . Others , like MMERGLN , ETnERV3 and IAPEz , underwent repression from D6 to D13 , but not to the full extent when compared to D0 . As a general rule , these data show that non-LTR ( LINEs and SINEs ) and LTR elements belonging to the K , L and 1 classes—albeit very different in terms of evolutionary origins and genomic structures—adopt common fates upon acute loss of DNA methylation . To examine whether the burst of transcription observed from bulk RNA profiling emanated from a few discrete elements or reflected a general trend within each family , we measured the transcriptional output of individual transposon copies by allowing unique read mapping only . We found that 7163 uniquely identifiable LINEs and 2372 ERVs showed activity throughout the conversion , which represented 1 . 2% and 3 . 8% of the total number of LINE1s and ERVs , respectively , or 19 . 3% and 15 . 2% of the intact elements of these families ( integrity score >0 . 8 ) . Importantly , these numbers are likely underestimated because active but identical copies cannot be discriminated , and are discarded from the analysis . Within all families , the number of significantly expressed elements was higher at D6 than at D0 or D13 of conversion ( Supplementary file 2A and Figure 2—figure supplement 4C ) . Generally , not only were more elements active , but individual copies also gained expression at D6 ( Figure 2F and Figure 2—figure supplement 4D , E ) . Finally , active transposons were evenly distributed along chromosomes , with no particular genomic hotspot ( Figure 2—figure supplement 4F ) . As a whole , the unique mapping analysis confirmed the class-specific features previously inferred from the familial analysis , regarding the degree of activation at D6 ( from modest for LINEs to intense for MMERGLN ) and silencing restoration at D13 ( strong for LINEs , intermediate for MMERGLN and nonexistent for IAPEy ) . Most importantly , it uncovered unprecedented details into the diverse regulation of individual transposons . Expression levels were the most homogeneous among elements of the same family during the D6 de-repression phase . Comparatively , at the D13 silencing restoration time-point , we observed heterogenic regulation at the inter- and intra-familial levels ( Figure 2F and Figure 2—figure supplement 4D ) . Some families , such as LINEs and MMERGLN , displayed collective behaviors , with the vast majority of elements simultaneously undergoing repression . In contrast , IAPEz , MERVL or ETn elements showed the widest distribution in individual expression . In particular , IAPEz elements were split into two categories at D13 , one that maintained high expression , and the other that underwent complete silencing . Globally , our analysis reveals that transposons undergo a transient relaxation of silencing upon DNA methylation loss followed by an expression reduction phase . However , family- and element-specific behaviors provide nuance to this general trend . It should be stressed here that a certain degree of heterogeneity is frequently inaccessible for young and highly conserved families of transposons , such as IAPEz and MMERVK10C , for which mapping reads to precise genomic locations is ambiguous , if not impossible . Compared to transposons , protein-coding genes followed different dynamics during 2i+vitC conversion ( Figure 3A and Figure 3—figure supplement 1A ) . The vast majority exhibited stable expression , while 3 , 301 genes were either up- or down-regulated; these numbers are similar to previous reports of a serum to 2i transcriptional switch ( Marks et al . , 2012 ) . While the general expression trend for transposons was biphasic , most differentially expressed genes displayed a monotonic pattern . A relevant example is the Dazl gene , which was continuously up-regulated from D0 to D13 ( Figure 3B ) , reflecting its sensitivity to vitC ( Blaschke et al . , 2013 ) . Conversely , expression of genes encoding transcription factors of the ZSCAN4 family progressively decreased during the conversion , with undetectable transcripts by D13 ( Figure 3B ) . As expression of these factors reflects a subpopulation of ES cells exhibiting a transcriptional profile akin to 2-cell stage embryos ( Macfarlan et al . , 2012 ) , our results imply that 2-cell-like cells exist in serum-based conditions but disappear in 2i+vitC medium . 10 . 7554/eLife . 11418 . 010Figure 3 . Gene expression analysis upon serum to 2i+vitC conversion . ( A ) Heatmap representation of genes ( n = 3301 ) that are significantly misregulated between at least two time points of D0 , D6 and D13 during medium conversion . Color codes as in Figure 2E . The arrow highlights a subset of genes whose expression transiently peaks at D6 . ( B ) Monotonic expression patterns of Dazl and Zscan4 family genes as measured by RNA-seq at D0 , D6 and D13 , expressed in RPKM ( read per kb per millions ) . Mean ± SEM between two biological replicates . ( C ) Dynamic expression of pluripotency transcription factor genes as measured by RNA-seq . Data is expressed as fold change to D0 and represent mean ± SEM between two biological replicates . ( D ) Expression of core pluripotency transcription factors by western blot . ( E ) RNA-seq track showing a chimeric transcript between a RLTR9E transposon element and the Mep1b gene specifically expressed at D6 . Data represent normalized read density . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 01010 . 7554/eLife . 11418 . 011Figure 3—figure supplement 1 . Gene expression analysis upon serum to 2i+vitC conversion . ( A ) Volcano plot representation of up- and down-regulated genes as measured by RNA-seq between D0 and D6 , D6 and D13 , and D0 and D13 . Red dots indicate significantly misregulated genes ( fold change >4 and p-value <0 . 01 ) . ( B ) RNA-seq track showing a chimeric transcript between an ORR1A2 transposon and the 2-cell-specific Ubtfl1 gene , specifically expressed at D6 . Data represent normalized read density ( C ) RNA-seq track showing uncoupled expression between the 2-cell-specific Zscan4d gene and the neighboring full-length MERVL transposon . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 011 The burst of expression at D6 appears specific to transposons , and is not a general trend of the genome . Nevertheless , 156 genes adopted a transposon-type pattern , with a peak at D6 followed by subsequent down-regulation at D13 ( Figure 3A and Supplementary file 1 ) . These genes were linked to ontology categories such as organismal development and were significantly enriched for transcription factors , most notably those related to pluripotency ( Supplementary file 2B ) . Further examination indicated that transcription of Nanog , Klf4 , Tbx3 and Prmd14 peaked at D6 ( Figure 3C ) ; enhanced production of pluripotency-related proteins was also observed by western blot within the first few days of 2i+vitC conversion ( Figure 3D ) . Therefore , the peak of transposon transcription at D6 coincides with a maximum availability of pluripotency regulators . It was previously shown that LTR sequences of ERVs can direct transcription of nearby genes in ES cells and early embryos , forming chimeric transcripts ( Karimi et al . , 2011b; Macfarlan et al . , 2012 ) . We detected several dozens of genes that used a promoter located in a transposable element ( ERV or LINE1 ) , independently of the medium composition ( Supplementary file 2C ) . A particular case was Mep1b , which clusters with the 156 'transposon-like' genes . This gene was induced ten-fold at D6 , concomitant with the activation of the RLTR9E element driving its expression , before returning to its initial level at D13 ( Figure 3E ) . Several 2-cell-specific genes have been shown to initiate from class L LTRs ( Macfarlan et al . , 2011 ) ; some of them , like Ubftl1 , were indeed specifically overexpressed at D6; in contrast , Zscan4d expression was uncoupled from the expression of the adjacent full-length MERVL element in our system ( Figure 3—figure supplement 1B , C ) . Although the burst of transposon expression at D6 can sporadically coordinate the transient activation of adjacent genes , it can be concluded that the genome-wide relaxation of transposons had generally a minimal effect on protein-coding gene expression . To gain insight into the basis for transposon regulatory dynamics , we examined the chromatin state of cells undergoing serum to 2i+vitC conversion . By western blot and immunostaining , we observed large-scale reorganization of histone modifications linked to transcriptional repression . While H3K9me3 marks remained globally constant , H3K9me2 levels were strongly reduced and , inversely , H3K27me3 levels increased from the first days of conversion ( Figure 4A and Figure 4—figure supplement 1A ) . The global dynamics of histone marks were not correlated with the changes in the availability—or lack thereof—of H3K9me2 modifiers and components of the polycomb machinery ( Figure 4—figure supplement 1B ) . 10 . 7554/eLife . 11418 . 012Figure 4 . Repressive chromatin reorganization upon loss of DNA methylation . ( A ) Western blot analysis of global levels of repressive histone modifications during the course of serum to 2i+vitC conversion . ( B ) H3K9me2 enrichment levels at three transposon families as measured by ChIP-qPCR . Quantitative data are expressed as the percentage of ChIP over Input . Data represents mean ± SEM of two biological replicates . ( C ) Genomic annotation of ChIP-seq peaks . Data represent the number of annotated peaks . H3K9me3 data are representative of two biological replicates , while H3K27me3 data represent only one , as peak calling could not be performed successfully on the second replicate . ( D ) H3K27me3 enrichment at major satellite repeats as measured by ChIP-qPCR . Data are represented as in 4B . ( E ) Heatmap and hierarchical clustering of average H3K9me3 and H3K27me3 levels in 69 transposons families at D0 , D6 and D15 . Colors represent the average read count for an element in a given family , relative to input ( average between two biological replicates ) . Only intact ( score>0 . 8 ) elements were considered . ( F ) Representative genomic region depicting evolution of H3K9me3 ( green ) and H3K27me3 ( red ) at L1-A ( category A ) , IAPEz ( category B ) and MERVL ( category C ) transposons . Data represent normalized read density . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 01210 . 7554/eLife . 11418 . 013Figure 4—figure supplement 1 . Repressive chromatin reorganization during conversion from serum to 2i+vitC . ( A ) Immunofluorescence experiment against H3K9me2 , H3K9me3 and H3K27me3 in J1 ES cells during serum to 2i+vitC conversion . ( B ) RPKM expression values of H3K9 HKMTs , Trim28 and PRC2 members as measured by RNA-seq . Mean ± SEM between two biological replicates . Transposon expression in Ehmt2-KO and TT2 ES cells ( WT ) as measured by Nanostring . Data are expressed as fold changes to WT D0 and represent mean ± SEM between two biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 01310 . 7554/eLife . 11418 . 014Figure 4—figure supplement 2 . Repressive chromatin reorganization during conversion from serum to 2i+vitC . ( A ) Pearson correlation and hierarchical clustering between samples after peak calling in H3K9me3 and H3K27me3 ChIP-seq experiments . ( B ) Number of H3K9me3 and H3K27me3 ChIP-seq read mapping to major satellite repeat consensus sequence . ( C ) Genomic annotation of ChIP-seq peaks in ES grown in serum but lacking DNA methylation ( Dnmt3-dKO , Dnmt1-KD ) . On the contrary to Figure 4C , peak calling was performed without taking input data into account . ChIP-seq datasets were taken from . Brinkman et al . , 2012 . ( D ) H3K4me3 enrichment levels at three transposon families as measured by ChIP-qPCR . Data are expressed as the percentage of ChIP over Input and represents mean ± SEM of two biological replicates . ( E ) H3K9me3 and H3K4me3 ChIP followed by bisulfite pyrosequencing . Mean ± SEM between two biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 014 10 . 7554/eLife . 11418 . 015Figure 4—figure supplement 3 . Screenshots of repressive chromatin reorganization at transposons . ( A ) Representative genomic region depicting evolution of H3K9me3 and H3K27me3 at RLTR4 repeats . Data represent normalized read density . ( B ) Representative genomic region depicting evolution of H3K9me3 and H3K27me3 at RLTR10 and IAPEy repeats . ( C ) Representative genomic region depicting evolution of H3K9me3 and H3K27me3 at various solo-LTRs . Example chosen here show isolated solo-LTRs that are not surrounded by other transposons . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 015 ChIP-qPCR measurement confirmed either the constitutive absence or the rapid removal of H3K9me2 at several transposon types upon 2i+vitC conversion ( Figure 4B ) , making it unlikely that this mark could participate to long-term transposon silencing in the absence of DNA methylation . To functionally exclude this possibility , we examined ES cells lacking the EHMT2 H3K9 dimethyltransferase ( Tachibana et al . , 2002 and Figure 4—figure supplement 1C ) . As previously described , when cultured in serum , Ehmt2-KO ES cells did not exhibit significant up-regulation of transposons as measured by Nanostring , with the exception of MERVL elements ( Macfarlan et al . , 2012; Maksakova et al . , 2013 ) . Upon 2i+vitC conversion , while LINE1 elements behaved as in WT cells , IAPEz and MERVL expression was enhanced around D6 in Ehmt2-KO cells ( Figure 4—figure supplement 1C ) . As H3K9me2 cannot be detected after D3 at these sequences ( Figure 4B ) , this relative up-regulation likely occurs through indirect effects . Most importantly , Ehmt2 mutants exhibited transposon re-silencing after D6 , which excludes a role for H3K9me2 in compensating the loss of DNA methylation-dependent repression . We then focused our analysis on the distribution of H3K9me3 and H3K27me3 marks by chromatin immunoprecipitation followed by deep sequencing ( ChIP-seq ) in biological replicates at D0 , D6 and D15 of conversion , allowing multiple mapping with random allocation ( Supplementary file 1 and Figure 4—figure supplement 2A ) . Neither the total number of H3K9me3 peaks ( 39424 at D0 and 38554 at D15 ) , nor their preferential occurrence on transposons was significantly altered during the conversion ( Figure 4C ) . In contrast , the number of H3K27me3-enriched regions raised four fold from D0 to D15 ( 9663 to 40098 ) . The vast majority of newly gained H3K27me3 peaks were located in ERV and LINE1 repeats , at the expense of gene promoters ( Figure 4C ) . We also observed a gradual H3K27me3 re-localization to pericentric heterochromatin during the conversion , by ChIP-qPCR at major satellite repeats ( Figure 4D ) , by immunostaining ( Figure 4—figure supplement 1A ) and by mapping ChIP-seq reads to the major satellite consensus sequence ( Figure 4—figure supplement 2B ) . Redistribution of H3K27me3 from gene promoters towards satellite repeats was previously reported in 2i-only conditions ( Marks et al . , 2012 ) . However , increased H3K27me3 levels and subsequent accumulation at different transposon repeats seems specific to the globally hypomethylated genome of 2i+vitC-cultured cells . Accordingly , hypomethylated Dnmt-tKO ES cells grown in serum displayed similar H3K27me3 redistribution towards transposons when we analyzed available ChIP-seq data ( Figure 4—figure supplement 2C ) . In agreement with the high expression of these elements , the 5’ UTR of LINEs ( L1-A and L1-T ) and the LTR of IAPEz were enriched in H3K4me3 , as assessed by ChIP-qPCR ( Figure 4—figure supplement 2D ) . Interestingly , bisulfite-pyrosequencing analysis of H3K4me3- or H3K9me3-immunoprecipitated chromatin confirmed , as expected , that active LINE1 elements ( marked by H3K4me3 ) had lower DNA methylation levels compared to inactive ones ( marked by H3K9me3 ) ( Figure 4—figure supplement 2E ) . This provides an additional documentation of the intra-familial heterogeneity of LINE1 elements at D6 . We next measured relative H3K9me3 and H3K27me3 levels over different transposon families , focusing our analysis on elements that were scored as intact . At D0 in serum , most transposon families were occupied by H3K9me3 to various extents , but lacked H3K27me3 ( Figure 4E ) . One noticeable exception was RLTR4 , which exhibited a strong H3K27me3 signal ( Figure 4—figure supplement 3A ) . Interestingly , this element is 90% identical to MuLV , which is one of the few transposons up-regulated in polycomb-deficient ES cells ( Leeb et al . , 2010 ) . Upon 2i+vitC conversion , H3K9me3 levels remained largely constant , although patterns observed in serum tended to be exacerbated: families with the initial highest enrichment ( IAPEz , RLTR6 and MMERVK10C ) were further enriched for this mark , while families with modest enrichment ( MERVL , MURVY or the MalR-class L ORR1A and ORR1B ) tended to lose it . Meanwhile , H3K27me3 progressively accumulated at most transposons ( Figure 4E ) , and this gain was variable among families: from inexistent for IAPEz to moderate for LINEs and VL30 , and to strong for various ERVs . Of note , when focusing on ERV elements , high levels of H3K9 or H3K27 methylation were found to be restricted to full-length intact elements: isolated solitary ( solo ) LTRs ( not interspersed with other repeats ) had most often no enrichment for H3K9 or H3K27 methylation ( Figure 4—figure supplement 3C ) . This suggests that internal ERV sequences are important for H3K9me3 and H3K27me3 recruitment . IAP elements were an exception , as their solo-LTRs were enriched for H3K9me3 on their own . Remarkably , different kinetics were observed for H3K9me3- and H3K27me3-related changes: H3K9me3 levels were rapidly modified between D0 and D6 , while H3K27m3 gain lagged behind , reaching its full extent between D6 and D15 . Although the whole picture is quite complex , it can be concluded that medium-induced DNA methylation profoundly remodels the repressive chromatin landscape of transposons . From a universal H3K9me3 occupancy in serum , transposon families exhibited three general trends in 2i+vitC: A ) co-occupancy of H3K9me3 and H3K27me3 ( LINEs , MMERGLN , RLTR6 , RLTR10 , IAPEy , VL30 ) , B ) exclusive H3K9me3 occupation ( IAPEz , MMERVK10C ) , and C ) complete switch from H3K9me2/3 to H3K27me3-regulated chromatin ( MERVL and MURVY ) ( Figure 4E , F and Figure 4—figure supplement 3B ) . Our analysis therefore provides a classification of the different transposon families into three main categories ( A , B , and C ) , according to the chromatin signature they adopt upon DNA methylation loss . To assess the behavior of individual elements among these three generic patterns , we attempted to analyze unique mappers , but the coverage on individual transposons was too low to extract reliable information . Nevertheless , to gain insight into the question of intra-familial heterogeneity , we plotted H3K9me3 and H3K27me3 enrichment for every intact transposons ( score>0 . 8 ) per family during the conversion . We found that elements of the B and C categories tended to be very homogeneous . IAPEz elements ( category B ) collectively gained H3K9me3 from D0 to D6; the MERVL and the Y-specific MURVY families ( category C ) also showed compact patterns , with individual elements transitioning together from H3K9me3 enrichment at D0 to H3K27me3 at D15 ( Figure 5A and Figure 5—figure supplement 1C ) . The A category , which is enriched for both H3K9me3 and H3K27me3 , was more diverse , with some families displaying homogeneous patterns , while others showed intra-familial dispersion in chromatin fates upon conversion . Within the MMERGLN , RLTR6 or RLTR10 families , all elements gained H3K27me3 while maintaining or gaining high levels of H3K9me3 . Within the L1-T and IAPEy families , the majority of elements gained H3K27me3 , but a subset maintained H3K9me3 without acquiring H3K27me3 ( Figure 5A and Figure 5—figure supplement 1A ) . Another case of intra-familial heterogeneity is provided by RLTR4 , which specifically carries H3K27me3 marks at D0 in serum: we demonstrate here that this enrichment was restricted to a small proportion of elements , as was previously suspected ( Reichmann et al . , 2012 and Figure 5—figure supplement 1D ) . By extracting single-element information from RNA-seq and ChIP-seq data , it is clear that both transcriptional and chromatin heterogeneity exists among some transposon families . Our analysis reveals that caution should be taken when interpreting average familial behaviors , as they may be representative of only a few individual elements inside a given family . 10 . 7554/eLife . 11418 . 016Figure 5 . H3K9me3 and H3K27me3 mark the same transposons but do not spatially overlap . ( A ) Normalized H3K9me3 and H3K27me3 enrichment over input at individual elements from different transposon families . Each dot represents a single element at D0 ( blue ) , D6 ( green ) and D15 ( red ) . Only intact ( integrity score>0 . 8 ) elements were considered . Data represent the average between two biological replicates . Analyzed numbers of distinct transposon copies per family appear into brackets . ( B ) Composite profile showing H3K9me3 ( green ) and H3K27me3 ( red ) coverage along different transposon sequences at D0 , D6 and D15 of medium conversion . ( C ) Representative genomic regions comprising LINE1 and ERV repeats that gain H3K27me3 in their 3’ end during the conversion , while maintaining H3K9me3 in the 5’ end . Data represent normalized read density . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 01610 . 7554/eLife . 11418 . 017Figure 5—figure supplement 1 . Intra-familial heterogeneity of H3K9me3 and H3K27me3 in transposons . Normalized H3K9me3 and H3K27me3 enrichment over input at individual elements from different transposon families . Each dot represents a single element at D0 ( blue ) , D6 ( green ) and D15 ( red ) . Only 'intact' ( integrity score>0 . 8 ) elements were considered . Data represent the average between two biological replicates . The number of distinct transposon copies considered is indicated next to the name of the family . ( A ) Member of the A category of transposons that are marked by both H3K9me3 and H3K27me3 in 2i+vitC . ( B ) B category , with exclusive H3K9me3 . ( C ) C category , with exclusive H3K27me3 . ( D ) RLTR4 is the only transposon marked by H3K27me3 in serum . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 01710 . 7554/eLife . 11418 . 018Figure 5—figure supplement 2 . H3K9me3 and H3K27me3 mark transposon sequences but do not spatially overlap . ( A ) Composite profile of H3K9me3 and H3K27me3 coverage on and +/- 5kb around different transposon sequences at D0 , D6 and D15 of 2i+vitC conversion . Only intact ( score >0 . 8 ) and full-length elements were analysed . For LINEs specifically , coverage was only represented on elements where an H3K27me3 peak was detected upon peak calling . ( B ) Composite profile of H3K27me3 coverage on and +/- 5kb around different transposon sequences in WT and Dnmt-tKO ( Dnmt3A/3B-dKO; Dnmt1 KD ) DNA methylation-deficient ES cells . ChIP-seq datasets were taken from Brinkman et al . , 2012 . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 018 H3K27me3 and H3K9me3 marks usually do not occur concomitantly ( Mikkelsen et al . , 2007 ) . We were therefore intrigued to observe that H3K27me3 and H3K9me3 were simultaneously enriched at transposon families of the A category in 2i+vitC medium ( Figure 4E and Figure 5A ) . To map the relative position of H3K9me3 and H3K27me3 , we determined their average profile over full-length individual elements of all transposon families , including their immediate genomic vicinity ( +/- 5 kb from the center of each element ) . Notably , H3K9me3 domains often spread out on adjacent genomic regions , whereas H3K27me3 was confined to transposon sequences ( Figure 5B , C and Figure 5—figure supplement 2A ) . It was previously described that H3K9me3 enrichment is restricted to the 5’ UTR of LINEs , while being evenly distributed along the entire length of ERVs ( Bulut-Karslioglu et al . , 2014; Pezic et al . , 2014 ) . In fact , we found this to be valid for specific ERVK elements only , namely IAPEz , IAPEy and MMERKV10C . Our most striking finding was the observation of a spatial separation between the two marks in category A transposons: H3K9me3 tended to occupy the 5’ end , while H3K27me3 preferentially targeted the 3’ end . This was observed for a significant proportion of LINEs and for several ERVs of the 1 or K classes ( MMERGLN , RLTR6 , MuRRS , RLTR10 ) ( Figure 5C for visual examples ) . However , some category A families showed H3K9me3 and H3K27me3 co-localization in their 5’ region ( VL30 , IAPEy , ETnERV ) . Notably , these transposon families harbor the greatest individual chromatin heterogeneity upon conversion ( Figure 5—figure supplement 1A ) : we presume that the metaplot figures likely represent an average among different individual elements and/or different cell populations . Having demonstrated that culture-induced DNA demethylation leads to increased and family-specific distribution of H3K27me3 on transposon sequences , we reasoned that similar features might occur upon genetically induced DNA demethylation . Fulfilling this prediction , analysis of available ChIP-seq datasets ( Brinkman et al . , 2012 ) showed concordant H3K27me3 patterns between serum-grown Dnmt-tKO ES cells and 2i+vitC-grown ES cells: entire coverage of MERVL sequences , 3’ localization in MMERGLN and RLTR6 , and 5’ localization in VL30 and ETnERV ( Figure 5—figure supplement 2B ) . These results support the notion that the pattern of H3K27me3 distribution on transposon sequences corresponds to an adaptation to the lack of DNA methylation . In summary , upon 2i+vitC-induced DNA demethylation , H3K27me3 and H3K9me3 can converge on category A transposon sequences , but they occupy different territories . IAPEz ( category B ) and MERVL ( C ) represent extreme cases of exclusivity , with the former being entirely covered by H3K9me3 , and the latter by H3K27me3 . Our study provides unprecedented evidence that H3K27me3 deposition at transposons is a default response to the absence of both DNA and H3K9 methylation . Our analysis reveals that H3K9me3 and H3K27me3 jointly or separately decorate transposon sequences upon DNA methylation loss . Through genetic analyses , we aimed to discern the functional relevance of these marks in controlling the three categories of transposons that we defined . Regarding H3K9me3-dependent pathways , we used CRISPR/Cas9 editing to generate a double-knockout ES cell line for the H3K9 trimethyltransferases , SUV39H1 and SUV39H2 ( Suv39h-dKO ) . Additionally we created haploinsufficient mutants for H3K9 trimethyltransferase SETDB1 ( Setdb1 +/- ) and its TRIM28 co-repressor ( Trim28 +/- ) ( Figure 6—figure supplement 1A , B ) ; complete SETDB1 or TRIM28 removal is not compatible with ES cell survival ( Dodge et al . , 2004; Rowe et al . , 2010 ) . The role of H3K27me3 was studied in mutant ES cells for the EED protein ( Eed-KO , Schoeftner et al . , 2006 ) , which is required for H3K27me3 catalysis by the Polycomb Repressive Complex 2 ( PRC2 ) ( Margueron and Reinberg , 2011 ) . Eed-KO ES cells experienced massive cell death around D8 of conversion , but slowly recovered in the following days ( data not shown ) . Finally , in order to study compensatory mechanisms between H3K9 and H3K27 pathways , we deleted EED in Suv39h-dKO ES cells , generating Suv39-Eed triple-knockout lines ( Suv39h-Eed-tKO ) ( Figure 6—figure supplement 1C ) . Suv39-Eed-tKO ES cells were viable with a reduced proliferation rate when grown in serum , but did not survive more than a week when converted to 2i+vitC ( data not shown ) . Nanostring quantification of transposon transcripts was performed upon serum to 2i+vitC transfer of these five cell lines , which , importantly , share the same J1 cell background . As representatives of transposon A category , young LINE1 elements maintain H3K9me3 while gaining H3K27me3 during medium conversion . Previous studies concluded that L1 repression in serum relies on SUV39H-dependent H3K9me3 ( Bulut-Karslioglu et al . , 2014 ) : through the analysis of our genetic mutants , we found this was the case for L1-A , and very modestly for L1-T elements ( Figure 6A and Figure 6—figure supplement 2A ) . Moreover , LINE1 expression was significantly upregulated at the end of the conversion of Setdb1 +/- ES cells , suggesting an important role for SETDB1-mediated repression upon DNA methylation loss . In regards to the role of polycomb , RNA-seq analysis of Eed-KO ES cells revealed that absence of polycomb had little effect on transposons in ES cells with a methylated genome ( at D0 of conversion ) . In contrast , many transposon families , including the different young LINE1 subtypes that belong to the A category , were strongly activated in fully 2i+vitC-converted Eed-KO cells ( Figure 7A , B and Figure 7—figure supplement 1 ) : polycomb-mediated silencing is therefore involved in controlling these transposons , specifically in absence of DNA methylation . Interestingly , ERV families that gain H3K27me3 uniquely in their 3’ end – such as MMERGLN , RLTR6 or RLTR10 – were not misregulated in absence of EED; this suggests that 3’ end deposition of H3K27me3 does not drive their transcriptional repression . 10 . 7554/eLife . 11418 . 019Figure 6 . Complex regulation of transposons by SUV39H , TRIM28 , SETDB1 and EED upon loss of DNA methylation . Expression levels in Suv39h-dKO , Trim28+/- , Setdb1+/- , Eed-KO , Suv39h-Eed-tKO and WT J1 ES cells for: ( A ) LINE1 ( category A ) ( B ) IAPEz ( category B ) ( C ) MERVL ( category C ) Expression levels were measured by Nanostring nCounter . Data are expressed as fold changes to WT D0 and represent mean ± SEM between two ( Trim28 and Setdb1 ) , three ( Suv39h and Suv39h-Eed-tKO ) , four ( Eed-KO ) and eight ( J1 ) biological replicates . *p<0 . 05 , **p<0 . 01 and ***p<0 . 001 ( unequal variances t-test between WT and mutants at a given day ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 01910 . 7554/eLife . 11418 . 020Figure 6—figure supplement 1 . Caracterization of genetic mutants . ( A ) Characterization of Suv39h-dKO ES cells: loss of H3K9me3 at pericentric heterochromatin assessed by immunofluorescence ( upper panel , compare with Figure 4—figure supplement 1A at D0 for WT ) and lack of SUV39H1 protein assessed by western blot ( lower panel ) . ( B ) Characterization of CRISPR/Cas9- generated Trim28+/- ES cells: western blot showing the level of reduction of TRIM28 protein in three independent clones . Clone F7 was used for subsequent experiments . ( C ) Characterization of CRISPR/Cas9- generated Suv39h-Eed-tKO ES cells: western blot showing lack of SUV39H1 , EED and H3K27me3 in independent clones . Clones D2 , F1 and A9 were used as biological replicates in subsequent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 02010 . 7554/eLife . 11418 . 021Figure 6—figure supplement 2 . Regulation of transposons by HDACs ( A ) Expression levels in Suv39h-dKO for L1-A and L1-T families measured by Nanostring nCounter . Data are expressed as fold changes to WT D0 and represent mean ± SEM between three ( Suv39h-KO ) and eight ( J1 ) biological replicates . ( B ) Expression levels in WT ES cells treated with the HDACs inhibitor Trichostatin A ( TSA ) for 24 hr measured by RT-qPCR . Values were normalized to Gapdh and Rplp0 . Data are expressed as the fold change to untreated cells and represent mean ± SEM between two ( untreated control ) and four biological replicates ( treated with 25 or 50 ng/mL of TSA , with undistinguishable effect between the two concentrations ) . *p<0 . 05 , **p<0 . 01 and ***p<0 . 001 ( unequal variances t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 02110 . 7554/eLife . 11418 . 022Figure 7 . Polycomb regulates transposons in absence of DNA methylation . ( A ) Volcano plot representation of up- and down-regulated transposons as measured by RNA-seq between WT and Eed-KO cells at D0 ( left ) and D13 of conversion ( right ) . Red dots indicate significantly misregulated repeats between two conditions ( fold change >2 and p-value <0 . 05 ) . ( B ) Heatmap representation and hierarchical clustering of expression changes for 69 transposon families at D0 , D6 and D13 , in WT and Eed-KO cells . Colors represent on a log2-scale the differential expression between a given condition and the average of the five conditions . *p<0 . 05 . ( C ) Expression of individual elements from different transposon families at D0 , D6 and D13 , in WT and Eed-KO cells , expressed in CPM . The black and red bars represent the median of the distribution , for WT and Eed-KO , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 02210 . 7554/eLife . 11418 . 023Figure 7—figure supplement 1 . RNA-seq of Eed-KO ES cells ( A ) Pearson correlation and hierarchical clustering between RNA-seq experiments . ( B ) Volcano plot representation of up- and down-regulated genes as measured by RNA-seq between WT and Eed-KO at D0 ( left ) and D13 ( right ) . Light and dark green dots represent genes with a H3K27me3 peak in promoter regions ( 2222 and 1870 genes at D0 and D13 , respectively ) . Red and dark green dots indicate significantly misregulated genes ( fold change >4 and p-value <0 . 01 ) . ( C ) Volcano plot representation of up- and down-regulated transposons as measured by RNA-seq between D0 and D13 in Eed-KO . Red dots indicate significantly misregulated repeats between two conditions ( fold change >2 and p-value <0 . 05 ) . RNA-seq mapping allowed multiple hits onto the genome . ( D ) Expression of individual elements from different transposon families at D0 , D6 and D13 , in WT and Eed-KO , in Count per Millions ( CPM ) . Each dot represents a single element . RNA-seq mapping allowed only unique hits in the reference genome; only elements with a minimum of 10 reads in at least one of the sample were conserved . The black and red bars represent the median of the distribution , for WT and Eed-KO , respectively . Analyzed numbers of distinct transposon copies per family appear into brackets . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 023 The category B of transposons is exemplified by IAPEz elements , which harbor exclusive H3K9me3 enrichment in all culture conditions . Although this profile would predict a continuous and exclusive dependence towards H3K9me3 upon medium adaptation , we observed complex behaviors in the different mutants ( Figure 6B ) . During conversion , Trim28+/- and Setdb1+/- cells showed enhanced IAPEz up-regulation and repression failure after D6 , in line with a major role of SETDB1-related H3K9me3 for controlling these elements in ES cells ( Matsui et al . , 2010 ) . However , SUV39H depletion led to an unexpected IAPEz suppression upon conversion . One possible explanation is that Suv39h-dKO cells have acquired long-term compensatory mechanisms that prevent transient IAPEz activation upon DNA methylation loss . Moreover , IAPEz elements were more strongly expressed in Eed-KO compared to WT cells during conversion ( Figures 6B and Figure 7A , B ) , which is at odds with their apparent lack of H3K27me3 enrichment in ChIP-seq data ( Figures 4E , F ) . Analysis of individual elements showed that this activation did not emanate from a few discrete elements but represented a general trend ( Figures 7C ) . These results could be due to indirect effects of the Eed deficiency . Finally , the H3K9me2/3- to H3K27me3-chromatin transition undergone by category C elements was very clearly illustrated in chromatin modifier mutants ( Figure 6C ) . MERVL elements are known to be primarily repressed by EHMT1/EHMT2-dependent H3K9me2 marks in serum ( Maksakova et al . , 2013 ) . We found that MERVL silencing also strongly relied on SUV39H-control in serum ( Figure 6C ) , which correlates with a modest H3K9me3 enrichment in our ChIP-seq data . SUV39H-dependent H3K9me3 became dispensable for MERVL silencing upon 2i+vitC conversion , and the switch towards H3K27me3 control was perfectly correlated with a 15-fold expression increase in Eed-KO cells ( Figure 6C ) . RNA-seq analysis indicates that MERVL was actually the most highly activated transposon in absence of EED , as a result of a collective upregulation of all individual elements ( Figure 7A–C ) . MERVL therefore represents a striking model of epigenetic switch from H3K9 to H3K27 methylation-based repression , which occurs subsequently to DNA methylation loss . Interestingly , we found that polycomb-dependent control can be implemented in absence of SUV39H , even in cells with high DNA methylation levels: while MERVL elements were upregulated by five-fold in Suv39h-dKO cells in serum grown-conditions , additional depletion of EED in Suv39-Eed-tKO cells led to a 20 fold increase compared to WT cells ( Figure 6C ) . Finally , MERVL silencing was previously shown to rely on other histone modifiers , such as histone deacetylases ( HDACs ) ( Macfarlan et al . , 2011 ) . Treatment with the HDAC inhibitor Trichostatin A ( TSA ) resulted in strong upregulation of MERVL in both serum and 2i+vitC conditions ( Figure 6—figure supplement 2B ) , suggesting that histone deacetylation is required for efficient silencing of MERVL , in presence or absence of DNA methylation .
Our study provides unprecedented insight into the dynamic adaptation of the pluripotent genome to a loss of DNA methylation-based control of transposons . This was achieved through detailed kinetic assessment of transcription and chromatin states during conversion of WT ES cells from serum to 2i+vitC media , as a way to reproduce the DNA methylation erasure that occurs during embryogenesis . Despite their heterogeneous origins and structures , we found that various transposon families residing in the mouse genome adopted a common regulatory fate: after an initial transcriptional burst , repression was re-established in a DNA methylation-independent manner . Distinct combinations of H3K9me2/3 and H3K27me3 were observed among transposon families , defining three functional categories of chromatin-based responses to DNA methylation loss: joint H3K9me3 and H3K27m3 ( A ) , H3K9me3-exclusive ( B ) , and H3K27me3-exclusive ( C ) ( Figure 8 ) . Importantly , Dnmt-tKO cells , which have endured long-term adaptation to a DNA methylation-free state , displayed similar transposon-specific chromatin patterns when grown in serum , which excludes a medium-related effect . In conclusion , our work revises the previous assumption that DNA methylation is dispensable for transposon silencing in ES cells; rather , we reveal here that various histone-based repression strategies are implemented upon DNA methylation loss , thereby safeguarding pluripotent cells against a multitude of heterogeneous transposon entities . 10 . 7554/eLife . 11418 . 024Figure 8 . Model for the acquisition of H3K27me3 at transposons during genome-wide demethylation . ( A ) Summary of chromatin and transcriptional changes during conversion from serum to 2i+vitC . DNA methylation and H3K9me2 are rapidly erased , H3K9me3 remains stable and H3K27me3 increases . Transposon expression peaks at D6 . ( B ) Model for the acquisition of H3K27me3 at transposons: upon loss of DNA methylation , H3K27me3 appears at GC-rich , H3K9me3 free-regions . Relative enrichments in H3K9me3 and H3K27me3 define three main types of repressive chromatin organization . Category A transposons are marked by H3K9me3 on their 5’ end and gain H3K27me3 on their 3’ region . Category B transposons are fully covered by H3K9me3 and do not gain H3K37me3 . Category C transposons lose H3K9me2 and H3K9me3 and acquire H3K27me3 decoration on their full length . At D6 of 2i+vitC conversion , the abundance of pluripotent transcription factors and the loose chromatin environment likely contribute to the burst of transposon expression . DOI: http://dx . doi . org/10 . 7554/eLife . 11418 . 024 Upon 2i+vitC-mediated DNA methylation loss , the repertoire of repressive histone marks is profoundly remodeled ( Figure 8A ) : H3K9me2 enrichment decreases , while H3K27me3 is enhanced and H3K9me3 levels are globally constant . Interestingly , repressive chromatin reorganization has also been cytologically observed in primordial germ cells ( PGCs ) undergoing genome-wide demethylation ( Hajkova et al . , 2008; Seki et al . , 2007 ) . Moreover , the relocalization of H3K27me3 at transposons and its co-occurrence with H3K9me3 was also reported in hypomethylated PGCs ( Liu et al . , 2014 ) . Our cellular system could therefore represent an adequate model to study in vivo events of chromatin reprogramming occurring at transposons upon DNA methylation loss . The persistence of H3K9me3 upon loss of DNA methylation highlights that DNA methylation does not exert significant control over H3K9me3-targeting of transposons in ES cells . Interestingly , we consistently observed that regions of persistent DNA methylation ( RMRs ) coincide with high H3K9me3 enrichment on transposon sequences in fully 2i+vitC-converted cells , e . g . on the 5’ end of LINE1 elements and throughout the length of ERVK elements . This supports previous evidence that H3K9me3 can confer protection against DNA demethylation ( Leung et al . , 2014 ) . Inversely , the rapid disappearance of H3K9me2 upon serum to 2i+vitC conversion could reflect a direct role of DNA methylation in the maintenance of these marks . Accordingly , H3K9me2 reduction was also observed in DNA methylation-free Dnmt-tKO ES cells grown in serum ( data not shown ) . Coupled losses of DNA methylation and H3K9me2 have also been previously reported in vivo , during normal primordial germ cell development ( Hajkova et al . , 2008; Seki et al . , 2005 ) and in DNA methylation-deficient spermatocytes ( Zamudio et al . , 2015 ) . A mechanism of H3K9me2 methyltransferase recruitment via DNA methylation has been resolved in plants ( Du et al . , 2015 ) ; the evolution of an analogous mechanism in mammals should be explored . Of particular importance to this study is our observation of an epigenetic switch occurring between DNA methylation- and H3K27me3-based control . H3K27me3 is barely detectable at transposons in DNA hypermethylated WT ES cells grown in serum; in contrast , transposons accumulate H3K27me3 in both 2i+vitC-converted cells and in serum-grown Dnmt-tKO cells . This is in line with the prevailing notion that DNA methylation and H3K27me3 are mutually exclusive genome-wide and that DNA methylation antagonizes H3K27me3 deposition ( Brinkman et al . , 2012; Jermann et al . , 2014; Tanay et al . , 2007 ) . Saliently , this raises the question as to how transposons acquire H3K27me3 upon DNA methylation loss . In mammalian genomes , polycomb is typically targeted to unmethylated GC-rich gene promoters ( Jermann et al . , 2014; Mendenhall et al . , 2010 ) . Notably , transposon sequences have a GC content superior to the genome average ( Figure 5—figure supplement 2B ) : this signature could be sufficient to attract polycomb-mediated H3K27me3 deposition in the absence of DNA methylation . Intermediate methyl-sensitive DNA binding proteins may be involved: the BEND3 protein was recently identified as a sensor of DNA methylation states at pericentromeric repeats , recruiting polycomb-dependent H3K27me3 marks in Dnmt-tKO ES cells ( Saksouk et al . , 2014 ) . Interestingly , we also observed H3K27me3 relocalization towards pericentromeric repeats in hypomethylated 2i+vitC ES cells . Similarly , the H3K36 demethylase KDM2B , which targets unmethylated CpG-rich sequences , was shown to recruit PRC1 , potentially leading to H3K27me3 deposition through PRC2 recruitment ( Blackledge et al . , 2014; Farcas et al . , 2012 ) . Comparable mechanisms might be at play for the recruitment of H3K27me3 at hypomethylated transposons , involving BEND3 , KDM2B and/or other methyl-sensitive DNA binding proteins . Thus , based on previous observations , we posit that H3K27me3 invades the transposon space left unmarked by DNA methylation upon 2i+vitC conversion . Moreover , we provide evidence that the three possible chromatin configurations that the different transposon families adopt are further determined by H3K9me3 occupancy ( Figure 8B ) . Mutual exclusion between H3K9me3 and H3K27me3 marks has been previously documented at gene promoters and pericentromeric repeats ( Mikkelsen et al . , 2007; Peters et al . , 2003 ) , but not at transposons . We found that category B transposons , which constantly maintain H3K9me3 marks throughout their entire length , do not acquire H3K27me3-based chromatin even though they lose DNA methylation . In contrast , category C transposons , exemplified by MERVL elements , become strongly enriched for H3K27me3 as H3K9me2/3 depletes during medium conversion . Finally , category A elements provide a striking illustration of the physical segregation of H3K9me3 and H3K27me3: as H3K9me3 constitutively marks the 5’ end of this transposon category , only their 3’ end is accessible to H3K27me3 deposition upon DNA methylation loss . The presence of H3K27me3 at the 3’ end of transcription units has not been described before and its functional significance was unknown . Our transcriptome analysis of 2i+vitC-converted Eed-KO cells reveals that 3’ end enrichment of H3K27me3 does not confer transcriptional repression of transposons of the A category: it may rather represent a passive response to the lack of DNA methylation and H3K9me3 at this position . The main message conveyed by our work is that compensatory histone-based mechanisms ensure transposon silencing when DNA methylation-based control is alleviated in ES cells . We cannot rule out that other mechanisms-such as small RNA-based post-transcriptional repression-could also participate to transposon control . Importantly , genetic analyses globally confirmed the functionality of the chromatin patterns that we identified . In particular , H3K27me3-dependency was very well illustrated by the failure to repress several transposon families – in particular MERVL and LINE1– in Eed-KO ES cells undergoing medium-based DNA methylation loss . However , the transposon category B ( IAPEz ) , which remains enriched for H3K9me3 throughout media conversion , gave complex , disparate phenotypes in the mutants of the different H3K9me3 pathways . While these elements failed to be repressed in Setdb1 and Trim28-deficient ES cells , the complete suppression of IAPEz reactivation in Suv39-dko cells was unexpected . We suspect that alternative repressive processes likely obscure IAPEz transcriptional responses to DNA methylation loss in this mutant . This is akin to Dnmt-tKO cells , which also exhibit global transposon repression . Thus , our analyses highlight the possible unexplained phenotypes of mutant cells that have adapted to long-term chromatin-based deficiencies . Finally , one important point to raise is that the epigenetic switch from a DNA methylation-dependent to -independent mode of transposon silencing is not perfectly synchronized: ES cells experience an acute burst of transposon expression at D6 of medium conversion . At this time point , we showed that DNA methylation has been mostly erased but H3K27me3 patterns have not been established yet . Interestingly , the stability of H3K9me3 marks at category A and B transposons is not sufficient to ensure their continuous silencing upon conversion . This may imply that H3K9me3 readers are transiently deficient in this system . The lag between DNA methylation loss and subsequent implementation of histone-based repression could create an opportunistic window for transposon reactivation , provided that adequate transcription factors are available . Several studies have previously pointed out that transposons are enriched in pluripotency transcription factor binding motifs ( Kunarso et al . , 2010; Wang et al . , 2014 ) , in particular for NANOG and OCT4 , and that upregulation of these transcription factors was sufficient to promote transposon expression ( Grow et al . , 2015 ) . We propose that the simultaneous disappearance of DNA methylation marks and increased availability of pluripotency activators create favorable conditions to transposon expression at D6 of serum to 2i+vitC conversion ( Figure 8B ) . After a brief silencing release , functional repressive chromatin is recovered , in an H3K9me3 and/or H3K27me3-dependent manner . Notably , we repeatedly observed a peak of massive cell death of H3K27me3-deficient Eed-KO ES cells between D6 and D10 of medium conversion , when DNA methylation has mostly disappeared . This phenotype was even exacerbated in Suv39h-Eed-tKO ES cells , which did not survive after a week of medium conversion . These observations support the critical role for H3K27me3 in supplementing DNA methylation-based control in ES cells .
J1 and Dnmt-tKO ES cells were a gift from M . Okano ( Tsumura et al . , 2006 ) . E14 ES cells were kindly provided by E . Heard . WT TT2 and Ehmt2-KO ES cells ( Tachibana et al . , 2002 ) , and Eed-KO ( Schoeftner et al . , 2006 ) ( on a J1 background ) were gifts from Y Shinkai and A Wutz , respectively . Trim28+/- , Setdb1+/- and Suv39-dKO were generated from J1 ES cells using CRISPR/Cas9 editing . Briefly , guide-RNAs specific to the target sequences were designed using the online CRISPR Design Tool ( Hsu et al . , 2013 and Supplementary file 2D ) and incorporated into the X330 backbone ( Cong et al . , 2013 ) . Five millions J1 ES cells grown in serum were transfected with 1–3 μg of plasmid using Amaxa 4d nucleofector ( Lonza ) and plated at a low density . Individual clones were picked and screened by PCR; mutated alleles were confirmed by Sanger sequencing . Suv39h-dKO cells were obtained by creating a frame-shift in Suv39h1 exon 4 and by deleting Suv39h2 exon 4; Trim28+/- cells were generated by deleting exon 3; Suv39h-Eed-tKO were obtained by deleting Eed exon 6 in Suv39h-dKO cells; Setdb1 +/- were generated by creating a frameshift in exon 16 . ES cells were grown in two different media , serum and 2i , defined as follow . Serum: Glasgow medium ( Sigma ) , 15% FBS ( Gibco ) , 2 mM L-Glutamine , 0 . 1 mM MEM non essential amino acids ( Gibco ) , 1 mM sodium pyruvate ( Gibco ) , 0 . 1 mM β-mercaptoethanol , 1000 U/mL leukemia inhibitory factor ( LIF , Miltenyi ) ; 2i: 50% neurobasal medium ( Gibco ) , 50% DMEM/F12 ( Gibco ) , 2 mM L-glutamine ( Gibco ) , 0 . 1 mM β-mercaptoethanol , Ndiff Neuro2 supplement ( Milipore ) , B27 serum-free supplement ( Gibco ) , 1000 U/mL LIF , 3 μM Gsk3 inhibitor CT-99021 , 1 μM MEK inhibitor PD0325901 . Vitamin C ( Sigma ) was added at a concentration of 100 ug/mL ( Blaschke et al . , 2013 ) . When in serum , J1 , Dnmt-tKO , E14 , and all CRISPR-generated mutant ES cells were grown in feeder-free conditions on gelatin-coated plates . TT2 , Ehmt2-KO were cultured on a monolayer of mitomycin C-treated mouse embryonic fibroblasts . ES cells were passaged with TrypLE Express Enzyme ( Life Technologies , Carlsbad , CA ) . All 2i ES cells were grown in gelatin-coated plates and passaged every two or three days with Accutase ( Life Technologies ) . Trichostatin A was added for 24 hr at concentration of 25 or 50 ng/mL Mycoplasma-free status was assessed using VenorGeM Classic mycoplasma detection kit ( Minerva Biolabs ) . Genomic DNA was isolated using the GenElute Mammalian Genomic DNA Miniprep Kit ( Sigma ) with RNase treatment . Global CpG methylation levels were assessed using LUminometric Methylation Assay ( LUMA ) as described previously ( Karimi et al . , 2011a; Richard Pilsner et al . , 2010 ) . Briefly , 500 ng of genomic DNA was digested with MspI/EcoRI and HpaII/EcoRI ( NEB ) in parallel reactions . HpaII is a methylation-sensitive restriction enzyme and MspI is its methylation insensitive isoschizomer . EcoRI is included as an internal reference . The overhangs created by the enzymatic digestion were quantified by Pyrosequencing ( PyroMark Q24 , Qiagen ) with the dispensation order: GTGTGTCACACAGTGTGT . Global CpG methylation levels were calculated from the peak heights at the position 7 , 8 , 13 , 14 as follows: 1-sqrt ( [p8*p14/p7*p13]HpaII /[p8*p14/p7*p13]MspI ) CpG methylation at specific loci was assessed by bisulfite-pyrosequencing using the Imprint DNA modification Kit ( Sigma ) for conversion . PCR and sequencing primers ( Supplementary file 2D ) were designed with the PyroMark Assay Design Software and quantification of DNA methylation was performed according to the recommended protocol . Whole-Genome Bisulfite Sequencing libraries were prepared from 50ng of bisulfite-converted genomic DNA using the EpiGnome/Truseq DNA Methylation Kit ( Illumina ) following the manufacturer instructions . Sequencing was performed in 100 pb paired-end reads at a 30X coverage using the Illumina HiSeq2000 platform ( Supplementary file 1 ) . Total RNA was extracted using Trizol ( Life Technologies ) . cDNAs were prepared after DNase treatment ( Turbo DNase , Ambion ) using random priming with Superscript III ( Life Technologies ) . Real-time quantitative PCR was performed using the SYBR Green Master Mix on the Viia7 thermal cycling system ( Applied Biosystem ) . Relative expression levels were normalized to the arithmetic mean of the housekeeping genes Gapdh and Rplp0 and to WT-D0 with the ΔΔCt method . Primers are given in Supplementary file 2D . Nanostring nCounter quantification was performed using 100ng of total RNA per sample on a custom expression Codeset ( target sequences in Supplementary file 2D ) . Actin , Ppia , Gapdh and Rplp0 were used for normalization . Data are presented as the fold change compared to WT-D0 . Expression data for the different mutants are presented next to WT data that were processed on the same Nanostring run . The same WT data can be used in several figures . When necessary and in order to calculate mean and standard error of the mean between replicates every two days , we extrapolated linearly the expression value of a given day using data of immediately adjacent time points ( for both RT-qPCR and Nanostring ) . RNA-seq libraries were prepared from 500ng of DNase-treated RNA with the TruSeq Stranded mRNA kit ( Illumina ) . Sequencing was performed in 100pb paired-end reads using the Illumina HiSeq2000 platform ( Supplementary file 1 ) . Cells were cross-linked directly in culture plates with 1% formaldehyde ( culture medium supplemented with 1% formaldehyde , 0 . 015 M NaCl , 0 . 15 mM EDTA , 0 . 075 mM EGTA , 15 mM Hepes pH 8 ) . After quenching with 0 . 125 M glycine , cells were washed in PBS and pelleted . Cells were then incubated at 4°C for 10 min in buffer 1 ( Hepes-KOH pH 7 . 5 50 mM , NaCl 140 mM , EDTA pH 8 . 0 1 mM , glycerol 10% NP-40 0 . 5% , Triton X-100 0 . 25% and the protease inhibitors: PMSF 1 mM , Aprotinin 10 μg/ml , leupeptin 1 μg/ml and pepstatin 1 μg/ml ) , then at room temperature for 10 min in buffer 2 ( NaCl 200 mM , EDTA pH 8 . 0 1 mM , EGTA pH 8 . 0 0 . 5 mM and 10 mM Tris pH 8 and the same protease inhibitors as buffer 1 ) and finally resuspended in buffer 3 ( EDTA pH 8 . 0 1 mM , EGTA pH 8 . 0 0 . 5 mM , Tris pH8 10 mM , N-lauroyl-sarcosine 0 . 5%; protease inhibitors as buffer 1 ) . Chromatin was sonicated with a Bioruptor ( Diagenode ) to reach a fragment size around 200 bp . Chromatin corresponding to 10 μg of DNA was incubated overnight at 4°C with 3–5 μg of antibody in incubation buffer ( buffer 3 supplemented with 0 . 5 volume of 3% Triton , 0 . 3% sodium deoxycholate , 15 mM EDTA; protease inhibitors ) . A fraction of chromatin extracts ( 5% ) were taken aside for inputs . Antibody-bound chromatin was recovered using Protein G Agarose Columns ( Active Motif ) . Briefly , the antibody-chromatin mix was incubated in the column for 4 hr , washed eight times with modified RIPA buffer ( Hepes pH7 . 6 50 mM , EDTA pH 8 . 0 10 mM , sodium deoxycholate 0 . 7% , NP-40 1% , LiCl 500 mM , PMSF 1 mM , 1 μg/ml leupeptin and 1 μg/ml pepstatin ) , and washed one last time with TE-NaCl ( 50mM Tris pH 8 . 0 , 10 mM EDTA , 50 mM NaCl ) . Chromatin was eluted with pre-warmed TE-SDS ( 50mM Tris pH 8 . 0 , 10 mM EDTA , 1% SDS ) . ChIP-enriched sample and inputs were then reverse cross-linked at 65°C overnight and treated with RNase A and proteinase K . DNA was extracted with phenol/chloroform/isoamyl alcohol , precipitated with glycogen in sodium acetate and ethanol and finally resuspended in TE . Enrichment compared to input was analyzed by qPCR . A quantity of 20 ng of ChIP- or input-DNA were used for ChIP-seq . Remaining large DNA fragments were first eliminated using SPRIselect beads ( Beckman Coulter ) and libraries were prepared using the TruSeq ChIP Sample Prep kit ( Illumina ) . Sequencing was performed in 50pb paired-end reads using the Illumina HiSeq2000 platform ( Supplementary file 1 ) . qPCR primers and antibodies are listed in Supplementary file 2D , E . To prepare total protein extracts , cells were resuspended in BC250 lysis buffer ( 25 mM Tris pH 7 . 9 , 0 . 2 mM EDTA , 20% Glycerol , 0 . 25 M KCl and protease inhibitor coktail from Roche ) , sonicated and centrifuged to pellet debris . To prepare nuclear protein extracts , cells were incubated for 10 min on ice in buffer A ( Hepes pH 7 . 9 10 mM , MgCl2 5 mM , Sucrose 0 . 25 M , NP40 0 . 1% , DTT 1mM and protease inhibitors ) and centrifuged . The pellet was resuspended in buffer B ( Hepes pH 7 . 9 25 mM , glycerol 20% , MgCl2 1 . 5 mM , EDTA 0 . 1 mM , NaCl 700 mM , DTT 1 mM and protease inhibitors ) , sonicated and centrifuged to pellet debris . Total and nuclear proteins were quantified by Bradford assay . Proteins ( 10–20 μg per gel lane ) were separated by electrophoresis in 8–15% poly-acrylamide gels and transferred onto nitrocellulose membranes using the Trans-Blot turbo transfer system ( Biorad ) . After incubation with primary antibodies and HRP-conjugated secondary antibodies , signal was detected using ECL prime kit ( Amersham ) and ImageQuant Las-4000 mini biomolecular Imager . Antibodies are listed in Supplementary file 2E . Cells were harvested with Trypsin or Accutase , resuspended in PBS and plated for 10 min on Poly-L-Lysine-coated glass cover slips . Cells were first fixed with 3% paraformaldehyde for 10 min at room temperature , then rinsed three times with PBS and permeabilized for 4 min with 0 . 5X Triton on ice . After blocking in 1% BSA for 15 min , samples were incubated at room temperature for 40 min with primary antibodies , 45 min with secondary antibodies and 3 min in 0 . 3 μg/mL DAPI . Slides were mounted with Prolong Gold mounting media ( Invitrogen ) . Images were obtained with an Upright Widefield microscope ( Leica ) or a Zeiss LSM700 inverted confocal microscope . Quantification of immunofluorescence intensity in individual cells was performed using custom ImageJ and R scripts . Between 2000 and 5000 cells were analyzed per sample . Antibodies are listed in Supplementary file 2E . Cells were cultured for two hours with 0 . 04 μg/mL colchicine and harvested by trypsinization . Cell pellets were incubated in hypotonic buffer ( 15% FBS in water ) for 7 min at 37°C and fixed with 66% acetic acid/33% ethanol . After centrifugation , cells were resuspended in 1 . 5 mL fixative and dropped from ~1 m height onto glass slides . Slides were dried and DNA was stained with DAPI . Chromosomes were counted with an Upright Widefield microscope ( Leica ) . Around 20 cells were analyzed per cell line . Absolute copy numbers of IAP and LINE1 were calculated by qPCR by establishing standard curves plotting absolute Ct values of genomic DNA against serial dilutions of PCR targets cloned into the pCR2 . 1-TOPO vector ( Life Technologies ) , as described in Zamudio et al . , 2015 . As described in Bailly-Bechet et al . , 2014 , a dictionary was constructed for LTR retrotransposons that associated elements corresponding to the internal sequence and those corresponding to LTR sequences . With the latter and the RepeatMasker database , fragments of transposable elements corresponding to the same copy were merged . Divergence , deletion and insertion percentages were recalculated from RepeatMasker and an integrity score for each transposon were calculated as follow: score = 1-average ( %divergence , % deletions , % insertions ) Whole-genome bisulfite sequencing reads generated in this study or recovered from available datasets were treated as follow . The first eight base pairs of the reads were trimmed using FASTX-Toolkit v0 . 0 . 13 ( http://hannonlab . cshl . edu/fastx_toolkit/index . html ) . Adapter sequences were removed with Cutadapt v1 . 3 ( https://code . google . com/p/cutadapt/ ) and reads shorter than 16 bp were discarded . Cleaned sequences were aligned onto the Mouse reference genome ( mm10 ) using Bismark v0 . 12 . 5 ( Krueger and Andrews , 2011 ) with Bowtie2-2 . 1 . 0 ( Langmead and Salzberg , 2012 ) and default parameters . Only reads mapping uniquely on the genome were conserved . Methylation calls were extracted after duplicate removal . Only CG dinucleotides covered by a minimum of 10 reads were conserved for the rest of the analysis . The R-package Methylkit v0 . 9 . 2 ( Akalin et al . , 2012 ) was used to provide Pearson’s correlation scores between samples . To analyze the distribution of CpG methylation in different genomic compartments , the mouse genome was divided into different partitions . The RefSeq gene annotation and the RepeatMasker database were downloaded from UCSC table browser and used for transcript and repeat annotations , respectively . Promoters were defined as the -1 kb to +100 pb region around transcription start sites . CpG islands ( CGIs ) were defined as in Illingworth et al . , 2010 . Intergenic partitions were defined as genomic regions that did not overlap with promoters , CGI , exons , introns or repeats . Whole-genome mapping of CpG methylation was then intersected with the different genomic compartments using Bedtools ( Quinlan and Hall , 2010 ) . Average CpG methylation on individual transposons was extracted from RepeatMasker with Bedtools , average CpG methylation in the different transposon families was calculated and plotted using R . Heatmap for average CpG methylation in Imprinted control regions ( ICRs ) was generated similarly after retrieving ICR genomic coordinates from the WAMIDEX database ( Schulz et al . , 2008 ) . Residually methylated regions ( RMRs ) in 2i+vitC samples were identified using the MethPipe pipeline ( Song et al . , 2013 ) with default parameters . RMRs located less than 1 kb from each others were concatenated . In order to quantify gene expression , Paired-end 2x100 bp reads were mapped onto mm10 using Tophat v2 . 0 . 6 and RefSeq gene annotation ( Kim et al . , 2013 ) allowing five mismatches . Gene-scaled quantification was performed with HTSeq v0 . 6 . 1 ( Anders et al . , 2014 ) . In order to quantify transposon expression , reads mapping to ribosomal RNA ( rRNA ) sequences ( GenBank identifiers: 18S NR_003278 . 3 , 28S NR_003279 . 1 , 5S D14832 . 1 , 5 . 8S K01367 . 1 ) were first removed with Bowtie v1 . 0 . 0 allowing three mismatches . The rRNA-depleted libraries were then mapped onto mm10 using Bowtie v1 . 0 . 0 allowing zero mismatch and 10000 best alignments per read . Exonic reads were removed . In order to count reads mapping to transposable elements , reads were weighted by the number of mapping sites and each library was intersected with the reconstructed RepeatMasker annotation , conserving only reads overlapping at least at 80% with a given transposon . For each library , read counts for genes and transposons were combined into a single table . TMM normalization from the edgeR package v3 . 6 . 2 ( Robinson and Oshlack , 2010 ) was first applied . As described in the guideline of limma R-package v3 . 20 . 4 , normalized counts were processed by the voom method ( Law et al . , 2014 ) to convert them into log2 counts per million with associated precision weights . The differential expression was estimated with the limma package . Genes and transposons were called differentially expressed when two criteria were met: 1 ) the fold-change between two conditions was higher than four and two , respectively , and 2 ) the adjusted p-value using the Benjamini Hochberg procedure was below 0 . 05 . For the analysis of RNA-seq libraries with uniquely mapped reads , the mapping was performed as previously with Bowtie v1 . 0 . 0 , except that only uniquely mapping reads were conserved . Read counts on individual reconstructed element were quantified using HTSeq v0 . 6 . 1 . Only elements with at least 10 reads in at least one sample were conserved for further analysis and read counts were subsequently normalized by the library size . Normalized read counts for individual elements belonging to different families were then plotted using custom R script . Tracks were created using HOMER software v4 . 7 ( Heinz et al . , 2010 ) . In order to identify and characterize chimeric transcripts , reads were mapped onto mm10 using Tophat v2 . 0 . 6 , without providing a gene annotation . Cufflinks v2 . 2 . 1 ( Trapnell et al . , 2010 ) was used to reconstruct the transcriptome and quantify the different isoforms . Transcripts were considered chimeric when the first exon overlapped with a transposon annotated in Repeatmasker and one of the other exon was annotated in RefSeq . Paired-end 2x50bp reads were mapped onto mm10 using Bowtie v1 . 0 . 0 allowing 3 mismatches . Reads mapping to multiple locations were randomly allocated . Duplicate reads were removed using Picard v1 . 65 ( http://broadinstitute . github . io/picard/ ) . Tracks were created using HOMER software v4 . 7 ( Heinz et al . , 2010 ) and Peak calling was performed with MACS2 v2 . 0 . 10 ( Zhang et al . , 2008 ) using the broad option and a 5% FDR threshold . Detected peaks were annotated using RefSeq and RepeatMasker databases . In order to construct the heatmap and the scatter plots , the total number of read counts for every annotated transposable element was computed using Bedtools and the reconstructed RepeatMasker annotation . Enrichment was normalized by the size of the element and Input data . Metaplots for average enrichment and GC content on and around different transposons were obtained using HOMER V4 . 7 . Only full-length ( >6 kb ) and intact ( integrity score >0 . 8 ) elements were used for the metaplots . | Transposons are sequences of DNA with the ability to mobilize and jump from one position to another . In the human genome , the number of transposons far surpasses the number of genes . Furthermore , while transposons have been beneficial for the evolution of the human genome , they can also alter genes and cause cancer and genetic diseases . The danger posed by transposons has led to numerous mechanisms to keep them under control . In particular , a natural biochemical modification of the DNA molecule called “DNA methylation” plays an important role in keeping transposons inactive or silent . However , during the early development of an embryo , the DNA methylation marks are erased throughout the entire genome . This provides an opportunity for transposons to be active , and it is not clear how the genome manages to control transposons in the absence of this essential protective mark . Walter et al . have now investigated this process by using a cell type that mimics the loss of DNA methylation that occurs during embryonic development – mouse embryonic stem cells grown in the laboratory . The experiments revealed that when DNA methylation is lost progressively , the transposons are reactivated at first but are later put back into a silent mode by alternative mechanisms . These mechanisms compensate for the disappearance of DNA methylation by encouraging the DNA around transposons to become compacted , which prevents the transposons from moving . Further analysis revealed that the different families of transposons that exist in the mouse genome can be classified into three groups , and in each group different proteins ensure the transposons remain repressed in the absence of DNA methylation . Together these findings reveal that multiple pathways cooperate to protect the genome against the activity of a variety of transposons . Finally , in mammals , DNA methylation is naturally erased both during the formation of sperm and egg cells and in the early embryo . As such , it will be important to verify whether the mechanisms discovered in the laboratory-grown cells also tame transposons during these critical developmental periods . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"chromosomes",
"and",
"gene",
"expression"
] | 2016 | An epigenetic switch ensures transposon repression upon dynamic loss of DNA methylation in embryonic stem cells |
We use psychophysics and MEG to test how sensitivity to input statistics facilitates auditory-scene-analysis ( ASA ) . Human subjects listened to ‘scenes’ comprised of concurrent tone-pip streams ( sources ) . On occasional trials a new source appeared partway . Listeners were more accurate and quicker to detect source appearance in scenes comprised of temporally-regular ( REG ) , rather than random ( RAND ) , sources . MEG in passive listeners and those actively detecting appearance events revealed increased sustained activity in auditory and parietal cortex in REG relative to RAND scenes , emerging ~400 ms of scene-onset . Over and above this , appearance in REG scenes was associated with increased responses relative to RAND scenes . The effect of temporal structure on appearance-evoked responses was delayed when listeners were focused on the scenes relative to when listening passively , consistent with the notion that attention reduces ‘surprise’ . Overall , the results implicate a mechanism that tracks predictability of multiple concurrent sources to facilitate active and passive ASA .
Natural scenes are highly structured , containing statistical regularities in both space and time and over multiple scales ( Julesz , 1981; Portilla and Simoncelli , 2000; Geisler , 2008; McDermott et al . , 2013; Theunissen and Elie , 2014 ) . A growing body of work suggests that the human brain is sensitive to this statistical structure ( Rao and Ballard , 1999; Näätänen et al . , 2001; Bar , 2004; Oliva and Torralba , 2007; Costa-Faidella et al . , 2011; Garrido et al . , 2013; Okazawa et al . , 2015; Barascud et al . , 2016 ) and uses it for efficient scene analysis ( Winkler et al . , 2009; Andreou et al . , 2011; Bendixen , 2014 ) . Uncovering the process by which this occurs , and how sensory predictability interacts with attention , is a key challenge in sensory neuroscience across modalities ( Winkler et al . , 2009; Summerfield and de Lange , 2014; Summerfield and Egner , 2016 ) . The current state of understanding is limited by at least two factors: ( 1 ) most studies of sensory predictability and its effects on behavior have used slow presentation rates thus enabling conscious reflection of stimulus expectancy . As a consequence , relatively little is known about the neural underpinning of predictability processing on the rapid time scales relevant to perception of natural objects . ( 2 ) In most cases , predictability has been studied when participants attend to a single object ( Murray et al . , 2002; Arnal et al . , 2011; Kok et al . , 2012; Chennu et al . , 2013; Bendixen , 2014 ) – a far cry from the complex scenes in which we normally operate . We therefore do not understand whether/how statistical structure is extracted from complex , crowded scenes . The present work addresses both of these issues in the context of an auditory scene . To understand how statistical structure facilitates perceptual analysis of acoustic scenes , we use an ecologically relevant paradigm ( change detection ) that captures the challenges of natural listening in crowded environments ( Cervantes Constantino et al . , 2012; Sohoglu and Chait , 2016 ) . In this paradigm , listeners are presented with multiple concurrent acoustic sources and on occasional trials , a new source appears partway into the ongoing scene ( see Figure 1A ) . By varying the temporal patterning of scene sources , we can create conditions in which the scenes are characterized by statistically regular or random structure and measure the effect of this manipulation on listeners’ ability to detect the appearance of new sources within the unfolding soundscape . 10 . 7554/eLife . 19113 . 003Figure 1 . Stimuli and behavior . ( A ) Examples of REG and RAND scenes . The plots represent ‘auditory’ spectrograms , equally spaced on a scale of ERB-rate ( Moore and Glasberg , 1983 ) . Channels are smoothed to obtain a temporal resolution similar to the Equivalent Rectangular Duration ( Plack and Moore , 1990 ) . Black arrows indicate appearing sources . In these examples , the appearing source is temporally regular . The stimulus set also included scenes in which the appearing source was temporally random ( see Materials and methods ) . ( B ) Behavioral results ( d’ and detection time ) as a function of scene temporal structure ( REG versus RAND ) . These are shown for each type of scene change ( when the appearing source was temporally regular or when random ) . Error bars represent within-subject standard error of the mean ( SEM; Loftus and Masson , 1994 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19113 . 003 The behavioral response pattern reveals that perceptual analysis of such scenes is enhanced by the presence of regular statistical structure , as assessed by listeners’ ability to detect source appearance . One possible explanation for this effect is that neural responses to regularly repeating scene components adapt ( decrease over time ) more than to random components . Indeed , perceptual influences of statistical structure have often been attributed to neural adaptation ( e . g . 'stimulus specific adaptation'; May et al . , 1999; Jääskeläinen et al . , 2004; Haenschel et al . , 2005; Costa-Faidella et al . , 2011; ; Khouri and Nelken , 2015 ) . Accordingly , the relative change in neural response to a new spectral component ( that is , the appearing source ) will be larger and thus more detectable in regular versus random scenes ( Summerfield et al . , 1987; Hartmann and Goupell , 2006; Erviti et al . , 2011 ) . By this account , statistical structure does not modulate the magnitude of neural response to a new event per se . Rather , improved detection is attributed exclusively to decreased neural responses occurring before the appearance of the new source . Indeed , in a mismatch negativity paradigm , Costa-Faidella ( 2011 ) demonstrated that neural responses to repeating ( ‘standard’ ) tones adapt more in temporally regular than random sequences without accompanying changes in response to new ( ‘deviant’ ) tones ( see also Schwartze et al . , 2011 , 2013; Tavano et al . , 2014 ) . However , other work has shown that statistically regular patterns can be associated with increased neural responses ( Haenschel et al . , 2005; Kok et al . , 2012; Chennu et al . , 2013; Hsu et al . , 2014; Kouider et al . , 2015; Barascud et al . , 2016 ) . These effects have been interpreted to reflect a mechanism that tracks the level of predictability or ‘precision’ of the sensory input , a measure inversely related to the uncertainty or entropy of a variable . This mechanism is hypothesized to enable the up-regulation of processing for information that is reliable and likely to indicate genuine events in the environment ( Feldman and Friston , 2010; Zhao et al . , 2013; Auksztulewicz and Friston , 2015; Barascud et al . , 2016 ) . Importantly , the up-regulation of neural processing in these accounts is hypothesized to lead to increased neural responses for regular scenes before source appearance , as well as an increased error ( ‘surprise’ ) response evoked by the new source . In the current study we adjudicate between adaptation and precision accounts using magnetoencephalography ( MEG ) recordings of brain activity . Given the ongoing debate about how the neural influence of statistical regularity might depend on attention ( Jones and Boltz , 1989; Näätänen et al . , 2001; Summerfield and Egner , 2009; Winkler et al . , 2009; Feldman and Friston , 2010; Kok et al . , 2012; Bendixen , 2014; Schröger et al . , 2015 ) , we do this in the context of passive listening ( listeners engaged in an unrelated visual task ) as well as active listening ( listeners actively detecting source appearance ) . Our results provide evidence in support of precision accounts: we show that brain responses to ongoing acoustic scenes , and to new sources appearing within those scenes , increase in the presence of regular statistical structure . Strikingly , the effect of regularity on appearance detection is delayed when listeners are actively focused on the scenes rather than listening passively . This latter finding suggests ( somewhat counter intuitively ) that active listening can counteract the influence of regularity but is consistent with attention acting to reduce ‘surprise’ ( Spratling , 2008; Chennu et al . , 2013 ) .
Listeners’ source appearance detection performance in the Active group is shown in Figure 1B . Listeners were more accurate and quicker to detect source apperance when the scene structure was temporally regular ( REG ) versus random ( RAND; d’ F ( 1 , 12 ) = 100 . 7 , p<0 . 001; detection times F ( 1 , 12 ) = 17 . 61 , p<0 . 01 ) . This effect occurred independently of the temporal structure of the appearing component ( d’ F ( 1 , 12 ) = 0 . 075 , p=0 . 789; detection times F ( 1 , 12 ) = 4 . 23 , p=0 . 062 ) . Additionally , listeners were quicker ( by ~27 ms ) to detect source appearance when it was temporally regular ( detection times F ( 1 , 12 ) = 5 . 70 , p<0 . 050 ) , although this effect did not extend to d’ ( F ( 1 , 12 ) = 2 . 29 , p=0 . 156 ) . Thus temporally regular scenes are associated with enhanced detection performance and in a manner independent of the temporal structure of the appearing source . Overall , the mean hit rate was high ( mean = 76 . 1% , ranging from 57 to 97% across listeners ) and mean false alarm rate low ( mean = 6 . 25% , ranging from 0 to 18 . 8% ) . The appearance-evoked response is shown in Figure 3 . Note that these data have been baseline corrected relative to the 200 ms period prior to the appearance event . Thus , effects reported in this section are specific to the appearance-evoked response and not merely a reflection of the pre-existing REG versus RAND effect observed for the scene-evoked response . 10 . 7554/eLife . 19113 . 005Figure 3 . Appearance-evoked response . ( A ) RMS time-course of the appearance-evoked response showing the main effect of scene temporal structure ( REG versus RAND ) . Thick horizontal green lines indicate time points for which there were significant differences in RMS between REG and RAND conditions ( p<0 . 05 FWE corrected at the cluster level; Thin light-green lines show uncorrected clusters ) . Purple lines indicate ( jackknife-estimated ) latencies of the onset of the REG versus RAND effect ( horizontal and vertical portions indicate mean and jackknife-corrected standard error , respectively ) . Also shown are topographical patterns at the time of the appearance-evoked M50 ( 72–112 ms ) , M100 ( 144–188 ms ) and M200 ( 232–360 ms ) components . ( B ) Mean RMS over the appearance-evoked M50 period ( 712–112 ms ) . Asterisk indicates the significant ( p<0 . 05 ) interaction ( [REG>RAND]>[Passive>Active] ) . Error bars represent within-subject standard error of the mean ( computed separately for Passive and Active groups . ( C ) Same as panel A but showing main effect of appearing source structure ( temporally regular versus random ) . See also Figure 3—figure supplement 1 for the MEG time-course averaged over selected sensors responsive to the appearance-evoked M50 component . DOI: http://dx . doi . org/10 . 7554/eLife . 19113 . 00510 . 7554/eLife . 19113 . 006Figure 3—figure supplement 1 . MEG time-course averaged over selected sensors responsive to the appearance-evoked M50 component . ( A ) MEG from sensors showing positive signal at the time of the appearance-evoked M50 component . Thick horizontal green lines indicate time points for which there were significant differences in MEG amplitude between REG and RAND conditions ( p<0 . 05 FWE corrected at the cluster level; Thin light-green lines show uncorrected clusters ) . ( B ) MEG from sensors showing negative signal at the time of the appearance-evoked M50 component . DOI: http://dx . doi . org/10 . 7554/eLife . 19113 . 006 In the Active group , the appearance-evoked response is characterized by a typical pattern of M50/M100/M200 deflections frequently observed at sound onset ( as seen above ) and following changes within an ongoing sound sequence ( Martin and Boothroyd , 2000; Gutschalk et al . , 2004; Chait et al . , 2008; Sohoglu and Chait , 2016 ) . Although the responses here are characterized by later latencies ( around 90 , 150 and 300 ms , respectively ) than those typically observed in other studies that report similar deflections . This may be due to the higher complexity of the present stimuli , which is known to lead to delayed responses ( see e . g . Chait et al . , 2008; Sohoglu and Chait , 2016 ) . M50 and M200 deflections are also observed in the Passive group but we note with interest the absence of a prominent M100 component , consistent with previous reports of this component being particularly sensitive to attention and/or task-related demands ( Ahveninen et al . , 2011; Ding and Simon , 2012; Königs and Gutschalk , 2012; Sohoglu and Chait , 2016 ) . As shown in Figure 3A , cluster-based statistics showed a significant effect of scene structure on the appearance-evoked response from 96 ms in the Passive group and from 260 ms in the Active group , both involving an increased neural response for REG versus RAND conditions . This effect was apparent in the Passive group already at the earliest M50 component while in the Active group , it was confined to later components of the evoked response ( M200 at corrected significance; M100 uncorrected ) . As shown in Figure 3A , the topographical patterns for REG and RAND conditions were qualitatively similar in both Passive and Active groups . As described previously , the appearance-evoked response was derived by baseline correcting relative to the 200 ms period prior to source appearance and therefore the measured effect of REG versus RAND is distinct to that observed for the scene-evoked response . To confirm this , we also analyzed matched trials in which there was no change ( no appearing source; shown as transparent traces in Figure 3A ) . For this analysis , no effect of REG versus RAND was observed , confirming that scene structure modulates the appearance-evoked response in addition to the scene-evoked response . The cluster-based permutation statistics above imply an interaction between scene structure and group involving an earlier effect of scene structure in Passive versus Active groups . To directly test this interaction , we estimated the onset latency of the scene structure effect using the jackknife procedure and assessed whether this latency differed significantly between groups . The scene structure effect was estimated to occur on average , 55 ms earlier in Passive versus Active groups ( mean onset latency = 87 ms for Passive; 142 ms for Active ) . This difference was confirmed significant using a jackknife adjusted independent samples test ( t ( 25 ) = −3 . 96 , p<0 . 001 ) . This analysis is consistent with the cluster-based permutation statistics above also suggesting a scene structure effect on the early M50 component only in the Passive group . To further characterize this scene structure by group interaction on the M50 peak , a post-hoc between-group t-test ( one-tailed ) was conducted on the difference in MEG response between REG and RAND conditions at the time of the appearance-evoked M50 ( 72–112 ms ) . As shown in Figure 3B , the difference in MEG response between REG and RAND conditions was significantly stronger in Passive versus Active groups ( t ( 25 ) = 1 . 78 , p<0 . 05 . ) . An alternative explanation of the interaction between group and scene structure is possible if the M50 and M100 peaks reflect independent but temporally and spatially overlapping components . By this account , the increased response for REG versus RAND scenes at the M50 does not differ between Passive and Active groups . Rather , REG scenes result in an increased M100 in Active listeners that causes a reduction in the M50 ( due to their opposite polarities; see topographic plots in Figure 3A ) . To explore this possibility , we selected the twenty most positive and twenty most negative channels at the time of the M50 deflection ( 72–112 ms; pooling over REG and RAND conditions ) . The MEG signal was then averaged across channels within these two ( positive and negative ) groupings and the resulting time-courses analyzed ( shown in Figure 3—figure supplement 1 ) . As this analysis is based on the mean ( rather than RMS ) neural response across channels , the polarity of the signal is preserved and thus potentially provides a more accurate representation of the underlying dynamics . Furthermore , the selection of channels based on the M50 deflection would be expected to attenuate interfering responses from the M100 component . As shown in Figure 3—figure supplement 1 , consistent with the earlier RMS analysis , the M50 showed a larger response for REG versus RAND scenes in Passive but not in Active listeners . This is despite the M100 showing no evidence of modulation by scene structure in Active listeners ( even at an uncorrected threshold of p<0 . 05 ) , making it unlikely the pattern of results reflect a suppression of the M50 by the M100 . The appearance-evoked response as a function of the temporal structure of the appearing source was also analyzed and is shown in Figure 3C . Despite listeners’ detection times being somewhat quicker when the appearing source was temporally regular versus random ( by ~27 ms on average across the group; shown earlier in Figure 1B ) , no significant differences in MEG response were observed in Passive or Active groups . Neither was there a significant interaction between scene and appearing source structure . However , we cannot rule out modulation of more temporally variable neural processes not captured by the evoked analysis employed here . Since the behavioral effects were only observed in detection times , it is also possible that the relevant brain activity is masked by motor response-related processes . Finally , we localized the neural generators of the scene structure effect . As shown in Figure 4A , the scene-evoked response ( averaged from 500 to 800 ms ) showed greater source power for REG versus RAND scenes in both hemispheres of the superior temporal lobe , including primary auditory cortex , planum temporale and the superior temporal gyrus ( peak voxel locations are reported in Table 1 ) . An additional distinct cluster of activation is observed in left post central gyrus of the superior parietal lobe . 10 . 7554/eLife . 19113 . 007Figure 4 . Source reconstruction . ( A ) Main effect of scene temporal structure at the time of the sustained portion of the scene-evoked response ( 500–800 ms post scene onset ) . Statistical map is overlaid onto an MNI space template brain , viewed over the left and right hemispheres . Color-bar indicates statistical threshold . ( B ) [REG>RAND]>[Passive>Active] interaction at the time of the appearance-evoked M50 component ( 72–112 ms post appearance ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19113 . 00710 . 7554/eLife . 19113 . 008Table 1 . Peak voxel locations ( in MNI space ) and summary statistics from source reconstruction . Activations for the scene-evoked analysis are for the REG>RAND contrast ( 500–800 ms post scene onset ) while those for the appearance-evoked analysis are for the [REG>RAND]>[Passive>Active] interaction contrast ( 72–112 ms post appearance ) . Activations have been thresholded using the same parameters as for Figure 4 ( p<0 . 001 for scene-evoked; p<0 . 01 for appearance-evoked ) but with an additional cluster extent threshold of n > 15 voxels ( for display purposes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19113 . 008MNI CoordinatesAnalysisRegionSideExtentt-valuexyzScene-evokedPlanum Temporale/Parietal OperculumLeft14185 . 2779−48−2816 ( 500-800 ms post scene onset ) 4 . 364−62−50143 . 9777−52−30-4Postcentral GyrusLeft2044 . 8082−32−3664Supramarginal GyrusRight7044 . 246964−24243 . 717644−616Planum TemporaleRight5823 . 945964−1663 . 925246−266Precentral GyrusRight193 . 605160618Appearance-evokedPrecentral GyrusLeft1903 . 2219−50−644 ( 72-112 ms post appearance ) 2 . 9902−34638Precentral Gyrus/Central OperculumRight7113 . 1966560103 . 0153564−102 . 910136−816Middle Temporal GyrusRight1572 . 985958−2−24Middle Temporal GyrusRight552 . 695152−548Precentral GyrusRight212 . 644454−440Postcentral GyrusLeft162 . 5982-30−3468 For the appearance-evoked response , we focused on the scene structure by group interaction ( [REG>RAND] > [Passive>Active] ) that emerged during the early M50 component . As shown in Figure 4B , this effect localized to similar regions as for the scene-evoked response: superior/middle temporal lobe ( albeit in the right hemisphere only ) and post central gyrus . Additional activation is observed more anteriorly in the pre central gyrus , extending into the middle frontal gyrus .
Around 400 ms following scene onset , we observed an increase in the sustained MEG response for scenes consisting of regularly structured , relative to randomly fluctuating sources . This finding is opposite to what would be expected based on adaptation i . e . decreased neural responses for temporally regular events , which has previously been observed for isolated tone sequences ( Costa-Faidella et al . , 2011; Schwartze et al . , 2013; Tavano et al . , 2014 ) . It is however consistent with a mechanism that infers the precision ( predictability ) of sensory input and uses this information to up-regulate neural processing towards more reliable sensory signals ( Feldman and Friston , 2010; Zhao et al . , 2013; Auksztulewicz and Friston , 2015; Barascud et al . , 2016 ) . Indeed , it has recently been demonstrated that the magnitude of sustained MEG activity ( from naïve distracted listeners ) tracks the predictability of rapid tone sequences ( Barascud et al . , 2016 ) . In that study , regularity was characterized by a spectral pattern repeating over time within a single ongoing tone sequence . Although distinct to the temporal regularity studied here , the ensuing effect on MEG response is strikingly similar to the sustained effect we observe . Importantly , the current findings demonstrate mechanisms that automatically ( irrespective of directed attention ) and rapidly ( within 400 ms of scene onset ) encode regularities distributed over many concurrent sources , typical of natural listening environments . If the auditory system can form precise models about the content of ongoing scenes , novel events that violate those models would evoke greater neural responses and be perceived as more salient . Indeed , listeners were better and faster at detecting an appearing source in regular versus random scenes . The MEG response in naïve , passively listening subjects revealed a large ( 22% ) increase in the evoked response starting from the very first response deflection ( M50 component ) following source appearance . Importantly , this effect occurred over and above that observed prior to the appearance event , demonstrating bottom-up driven ‘surprise’ responses tightly linked to the predictability of the ongoing scene context . Interestingly , the REG>RAND effect emerged substantially later when participants were actively attending to the appearance events . More discussion of that is below . Overall , the results demonstrate that the enhanced detection performance observed in behavior is not solely the result of changes in neural responses occurring prior to source appearance ( cf . adaptation accounts; May et al . , 1999; Jääskeläinen et al . , 2004 ) but also due to enhanced neural responses to novel events themselves . This is again what would be expected based on precision accounts and is also consistent with animal physiology work showing that the magnitude of responses in single neurons of auditory cortex to new ( ‘deviant’ ) tones is larger than expected based on simple adaptation to previously repeated ( ‘standard’ ) tones alone ( Khouri and Nelken , 2015 ) . Source reconstruction suggests that neural responses in a network of brain regions are modulated by scene temporal structure , including early auditory regions in the superior temporal lobe but also left parietal cortex ( post central gyrus ) . This is consistent with evidence from neuroimaging ( Rao et al . , 2001; Coull and Nobre , 2008; Andreou et al . , 2015 ) , electrophysiology ( Leon et al . , 2003; Janssen and Shadlen , 2005 ) and lesion studies ( Harrington et al . , 1998; Battelli et al . , 2008 ) implicating a specific role for left parietal cortex in temporal processing . Parietal cortex has also been associated with figure-ground processing when the figure is defined by temporally repeatable spectral components in an otherwise randomly structured background ( Teki et al . , 2011; Teki et al . , 2016 ) . Thus , together the current study and previous findings suggest parietal cortex may be part of a wider network ( along with auditory cortical regions ) that codes the temporal structure of acoustic scenes . Alternatively , the parietal activity changes we observe may reflect a more domain-general increase in bottom-up saliency attributable to regularity ( Corbetta and Shulman , 2002; Zhao et al . , 2013 ) . We note however that although Barascud et al . ( 2016 ) report effects of statistical structure in early auditory regions ( like the current findings ) , they did not observe changes in MEG and fMRI responses in parietal cortex . Instead , spectral regularity modulated activity in the inferior frontal gyrus . This is may suggest a degree of neural specialization for the particular type of regularity encoded e . g . temporal-based involving parietal cortex versus spectral-based involving inferior frontal regions . Future work is required however to determine whether temporal and spectral regularities are encoded by distinct neural substrates ( e . g . by contrasting neural effects of temporal and spectral regularities in the same experiment ) . Following scene onset ( prior to new source appearance ) , the neural influence of regularity showed no evidence of attentional modulation ( the strength of the scene-evoked response to regularly and randomly structured scenes was statistically indistinguishable in passive compared with active listening subjects ) . This suggests that the brain automatically encodes scene regularities , irrespective of directed attention . After the appearance of a new source , however , regularity and attention had an interactive influence on the evoked response; whereas the first cortical deflection of the appearance-evoked response ( M50 ) increased in regular scenes during passive listening , this effect was confined to later deflections ( M100 and M200 ) when listeners actively detected source appearance . How the neural influence of regularity might depend on attention is the subject of ongoing debate ( Jones and Boltz , 1989; Näätänen et al . , 2001; Summerfield and Egner , 2009 , 2016; Winkler et al . , 2009; Feldman and Friston , 2010; Kok et al . , 2012; Bendixen , 2014; Summerfield and de Lange , 2014; Schröger et al . , 2015 ) . One proposal is that attention ( like regularity ) acts to determine the inferred precision of sensory input ( Friston , 2009; Barascud et al . , 2016 ) . In this view , attention increases precision ( and neural responses ) when sensory signals are task-relevant . In this case inferred precision is changed not by the intrinsic structure of the stimulus ( e . g . whether temporally regular or random ) but by the behavioral goals of the listener . As precision is hypothesized to have a multiplicative ( gain ) influence on neural activity , this account would have predicted attentional enhancement of the regularity effect . Indeed , Hsu et al . ( 2014 ) demonstrated greater EEG responses for predictable ( ascending ) pitch patterns , which was most apparent when those patterns were embedded in an attended stream . In contrast to this pattern , attention in our study delayed the influence of regularity on appearance-related responses . How then might the current attentional effect be explained ? We suggest that attention in our study acted as a form of expectation . That is , when listeners actively detected source appearance , scene changes were relatively more expected . If change-related responses reflect the amount of ‘surprise’ given the preceding stimulus context , then they should diminish when change is expected and counteract the precision-mediated increase from regularity . Thus , although counterintuitive , the later benefit from regularity when listeners are actively seeking source appearance is consistent with attention acting to reduce surprise . In the auditory modality , the mismatch negativity ( MMN ) response is often interpreted as reflecting ‘surprise’ ( Näätänen et al . , 2007; Garrido et al . , 2009 ) . The M50 effect we observe occurs earlier and with a distinct topography to the MMN , but may relate to novelty effects on the so-called ‘middle-latency’ responses ( ~40 ms ) revealed in other work ( Chait et al . , 2007 , 2008; Grimm et al . , 2011; Recasens et al . , 2014 ) . While the proposal that attention acts to reduce surprise may appear at odds with the widespread view of attention playing a distinct functional role to expectation ( Summerfield and Egner , 2009 , 2016 ) , one associated with enhanced neural processing of attended signals ( Desimone and Duncan , 1995; Fritz et al . , 2003 ) , it is consistent with previous observations . In Chennu et al . ( 2013 ) , listeners were presented with tone sequences containing regularities unfolding over multiple ( local and global ) timescales . When listeners were instructed to detect deviant tones on a local timescale , the mismatch negativity component indexing that local regularity was attenuated compared with when listeners detected global deviants . The authors interpreted this suppression effect as reflecting reduced surprise from 'top-down expectation ( or bias ) and consequent attentional focus' . Similarly , Spratling ( 2008 ) argues that attention and expectation are part of the same general class of top-down signal that act in a similar fashion to modulate perceptual processing . Thus , in this view , the distinction between attention and expectation is blurred and whether these phenomena result in increased or decreased neural processes will depend on the precise details of the stimuli and behavioral demands ( Schröger et al . , 2015; Henson , 2016 ) . In this respect , we note that previous investigations of regularity and attention employed static or relatively slow-evolving stimuli ( 1–5 Hz ) and often containing a single perceptual object ( image of a face or tone sequence ) . This may have enabled conscious awareness of stimulus content , involving distinct processes to those relevant to the rapidly evolving and complex scenes employed here and , arguably , to the perceptual challenges faced in natural environments .
Two groups of participants were tested after being informed of the study’s procedure , which was approved by the research ethics committee of University College London . The two groups differed in whether participants’ attention was directed away ( ‘Passive’ group ) or towards ( ‘Active’ group ) auditory stimulation ( see Procedure section below ) . The Passive group comprised 14 ( 6 female ) participants aged between 19 and 34 years ( mean = 23 . 6 , SD = 4 . 68 ) . All but one of these participants was right-handed . The Active group comprised 13 ( 7 female ) , different , right-handed participants aged between 18 and 33 years ( mean = 24 . 3 , SD = 4 . 91 ) . All reported normal hearing , normal or corrected-to-normal vision , and had no history of neurological disorders . There were no significant differences between groups in terms of gender ( two-tailed χ ( 1 ) = . 326 , p=0 . 568 ) or age ( two-tailed t ( 25 ) = 0 . 35 , p= 0 . 73 ) . d’ scores were obtained for the Active group by first computing for each subject and condition , the hit rate ( proportion of source appearances correctly detected ) and false alarm rate ( proportion of ‘No Change’ trials for which responses were made ) . Following this , each d’ score was computed as the difference in the z-transformed hit rate and false alarm rate . Detection time was measured between the time of new source appearance and the subject’s key press . Magnetic fields were recorded with a CTF-275 MEG system , with 274 functioning axial gradiometers arranged in a helmet shaped array . Electrical coils were attached to three anatomical fiducial points ( nasion and left and right pre-auricular ) , in order to continuously monitor the position of each participant’s head with respect to the MEG sensors . The MEG data were analyzed in SPM12 ( Wellcome Trust Centre for Neuroimaging , London , UK ) and FieldTrip ( Donders Institute for Brain , Cognition and Behaviour , Radboud University Nijmegen , the Netherlands ) software implemented in Matlab . The data were downsampled to 250 Hz , low-pass filtered at 30 Hz and epoched −200 to 800 ms relative to scene onset ( to obtain the scene-evoked response ) or −200 to 400 ms relative to the time of the appearance event ( to obtain the appearance-evoked response ) . This epoch encompassed detection-related brain processes leading up to the initiation of the behavioral response in the Active group , which ranged from 465 to 911 ms across participants and conditions . After epoching , the data were baseline-corrected relative to the 200 ms period prior to scene onset ( for the scene-evoked data ) or prior to the time of source appearance ( for the appearance-evoked data ) . Subsequent preprocessing differed depending on whether the analysis was conducted in sensor- or source-space . For sensor-space analysis , any trials in which the data deviated by more than three standard deviations from the mean were discarded . Following outlier removal , Denoising Source Separation ( DSS ) was applied to maximize reproducibility of the evoked response across trials ( de Cheveigné and Simon , 2008; de Cheveigné and Parra , 2014 ) . For each subject , the first two DSS components ( i . e . , the two ‘most reproducible’ components; determined −200 to 800 ms relative to scene onset ) were retained and used to project both the scene-evoked and appearance-evoked data back into sensor-space , which were then averaged across trials . For source-space analysis , DSS was not performed . Instead , the data were robust averaged across trials to downweight outlying samples ( Wager et al . , 2005; Litvak et al . , 2011 ) . To remove any high-frequency components that were introduced to the data by the robust averaging procedure , low-pass filtering was repeated after averaging . Note that although images were presented only in the Passive group , auditory and visual events were temporally uncorrelated . Thus , in both Passive and Active groups , our MEG measures are expected to reflect primarily auditory ( and not visual ) evoked activity . MEG data across the sensor array were summarized as the root mean square ( RMS ) across sensors for each time sample within the epoch period , reflecting the instantaneous magnitude of neuronal responses . Group-level paired t-tests were performed for each time sample while controlling the family-wise error ( FWE ) rate using a non-parametric ( cluster-based ) permutation procedure based on 5000 iterations ( Maris and Oostenveld , 2007 ) . Reported effects were obtained by using a cluster defining height threshold of p<0 . 05 with a cluster size threshold of p<0 . 05 ( FWE corrected ) , unless otherwise stated . Statistical tests of evoked response latency differences were conducted on subsamples of the grand averaged RMS time-course using the jackknife procedure ( Efron , 1981 ) . In the jackknife procedure , the grand averaged data are resampled n times ( with n being the number of participants ) while omitting one participant from each subsample . Statistical reliability of an effect can then be assessed using standard tests ( e . g . t-test ) , not across individual participants , but across subsamples of the grand average . This technique has been shown to be superior to computing latency differences from individual participant data because of the higher signal-to-noise ratio associated with grand averages ( Miller et al . , 1998; Ulrich and Miller , 2001 ) . Jackknife-estimated latencies of the scene structure effect were determined by first computing the difference waveform between REG and RAND scenes and then for each jackknife subsample , computing the first latency at which the magnitude of the difference waveform deviated by more than three standard deviations from the mean RMS across time in the baseline period ( −200 to 0 ms ) . When using the jackknife procedure , t-statistics were corrected following the procedure in Miller et al . ( 1998 ) ( multiplication of the subsample standard error by a factor of n-1 ) . To determine the underlying brain sources of the sensor-space effects , we used a distributed method of source reconstruction , implemented within the parametric empirical Bayes framework of SPM12 ( Phillips et al . , 2005; Litvak and Friston , 2008; Henson et al . , 2011 ) . Participant-specific forward models were computed using a Single Shell model and sensor positions projected onto an MNI space template brain by minimizing the sum of squared differences between the digitized fiducials and the MNI template . For inversion of the forward model , we used the ‘LOR’ routine in SPM12 , which assumes that all sources are activated with equal apriori probability and with weak correlation to neighboring sources . This was applied to the entire epoch ( −200 to 800 ms for scene-evoked data; −200 to 400 ms for appearance-evoked data ) . Source solutions were constrained to be consistent across subjects ( pooled over Passive and Active groups ) , which has been shown to improve group-level statistical power ( Litvak and Friston , 2008; Henson et al . , 2011 ) . In brief , this procedure involves 1 ) realigning and concatenating sensor-level data across subjects 2 ) estimating a single source solution for all subjects 3 ) using the resulting group solution as a Bayesian prior on individual subject inversions . Thus , this method exploits the availability of repeated measurements ( from different subjects ) to constrain source reconstruction . Importantly , however , this procedure does not bias activation differences between conditions in a given source . Significant effects from sensor-space were localized within the brain ( in MNI space , constrained to gray matter ) after summarizing source power in the 0–30 Hz range for each participant and time-window of interest using a Morlet wavelet projector ( Friston et al . , 2006 ) . Given that the goal of source reconstruction was to localize the neural generators of sensor-space effects previously identified as significant , statistical maps of source activity are displayed with uncorrected voxelwise thresholds ( Gross et al . , 2012 ) . | Everyday environments like a busy street bombard our ears with information . Yet most of the time , the human brain quickly and effortlessly makes sense of this information in a process known as auditory scene analysis . According to one popular theory , the brain is particularly sensitive to regularly repeating features in sensory signals , and uses those regularities to guide scene analysis . Indeed , many biological sounds contain such regularities , like the pitter-patter of footsteps or the fluttering of bird wings . In most previous studies that investigated whether regularity guides auditory scene analysis in humans , listeners attended to one sound stream that repeated slowly . Thus , it was unclear how regularity might benefit scene analysis in more realistic settings that feature many sounds that quickly change over time . Sohoglu and Chait presented listeners with cluttered , artificial auditory scenes comprised of several sources of sound . If the scenes contained regularly repeating sound sources , the listeners were better able to detect new sounds that appeared partway through the scenes . This shows that auditory scene analysis benefits from sound regularity . To understand the neurobiological basis of this effect , Sohoglu and Chait also recorded the brain activity of the listeners using a non-invasive technique called magnetoencephalography . This activity increased when the sound scenes featured regularly repeating sounds . It therefore appears that the brain prioritized the repeating sounds , and this improved the ability of the listeners to detect new sound sources . When the listeners actively focused on listening to the regular sounds , their brain response to new sounds occurred later than seen in volunteers who were not actively listening to the scene . This was unexpected as delayed brain responses are not usually associated with active focusing . However , this effect can be explained if active focusing increases the expectation of new sounds appearing , because previous research has shown that expectation reduces brain responses . The experiments performed by Sohoglu and Chait used a relatively simple form of sound regularity ( tone pips repeating at equal time intervals ) . Future work will investigate more complex forms of regularity to understand the kinds of sensory patterns to which the brain is sensitive . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Detecting and representing predictable structure during auditory scene analysis |
Autotransporters are a superfamily of bacterial virulence factors consisting of an N-terminal extracellular ( ‘passenger’ ) domain and a C-terminal β barrel ( ‘β’ ) domain that resides in the outer membrane ( OM ) . The mechanism by which the passenger domain is secreted is poorly understood . Here we show that a conserved OM protein insertase ( the Bam complex ) and a molecular chaperone ( SurA ) are both necessary and sufficient to promote the complete assembly of the Escherichia coli O157:H7 autotransporter EspP in vitro . Our results indicate that the membrane integration of the β domain is the rate-limiting step in autotransporter assembly and that passenger domain translocation does not require the input of external energy . Furthermore , experiments using nanodiscs strongly suggest that autotransporter assembly is catalyzed by a single copy of the Bam complex . Finally , we describe a method to purify a highly active form of the Bam complex that should facilitate the elucidation of its function .
The autotransporter ( type Va ) pathway is the most widespread virulence factor secretion pathway in Gram-negative bacteria ( Leyton et al . , 2012 ) . Autotransporters are single polypeptides that consist of two domains , an N-terminal domain ( ‘passenger domain’ ) that is exposed on the cell surface and that often mediates a virulence function , and a C-terminal domain ( ‘β domain’ ) that resides in the outer membrane ( OM ) . Following their translocation across the OM , many passenger domains are released from the cell surface by a proteolytic cleavage . Passenger domains vary widely in sequence and size ( ∼20–400 kD ) , but almost always form an unusual repetitive structure known as a β helix ( Junker et al . , 2006 ) . β domains are typically ∼30 kD in size and , like the vast majority of bacterial integral OM proteins , fold into a β barrel structure . While the sequences of β domains are also very heterogeneous , the structures of all of the β domains that have been solved to date are nearly superimposable ( Oomen et al . , 2004; Barnard et al . , 2007; van den Berg , 2010; Zhai et al . , 2011 ) . After autotransporters are translocated across the inner membrane ( IM ) through the Sec complex they interact with molecular chaperones that presumably maintain them in an assembly-competent state ( Ieva and Bernstein , 2009; Ruiz-Perez et al . , 2009; Ieva et al . , 2011 ) . Subsequently the β domain is targeted to the Bam complex , an essential heterooligomer consisting of an integral OM protein ( BamA ) and four lipoproteins ( BamBCDE ) that catalyzes the membrane insertion of β barrel proteins ( Voulhoux et al . , 2003; Wu et al . , 2005; Sauri et al . , 2009; Hagan et al . , 2010; Ieva et al . , 2011 ) . X-ray crystallographic analysis suggests that the BamA β barrel domain is unstable and may open laterally to allow the escape of client proteins into the lipid bilayer ( Noinaj et al . , 2013 , 2014 ) . Although it has been shown that the passenger domain is translocated across the OM in a C-to-N-terminal fashion ( Ieva and Bernstein , 2009; Junker et al . , 2009 ) , the mechanism of translocation is poorly understood . Based on the observation that deletion of the β domain abolishes passenger domain translocation , it was originally proposed that the β domain serves as the transport channel for the passenger domain ( whence the name ‘autotransporter’ ) ( Pohlner et al . , 1987 ) . At first glance , this hypothesis seems to be supported by crystallographic studies showing that upon completion of translocation the two domains are connected by an α-helical linker that is embedded inside the β domain pore ( Oomen et al . , 2004; Barnard et al . , 2007; van den Berg , 2010 ) . The same studies , however , revealed that the β domain pore is only ∼10 Å in diameter . Because the directionality of translocation presumably requires the formation of a C-terminal hairpin followed by the sliding of medial and N-terminal segments past a static strand , both strands of the hairpin would have to be in a fully extended conformation to fit inside the pore . On the contrary , available evidence indicates that the polypeptide that resides inside the β domain forms an α helix prior to the completion of translocation ( Ieva et al . , 2008; Peterson et al . , 2010 ) . Furthermore , small folded polypeptides have been shown to be secreted efficiently by the autotransporter pathway and a subset of naturally occurring passenger domains undergo disulfide bonding in the periplasm ( Skillman et al . , 2005; Jong et al . , 2007 ) . Finally , BamA has been shown to interact with the passenger domain during the translocation reaction ( Ieva and Bernstein , 2009 ) . While the β domain does appear to play a role in translocation ( Saurí et al . , 2011; Pavlova et al . , 2013 ) , these and other results strongly suggest that autotransporter assembly is more complex than originally envisioned . In one scenario , the Bam complex might facilitate the membrane insertion of the β domain and the translocation of the passenger domain in a concerted reaction , but it is also possible that other factors are also required to transport the passenger domain across the OM . The energy source for passenger domain translocation has also remained poorly understood . Although most protein translocation reactions require external energy in the form of ATP or a membrane potential , the periplasm is devoid of ATP and there is no membrane potential across the bacterial OM . It is conceivable that an IM protein that derives energy from the hydrolysis of cytoplasmic ATP or from the electrochemical gradient across the IM interacts with the passenger domain and drives translocation , but such a protein has never been identified . To explain the energetics of secretion , it has been proposed that the sequential folding of segments of the β helix in the extracellular space drives translocation and prevents retrograde movement into the periplasm ( Klauser et al . , 1992; Junker et al . , 2006 ) . Consistent with this vectorial folding model , mutations that impair the folding of C-terminal segments of two different passenger domains significantly reduce the efficiency of secretion ( Peterson et al . , 2010; Renn et al . , 2012 ) . Mutations that perturb the folding of medial and N-terminal segments of the passenger domain , however , produce only a modest translocation defect ( Kang’ethe and Bernstein , 2013b ) . Furthermore , the intrinsically disordered receptor domain ( RD ) of the Bordetella pertussis RTX toxin has been shown to be secreted efficiently by the autotransporter pathway ( Kang’ethe and Bernstein , 2013a ) . Interestingly , the presence of a large number of acidic residues is critical for the secretion of the RD . This observation , together with the finding that naturally occurring passenger domains are predominantly acidic , suggests that charge interactions may play a significant role in driving the translocation reaction . To gain further insight into both the mechanism and energetics of passenger domain translocation , we sought to reconstitute autotransporter assembly in vitro using purified components . In our experiments , we used the Escherichia coli O157:H7 autotransporter EspP , as a model protein . Following the completion of translocation , the passenger domain of EspP and other members of the SPATE ( serine protease autotransporters of Enterobacteriaceae ) family is released in an unusual intrabarrel cleavage reaction that requires accurate folding of the β domain and precise positioning of the passenger domain-β domain junction ( Dautin et al . , 2007; Barnard et al . , 2012 ) . Although it was previously shown that the insertion of small E . coli β barrel proteins such as OmpT and OmpA into proteoliposomes can be catalyzed by the purified Bam complex and the periplasmic chaperone SurA ( Hagan et al . , 2010; Hagan and Kahne , 2011; Hagan et al . , 2013 ) , we did not observe assembly of EspP using the same methodology . Using an alternative expression and purification strategy , however , we obtained an apparently more active form of the Bam complex that , together with SurA , was both necessary and sufficient to promote the membrane integration of the EspP β domain and the translocation and cleavage of the EspP passenger domain . Remarkably , passenger domain translocation did not require the input of any additional energy . In addition to defining the minimal set of factors required for autotransporter assembly , our work provides a valuable resource for future studies on the function of the Bam complex and its role in the biogenesis of the broader class of OM proteins .
It has long been known that large N-terminal segments of autotransporter passenger domains can be deleted without affecting the integration of the β domain into the OM or the secretion of the remaining passenger domain fragment ( Dautin and Bernstein , 2007 ) . The native EspP passenger domain is 968 residues in length ( Figure 1A ) , but a derivative that contains only 26 residues of the ∼28 residue C-terminal segment that reside inside the β domain in an α-helical conformation ( EspPΔ5 ) is assembled as efficiently as the wild-type protein in vivo ( Pavlova et al . , 2013 ) . The modification or deletion of residues in the α-helical segment , however , can profoundly perturb the folding and integration of the β domain into the OM and/or the proteolytic release of the passenger domain , which requires precise alignment of the cleavage junction with key catalytic residues ( Dautin et al . , 2007; Ieva et al . , 2008; Barnard et al . , 2012 ) . Because EspPΔ5 and other derivatives that expose no more than a short segment on the cell surface are structurally similar to generic β barrel proteins such as OmpT , it might be expected that they would have similar assembly requirements . Indeed any autotransporter-specific assembly factors might only be required once the passenger domain reaches a threshold size . Based on this reasoning , we first analyzed the assembly of an EspP derivative designated EspP ( 46+β ) that consists of EspPΔ5 plus 20 N-terminal residues derived from the cloning vector ( Figure 1B ) . We expected that the accurate assembly of EspP ( 46+β ) would result in the proteolytic release of the 46 residue passenger domain and the accumulation of a folded ∼30 kD β domain that , as previously shown , is resistant to SDS denaturation unless heated ( Skillman et al . , 2005 ) . 10 . 7554/eLife . 04234 . 003Figure 1 . Domain structure of EspP and model for passenger domain translocation . ( A ) EspP consists of a signal peptide ( SP; residues 1–55 ) , an extracellular ( ‘passenger’ ) domain ( residues 56–1023 ) and a β barrel ( ‘β’ ) domain ( residues 1024–1300 ) ( Brunder et al . , 1997 ) . While most of the passenger domain forms a long β helix , the N-terminus ( residues 56–313 ) forms a discrete globular domain ( Khan et al . , 2011 ) . ( B ) Illustration of EspP ( 46+β ) . Prior to the release of the passenger domain in an intrabarrel cleavage reaction , ∼28 residues of the 46 residue passenger domain are embedded inside the β domain pore . ( C ) Available evidence indicates that the EspP passenger domain is secreted in a hairpin conformation while distinct regions of the β domain interact with BamA , BamB and BamD ( Ieva et al . , 2011 ) . Components of the transport channel likely include the open β domain and/or the BamA β barrel , which has been proposed to open laterally ( Noinaj et al . , 2013 , 2014 ) . In any case , the finding that proteolytic maturation and the release of the β domain from the Bam complex both require the completion of translocation ( Peterson et al . , 2010; Ieva et al . , 2011; Pavlova et al . , 2013 ) strongly suggests that the active site cannot form during the passage of the passenger domain across the OM . BamC and BamE hve been omitted from the model for the sake of clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 003 Initially we tested whether EspP ( 46+β ) would assemble in vitro using a previously described OM protein assembly assay ( Hagan et al . , 2010 ) . In this assay , E . coli BamAB and BamCDE are first expressed and isolated independently . The two subcomplexes are then mixed together to form a holocomplex and reconstituted into proteoliposomes . In the presence of the proteoliposomes and a molar excess of SurA , denatured OmpT folds into a stable structure that is enzymatically active and resistant to SDS denaturation . We attempted to express and purify the Bam complex exactly as described , but for reasons that are unclear we did not observe efficient formation of the Bam holocomplex unless we purified BamAB through an additional step . Nevertheless , we ultimately obtained a single peak on a gel filtration column that was highly enriched in the reconstructed complex , which we designated Bam ( AB ) ( CDE ) ( Figure 2—figure supplement 1A , B ) . After the peak fractions were pooled , most of the Bam holocomplex could be reconstituted into proteoliposomes ( Figure 2—figure supplement 1C ) . Subsequently we purified SurA to homogeneity as described ( Hagan et al . , 2010 ) ( Figure 2—figure supplement 2 ) . To assess the folding of OmpT , we used an activity assay in which the cleavage of a fluorogenic peptide leads to an increase in fluorescence intensity over time . Consistent with previous results , we observed a fluorescent signal when OmpT was incubated in the presence of Bam ( AB ) ( CDE ) and SurA ( Figure 2E , purple curve ) . EspP ( 46+β ) purified from inclusion bodies ( Figure 3—figure supplement 1 ) was mixed with the same components , and the proteolytic maturation of the protein was assessed by Western blotting using an antiserum against an EspP C-terminal peptide . No free β domain was detected , however , even after a prolonged incubation ( Figure 2—figure supplement 1D ) . 10 . 7554/eLife . 04234 . 004Figure 2 . Purification and functional test of BamABCDE . ( A ) Chromatogram of BamABCDE on S-200 gel filtration column . ( B ) SDS-PAGE analysis of the peak fractions in ( A ) . Proteins were visualized by Coomassie Blue staining . ( C ) The peak fractions in ( B ) were pooled and analyzed by Blue Native PAGE . ( D ) Purified BamABCDE was centrifuged in the absence of liposomes or after reconstitution into liposomes . The pellet ( P ) and supernatant ( S ) fractions were analyzed by SDS-PAGE . ( E ) Urea denatured OmpT was diluted and incubated with SurA and proteoliposomes containing either BamABCDE ( green ) or Bam ( AB ) ( CDE ) ( purple ) , proteoliposomes containing BamABCDE alone ( red ) or SurA alone ( blue ) . OmpT activity was assessed by measuring the fluorescent signal generated by the cleavage of a fluorogenic peptide . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 00410 . 7554/eLife . 04234 . 005Figure 2—figure supplement 1 . Test of purified Bam ( AB ) ( CDE ) in EspP assembly assay . ( A ) Chromatogram of Bam ( AB ) ( CDE ) on S-200 gel filtration column . ( B ) The peak fractions in part A were pooled and analyzed by SDS-PAGE . Proteins were visualized by Coomassie Blue staining . ( C ) Purified Bam ( AB ) ( CDE ) was centrifuged in the absence of liposomes or after reconstitution into liposomes . The pellet ( P ) and supernatant ( S ) fractions were analyzed by SDS-PAGE . ( D ) Urea-denatured EspP ( 46+β ) was incubated with SurA and proteoliposomes containing Bam ( AB ) ( CDE ) . Aliquots were placed on ice at various time points , heated to 95°C or maintained at room temperature after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 00510 . 7554/eLife . 04234 . 006Figure 2—figure supplement 2 . SDS-PAGE analysis of purified SurA . His-tagged SurA was purified by affinity chromatography and analyzed by SDS-PAGE . The protein was visualized by Coomassie Blue staining . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 006 Partly because we had difficulty reconstructing the Bam complex from BamAB and BamCDE subcomplexes , we next cloned the genes encoding all five subunits into a single expression plasmid . An octahistidine tag was introduced at the C-terminus of BamE to facilitate Bam complex purification . All of the proteins co-eluted on gel filtration columns ( Figure 2A , B ) . This observation strongly suggests that Bam holocomplexes were formed efficiently in vivo . When the peak fractions were pooled and analyzed by Blue Native PAGE , a single band that migrated slightly slower than the 242 kDa marker was seen ( Figure 2C ) . The Bam complex is only ∼200 kDa , but it has previously been shown to migrate slower than its actual molecular weight on Blue Native gels ( Hagan et al . , 2013 ) . In the presence of E . coli phospolipids , almost all of the purified protein , which we designated BamABCDE , could be reconstituted into proteoliposomes ( Figure 2D ) . Remarkably , the intensity of the fluorescent signal in the fluorogenic peptide cleavage assay indicated that OmpT folded much more efficiently in the presence of BamABCDE than Bam ( AB ) ( CDE ) ( Figure 2E , compare green and purple curves ) . These results suggested that the activity of the Bam complex is optimized when all of the subunits are expressed together . Interestingly , we found that BamABCDE also stimulated the assembly of EspP ( 46+β ) . The urea-denatured EspP derivative was incubated at 30°C with proteoliposomes containing the Bam complex and SurA , and samples were removed at various time points . Folding of the protein and subsequent proteolytic maturation were monitored by Western blot using the anti-EspP C-terminal antiserum and a fluorescently-labeled secondary antibody . A rapidly migrating ( ∼27 kD ) band that corresponds to the folded form of the free EspP β domain appeared within 5 min and was more prominent at later time points when the samples were not heated ( Figure 3A ) . When the samples were heated , a much more intense ∼30 kD band that corresponds to the unfolded β domain was observed . This marked difference in intensity has been observed previously ( Barnard et al . , 2007; Pavlova et al . , 2013 ) and presumably results from the occlusion of the C-terminal epitope in the folded form of the β domain . No proteolytic processing was observed if either SurA or BamABCDE was omitted from the reaction or if BamAB or BamCDE subomplexes were used in place of the holocomplex ( Figure 3B ) . Single plasmids were also used to produce Bam complexes that lack either BamB or BamC , but curiously only ∼50–70% of the purified partial complexes could be reconstituted into proteoliposomes ( Figure 3—figure supplement 2A ) . BamABDE and BamACDE both promoted the assembly EspP ( 46+β ) , but less effectively than BamABCDE ( Figure 3—figure supplement 2B ) . Finally , the addition of purified Skp to reactions containing BamABCDE and SurA inhibited EspP ( 46+β ) assembly ( Figure 3—figure supplement 3 ) and ATP ( 0 . 1 mM ) blocked assembly altogether ( data not shown ) . 10 . 7554/eLife . 04234 . 007Figure 3 . BamABCDE and SurA catalyze the assembly of EspP ( 46+β ) . ( A ) Urea-denatured EspP ( 46+β ) was incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C or maintained at room temperature after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . ( B ) Urea-denatured EspP ( 46+β ) was incubated with the indicated factors . Aliquots were removed after 0 and 30 min and analyzed as in ( A ) . ( C ) Urea-denatured EspP* ( 46+β ) was diluted and incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were removed at various time points and analyzed as in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 00710 . 7554/eLife . 04234 . 008Figure 3—figure supplement 1 . SDS-PAGE analysis of purified EspP derivatives . The indicated EspP derivatives were purified from inclusion bodies and analyzed by SDS-PAGE . The proteins were visualized by Coomassie Blue staining . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 00810 . 7554/eLife . 04234 . 009Figure 3—figure supplement 2 . Bam complexes lacking BamB or BamC catalyze the assembly of EspP ( 46+β ) less effectively than the holocomplex . ( A ) Purified BamABDE and BamACDE were centrifuged after reconstitution into liposomes . The pellet ( P ) and supernatant ( S ) fractions were analyzed by SDS-PAGE . ( B ) Urea-denatured EspP ( 46+β ) was incubated with SurA and proteoliposomes containing BamABCDE , BamABDE or BamACDE . Aliquots were placed on ice at various time points , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . A quantitation of the data is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 00910 . 7554/eLife . 04234 . 010Figure 3—figure supplement 3 . Skp inhibits the assembly of EspP ( 46+β ) . Urea-denatured EspP ( 46+β ) was incubated with SurA , proteoliposomes containing BamABCDE , and the indicated concentration of Skp . Aliquots were placed on ice after 0 and 30 min , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 01010 . 7554/eLife . 04234 . 011Figure 3—figure supplement 4 . The efficiency of EspP ( 46+β ) assembly is limited by the ability of the protein to remain folding-competent . ( A ) Urea-denatured EspP ( 46+β ) was incubated with SurA and proteoliposomes containing the indicated concentration of BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . A quantitation of the data is shown . ( B ) Urea-denatured EspP ( 46+β ) ( 0 . 1 μM ) was mixed with SurA and proteoliposomes containing 0 . 2 μM BamABCDE , and the mixture was divided into two equal reactions . After 5 min fresh EspP ( 46+β ) ( 0 . 2 μM ) was added to one reaction . Aliquots from both reactions were placed on ice at various time points , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . The graph shows a quantitation of the signal produced by the binding of a fluorescently labeled secondary antibody to the cleaved β domain on the Western blot . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 011 To confirm that the ∼30 kDa fragment we observed was identical to the polypeptide that results from proteolytic maturation in vivo , we examined the assembly of a non-cleavable version of EspP ( 46+β ) . This derivative , which is designated EspP* ( 46+β ) , contains a mutation at the cleavage site that abolishes proteolysis but does not affect folding ( Skillman et al . , 2005 ) . A polypeptide that migrated slower than the folded β domain ( ∼30 kDa vs ∼27 kDa ) and that corresponds to the folded form of EspP* ( 46+β ) was observed when the samples were not heated ( Figure 3C ) . As expected , this band disappeared when the samples were heated because in the absence of cleavage the denatured form of EspP* ( 46+β ) co-migrates with the unassembled protein . Quantitation of the relative fluorescent signal produced by the denatured form of the EspP β domain and unprocessed EspP ( 46+β ) on Western blots indicated that ∼10–20% of the protein was typically assembled under our experimental conditions . The efficiency of assembly did not appear to be limited by the concentration of either reconstituted BamABCDE or SurA . A moderate increase in efficiency was observed when the BamABCDE concentration was increased from 0 . 05 μM to 0 . 2 μM ( the concentration used in the experiments described here ) , but no increase was observed at higher concentrations ( Figure 3—figure supplement 4A ) . Interestingly , if we added fresh substrate 5ʹ after the start of the incubation , the total yield of assembled EspP ( 46+β ) increased in proportion to the amount of added substrate ( Figure 3—figure supplement 4B ) . This observation is consistent with other evidence that the Bam complex can catalyze multiple rounds of assembly in vitro ( Hagan and Kahne , 2011 ) . In light of our results , it is likely that the efficiency of EspP ( 46+β ) assembly was limited by the inherent ability of the denatured protein to remain assembly competent after dilution out of 8 M urea . We next wished to determine whether the factors that promote the assembly of the EspP β domain can also promote the translocation of substantial passenger domain fragments . To this end , we expressed and purified EspP derivatives that contain progressively longer portions of the native passenger domain plus 20 N-terminal residues derived from the cloning vector ( Figure 3—figure supplement 1 ) . Because proteolytic maturation of EspP is absolutely dependent on the completion of passenger domain translocation in vivo ( Ieva and Bernstein , 2009; Peterson et al . , 2010; Pavlova et al . , 2013 ) , we initially assessed translocation by monitoring the appearance of the cleaved β domain . Indeed available evidence strongly suggests that the completion of β domain assembly ( and the assembly of the active site ) is a late event in autotransporter biogenesis ( Ieva et al . , 2011 ) , possibly because the β domain forms at least part of the passenger domain transport channel ( Figure 1C ) . As Western blot analysis using the anti-EspP C-terminal antiserum indicated , EspP derivatives that have passenger domains ranging in size from 72–734 residues were all processed with approximately the same efficiency as EspP ( 46+β ) ( Figure 4 ) . The largest derivative lacks the N-terminal globular segment of the EspP passenger domain , but contains the entire β helical segment ( Figure 1A ) . These results not only suggest that BamABCDE and SurA are sufficient to promote passenger domain translocation , but also suggest that the passenger domain does not readily adopt a translocation-incompetent conformation that independently limits the efficiency of autotransporter assembly . 10 . 7554/eLife . 04234 . 012Figure 4 . BamABCDE and SurA catalyze the assembly of longer EspP derivatives . The indicated urea-denatured EspP derivative was incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were removed at various time points , heated to 95°C or maintained at room temperature after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . The ∼27 kDa polypeptide observed at the 0 min time point on the bottom two gels is an unidentified background band . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 012 To obtain additional evidence that EspP derivatives that contain significant passenger domain fragments assemble correctly in the in vitro assay , we selected a few derivatives for further analysis . Initially we found that after incubating EspP ( 96+β ) with proteoliposomes containing BamABCDE and SurA for 30 min , the cleaved β domain was both subject to heat denaturation in SDS and resistant to proteinase K ( PK ) digestion ( Figure 5A ) . This result provided direct evidence that the β domain was not only properly folded , but also inserted into the proteoliposomes . Subsequently we examined the assembly of EspP* ( 96+β ) , a non-cleavable version of EspP ( 96+β ) , under the same conditions . In the absence of PK , a band that migrates slower than the folded β domain ( ∼35 kDa vs ∼27 kDa ) and that corresponds to the folded form of the full-length protein was observed ( Figure 5B ) . The finding that the ∼35 kDa polypeptide was resistant to PK digestion unless detergent was added showed that the passenger domain was translocated into the lumen of the proteoliposomes . We also performed an initial analysis of the assembly of EspP ( HA-251+β ) , a derivative that contains 251 residues of the EspP passenger domain plus an N-terminal HA tag ( 14 additional residues ) . After we incubated the protein with proteoliposomes containing BamABCDE and SurA we detected the β domain and the cleaved passenger domain on Western blots probed with anti-EspP C-terminal and anti-HA antisera , respectively , and both fragments were resistant to PK digestion unless detergent was added ( Figure 5C ) . Because the epitope tag is the last segment of the protein that traverses the membrane , these results confirm that the translocation reaction went to completion . 10 . 7554/eLife . 04234 . 013Figure 5 . EspP derivatives are correctly assembled into proteoliposomes . ( A ) Urea-denatured EspP ( 96+β ) was incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were placed on ice after 0 and 30 min and either treated with PK or left untreated . After the addition of SDS-PAGE buffer samples were heated to 95°C or maintained at room temperature and analyzed by Western blot using an anti-EspP C-terminal peptide . ( B ) The experiment described in ( A ) was repeated with EspP* ( 96+β ) , except that n-dodecyl β-D-maltoside ( DDM ) was added to one sample prior to PK treatment , and none of the samples were heated after the addition of SDS-PAGE buffer . ( C ) The experiment described in ( A ) was repeated with EspP ( HA-251+β ) , except that DDM was added to one sample prior to PK treatment , and all of the samples were heated to 95°C after the addition of SDS-PAGE buffer . Samples were divided in half and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum or an anti-HA antiserum . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 013 To examine the biogenesis of EspP derivatives that have different length passenger domains in more detail , we next performed a kinetic analysis of the assembly of EspP ( 46+β ) and EspP ( HA-251+β ) . We used 0 . 2 μM substrate in these experiments because we found that doubling the EspP concentration moderately increased the efficiency of assembly . As suggested by a study on OmpT , the enhancement of assembly might be a fortuitous effect of increasing the concentration of urea , which presumably helps to maintain the folding-competence of the substrate ( Hagan and Kahne , 2011 ) . Western blot analysis using the anti-EspP C-terminal antiserum showed that EspP ( 46+β ) underwent substantial proteolytic processing within 2 min ( Figure 6A , top and Figure 6B , blue curve ) . Interestingly , EspP ( HA-251+β ) was processed at a very similar rate ( Figure 6C , top and Figure 6D , blue curve ) . Indeed the rate constants obtained by fitting the assembly data for the two EspP derivatives to either a single exponential or two-step model were nearly identical ( Supplementary file 1 ) . In the case of the EspP ( HA-251+β ) derivative , a PK-resistant ∼30 kD polypeptide that corresponds to the cleaved passenger domain accumulated in parallel with the cleaved β domain ( Figure 6E , top ) . This observation confirmed that the passenger domain was translocated into the lumen of the proteoliposomes during the assembly reaction . We also introduced a mutation ( G1066A ) that slightly impairs folding of the β domain and thereby delays the initiation of passenger domain translocation by ∼1 min in vivo ( Pavlova et al . , 2013 ) into EspP ( 46+β ) and EspP ( HA-251+β ) . The mutation caused a similar delay in the assembly of both EspP derivatives in vitro ( Figure 6 and Supplementary file 1 ) . This observation corroborates the conclusion that defects in the folding of the β domain affect the ability of the Bam complex to catalyze subsequent steps in EspP assembly . 10 . 7554/eLife . 04234 . 014Figure 6 . The assembly of the β domain is the rate-limiting step in EspP biogenesis . ( A ) Urea-denatured EspP ( 46+β ) or EspP ( 46+β ) G1066A was incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . ( B ) Quantitation of the data shown in ( A ) . ( C ) The experiment shown in ( A ) was repeated with EspP ( HA-251+β ) or EspP ( HA-251+β ) . ( D ) Quantitation of the data shown in ( C ) . ( E ) A second aliquot from the experiment shown in ( C ) was placed on ice at each time point and treated with PK . After the addition of SDS-PAGE buffer the samples were heated to 95°C and analyzed by Western blot using an anti-HA antiserum . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 014 Interestingly , we found that an EspP derivative that contains an HA-tagged version of the full-length β helix [EspP ( HA-714+β ) ] was also assembled very rapidly . Despite the presence of a much larger passenger domain , EspP ( HA-714+β ) was processed only slightly more slowly than EspP ( 46+β ) ( Figure 7A , B ) . Furthermore , the assembly data could be fit equally well to the same kinetic models and did not suggest the existence of an additional slow step ( Supplementary file 1 ) . A PK-resistant ∼80 kDa polypeptide that corresponds to the cleaved passenger domain accumulated during the assembly reaction ( Figure 7C ) . As expected , this polypeptide was degraded when the proteoliposomes were solubilized with detergent ( Figure 7D ) . Taken together , the results indicate that passenger domain translocation is relatively fast , and that the assembly of the β domain is the rate-limiting step in autotransporter biogenesis . Curiously , the introduction of a short linker into EspP ( HA-714+β ) that impairs passenger domain folding and stalls translocation in vivo ( Ieva and Bernstein , 2009 ) only modestly affected assembly in vitro ( Figure 7—figure supplement 1 ) . This observation , along with the finding that an EspP chimera containing a >200 residue intrinsically disordered passenger domain segment that is secreted efficiently in vivo could also assemble in the in vitro assay ( Figure 7—figure supplement 2 ) , suggests that in the absence of an exogenous energy source translocation is not driven exclusively by protein folding . 10 . 7554/eLife . 04234 . 015Figure 7 . Assembly kinetics of an EspP derivative containing the full-length β helix . ( A ) Urea-denatured EspP ( HA-714+β ) was incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . ( B ) Quantitation of the data shown in ( A ) . ( C ) A second aliquot from the experiment shown in ( A ) was placed on ice at each time point and treated with PK . After the addition of SDS-PAGE buffer the samples were heated to 95°C and analyzed by Western blot using an anti-HA antiserum . ( D ) The experiment described in ( A ) was repeated . Two equal aliquots were placed on ice after 30 min and DDM was added to one aliquot . The samples were then treated with PK and analyzed by Western blot as described in ( C ) . In ( C and D ) the asterisk denotes an unidentified background band . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 01510 . 7554/eLife . 04234 . 016Figure 7—figure supplement 1 . Effect of a linker insertion in the passenger domain on the assembly of EspP ( HA-714+β ) . ( A ) Urea-denatured EspP ( HA-714+β ) TEV586 was incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . ( B ) Quantitation of the data shown in part A ( red curve ) and Figure 7A ( blue curve ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 01610 . 7554/eLife . 04234 . 017Figure 7—figure supplement 2 . BamABCDE and SurA catalyze the assembly of RD-EspP chimeras . Urea-denatured RD1512–1619-EspP ( 51+β ) and RD1457–1677-EspP ( 51+β ) , which contain 108 and 221 residue fragments of the intrinsically disordered receptor domain ( RD ) of the B . pertussis RTX toxin fused to the C-terminus of EspP , were incubated with SurA and proteoliposomes containing BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C or maintained at room temperature after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . Chimeric RD-EspP passenger domains have been shown to be secreted efficiently in vivo ( Kang’ethe and Bernstein , 2013a ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 017 Although it is often assumed that OM protein assembly is catalyzed by a Bam complex monomer , this notion has never been tested . Furthermore , given that autotransporter assembly is inherently more complicated than the assembly of generic β barrel proteins , it is conceivable that the Bam complex exists in a distinctive oligomeric state during its interaction with autotransporters . To determine whether a single copy of the Bam complex can mediate the assembly of EspP derivatives , we reconstituted BamABCDE into ∼10–12 nm nanodiscs using conditions that facilitate the incorporation of a single copy of the ABC transporter MalFGK2 ( Bao et al . , 2012 ) . The nanodiscs eluted at about the same position on gel filtration columns as BamABCDE monomers ( Figure 8A ) . Analysis of the peak fractions by SDS-PAGE suggested that the scaffold protein MSP1D1 was present in excess over the Bam complex proteins ( Figure 8B ) . Because nanodiscs are stabilized by two copies of the scaffold protein ( Denisov et al . , 2004 ) , however , this observation is consistent with the prediction that they contained a single copy of the Bam complex . In addition , the nanodiscs produced a single band on Blue Native PAGE that migrated only slightly slower than a BamABCDE monomer ( compare Figure 8C to Figure 1C ) . Western blot analysis using the anti-EspP C-terminal antiserum revealed that both EspP ( 46+β ) and EspP ( HA-251+β ) were assembled into nanodiscs about as efficiently as they were assembled into proteoliposomes ( Figure 8D ) . Although the non-vesicular nature of nanodiscs precluded assessment of passenger domain translocation using a PK-sensitivity assay , we found that the ∼30 kDa passenger domain of EspP ( HA-251+β ) accumulated in parallel with the cleaved β domain ( Figure 8E ) . These results not only provide evidence that a Bam complex monomer is sufficient to catalyze autotransporter assembly , but also show that the curvature of the proteoliposome membrane is not itself required for the assembly process . 10 . 7554/eLife . 04234 . 018Figure 8 . BamABCDE and SurA catalyze the assembly of EspP derivatives into nanodiscs . ( A ) Chromatograms of nanodiscs and BamABCDE in DDM on Superdex 75 gel filtration column . ( B ) SDS-PAGE analysis of the peak fractions in ( A ) . Proteins were visualized by staining the gel with Coomassie Blue . ( C ) The peak fractions in ( B ) were pooled and analyzed by Blue Native PAGE . ( D ) Urea-denatured EspP ( HA-251+β ) or EspP ( 46+β ) was incubated with SurA and nanodiscs containing BamABCDE . Aliquots were placed on ice at various time points , heated to 95°C or maintained at room temperature after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-EspP C-terminal peptide antiserum . The ∼27 kDa polypeptide observed at the 0 min time point on the top gel is an unidentified background band . ( E ) An additional aliquot from the EspP ( HA-251+β ) assembly reaction shown in ( D ) was removed at each time point , heated to 95°C after the addition of SDS-PAGE buffer , and analyzed by Western blot using an anti-HA antiserum . DOI: http://dx . doi . org/10 . 7554/eLife . 04234 . 018
In this study we identified the minimal set of factors that are required for the complete assembly of the model autotransporter EspP . Initially we showed that the Bam complex reconstituted into proteoliposomes and SurA and both necessary and sufficient to catalyze the folding and proteolytic processing of a simplified version of EspP that consists of the β domain plus a short passenger domain fragment that protrudes only slightly from the β domain pore . This derivative is similar to generic β barrel proteins such as OmpT that reside almost entirely in the OM , except that it contains an internal α-helical segment . In subsequent experiments we found that the same two factors are sufficient to catalyze the translocation of passenger domain fragments associated with much larger EspP derivatives . Interestingly , the efficiency and kinetics of assembly was largely independent of the length of the derivative . While this observation does not exclude the possibility that there are other factors that stimulate passenger domain secretion in vivo , it does suggest that they would either play a limited role in the translocation reaction per se or serve to prevent misfolding in the crowded periplasmic space . Finally , experiments using nanodiscs provided evidence that EspP assembly is catalyzed by a single copy of the Bam complex . Overall , our results are consistent with a model in which the passenger domain is secreted through a channel consisting of the open β domain and/or the BamA β barrel in an open conformation ( Figure 1C ) . Our work demonstrates that passenger domain secretion does not require either an input of exogenous energy or the presence of IM proteins that transduce energy from the cytoplasm or the membrane potential . The results are surprising because ATP hydrolysis is required to drive protein translocation through the Sec complex in an analogous reconstituted assay system ( Brundage et al . , 1990 ) and because other types of protein translocation reactions appear to require a significant energy expenditure ( Alder and Theg , 2003; Shi and Theg , 2013 ) . The high energy cost may result in part from a tendency of proteins to slide backwards at specific stages of the translocation reaction ( Schiebel et al . , 1991; Bauer et al . , 2014 ) . In the autotransporter pathway , the folding of the passenger domain in the extracellular space has been shown to play a role in driving translocation in vivo , and folding may also promote translocation in the in vitro assay . Available evidence , however , neither supports the notion that stepwise folding alone drives translocation nor explains the efficient secretion of the intrinsically disordered RD domain or polypeptides that fold in the periplasm . As previously suggested ( Kang’ethe and Bernstein , 2013a ) , electrostatic or other types of interactions between passenger domains and the Bam complex or membrane lipids may facilitate translocation . Like the FimD protein that secretes type I pilus subunits , the Bam complex might also guide passenger domains along a low-energy pathway ( Geibel et al . , 2013 ) . Alternatively , the Bam complex might catalyze at least the initial stages of translocation and the assembly of the β domain in a concerted fashion . Indeed a model in which a polypeptide segment is pushed across the membrane during the membrane integration of the β domain might explain the efficient secretion of ∼100 residue fragments that are too short to fold ( Skillman et al . , 2005; Pavlova , et al . , 2013 ) . While the coupling of translocation to β domain assembly might also explain the assembly of a folding-deficient mutant such as EspP ( HA-714+β ) 586TEV in vitro , the observation that stalled translocation reactions can be restarted in vivo ( Ieva and Bernstein , 2009; JHP and HDB , unpublished results ) strongly suggests that energy can be harnessed after the insertion of the β domain is largely complete . Our results also indicate that the assembly of the β domain is the rate-limiting step in autotransporter biogenesis . Pulse-chase labeling and photocrosslinking experiments have previously provided evidence that β domain assembly is slower than passenger domain translocation in vivo ( Ieva et al . , 2011 ) , but the interpretation of these experiments is complicated by the fact that the synthesis of the 1300 residue EspP protein is itself rather slow ( ∼45 s ) . The use of fully synthesized EspP derivatives in the in vitro assay has enabled us to circumvent this problem and assess the contribution of the translocation step to the overall reaction kinetics more effectively . It is noteworthy that the ability of the β domain to remain assembly-competent also appeared to limit the efficiency of the assembly reaction . This observation corroborates the conclusion that even long passenger domain segments fold slowly and consequently resist aggregation , at least in vitro ( Junker et al . , 2006 ) . Furthermore , given that the β domain clearly interacts with Skp in vivo ( Ieva et al . , 2011; Pavlova et al . , 2013 ) , it is striking that the presence of the chaperone inhibited assembly in vitro . Presumably we could not recapitulate the productive interaction between Skp and the β domain that occurs in vivo , or a factor that is required to release Skp from client proteins was not present ( or was present but not fully functional ) in our assay . Finally , the strategy that we have described to purify the Bam complex may facilitate future studies on the assembly of both autotransporters and many other OM proteins . It is striking that , at least in our hands , BamABCDE was more active than Bam ( AB ) ( CDE ) . While this disparity may simply be due to technical issues , our results also raise the possibility that the structure of the Bam complex assembled in vivo differs from that of the Bam complex reconstructed from BamAB and BamCDE subcomplexes in vitro . Indeed it is conceivable that while a stable heterooligomer can be formed from the two subcomplexes , the Bam complex is actually assembled by a different pathway inside living cells . In this regard it should be noted that in preliminary experiments we obtained evidence that the Bam complex remains intact in vivo and does not undergo a dynamic cycle in which BamAB and BamCDE subcomplexes rapidly dissocate and reassociate ( JHP and HDB , unpublished results ) .
Previously described plasmids ( Hagan et al . , 2010 ) that encode the BamAB and BamCD subcomplexes , His-tagged BamE , SurA and the His-tagged OmpT G236K/K237G mutant ( pSK38 , pSK46 , pBamE-His , pSK257 and pCH28 ) were reconstructed . To generate plasmid pJH113 , a new Nde I site was introduced into the polylinker of a derivative of pTRC99a that lacks the endogenous Nde I site ( Szabady et al . , 2005 ) using the QuikChange Mutagenesis Kit ( Agilent , Santa Clara , CA ) with the oligonucleotide pTRC/Nde and its complement ( all oligonucleotides used in this study are listed in Supplementary file 2 ) . The genes that encode all five subunits of the Bam complex were then cloned sequentially into pJH113 to create plasmid pJH114 . Each gene was amplified by PCR using genomic DNA from E . coli strain AD202 as a template ( Akiyama and Ito , 1990 ) . BamA was first cloned into the Nde I and BamH I sites of pJH113 , and the other genes were then cloned into the BamH I site of the resulting plasmid . In the final round of cloning an octahistidine tag was added to the C terminus of BamE during PCR amplification . Plasmids pJH115 , pJH116 and pJH117 , which encode BamACDE ( His8 ) , BamABDE ( His8 ) , and BamCDE ( His8 ) , respectively , were constructed in the same fashion except that one or more of the cloning steps were omitted . Plasmid pJH118 , which encodes ( His6 ) BamAB , was made by first introducing an Eag I site into pJH113-bamA using oligonucleotide BamA . Eag ( + ) and its complement . A hexahistidine tag was then introduced at the N-terminus of the mature region of BamA using oligonucleotides BamAHis ( + ) and BamAHis ( − ) . Finally , a Kpn I-Xba I fragment encoding the C-terminus of BamA and BamB was excised from pJH113-bamAB and cloned into the cognate sites of this plasmid . A plasmid encoding the MSP1D1 protein was previously described ( Denisov et al . , 2004 ) and was obtained from Addgene ( plasmid 20061 ) . To make plasmids that express His6-tagged EspP derivatives , fragments of espP were amplified by PCR using the primer EspP ( − ) and an appropriate upstream oligonucleotide and pRLS5 , pJH62 or pKMS3 as a template ( Skillman et al . , 2005; Szabady et al . , 2005 ) . The resulting PCR products were digested with Nde I and BamH I and cloned into the cognate sites of pET28b . To construct plasmids encoding HA-tagged EspP derivatives , the oligonucleotides HA tag ( + ) and HA tag ( − ) ( Supplementary file 2 ) were cloned into the Nco I and Nde I sites of plasmids encoding EspP ( 271+β ) and EspP ( 734+β ) . To make plasmids expressing RD-EspP chimeras , fragments of the RD gene were amplified using appropriate primers and pWK2 ( Kang’ethe and Bernstein , 2013a ) as a template . PCR products were then digested with Nco I and Nde I and cloned into the cognate sites of pET28b encoding EspP ( 51+β ) . BamAB and BamCDE ( His8 ) were produced independently essentially as described ( Hagan et al . , 2010 ) . BL21 ( DE3 ) transformed with pSK38 were grown at 37°C in a 1 l volume to OD600 = 0 . 3 . Cultures were shifted to 25°C over 30 min and bamAB overexpression was induced by the addition of 0 . 1 mM IPTG at OD600 = 0 . 5–0 . 6 . Cells were then incubated in the presence of the inducer for 3 hr . BL21 ( DE3 ) transformed with both pSK46 and pBamE-His were grown at 37°C in a 1 l volume to OD600 = 0 . 5–0 . 6 . BamCD and bamE ( His8 ) expression was induced by the addition of 0 . 1 mM IPTG , and incubation was continued for an additional 3–4 hr . Cells were pelleted in a Beckman JLA-8 . 1000 rotor ( 4000×g , 15 min ) , resuspended in 10 ml cold 20 mM Tris–HCl pH 8 and lysed using an Avestin EmulsiFlex C3 ( 3–4 passes ) . Lysates were centrifuged at 5000×g at 4°C for 10 min to remove unbroken cells . The supernatants were centrifuged in a Beckman Ti70 rotor ( 50 , 000 rpm , 30 min , 4°C ) to pellet the membranes , which were solubilized in 10 ml TBS ( 25 mM Tris , pH 7 . 4137 mM NaCl , 3 mM KCl ) containing 2% Triton X-100 , 10 mM EDTA , and 10 μg/ml lysozyme for 30 min at room temperature . The ultracentrifugation step was repeated and the resulting supernatants were dialyzed overnight against cold buffer A ( TBS/0 . 5% Triton X-100 ) . BamAB was then further enriched by gel filtration on an S-200 column ( GE Healthcare ) equilibrated with buffer A . Partially purified BamAB and BamCDE were mixed in at least a 5:1 ratio and rotated in the presence of 2 ml Ni-NTA agarose for 1 hr at 4°C . The Ni-NTA beads were washed with one column volume buffer A containing 50 mM imidazole . The assembled Bam ( AB ) ( CDE ) complex was eluted with 3 . 5 ml buffer A containing 500 mM Imidazole and injected into an S-200 column equilibrated with TBS pH 8 , 0 . 03% n-dodecyl-β-D-maltoside ( DDM ) , 1 mM tris ( 2-carboxyethyl ) phosphine ( TCEP ) . The column was run at 0 . 5 ml/min and 1 ml fractions were collected . Portions of each fraction were analyzed by SDS-PAGE to identify fractions that contained complete Bam complexes . Typically 5–7 fractions were pooled and concentrated ∼10-fold using Amicon Ultra-15 centrifugal filters ( Millipore , Billerica , MA ) . The concentration of the purified Bam complex was then determined using the Bio-Rad DC Protein Assay following the manufacturer's instructions . His-tagged SurA was overproduced and purified essentially as described ( Hagan et al . , 2010 ) . In brief , 1 l cultures of BL21 ( DE3 ) transformed with pSK257 were grown at 37°C to OD600 = 1 and shifted to 16°C . Cultures were incubated overnight after the addition of 0 . 1 mM IPTG ( 0 . 1 mM ) . Cells were then harvested as described above , resuspended in 10 ml cold 20 mM Tris HCl pH 8 , and lysed using an EmulsiFlex C3 . Lysates were clarified by centrifugation at 35 , 000×g at 4°C for 20 min . SurA was purified from the resulting supernatant using two consecutive rounds of TALON affinity chromatography ( Clontech , Mountain View , CA ) following the manufacturer's instructions . The purified protein was dialyzed overnight against 20 mM Tris–HCl pH 8 to remove the imidazole . E . coli strain HDB150 ( MC4100 ompT::spc ΔaraBAD leuD::kan ) transformed with pJH114 was grown overnight at 37°C in LB containing 100 μg/ml ampicillin . The cells were washed and diluted 1:80 into 1–2 l of fresh medium . When the cultures reached OD600 = 0 . 5–0 . 6 , 0 . 4 mM IPTG was added to induce the expression of bamABCDE . Cells were grown in the presence of the inducer for 1 . 5 hr and then harvested as described above . Cell pellets were resuspended in 10 ml/l cold 20 mM Tris–HCl pH 8 and cells were lysed using an EmulsiFlex C3 . Lysates were centrifuged at 6000×g at 4°C for 10 min . The supernatants were centrifuged in a Ti 70 rotor as described above to isolate total membranes . After the pellets were incubated in 10 ml/l cold 50 mM Tris pH 8 , 150 mM NaCl , 1% DDM on ice for 1 hr the centrifugation step was repeated . Supernatants containing the soluble membrane proteins were then rotated in the presence of 2 ml/l Ni-NTA agarose for 1 . 5 hr at 4°C . Ni-NTA beads were washed with one column volume buffer B ( 50 mM Tris pH 8 , 150 mM NaCl , 0 . 03% DDM ) containing 50 mM imidazole . BamABCDE was then eluted in 3 . 5 ml buffer A containing 500 mM imidazole and injected onto a S-200 column equilibrated with buffer A . The column was run at 0 . 5 ml/min and 1 ml fractions were collected . Fractions that contained complete BamABCDE complexes were identified , pooled and concentrated , and the concentration of the purified protein was determined as described above . NativePAGE ( Blue Native ) 4–16% Bis-Tris gels ( Life Technologies , Grand Island , NY ) were run to verify that the complex was intact . The same protocol was used to produce and purify BamACDE , BamABDE and the BamCDE and BamAB subcomplexes used in Figure 3B and Figure 3—figure supplement 2 , except that cells were transformed with pJH115 , pJH116 , pJH117 or pJH118 . E . coli phospholipids ( E . coli Polar Lipid Extract , Avanti Polar Lipids , Alabaster , AL ) were suspended in water at a concentration of 20 mg/ml and sonicated until well dispersed . A portion of the phospholipid suspension ( 40 μl ) was added to 200 μl of the purified Bam ( AB ) ( CDE ) or BamABCDE ( 20 μM ) and incubated on ice for 5 min . The mixture was then diluted with 4 ml of 20 mM Tris HCl pH 8 and incubated on ice for 30 min to reduce the detergent concentration and promote proteoliposome formation . The proteoliposomes were pelleted in a Beckman TLA100 . 4 rotor ( 50 , 000 rpm , 4°C , 30 min ) and resuspended in 200 μl 20 mM Tris HCl pH 8 . Aliquots of the proteoliposomes were flash frozen in liquid nitrogen and stored at −80°C . BL21 ( DE3 ) transformed with pMSP1D1 ( Denisov et al . , 2004 ) were grown in 1 l cultures at 37°C , and 0 . 5 mM IPTG was added at OD600 = 0 . 7–0 . 8 . Cultures were incubated for 3 hr after the addition of the inducer . Cells were harvested , resuspended in 10 ml cold 20 mM Tris–HCl pH 8 and lysed as described above , and lysates were centrifuged at 35 , 000×g at 4°C for 20 min . Clarified supernatants were mixed with 2 ml NiNTA agarose and rotated for 1 . 5 hr at 4°C . The beads were washed with one column volume buffer D ( 40 mM Tris HCl pH 8 , 300 mM NaCl ) containing 50 mM imidazole . MSP1D1 was then eluted with 5 ml buffer D containing 500 mM imidazole and dialyzed overnight against cold buffer E ( 20 mM Tris HCl pH 7 . 4 , 100 mM NaCl , 0 . 5 mM EDTA ) . The purified protein was then flash frozen and stored at −80°C . The concentration of MSP1D1 was determined spectrophotometrically using a previously determined ε280 value ( http://sligarlab . life . uiuc . edu/nanodisc/protocols . html ) . Before the protein was used to assemble nanodiscs it was diluted to 0 . 3 mM in TSGD buffer ( 50 mM Tris HCl pH 8 , 100 mM NaCl , 0 . 03% DDM , 10% glycerol ) ( Bao et al . , 2012 ) . Initially 10–15 ml ( dry volume ) Bio-Beads ( Bio-Beads SM-2 Adsorbent , Bio-Rad , Hercules , CA ) were washed and stored at 4°C in TS buffer ( 50 mM Tris HCl pH 8 , 50 mM NaCl ) as previously described ( Bao et al . , 2012 ) . Nanodisc reconstitution was performed in a final volume of 300 μl essentially as described ( Bao et al . , 2012 ) by adding 6 μM purified Bam complex , 18 μM MSP1D1 and 360 μM phospholipids ( E . coli Polar Lipid Extract , Avanti Polar Lipids ) to TSGD buffer . The mixture was then incubated with Bio-Beads ( 50 μl ) overnight at 4°C on a rocking platform . The Bio-Beads were then allowed to settle out , and the supernatant was loaded onto a Superdex 75 gel filtration column equilibrated with buffer F ( 50 mM Tris HCl pH8 , 100 mM NaCl , 10% glycerol ) . The column was run at a flow rate was 0 . 5 ml/min , and 0 . 5 ml fractions were collected . Peak fractions were pooled and concentrated to a volume of 300 μl using Amicon Ultra-15 centrifugal filters . OmpT and all EspP derivatives were synthesized without their signal peptides and purified from inclusion bodies . BL21 ( DE3 ) transformed with an appropriate plasmid were grown overnight at 37°C in LB containing 50 μg/ml kanamycin and diluted 1:80 into 1 l fresh medium . When the cultures reached OD600 = 0 . 7 , 0 . 5 mM IPTG was added . The cultures were incubated in the presence of the inducer for 3 hr . The cells were then harvested , resuspended in 10 ml cold TBS , and lysed as described above . Lysates were centrifuged at 5000×g at 4°C for 10 min . Pellets containing the inclusion bodies were resuspended in 10 ml TBS and washed twice by repeating the centrifugation step . After the final wash , pellets were resuspended in 5 ml 8 M urea and incubated at room temperature for 1 hr . Samples were then chilled briefly on ice and centrifuged in a Beckman TLA100 . 4 rotor ( 26 , 000 rpm , 4°C , 20 min ) . The concentration of the overexpressed OM proteins , which were highly enriched in the supernatants , were determined using the Bio-Rad DC Protein Assay . The folding of OmpT was monitored by slightly modifying a previously described assay ( Hagan et al . , 2010 ) . OmpT was diluted from a 293 μM stock to a final concentration of 20 μM and incubated with SurA ( final concentration 140 μM ) in 50 μl 20 mM Tris pH 6 . 5 for 10 min at room temperature . The Bam complex and the fluorogenic peptide Abz-Ala-Arg-Arg-Ala-Tyr ( NO2 ) -NH2 ( New England Peptide , Gardner , MA ) were diluted in 50 μl of the same buffer to a final concentration of 16 μM and 2 mM , respectively . The two sub-reactions were then mixed together and the increase in fluorescence at 30°C was monitored on a Spectramax M5 fluorescent plate reader for 1 hr with readings every 20 s . Fluorescence emission was recorded at 430 nm following excitation at 325 nm . An appropriate EspP derivative , SurA and proteoliposomes containing Bam complex were added successively to 20 mM Tris HCl pH 8 . Reaction components were mixed after each addition . In the experiment shown in Figure 3—figure supplement 3 , Skp ( obtained as a highly purified native trimer from MyBioSource . com , San Diego , CA ) was added at the indicated concentration before SurA . In general the final concentration of reaction components was 0 . 1 μM EspP ( diluted from a 6 μM stock solution in 8 M urea ) , 1 μM SurA and 0 . 2 μM Bam complex . In the experiments shown in Figures 6 and 7 and Figure 7—figure supplement 1 , the EspP and SurA concentrations were raised to 0 . 2 μM and 2 μM , respectively . All assembly reactions were performed at 30°C . At each time point , aliquots were removed and either mixed with an equal volume of 2× SDS-PAGE sample buffer or incubated on ice for 5 min with PK ( 5 μg/ml ) . In some experiments , 1% DDM was added prior to PK treatment . Protease digestions were stopped by the addition of 1 mM PMSF and 2× SDS-PAGE sample buffer . To assess the sensitivity of the EspP β domain to SDS denaturation , half of each sample was heated at 95°C for 5 min while the other half was maintained at room temperature . Proteins were resolved by SDS-PAGE and transferred to nitrocellulose using an iBlot apparatus ( Life Technologies ) . Western blotting was conducted using antisera generated against an EspP C-terminal peptide ( Szabady et al . , 2005 ) or the HA ( Y-11 ) peptide ( Santa Cruz Biotechnology , Dallas , TX ) . Antibody-antigen complexes were detected by incubating filters with an IRDye 680-conjugated goat anti-rabbit antiserum and monitoring fluorescence at 700 nM using an Odyssey infrared imaging system ( Licor , Lincoln , NE ) . Percent substrate cleavage at each time point was defined as 100 × ( cleaved β domain/unprocessed substrate + cleaved β domain ) . Novex 8–16% minigels ( Life Technologies ) were used to monitor protein purifications and to resolve the products of all EspP assembly reactions . Data from EspP assembly experiments was fit to single exponential and lag-phase kinetic models using KaleidaGraph 4 . 1 ( Synergy Software , Reading , PA ) . The percent substrate cleavage at the 30 min time point was defined as maximum assembly and was used to determine a rate constant k1 ( or two rate constants k1 and k2 for a sequential two-step model ) . As in classical enzyme kinetics , t1/2 = ln2/k . | Disease-causing bacteria release molecules called virulence factors to help them infect their host . These virulence factors need to pass through the membrane that surrounds the cell . Indeed , some bacteria , such as Escherichia coli , have two membranes , so some virulence factors need to pass through an extra membrane . One group of virulence factors found in E . coli are called autotransporters . These proteins have two sections: the passenger domain , which is the main part of the virulence factor , and the β domain , which anchors the autotransporter in the outer membrane . Once the passenger domain is outside the cell , the link to the β domain can be broken to release the virulence factor . However , we do not know how the passenger domain passes through the outer membrane . By studying an E . coli autotransporter called EspP , Roman-Hernandez et al . have now identified the other proteins that are required for the β domain to insert into an artificial membrane , and allow the passenger domain to pass through the membrane . These other proteins are a group of proteins called the Bam complex and a chaperone protein called SurA . The experiments also show that an external source of energy is not needed to drive this process , and they suggest that the passenger domain moves through a hole in the outer membrane formed by the β domain and/or the Bam complex . Roman-Hernandez et al . also developed a new way to purify the Bam complex that should help all researchers working on this set of proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2014 | Reconstitution of bacterial autotransporter assembly using purified components |
The DNA methyltransferase Dnmt3a suppresses tumorigenesis in models of leukemia and lung cancer . Conversely , deregulation of Dnmt3b is thought to generally promote tumorigenesis . However , the role of Dnmt3a and Dnmt3b in many types of cancer remains undefined . Here , we show that Dnmt3a and Dnmt3b are dispensable for homeostasis of the murine epidermis . However , loss of Dnmt3a-but not Dnmt3b-increases the number of carcinogen-induced squamous tumors , without affecting tumor progression . Only upon combined deletion of Dnmt3a and Dnmt3b , squamous carcinomas become more aggressive and metastatic . Mechanistically , Dnmt3a promotes the expression of epidermal differentiation genes by interacting with their enhancers and inhibits the expression of lipid metabolism genes , including PPAR-γ , by directly methylating their promoters . Importantly , inhibition of PPAR-γ partially prevents the increase in tumorigenesis upon deletion of Dnmt3a . Altogether , we demonstrate that Dnmt3a and Dnmt3b protect the epidermis from tumorigenesis and that squamous carcinomas are sensitive to inhibition of PPAR-γ .
DNA methylation is an epigenetic mechanism that regulates several aspects of gene expression , such as long-term gene silencing , transcriptional elongation , and maintenance of genomic stability ( Allis and Jenuwein , 2016; Avgustinova and Benitah , 2016; Rinaldi and Benitah , 2015 ) . It is found throughout the vertebrate genome and is deposited by DNA methyltransferases on the fifth position of cytosine ( 5-mC ) , predominantly at CpG dinucleotides . The role of DNA methylation in establishing different cell fates during embryogenesis is fairly well understood . However , if and how DNA methylation is necessary to stably maintain the identity of adult stem cells , and how this process is disrupted during oncogenic transformation , is under intense investigation ( Shen and Laird , 2013 ) . Three DNA methyltransferases are encoded in the vertebrate genome . Dnmt1 is predominantly associated with the maintenance of DNA methylation following cell division due to its high affinity for hemimethylated DNA . Consequently , depletion of Dnmt1 leads to a significant reduction of the global levels of 5-mC ( Li et al . , 1992; Lei et al . , 1996 ) . Dnmt3a and Dnmt3b are de novo DNA methyltransferases that establish genome-wide DNA methylation during mammalian embryogenesis and adult stem cell homeostasis ( Okano et al . , 1999 ) . In mouse embryonic stem cells , the combined loss of Dnmt3a and Dnmt3b leads to the progressive loss of DNA methylation , suggesting that these enzymes are additionally involved in maintaining 5-mC levels ( Chen et al . , 2003 ) . Since Dnmt3a-null mice die perinatally , and ablation of Dnmt1 and Dnmt3b results in embryonic lethality , conditional deletion mouse models have been necessary to study their functions in adulthood ( Ueda et al . , 2006; Okano et al . , 1999 ) . Hematopoietic stem cells ( HSCs ) lacking Dnmt3a cannot differentiate correctly upon serial transplantation , and end up developing a range of severe myeloid and lymphoid malignancies in aged animals ( Challen et al . , 2011; Mayle et al . , 2015 ) . Conversely , HSCs-depleted of Dnmt3b show no phenotypical differences with respect to wild-type controls , whereas combined ablation of Dnmt3a and Dnmt3b in HSCs result in an enhanced block of hematopoietic differentiation as compared to Dnmt3a loss alone ( Challen et al . , 2014 ) . Interestingly , the observed phenotype seems specific to stem cells , as the fully differentiated cardiac myocytes carrying a combined deletion of Dnmt3a and Dnmt3b are indistinguishable from wild-type controls ( Nührenberg et al . , 2015 ) . Similarly , to HSCs , purified murine neural stem cells ( SCs ) lacking Dnmt3a do not show problems with self-renewal but fail to differentiate properly ( Wu et al . , 2010 ) . In addition , Dnmt3a acts as a potent tumor suppressor in lung tumorigenesis , to promote adenoma progression but not initiation , downstream of oncogenic K-Ras ( Gao et al . , 2011 ) . This is in contrast with the pro-tumorigenic activity of Dnmt3b , which at least in the murine intestinal epithelium cooperates with the loss of APC to drive adenoma initiation and growth ( Steine et al . , 2011; Lin et al . , 2006 ) Recently , progress has been made in identifying the molecular mechanisms underlying the biological functions of Dnmt3a and Dnmt3b by studying their genome-wide localization . For instance , Dnmt3b associates with and methylates the gene bodies of actively transcribed genes in murine embryonic SCs and human embryonic carcinoma cells ( Baubec et al . , 2015; Jin et al . , 2012; Morselli et al . , 2015 ) . Likewise , it has been proposed that gene body methylation is responsible of most of the transcriptional changes underlying the ability of Dnmt3a to promote neural SCs differentiation , and in protecting the lung epithelium from tumor progression ( Wu et al . , 2010; Gao et al . , 2011 ) . We have recently reported that Dnmt3a and Dnmt3b are required for the self-renewal of human keratinocyte progenitors , whereas Dnmt3a is also required for their proper differentiation ( Rinaldi et al . , 2016 ) . Mechanistically , Dnmt3a and Dnmt3b bind to and promote the activity of enhancers in both human epidermal progenitors and differentiated keratinocytes ( although Dnmt3a having a stronger affinity for enhancers in differentiated keratinocytes ) . Interestingly , both proteins preferentially associate to super-enhancers rather than typical enhancers . Nonetheless , they differ in their mechanism of action , since Dnmt3a ( together with Tet2 ) is essential to maintain high levels of 5-hydroxymethylcytosine ( 5-hmC ) at the center of its target enhancers , while Dnmt3b promotes 5-mC along the body of the enhancer . These regulatory regions dictate the transcription of essential genes necessary for epidermal stem cell identity and maintenance , such as FOS , ITGA6 , TP63 , and KRT5 . Similar to its role in mouse ES cells , Dnmt3b also binds to and methylates the gene bodies of these genes to reinforce their expression ( Rinaldi et al . , 2016 ) . Dnmt3a also associates to the enhancers regulating the expression of genes such as IVL , LOR , FLG2 , and KRT1 which drive the differentiation of SCs into mature keratinocytes ( Rinaldi et al . , 2016 ) . Interestingly , DNA methylation at active enhancers has also been recently reported in normal and cancer-derived human colon , mammary and prostate epithelial cells ( Charlet et al . , 2016 ) . However , to date , no in vivo studies have investigated the roles of Dnmt3a and Dnmt3b in adult epidermal function and malignant transformation . Using mouse models carrying an epidermis-specific ablation of either Dnmt3a or Dnmt3b , or both , we demonstrate that Dnmt3a and Dnmt3b are dispensable for skin homeostasis . However , Dnmt3a has a critical role in suppressing carcinogen-induced squamous tumor initiation , but not progression , while both Dnmt3a and Dnmt3b concertedly prevent tumor progression .
We first studied the pattern of expression of Dnmt3a and Dnmt3b during epidermal development , and in adulthood . At E14 . 5 , Dnmt3a was expressed in the entire Keratin-14+ compartment comprising the basal layer of the embryonic epidermis and the hair placodes ( Figure 1—figure supplement 1A ) . At P0 , all Keratin-14+ basal cells were positive for Dnmt3a with the exception of the highly proliferative hair follicle bulb cells ( Figure 1—figure supplement 1A–B ) . By the time animals reached adulthood , Dnmt3a levels remained high in the hair follicle bulge where most hair follicle stem cells reside ( Solanas and Benitah , 2013 ) , and decreased in the interfollicular epidermis , although some basal IFE cells expressed high levels ( Figure 1—figure supplement 1A–D ) . On the other hand , we were not capable of detecting Dnmt3b by immunofluorescence staining in sections of developing or adult mouse epidermis ( not shown ) , suggesting that Dnmt3b is expressed at low levels ( Challen et al . , 2014 ) . In fact , RNA-seq data confirmed that Dnmt3a was enriched almost fivefold as compared to Dnmt3b in adult basal IFE keratinocytes ( Figure 1—figure supplement 1C ) . However , Dnmt1 , the main DNA methyltransferase , was the most abundant DNA methyltransferase , both in interfollicular epidermis and in hair follicle bulge cells ( Figure 1—figure supplement 1C ) . To gain insight to the roles of Dnmt3a and Dnmt3b in epidermal tissue function , we generated epidermis-specific conditional knockout ( cKO ) mice by crossing animals containing the Dnmt3a or Dnmt3b gene flanked by loxP sites with animals carrying the Keratin14-CRE-ROSA26-YFP-cassette ( hereafter referred to as Dnmt3a/3b-cKO ) ( Gao et al . , 2011 ) . Surprisingly , neither Dnmt3a- nor Dnmt3b-cKO displayed noteworthy epidermal phenotypical differences as compared to their wild-type littermates at different postnatal ages ( Figure 1—figure supplement 1D–E and Figure 2—figure supplement 1A ) . Likewise , despite its strong expression in hair follicle stem cells , the loss of Dnmt3a did not result in evident changes in hair follicle cycling and pelage growth ( Figure 1—figure supplement 1E–F ) . Deregulation of DNA methylation can alter gene expression , leading to tumor suppressor silencing or oncogene activation ( Witte et al . , 2014 ) , and mutation/deregulation of Dnmt3a and Dnmt3b has been observed in several tumor types ( Leppert and Matarazzo , 2014; Subramaniam et al . , 2014 ) . Recently , Dnmt3a has attracted much attention , as it is one of the most frequently mutated genes in cancer ( Kim et al . , 2013 ) , especially in acute myeloid leukemia ( Garg et al . , 2015; Ley et al . , 2010 ) . In fact , a loss-of-function mutation of Dnmt3a is one of the earliest mutations that occurs in human acute myeloid leukemia ( Shlush et al . , 2014 ) . Importantly , these mutations are functional since knock-in mice that model it develop a range of severe myeloid and lymphoid malignancies ( Challen et al . , 2011; Mayle et al . , 2015 ) . In addition , HSCs harboring inactivating mutations of Dnmt3a are clonally selected in ageing humans ( Shlush et al . , 2014 ) . However , much less is known about how deregulation of Dnmt3a and Dnmt3b affect tumorigenesis in epithelial tissues . To elucidate the roles of Dnmt3a and Dnm3b in skin tumorigenesis , we first generated tumors from the epidermis using the chemically induced carcinogenesis protocol based on DMBA/TPA ( Ewing et al . , 1988 ) . The first epidermal squamous malignancies appeared significantly sooner in Dnmt3a-cKO than in their wild-type littermates , after only 2 months from the first DMBA treatment , indicating that Dnmt3a acts as a barrier against tumor initiation ( Figure 1A , B ) . Dnmt3a-cKO animals also showed a significant increase in tumor burden , with an average of 17 tumors per animal compared to three tumors per wild-type animal after 6 months of initiating the experiment ( Figure 1C and Figure 1—source data 1 ) . 10 . 7554/eLife . 21697 . 003Figure 1 . Dnmt3a loss shortens the onset of carcinogen-induced skin neoplasia , and increases tumor burden . ( A ) Representative pictures of wild-type and Dnmt3a-cKO animals after 5 months of treatment with DMBA/TPA . Graph in panel A represents the percentage of animals WT ( n = 6 ) or Dnmt3a-cKO ( n = 6 ) that entered into anagen after 2 weeks of treatment of DMBA/TPA , p=0 . 02 , Chi-Square test . ( B ) Time of appearance , expressed in percentages of skin tumors on wild-type or Dnmt3a-cKO animals , p=0 . 005 . ( C ) Number of skin tumors after 3 or 6 months of DMBA/TPA treatment , p=0 . 001 and p=0 . 0007 . ( D ) Representative images ( hematoxylin/eosin staining ) of different subtypes of skin tumors . ( E ) Histopathological analysis of the different subsets of skin tumors that appeared after DMBA/TPA treatment of wild-type or Dnmt3a-cKO animals . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 00310 . 7554/eLife . 21697 . 004Figure 1—source data 1 . Data related to Figure 1B and Figure 1C . Data showing the days from the DMBA/TPA treatment to the appearance of the first tumors in each wild type and Dnmt3a-cKO mouse ( Figure 1B ) . Data showing the number of tumors counted on the backskin of the wild type and Dnmt3a-cKO after three or 6 months of DMBA/TPA treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 00410 . 7554/eLife . 21697 . 005Figure 1—figure supplement 1 . Dnmt3a is highly expressed in the basal cells of the interfollicular epidermis ( IFE ) , and in the bulge of hair follicles in young mice . ( A–B ) Immunofluorescence staining for Dnmt3a , Keratin 14 and Nuclei of wild type back skin ( A ) and tail skin ( B ) isolated at different ages . ( C ) Fpkm values of the Dnmts from RNA-seq data performed in wild type interfollicular epidermal stem cells ( IFE , n = 4 ) and hair follicle stem cells ( Bulge , n = 3 ) FACS sorted after 6 weeks of DMBA/TPA treatment ( D ) Immunofluorescence staining for Dnmt3a and keratin 14 of the back skin from wild-type or Dnmt3a-cKO animals . ( E ) Representative images ( hematoxylin/eosin staining ) of the back skin from wild-type and Dnmt3a-cKO littermates at different ages . ( F ) Representative Images of one wild type and one Dnmt3a-cKO littermate ( both females ) age of five weeks . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 005 Although Dnmt3a-cKO animals showed a strong increase in tumor initiation and burden , they developed the same percentage of squamous cell carcinomas than wild-type mice ( Figure 1D , E ) . Indeed , a detailed histological analysis of the tumors collected from Dnmt3a-cKO and wild-type animals indicated that Dnmt3a-cKO mice developed the same percentage of benign tumors , such as keratoacanthomas and papillomas , as well as of malignant invasive papillomas and squamous cell carcinomas ( SCCs ) ( Figure 1E ) . Dnmt3a-cKO mice only developed an increase in the percentage of sebaceous adenomas ( Figure 1E ) . No metastases were scored in any of the animals , as expected using this protocol in mice with a C57/Bl6 genetic background ( Sundberg et al . , 1997 ) . Altogether , these results indicate that loss of Dnmt3a dramatically increases initiation of epidermal squamous tumors without affecting their malignant progression , and slightly skews the histology of tumors towards the sebaceous lineage . Dnmt3a suppresses K-Ras-driven lung tumor progression , whereas Dnmt3b is pro-tumorigenic in APC-deficient colorectal adenomas ( Gao et al . , 2011; Lin et al . , 2006 ) . Hence , we next tested whether Dnmt3a and Dnmt3b also exert opposing effects regarding tumorigenesis in the epidermis . Interestingly , there were no differences between wild-type and Dnmt3b-cKO mice with respect to either the timing of tumor initiation or tumor burden upon treatment with DMBA/TPA ( Figure 2A-right panel ) . There were no significant changes in the histological appearance of the tumors , or the number of basal cells proliferating or undergoing apoptosis , between Dnmt3b-cKO and the wild-type controls ( Figure 2—figure supplement 2 and Figure 2—figure supplement 3A–B ) . 10 . 7554/eLife . 21697 . 006Figure 2 . Dnmt3a and Dnmt3b double cKO animals develop more aggressive tumors than wild-type , Dnmt3a-cKO and Dnmt3b-cKO mice . ( A ) Left , representative images ( hematoxylin/eosin staining ) of skin tumors isolated from wild type and Dnmt3b-cKO littermates after 6 months of DMBA/TPA treatment . Right , time of appearance of tumors shown as percentages in wild-type and Dnmt3b-cKO animals , and number of skin tumors after 3 or 6 months of DMBA/TPA treatment . ( B ) Left , representative images ( hematoxylin/eosin staining ) of skin tumors isolated from wild type and Dnmt3a/Dnmt3b DcKO littermates after 6 months of DMBA/TPA treatment . Right , time of appearance of tumors represented as percentages in wild-type and Dnmt3a/Dnmt3b DcKO animals , and number of skin tumors after 3 or 6 months of treatment with DMBA/TPA . ( C–D ) Number of tumors ( left ) and time of appearance ( right ) expressed as percentages , in wild type , Dnmt3a-cKO and DcKO animals after 6 months of DMBA/TPA treatment . ( E ) Histopathological analysis of the different subsets of skin tumors that appeared after DMBA/TPA treatment of wild type or DcKO animals . ( F ) Representative images of metastatic nodules identified only in a percentage ( 33% ) of the lungs of DcKO animals , scale bar = 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 00610 . 7554/eLife . 21697 . 007Figure 2—figure supplement 1 . Deletion of Dnmt3b does not affect epidermal and hair follicle homeostasis . Representative images ( hematoxylin/eosin staining ) of back skin and tail skin from wild type and Dnmt3b-cKO littermates at different ages . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 00710 . 7554/eLife . 21697 . 008Figure 2—figure supplement 2 . Dnmt3b-KO and wild-type skin tumors are histologically indistinguishable . Representative images ( hematoxylin/eosin staining ) of different skin tumors isolated from wild type and Dnmt3b-cKO after 6 months of DMBA/TPA treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 00810 . 7554/eLife . 21697 . 009Figure 2—figure supplement 3 . Dnmt3b-KO tumors do not show changes in proliferation or apoptosis compared to their wild-type counterparts . ( A ) Left- Representative images of immunohistochemistry staining against the cell proliferation marker KI67 in skin tumors isolated from wild type and Dnmt3b-cKO after 6 months of DMBA/TPA treatment . Right- Quantification of KI67 staining in wild type ( n = 5 ) and Dnmt3b-KO ( n = 5 ) skin tumors using the TMarker software . ( B ) Left-Representative images of TUNEL staining to detect apoptosis in skin tumors isolated from wild type and Dnmt3b-cKO after 6 months of DMBA/TPA treatment . Right-Quantification of Tunel staining in wild type ( n = 5 ) and Dnmt3b-KO ( n = 5 ) skin tumors using the TMarker software . Unpaired T-Test was used for statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 00910 . 7554/eLife . 21697 . 010Figure 2—figure supplement 4 . The combined deletion of Dnmt3a and Dnmt3b does not affect epidermal homeostasis . ( A ) Representative images ( hematoxylin/eosin staining ) of back skin and tail skin from adult and aged wild type and DcKO littermates . ( B ) Immunofluorescence staining for 5-methylcytosine and Keratin 14 in aged ( over 70 weeks old ) wild type and DcKO littermates . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01010 . 7554/eLife . 21697 . 011Figure 2—figure supplement 5 . Squamous cell carcinomas in Dnmt3a/Dnmt3b double KO mice express lower levels of epithelial markers compared to wilt-type tumors . ( A ) Representative confocal images for E-Cadherin , Keratin 14 and DAPI in wild type , single Dnmt3a KO and double Dnmt3a/Dnmt3b KO squamous cell carcinomas . ( B ) Representative confocal images for Vimentin , Keratin 14 and DAPI in wild type , single Dnmt3a-cKO and double Dnmt3a/Dnmt3b cKO squamous cell carcinomas . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01110 . 7554/eLife . 21697 . 012Figure 2—figure supplement 6 . The combined deletion of Dnmt3a and Dnmt3b favors the development of skin tumors with features of spindle cell carcinomas . ( A–B ) Hematoxylin/eosin staining and confocal images of two different spindle cell carcinomas developed in two DcKO animals . Representative images of immunofluorescence staining to detect the expression of Vimentin , YFP and Keratin14 . ( C ) Hematoxylin/eosin staining and confocal images of the stroma of a squamous cell carcinoma developed by a wild-type animal showing the absence of expression of Vimentin in the epithelial compartment of the tumor . Immunofluorescence staining shown correspond to Vimentin ( green ) , YFP ( grey ) , and Keratin14 ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 012 To assess whether Dnmt3a and Dnmt3b potentially play redundant roles during tumorigenesis , we also induced tumors using the DMBA/TPA protocol in animals carrying an epidermis-specific deletion for both Dnmt3a and Dnmt3b in combination ( DcKO ) . Strikingly , although DcKO animals had a severe depletion of DNA methylation in their epidermises , they formed a morphologically normal skin with all its appendages and did not develop any epidermal abnormality even up to 70 weeks of age ( Figure 2—figure supplement 4A–B ) . This strongly suggests that de novo DNA methylation is dispensable for the long-term homeostasis of undamaged epidermis . However , when subjected to tumorigenesis , DcKO animals displayed a significantly shortened latency and significant higher tumor burden than wild-type mice ( Figure 2B–F ) . Although these differences were similar to the ones observed in single Dnmt3a-cKO mice ( Figure 2C–D ) , DcKO mice formed aggressive squamous cell carcinomas at a higher frequency as compared to the single cKOs of Dnmt3a or Dnmt3b ( Figure 2E ) . In addition , metastatic nodules in the lungs were observed in 30% of DcKO animals , but in none of the wild-type , Dnmt3a-cKO , or Dnmt3b-cKO animals ( Figure 2F ) . Recent reports show that epidermal squamous cell carcinomas that harbor cells undergoing epithelial to mesenchymal transitions are more metastatic than those that remain predominantly epithelial in nature ( Latil et al . , 2017; da Silva-Diz et al . 2016 ) . Interestingly , DcKO tumors contained large areas with spindle-shaped cells that expressed lower levels of the epithelial markers E-Cadherin and Keratin14 , compared to the wild type and to Dnmt3a-cKO tumors ( Figure 2E , and Figure 2—figure supplement 5 ) . These cells also expressed the mesenchymal marker Vimentin ( Figure 2—figure supplement 6A–C ) . Importantly , these cells that had undergone a mesenchymal transition were still YFP+ , thus deriving from the K14+ origin of the tumor ( Figure 2—figure supplement 6 ) . Taken together , these results indicate that Dnmt3a and Dnmt3b are dispensable for epidermal homeostasis , and that Dnmt3a , but not Dnmt3b , suppresses skin squamous tumor initiation . However , both Dnmt3a and Dnmt3b repress the malignant transformation of epidermal cells into aggressive squamous cell carcinomas . We next wanted to characterize the molecular mechanisms that might underlie the tumor-suppressive function of Dnmt3a in the epidermis . To this end , we isolated by FACS-based cell sorting the basal integrin α6bright tumor cells from four wild-type and eight Dnmt3a-cKO tumors , and performed whole-genome transcriptome profiling by RNA-seq ( Figure 3A ) . It is important to note that our mouse pathologists scored these tumors as squamous cell carcinomas ( SCCs ) , and that ADFP ( Perilipin-2 ) expressing sebaceous adenomas were not included in this transcriptome study ( Figure 3—figure supplement 1A–C ) . 10 . 7554/eLife . 21697 . 013Figure 3 . Deletion of Dnmt3a results in increased tumor heterogeneity , and upregulation of genes related to lipid metabolism . ( A ) Schematic representation of FACS sorting strategy to isolate both RNA and DNA from Itga6pos cells within the tumors . ( B ) Heatmaps representing gene expression ( rlog transformed values ) of the 391 differentially expressed genes between wild type and Dnmt3a-cKO sorted tumor cells . ( C ) Two-dimensional principal-component analysis ( PCA ) of RNA-seq samples from wild-type ( n = 4 ) and Dnmt3a-cKO ( n = 8 ) Itga6bright sorted tumor cells . ( D ) Gene ontology analysis using Genomatix Online Software of the 114 downregulated and 277 upregulated genes in Dnmt3a-cKO tumors , divided by biological processes and over-represented signal transduction pathways . ( E ) Immunofluorescence staining for Krt14 and PPAR-γ of skin tumors from wildtype and Dnmt3a-cKO animals . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01310 . 7554/eLife . 21697 . 014Figure 3—figure supplement 1 . RNA samples submitted for sequencing were obtained from tumors scored predominantly as squamous cell carcinomas in wild-type and Dnmt3a-cKO mice . ( A ) Hematoxylin/eosin staining from the four wild-type tumors used for RNA-seq . ( B ) Hematoxylin/eosin staining from the eight Dnmt3a-cKO tumors analyzed for RNA-seq . In A and B , immunofluorescence staining shown correspond to DAPI , ADFP ( to ensure that no sebaceous adenomas were collected ) , and Krt14 . ( C ) Representative hematoxylin/eosin staining and immunofluorescence of DAPI , ADFP and Krt14 , of sebaceous adenomas eliminated from the RNA-seq study . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01410 . 7554/eLife . 21697 . 015Figure 3—figure supplement 2 . Loss of Dnmt3a results in a reduction of apoptosis in skin tumors . ( A ) Representative images for TUNEL staining to detect apoptotic cells in skin tumors isolated from wild type and Dnmt3a-cKO animals . The right graph shows the quantification of the Tunel staining in wild type ( n = 12 ) and Dnmt3a-cKO ( n = 17 ) tumors . ( B ) Representative images for active Caspase-3 staining to visualize apoptotic cells in skin tumors isolated from wild-type and Dnmt3a-cKO animals . The right panel shows the quantification of the staining in wild-type ( n = 6 ) and Dnmt3a-cKO ( n = 6 ) tumors . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01510 . 7554/eLife . 21697 . 016Figure 3—figure supplement 2—Source Data 1 . Data related to Figure 3—figure supplement 2A–B . Number of apoptotic cells ( expressed in percentage of TUNEL or CASPASE-3 positive cells/DAPI positive cells ) in wild-type and Dnmt3a-cKO tumors . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01610 . 7554/eLife . 21697 . 017Figure 3—figure supplement 3 . DMBA/TPA treatment induces an increase in cellular cell proliferation in Dnmt3a-cKO animals . ( A ) Representative images of KI67 staining in treated or untreated back skin , and in skin tumors , of Dnmt3a-cKO and wild-type littermates . ( B ) Quantification of KI67 staining using the TMarker software , showing the percentages of KI67-positive cells in the different conditions studied and normalized to the proliferation in the interfollicular epidermis of wild-type mice . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01710 . 7554/eLife . 21697 . 018Figure 3—figure supplement 4 . Dnmt3a-KO tumors express high levels of PPAR-γ . ( A ) CPM ( Count per Million Read ) values of the mRNA encoding for PPAR-γ obtained from the RNA-sequencing of the 12 tumors studied . ( B ) Representative immunofluorescence staining for DAPI , PPAR-γ and Krt14 in all the 12 tumors used for the RNA-sequencing experiment . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 01810 . 7554/eLife . 21697 . 019Figure 3—figure supplement 4—Source Data 1 . Data related to Figure 3—figure supplement 4B . Number of proliferative cells ( expressed as the percentage of KI67+/all nuclei ) in the untreated epidermis , DMBA-treated epidermis , and in the tumors , from wild-type and Dnmt3a-cKO mice . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 019 The PCA analysis of the RNA-seq samples showed that the four wild-type tumors clustered together , indicating that overall their transcriptomes were defined by common genes ( Figure 3B , C ) . In contrast , the transcriptomes of Dnmt3a-cKO tumors were substantially more heterogeneous , suggesting that the loss of Dnmt3a could result in the deregulation of numerous different pathways in cancer cells , or that in the context of Dnmt3a loss , different cell of origins ( i . e . basal IFE cells , hair follicle stem cells , or Lrig+ stem cells ) might be more prone to generate more transcriptionally divergent tumors . Nevertheless , 391 genes were consistently differentially expressed between wild-type and Dnmt3a-cKO tumors , of which 114 were downregulated and 277 were upregulated in the latter ( Supplementary file 1 ) . The downregulated genes were mainly associated with apoptosis , suggesting that loss of Dnmt3a promotes cell survival and protects against programmed cell death ( Figure 3D ) ; consequently , TUNEL and active Caspase3 staining confirmed that Dnmt3a-cKO tumors had fewer apoptotic cells as compared to wild-type tumors ( Figure 3—figure supplement 2A–B and Figure 3—figure supplement 2—Source data 1 ) . Dnmt3a-cKO tumors also expressed higher levels of several genes involved in cell proliferation ( Figure 3D ) . Interestingly , proliferation was only significantly increased in pre-cancerous DMBA/TPA-treated Dnmt3a-cKO epidermis ( Figure 3—figure supplement 3A–B ) , while no differences in proliferation were evident between the homeostatic epidermis and tumors of wild-type and Dnmt3a-cKO mice ( Figure 3—figure supplement 3A–B ) . Altogether , this suggests that the loss of Dnmt3a endows pre-cancerous mutant basal cells with a survival and proliferative advantage , which could account for the increased number of tumors these mice develop . However , once tumors are formed , they progress with the same kinetics as wild-type tumors ( Figure 3—figure supplement 3C and Figure 3—figure supplement 4—Source data 1 ) . Gene ontology ( GO ) analysis of the 277 genes that were upregulated in Dnmt3a-cKO basal tumor cells highlighted two principal pathways that were over-represented in all eight Dnmt3a-cKO tumors: Wnt signaling ( including ligands and receptors ) , and more predominantly , lipid metabolism ( Figure 3D ) . Interestingly , recent reports have associated an increase in lipid metabolism with increased tumorigenesis of chronic myeloid leukemia , as well as colorectal , liver , oral , and breast cancer ( Beyaz et al . , 2016; Ma et al . , 2016; Camarda et al . , 2016; Ye et al . , 2016; Corbet et al . , 2016; Schug et al . , 2015; Pascual et al . , 2017; Wahl et al . , 2017; Bensaad et al . , 2014; Luo and Puigserver , 2016 ) . A number of genes associated with fatty acid and lipid metabolism were upregulated in Dnmt3a-cKO tumors ( Figure 3D , Supplementary file 1 ) . Among these , the most upregulated ones encoded the key pro-adipogenic transcription factors PPAR-α and PPAR-γ , which promote adipocyte differentiation and the expression of genes involved in fatty acid metabolism , and which are not expressed in homeostatic epidermal cells ( Fajas et al . , 2001 ) . The role of these transcription factors in cancer is still poorly understood , although they tend to be upregulated in many types of human tumors ( Fajas et al . , 2001 ) . Importantly , PPAR-γ was upregulated at the RNA and protein levels in all the Dnmt3a-cKO sequenced tumors ( Figure 3E and Figure 3—figure supplement 4A–B ) . Interestingly , the expression of PPAR-γ has been extensively reported to be under epigenetic control by repressive mechanisms such as H3K9 methylation and DNA methylation ( Wang et al . , 2013; Zhao et al . , 2013 ) . To further dissect the early molecular changes that might result in the tumor-suppressing role of Dnmt3a in the epidermis , we did a short ( 6-week long ) DMBA/TPA carcinogenesis treatment ( Figure 4A ) . We then FACS-isolated Itga6brightCD34neg cells , consisting mostly of epidermal basal cells ( IFE ) , and hair follicle stem cells ( Bulge; Itga6brightCD34pos ) from pre-cancerous back skin of wild-type or Dnmt3a-cKO animals for RNA-seq analysis ( Supplementary file 2 ) . After this short DMBA/TPA treatment , most of the upregulated genes in epidermal cells ( IFE ) were already predominantly linked to lipid metabolism and cell proliferation , whereas they related mostly to cell proliferation , and Wnt signaling in bulge stem cells ( Figure 4—figure supplement 1A–B ) . Interestingly , we did not observe a diminished expression of genes regulating apoptosis , as we did in tumor cells . Hence , these results suggest that most of the transcriptome changes observed in tumors upon deletion of Dnmt3a occur early , and that the transition from the pre-cancerous epithelium to tumor growth occurs subsequently by bypassing apoptosis . 10 . 7554/eLife . 21697 . 020Figure 4 . Dnmt3a binds a subset of enhancers in tumor cells . ( A ) Schematic representation of a short treatment of DMBA/TPA in wild-type and Dnmt3a-cKO animals . ( B ) Genomic localizations of Dnmt3a determined by ChIP-seq of Dnmt3a in epidermal cells isolated from wild-type animals after 6 weeks of DMBA/TPA treatment . ( C ) Gene ontology analysis of the 363 H3K27ac-enriched regions ( located at least 4 kb away from the TSS ) also bound by Dnmt3a in isolated epidermis from wild-type animals after 6 weeks of DMBA/TPA . ( D ) Screenshot of enhancers bound by Dnmt3a in DMBA/TPA-treated skin in the FOS locus . All tracks are normalized to the number of mapped reads . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 02010 . 7554/eLife . 21697 . 021Figure 4—figure supplement 1 . Deletion of Dnmt3a alters the expression of genes involved in proliferation , lipid metabolism , epidermal differentiation , and Wnt signaling , after 6 weeks of DMBA/TPA treatment . ( A ) Left panel , heatmaps representing gene expression ( rlog transformed values ) of the 498 genes in sorted bulge hair follicle stem cells ( Bulge ) ( Itga6bright/CD34pos ) that were differentially expressed between wild-type ( n = 3 ) and Dnmt3a-cKO ( n = 3 ) . Right panel , gene ontology analysis of the 498 differentially expressed genes up- or downregulated in Dnmt3a-cKO mice as compared to their wild-type littermates . ( B ) Left panel , heatmaps representing gene expression ( rlog transformed values ) of the 188 differential expressed genes between wild type ( n = 4 ) and Dnmt3a-cKO ( n = 4 ) sorted interfollicular epidermal ( IFE ) basal cells ( Itga6bright/CD34neg ) . Right panel , gene ontology analysis of the 188 differentially expressed genes that were up- or downregulated in Dnmt3a-cKO mice as compared to their wild-type littermates . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 021 Dnmt3a is responsible for establishing and maintaining the levels of both 5-mC and 5-hmC around enhancers and promoters ( Colquitt et al . , 2014; Yang et al . , 2016 ) . In addition , Dnmt3a directly methylates the center of its target enhancers resulting in their subsequent hydroxymethylation via Tet2 in human epidermal keratinocytes ( Rinaldi et al . , 2016 ) . To study which targets are regulated directly by Dnmt3a during transformation of murine epidermis , we performed ChIP-Seq for Dnmt3a in DMBA/TPA-treated pre-cancerous back skin epidermises from wild-type or Dnmt3a-cKO animals ( Figure 4A ) . We also compared the ChIP-seq data obtained with MeDIP-seq and hMeDIP-seq performed on FACS-sorted tumor cells . The profiles of MeDIP-seq and hMeDIP-seq around regulatory regions ( transcription start sites ( TSS ) and enhancers ) agreed with published data ( Figure 5—figure supplement 1A ) , and the CG content in our MeDIP-seq/hMeDIP-seq was highly enriched as compared to the input , both of which are measures of good quality data ( Figure 5—figure supplement 1B ) . We detected 16 , 483 genomic locations bound by Dnmt3a in wild-type animals , but only 64 in Dnmt3a-cKO , confirming the specificity of the Dnmt3a antibody ( Figure 4B and Supplementary file 3 ) . Of the bound regions in the wild-type epidermis , more than 20% corresponded to intergenic regions ( Figure 4B ) . ChIP-Seq for H3K27ac using the same samples allowed us to identify 3097 intergenic regions enriched for H3K27ac that corresponded to active enhancers , 10% of which were bound by Dnmt3a in wild-type cells ( Figure 4A–C , Supplementary file 3 ) . Interestingly , proximity-based analysis revealed that the active enhancers bound by Dnmt3a predominantly corresponded to genes essential for keratinocyte differentiation and transcriptional regulation , such as Evpl ( encoding for Envoplakin ) , Ppl ( encoding for Periplakin ) , Fos , Myc , Cebpa , and Fosl2 ( Figure 4C–D ) , similarly to what we have previously reported in human epidermal keratinocytes ( Rinaldi et al . , 2016 ) . The active enhancers bound by Dnmt3a contained higher levels of DNA methylation and hydroxymethylation than those not bound by it ( Figure 5A , C ) . Importantly , loss of Dnmt3a significantly reduced their DNA methylation and hydroxymethylation ( Figure 5A , C ) . Intriguingly , a significant reduction in DNA methylation also occurred in enhancers not bound by Dnmt3a , albeit to a statistically significantly lesser extent than those directly targeted by Dnmt3a in wild-type cells ( Figure 5A , B ) . Upon deletion of Dnmt3a , DNA hydroxymethylation was also significantly reduced in its target enhancers , and to a lesser extent in non-Dnmt3a-bound enhancers ( Figure 5C ) . However , the ratio of 5-hmC levels at enhancers bound by Dnmt3a between wild-type and Dnmt3a-cKO epidermal cells is significantly higher as compared to the ratio of 5-hmC levels between the enhancers that are not normally bound by Dnmt3a ( Figure 5D ) . This indicates that the presence of Dnmt3a correlates with significantly higher 5-hmC levels , likely because Dnmt3a provides 5-mC as a substrate for generating 5-hmC , as we have previously shown in human keratinocytes ( Rinaldi et al . , 2016 ) . 10 . 7554/eLife . 21697 . 022Figure 5 . Depletion of Dnmt3a leads to loss of DNA methylation and hydroxymethylation around its target enhancers . ( A ) Relative methylation score ( CpG count ) measured around 363 enhancers bound by Dnmt3a ( –5 kb , +5 kb ) from independent biological replicates of FACS sorted tumor cells from wild type ( n = 2 ) and Dnmt3a-cKO ( n = 2 ) ( p<2 . 2 × 10−16 ) . ( B ) Relative methylation score ( CpG count ) measured around 2734 enhancers not bound by Dnmt3a ( –5 kb , +5 kb ) from independent biological replicates of FACS-sorted tumor cells from wild-type ( n = 2 ) and Dnmt3a-cKO ( n = 2 ) animals ( p=2 . 374e−5 ) . ( C ) Global levels of 5-hmC at enhancer center ( –2Kb , + 2 Kb ) were quantified using HOMER software in independent biological replicates of FACS sorted tumor cells from wild-type ( n = 2 ) and Dnmt3a-cKO ( n = 2 ) mice at enhancers bound or not by Dnmt3a . ( D ) Ratio between the 5-hmC levels at enhancers bound or not by Dnmt3a in wild-type and Dnmt3a-cKO tumor cells . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 02210 . 7554/eLife . 21697 . 023Figure 5—figure supplement 1 . MeDIP-seq and hMeDIP-seq analysis from sorted tumor cells . ( A ) CpG count reads versus theoretical distribution in MeDIP and hMeDIP samples from wild type and Dnmt3a-cKO tumors . ( B ) MeDIP-seq signals around active and non-active TSSs in wild-type Itga6bright tumor cells . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 023 In addition to active enhancers , a significant proportion ( 19% ) of the enriched regions for Dnmt3a corresponded to promoters/TSSs ( Figure 4B and Supplement file 3 ) . To understand if Dnmt3a was methylating these promoters , we overlaid the Dnmt3a ChIP-seq with the MeDIP-seq data . Notably , the promoters bound by Dnmt3a showed a strong and statistically significant loss of DNA methylation around the corresponding TSS ( Figure 6A ) . The levels of DNA methylation were not significantly changed at promoters not bound by Dnmt3a ( Figure 6B ) . Of note , Dnmt3a-target TSSs were not enriched for 5-hMC ( not shown ) . The loss of DNA methylation at the promoters/TSSs bound by Dnmt3a was also accompanied by a general increase in the transcription of these genes , measured by RNA-seq in the tumors ( Figure 6C ) . Altogether , these data suggest that Dnmt3a directly represses the expression of a specific subset of genes by methylating their promoters/TSSs . 10 . 7554/eLife . 21697 . 024Figure 6 . Dnmt3a binds and methylates a subset of promoters of genes involved in lipid metabolism in DMBA/TPA-treated epidermal cells . ( A ) Relative methylation score ( CpG count ) measured around active and silenced promoters bound by Dnmt3a ( –5 kb , +5 kb ) from independent biological replicates of FACS-sorted tumor cells from wild type ( n = 2 ) and Dnmt3a-cKO ( n = 2 ) animals . ( B ) Relative methylation score ( CpG count ) measured around promoters not bound by Dnmt3a ( –5 kb , +5 kb ) from independent biological replicates of FACS-sorted tumor cells from wild-type ( n = 2 ) and Dnmt3a-cKO ( n = 2 ) animals ( p=0 . 104 ) . ( C ) CPM ( Counts por Million ) values of genes bound at the TSS by Dnmt3a in DMBA skin tumors from wild-type or Dnmt3a-cKO animals . ( D ) Gene ontology analysis , using Enrichr online software , of the 3521 genes bound at their promoter by Dnmt3a . ( E ) Screenshot of PPAR-γ gene , with all tracks normalized . ( F ) Normalized methylation score measured around TSS of Ppar-γ ( –1 kb to +1 kb ) bound by Dnmt3a . ( G ) CPM ( Counts por Million ) values of PPAR-γ measured by RNA-seq in sorted Itga6bright cells from DMBA/TPA-treated IFE and from DMBA skin tumors in wild-type and Dnmt3a-cKO mice . ( H ) Immunofluorescence staining for Krt14 and PPAR-γ of DMBA/TPA-treated skin and skin tumors from wild-type and Dnmt3a-cKO animals . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 02410 . 7554/eLife . 21697 . 025Figure 6—source data 1 . Data related to Figure 6G . RNA-sequencing CPM values ( Counts por Million Reads ) of Pparg expression in the DMBA/TPA epidermis and in the tumors from wild-type and Dnmt3a-cKO mice . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 025 Interestingly , a GO analysis of the promoters bound by Dnmt3a indicated that they regulated the expression of genes predominantly involved in cell proliferation and lipid metabolism , consistent with our RNA-seq results ( Figure 6D and Supplement file 3 ) . Interestingly , Dnmt3a bound the promoters of Ppar-α and Ppar-γ in wild-type but not Dnmt3a-cKO epidermis ( Figure 6E and Supplement file 3 ) . Furthermore , 5-mC levels were lower in the TSS of the PPAR-γ gene in Dnmt3a-cKO as compared to wild-type tumors , indicating a DNA methylation-dependent mechanism of transcriptional repression ( Figure 6F ) . Consistent with a transcriptional derepression of the locus following loss of DNA methylation , PPAR-γ mRNA and protein levels were upregulated both in pre-cancerous interfollicular epidermis and in tumors lacking Dnmt3a , suggesting that the upregulation of PPAR-γ is acquired at the pre-cancerous stage , even before overt tumors appear ( Figure 6G , H and Figure 6—source data 1 ) . We next tested whether the increase in the expression of genes involved in lipid metabolism was required for the earlier onset of tumorigenesis and increased tumor burden in Dnmt3a-cKO mice . To this end , wild-type and Dnmt3a-cKO mice were subjected to the DMBA/TPA skin carcinogenesis protocol , but were separated into two cohorts , one treated topically with a PPAR-γ chemical inhibitor in combination with DMBA/TPA , and the other with the vehicle ( Figure 7A-diagram ) ( Sahu et al . , 2012; Grabacka et al . , 2006 ) . Interestingly , inhibition of PPAR-γ significantly delayed the onset of tumor appearance in Dnmt3a-cKO mice , and reduced the number of tumors developed by the Dnmt3a-cKO ( Figure 7B–C and Figure 7—source data 1 ) . However , the average size of the tumors was not affected by the inhibition of PPAR-γ ( Figure 7D ) . Thus , inhibition of PPAR-γ could be a potential new therapy for cutaneous squamous cell carcinomas harboring low levels of Dnmt3a . 10 . 7554/eLife . 21697 . 026Figure 7 . PPAR-γ inhibition revert the tumor initiation phenotype of the Dnmt3a-cKO . ( A ) Schematic representation of the DMBA/TPA orthotopic treatment together PPAR-γ inhibitor ( Sigma GW9662 ) treatment onto wild-type and Dnmt3a-cKO animals . ( B ) Time of appearance , expressed in percentages of skin tumors on wild-type or Dnmt3a-cKO animals ( vehicle and GW9662 treated ) : p=0 . 008 , Chi-Square Test . ( C ) Number of skin tumors after 3 months of DMBA/TPA treatment plus GW9662 treatment , p=0 . 007 ( Unpaired T-Test ) . ( D ) Tumors sizes expressed in millimeters ( mm ) after 3 months of DMBA/TPA plus GW9662 treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 02610 . 7554/eLife . 21697 . 027Figure 7—source data 1 . Data related to Figure 7C . Data showing the number of tumors counted on the backskin of the wild-type and Dnmt3a-cKO mice after 120 days of DMBA/TPA plus vehicle or plus GW9662 ( PPAR-G inhibitor ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 02710 . 7554/eLife . 21697 . 028Figure 7—figure supplement 1 . The mRNA of Dnmt3a is downregulated in human cutaneous squamous cell carcinomas compared to normal human epidermis . MRNA expression of Dnmt3a in human healthy epidermis compared to actinic keratoses and Squamous Cell Carcinomas ( SCC ) quantified using GEO2R platform of the published databases ( GSE2503 , GSE42677 , GSE45164 , and GSE53462 ) . Unpaired parametric T-Test was used for statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 02810 . 7554/eLife . 21697 . 029Figure 7—figure supplement 1—source data 1 . Data related to Figure 7—figure supplement 1 . Dnmt3a mRNA expression obtained using the online platform GEO2R from four published datasets of human healthy epidermis , actinic keratosis and squamous cell carcinomas samples . DOI: http://dx . doi . org/10 . 7554/eLife . 21697 . 029 At last , using public available data from four different studies we determined that the expression of Dnmt3a is significantly reduced in squamous cell carcinomas and actinic keratosis , the premalignant stage of squamous tumors , compared to human healthy epidermis ( Figure 7—figure supplement 1 and Figure 7—figure supplement 1—Source data 1 ) .
Dnmt3a modifies cytosine at CpG dinucleotides and is responsible for the proper differentiation of murine hematopoietic stem cells and murine neural stem cells ( Challen et al . , 2011; Mayle et al . , 2015; Shlush et al . , 2014 ) . Recently , we and others have shown in human epidermal keratinocytes and murine olfactory sensory neurons , respectively , that Dnmt3a regulates gene expression by cooperating with Tet to maintain high levels of 5-hmC at enhancers ( Colquitt et al . , 2014; Rinaldi et al . , 2016 ) . Importantly , Dnmt3a is not only frequently mutated in human tumors ( Kim et al . , 2013 ) but is perhaps one of the first mutations to occur during tumorigenesis ( Shlush et al . , 2014 ) . Using knockout mouse models , we now have demonstrated that Dnmt3a is also tumor-suppressive toward carcinogen-induced epidermal squamous neoplasia . Its loss not only accelerated the onset of tumors , but also increased tumor burden . However , once formed , Dnmt3a-deficient tumors grew , and progressed to carcinomas with the same kinetics and proportions , respectively , as their wild-type counterparts . Recent works have shown that the absence of Dnmt3a or Tet2 in hematopoietic stem cells predisposes to leukemia formation ( Yang et al . , 2016; Rasmussen et al . , 2015 ) , but that restoring the expression of Dnmt3a after the leukemia had developed did not revert the phenotype ( Yang et al . , 2016 ) . That said , the role of Dnmt3a in tumorigenesis is tissue specific , since in the lung it does not affect tumor initiation but rather tumor progression ( Gao et al . , 2011 ) . Interestingly , in the work of Gao et al . , most of the changes in gene expression in Dnmt3a-depleted cells were attributed to alterations in gene body methylation , rather than at promoters . Conversely , in our model , we see significant changes at regulatory elements ( i . e . promoters and enhancers ) that lead to changes in gene expression in Dnmt3a-depleted epidermal tumors . Our results , together with accumulating evidence from other groups , demonstrates a clear relationship between the levels of Dnmt3a–Tet–5-hmC and tumorigenesis: the inactivation of this axis in adult stem cells predisposes them to tumor initiation . Interestingly , a global reduction of 5-hmC is a hallmark of several cancer types , including squamous cell carcinoma , and is often correlated with poor prognosis ( Zhang et al . , 2016; Liao et al . , 2016; Shi et al . , 2016; Ficz and Gribben , 2014; Lian et al . , 2012 ) . Our results indicate that Dnmt3a drives the DNA-methylation , and subsequent hydroxymethylation , of a subset of enhancers that regulate the expression of genes involved differentiation . Conversely , Dnmt3a binds to , and DNA methylates , the promoters of cell proliferation and lipid metabolism genes to repress their expression . Interestingly , however , deletion of Dnmt3a does not result in changes in the specification of the different keratinocyte lineages in the skin ( epidermis , hair follicles , and sebaceous glands ) , nor their homeostasis in adulthood . What is more , even the combined deletion of Dnmt3a and Dnmt3b , albeit significantly reducing overall DNA methylation levels , did not result in any skin phenotype even in aged mice . On the other hand , we have recently shown that Dnmt3a and Dnmt3b are necessary for the self-renewal and differentiation of primary human keratinocytes ( Rinaldi et al . , 2016 ) . This apparent contradiction in the phenotypes observed might be due to the fact that our work with human keratinocytes relied on culturing the cells , which was recently shown in murine skin keratinocytes to induce a wound healing damaged-like reversible state that affects their epigenome and transcriptome ( Adam et al . , 2015 ) . Thus , in vivo deletion of Dnmt3a might not be sufficient to alter the homeostasis of undamaged skin , but renders the epidermis more susceptible to situations of damage . Accordingly , the epidermis of Dnmt3a-cKO mice responded to the treatment of DMBA/TPA in a much more pronounced manner . This effect might not be specific to Dnmt3a . For instance , Dnmt1 is the DNA methyltransferase most highly expressed in epidermal cells and is responsible for about 70% of DNA methylation levels ( Li et al . , 1992 ) . Its depletion causes a strong loss of self-renewal of primary human basal keratinocytes ( Sen et al . , 2010 ) . However , its loss in murine epidermis leads to a mild increase in proliferation , and to a partial alopecia , only in very aged mice ( Li et al . , 2012 ) . Intriguingly , these results also suggest that Dnmt1 and Dnmt3a/3b exert different functions in epidermal homeostasis , although future work will be required to study these putative differences in depth . Notwithstanding the differences between in vivo and ex vivo studies , our results show that the genomic localization of Dnmt3a is very similar between intact murine keratinocytes and cultured human keratinocytes . Besides its localization at active enhancers of genes involved in epidermal differentiation , Dnmt3a also bound to , and methylated , promoters of genes that regulate cell proliferation and lipid metabolism to repress their expression . Among these genes , were the master regulators of lipid metabolism and adipogenesis PPAR-α and PPAR-γ . Interestingly , a number of recent studies have highlighted the importance of a persistent lipid metabolism in promoting tumor transformation , and tumor metastasis in colorectal , liver , breast and oral squamous carcinomas , as well as for enhancing chemoresistance of leukemia stem cells ( Pascual et al . , 2017; Ma et al . , 2016; Beyaz et al . , 2016; Ye et al . , 2016 ) . The upregulation of these transcription factors upon deregulation of Dnmt3a might predispose the epidermis to develop more tumors , suggesting that an intriguing mechanistic link between lipid metabolism and the epigenetic regulation of tissue homeostasis through DNA methylation , might exist . A recent large clinical association study has already pointed to this by establishing a correlation between the expression of obesity-related genes and changes in the content of DNA methylation ( Wahl et al . , 2017 ) . Importantly , our results show that PPAR-γ is partially responsible for promoting tumorigenesis in Dnmt3a-deficient epidermis , which considering that human skin tumors express lower levels of Dnmt3a , might provide us with a new therapeutic antitumor avenue against squamous cell carcinomas .
The work with mice was approved by the Ethical Committee for Animal Experimentation ( CEEA ) of the Scientific Park of Barcelona ( PCB ) , and the Government of Catalunya . Inbred male or female Dnmt3a flox/flox ( C57/Bl6 ) backcrossed to Krt14-CRE-YFP ( C57/Bl6 ) for six to nine generations were used for all animal experiments . Chemically-induced skin carcinogenesis was performed as previously described ( Nassar et al . , 2015; Abel et al . , 2009 ) , with a slight modification to yield high-frequency SCCs in the C57/Bl6 genetic background . Briefly , the back skin of 8-week-old mice—at which time hair follicles are in their resting phase ( telogen ) —was shaved and treated with the mutagen 7 , 12-dimethylbenz[a]anthracene ( DMBA; 200 µl of 0 . 25 mg/ml solution in acetone ) and the pro-inflammatory and pro-proliferation agent 12-O-tetradecanoyl phorbol-13-acetate ( TPA; 200 µl of 0 . 02 mg/ml solution in acetone ) once weekly for 6 weeks . Specifically , DMBA was given on Monday and TPA always on the Friday of the same week . For short DMBA experiments , animals were sacrificed and back skins were processed 3 days after the sixth TPA application . For tumor formation studies , treatment continued twice weekly with TPA ( 200 µl of 20 µg/ml solution in acetone ) for up to 20 weeks , or until the largest tumor of each mouse reached 1 . 5 mm diameter , at which point animals were sacrificed . In total , 12 wild type and 15 Dnmt3a-cKO tumors from 6 and 8 mice , respectively , were included for tumor analyses . The chemically-induced skin carcinogenesis was performed as previously described above ( Nassar et al . , 2015; Abel et al . , 2009 ) . We used the chemical 2-Chloro-5-nitro-N-phenylbenzamide ( Sigma Aldrich: GW9662 ) described as potent PPAR- γ inhibitor . Briefly , the shaved dorsal epidermis of wild-type and Dnmt3a-cKO mice was treated twice a week topically with 200 µl of 100 nmoles of GW9662 solubilized in acetone . We applied the PPAR-γ inhibitor together with the first DMBA treatment , and subsequently administered it at every DMBA or TPA treatment . The PPAR-γ inhibitor was applied always 2 min before every administration of DMBA or TPA ( Sahu et al . , 2012; Grabacka et al . , 2006 ) . To isolate pre-cancerous epidermal cells following short DMBA/TPA treatment , back skins were dissected and processed to single-cell suspensions as previously described ( Jensen et al . , 2010 ) . To purify tumor cells , DMBA/TPA-induced SCCs were mechanically dissociated using a McIlwain Tissue Chopper ( The Mickle Laboratory Engineering Co . LTD ) . Minced tumor tissue was digested under agitation in serum-free EMEM medium without calcium containing 2 . 5 mg/ml Collagenase I ( Sigma Aldrich , St . Louis , Missouri ) , and 0 . 75 mg/ml trypsin ( Life Technologies ) for 90 min at 37°C . Cells were pelleted , suspended in 1–2 ml of 0 . 25% pre-warmed trypsin/EDTA ( Life Technologies ) containing 100 µg/ml ( Aldrich ) per tumor , and incubated at 37°C for 2 min . Trypsin was inactivated by adding EMEM without calcium containing 10% chelated FBS . Cells were washed twice in PBS and filtered sequentially through 100 µm and 40 µm cell strainers . For ChIP-seq , single-cell suspensions were cross-linked for 10 min at room temperature with 1% formaldehyde ( methanol-free; Thermofisher , 28906 ) and quenched for 5 min to a final concentration of 0 . 125M of glycine . Cells were washed 2× with cold PBS and frozen at –80°C . For flow cytometry analysis , epidermal or tumor cells were re-suspended at 1 × 107 cells/ml in PBS and labeled with CD49f-PE ( clone NKI-GoH3 , 1:200 , AbD Serotec ) and CD34-biotin ( clone RAM34 , 1:50 , eBioscience ) followed by streptavidin-APC ( 1:400 , BD Biosciences ) . Tumor cell suspensions were additionally labeled with lineage-BV605 ( CD31 , clone 390; CD45 , clone 30-F11; TER119 , clone TER119; all 1:100 ) ( Biolegend ) to exclude stromal cell contamination . Both epidermal and tumor cells were positive for YFP due to the presence of the Rosa26-YFP allele in the mice . Tumor cells ( YFPbright/lineageneg cells ) , pre-cancerous epidermis of interfollicular epidermis ( YFPpos/CD49fhigh/CD34neg cells ) , and bulge hair follicle stem cells ( YFPbright/CD49fhigh/CD34pos cells ) were FACS-sorted using a BD FACSAria Fusion flow cytometer ( BD Biosciences ) . Approximately , 3–20 × 104 cells were sorted and lysed in 1 ml of TRIzol for RNA and DNA isolation . After adding 200 µl chloroform , samples were vortexed for 30 s and then centrifuged at 12 , 000 g to separate the RNA-containing supernatant from the organic phase . RNA was precipitated with 1× volume of isopropanol , washed twice with 70% ethanol , and then used for library preparation . The interphase of the TRIzol solution ( after removal of the supernatant ) was precipitated adding 1× volume of isopropanol , centrifuged for 1 hr at 4°C at 13 , 000 g , washed twice with ethanol , and digested overnight at 55°C with proteinase K ( 10 mg/ml ) in TE 1× buffer . The following day , digested material was incubated 1 hr at 37°C with RNase A and purified using a conventional phenol/chloroform separation . The DNA pellet was quantified , and DNA was used for library preparation for MeDIP-seq and hMeDIP-seq experiments . Purified genomic DNA ( 250 ng ) from tumor cells was sonicated to obtain fragments of 300–700 bp . Adaptors from the NEBNext Ultra DNA Library Prep Kit for Illumina were added to the fragmented DNA . DNA was denatured for 10 min at 99°C and cooled to avoid re-annealing . Fragmented DNA was incubated overnight with 1 µg of antibodies ( 5-methylcytosine , Abcam cat . # ab10805; 5-hydroxymethylcytosine , Active Motif , cat . # 39769 , RRID: AB_10013602 ) previously cross-linked with 15 µl of Dynabeads Protein A ( Life Technologies ) . Immunocomplexes were recovered using 8 µl for 2 hr . The following morning , DNA was washed three times for 10 min each , and purified DNA was extracted using QIAquick MinElute ( Qiagen ) . Amplified libraries were prepared using NEBNext Ultra DNA Library Prep Kit for Illumina ( E7370L ) following the manufacturer's instructions . The libraries of total RNA from wild type and Dnmt3a-cKO tumors was prepared using the TruSeqStranded Total Sample Preparation kit ( Illumina Inc . ) according to the manufacturer’s protocol . Each library was sequenced using TruSeq SBS Kit v3-HS , in paired end-mode with the read length 2 × 76 bp . A minimal of 137 million paired-end reads was generated for each sample run in one sequencing lane on HiSeq2000 ( Illumina , Inc ) following the manufacturer’s protocol . Images analysis , base calling , and quality scoring of the run were processed using the manufacturer’s software Real-Time Analysis ( RTA 1 . 13 . 48 ) and followed by generation of FASTQ sequence files by CASAVA . RNA-seq datasets were pre-processed by removing both low-quality bases from the 3′- ends of the reads and adapter sequences using Trimmomatic ( version 0 . 33 ) ( Bolger et al . , 2014 ) . The trimmed reads were aligned to the mouse genome ( UCSC mm10 ) using TopHat ( version 2 . 0 . 13 ) ( Trapnell et al . , 2009 ) , with default parameters and –g 5 . Gene and transcript expression levels were quantified with HTSeq ( version 0 . 6 . 1p1 ) ( Anders et al . , 2015 ) . From the raw counts , counts per million ( cpm ) and fragments per kilobase of transcript per million mapped reads ( fpkm ) values were calculated . Differential expression analysis was performed using DESeq2 ( Love et al . , 2014 ) using a q-value cutoff of 0 . 05 and a fold-change cutoff of 1 . 5 to identify differentially expressed genes . ChIP was performed as previously described ( Morey et al . , 2012 ) . Briefly , frozen pelleted were lysed in 1 ml ChIP buffer ( 150 mM NaCl , 10 mM Tris-HCl , 5 mM EDTA , 1% SDS , 0 . 5 mM DTT , and 1% Triton X-100 ) and sonicated for 30 min in a Bioruptor Pico ( Diagenode ) . DNA fragments were de-crosslinked overnight at 65°C and checked with a bioanalyzer . After a DNA check , chromatin was diluted 1:5 with ChIP buffer with no SDS ( 150 mM NaCl , 10 mM Tris-HCl , 5 mM EDTA , 0 . 5 mM DTT , and 1% Triton X-100 ) . Immunoprecipitation experiments for transcription factors used 30 µg of chromatin , and those for H3K27ac , 3 µg of chromatin . Antibodies ( 10 µg for Dnmt3a and 3 µg for H3K27ac ) were incubated overnight with the chromatin in ChIP buffer . Immunocomplexes were recovered with 40 µl of protein A bead slurry ( Healthcare , cat . # 17-5280-01 ) . Immunoprecipitated material was washed three times with low-salt buffer ( 50 mM HEPES pH 7 . 5 , 140 mM NaCl , 1% Triton ) and 1× with high-salt buffer ( 50 mM HEPES pH 7 . 5 , 500 mM NaCl , 1% Triton ) . DNA complexes were de-crosslinked at 65°C overnight , and DNA was then eluted in 50 µl of water using the PCR purification kit ( QIAGEN ) . Antibodies used for ChIP were Dnmt3a ( SantaCruz H-295; RRID: AB_2093990 ) and H3K27ac ( Merck Millipore , cat . # 07–360; RRID: AB_310550 ) . Libraries for sequencing were prepared using NEBNext Ultra DNA Library Prep Kit from Illumina ( E7370L ) following the manufacturer's instructions . ChIP-seq datasets were aligned to the mouse genome build mm10 using BowTie ( version 1 . 0 . 1 ) ( Langmead et al . , 2009 ) ; the parameters used were –k 1 , –m 1 , and –n 2 . UCSC browser tracks ( Kent et al . , 2002 ) were created from the mapped bam file after converting it to bedGraph ( normalized to 10 million reads ) and subsequently bigWig format . Peak calling of Dnmt3A to determine regions of ChIP-seq enrichment over the background was done with the MACS version 1 . 4 . 1 . Peaks of the methylation and hydroxymethylation datasets were determined similarly . For histone marks , MACS version two was used with parameters –broad , -q 0 . 01 , and –g mm . ChIP-seq peaks were annotated using the annotatePeaks . pl script of the HOMER suite ( version 4 . 6 ) ( Heinz et al . , 2010 ) using the UCSC mm10 annotation . The coverage depths of different ChIP-seq experiments at specified regions were also calculated using the annotatePeaks . pl script . This generated a normalized coverage value of different sequencing experiments at equally spaced bins spanning the region of interest . Bin size was set to 1 bp . For the differential regulation analysis of MeDIP-seq data with replicates , common peaks were first determined among the replicates of the wild type and KO samples separately . A consensus peakset was then created from the two common peaksets , and the read counts were calculated for all the peaks of the consensus peakset . DESeq2 ( Love et al . , 2014 ) was applied to calculate the differentially bound peaks using an adjusted p value of < 0 . 05 . Skin and tumors were isolated from mice , fixed in formalin 10% for 2 hr at room temperature , and embed in paraffin . Sections were cut and stained on glass coverslips . After deparaffinization sections were permeabilized with 0 . 5% Triton/PBS for 10 min , blocked with 10% goat serum , and stained overnight at 4°C with primary antibodies diluted in 1% goat serum . The morning after , sections were washed three times with PBS 1X with 10 min for each wash and stained with secondary antibody ( 1/1000 ) . Nuclei were counterstained with DAPI ( Sigma , D9542 ) . Primary antibodies were anti-Dnmt3a ( 1:100 , SantaCruz H-295: RRID: AB_2093990 ) , anti-PPAR-γ ( 1:100 Santa Cruz , sc-7196: RRID: AB_654710 ) , and anti-keratin 14 ( 1:500 , Biolegend SIG-3476: RRID:AB_10718041 ) ; secondary antibodies were anti-rabbit Alexa Fluor 488 ( RRID: AB_141708 ) and anti-mouse Alexa Fluor 647 ( 1:500 , Molecular Probes; RRID: AB_162542 ) . For immunofluorescence staining anti-5-Methylcytosine ( 1:100 , Abcam10805 , clone 33D3: RRID: AB_442823 ) , sections ( after deparaffinization and before Triton incubation ) were incubated for 15 min with 2N HCL to further denature DNA . Adipophilin ( ADFP , ab37516 , 1:100 dilution; RRID: AB_722641 ) . Tunel Staining was performed using the Promega DeadEnd Tunel System following the manufacturer's instructions . Pictures were acquired using a Leica TCS SP5 confocal microscope . To compare tumor burden between genotypes , we used a T-Test with 95% confidence . To compare free tumor survival differences and anagen entry differences , we used a Chi-Square test . To compare Relative Methylation Score ( RMS ) levels and to compare normalized 5-hmC levels between wild type and Dnmt3a-cKO sorted tumor cells we used a paired Wilcoxon Test . The same paired-Wilcoxon test was used to measure differences in RNA expression . Skin and tumor sections were stained after deparaffinization with KI67 ( Abcam ab15580; RRID: AB_443209 ) for 60 min . After two washes , section were incubated with Power Vision Rabbit ( InmunoLogic ) for 45 min . Positive staining was revealed using a chromogen DAB for 5 min ( Dako ) . Counterstain for hematoxylin was incubated for 3 min ( Dako ) . Stained sections were scanned using a high-resolution NanoZoomer 2 . 0 HT ( Hamamatsu ) . KI67-positive nuclei in the interfollicular epidermis were measured using the TMarker software ( Schüffler et al . , 2013 ) . Positive and negative nuclei for the staining were trained using the color deconvolution plugin and quantified using the cancer nucleus classification plugin . The total number of positive nuclei was normalized to the total number of nuclei in the area considered . Unpaired parametric T-test was used to measure statistical difference among groups and genotypes . | Most of the cells in our body contain the same DNA . However , our bodies are made of many different types of cell , such as nerve cells or skin cells , which perform very different jobs . In each cell type only certain sets of genes encoded by the DNA are active . Proteins known as epigenetic regulators are responsible for producing the different patterns of gene activity . If epigenetic regulators are switched on or off at the wrong time , they can contribute to ageing and diseases such as cancer . Enzymes known as DNA methyltransferases are one group of epigenetic regulators . DNA methyltransferases control the activity of genes by adding small chemical groups known as methyl groups to the DNA . Two of these enzymes – known as Dnmt3a and Dnmt3b – are important during development to help cells mature and specialize into different types . Mice that lack both of these enzymes either die as embryos or just after birth . Furthermore , these enzymes are mutated or less active in some skin cancers and various other human cancers . Here , Rinaldi et al . investigated the role these enzymes play in adult mice . The experiments show that under ordinary laboratory conditions , mutant mice that lacked Dnmt3a and Dnmt3b were as healthy as normal mice . However , when the mice were exposed to chemicals that promote tumor growth , which mimics skin exposure to UV light , the mutant mice developed many more skin tumors than the normal mice . Furthermore , the tumors in the mutant mice were more likely to form secondary tumors in the lung . Rinaldi et al . found that Dnmt3a reduced the production of a protein called PPAR-γ , which helps to break down some types of fat molecules . Treating the mutant mice with a drug that inhibits PPAR-γ activity slowed the growth of the tumors . Overall , these experiments show a new way in which DNA methyltransferases act in adult animals . Future research will investigate whether drugs that inhibit the breakdown of fats could help to treat cancers in which the Dnmt3a and Dnmt3b proteins are mutated or less active . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cancer",
"biology"
] | 2017 | Loss of Dnmt3a and Dnmt3b does not affect epidermal homeostasis but promotes squamous transformation through PPAR-γ |
Meiotic drivers are genetic elements that break Mendel’s law of segregation to be transmitted into more than half of the offspring produced by a heterozygote . The success of a driver relies on outcrossing ( mating between individuals from distinct lineages ) because drivers gain their advantage in heterozygotes . It is , therefore , curious that Schizosaccharomyces pombe , a species reported to rarely outcross , harbors many meiotic drivers . To address this paradox , we measured mating phenotypes in S . pombe natural isolates . We found that the propensity for cells from distinct clonal lineages to mate varies between natural isolates and can be affected both by cell density and by the available sexual partners . Additionally , we found that the observed levels of preferential mating between cells from the same clonal lineage can slow , but not prevent , the spread of a wtf meiotic driver in the absence of additional fitness costs linked to the driver . These analyses reveal parameters critical to understanding the evolution of S . pombe and help explain the success of meiotic drivers in this species .
Mating behaviors have long been a focus of art , literature , and formal scientific inquiry . This interest stems , in part , from the remarkable importance of mate choice on the evolution of species . Outcrossing and inbreeding represent distinct mating strategies that both have potential evolutionary benefits and costs ( Glémin et al . , 2019; Muller , 1932; Otto and Lenormand , 2002 ) . For example , outcrossing ( mating between individuals from distinct lineages ) can be beneficial because it can help purge deleterious alleles from a line of descent , but it can also be costly as it can promote the spread of selfish genes ( Burt and Trivers , 1998; Crow , 1988; Hurst and Werren , 2001; McDonald et al . , 2016; Zeyl et al . , 1996 ) . Meiotic drivers represent one type of selfish genetic element that relies on outcrossing to persist and spread in a population ( Lindholm et al . , 2016; Sandler and Novitski , 1957 ) . These loci can manipulate the process of gametogenesis to bias their own transmission into gametes at the expense of the rest of the genome ( Burt and Trivers , 2006 ) . Meiotic drivers are often considered selfish or parasitic genes because they generally offer no fitness benefits to their hosts and are instead often deleterious or linked to deleterious alleles ( Dyer et al . , 2007; Higgins et al . , 2018; Klein et al . , 1984; Rick , 1966; Schimenti et al . , 2005; Taylor et al . , 1999; Wilkinson and Fry , 2001 ) . As inbreeding is thought to inhibit the spread of selfish genes like drivers , drivers are predicted to be unsuccessful in species that rarely outcross ( Burt and Trivers , 1998; Hurst and Werren , 2001 ) . This assumption appears to be challenged in several fungal species , including the fission yeast Schizosaccharomyces pombe ( Grognet et al . , 2014; Hammond et al . , 2012; Svedberg et al . , 2021; van der Gaag et al . , 2000; Vogan et al . , 2021 ) . Generally , S . pombe cells grow as haploids and their mating ( fusion of two haploids of opposite mating type ) generates diploid zygotes that can then undergo meiosis to generate haploid progeny known as spores ( Figure 1—figure supplement 1A ) . Mating in S . pombe is widely thought to preferentially occur between daughter cells clonally derived from a common progenitor via a recent mitotic division , which we refer to as ‘same-clone mating’ ( Figure 1—figure supplement 1B; Egel , 1977; Gutz and Doe , 1975; Miyata and Miyata , 1981; Billiard et al . , 2012; Perrin , 2011 ) . The support for this idea is described below . Despite this notion that S . pombe cells preferentially undergo same-clone mating ( leading to inbreeding ) , S . pombe isolates host multiple meiotic drivers ( Bravo Núñez et al . , 2020b; Eickbush et al . , 2019; Farlow et al . , 2015; Hu et al . , 2017; Nuckolls et al . , 2017; Zanders et al . , 2014 ) . In the wild , a minority of S . pombe strains are heterothallic , meaning they have a fixed mating type ( h+ or h-; Gutz and Doe , 1975; Nieuwenhuis and Immler , 2016; Schlake and Gutz , 1993 ) . Heterothallic S . pombe strains must mate with non-clonally related cells of the opposite mating type to complete sexual reproduction ( Egel , 1977; Gutz and Doe , 1975; Leupold , 1949; Miyata and Miyata , 1981; Nieuwenhuis and Immler , 2016; Osterwalder , 1924; Schlake and Gutz , 1993 ) . However , most isolates of S . pombe are homothallic , that is , they can switch between the two mating types ( h+ and h- ) during clonal expansion via mitosis ( Figure 1—figure supplement 1B; Egel and Eie , 1987; Gutz and Doe , 1975; Klar , 1990; Nieuwenhuis et al . , 2018; Singh and Klar , 2003 ) . Homothallism in S . pombe enables mating to occur between cells clonally derived from the same progenitor via mitosis ( Billiard et al . , 2012 ) . Mating can even occur between two cells produced by a single mitotic cell division ( Egel , 1977; Gutz and Doe , 1975; Miyata and Miyata , 1981 ) . It is often assumed that homothallic fungi , like S . pombe , do not generally mate with non-clonally related cells . It is important to note , however , that homothallism in fungi does not inherently lead to mating between clonally related cells ( Giraud et al . , 2008 ) . Under some conditions , such as when gametes are dispersed and finding a mate is costly , homothallism can be a beneficial strategy for outcrossing species as it ensures gamete compatibility ( Billiard et al . , 2012 ) . In addition , population genetic analyses of one homothallic fungus , Sclerotinia sclerotiorum , support that homothallism is compatible with frequent outcrossing ( Attanayake et al . , 2014 ) . Very little is known about the ecology of S . pombe in the wild , including how often S . pombe cells outcross ( Jeffares , 2018 ) . In the lab , the mating propensities of S . pombe have mostly been investigated in derivatives of a strain first isolated from French wine in 1921 ( Osterwalder , 1924 ) . In this work , we will refer to derivatives of this isolate as ‘Sp’ . Distinct homothallic Sp strains can be induced to mate in the lab ( i . e . , outcrossed ) , but such crosses are significantly more challenging to execute than crosses between heterothallic isolates . The relative difficulty associated with outcrossing homothallic strains has been attributed to preferential same-clone mating ( Ekwall and Thon , 2017; Forsburg and Rhind , 2006 ) . Microscopy experiments have supported the idea that homothallic Sp haploids tend to undergo same-clone mating ( Bendezú and Martin , 2013; Egel , 1977; Leupold , 1949; Miyata and Miyata , 1981 ) . However , the precise level of same-clone mating when homothallic S . pombe cells are among nonclonal sexual partners has not , to our knowledge , been formally reported for any isolate . Population genetic analyses have provided additional support for the notion that S . pombe inbreeds . There are two major S . pombe lineages that diverged between 2300 and 78 , 000 years ago ( Tao et al . , 2019; Tusso et al . , 2019 ) . Much of the sampled variation within the species represents different admixed hybrids of those two ancestral lineages resulting from an estimated 20–60 outcrossing events ( Tusso et al . , 2019 ) . This low outcrossing rate could result from limited opportunities to outcross , preferential same-clone mating , or a combination of the two . It is important to note , however , that the outcrossing rate estimates in S . pombe are likely low as pervasive meiotic drive and decreased recombination in heterozygotes suppress the genetic signatures that are used to infer outcrossing ( Bravo Núñez et al . , 2020b; Hu et al . , 2017; Nuckolls et al . , 2017; Zanders et al . , 2014 ) . Accounting for all these factors to generate a more accurate S . pombe outcrossing rate from population genetic data is currently an insurmountable obstacle for several reasons . First , the landscape of meiotic drivers and their suppressors is complex in that more than four drivers and more than eight suppressors of drive could be present in a given heterozygote produced by outcrossing ( Eickbush et al . , 2019; Bravo Núñez et al . , 2020a ) . Predicting how those factors will affect allele transmission is well beyond the capabilities of current population genetic models , even if the epistatic relationships within the drive systems were understood , which they currently are not . In addition to the drivers , S . pombe isolates contain an array of chromosomal rearrangements and other unknown barriers to recombination ( Avelar et al . , 2013; Brown et al . , 2011; Jeffares et al . , 2017; Tusso et al . , 2019; Zanders et al . , 2014 ) . Importantly , recombination rates are also inextricably linked to the complex drive landscape because the presence of drivers also affects the recombination rate within viable offspring , independent of chromosomal rearrangements ( Bravo Núñez et al . , 2020b ) . Given those limitations , the current laboratory knowledge and outcrossing rate estimates suggest that drivers would infrequently have the opportunity to act in S . pombe ( Egel , 1977; Ekwall and Thon , 2017; Forsburg and Rhind , 2006; Gutz and Doe , 1975; Miyata and Miyata , 1981; Tusso et al . , 2019 ) . In fact , the notion that S . pombe infrequently outcrosses has even been used to challenge the idea that S . pombe meiotic drivers are truly selfish genes that persist due to meiotic drive ( Sweigart et al . , 2019 ) . Nonetheless , the S . pombe genome houses numerous meiotic drive genes from the wtf gene family ( Bravo Núñez et al . , 2018; Bravo Núñez et al . , 2020a; Eickbush et al . , 2019; Hu et al . , 2017; Nuckolls et al . , 2017 ) . The wtf drivers destroy the meiotic products ( spores ) that do not inherit the driver from a heterozygote . Each wtf drive gene encodes both a Wtfpoison and a Wtfantidote protein that , together , execute targeted killing of the spores that do not inherit the wtf driver ( Hu et al . , 2017; Nuckolls et al . , 2017 ) . In the characterized wtf4 driver , the Wtf4poison protein assembles into toxic protein aggregates that are packaged into all developing spores . The Wtf4antidote protein co-assembles with Wtf4poison only in the spores that inherit wtf4 and likely neutralizes the poison by promoting its trafficking to the vacuole ( Nuckolls et al . , 2020 ) . Spore killing by wtf drivers leads to the loss of about half of the spores and almost exclusive transmission ( >90% ) of the wtf driver from a heterozygote ( Hu et al . , 2017; Nuckolls et al . , 2017 ) . Despite their heavy costs to the fitness of heterozygotes , the drivers are successful in that all assayed S . pombe isolates contain multiple wtf drivers ( 4–14 drivers; Bravo Núñez et al . , 2020a; Eickbush et al . , 2019; Hu et al . , 2017 ) . In addition to those drivers , the S . pombe isolates also contain between 8 and 17 suppressors of drive that encode only a Wtfantidote protein . In this work , we exploited the tractability of S . pombe to better understand how mating phenotypes , particularly inbreeding propensity , could affect the spread of a meiotic driver in this species . Despite limited genetic diversity among isolates , we observed natural variation in inbreeding propensity and other mating phenotypes . Some natural isolates preferentially undergo same-clone mating in the presence of a potential outcrossing partner , whereas others mate more randomly . Additionally , we found that the level of same-clone mating can be altered by cell density and affected by the available sexual partners . This is important as it highlights that our measured values in the lab are not meant to indicate precise levels of outcrossing that would occur under unknown natural conditions . To explore the effects that varying mating phenotypes could have on the spread of a wtf driver in a population , we used both mathematical modeling and an experimental evolution approach . We found that , while the spread of a wtf driver could be slowed by the observed levels of same-clone mating , the driver could still spread in the absence of linked deleterious traits . We incorporated our observations into a model in which rapid wtf gene evolution and occasional outcrossing facilitate the maintenance of wtf drivers . More broadly , this study illustrates how the success of drive systems is impacted by mating phenotypes .
To quantify mating propensities of homothallic S . pombe strains in the presence of other potential homothallic sexual partners , we first generated fluorescently tagged strains to easily observe mating via microscopy ( Figure 1A ) . We marked strains with either GFP or mCherry ( both constitutively expressed and integrated at the ura4 locus ) . We then mixed equal proportions of GFP-expressing and mCherry-expressing haploid cells and plated them on a medium ( SPA ) that induces one round of mating and meiosis . We imaged the cells immediately after plating to measure the starting frequency of both parent types . We then imaged again 24–48 hr later when many cells in the population had mated to either form zygotes or fully developed spores . We inferred the genotypes ( homozygous or heterozygous ) of each zygote and ascus ( spore sac ) based on their fluorescence ( Figure 1A ) . Homozygotes were produced by mating of two cells carrying the same fluorophore , while heterozygotes were produced by mating between a GFP-labeled and an mCherry-labeled cell ( Figure 1B and C ) . Finally , we calculated the inbreeding coefficient ( F ) by comparing the observed frequency of heterozygotes to the frequency expected if mating was random ( F = 1–observed heterozygotes/expected heterozygotes; Figure 1A ) . Exclusive mating between cells with the same fluorophore , random mating , and exclusive mating between cells with different fluorophores would yield coefficients of 1 , 0 , and –1 , respectively ( Hartl and Clark , 2007 ) . This assay occurs in a specific laboratory context , so the conditions differ from those in the wild . Importantly , however , we assay all strains under the same conditions , so the assay is sufficient to explore possible variation in mating phenotypes between different isolates of S . pombe . In homothallic cells of the common lab isolate , Sp , we measured an average inbreeding coefficient of 0 . 57 using our microscopy assay ( Figure 1B and D ) . As a control , we also assayed a mixed heterothallic population containing roughly equal amounts of GFP-expressing and mCherry-expressing cells of both mating types . We expected this control population to exhibit random mating between GFP- and mCherry-expressing cells , as heterothallic cells cannot undergo same-clone mating . We did observe random mating of this mixed heterothallic population ( F = −0 . 05 ) , which helps validate our assay ( Figure 1D ) . To further validate our microscopy results , we also assayed Sp cells using an orthogonal approach employing traditional genetic markers . For this analysis , we mixed haploid cells on supplemented SPA medium ( SPAS ) to induce them to mate and undergo meiosis . We then manually genotyped the progeny and used the fraction of recombinant progeny to calculate inbreeding coefficients ( Figure 1—figure supplement 2A ) . The average inbreeding coefficients measured using the genetic assay were very similar to the values we measured using the microscopy assay ( 0 . 49 for homothallic Sp cells; Figure 1—figure supplement 2B ) . Together , our results confirm and quantify previous observations of non-random mating in homothallic Sp cells ( Bendezú and Martin , 2013; Egel , 1977 ) . In addition , we demonstrate that our fluorescence assay provides a powerful tool to quantify mating events . We next extended our analyses to other S . pombe natural isolates collected from the wild . We assayed six additional isolates , FY29043 , FY29022 , FY28981 , FY28974 , FY29044 , and Schizosaccharomyces kambucha ( Sk ) , using our fluorescence microscopy assay . We chose these isolates because they each contain different fractions of the two inferred ancestral S . pombe lineages ( Tusso et al . , 2019 ) . In addition , the strains we chose are homothallic , sporulated well , were non-clumping , and we were able to transform them with the GFP and mCherry markers described above ( Figure 1—source data 1 , Figure 1—figure supplement 3; Jeffares et al . , 2015 ) . We found that the measured inbreeding coefficients varied significantly between the different natural isolates ( Figure 1D ) . The phenotype of FY29043 was similar to Sp , but other S . pombe isolates , including Sk , mated more randomly ( Figure 1D ) . We also observed variation in mating efficiency ranging from 10% of cells mating in FY28981 to 50% of cells mating in Sk ( Figure 1E ) . Given that S . pombe cells are immobile , we thought that cell density could affect their propensity to undergo same-clone mating . To test this , we compared the inbreeding coefficients of both homothallic Sp and Sk isolates at three different starting cell densities: our standard mating density ( 1× ) , high density ( 10× ) , and low density ( 0 . 1× ) . Because crowding prevented us from assaying high-density cells using our microscopy approach , we used the genetic assay for each condition . We found that inbreeding coefficients were higher in both Sp and Sk isolates when cell densities were lower ( Figure 1—figure supplement 2C and D ) . This is likely because cells plated at low density tended to be physically distant from potential sexual partners that were not part of the same clonally growing patch of cells ( Figure 1—figure supplement 4 ) . However , a control population consisting of mixed heterothallic Sp cells ( h+ and h- cells of two different genotypes mixed in equal proportions ) that cannot mate within a clonal patch of cells showed near random mating between the two genotypes at all cell densities assayed ( Figure 1—figure supplement 2C ) . Overall , these experiments demonstrate that inbreeding coefficients vary within S . pombe homothallic isolates and can be affected by cell density . We next used time-lapse imaging to explore the origins of the different inbreeding coefficients , focusing on the Sp and Sk isolates . Previous work claimed that Sk cells have reduced mating-type switching efficiency , based on levels of the DNA break ( double-strand break site [DSB] ) that initiates switching ( Singh and Klar , 2002 ) . A mutation at the mat-M imprint site was proposed to be responsible for the reduced level of DSBs ( Singh and Klar , 2003 ) . We reasoned that less mating-type switching could lead to less same-clone mating , as a small population of cells clonally derived from the same progenitor via mitosis would be less likely to contain cells of compatible mating types . This would promote cells mating outside their clonal lineage . One can see evidence consistent with this phenomenon in our experiments varying the cell density of homothallic Sp cells ( Figure 1—figure supplement 2C-D ) . Specifically , we observed more random mating when cells were at higher density , which likely results from more opportunity for cells to mate outside their clonal lineage . We aspired to directly compare mating type switching rates between Sp and Sk cells , although such direct assays have not , to our knowledge , been done in any strain background . We attempted to develop a direct assay using previously described cytological reporters of mating type . The reporters have been used to assay ratios of h + and h- cells in a population ( Jakočiūnas et al . , 2013 , Maki et al . , 2018 ) , but we found they were not well-suited for following switching events live in time-lapse imaging in our hands ( Figure 2—figure supplement 1A ) . Because of this , we decided to further explore the possible differences in mating-type switching frequencies indirectly using time-lapse assays , similar to those used to originally determine the patterns of mating-type switching in Sp ( Miyata and Miyata , 1981 ) . As in the classic study of Miyata and Miyata , 1981 , we used the first mating event as an imperfect proxy for when cells were capable of mating ( i . e . , they had a partner of the opposite mating type ) . For these assays , we tracked the fate of individual homothallic founder cells plated on SPA at low density ( 0 . 25× to our standard mating density used above ) to quantify how many mitotic generations occurred prior to the first mating event . When the first mating event occurred , we recorded the proportion of the cells present that mated ( prior to the appearance of cells from next mitotic generation ) . We also recorded the relationships between the cells that did mate ( Figure 2A ) . Specifically , we scored whether the mated cells were the product of a single mitotic division . Historically , these have been called ‘sister cells’ in the S . pombe mating literature , although we use the term ‘sibling cells’ to be gender neutral ( e . g . , Miyata and Miyata , 1981 ) . We did not consider cells that were born in mitotic generations past the one in which mating first occurred ( Figure 2A ) . For example , if cells within a given lineage first mated at generation 3 , we scored the relationships between those mated cells and recorded how many unmated cells remained at the end of generation 3 ( even if descendants of the unmated cells eventually mated ) . We then did not consider that lineage any further , so lineages in which cells first mated at generation 2 are not represented in the generation 3 data . Classic work characterizing mating-type switching patterns in Sp found that one in a group of four clonally derived cells will have the opposite mating type relative to the other three cells . A newly switched cell will be compatible to mate with close relatives , most typically its sibling cell ( Figure 2A; Miyata and Miyata , 1981 ) . Under the same switching model , a smaller portion of cells derived from a single division could be of opposite mating types and thus sexually compatible . For example , if one considers switchable cells ( e . g . , Ps in Figure 2A ) , they must divide only once to generate a pair of mating-competent cells . In Sp cells , we observed mating among the clonal descendants of some progenitor cells after a single mitotic division ( i . e . , at generation 2 ) . By the third generation , we observed mating among the descendants of more than half of the progenitor cells . Almost all the observed mating events were between sibling cells ( Figure 2A and B ) . These observations are consistent with published work assaying mating patterns within lineages of clonal Sp cells ( Bendezú and Martin , 2013; Klar , 1990; Miyata and Miyata , 1981 ) . Sk cells plated at 0 . 25× density on SPA divided significantly more than Sp cells prior to the first mating ( Wilcox rank sum test; p < 0 . 005 Figure 2B ) . Sk progenitor cells most frequently started mating at the fourth mitotic generation ( Figure 2B ) . This phenotype is consistent with less mating-type switching as more generations would be required on average to produce a cell with the opposite mating type ( Singh and Klar , 2002 ) . In addition , many mating events were between non-sibling cells . This phenotype can also be explained indirectly by reduced mating-type switching . Specifically , after cells undergo several divisions , they generate non-linear cell clusters in which comparably more non-sibling cells are in close proximity ( Figure 2—figure supplement 1A-B ) . This clustering could lead to more non-sibling mating than when cells are mating-competent after fewer divisions and the low number of cells are largely arranged linearly . To better understand the differences between Sp and Sk , we compared the sequence of the mating-type locus in the two isolates as many distinct alleles have been identified ( Nieuwenhuis et al . , 2018 ) . Consistent with previous work , we found that the mating-type regions of Sp and Sk are highly similar ( Singh and Klar , 2002 ) . However , using a previously published mate-pair sequencing dataset , we discovered an ~5 kb insertion of nested Tf transposon sequences in the Sk mating-type region ( Eickbush et al . , 2019 ) . We confirmed the presence of the insertion using PCR ( Figure 2—figure supplement 2A-B ) . We also found evidence consistent with the same insertion in FY28981 , which also mates more randomly than Sp ( Figure 2—figure supplement 2B , Figure 1D ) . We did not , however , formally test if the insertion affects mating phenotypes . Even if it does have an effect , it is insufficient to explain all the mating variation we observed as FY29044 mates randomly , yet it lacks the insertion ( Figure 1D and Figure 2—figure supplement 2B ) . To further explore the hypothesis that decreased mating-type switching efficiency in Sk could contribute to the mating differences we observed ( Figure 2B ) , we carried out time-lapse analyses of cells , at our standard 1× cell mating density . We reasoned that at this density , any given cell is likely to have a cell of opposite mating type nearby , even if mating-type switching is infrequent . We again used mixed populations of GFP- and mCherry-expressing cells to facilitate the scoring of mating patterns ( Figure 2C , Figure 2—video 1 supplement 1 ) . We found that the Sp cells predominantly mated in the second and third mitotic generations and most mating events were between sibling cells ( Figure 2C ) . The mating behavior of Sk cells changed more dramatically between 0 . 25× density and the higher 1× density . Whereas Sk cells tended to first mate in the fourth generation at 0 . 25× density , at 1× density Sk cells , like Sp , generally mated in the second and third mitotic generations ( Figure 2C , Figure 2—video 2 supplement 2 ) . Additionally , we observed significantly reduced levels of mating between Sk sibling cells at 1× density relative to 0 . 25× ( 10% and 56% , respectively; Figure 2E and B ) . These phenotypes are consistent with reduced mating-type switching in Sk . Specifically , our data suggest that Sk cells do not need to undergo more divisions before they are competent to mate . Rather , the additional divisions that occurred at 0 . 25× density in Sk could have been necessary to produce a pair of cells with opposite mating types . At 1× density , additional divisions are not expected to be required as additional non-sibling compatible partners are available . It is important to note , however , that our experiments combined with the previous work showing fewer switching-initiating DSBs in Sk ( Singh and Klar , 2002 ) , support , but do not conclusively demonstrate that mating-type switching occurs less frequently in Sk . In particular , our assays to understand switching frequencies use mating as a proxy for switching , which is not ideal . Therefore , reduced switching in Sk represents a promising model that remains to be tested . Still , our results conclusively demonstrate the key point that the mating phenotypes previously measured in Sp do not apply to all S . pombe isolates . Despite very little genetic diversity , S . pombe isolates maintain significant natural variation in key mating phenotypes ( Jeffares et al . , 2017 ) . While assaying inbreeding cytologically , we noticed that the Sk natural isolate displayed tremendous diversity in ascus size and shape ( Figure 1C , Figure 2—figure supplement 3A , Figure 2—video 2 ) . This was due to high variability in the size of the mating projections , known as shmoos . Sk produced long shmoos only in response to cells of the opposite mating type and not as a response to nitrogen starvation alone ( Figure 2—figure supplement 3B ) . The long Sk shmoos motivated us to quantify asci length across all the natural isolates described above . We found that most isolates generated zygotes or asci that were ~10–15 μm , similar to Sp . The majority of Sk zygotes and asci also fell within this range , but ~25% of Sk zygotes and asci were longer than 15 μm , with some exceeding 30 μm ( Figure 2—figure supplement 3C ) . We also assayed zygote/ascus length in an additional natural isolate in which we were unable to quantify inbreeding due to a clumping phenotype ( FY29033 ) . This isolate also showed populations of long asci , like Sk ( Figure 2—figure supplement 3C ) . Additionally , we occasionally noticed a fused asci phenotype in Sk ( Figure 2—figure supplement 3D ) . Time-lapse analyses of mating patterns , described above , revealed these fused asci can result from an occasional disconnect between mitotic cycles and the physical separation of cells ( Figure 2—figure supplement 3E ) . This phenotype is reminiscent of adg1 , adg2 , adg3 , and agn1 mutants in Sp that have defects in cell fission ( Alonso-Nuñez et al . , 2005; Gould and Simanis , 1997; Sipiczki , 2007 ) . Although we observed this phenotype in all time-lapse experiments using Sk cells , the prevalence of this phenotype varied greatly between experiments . We rarely observed this phenotype in Sp cells . We did not analyze time-lapse images of the other natural isolates , where this phenotype is most easily observed , so it is unclear if this septation phenotype occurs in other natural isolates . We next assayed if the mating preferences of S . pombe isolates Sp and Sk were invariable , or if they could be affected by the available mating partners due to mating incompatibilities ( Seike et al . , 2019b ) . To test this idea , we used both still and time-lapse imaging of cells mated at 1× density on SPA . For these experiments , we mixed fluorescently labeled Sp and Sk cells in equal frequencies . We observed in experiments employing still images that the overall inbreeding coefficient of the mixed Sk/Sp population of cells was intermediate between single-isolate crosses and mixed crosses ( Figure 2D ) . In time-lapse experiments , we observed that Sk cells maintained low levels of mating between sibling cells in the mixed Sk/Sp population ( 9 . 2% compared to 9 . 8% in a homogeneous population; Figure 2E ) . Among the Sp cells , mating between sibling cells decreased significantly from 56 . 8% to 29 . 7% in the mixed mating environment ( Figure 2E; t-test , p = 0 . 04 ) . Together , these results suggest that Sk cells can interfere with the ability of Sp cells to undergo same-clone mating . Although sibling cell mating preference changed , we did not observe a significant decrease in the mating efficiency of Sp cells in a mixed Sp/Sk population relative to a pure Sp population ( Figure 2—figure supplement 4A ) . Instead , the mating efficiency in the mixed Sp/Sk population was intermediate of those observed in pure Sp and Sk populations , indicating these isolates do not affect each other’s ability to mate . We were intrigued by the idea that long shmoos ( mating projections ) of Sk could contribute to its ability to disrupt Sp sibling mating . We were unable to address this idea directly . We did , however , find that Sk/Sp matings produce significantly longer zygotes/asci than either Sk/Sk or Sp/Sp matings ( Figure 2—figure supplement 4B ) . This was true even when we compared Sk/Sp zygote/ascus length to the length of heterozygous Sk/Sk or heterozygous Sp/Sp zygotes/asci . While this result does not prove that long Sk shmoos disrupt Sp sibling mating , it does show that long shmoos tend to be used in these outcrossing events . We next extended our analyses by assaying mating efficiency and inbreeding coefficients in all pairwise combinations of Sp , Sk , FY29043 , and FY29044 using still images of mated cells . After adjusting for mating efficiencies and parental inbreeding coefficients ( see Materials and methods ) , the phenotypes we observed in these crosses were mostly additive , in that they were intermediate to the pure parental strain phenotypes ( Figure 2—figure supplement 5 ) . The two exceptions were in the crosses between Sk and the isolates FY20943 and FY20944 . Sk formed more Sk/Sk homozygotes than expected in the two crosses ( one-tailed t-test , p = 0 . 043 and p = 0 . 038 , respectively ) , suggesting that Sk cells may not be fully sexually compatible with FY20943 and FY20944 ( Figure 2—figure supplement 5 ) . However , the magnitude of the effect was small in both cases and the statistical significance is lost after Bonferroni correction for multiple testing . Overall , our observations indicate that mating phenotypes of a given isolate can be affected by different mating partners . Importantly , however , our results suggest mating incompatibilities are unlikely to have a major role in limiting outcrossing within S . pombe . We next wanted to test how the observed range of inbreeding coefficients would affect the spread of a wtf driver in a population . To do this , we first used population genetic modeling . We used the meiotic drive model presented by J Crow , but we also introduced an inbreeding coefficient ( Hartl and Clark , 2007; Crow , 1991 ) ( see Materials and methods for a full description of the model ) . The model considers a population with two possible alleles at the queried locus and no genetic drift . We assumed a wtf driver would exhibit 98% drive ( transmission to 98% of spores ) in heterozygotes based on measured values for the Sk wtf4 driver ( Nuckolls et al . , 2017 ) . We assumed that all genotypes have the same fitness during haploid cell growth . For diploid cells induced to produce spores , we assumed homozygotes have a fitness of 1 ( e . g . , maximal fitness ) , whereas wtf driver heterozygotes have a fitness of 0 . 51 , since meiotic drive destroys nearly half of the spores ( Nuckolls et al . , 2017 ) . The inbreeding coefficient dictates the frequency of heterozygotes and thus the frequency at which the wtf driver can act . We varied the inbreeding coefficient ( F ) from 1 ( all matings generate homozygotes ) to –1 ( all matings generate heterozygotes ) . We used the model to calculate the predicted change in the frequency of a wtf driver after only one sexual generation ( Figure 3A ) . We also calculated the spread of a wtf driver in a population from a 5% starting frequency ( Figure 3B ) and from lower starting frequencies ( Figure 3—figure supplement 1A ) over generations of sexual reproduction . If we used an inbreeding coefficient of 1 , the frequency of the driver does not increase after sexual reproduction or spread in a population over time ( Figure 3A and B ) . No change in driver frequency was expected because no heterozygotes are produced under this condition , so no drive can occur . The wtf driver has the greatest advantage if the inbreeding coefficient is –1 , as all matings generate heterozygotes . Under all other conditions , including the range of inbreeding coefficients we measured experimentally in S . pombe natural isolates , some heterozygotes form . The wtf driver thus increases in frequency over generations of sexual reproduction , even when the driver starts at very low ( anything greater than 0 ) frequencies ( Figure 3A and B and Figure 3—figure supplement 1A , see Materials and methods ) ( Crow , 1991 ) . This model predicts that wtf drivers can spread under the modeled conditions if some non-same-clone mating occurs , even if it is infrequent . This observation is consistent with previous theoretical analyses demonstrating that inbreeding can slow the spread of meiotic drivers ( Martinossi-Allibert et al . , 2021 ) . We next wanted to consider the spread of a wtf driver under conditions in which genetic drift could occur ( Figure 3—figure supplement 1B and C ) . To do this , we simulated populations of different sizes ( 10–100 , 000 total individuals ) that started with one driver . We followed the population for 1000 generations of mating and spore formation . We randomly selected surviving ‘haploids’ to populate the next generation to keep the population size fixed . We assumed that all genotypes have equal fitness during haploid cell growth . For mathematical simplicity , we assumed complete drive and a corresponding fitness of 0 . 5 in heterozygous diploids . We again assumed that both types of homozygous diploids had a fitness of 1 . We also varied the inbreeding coefficient from 0 ( random mating ) to 1 ( all matings generate homozygotes ) . Surprisingly , we found that the probability of a driver’s success was related to population size , similar to the recent results of Martinossi-Allibert et al . , 2021 . Specifically , the driver was most likely to be maintained in the smaller populations and in populations with more random mating due to the driver’s positive effect on its own allele transmission ( Figure 3—figure supplement 1C ) . In larger populations in which the drivers started at a lower initial frequency , the drive allele generally took a long time to increase its frequency compared to the alternate allele , or it was lost due to drift . This was especially true when inbreeding coefficients were high ( Figure 3—figure supplement 1C ) . We next wanted to test if our predictions reflect the behavior of wtf drive alleles in a laboratory population of Sp cells over many generations . To do this , we constructed an experimental evolution system employing the GFP and mCherry fluorescent markers described above to measure changes in allele frequencies in a population over time using cytometry . To mark drive alleles , we linked the fluorescent markers with the Sk wtf4 driver and integrated the whole construct at the ura4 locus in Sp ( Nuckolls et al . , 2017 ) . For non-driving alleles , we used GFP or mCherry integrated at the ura4 locus without a linked wtf gene . We call the non-wtf alleles ‘empty vector’ . We started the experimental evolution populations with a defined ratio of GFP- and mCherry-expressing cells . We then induced a subset of the population to mate and sporulate followed by collection and culturing of the progeny ( spores ) . From these cells , we remeasured GFP and mCherry frequencies using flow cytometry , and we initiated the next round of mating and meiosis ( Figure 4A ) . Because our experiments rely on comparing the frequency of GFP- and mCherry-expressing cells over time , we needed to test the relative fitness of the markers . We found that both fluorescent markers were lost from all our experimental populations over time ( Figure 4—figure supplement 1A-B ) . This was likely because insertion of the markers disrupted the ura4 gene and cells that excised the marker reverted the ura4 mutation and thereby gained a fitness benefit . We therefore only considered fluorescent cells for our analyses and stopped the experiments when more than 95% cells lacked a fluorescent marker . In addition , in one set of experiments we also sorted cells at defined timepoints to remove non-fluorescent cells from our populations ( Figure 4D–E , Figure 4—figure supplement 1B and Figure 4—figure supplement 2C-D , described below ) . To assay for potential differences in the fitness costs of GFP and mCherry markers , we carried out our analyses in two control populations without drive . One control population lacked the Sk wtf4 driver while the other had Sk wtf4 linked to both fluorescent markers . For both types of controls , we analyzed homothallic ( inbreeding coefficient ~0 . 5 ) and mixed heterothallic ( inbreeding coefficient ~0 ) cell populations ( Figure 1D ) . We found in most cases that the number of mCherry-expressing cells increased at the expense of GFP-expressing cells over time ( Figure 4—figure supplement 2A-D ) . The notable exception was in heterothallic populations containing Sk wtf4 linked to both fluorophore alleles , where we did not observe a different cost of the GFP allele compared to mCherry ( Figure 4—figure supplement 2D ) . The origin of the differential fitness between mCherry and GFP alleles and why this cost was not observed in the one heterothallic population are both unclear . We did not determine the cause of the differences between them . To allow us to predict expected allele frequencies more accurately , we wanted to obtain a gross estimate of the fitness cost of the GFP-marked alleles relative to the mCherry-marked alleles in our experiments . To do this , we used the first six generations of data from our control crosses ( Figure 4—figure supplement 2 ) to fit a maximum likelihood model in which all parameters were fixed except the fitness values of GFP/mCherry heterozygotes and GFP/GFP homozygotes . We found that the fitness cost of the GFP in homozygotes was 0 . 234 ( see Materials and methods ) . We found the dominance of the GFP cost was 0 . 083 ( low fitness cost of GFP in GFP/mCherry heterozygotes ) . We then used these costs , calculated from our controls without meiotic drive , to calculate expected values in our experimental analyses in which drive can occur ( Figure 4B–E ) . For our experiments competing Sk wtf4 with an empty vector allele , we first assayed populations in which the alleles both started at 50% frequency . In homothallic ( inbreeding coefficient ~0 . 5 ) populations , we observed that wtf4 alleles spread in the population over several generations of sexual reproduction . The driver spread faster when linked to mCherry than when linked to GFP , presumably due to the aforementioned fitness costs linked to GFP ( Figure 4B–C ) . In both cases , the rate of spread of the allele was very close to our model’s predictions if we assumed an inbreeding coefficient of 0 . 5 ( black lines , Figure 4B–C ) and differed considerably from the model’s predictions assuming random mating ( inbreeding coefficient = 0; gray lines , Figure 4B–C ) . We saw similar spread of wtf4 in homothallic populations in a set of repeat experiments in which we sorted the cell populations twice to remove non-fluorescent cells ( Figure 4D and E ) . In these experiments , we also assayed mixed heterothallic populations with an equal mix of GFP- and mCherry-marked cells from both mating types . As described above , these mixed heterothallic populations show random mating ( inbreeding coefficient ~0 ) between GFP- and mCherry-labeled cells ( Figure 1D ) . In the mixed heterothallic populations , the wtf4 driver spread significantly faster than in homothallic cells . In generations 1–6 , the spread of wtf4 was very similar to that predicted by our model if we assumed random mating ( inbreeding coefficient = 0 ) . In later generations , our observations did not fit the model well . We suspect extensive loss of fluorescent cells , especially those with mCherry , and the resulting decrease in population size could contribute to this effect ( Figure 4D and E; Figure 4—figure supplement 1 ) . We compared our data to a model in which driver heterozygotes have a fitness of 0 . 51 , but we also considered variants of the model in which heterozygotes have fitness greater than 0 . 5 , as can occur with spore killers in Podospora anserina ( Vogan et al . , 2021; Martinossi-Allibert et al . , 2021 ) . In S . pombe , an increase in driver heterozygote fitness could occur if the driving alleles benefit from drive beyond the benefits gained directly by killing spores bearing the alternate allele . For example , the surviving meiotic products could theoretically gain fitness during spore development by scavenging increased resources from the killed meiotic products ( Nauta and Hoekstra , 1993 ) . We found that increasing heterozygote fitness beyond 0 . 51 decreased the fit of our data to the model ( Figure 5—figure supplement 2 ) , suggesting wtf drivers do not gain additional benefits beyond killing spores bearing the alternate allele . Overall , our results demonstrate that our population genetics model with driver heterozygote fitness at 0 . 51 is good at describing the spread of wtf4 in our experimental population , particularly in the first few generations . Our results also confirm that inbreeding coefficients near 0 . 5 slow , but do not stop , the spread of drivers in a population . Meiotic drivers tend to accumulate linked deleterious alleles in nature ( Atlan et al . , 2004; Dyer et al . , 2007; Finnegan et al . , 2019; Fishman and Saunders , 2008; Higgins et al . , 2018; Lyon , 2003; Olds-Clarke , 1997; Schimenti et al . , 2005; Unckless et al . , 2015; Wilkinson and Fry , 2001; Wu , 1983 ) . While the linkage of individual drivers to deleterious alleles in S . pombe has not been extensively investigated , driving alleles in the Sk isolate are linked to a chromosomal translocation that decreases fitness in heterozygotes ( Zanders et al . , 2014 ) . Similar chromosomal rearrangements involving chromosome 3 , which houses most wtf genes , are common in S . pombe , so it is likely drivers are frequently linked to rearrangements ( Bowen et al . , 2003; Brown et al . , 2011; Avelar et al . , 2013 ) . We therefore used the population genetic model to calculate the ability of a driver to spread when tightly linked to alleles with fitness costs ranging from 0 to 0 . 8 . We also varied the inbreeding coefficient from 0 ( random mating ) to 1 ( no heterozygotes are produced ) . We found that , in the absence of additional linked costs , wtf drivers are predicted to spread in a population at all initial frequencies greater than 0 ( Figure 5A ) . As described above , this spread is slowed by inbreeding , but is not stopped until inbreeding coefficients reach 1 ( no heterozygotes are produced ) . When the driver is burdened by additional fitness costs , it can still spread in a population . Importantly , however , as the fitness costs of linked deleterious alleles increase , the driver must start at a higher initial frequency to spread . If the fitness costs are recessive or close to recessive ( dominance coefficient h near 0 ) , a driver can invade at low frequencies in a randomly mating population , but a driver would require a higher initial frequency to spread in an inbreeding population ( inbreeding coefficients > 0; Figure 5A , Figure 5—figure supplement 1A ) . If the fitness costs are partially dominant ( 50% dominance ) , the drivers require a higher initial frequency to spread , even with random mating ( Figure 5—figure supplement 1B ) . In finite populations in which drift can occur , we also observed that linked costs could limit the maintenance of a driver particularly in populations that inbreed ( inbreeding coefficients > 0; Figure 5—figure supplement 1C , D ) . This is in line with previous modeling and can be explained because the cost of the deleterious allele is fixed , but the benefit a driver gains from drive is frequency dependent ( Drury et al . , 2017; Martinossi-Allibert et al . , 2021; Nauta and Hoekstra , 1993 ) . We next tested these predictions experimentally using the Sk wtf4 allele linked to GFP in homothallic cells . As described above , the GFP allele is linked to an unknown deleterious trait ( estimated 1 . 9% and 23 . 4% cost in heterozygotes and homozygotes , respectively ) . We varied the inbreeding coefficients of the populations by assaying cells mated at 1× and 0 . 1× density . As reported above , homothallic cells mated at 1× density exhibit an inbreeding coefficient of 0 . 5–0 . 57 , but that is increased to 0 . 8–0 . 95 by mating the cells at low ( 0 . 1× ) density ( Figure 1D , Figure 1—figure supplement 2B and C ) . Consistent with the predictions of our model , we observed in the experimental populations that the driver failed to spread when the initial frequency was less than 0 . 25 ( Figure 5B ) . When the wtf4:GFP allele was found in roughly half of the population , it could spread when the inbreeding coefficient was low but decreased in frequency when the inbreeding coefficient was increased ( Figure 5B ) . Similar , but less dramatic , effects were observed at higher initial frequencies of the wtf4:GFP allele . Altogether , our experimental analyses are consistent with the predictions of our model and show that both inbreeding and linked deleterious alleles can impede the spread of a wtf meiotic driver .
Mating phenotypes , particularly the outcrossing rate , are key parameters that affect the evolution of species ( Muller , 1932; Otto and Lenormand , 2002 ) . We sought to explore mating phenotypes in S . pombe to better understand the evolution of the wtf gene family found in this species . Although genetic variation within S . pombe is limited , past studies found variation in mating efficiency and uncovered genetic diversity of the mating-type locus ( Jeffares et al . , 2015; Nieuwenhuis et al . , 2018; Rhind et al . , 2011; Singh and Klar , 2002 ) . In this work , we assayed mating in an array of homothallic S . pombe natural isolates under a variety of laboratory conditions . Similar to previous work , we found Sp mating efficiency close to 40% and observed variable mating efficiencies for natural isolates ( Merlini et al . , 2016; Seike et al . , 2019a ) . In addition , we quantified the propensity of natural isolates to undergo same-clone mating when given the opportunity to mate with non-clonally related cells . We found this trait , as measured using an inbreeding coefficient , was variable between natural isolates and could be affected by cell density or available sexual partners . All of the S . pombe natural isolates we studied are homothallic and thus capable of same-clone mating . Despite this , we found that all isolates underwent some non-same-clone mating . In addition , some isolates , like Sk , showed considerable non-same-clone mating ( inbreeding coefficient near 0 ) under standard mating conditions . This provides additional data supporting homothallism is compatible with outcrossing between non-clonally related isolates ( Attanayake et al . , 2014 ) . In addition , S . pombe asci undergo a programmed degeneration process ( endolysis ) shortly after spore formation ( Encinar del Dedo et al . , 2009 ) . This presumably frees spores to potentially distribute ( in wind , water , or associated with animals ) separate from the other spores produced in the same meiosis . The physical independence of S . pombe spores could also facilitate outcrossing in this homothallic fungus ( Billiard et al . , 2012 ) . We did not definitively identify the molecular mechanisms underlying the variation in mating phenotypes we observed . Our data is , however , consistent with a model in which less frequent mating-type switching in the Sk isolate contributes to more random mating in Sk than in the common lab isolate , Sp . Specifically , if switching is less frequent in Sk , sibling cells are less likely to be compatible to mate . Incompatibility of sibling cells opens the possibility for mating with other , perhaps non-clonally derived , cells in the population . Singh and Klar were the first to propose that Sk switched less frequently than Sp when they noticed less of the DNA break that initiates switching ( Singh and Klar , 2002 ) . We discovered a large , nested insertion of transposon sequences in the mating-type locus of Sk , and we posit that this insertion could contribute to reduced DNA break formation and , potentially , decreased mating-type switching . It is important to note , however , that not all strains that mate randomly share this transposon insertion . We also stress that , as in previous work ( e . g . , Miyata and Miyata , 1981 ) , we did not directly assay switching rates and instead used mating events to infer information about switching . The long shmoos we observed in Sk may also contribute to more random mating in this isolate , as the long shmoo may increase the available number of partners within range . Additional previously described natural variation that we did not functionally explore may also contribute to differences in inbreeding propensity in S . pombe . For example , heterothallic natural isolates are predicted to exclusively undergo non-same-clone mating with cells of the opposite mating type ( Jeffares et al . , 2015; Nieuwenhuis et al . , 2018 ) . In addition , homothallic isolates with atypical mating-type loci with extra copies of the mat cassettes could grow into populations that are biased toward one mating type ( Nieuwenhuis and Immler , 2016 ) . Indeed , we analyzed the presumably expressed mating-type locus ( mat1 ) in several isolates for which we had nanopore sequencing data and found an approximate 3:1 excess of the h+ allele in FY29033 ( Figure 2—figure supplement 2C ) . The excess of one mating type is predicted to also facilitate non-same-clone mating . It is important to note that our study does not address the actual frequency of outcrossing in S . pombe populations in the wild . Very little is known about the ecology of fission yeast , including how frequently genetically distinct isolates are found in close enough proximity to mate ( e . g . , closer than ~40 µm apart ) ( Jeffares , 2018 ) . Outcrossing rates have been estimated using genomic data , but those estimates generally assume both that heterozygous recombination rates will match those observed in pure Sp and that allele transmission is Mendelian ( Farlow et al . , 2015; Tusso et al . , 2019 ) . Although these genomic estimates are reasonable , neither of these assumptions is consistent with empirical analyses ( Bravo Núñez et al . , 2020b; Hu et al . , 2017; Zanders et al . , 2014 ) . These assumptions have , therefore , likely led to an underestimation of the true outcrossing rate . To understand the evolution of the wtf drive genes , it is not necessarily essential to understand how frequently significantly diverged natural isolates , like those assayed in this work , mate . Instead , it is important to understand how often a driver is found in a heterozygous state . We suspect that heterozygosity for wtf drivers does not absolutely require outcrossing between more distantly related strains . This is because the wtf gene family , particularly the genes involved in meiotic drive , exhibit extremely rapid evolution . Even though genetic diversity within S . pombe is low ( <1% average DNA sequence divergence in non-repetitive regions ) , the wtf genes present in different isolates tend to be largely distinct ( Eickbush et al . , 2019; Hu et al . , 2017; Jeffares et al . , 2015; Rhind et al . , 2011 ) . The number of wtf genes per isolate varies from 25 to 38 wtf genes ( including pseudogenes ) , and even genes found at the same locus can be dramatically different ( e . g . , <61% coding sequence identity between alleles of wtf24 ) ( Eickbush et al . , 2019 ) . For example , wtf22 is a predicted pseudogene in one strain , a predicted antidote in another strain , and is predicted to encode distinct drivers ( i . e . , mutually killing ) in two more strains . There is only one wtf locus where all four strains that were surveyed each contain a meiotic driver , wtf4 ( Eickbush et al . , 2019 ) . Still , we do not consider this a fixed driver because the sequence of wtf4 is different in each strain , which , in all cases tested , leads to a distinct drive phenotype ( Bravo Núñez et al . , 2018; Bravo Núñez et al . , 2020a; Bravo Núñez et al . , 2020b; Eickbush et al . , 2019; Hu et al . , 2017 ) . For example , because Sk wtf4 and Sp wtf4 have different sequences , the antidote of Sk wtf4 does not neutralize Sp wtf4 and vice versa ( Bravo Núñez et al . , 2020a ) . The rapid evolution of wtf genes is driven largely by non-allelic gene conversion within the family and expansion or contraction of repetitive sequences within the coding sequences of the genes ( Eickbush et al . , 2019 ) . Because wtf genes generally provide no protection against wtf drivers with distinct sequences , the variation in wtf gene sequences has profound consequences ( Bravo Núñez et al . , 2018; Bravo Núñez et al . , 2020a; Hu et al . , 2017 ) . Even small sequence changes in wtf drivers can cause the birth and death of drivers . When a cell bearing a novel wtf driver mutation mates with a cell without the mutation , the driver is heterozygous , and thus , drive can occur and the novel allele can potentially spread through the otherwise largely homogeneous population . Given fission yeast cells have no inherent mobility , the novel drive alleles could arise within small isolated subpopulations ( demes ) in which drivers have an increased possibility of establishing , as was formally described by Martinossi-Allibert et al . , 2021 . Previous work assayed the strength of drive and the associated fitness reduction due to heterozygous wtf drivers ( Bravo Núñez et al . , 2018; Bravo Núñez et al . , 2020a; Hu et al . , 2017; Nuckolls et al . , 2017 ) . Those data , along with the inbreeding coefficients measured in this study , allowed us to mathematically model the spread of a wtf meiotic driver in an S . pombe population . Our modeling showed that the inbreeding coefficients we observed in S . pombe could slow the spread of a wtf driver . In the absence of drift , however , even the highest inbreeding coefficients we observed in S . pombe do not halt the spread of a driver , except in cases where the driver is found in low frequencies and linked to a deleterious allele . Given the tractability of S . pombe , we were also able to test the predictions of the model experimentally . Overall , our experimental results were quite similar to the model’s predictions discussed above . This suggests that our model encompasses all critical parameters . In addition , our experiments show how the wtf drivers can persist and spread in S . pombe , even if outcrossing is infrequent . The variation of mating phenotypes also indicates that the rate of spread of a wtf driver is expected to vary between different populations of S . pombe . Overall , our results are consistent with previous empirical and modeling studies of meiotic driver dynamics in populations . For example , like our fortuitously deleterious GFP allele , meiotic drivers are often linked to deleterious alleles that can hitchhike with the driver ( Atlan et al . , 2004; Dyer et al . , 2007; Finnegan et al . , 2019; Fishman and Saunders , 2008; Higgins et al . , 2018; Lyon , 2003; Olds-Clarke , 1997; Schimenti et al . , 2005; Unckless et al . , 2015; Wilkinson and Fry , 2001; Wu , 1983 ) . The added costs reduce the spread of drivers , which can lead a population to harbor a driver at stable intermediate frequency ( Dyer and Hall , 2019; Finnegan et al . , 2019; Fishman and Kelly , 2015; Hall and Dawe , 2018; Manser et al . , 2011 ) . Additionally , inbreeding can be selected as it increases fitness in a population when a driver is recessive lethal ( cost ~1 , h = 0 ) , such as in synthetic drive systems ( Bull , 2016 ) . To conclude , we would like to highlight the potential usefulness of the S . pombe experimental evolution approach developed for this study . With this system , we were able to observe the effects of altering allele frequencies , inbreeding rate , and fitness of a driving haplotype . In the future , this system could be used to experimentally explore additional questions about drive systems . For example , one could experimentally model meiotic drivers that bias sex ratios by linking the driver to the mating-type locus in a heterothallic population . In addition , one could explore the evolution of complex multi-locus drive systems employing combinations of multiple wtf meiotic drivers or drivers and suppressors . This tool could lead to novel insights about natural drivers , but it may also be particularly useful for exploring potential evolutionary trajectories of artificial gene drive systems ( Burt and Crisanti , 2018; Drury et al . , 2017; Price et al . , 2020; Wedell et al . , 2019 ) .
We introduced the fluorescent genetic markers into the genome using plasmids that integrated at the ura4 locus . To generate the integrating plasmids , we first ordered gBlocks from IDT ( Coralville , IA ) that contained mCherry or GFP under the control of a TEF promoter and ADH1 terminator ( Hailey et al . , 2002; Sheff and Thorn , 2004 ) . We digested the gBlocks with SpeI and ligated the GFP cassette into the SpeI site of pSZB331 and the mCherry cassette into the SpeI site of pSEZB332 ( alternate clone of pSZB331; Bravo Núñez et al . , 2020a; Bravo Núñez et al . , 2020b ) to generate pSZB437 and pSZB882 , respectively . We then linearized the plasmids with KpnI and transformed them into S . pombe using the standard lithium acetate protocol ( Schiestl and Gietz , 1989 ) . We independently transformed the isolates GP50 ( S . pombe ) , S . kambucha , FY28974 , FY28981 , FY29022 , FY29033 , FY29043 , and FY29044 . We were unsuccessful in transforming FY28969 , FY29048 , and FY29068 . FY29033 was not included in the inbreeding analyses due to its proclivity to clump . The homothallic and heterothallic strains carrying mCherry or GFP were transformed using the same method . To add Sk wtf4 to the Sp genome , we again used a ura4-integrating plasmid . To generate this plasmid , we amplified Sk wtf4 from SZY13 using the oligos 688 and 686 . We digested the amplicon with SacI and ligated into the SacI site of pSZB332 to generate pSZB716 ( Bravo Núñez et al . , 2020a; Bravo Núñez et al . , 2020b ) . We then separately introduced the GFP and mCherry gBlocks into the SpeI site of pSZB716 to generate pSZB904 and pSZB909 , respectively . We introduced the resulting plasmids into yeast as described above . We performed crosses using standard approaches ( Smith , 2009 ) . We cultured each haploid parent to saturation in 3 mL YEL ( 0 . 5% yeast extract , 3% dextrose , and 250 mg/L adenine , histidine , leucine , lysine , and uracil ) for 24 hr at 32°C . We then mixed an equal volume of each parent ( 700 μL each for individual homothallic strain , 350 μL for heterothallic parents ) , pelleted and resuspended in an equal volume of ddH2O ( 1 . 4 mL total ) , then plated 200 μL on SPA ( 1% glucose , 7 . 3 mM KH2PO4 , vitamins , and agar ) for microscopy experiments or SPAS ( SPA +45 mg/L adenine , histidine , leucine , lysine , and uracil ) for genetics experiments . We incubated the plates at 25°C for 1–4 days , depending on the experiment ( see figure legends for exact timing ) . When we genotyped spore progeny , we scraped cells off of the plates and isolated spores after treatment with B-Glucuronidase ( Sigma ) and ethanol as described in Smith , 2009 . We grew haploid isolates to saturation in 3 mL YEL overnight at 32°C . We washed the cells once with ddH2O then resuspended them in an equal volume ddH2O . We then spotted 10 μL of each strain onto an SPAS plate , which we then incubated at 25°C for 4 days prior to staining with iodine ( VWR iodine crystals ) vapor ( Forsburg and Rhind , 2006 ) . We used mate-pair Illumina sequencing reads to assemble the mating-type locus of S . kambucha with previously published data ( Eickbush et al . , 2019 ) . We assembled the mating-type locus using Geneious Prime software ( https://www . geneious . com; last accessed March 18 , 2019 ) using an analogous approach to that described to assemble wtf loci ( Eickbush et al . , 2019 ) . To extract DNA for nanopore sequencing , we used a modified version of a previously developed protocol ( Jain et al . , 2018 ) . We pelleted 50 mL of a saturated culture and proceeded as described , with the addition of 0 . 5 mg/mL zymolyase to the TLB buffer immediately prior to use . We used a MinION instrument and R9 MinION flow cells for sequencing . For library preparation , we used the standard ligation sequencing prep ( SQK-LSK109 ) , including end repair using the NEB End Prep Enzyme , FFPE prep using the NEB FFPE DNA repair mix , and ligation using NEB Quick Ligase . We did not barcode samples and thus used each flow cell for a single genome . We used guppy v2 . 1 . 3 for base calling . We removed sequencing adapters from the reads using porechop v0 . 2 . 2 and then filtered the reads using filtlong v0 . 2 . 0 to keep the 100× longest reads . We then error corrected those reads , trimmed the reads and de novo assembled them using canu v1 . 8 and the ovl overlapper with a predicted genome size of 13 mb and a corrected error rate of 0 . 12 ( Koren S et al . , 2017 ) . Base called reads are available as fastq files at the SRA under project accession number PRJNA732453 . In order to count allele frequency within the active mat locus , we mapped raw reads back to the corresponding de novo assembly using graphmap v0 . 5 . 2 and processed using samtools v1 . 12 ( Li et al . , 2009; Sović et al . , 2016 ) . We then visually observed the reference-based assemblies using IGV v2 . 3 . 97 to count the number of h + and h- alleles present at the active mating type locus with anchors to unique sequence outside the mat locus ( Robinson et al . , 2011 ) . We mixed haploid parents ( a GFP- and an mCherry-expressing strain ) in equal proportions on SPA as described above . We then left the plate to dry for 30 min and then took a punch of agar from the plate using a 1271E Arch Punch ( General Tools , Amazon ) . We then inverted the punch of agar into a 35 mm glass bottomed dish ( No 1 . 5 MatTek Corporation ) . We used this sample to count the initial frequency of the two parental types . We then imaged a second punch of agar taken from the same SPA plate after 24 hr incubation at 25°C for homothallic cells and 48 hr for heterothallic cells . Each biological replicate was constituted by a separate cross . To image the cells , we used an AXIO Observer . Z1 ( Zeiss ) widefield microscope with a 40× C-Apochromat ( 1 . 2 NA ) water-immersion objective . We excited mCherry using a 530–585 nm bandpass filter which was reflected off an FT 600 dichroic filter into the objective and collected emission using a long-pass 615 nm filter . To excite GFP , we used a 440–470 nm bandpass filter , reflected the beam off an FT 495 nm dichroic filter into the objective and collected emission using a 525–550 nm bandpass filter . We collected emission onto a Hamamatsu ORCA Flash 4 . 0 using µManager software . We imaged at least three different fields for each sample as each technical replicate . We used cell shape to identify mated cells ( zygotes and asci ) and used fluorescence to identify the genotype of each haploid parent . To measure both fluorescence and the length of asci , we used Fiji ( https://imagej . net/Fiji ) software to hand-draw five pixel-width lines through the length of each zygote or ascus . After subtracting background using a rolling ball background subtraction with width 50 pixels , we then measured the average intensity for the GFP and mCherry channels . When measuring the log10 ratio of GFP over mCherry , the mCherry homozygotes have the lowest ratio , homozygotes for GFP the highest ratio , and heterozygotes intermediate . To calculate the inbreeding coefficient , we used the formula F = 1 − ( observed heterozygotes/expected heterozygotes ) . We used Hardy-Weinberg expectations to calculate the expected frequency of heterozygotes ( 2p ( 1p ) ) for each sample , where ‘p’ is the fraction of mCherry+ cells and ( 1-p ) is the fraction of GFP+ cells measured prior to mating ( Hartl and Clark , 2007 ) . For time-lapse imaging of cells mated at 1× density ( Figure 2C ) , we prepared cells using the agar punch method described above . For cells at 0 . 25× density ( Figure 2B ) , we used the same approach , except we cultured cells in 3 mL EMM ( 14 . 7 mM C8H5KO4 , 15 . 5 mM Na2HPO4 , 93 . 5 mM NH4Cl , 2% w/v glucose , salts , vitamins , minerals ) then washed three times with PM-N ( 8 mM NA2HPO4 , 1% glucose , EMM2 salts , vitamins , and minerals ) before plating cells to SPA . While imaging the cells , we added a moistened kimwipe to the MatTek dish to maintain humidity . We sealed the dish lids on with high-vacuum grease ( Corning ) . We imaged cells using either a Ti Eclipse ( Nikon ) coupled to a CSU W1 Spinning Disk ( Yokagawa ) , or a Ti2 ( Nikon ) widefield using the 60× oil immersion objective ( NA 1 . 45 ) , acquiring images every ten minutes for 24–48 hr , using a 5 × 5 grid pattern with 10% overlap between fields . The Ti Eclipse was used for one replicate each of the 1× crosses and the Ti2 was used for all remaining experiments . We used an Okolab stage top incubator to maintain the temperature at 25°C . For the Ti2 ( widefield ) data we excited GFP through a 470/24 nm excitation filter and collected through an ET515/30 m emission filter . For mCherry on this system , we excited through a 550/15 nm excitation filter and collected through an ET595/40 m emission filter . For the Ti Eclipse ( confocal ) data , we excited GFP with a 488 nm laser and collected its emission through an ET525/36 m emission filter . For mCherry on this system , we excited with a 561 nm laser and collected through an ET605/70 m emission filter . To monitor mating in 1× crosses ( Figure 2C-E ) , we recorded the number of divisions and mating choice of the progeny of 286 cells until an expected mating efficiency for the population being filmed was attained . The expected mating efficiency was calculated from still images of the same crosses . We recorded two videos of each cross . To monitor the number of divisions required before mating could occur in 0 . 25× cultures ( Figure 2B ) , around 200 individual cells were monitored through the duration of the generation in which the first mating event occurred . If cells failed to mate , they were monitored throughout the duration of the movie . If a cell or its mitotic offspring interacted with a neighboring cell cluster , it was not included in the analysis . We recorded two videos for each isolate . We calculated mating efficiency from microscopic images using the following formula:Mating Efficiency%= 2Z+2A+S2V+2Z+2A+S2*100 where Z represents the number of zygotes , A represents the number of asci , S represents the number of free spores , and V represents the number of vegetative cells ( Seike and Niki , 2017 ) . In an attempt to directly observe mating-type switching , we used the strain TP220 ( Jakočiūnas et al . , 2013 ) . This strain contains a dual reporter system with YFP under control of the h- cell-specific mfm3 promoter and CFP under control of the h + cell-specific map2 promoter . Cells were cultured for 24 hr in PM-N+ ade + leu and then plated to SPA+ ade + leu ( at 0 . 5× density ) or SPAS ( at 1× density ) ( see figure legends ) . Plates were incubated at 25°C for 1 or 12 hr ( SPA+ ade + leu and SPAS , respectively ) before we prepared cells for microscopy using the agar punch method described above . Two microscopes were used for this analysis . In one case , a Nikon Ti microscope coupled to a Yokogawa CSU W1 spinning disc was used . CFP was excited with 445 nm and YFP was excited with 515 nm laser light . Cells were imaged with a 60× Plan Apochromat Objective ( NA 1 . 4 ) . CFP fluorescence was filtered through a 480/30 bandpass filter and YFP was filtered through an ET535/36 m bandpass filter . Fluorescence was recorded with an iXon DU897 EMCCD ( Andor ) . In a separate experiment , a Nikon Ti2-E widefield microscope was used . Here , a bulb was filtered to excite CFP with a 440/20 nm bandpass filter while YFP was excited with a 510/25 nm bandpass filter . The cells were imaged with a 60× Plan Apochromat phase three objective ( NA 1 . 4 ) . The fluorescence of CFP was filtered through an ET472/24 nm bandpass filter and YFP was filtered through a 542/21 nm bandpass filter . Here , images were acquired with a Prime 95 sCMOS camera ( Photometrics ) . In both datasets , images were acquired every 20 min for 24 hr . Images from these datasets were Gaussian blurred one pixel , then rolling ball background subtracted with a radius of 50 pixels . Sample movement was eliminated by registering the images using the plugin ‘StackRegJ_’ in Fiji . Subsections of the image were created by duplicating out 500 × 500 pixel regions from this large image . These were then Gaussian blurred one pixel again and brightness and contrast adjusted as needed to illustrate the fluorescence . Finally , these small fields of view were cropped again to yield the images in Figure 2—figure supplement 1 . We used a high-throughput system to genotype the progeny from each cross . First , we crossed two parental populations to generate spore progeny as described above . In addition to placing the mixed haploid cells on SPAS , we also diluted a subset of the mix and plated it onto YEAS ( YEA +45 mg/L adenine , histidine , leucine , lysine , and uracil ) . We genotyped the colonies that grew on that YEAS plate to measure the starting frequency of each parental strain in the cross . We plated the spores produced by the cross on YEAS and grew them at 32°C for 4 days . We picked the colonies using a Qpix 420 Colony Picking System and cultured them in YEL for 24 hr at 30°C in 96-well round-bottom plates ( Axygen ) . We then used a Singer RoTor robot to spot the cultures to YNP dropout and YEAS drug plates and incubated them at 32°C for 3 days . We then imaged the plates using an S&P robotics SPImager with a Canon EOS Rebel T3i camera . We analyzed each picture using the subtract background feature in Fiji and assigned regions of interest to the 384 spots where cells were pinned . We then measured the average intensity of each spot and classified cells as grown or not by a heuristic threshold . We genotyped some cross progeny manually using standard techniques due to robot unavailability , with indistinguishable results . A single cross was considered as one biological replicate . We then inferred the frequency of mating between cells of distinct genotypes based on the frequency of recombinant progeny using a combination of either two or three unlinked genetic markers . If mating was random , we expect the progeny to reflect Hardy-Weinberg expectations ( p2+2p1-p+1-p2=1 ) , where p2+1-p2 reflect the expected frequency of homozygotes and 2p ( 1-p ) reflects the expected frequency of outcrossing . Homozygotes of either parental genotype can only produce offspring with the parental genotypes . If the two parental strains outcross to generate heterozygotes , they will make the parental genotypes and recombinant genotypes all in equal frequencies ( 2n total genotypes where n is the number of segregating markers ) . For our crosses with three markers , we therefore expected the true ‘observed’ frequency of progeny produced by outcrossing to be equal to the number of observed recombinants divided by 6/8 . For our crosses with two unlinked markers , we divided by 2/4 . We then calculated the inbreeding coefficient using the formula , F=1-the true observed fraction of progeny produced by outcrossing2p1-p . We calculated the expected zygote frequencies when isolates were mated with different isolates on SPA ( see Measuring inbreeding coefficients by microscopy ) using an additive model that incorporated mating efficiencies and inbreeding coefficients measured from the isogenic crosses . The model assumed that each strain contributes equally to the cross , and that they do not change their own mating in response to the mating partner . We calculated the expected frequency of homozygotes for parental strain one as:=p2+p1-pFI11-us1 , where the inbreeding coefficient is FI1 and the mating efficiency is 1-us1 for parental strain 1 ( s1 ) , considering its initial frequency p . The expected heterozygote frequency is:=2p1-p1- FI1+FI221-us1+us22 . The expected fraction of homozygotes for parental strain two is:=1- ( Homozygote frequency strain 1+ Heterozygote frequency ) To model the expected changes in allele frequencies in a randomly mating population over time , we used the equations described by Crow , 1991 . For nonrandomly mating populations , we included the ‘F’ inbreeding coefficient in the equations , similar to Hartl and Clark , 2007 . We assumed that wtf4 had a drive strength ( ‘k' ) of 0 . 98 based on experimental observations ( Nuckolls et al . , 2017 ) . We assumed the control allele exhibited Mendelian transmission ( k = 0 . 5 ) . Simulations for the spread of a driver in Figure 3 only considered drive and inbreeding coefficients . To simulate drive in fluorescent populations , the starting frequencies of each allele ( i . e . , ‘p’ ) were determined empirically for each experiment using either traditional genetic approaches or cytometry . Predicted values were calculated from each biological replicate ( independent cross ) , composed of three technical replicates measured via cytometry per cross . We assumed all genotypes had equal fitness as haploids . For relative fitness of diploids , we assigned mCherry/mCherry homozygotes a fitness of w11=1 , regardless of whether they were EV/EV or wtf4/wtf4 homozygotes . In all but one cross ( see below ) , we observed a fitness cost linked to the GFP alleles relative to mCherry alleles during sexual reproduction , regardless of wtf4 . We therefore used our data ( see below ) to calculate 0 . 234 as the fitness cost of the GFP-linked variant . Because of that , we assigned a fitness value of 0 . 766 to GFP homozygotes , w22 . We found the fitness cost linked to GFP had low dominance ( see below ) and thus assigned a fitness value of 0 . 98 for GFP/mCherry homozygous for wtf4 or empty vector , w12 . For GFP/mCherry heterozygotes that were also heterozygous for wtf4 , we assigned a fitness of ( 0 . 5 ) that accounts for the death of spores killed by drive ( 0 . 51 ) and the GFP cost ( 0 . 234 ) and dominance ( 0 . 083 ) . The calculation for allele frequency for a wtf meiotic driver in consecutive sexual cycles from haploid populations is:pg+1=p2g+Fpg ( 1-p ) gw11+2pg1-pgk ( 1-F ) w12W-g . where the mean fitness of the population at each generation ( g ) W-g=p2gw11+2pg1-pgw12+ ( 1-p ) 2gw22+Fpg1-pgw11+w22-2w12 . When the fitness reduction in heterozygotes is only due to dead cells by drive with no additional costs , w12=12k , the drivers spread under any initial frequency greater than 0 ( Crow , 1991 ) . To determine the fitness cost of the GFP-linked variant that was present in most of our experiments , we used the L-BFGS-B algorithm to find a fitness that maximized the likelihood that our observed allele frequencies varied only due to the linked cost of GFP ( c ) and dominance ( h ) in the first six generations of the control experiments shown in Figure 4—figure supplement 2 ( Byrd et al . , 1995 ) . To do this , we used the mle function from the R stats4 package ( Team , 2019 ) . We assumed the fitness cost alters the relative fitness of both homozygotes w22=1-c and heterozygotes w12=1-ch . The fitting was done using only the initial six generations due to a rapid loss of fluorescent cells from seven to ten generations ( Figure 4—figure supplement 1 ) . To calculate the minimum initial frequency of driver required to spread in a population when linked to alleles with varying additional fitness costs , we use the additional linked cost ( c ) and their dominance ( h ) associated to each genotype and solved using ( Wolfram , 1987 ) . For simplicity we also assumed complete transmission bias , k=1 . A wtf meiotic driver linked to a deleterious allele can spread under the condition ( critical value ) thatFcritical<ch ( p−1 ) −cp+pc ( h−1 ) ( p−1 ) +p . where p is the initial frequency of a wtf driver , c is the cost linked to the driver , and h is the dominance coefficient . For this case the assigned relative fitness for homozygote carrying the wtf driver is w11=1-c . To determine the possibility of a meiotic driver to spread and be fixed in a finite population , we ran stochastic simulations using a Wright-Fisher model considering multiple population sizes . We assumed no fitness differences between genotypes during haploid cell growth , complete drive ( k = 1 ) , gamete killing proportional to drive in heterozygotes , no linked deleterious alleles , and non-overlapping generations . We used characters of ‘0’ for non-driving cells and ‘1’ for driving cells ( Figure 3—figure supplement 1B ) . Initially the population starts with population size ( N ) . We next sampled a fixed population size ( n ) with replacement , assuming infinite gamete supply used for mating ( e . g . , as if mitotic expansion of the population occurred prior to mating ) . In our simulations N = n . We sampled mating groups without replacement for an inbreeding population ( Spop ) with a fraction F and the rest of cells ( n − Spop ) mate randomly . Populations that mate randomly are proportional to Hardy-Weinberg genotype frequencies with approximated integer values . Homozygotes ( ‘0 , 0’ and ‘1 , 1’ ) produce four progeny ( ‘0 , 0 , 0 , 0’ and ‘1 , 1 , 1 , 1’ , respectively ) , while heterozygotes produce two progeny that inherit the drive allele ( ‘1 , 1’ ) . When a fitness cost is associated with a genotype , we decimated the gametic products in proportion to the cost . The collection of 0’s and 1’s after this process constituted the next generation . We repeated 1000 generations for 1000 iterations for each tested condition . We performed the crosses and collected spores as described above . We then started the next generation by culturing 60 µL of spores from each cross in each of three different wells with 600 µL fresh YEL media in 96 deep-well-round-bottom plates ( Axygen ) and cultured for 24 hr at 1200 rpm at 32°C . We then transferred 60 µL from each culture to a new plate with 600 µL YEL and cultured for 12–14 hr at 1200 rpm at 32°C . We then pooled the culture replicates in an Eppendorf tube , spun down , and resuspended them in an equal volume of ddH2O . We then took 100 µL of this sample to assay via cytometry ( described below ) . We also plated 200 µL of each sample on SPAS plates and incubated at 25°C for 5 days to allow the cells to mate and sporulate . To detect and quantify fluorescent cells via flow cytometry , we used the ZE5 Cell Analyzer ( Bio-Rad ) . We spun down 100 µL of each culture , washed the cell pellet with water , spun down again , and resuspended the cells in 200 µL of 1× PBS ( phosphate buffered saline ) with 1 . 5 μL 4′ , 6-diamidino-2-phenylindole ( DAPI , Sigma-Aldrich , 100 ng/mL ) . DAPI stains dead cells , so we considered DAPI-negative cells as live cells . To image DAPI , we used 355 nm laser excitation and a 447/60 nm detector . To image GFP , we excited using a 488 nm laser and detected emission with a 525/35 nm filter . To image mCherry , we excited using a 561 nm laser and detected emission with a 615/24 nm filter . We used 405 and 488 nm ( FSC 405 and FSC 488 ) lasers for forward scatter and 488 nm laser for side scatter ( SSC 488 ) . To quantify the frequency of GFP- or mCherry-positive cells , we analyzed the Flow Cytometry Standard files in R/Bioconductor using the packages FlowTrans and FlowClust ( Lo et al . , 2009 ) . We first separated the cells that had round and uniform shape and similar granularity . This step allowed us to detect a more uniform population of single cells . We then discarded DAPI-positive cells . To determine the GFP- and mCherry-positive cells , we used limits for each channel . The limits varied throughout the experiment due to reconfiguration in the flow cytometer . For every measurement , we corrected with standard samples of cells that only expressed either GFP or mCherry . Fluorescent GFP- and mCherry-positive cells always showed non-overlapping cell populations . We quantified cells that were not classified as non-fluorescent cells . In the experimental evolution experiments ( Figures 4 and 5B , and Figure 4—figure supplement 2 ) , a large proportion of cells lost their fluorescent marker over time ( Figure 4—figure supplement 1A-B ) . We assume this is because we introduced the markers using an integrating plasmid that enters the genome following a single crossover event . Because of this , a single crossover can then pop the marker and the associated vector out of the genome . We repeatedly observed that the loss occurred faster with the mCherry markers . Consistent with this model , cells that lost fluorescence generally also lost the associated drug resistance marker present on the integrating vector . We only considered fluorescent cells in our analyses . In the experiments ( Figure 4B–C , Figure 4—figure supplement 1A-B , and Figure 5B ) , we extended the evolution up to 10 generations . In other experiments ( Figure 4D–E and Figure 4—figure supplement 2C-D ) , we removed non-fluorescent cells by cell sorting after generations 2 and 5 . To sort cells , we first collected and cultured spores as described above with the first culture done for 12 hr . We then transferred 60 µL of germinated cells into 600 µL fresh YEL media . We used three cultures for each experimental line and pooled all in equal amounts to have 1 . 4 mL of each line . We spun each sample down and resuspended the cells in 5 mL ddH2O prior to sorting . We removed non-fluorescent cells and retained GFP-positive ( 488 nm laser for excitation and a filter 507 nm ) and mCherry-positive ( 561 nm laser for excitation and a 582 nm filter ) cells using the laser the BD FACSMelody cell sorter software . We collected 1 . 2 million cells for each line into 1× PBS . We then spun down the cells and resuspended the pellets in 200 µL YEL . We then took 60 µL from each sample and diluted the cells into 600 µL YEL and continued with the time course as described above . This restored fluorescently labeled cells populations as expected for homothallic and heterothallic lines ( Figure 4—figure supplement 1B ) . Cell sorting did not affect our results as we observed the same patterns in replicate experiments in which we did not remove the non-fluorescent cells ( Figure 4 ) . Analysis performed to filter , measure , and compare flow cytometry with simulations are located in the Zanders Lab Github repository , https://github . com/Zanders-Lab/Diverse_mating_phenotypes_impact_the_spread_of_wtf_meiotic_drivers_in_S . -pombe; Zanders , 2021 , http://www . stowers . org/research/publications/libpb-1625 . | The fission yeast , Schizosaccharomyces pombe , is a haploid organism , meaning it has a single copy of each of its genes . S . pombe cells generally carry one copy of each chromosome and can reproduce clonally by duplicating these chromosomes and then dividing into two cells . However , when the yeast are starving , they can reproduce sexually . This involves two cells mating by fusing together to create a ‘diploid zygote’ , which contains two copies of each gene . The zygote then undergoes ‘meiosis’ , a special type of cell division in which the zygote first duplicates its genome and then divides twice . This results in four haploid spores which are analogous to sperm and eggs in humans that each contain one copy of the genome . The spores will grow and divide normally when conditions improve . The genes carried by each of the haploid spores depend on the cells that formed the zygote . If the two ‘parent’ yeast had the same version or ‘allele’ of a gene , all four spores will have it in their genome . However , if the two parents have different alleles , only 50% of the offspring will carry each version . Although this is usually the case , there are certain alleles , called meiotic drivers , that are transmitted to all offspring even in situations where it is only carried by one parent . Meiotic drivers can be found in many organisms , including mammals , but their behavior is easiest to study in yeast . Meiotic drivers known as killers achieve this by disposing of any ‘sister’ spores that do not inherit the same allele of this gene . This ‘killing’ can only happen when only one of the ‘parents’ carries the driver . This scenario is thought to rarely occur in species that inbreed , as inbreeding leads to both gene copies being the same . However , this does not appear to be the case for S . pombe , which contain a whole family of killer meiotic drivers , the wtf genes , despite also being reported to mainly inbreed . To investigate this contradiction , López Hernández et al . isolated several genetically distinct populations of S . pombe . These isolates were grown together to determine how often the each one would outcross ( mate with an individual from a different population ) or inbreed . The results found that levels of inbreeding varied between isolates . Next , López Hernández et al . used mathematical modelling and experimental evolution analyses to study how wtf drivers spread amongst these populations . This revealed that wtf genes spread faster in populations with more outcrossing . In some instances , the wtf driver was linked to a gene that could harm the population . In these cases , López Hernández et al . found than inbreeding could purge these drivers and stop them from spreading the dangerous alleles through the population . López Hernández et al . establish a simple experimental system to model driver evolution and experimentally demonstrate how key parameters , such as outcrossing rates , affect the spread of these genes . Understanding how meiotic drivers spread is important , as these systems could potentially be used to modify populations important to humans , such as crops or disease vectors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"genetics",
"and",
"genomics"
] | 2021 | Diverse mating phenotypes impact the spread of wtf meiotic drivers in Schizosaccharomyces pombe |
Ultraviolet-protective compounds , such as mycosporine-like amino acids ( MAAs ) and related gadusols produced by some bacteria , fungi , algae , and marine invertebrates , are critical for the survival of reef-building corals and other marine organisms exposed to high-solar irradiance . These compounds have also been found in marine fish , where their accumulation is thought to be of dietary or symbiont origin . In this study , we report the unexpected discovery that fish can synthesize gadusol de novo and that the analogous pathways are also present in amphibians , reptiles , and birds . Furthermore , we demonstrate that engineered yeast containing the fish genes can produce and secrete gadusol . The discovery of the gadusol pathway in vertebrates provides a platform for understanding its role in these animals , and the possibility of engineering yeast to efficiently produce a natural sunscreen and antioxidant presents an avenue for its large-scale production for possible use in pharmaceuticals and cosmetics .
The sunscreen compounds , mycosporine-like amino acids ( MAAs ) and related gadusols , commonly found in bacteria , fungi , algae , and marine invertebrates ( Shick and Dunlap , 2002; Miyamoto et al . , 2014 ) , have been proposed to fulfill a variety of functions , such as sunscreen , antioxidant , stress response , intracellular nitrogen reservoir , and/or optical filter ( Gao and Garcia-Pichel , 2011; Bok et al . , 2014 ) . Although their formation had long been proposed to originate from the shikimate pathway , more recent bioinformatic and biochemical studies revealed that in cyanobacteria , MAAs are synthesized by desmethyl-4-deoxygadusol synthase ( DDGS ) , a dehydroquinate synthase ( DHQS ) homolog ( Wu et al . , 2007; Balskus and Walsh , 2010; Singh et al . , 2010; Asamizu et al . , 2012 ) . Interestingly , inactivation of the DDGS gene in Anabaena variabilis ATCC 29413 did not abolish the production of MAAs , suggesting that additional pathways exist for the biosynthesis of MAAs ( Spence et al . , 2012 ) . DDGS converts sedoheptulose 7-phosphate ( SH7P ) to desmethyl-4-deoxygadusol via a unique sequence of dephosphorylation , aldol condensation , enolization , dehydration , reduction , and tautomerization reactions ( Figure 1—figure supplement 1 ) ( Balskus and Walsh , 2010; Asamizu et al . , 2012 ) . The product is subsequently converted by a methyltransferase to 4-deoxygadusol , the building block of MAAs . 4-Deoxygadusol has also been proposed to be the precursor of gadusol ( Starcevic et al . , 2010; Rosic and Dove , 2011 ) , a related compound initially isolated from cod roe ( Gadus morhua L . ) ( Plack et al . , 1981 ) , but also found in roes of other marine fish ( Plack et al . , 1981; Arbeloa et al . , 2010 ) , sea urchin eggs ( Chioccara et al . , 1986 ) , cysts and nauplii of brine shrimp ( Grant et al . , 1985 ) , mantis shrimp crystalline cones ( Bok et al . , 2014 ) , and sponges ( Bandaranayake et al . , 1997 ) . As genes responsible for the production of 4-deoxygadusol and MAAs are commonly found in bacteria ( e . g . , cyanobacteria ) , algae , and other marine microorganisms ( Shick and Dunlap , 2002; Miyamoto et al . , 2014 ) , the accumulation of these compounds in marine animals has been proposed to be of dietary or symbiont origin ( Arbeloa et al . , 2010; Gao and Garcia-Pichel , 2011; Loew , 2014 ) . On the other hand , a gene cluster like that in cyanobacteria apparently encoding a four-step DDGS-based pathway for converting SH7P to MAAs was discovered in the genomes of a coral ( Acropora digitifera ) and sea anemone ( Nematostella vectensis ) , suggesting that these invertebrates can produce MAAs autonomously ( Rosic and Dove , 2011; Shinzato et al . , 2011 ) . DDGS is a member of the sugar phosphate cyclase ( SPC ) superfamily ( Wu et al . , 2007 ) . In addition to DHQS , this superfamily includes four enzymes known for their roles in the biosynthesis of natural products with therapeutic application: 2-epi-5-epi-valiolone synthase ( EEVS ) , 2-epi-valiolone synthase , aminoDHQS , and 2-deoxy-scyllo-inosose synthase ( DOIS ) ( Wu et al . , 2007; Mahmud , 2009; Asamizu et al . , 2012; Kang et al . , 2012 ) . EEVS catalyzes the entry step to the biosynthesis of pseudosugar-containing natural products , such as the antidiabetic drug acarbose and the crop protectant validamycin A , and has so far only been identified and characterized in bacteria ( Mahmud , 2009 ) . Recently , we surprisingly also found genes that encode EEVS-like proteins ( annotated as ‘PREDICTED: pentafunctional AROM polypeptide-like’ ) in the genomes of fish , amphibians , reptiles , and birds ( Figure 1A , Figure 1—figure supplement 2 , Table 1 , Figure 1—source data 1 ) . This gene is located in a cluster with another functionally unknown gene ( labeled as MT-Ox hereafter ) and flanked by a suite of transcription factor genes encoding FRMD4B , MitF , MDFIC , and FoxP1 ( Figure 1B ) . These transcription factors have been known to regulate the expression of genes with essential roles in cell differentiation , proliferation , and survival ( Levy et al . , 2006; Rice et al . , 2012; Garner et al . , 2014; Zhao et al . , 2015 ) . However , whether they have a direct functional relationship with the EEVS and MT-Ox genes remains an open question . Whereas humans and other mammals also have FRMD4B , MitF , MDFIC , and FoxP1 homologs , they lack the EEVS-like and the MT-Ox genes ( Figure 1B ) . Therefore , the presence of the EEVS-like gene , previously only found in bacteria , in the genomes of fish , and other egg-laying vertebrates is puzzling and evolutionarily intriguing . Here , we report bioinformatics and functional studies of zebrafish ( Danio rerio ) EEVS-like and MT-Ox genes and the expression of this two-enzyme pathway in yeast , and discuss the evolutionary origin of this pathway in fish , amphibians , reptiles , and birds . 10 . 7554/eLife . 05919 . 003Figure 1 . Bioinformatic analysis of sugar phosphate cyclases in prokaryotes and eukaryotes and biochemical characterization of gadusol biosynthetic enzymes . ( A ) Bayesian phylogenetic tree of SPCs . Numbers represent posterior probability . The stramenopile Aureococcus anophagefferens , denoted by the blue star , has EEVS and MT-Ox proteins strikingly similar ( over 50% identical ) to those in vertebrates . The micro algae Coccomyxa subellipsoidea , denoted by the red star , also has EEVS and MT-Ox . ( B ) Genetic organizations of EEVS and MT-Ox genes in fish , amphibians , birds , and reptiles . Humans and other mammals lack these genes ( indicated in dashed red box ) . For a complete list of vertebrates whose genomes are known to contain EEVS and MT-Ox genes , see Table 1 . FRMD4B , FERM domain-containing protein 4B; MitF , microphtalmia-associated transcription factor; MDFIC , MyoD-family inhibitor domain-containing protein-like; and FoxP1 , Forkhead-related transcription factor 1 . ( C ) Biochemical characterization of recombinant LOC100003999 and zgc:113054 proteins . ( D ) WebLogo ( Crooks et al . , 2004 ) images of residue conservation patterns at the three metal ligand positions and two active site fingerprint sites known ( Kean et al . , 2014 ) to distinguish bacterial EEVSs from DDGSs . The residue numbers given correspond to the reference proteins D . rerio EEVS , ValA , and A . variabilis DDGS , respectively . WebLogos were based on 126 vertebrate , 63 bacterial EEVS , and 160 bacterial DDGS sequences , respectively , that had BLASTP E-values <10−120 in searches using the reference proteins noted above as queries . Each group was aligned using ProMals ( Pei and Grishin , 2007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00310 . 7554/eLife . 05919 . 004Figure 1—source data 1 . Amino acid sequences of the Sugar Phosphate Cyclases ( SPCs ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00410 . 7554/eLife . 05919 . 005Figure 1—figure supplement 1 . Biosynthetic pathway to shinorine in Anabaena variabilis ( Balskus and Walsh , 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00510 . 7554/eLife . 05919 . 006Figure 1—figure supplement 2 . Maximum likelihood phylogenetic tree of sugar phosphate cyclases . Numbers represent bootstrap confidence values . Blue star denotes the stramenopile A . anophagefferens , and red star denotes the micro algae C . subellipsoidea . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00610 . 7554/eLife . 05919 . 007Figure 1—figure supplement 3 . SDS PAGE of EEVS and MT-Ox proteins and thin-layer chromatography ( TLC ) analysis of their products . ( A ) SDS PAGE of purified recombinant ValA , LOC100003999 , and zgc:113054 , and E . coli cell free extracts containing the enzymes . ( B ) TLC analysis of enzymatic products of purified ValA and purified LOC100003999 . ( C ) TLC analysis of EEV and the zgc:113054 product ( gadusol ) using FeCl3 solution as a spraying agent . Only gadusol , but not EEV , can be detected as a maroon spot . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00710 . 7554/eLife . 05919 . 008Figure 1—figure supplement 4 . GC-MS analysis of LOC100003999 and ValA reaction products . ( A ) SH7P with E . coli cell free extracts containing LOC100003999 . ( B ) SH7P with purified ValA . ( C ) Authentic EEV . Samples were converted to their trimethylsilyl derivatives using Tri-Sil HTP ( Thermo Scientific ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00810 . 7554/eLife . 05919 . 009Figure 1—figure supplement 5 . Sequences of the vertebrate clade of EEVS-related proteins have characteristic features of an EEVS . Alignment of the sequence of the only structurally known EEVS–ValA from Streptomyces hygroscopicus ( ShEEVS; PDB entry 4P53 ) with representative related enzymes . The family consensus secondary structure elements shown schematically shown above the sequences and within each sequence the residues involved in secondary structures are color coded: β-strands ( yellow ) , α-helices ( teal ) , 310 helices ( blue ) , and π helices ( orange ) . ValA is shown first , and other sequences in order are DrEEVS ( Danio rerio EEVS ) , CsEEVS ( C . subellipsoidea EEVS ) , AaEEVS ( A . anophagefferens EEVS ) , AvDDGS ( A . variabilis DDGS; Ava_3858 ) , AmEVS ( Actinosynnema mirum EVS; Amir_2000 ) , PDB entry 1DQS ( Aspergillus nidulans DHQS ) , and PDB entry 2D2X ( Bacillus circulans DOIS ) . The 14 ‘fingerprint’ active site positions identified by Kean et al . ( Kean et al . , 2014 ) are indicated by an asterisk ( * ) below the sequences . Above the sequences are indicated a subset of those residues that ligate the catalytic metal ( m ) and another subset with notable variation between the family members that have different types of activity ( ↓ ) . The DrEEVS sequence and those of all known vertebrate homologs ( not shown ) match ShEEVS at all fourteen active site positions . The putative algal EEVS enzymes , AaEEVS and CsEEVS , sequences match the EEVS residues at the positions thought to be important for distinguishing the EEVS activity , but respectively differ at one and two positions; these positions are the ones that are of unknown importance for EEVS activity . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 00910 . 7554/eLife . 05919 . 010Figure 1—figure supplement 6 . Stereoview for the modeled active site geometry of LOC100003999 . Shown are the 13 active-site residues conserved in all EEVS enzymes ( pale purple carbons ) , the NAD+ ( gray carbons ) , and the Zn2+ atom ( silver sphere with black coordination bonds ) , along with a mesh that delineates the pocket suitable for binding an SH7P substrate . Residue numbers in LOC100003999 identifying twelve of the active site residues are shown; the third metal ligand , H366 , is behind the zinc and is not labeled . For the two positions that differ between EEVS and DDGS , the residue type present in DDGS is shown in green in parentheses . Figure made with Pymol ( Schrödinger L . The PyMOL Molecular Graphics System , Version 1 . 3 . ) using the coordinates of ValA , an EEVS from S . hygroscopicus subsp . jinggangensis 5008 ( Kean et al . , 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01010 . 7554/eLife . 05919 . 011Table 1 . Genetic analysis of EEVS and MT-Ox genes in some invertebrates and chordatesDOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 011ClassSpeciesCommon nameFRMD4BMitFEEVSMT-OxMDFICFoxP1CnidariaAcropora digitiferaCoral*––––––Nematostella vectensisStarlet sea anemone*√√–––√Hydra vulgarisCommon hydra–√–––√Invertebrate chordatesCiona intestinalisVase tunicate√√–––√Branchiostoma floridaeFlorida lancelet√√–––√AgnathaPetromyzon marinusLamprey†‡√‡‡‡√ChondrichthyesCallorhinchus miliiAustralian ghostshark†√√‡‡√√Osteichthyes ( Actinopterygii ) Astyanax mexicanusMexican tetra√√√√√√Danio rerioZebra fish√√√√√√Dicentrarchus labraxEuropean seabass√√√√√√Fundulus heteroclitusMummichog√√√√√√Gadus morhuaAtlantic cod√√√√√√Gasterosteus aculeatusThree-spined stickleback√√√√√√Lepisosteus oculatusSpotted gar√√√√√√Maylandia zebraZebra mbuna√√√√√√Neolamprologus brichardiAfrican cichlid√√√√√√Oncorhynchus mykissRainbow trout√√√√√√Orechromis niloticusNile tilapia√√√√√√Oryzias latipesJapanese rice fish√√√√√√Poecilia formosaAmazon molly√√√√√√Pundamilia nyerereiAfrican cichlid√√√√√√Salmo salarAtlantic salmon√√√√√√Takifugu rubripesJapanese puffer√√–––√Tetraodon nigroviridisGreen spotted puffer√√–––√Xiphophorus maculatusSouthern platyfish√√√√√√SarcopterygiiLatimeria chalumnaeWest African Coelacanth√√––√√AmphibiaXenopus tropicalisWestern clawed frog√√√√√√ReptiliaAlligator mississippiensisAmerican alligator√√√√√√Anolis carolinensisGreen anole√√√√√√Chelonia mydasGreen sea turtle√√√√√√Chrysemys picta belliiWestern painted turtle√√√√√√Ophiophagus hannahKing cobra†√√‡‡√√Pelodiscus sinensisChinese softshell turtle√√√√√√Python bivittatusBurmese python†√√‡‡√√AvesAnas platyrhynchosMallard√√√√√√Columba liviaRock dove√√√√√√Falco cherrugSaker falcon√√√√√√Ficedula albicollisCollard flycatcher√√√√√√Gallus gallusChicken√√√√√√Geospiza fortisMedium ground-finch√√√√√√Meleagris gallopavoNorth American wild turkey√√√√√√Melopsittacus undulatesCommon pet parakeet√√√√√√Pseudopodoces humilisTibetan ground-tit√√√√√√Taeniopygia guttataZebra finch√√√√‡√Zonotrichia albicollisWhite-throated sparrow√√√√√√MammaliaBalaenoptera acutorostrata scammoniMinke whale√√–––√Bos taurusCattle√√–––√Callithrix jacchusWhite-tufted-ear marmoset√√–––√Capra hircusGoat√√–––√Ceratotherium simum simumSouthern white rhinoceros√√–––√Chinchilla lanigeraLong-tailed chinchilla√√––√√Chlorocebus sabaeusGreen monkey√√––√√Cricetulus griseusChinese hamster√√––√√Eptesicus fuscusBig brown bat√√–––√Equus caballusHorse√√–––√Erinaceus europaeusWestern European hedgehog√√––√√Felis catusCat√√–––√Gorilla gorillaWestern gorilla√√–––√Heterocephalus glaberNaked mole-rat√√––√√Homo sapiensHuman√√–––√Lipotes vexilliferYangtze river dolphin√√–––√Loxodonta africanaAfrican savanna elephant√√–––√Macaca fascicularisCrab-eating macaque√√––√√Mesocricetus auratusGolden hamster√√––√√Monodelphis domesticaGray short-tailed opossum√√––√√Mus musculusHouse mouse√√––√√Mustela putorius furoDomestic Ferret√√–––√Myotis davidiiMouse-eared bat√√–––√Myotis lucifugusLittle-brown bat√√––√√Nomascus leucogenysNorthern white-cheeked gibbon√√–––√Odobenus rosmarus divergensPacific walrus√√––√√Orcinus orcaKiller whale√√–––√Ornithorhynchus anatinusDuck-billed platypus†√√–––√Orycteropus aferArdvark√√–––√Otolemur garnettiiSmall-eared galago√√–––√Ovis ariesSheep√√–––√Pan troglodytesChimpanzee√√–––√Papio anubisOlive baboon√√–––√Peromyscus maniculatus bairdiiPrairie deer mouse√√––√√Pteropus alectoBlack flying fox√√–––√Rattus norvegicusNorway rat√√–––√Sarcophilus harrisiiTasmanian devil√√––√√Trichechus manatus latirostrisFlorida manatee√√–––√Vicugna pacosAlpaca√√–––√*Harbors genes for MAA biosynthesis . †Incomplete genome sequence; no EEVS and MT-Ox genes identified . ‡The presence of EEVS or MT-Ox genes is unknown due to missing contigs or sequence information .
To investigate the function of the vertebrate EEVS-like genes , the protein encoded by the zebrafish EEVS-like gene ( LOC100003999 ) was expressed in Escherichia coli . Incubation of the recombinant protein with SH7P gave a product , which was confirmed by thin-layer chromatography ( TLC ) and GC-MS to be EEV ( Figure 1C , Figure 1—figure supplements 3 , 4 ) , identifying the protein as an EEVS . The best-characterized bacterial EEVS is ValA from the validamycin pathway in Streptomyces hygroscopicus subsp . jinggangensis 5008 ( Bai et al . , 2006 ) , and the crystal structure of ValA ( Protein Data Bank [PDB] entry 4P53 ) ( Kean et al . , 2014 ) , allowed identification of a fingerprint set of 14 active-site residues with characteristic variations that could differentiate the various SH7P cyclases . Further supporting the assignment of the LOC100003999-encoded protein as an EEVS , sequence comparisons show that all animal EEVS-like proteins are highly similar ( 60–72% identity ) and also match the sequence of ValA at all 14 fingerprint sites ( Figure 1D , Figure 1—figure supplements 5 , 6 ) . This firmly establishes the presence of EEVS activity in animals . The second gene , MT-Ox ( zgc:113054 ) , is predicted to encode a protein that contains two domains: the N-terminal domain is similar to S-adenosylmethionine ( SAM ) –dependent methyltransferases , and the C-terminal domain is similar to NAD+-dependent oxidoreductases . Although its function in zebrafish is unknown , the transcription of this gene in larvae is upregulated by light , leading to a prediction of its involvement in circadian clock regulation ( Weger et al . , 2011 ) . In contrast , we hypothesized that this bifunctional protein is involved in modifying EEV to yield an oxidized and methylated product ( Figure 1C ) . To test this hypothesis , recombinant MT-Ox protein encoded by zgc:113054 was incubated with EEV in the presence of SAM and NAD+ . Following incubation , a product with λmax of 294 nm ( pH 7 ) and 270 ( pH 2 . 5 ) was detected ( Figure 2A–B ) . Further analysis of the product by ( − ) -ESI-MS ( m/z 203 [M-H]− ) and 1H NMR confirmed its identity as gadusol ( Figure 2—figure supplements 1 , 2 ) . We postulate that the conversion of EEV to gadusol by MT-Ox takes place via oxidation of the C-2 or C-3 OH , followed by enolization and methylation of the resulting C-2 OH ( Figure 2—figure supplement 3 ) . This new pathway to the UV-absorbing vinylogous acid functional group shared by gadusol and 4-deoxygadusol is distinct from that used in the biosynthesis of the MAAs ( Balskus and Walsh , 2010; Spence et al . , 2012 ) , and the existence of multiple pathways for generating this scaffold is intriguing and may indicate that this is a privileged chemical scaffold in living organisms . 10 . 7554/eLife . 05919 . 012Figure 2 . Production of gadusol in zebrafish and yeast and its sunscreen activity . ( A–B ) HPLC traces and UV absorptions of gadusol produced from Escherichia coli cell-free extract containing EEVS and purified MT-Ox protein at pH 7 . 0 and 2 . 5 . ( C–D ) Transcription patterns of EEVS and MT-Ox genes during zebrafish embryonic development . qRT-PCR analysis of mRNA isolated from zebrafish embryos ( n = 3 ) at 12 , 24 , 48 , 72 , 96 , and 120 hpf . ( E ) Time course of gadusol production in yeast harboring the zebrafish genes . The yeast was cultured in YNB + 2% glucose supplemented with leucine and lysine at 30°C for 2 days , and growth was monitored as A600 values ( control , dotted blue line; gadusol producer , solid red line ) . Gadusol concentration in the supernatant of 20 ml cultures ( n = 3 ) was monitored as A296 values in 50 mM phosphate buffer , pH 7 . 0 ( dashed green line ) corrected for non-gadusol background absorbance in the control supernatant , normalized to A600 value . Gadusol was quantified based on an extinction coefficient of 21 , 800 M−1 cm−1 in 50 mM phosphate buffer , pH 7 . ( F–H ) Comparative HPLC analysis of gadusol from recombinant enzymatic reaction , zebrafish extract , and yeast extract . ( I ) Gadusol suppresses the UVB sensitivity of a rad1∆ yeast mutant; and ( J ) Gadusol increases the UVB tolerance of a wild-type ( RAD1 ) strain . Cells suspended in control supernatant ( CS ) or gadusol+ supernatant ( G+S ) were irradiated with UVB and subsequently spotted in 3 µl aliquots ( n = 4 ) onto YEPD plates , which were incubated at 30°C for 24 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01210 . 7554/eLife . 05919 . 013Figure 2—figure supplement 1 . ( − ) -ESI-MS analysis of LOC100003999 and zgc:113054 reaction products . ( A ) SH7P incubated with E . coli cell free extracts containing LOC100003999 ( zebrafish EEVS ) and boiled zgc:113054 ( MT-Ox ) . ( B ) SH7P incubated with E . coli cell free extracts containing EEVS and purified MT-Ox . ( C ) SH7P incubated with purified ValA ( a bacterial EEVS ) and boiled MT-Ox . ( D ) Synthetic EEV incubated with purified MT-Ox . ( E ) Deuterated EEV incubated with purified MT-Ox . ( F ) MS/MS spectrum of gadusol obtained from the EEVS and MT-Ox reaction . ( G ) Proposed ion species of gadusol observed in F . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01310 . 7554/eLife . 05919 . 014Figure 2—figure supplement 2 . 1H NMR spectrum of gadusol obtained from E . coli cell free extracts containing LOC100003999 and zgc:113054 reactions . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01410 . 7554/eLife . 05919 . 015Figure 2—figure supplement 3 . Proposed mode of formation of gadusol in vertebrates . The dotted arrows show the three steps thought to be a part of the MT-Ox catalyzed reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01510 . 7554/eLife . 05919 . 016Figure 2—figure supplement 4 . The Pentose Phosphate Pathway ( Asamizu et al . , 2012 ) and the shunt pathway to gadusol . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 016 In zebrafish , both of the LOC100003999 and zgc:113054 genes are expressed during embryonic development . qRT-PCR analysis of mRNA isolated from zebrafish embryos at 12 , 24 , 48 , 72 , 96 , and 120 hpf showed maximal expression at 72 hpf ( Figure 2C–D ) . To demonstrate de novo synthesis of gadusol in zebrafish , the embryos were collected at 72 hpf , lyophilized and extracted with methanol , and the extract was analyzed by HPLC and ESI-MS ( Figure 2G ) . Our finding of gadusol in the extract unambiguously confirms the ability of zebrafish to synthesize gadusol and amends the current perception that gadusol found in fish and other vertebrates is necessarily of dietary or symbiont origin . However , as MAAs are synthesized via a different pathway and there is no evidence that fish have those biosynthetic enzymes , the accumulation of MAAs in fish would still appear to be of dietary origin ( Mason et al . , 1998; Zamzow , 2004 ) . To show that the recombinant LOC100003999 and zgc:113054 genes are sufficient for encoding gadusol synthesis , they were cloned into a yeast expression vector and transferred into a Saccharomyces cerevisiae strain , in which the transaldolase gene TAL1 had been deleted . Yeast possesses a robust pentose–phosphate pathway ( Figure 2—figure supplement 4 ) , and by removing the transaldolase enzyme , which normally metabolizes SH7P , and adding EEVS and MT-Ox , we expected to facilitate an effective shunt pathway from SH7P to gadusol . Analysis of the culture broth by HPLC , ESI-MS , and UV spectrophotometry revealed the presence of gadusol ( Figure 2H ) and its accumulation to ∼20 mg/l after 5 days ( Figure 2E ) . The results not only demonstrate the ability of the engineered yeast to produce and secrete gadusol but also present a new avenue for large-scale production of the compound for possible commercial uses , for example , sunscreen and/or antioxidant ( Plack et al . , 1981; Schmid et al . , 2006; Cardozo et al . , 2007; Arbeloa et al . , 2010 ) . To test the UV-protective activity of gadusol , a yeast rad1∆ mutant , which is sensitive to UVB , was suspended at approximately 107 cells/mL in the concentrated supernatant from the gadusol-producing yeast strain or from an otherwise isogenic control strain that did not produce gadusol . The gadusol-containing supernatant suppressed the UVB-sensitivity of the rad1∆ mutant ( Figure 2I ) , confirming the UVB-protective activity of gadusol . Analogous experiments with a wild-type strain ( RAD1 ) at higher doses of UVB showed comparable results ( Figure 2J ) , consistent with UVB protective activity . As noted above , the SPCs EEVS , EVS , DDGS , aminoDHQS , and DOIS are all related to DHQS and are widespread in bacteria and fungi , but other than this report , are not known to exist in vertebrates or prevertebrates . We suggest that the vertebrate EEVS and MT-Ox genes were most plausibly acquired via horizontal gene transfer . Interestingly , searches identify the stramenopile Aureococcus anophagefferens and the microalgae Coccomyxa subellipsoidea , as the only non-vertebrate organisms in current databases that harbor a similar bifunctional MT-Ox gene , and both organisms have a predicted EEVS gene adjacent to that of MT-Ox . As algae are known to be active horizontal gene transfer agents ( Ni et al . , 2012 ) , algae such as these become a plausible place both for the development of this alternate pathway for gadusol production and as a source of the genes found in vertebrates . Further supporting such a relationship , the A . anophagefferens EEVS protein is substantially more similar to the vertebrate EEVSs than it is to bacterial EEVSs ( Figure 1A and Figure 1—figure supplement 2 , denoted by the blue star ) . Further bioinformatics studies also showed that the tunicates and lancelets lack the EEVS and MT-Ox genes , suggesting that the gene transfer occurred sometime during the evolution of primitive chordates to bony fishes ( Figure 3 ) . While the EEVS and MT-Ox genes are retained in modern ray-finned fish ( with the exception of puffer fish ) , as well as in amphibians , reptiles , and birds , they were lost in mammals , including the egg-laying mammal platypus , indicating the lack of a direct link between gadusol and the mode of reproduction . The West African coelacanth genome ( Amemiya et al . , 2013 ) also appears to lack the EEVS and MT-Ox genes ( Figure 3 , Table 1 ) . This rare ovoviviparous fish lives in caves 100–500 meters deep and feeds at night , and its lack of gadusol production ability may be directly related to its limited exposure to UV light and/or oxidative stress . Other than the well-documented presence of gadusol in fish eggs , where it can serve to protect the roe from UV damage ( Arbeloa et al . , 2010; Colleter et al . , 2014 ) , nothing is known of its role ( s ) in fish , reptiles , amphibians , and birds . Exploring its function in these organisms will increase our understanding of their physiology and ecology . 10 . 7554/eLife . 05919 . 017Figure 3 . Model timeline for the evolution of the EEVS and MT-Ox genes in vertebrates . The genes entered an early vertebrate genome as a linked pair ( vertical blue arrow ) and were retained in the modern ray-finned fish , amphibians , reptiles , and birds as indicated by thick dark cyan arrows . Coelacanths and mammals lost the genes ( thick red arrows ) . No full genome sequence is available for assessing the presence of EEVS and MT-Ox in lungfish . The phylogenetic trees of the EEVS and MT-Ox proteins or mRNA from a selected set of vertebrates can be found in Figure 3—figure supplements 1–3 and Supplementary files 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01710 . 7554/eLife . 05919 . 018Figure 3—figure supplement 1 . Phylogenetic tree of the EEVS proteins from a selected set of vertebrates . The distantly related E . coli protein CglD was used as an out-group . Numbers represent bootstrap confidence values . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01810 . 7554/eLife . 05919 . 019Figure 3—figure supplement 2 . Phylogenetic tree of the MT-Ox proteins from a selected set of vertebrates . E . coli K12 dam was used as an out-group . Numbers represent bootstrap confidence values . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 01910 . 7554/eLife . 05919 . 020Figure 3—figure supplement 3 . Maximum likelihood tree of vertebrate EEVS mRNA sequences . E . coli cglD was used as an out-group . Numbers represent bootstrap confidence values . DOI: http://dx . doi . org/10 . 7554/eLife . 05919 . 020
For phylogenetic analysis , full-length amino acid sequences and vertebrate mRNA sequences were analyzed . A reciprocal BLAST hit analysis was performed with the EEVS protein ( see Supplementary file 3 ) . Sequences were aligned using MUSCLE . ProtTest was used to determine the best model of protein evolution ( LG+G ) ( Darriba et al . , 2011 ) , and MEGA6 was used to determine the best fit nucleic acid evolutionary model ( K80+G ) ( Tamura et al . , 2013 ) . RAxML was used for maximum likelihood analysis , and the robustness of the trees was assessed by bootstrap analysis ( 1000 replicates ) ( Stamatakis , 2014 ) . Bayesian analysis was performed by MrBayes ( version 3 . 2 . 3 ) , using a random starting tree , running eight chains for 4 , 000 , 000 generations , sampling every 250 trees ( Ronquist et al . , 2012 ) . The first 5000 trees were discarded as the burnin , with the remaining trees used to calculate posterior probability . RAxML and MrBayes were run on the CIPRES science gateway ( Miller et al . , 2010 ) . MEGA6 was used for maximum likelihood analysis of vertebrate mRNA sequences with tree robustness assessed by bootstrap ( 500 replicates ) . Sources of proteins for the analyses are listed in Supplementary file 1 . The LOC100003999 gene was codon optimized for E . coli and synthesized commercially ( GeneScript USA Inc . , Piscataway , NJ ) . The product was cloned into EcoRV site of pUC57-kan vector . The plasmid was digested with BglII and EcoRI and ligated into BamHI and EcoRI site of pRSET-B ( Life Technologies , Carlsbad , CA ) for the expression of N-terminal hexa-histidine-tagged protein . The zgc:113054 gene was also codon optimized for E . coli and commercially synthesized ( GeneScript USA Inc . ) . The product was cloned into EcoRV site of pUC57-amp vector . The plasmid was digested with BglII and EcoRI and ligated into BamHI and EcoRI site of pRSET-B ( Life Technologies ) for the expression of N-terminal hexa-histidine-tagged protein . pRSETB-valA , pRSETB-LOC100003999 , and pRSETB-zgc:113054 plasmids were individually used to transform E . coli BL21 GOLD ( DE3 ) pLysS . Transformants were grown overnight at 37°C on LB agar plate containing ampicillin ( 100 μg/ml ) and chloramphenicol ( 25 μg/ml ) . A single colony was inoculated into LB medium ( 2 ml ) containing the above antibiotics and cultured at 37°C for 8 hr . The seed culture ( 1 ml ) was transferred into LB medium ( 100 ml ) in a 500-ml flask and grown at 30°C until OD600 reached 0 . 6 . Then , the temperature was reduced to 18°C . After 1-hr adaptation , isopropyl-D-1-thiogalactopyranoside ( IPTG ) ( 0 . 1 mM ) was added to induce the N-terminal hexa-histidine-tagged proteins . After further growth for 16 hr , the cells were harvested by centrifugation ( 5000 rpm , 10 min , 4°C ) , washed twice with cold water , and stored at −80°C until used . Cell pellets from a 400-ml culture of E . coli BL21 GOLD ( DE3 ) pLysS containing pRSETB-valA , pRSETB-LOC100003999 , or pRSETB-zgc:113054 plasmids were resuspended in 20 ml of B buffer ( 40 mM Tris-HCl , 300 mM NaCl , 10 mM imidazole , pH 7 . 5 ) . Cells were disrupted by sonication for 1 min ( 4 times , 2 min interval ) at 13 watts on ice ( Probe sonicator , Misonix , Farmingdale , NY ) . 20 ml of lysate was divided into 2-ml tubes and centrifuged ( 14 , 500 rpm , 20 min , 4°C ) . Soluble fractions were collected and transferred into a 50-ml tube . Ni-NTA ( QIAGEN , Valencia , CA ) resin ( 5 ml ) was applied into 10-ml vol empty column , and the Ni-NTA resin was equilibrated with B buffer ( 50 ml , 10 CV ) . About 20 ml of supernatant from cell lysate was applied to the column ( flow rate; 0 . 8 ml/min ) . The column was then washed with 100 ml ( 20 CV ) of W buffer ( 40 mM Tris-HCl , 300 mM NaCl , 20 mM imidazole , pH 7 . 5 ) at 0 . 8 ml/min . The hexa-histidine-tagged proteins were eluted by imidazole addition using a gradient mixer containing 100 ml of W buffer and 100 ml of E buffer ( 40 mM Tris-HCl , 300 mM NaCl , 300 mM imidazole , pH 7 . 5 ) . The fractions ( 150 drops or about 5 ml ) were collected and checked by SDS-PAGE ( Coomassie Blue staining ) . Fractions containing pure proteins were combined ( 25 ml ) and dialyzed against 2 l of D buffer ( 10 mM Tris-HCl , pH 7 . 5 ) 3 times ( every 3 hr ) . Dialyzed protein solution was concentrated by ultrafiltration ( MWCO 10 K ) to 200 μM and flash frozen in liquid N2 prior to storage at −80°C . The yields of the purified proteins were 57 mg/l for ValA , 18 mg/l for LOC100003999 , and 79 mg/l for zgc:113054 . Each reaction mixture ( 25 μl ) contained Tris-HCl buffer ( 20 mM , pH 7 . 5 ) , NAD+ ( 1 mM ) , CoCl2 , or ZnSO4 ( 0 . 1 mM ) , SH7P ( 4 mM ) , and purified enzymes ( 0 . 12 mM ) . The mixture was incubated at 30°C for 2 hr . ValA ( instead of LOC100003999 ) was used as a positive control . No enzyme ( buffer only ) was used as a negative control . Each reaction mixture ( 50 μl ) contained potassium phosphate buffer ( 10 mM , pH 7 . 4 ) , NAD+ ( 2 mM ) , CoCl2 ( 0 . 2 mM ) , SH7P ( 4 mM ) , and LOC100003999 cell-free extract ( 20 μl ) was incubated at 30°C . After 6 hr , S-adenosylmethionine ( 5 mM ) and purified zgc:113054 ( 0 . 1 mM ) were added . The mixture was incubated at 30°C for another 6 hr . ValA was used ( instead of LOC100003999 ) as a positive control . Extract of E . coli harboring pRSET B empty vector was used as a negative control . A reaction mixture ( 25 μl ) containing potassium phosphate buffer ( 10 mM , pH 7 . 4 ) , NAD+ ( 2 mM ) , CoCl2 ( 0 . 2 mM ) , S-adenosylmethionine ( 5 mM ) , [6 , 6-2H2]-EEV ( 4 mM ) , and purified zgc:113054 ( 0 . 1 mM ) was incubated at 30°C for 2 hr . Boiled zgc:113054 was used as a negative control . Analytical TLC was performed using silica gel plates ( 60 Å ) with a fluorescent indicator ( 254 nm ) , which were visualized with a UV lamp and ceric ammonium molybdate ( CAM ) or 5% FeCl3 in MeOH-H2O ( 1:1 ) solutions . The enzymatic reaction mixtures were lyophilized , and the products were extracted with MeOH . The MeOH extract was then dried and Tri-Sil HTP ( Thermo Scientific , Waltham , MA ) ( 100 μl ) was added and left to stand for 20 min . The solvent was removed in a flow of argon gas , and the silylated products were extracted with hexanes ( 100 μl ) and injected into the GC-MS ( Hewlett Packard 5890 SERIES II Gas chromatograph ) . Fifty eppendorf tubes containing reaction mixtures ( 100 μl each ) , which consist of potassium phosphate buffer ( 10 mM , pH 7 . 4 ) , SH7P ( 5 mM ) , NAD+ ( 2 mM ) , CoCl2 ( 0 . 2 mM ) , and LOC100003999 cell-free extract ( 40 μl ) , were incubated at 30°C . After 6 hr , S-adenosylmethionine ( 5 . 5 mM ) and zgc:113054 cell-free extracts ( 30 μl ) were added . The reaction mixtures were incubated at 30°C for another 6 hr . The reaction mixtures were quenched with 2 vol of MeOH , left to stand at −20°C for 20 min , then centrifuged at 14 , 500 rpm for 20 min . The supernatants were pooled and dried under vacuum . The residual water was frozen and lyophilized . The crude sample was dissolved in water ( 1 ml ) and subjected to Sephadex LH-20 column chromatography using phosphate buffer ( 2 . 5 mM , pH 7 ) as an eluant . Fractions containing the product as judged by MS were combined and lyophilized . Furthermore , the product was purified by HPLC ( Shimadzu LC-20AD , C18 column [YMC] , 250 × 10 mm , 4 μm , flow rate 1 ml/min ) . Solvent system: MeOH—phosphate buffer ( 5 mM , pH 7 ) , gradient 1–100% of MeOH ( 0–40 min ) . Peak at 12 . 74 min was collected and dried to give gadusol ( 0 . 4 mg ) . 1H NMR ( 700 MHz , D2O , cryo-probe ) : δ 4 . 10 ( s , 1H , H-4 ) , 3 . 71 ( d , J = 12 Hz , H-7α ) , 3 . 56 ( d , J = 12 Hz , H-7β ) , 3 . 49 ( s , 3H , OCH3 ) , 2 . 68 ( d , J = 17 Hz , H-6α ) , 2 . 38 ( d , J = 17 Hz , H-6β ) . HR-MS ( ESI-TOF ) ( m/z ) : ( M+H ) + calculated for C8H13O6 , 205 . 0707; found , 205 . 0709 . Adult wild-type 5D zebrafish were housed at the Sinnhuber Aquatic Research Laboratory on a recirculating system maintained at 28 ± 1°C with a 14 hr light per 10 hr dark schedule . Embryos were collected from group spawns of adult zebrafish as described previously ( Reimers et al . , 2006 ) , and all experiments were conducted with fertilized embryos according to Oregon State University Institutional Animal Care and Use Protocols . Embryos were staged accordingly as previously described ( Kimmel et al . , 1995 ) and collected by hand for all experiments . Embryos were reared in media consisting of 15 mM NaCl , 0 . 5 mM KCl , 1 mM MgSO4 , 0 . 15 mM KH2PO4 , 0 . 05 mM Na2HPO4 , and 0 . 7 mM NaHCO3 ( Westerfield , 2000 ) . All polymerase chain reaction ( PCR ) reactions were performed according to manufacturer's specifications . Cycling conditions: 96°C for 3 min , 95°C for 1 min , 65°C for 1 min , and 72°C for 1 min per kB DNA; 35 cycles were used followed by 10 min at 72°C . All PCR products were characterized on an agarose gel . If needed , the PCR product was excised from the gel and purified using the E . Z . N . A . Gel Extraction Kit ( Omega Bio-tek , Norcross , GA ) . qPCR was performed on a Applied Biosystems StepOnePlus machine . The super mix PerfeCTa SYBR Green FastMix , ROX ( Quanta biosciences , Gaithersburg , MD ) was used . cDNA ( 100 ng ) from time points at 6 , 12 , 24 , 48 , 72 , 96 , and 120 hpf was used . Super mix ( 18 µl ) was added to bring the final volume to 20 µl . PCR conditions suggested by the supplier were used . For total RNA isolation , 30 embryos were homogenized in RNAzol ( Molecular Research Center , Cincinnati , OH ) ; RNA was purified according to the manufacturer's protocol . RNA was quantified by A260/280 ratios measured using a SynergyMx microplate reader ( Biotek , Winooski , VT ) and analyzed with the Gen5 Take3 module . 1 µg of RNA was used for cDNA synthesis . Superscript III First-Strand Synthesis ( Life Technologies ) and oligo d ( T ) primers were used to synthesize cDNA from the total RNA . Embryos were collected and euthanized at 72 hpf by induced hypoxia through rapid chilling on ice for 30 min . Embryo media were removed until about 5 ml were left and frozen at −80°C . Embryos were lyophilized overnight . The freeze-dried embryos were then ground with a pestle and mortar under liquid nitrogen . The powder was collected and placed in a pre-weighed glass vial . The mortar was washed with MeOH-H2O ( 80:20 ) , and the solvent was added to the powder . The solvent was evaporated , and powder was weighed . The embryo powder was extracted twice with MeOH-H2O ( 80:20 ) . The two extracts were combined , dried , and weighed . The extract was suspended in MeOH-H2O ( 80:20 ) ( 1 ml ) and extracted twice with hexanes . The aqueous layer was recovered , dried , and weighed . The extract was suspended in MeOH for analysis by mass spectrometry . The extract was dissolved in phosphate buffer pH 7 . 0 for identification by HPLC ( Shimadzu SPD-20A system , YMC ODS-A column ( 4 . 6 id × 250 mm ) , MeOH—5 mM phosphate buffer ( 1% MeOH for 20 min followed by a gradient from 1 to 95% MeOH in 20 min ) , flow rate 0 . 3 ml/min , 296 nm . The isolated gadusol was analyzed by MS ( ThermoFinnigan LCQ Advantage system ) and NMR ( in D2O; Bruker Unity 300 [300 . 15 MHz] spectrometer ) . The yeast strains used are listed in Supplementary file 4 . The TRP1 gene was replaced in BY4742 tal1∆::KanMX4 with a wild-type URA3 allele from S288c by standard methods ( Baudin et al . , 1993 ) . The deletion was confirmed by PCR using primer pairs TRP1DisUP/TRP1DisLO and URA3DisUP/TRP1DisLO . The BY4742 tal1∆::KanMX4 trp1∆::URA3 strain was then co-transformed ( Gietz et al . , 1992 ) with pXP416 and pXP420 to generate an empty vector control strain and with pXP420-EEVS and pXP416-MT-Ox to generate a gadusol-producing strain . The RAD1 gene was replaced in BY4742 tal1∆::KanMX4 trp1∆::URA3 with a wild-type LEU2 allele from S288c by standard methods ( Reynolds et al . , 1987 ) . The deletion was confirmed by PCR using primer pairs RAD1UP/RAD1LO . The resultant BY4742 tal1∆::KanMX4 trp1∆::URA3 rad1Δ::LEU2 strain was then co-transformed with pXP416 and pXP420 . Cells were pre-grown in YEPD ( 1% yeast extract , 2% peptone , and 2% glucose ) for transformations , and in YNB ( Bacto yeast nitrogen base without amino acids ) + 2% glucose supplemented with 30 µg/ml leucine and 30 µg/ml lysine to select for transformants and to produce gadusol . Liquid media were sterilized by filtration using a 0 . 45-µm filter , and agar-based media were sterilized by autoclaving . Liquid cultures were grown at 30°C for 48 hr and 200 rpm; plates were incubated at 30°C . Plasmids are listed in Supplementary file 5 . Primers used for PCR are listed in Supplementary file 6 . PCR amplicons with SpeI and XhoI terminal restriction sites were generated for the EEVS gene and MT-Ox gene using pRSETB-EEVS and pRSETB-MTOx as templates , respectively . The EEVS and MT-Ox amplicons were then digested with SpeI and XhoI and ligated into SpeI-digested pXP420 and XhoI-digested pXP420 and pXP416 , respectively , and introduced into competent E . coli ( Top 10; Life Technologies ) by transformation . E . coli transformants were selected on LB plates supplemented with ampicillin ( 100 µg/ml ) . Transformants were then screened by digesting plasmid DNA with SpeI and XhoI restriction enzymes and analyzing fragments by agarose gel electrophoresis . S . cerevisiae cell pellets from 5 ml cultures were extracted with MeOH , and the supernatant was extracted with nBuOH . Extracts were concentrated and analyzed by HPLC ( Shimadzu SPD-20A system , YMC ODS-A column [4 . 6 id × 250 mm] , MeOH—5 mM phosphate buffer ( 1% MeOH for 20 min followed by a gradient from 1 to 95% MeOH in 20 min ) , flow rate 0 . 3 ml/min , 296 nm . A rad1∆ mutant ( MATα his3∆1 leu2∆0 lys2∆0 trp1∆::URA3 ura3∆0 rad1∆::LEU2 tal1∆::KanMX4/pXP416 , pXP420 ) or wild-type RAD1 strain ( S288c , MATα SUC2 gal2 mal2 mel flo1 flo8-1 hap1 ho bio1 bio6 ) was grown at 30°C and 200 rpm in YNB + 2% glucose + 30 µg/ml leu + 30 µg/ml lys . Cells were harvested after 24 hr by centrifugation , washed twice in the ninefold concentrated supernatant of either the gadusol-producing strain BY4742 tal1∆ trp1∆/pXP416-MTOx , pXP420-EEVS or of the control strain BY4742 tal1∆ trp1∆/pXP416 , pXP420 , and suspended in the respective concentrated supernatants at 107 cells/ml . Cells ( 375 µl ) were irradiated with UVB ( 302 nm ) at the indicated doses in wells of a 24-well microtiter plate shaken at 900 rpm . 3 µl aliquots of cells were then spotted onto a YEPD plate , which was incubated 24 hr at 30°C prior to being photographed . The supernatants of the gadusol producing and control strains were obtained by centrifugation following 5 day of growth in YNB + 2% glucose + 30 µg/ml leucine + 30 µg/ml lysine at 30°C and 200 rpm . Supernatants were freeze-dried , dissolved in a volume of distilled water 1/10 of the initial culture volume , and stored at 4°C until use . Just prior to suspension of cells , the concentrated supernatant was adjusted to 50 mM phosphate , pH 7 . 0 resulting in a final ninefold concentrate . | Sunlight is the Earth's primary energy source and is exploited by an array of natural and man-made processes . Photosynthetic plants harness solar energy to convert carbon dioxide and water into biomass , and solar panels capture light and convert it to electricity . Sunlight is critical to life on Earth , and yet excessive exposure to sunlight can cause serious harm as it contains ultraviolet ( UV ) radiation , which damages the DNA of cells . In humans , this damage can lead to conditions such as cataracts and skin cancer . The marine organisms and animals that live in the upper ocean and on reefs are subject to intense and unrelenting sunlight . In their effort to protect against potentially deadly UV radiation , many small and particularly vulnerable marine organisms , such as bacteria and algae , produce UV-protective sunscreens . While UV-protective compounds have also been found in larger organisms , including fish and their eggs , the presence of these sunscreens has always been attributed to the animal sequestering the compounds from their environment or partnering with a sunscreen-producing microorganism . Now , Osborn , Almabruk , Holzwarth et al . have discovered a fish that is able to produce such a UV-protective compound completely on its own . After identifying the full set of genes—or pathway—responsible for generating the UV-protective compound , the same pathway was detected in a variety of diverse animals , including amphibians , reptiles , and birds . This opens up a new area of study , because besides providing UV protection , no one yet knows what other roles the molecule may have in these animals . Furthermore , introducing the complete pathway into yeast enabled these cells to produce the sunscreen . In the future , engineering a yeast population to produce large quantities of the natural sunscreen could lead to large-scale production of the UV-protective compound so it can be used in pharmaceuticals and cosmetics . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"short",
"report",
"biochemistry",
"and",
"chemical",
"biology"
] | 2015 | De novo synthesis of a sunscreen compound in vertebrates |
Heterogeneity of lymphatic vessels during embryogenesis is critical for organ-specific lymphatic function . Little is known about lymphatics in the developing kidney , despite their established roles in pathology of the mature organ . We performed three-dimensional imaging to characterize lymphatic vessel formation in the mammalian embryonic kidney at single-cell resolution . In mouse , we visually and quantitatively assessed the development of kidney lymphatic vessels , remodeling from a ring-like anastomosis under the nascent renal pelvis; a site of VEGF-C expression , to form a patent vascular plexus . We identified a heterogenous population of lymphatic endothelial cell clusters in mouse and human embryonic kidneys . Exogenous VEGF-C expanded the lymphatic population in explanted mouse embryonic kidneys . Finally , we characterized complex kidney lymphatic abnormalities in a genetic mouse model of polycystic kidney disease . Our study provides novel insights into the development of kidney lymphatic vasculature; a system which likely has fundamental roles in renal development , physiology and disease .
The lymphatic vasculature is a blind-ended network of vessels that is essential for tissue fluid balance , absorbance of dietary lipids and surveillance of the immune system in vertebrate homeostasis and has been implicated in cardiovascular disease ( Vieira et al . , 2018 ) , neurodegenerative disorders ( Da Mesquita et al . , 2018 ) , autoimmunity ( Louveau et al . , 2018 ) and cancer metastasis ( Le et al . , 2016 ) . While the formation of systemic lymphatic vasculature , emerging from venous endothelium at around embryonic day ( E ) 10 in mouse , has received considerable attention , how organ-specific lymphatics arise has only recently emerged as a focus of study , including the characterization of lymphatic development in the mouse mesentery ( Stanczuk et al . , 2015 ) , intestine ( Mahadevan et al . , 2014 ) , dermis ( Pichol-Thievend et al . , 2018; Martinez-Corral et al . , 2015 ) , meninges ( Antila et al . , 2017 ) and heart ( Klotz et al . , 2015; Stone and Stainier , 2019; Maruyama et al . , 2019; Gancz et al . , 2019 ) . In contrast , how lymphatic vessels develop within the mouse or human kidney is not well established , despite their functional importance in pathological processes within the mature organ , including renal fibrosis and transplant rejection ( Sakamoto et al . , 2009; Kerjaschki et al . , 2004; Pei et al . , 2019 ) . The development of the metanephros , the direct precursor of the mature kidney , commences at around E10 . 5 in mouse and five post-conceptional weeks ( PCW ) in humans . Kidney development is a complex process , orchestrated by interactions between epithelial , stromal , immune and blood vascular endothelial cells ( McMahon , 2016 ) . How lymphatic vessels fit within this process is unclear . Our current understanding of kidney lymphatic development is limited to two descriptive studies; a paucity of insight which reflects their apparent scarcity ( Petrova and Koh , 2018 ) and generally deep location within the kidney . Both prior studies draw conclusions based on staining tissue sections of rodent embryonic kidney with single lymphatic markers ( Lee et al . , 2011; Tanabe et al . , 2012 ) . However , the recent description of hybrid renal vessels with both blood vascular and lymphatic properties ( Huang et al . , 2016; Kenig-Kozlovsky et al . , 2018 ) and non-endothelial renal cell types expressing lymphatic markers ( Hochane et al . , 2019; Kim et al . , 2015 ) necessitate the simultaneous use of multiple markers to reliably identify lymphatic endothelium in the kidney . Here , we have characterized the spatiotemporal development of the kidney lymphatic vasculature using a combination of wholemount immunofluorescence , optical clearing and high-resolution confocal microscopy to obtain three-dimensional ( 3D ) images of intact mouse and human fetal kidneys at single-cell resolution . Using this approach , we provide novel insights into lymphatic vessel formation in mammalian kidney development , including a first description of the forming lymphatics in human fetal kidneys . Further , we identified lymphatic anomalies in a mouse model of polycystic kidney disease ( PKD ) , the most common genetic cause of kidney failure .
To observe lymphatic vessels during kidney development , we isolated metanephroi from CD-1 outbred wildtype mouse embryos at a range of developmental stages , and wholemount immunolabelled them for the early lymphatic endothelial cell markers ( Banerji et al . , 1999; Wigle and Oliver , 1999 ) prospero homeobox protein 1 ( PROX1 ) and lymphatic vessel endothelial hyaluronan receptor 1 ( LYVE-1 ) . We then optically cleared and imaged entire labelled kidneys using confocal microscopy . 3D reconstructions and z-stacks of confocal images revealed a cellular network of PROX1+/LYVE-1+ lymphatic endothelial cells in the E14 . 5 mouse kidney ( Figure 1A ) . At this stage in development , the ureteric bud , has already undergone 8–9 generations of branching ( Short et al . , 2014 ) and the kidney contains a rich , perfused blood vasculature ( Loughna et al . , 1997; Munro et al . , 2017; Rymer et al . , 2014 ) . No definitive PROX1+/LYVE-1+ vessels were observed in the kidney prior to E14 . 5 ( data not shown ) . By E16 . 5 , the kidney lymphatics constituted a vascular plexus , that contained some lumenized vessel segments ( Figure 1—video 1 ) . By E18 . 5 , lymphatic vessel branches were observed distally in the kidney . We confirmed the lymphatic identity of the plexus ( Figure 1—figure supplement 1 ) by its prominent expression of vascular endothelial growth factor receptor 3 ( VEGFR-3 ) and podoplanin ( Breiteneder-Geleff et al . , 1997; Kaipainen et al . , 1995 ) at E15 . 5 and weak expression of the blood endothelial marker endomucin relative to surrounding blood vasculature ( Stanczuk et al . , 2015; Hägerling et al . , 2013 ) . Using a computational approach for the segmentation and volume rendering of PROX1+/LYVE-1+ vessels , we created 3D models of lymphatic development in the mouse embryonic kidney ( Figure 1B , and Figure 1—video 2 ) . These models visually depict the progressive remodeling of kidney lymphatic vessels from E14 . 5 through to E18 . 5 . At all stages we observed a ring-like anastomosis in the renal hilum , from which lymphatic vessels originated . These vessels branched and increased in length over the course of renal development . The 3D models conveyed that large volumes of the embryonic kidney were unoccupied by lymphatic vessels , demonstrating that kidney lymphatics are sparse compared to blood vasculature at equivalent stages ( Munro et al . , 2017; Daniel et al . , 2018; Sequeira Lopez and Gomez , 2011 ) . We then used uroplakin ( UPK3A ) , platelet endothelial cell adhesion molecule-1 ( PECAM-1 ) , aquaporin 2 ( AQP2 ) and alpha smooth muscle actin ( αSMA ) to clarify the spatial relationship of lymphatic vessels to other structures within the developing kidney ( Figure 1—figure supplement 2 ) . We found that the ring-like lymphatic anastomosis resided under the nascent renal pelvis; identified by the urothelial marker UPK3A ( Figure 1—video 3 ) . PECAM-1+ blood endothelium and lymphatic vasculature run alongside each other in the kidney , as shown in other organs ( Mahadevan et al . , 2014; Klotz et al . , 2015 ) . Whereas αSMA was expressed around the PECAM-1+ renal arterial system at E18 . 5 , as previously described ( Pitera et al . , 2004 ) , we did not detect αSMA around PROX1+ vessels . This suggests that pre-collectors or collector vessels have not yet formed by this time-point or that lymphatics within the kidney are capillaries that lack mural cell coverage ( Wang et al . , 2017 ) . We labelled collecting duct epithelium with AQP2 to delineate the renal medulla and found no lymphatic vessels in this region of the kidney ( Figure 1—figure supplement 2 ) . At no point were lymphatics observed near or within the renal capsule . Quantitative analysis of 3D imaging has provided novel insights into mouse kidney development , such as the stereotypical nature of ureteric bud branching and the cellular dynamics of the progenitor niches that eventually form the nephron , the functional unit of the kidney ( Short et al . , 2014; Short et al . , 2018; O'Brien et al . , 2018 ) . To perform the first quantitative 3D analysis of kidney lymphatic vessel development , we segmented and analyzed PROX1+/LYVE-1+ vessels from entire mouse embryonic kidneys using filament tracing software to measure the number of kidney lymphatic vessel branches including lengths , mean diameters and volumes for each vessel branch . From these data , we quantified vascular parameters: the range of vessel branch lengths , the maximum vessel branch diameter and the total volume of all branches constituting the lymphatic network between E14 . 5 and E18 . 5 ( Figure 1C ) . Overall , between E14 . 4 and E18 . 5 , the total number of lymphatic vessel branches increased over five-fold ( p=0 . 0003 ) ; the range of vessel lengths increased approximately three-fold ( p<0 . 0001 ) ; the maximum diameter approximately doubled ( p=0 . 003 ) ; and the total network volume increased by a factor of 18 ( p=0 . 0006 ) . During this time , the volume of the kidney expands over 40-fold ( Short et al . , 2014 ) and mature renal cell types emerge in the kidney including pericytes and mesangial cells ( Mugford et al . , 2008 ) , podocytes and all the components of the differentiated nephron ( Combes et al . , 2019 ) . We then compared kidneys at intermediate timepoints between E14 . 5 and E18 . 5 . No lymphatic vascular parameters increased significantly between E14 . 5 and E15 . 5 . In contrast , between E15 . 5 and E18 . 5 , we observed a significant increase in the total number of vessel branches ( p=0 . 0038 ) , their range of lengths ( p<0 . 0001 ) , maximum diameter ( p=0 . 0166 ) and total network volume ( p=0 . 0429 ) . There were further significant increases in the number of vessel branches between E15 . 5 and E17 . 5 ( p=0 . 0322 ) and the range of vessel lengths between E16 . 5 ( p=0 . 0039 ) or E17 . 5 ( p=0 . 0154 ) when compared to E18 . 5 . By pooling values for lymphatic vessel branch parameters for all kidneys at each timepoint , we generated histograms conveying the increase in vessel number and branch length , diameter and volume at each timepoint . The change in shape of histograms between each timepoint were less pronounced between E14 . 5 and E15 . 5 , suggesting rapid remodeling of the plexus after E15 . 5 ( Figure 1—figure supplement 3 ) . Together , our quantitative approach suggests a period of quiescence in the early stages of kidney lymphatic vessel development , followed by a progressive increase in lymphatic expansion after E15 . 5 . This timing just precedes the initiation of excretory function by the kidney ( McMahon , 2016; Caubit et al . , 2008 ) and also coincides with the appearance of primitive erythrocytes in renal vasculature ( Munro et al . , 2017 ) and a wave of development of vascularized glomerular capillary loops ( Hu et al . , 2016 ) . The rapid expansion of the lymphatic plexus from E15 . 5 onwards could be driven by the accumulation of interstitial fluid , which has extravasated from the renal vasculature and stimulates lymphatic expansion ( Planas-Paz et al . , 2012 ) . Additionally , either the recruitment of progenitor cells , as previously suggested for other organs ( Escobedo and Oliver , 2016; Kazenwadel and Harvey , 2016 ) , or cellular and paracrine factors may be responsible ( Vaahtomeri et al . , 2017 ) . One critical paracrine factor is vascular endothelial growth factor C ( VEGF-C ) , which binds to VEGFR-3 and promotes the sprouting and migration of lymphatic endothelial cells ( Pichol-Thievend et al . , 2018; Hägerling et al . , 2013; Karkkainen et al . , 2004 ) . To explore the expression of VEGF-C in the developing kidney , we used mice carrying LacZ under the control of the endogenous Vegfc regulatory region ( Karkkainen et al . , 2004 ) ( Figure 1—figure supplement 4 ) . Wholemount X-gal staining of E15 . 5 kidneys from heterozygous embryos ( VegfcLacZ/+ ) demonstrated β-galactosidase ( β-gal ) activity in a branching pattern , mimicking the appearance of the renal arterial tree . Sections of X-gal-stained E16 . 5 VegfcLacZ/+ kidneys showed β-gal activity to be restricted to interstitial cells beneath the pelvis and adjacent arterioles . We further stained these sections for LYVE-1 and observed lumenized LYVE-1+ vessels in the hilum surrounded by β-gal-expressing interstitial cells . Together , these findings convey that the renal hilum; where kidney lymphatics first arise , is a VEGF-C-rich niche . During embryonic development and in the early postnatal period , lymphatics form by sprouting from veins and pre-existing lymphatics; a process termed lymphangiogenesis , and the assembly of lymphatic progenitors; a process termed lymphvasculogenesis ( Potente and Mäkinen , 2017 ) . A hallmark of lymphvasculogenesis is the presence of isolated clusters of lymphatic endothelial cells , as observed during the development of mesenteric , meningeal , dermal and cardiac lymphatic vasculature ( Stanczuk et al . , 2015; Pichol-Thievend et al . , 2018; Martinez-Corral et al . , 2015; Antila et al . , 2017; Stone and Stainier , 2019; Gancz et al . , 2019 ) . By inspecting confocal image stacks of E16 . 5 mouse embryonic kidneys , we found PROX1+/LYVE-1+ cellular clusters , which were structurally distinct from the lymphatic vessel plexus ( Figure 2A and Figure 2—video 1 ) . We confirmed the lymphatic identity of these clusters by their expression of VEGFR-3 and podoplanin at E15 . 5 and E16 . 5 ( Figure 2B ) . LYVE-1 , VEGFR-3 and podoplanin all highlighted filopodia-like processes extending from lymphatic clusters in the kidney , likely analogous to the migratory tips that extend from nascent lymphatic endothelium ( Xu et al . , 2010 ) . We performed further immunolabelling to characterize the molecular profile of the clusters ( Figure 2B , Figure 2—figure supplement 1 ) . The PROX1+/LYVE-1+ clusters did not express the murine macrophage marker F4/80 ( Munro et al . , 2019 ) . Relative to renal blood vasculature , PECAM-1 and endomucin were weakly expressed by lymphatic clusters , supporting their non-blood vascular endothelial identity ( Podgrabinska et al . , 2002 ) . As our imaging technique captured entire mouse embryonic kidneys at single-cell resolution , we were able to use a quantitative approach to assess the dynamics of lymphatic endothelial cell clusters during kidney development . For each embryonic kidney , we counted the frequency of PROX1+/LYVE-1+ clusters and quantified the total number of cells constituting all clusters within each kidney ( Figure 2C ) . Between E14 . 5 and E16 . 5 there was a significant increase in both the frequency of clusters ( p=0 . 0001 ) and number of total cluster cells ( p<0 . 0001 ) . Both parameters peaked at E16 . 5 but after E16 . 5 , both frequency ( p<0 . 0064 ) and number of total cluster cells ( p<0 . 0064 ) declined significantly . This decline coincides with a reduction in the generation of new ureteric bud branches and a wave of new nephron formation ( Short et al . , 2014 ) , and agrees with the kidney lymphatic vessel expansion we observed after E15 . 5 . We also assessed the morphology of clusters within mouse embryonic kidneys using 3D segmentation and volume rendering ( Figure 2D ) . At each timepoint , we arranged volume-rendered clusters from any single kidney into a hierarchy , starting with isolated PROX1+/LYVE-1+ cells through to large islands containing eight cells or more . To investigate whether VEGF-C might also be involved in the formation of kidney lymphatic clusters , we isolated and cultured intact E14 . 5 metanephroi ex vivo ( Figure 2E ) . 3D imaging demonstrated scattered PROX1+/LYVE-1+ clusters in control metanephroi after 48 hr in culture , and their abundance increased significantly upon supplementation of the culture media with recombinant VEGF-C ( p=0 . 0184; Figure 2—video 2 ) . As multiple stages of cluster development exist within any single kidney and the decline in cluster frequency coincides with rapid lymphatic vessel expansion , we propose that both lymphangiogenesis and lymphvasculogenesis contribute to the formation of the mature kidney lymphatic vasculature and that both are regulated by VEGF-C . Lymphatic development has been described predominantly in model organisms such as mouse and zebrafish ( Semo et al . , 2016 ) . Prior studies addressing normal human lymphatic development ( von Kaisenberg et al . , 2010; Cho et al . , 2012; Jin et al . , 2010; Belle et al . , 2017 ) have mostly examined systemic lymphatics and have only used single markers such as podoplanin . Thus , it is still not clear whether events observed in animal models can be extrapolated to lymphatic development in human organs . Moreover , human kidney development has been described and thoroughly compared with that of mouse ( Lindström et al . , 2018a; Lindström et al . , 2018b ) , though the existence of lymphatic vessels during this process has not yet been examined . To image lymphatic vessels during human kidney development , we acquired millimeter-thick slices from human fetal kidneys at 12PCW . This stage marks the end of the first trimester in human gestation , designated the fetal stage of human organogenesis , and the human kidney at this stage is approximately equivalent to the E15 . 5 mouse kidney ( Lindström et al . , 2018c ) . We used 3D reconstructions of 12PCW human kidneys to identify PROX1+/podoplanin+ lymphatic vessels , including large patent lymphatic vessels in the renal hilum and extensive networks of vessels in the maturing cortex ( Figure 3A ) . Analogous to mouse , we identified a population of PROX1+/podoplanin+ lymphatic endothelial cell clusters in 12PCW human kidneys ( Figure 3B and Figure 3—video 1 ) that were structurally distinct from bona fide lymphatic vessels . Thus , the existence of lymphatic endothelial cell clusters in the kidney is likely a conserved feature in mammalian development . These clusters , akin to those previously described in the dermis , mesentery , meninges and heart , may represent a tissue-specific population of lymphatic progenitor cells , and our work provides the first demonstration of their existence in a developing human organ . We then sought to translate our 3D imaging and analysis strategy to a pathological setting in which kidney lymphatic development might be affected . We examined a mouse model of the most prevalent genetic renal anomaly , polycystic kidney disease ( PKD ) , which features the accumulation of fluid-filled epithelial cysts within the kidney that drive renal inflammation and fibrosis ( Bergmann et al . , 2018 ) . We had several reasons for hypothesizing that kidney lymphatics are altered in PKD . Firstly , the gene responsible for the majority of cases of autosomal dominant ( AD ) PKD , Pkd1 , is known to regulate extra-renal lymphatic endothelial cell migration and lymphatic vessel morphogenesis , as edema and structural defects in lymphatic vasculature have been observed in Pkd1 mutant zebrafish and mice ( Coxam et al . , 2014; Outeda et al . , 2014 ) . Secondly , the endothelial knockout of Pkd1 in mice decreased branching and increased diameter of dermal lymphatic vessels ( Lindström et al . , 2018a ) . Finally , delivery of recombinant VEGF-C increased the abundance of lymphatics within the kidney and reduced disease severity in two mouse models of PKD ( Huang et al . , 2016 ) . However , despite a putative link to PKD , the phenotype of lymphatics in polycystic kidneys and their relationship to cysts have not yet been examined . We assessed the lymphatic vasculature in kidneys from mouse embryos carrying a Pkd1 p . R3277C allele ( Pkd1RC ) . This incompletely penetrant allele in homozygosity causes adult onset ADPKD ( Hopp et al . , 2012; Rossetti et al . , 2009 ) . In homozygous mice ( Pkd1RC/RC ) , the slow growth of kidney cysts is considered to mimic the temporal progression of human ADPKD compared with other more rapidly progressive mouse models ( Happé and Peters , 2014 ) . Similar to knockouts of Pkd1 ( Coxam et al . , 2014 ) or mice lacking key genes involved in lymphatic development ( Wigle and Oliver , 1999; Karkkainen et al . , 2004; Bos et al . , 2011; François et al . , 2008 ) , we found that Pkd1RC/RC homozygous embryos presented with subcutaneous edema at E15 . 5 ( Figure 4—figure supplement 1 ) . Using conventional histology ( Figure 4A ) we discerned corticomedullary cysts lined by tubular epithelial cells in E18 . 5 Pkd1RC/RC kidneys but not in wildtype littermate controls ( Pkd1+/+ ) , consistent with early stages of cyst formation . These cysts were in close proximity to cortical lymphatic vessels ( Figure 4—video 1 ) . The proximity between lymphatics and cysts suggests that excess tissue fluid in PKD may be able to enter lymphatic vessels via paracellular transport ( Triacca et al . , 2017 ) and that there might be direct or indirect molecular interactions between cyst epithelium and lymphatic endothelial cells ( Huang et al . , 2016 ) . To visualize the renal lymphatics in the Pkd1RC model , we generated 3D visual models of PROX1+/LYVE-1+ segmented lymphatic networks and observed a stunted appearance of lymphatics in Pkd1RC/RC compared with Pkd1+/+ kidneys at E18 . 5 ( Figure 4B ) . Interestingly , we found that the lymphatic network of E18 . 5 Pkd1+/+ kidneys , raised on a C57Bl/6 background , was more extensive than that of CD-1 mice assessed at the same embryonic day ( see Figure 1B ) . Differences between mouse strains have also been observed when analyzing the incidence of subcutaneous edema , hemorrhage and embryonic lethality upon knockout of key transcription factors or micro-RNAs governing lymphatic development ( François et al . , 2008; Kontarakis et al . , 2018 ) . To screen for more subtle lymphatic abnormalities in the kidneys of Pkd1RC/RC mice , we used filament tracing software to generate color-coded models of vessel diameter at E18 . 5 ( Figure 4C and Figure 4—video 2 ) . We found large caliber vessels ( diameter >70 μm ) , present at the base of the lymphatic plexus in Pkd1+/+ kidneys , constituting the ring-like anastomosis in the renal hilum . There were no lymphatic vessels of this large diameter in the equivalent region within Pkd1RC/RC kidneys at E18 . 5 . Quantitative analysis confirmed this phenotype ( Figure 4D ) , finding a significant reduction by 18% in the mean diameter of the largest lymphatic vessel in Pkd1RC/RC kidneys compared to Pkd1+/+ controls ( p=0 . 0087 ) . In contrast , no significant differences were observed in the total number of vessel branches , the range of vessel branch lengths , the total volume of the lymphatic network between homozygous mutants and wildtype controls ( Figure 4—figure supplement 2 ) . Although the kidney volume was not significantly different between the two groups , we derived that the number of lymphatic vessel branches per unit kidney volume ( p=0 . 0418 ) and the proportion of kidney volume occupied by lymphatics ( p=0 . 0367 ) were significantly reduced in Pkd1RC/RC kidneys compared to Pkd1+/+ controls ( Figure 4E ) . The complex global defect in kidney lymphatic vessel development in the Pkd1RC model , together with the close cyst-lymphatic relationship , suggests that inadequate clearance of tissue fluid by lymphatic vessels may contribute to the expansion of cysts in PKD . In summary , we have combined advanced 3D imaging with structural rendering to provide novel spatial , temporal and quantitative insights into the formation of lymphatic vessels in developing mouse and human kidneys . By creating 3D images of kidney lymphatic vessels at single-cell resolution , we were able to visually and quantitatively demonstrate a period of quiescence , followed by extensive remodeling and expansion of the lymphatic plexus during renal development in mouse , in a process that is likely promoted by VEGF-C . The timing of plexus expansion in mouse kidneys coincides with the appearance of a highly dynamic population of lymphatic endothelial cell clusters , structurally distinct from bona fide lymphatic vessels . These clusters responded to VEGF-C and were conserved in human fetal kidneys . Our imaging approach further revealed defects in lymphatic architecture in polycystic kidneys . Together , we present a comprehensive study of the lymphatic vasculature in the developing mammalian kidney , provide the first evidence that lymphvasculogenesis is conserved in humans and implicate lymphatic anomalies in the pathophysiology of PKD .
All experiments were carried out according to a UK Home Office project license ( PPL: PE52D8C09 ) and were compliant with the UK Animals ( Scientific Procedures ) Act 1986 . To assess lymphatic development in wildtype mouse kidneys , we collected embryos from outbred CD-1 mice; maintained as a random bred closed colony at our facility . To determine the expression of VEGF-C during kidney development , we mated VegfcLacZ/+ males ( Karkkainen et al . , 2004 ) , backcrossed to a CD-1 background , with CD-1 wildtype females and collected embryos at desired timepoints . For analysis of lymphatic development in a mouse model of PKD , we performed matings between mice heterozygous for the Pkd1RC mutation ( Hopp et al . , 2012 ) on a C57BL/6 background to generate experimental litters . For all mouse experiments , matings were set up in the evening and copulation plugs found the following morning at 9am were designated E0 . 5 . At the desired timepoint , pregnant dams were sacrificed by CO2 inhalation and death confirmed by cervical dislocation . Mouse embryos were excised from the uterus and dissected in 1 x phosphate-buffered saline ( PBS ) . Tails were acquired and stored at −20°C from embryos obtained from pregnant Pkd1RC/+ heterozygous females for genotyping , which was performed as previously described ( Hopp et al . , 2012 ) . At this point , embryos generated from Pkd1RC/+ crosses were imaged using a Zeiss Axio Lumar V12 stereomicroscope . Human fetal kidneys were obtained from the Human Developmental Biology Resource ( http://www . hdbr . org ) , which obtains written consent from donors to collect , store and distribute human fetal material between 4-20PCW ( Gerrelli et al . , 2015 ) . Following dissection , human fetal material was maintained in tissue culture medium at 4°C . With the exception of VegfcLacZ/+ kidneys , we immediately washed all biological material in washed in 1 x PBS to remove blood , fixed the tissues overnight in 4% ( w/v ) paraformaldehyde ( PFA ) in PBS at 4°C , and washed tissues twice in PBS the next day . Human fetal kidneys were then incubated in 30% ( w/v ) sucrose in PBS overnight at 4°C , prior to embedding in 4% ( w/v ) agarose in PBS and slicing at 1 mm on a vibrating blade microtome ( VT1000S , Leica Biosystems ) . We subsequently stored all material in PBS with 0 . 02% ( w/v ) sodium azide at 4°C . VegfcLacZ/+ or Vegfc+/+ kidneys were fixed in 2% PFA for 20 min on ice . They were then washed thrice for five mins each in PBS , before incubating at 37°C in stain solution ( 5 mM potassium ferricyanide , 5 mM potassium hexacyanoferrate ( III ) trihydrate , 2 mM magnesium chloride and 0 . 01% ( w/v ) sodium deoxycholate in PBS ) with X-gal in a 1:40 ratio . The X-gal reaction was quenched by washing twice in PBS with 0 . 02% sodium azide . Wholemount X-gal images were captured using a Leica MZFLIII stereomicroscope with an IDS UI-3080CP-C-HQ camera and micromanager software . Pkd1RC/RC , Pkd1+/+ or X-gal-stained VegfcLacZ/+ kidneys were dehydrated in an ethanol series prior to incubation in Histo-Clear II ( National Diagnostics ) and embedding in paraffin wax . Sections were cut at 5 μm using a microtome . Deparaffinization and staining with periodic acid-Schiff was performed as previously described ( Brzóska et al . , 2016 ) . Colorimetric immunostaining of VegfcLacZ/+ kidney sections for LYVE-1 was performed using a goat anti-mouse LYVE-1 with horseradish peroxidase rabbit anti-goat secondary antibody . Brightfield imaging was performed using an upright Zeiss Axioplan microscope equipped with a Zeiss Axiocam HRc camera and Axiovision software . Freshly isolated kidneys were obtained from E14 . 5 CD-1 mice , explanted on Millicell cell culture inserts ( Millipore ) and incubated at 37°C in 20% O2 and 5% CO2 . Explants were grown at the air-liquid interface with Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 ( Gibco ) , supplemented with or without 40 ng/ml recombinant VEGF-C . The culture medium was replaced 24 hr after initiation of the experiment . After 48 hr in culture , explants were collected in PBS , fixed in 4% PFA overnight and processed for wholemount immunolabelling , optical clearing and confocal microscopy ( see below ) . Infected explants and those that had not adhered to the cell culture membrane were discarded . Data were pooled from two independent experiments , each from a separate litter of CD-1 embryos . For mouse embryonic kidneys , a modified version of the iDISCO pipeline ( Renier et al . , 2014 ) was performed for wholemount indirect immunolabelling using antibodies and concentrations indicated in the Key resources table , with all steps performed on rotating shakers ( Jafree et al . , 2020 ) . We first dehydrated mouse embryonic kidneys in a methanol series at room temperature ( RT ) and bleached in methanol with 5% ( v/v ) of 30% hydrogen peroxide ( H2O2 ) solution overnight at 4°C . They were then rehydrated at RT and permeabilized in PBS with 0 . 2% ( v/v ) Triton X-100 , 2 . 3% ( w/v ) glycine and 20% ( v/v ) dimethyl sulfoxide ( DMSO ) overnight at 4°C . Blocking was then performed in PBS with 0 . 2% Triton X-100 , 6% ( v/v ) donkey serum and 10% DMSO at RT for one day , and incubated in 500 μl of wash buffer ( PBS solution with 0 . 2% ( v/v ) Tween-20 , 0 . 1% ( v/v ) of 10 mg/ml heparin stock solution ) with 3% donkey serum , 5% DMSO and primary antibody for three days at 4°C . The labelled kidneys were then incubated in washing buffer 4–6 times for 1 hr each at RT . Secondary antibodies , either AlexaFluor 488 , 546 or 633 ( ThermoFisher Scientific ) were then added at a concentration of 1:200 in 500 μl of wash buffer with 3% donkey serum and 5% DMSO and incubated at 4°C overnight , before 4–6 applications of washing buffer alone for 1 hr each . Hoechst was added with secondary antibodies for selected Pkd1RC/RC or Pkd1+/+ kidneys . The wholemount immunofluorescence process was similar for millimeter-thick vibratome slices of 12PCW human kidneys , but instead adapting a protocol optimized for human embryonic and fetal material ( Belle et al . , 2017 ) . The protocol was identical to above , but involved the following changes: ( 1 ) bleaching was performed in 10% of 30% H2O2 solution; ( 2 ) permeabilization was performed at RT in 1 x PBS with 0 . 2% ( w/v ) bovine gelatin with 0 . 2% Triton X-100 ( PBSGT ) with 20% DMSO overnight; ( 3 ) all wash steps were performed in PBSGT; ( 4 ) Blocking was performed in PBSGT with 10% DMSO and 3% donkey serum and ( 5 ) Primary and secondary antibody incubations were performed in 1 ml of PBSGT with 0 . 1% ( w/v ) saponin , 5% DMSO and 3% donkey serum at 37°C for six days , and primary antibodies were replenished after three days . After immunolabelling , both mouse embryonic kidneys and human kidney slices were dehydrated in a methanol series at RT , adapted for clearing using a 1:1 mixture of methanol and 1:2 benzyl alcohol:benzyl benzoate , termed BABB ( Combes et al . , 2014; Dodt et al . , 2007 ) , and finally cleared in BABB alone until optically transparent . All confocal images were acquired using a Zeiss LSM 880 Upright Confocal Multiphoton microscope with gallium arsenide phosphide internal and external detectors and a 10x/numerical aperture 0 . 5 water dipping objective ( working distance: 3 . 7 mm ) with 2 . 77 μm z-step . To protect the microscope objective from BABB , we suspended smaller tissues ( E14 . 5–16 . 5 mouse kidneys ) in a drop of BABB within a glass-bottomed FluoroDish ( World Precision Instruments ) , which was inverted for upright confocal imaging . Larger tissues ( E17 . 5–18 . 5 mouse kidneys and 12PCW human kidney slices ) were placed in BABB within a custom imaging chamber , consisting of an FKM rubber ring ( Polymax ) between a glass slide and coverslip . Prior to imaging , a drop of distilled water was placed on top of the inverted FluoroDish or coverslip , into which the objective was placed . For single-photon excitation we used laser lines at 488 , 561 and 633 nm . For two-photon excitation , a Mai Tai eHP DeepSee multiphoton laser ( SpectraPhysics , 800 nm ) was used . Confocal z-stacks were acquired as 8-bit images with pixel resolutions of 512 × 512 or 1024 × 1024 . For E16 . 5–18 . 5 mouse embryonic kidneys and 12PCW human fetal kidney slices , tile scanning was performed with a 10% overlap between each image tile . All images were saved as CZI files . We use the Stitching tool in Zen software ( Zeiss ) to stitch together tile scans acquired from confocal imaging . In kidneys from the Pkd1RC mouse model stained with Hoechst or in highly autofluorescence 12PCW human kidney slices , spectral unmixing was performed in Zen to separate nuclear counterstaining and autofluorescence respectively from spill-over into other fluorescence channels . Images were then exported to FIJI . Confocal image stacks were separated into individual fluorescence channels and Despeckle and Sharpen tools were used to reduce non-specific background fluorescence . Where maximum intensity projections or optical z-sections were required , scalebars applied and exported as TIFF files . TIFF files were further imported into Imaris ( Bitplane ) where 3D reconstruction , segmentation or analysis , volume rendering or videos were required . To segment lymphatic vessel and clusters from mouse embryonic kidneys , we used the Isosurface Rendering tool in Imaris . Segmentation of the LYVE-1 channel was performed based on fluorescence intensity , and thresholds were manually selected to capture PROX1+/LYVE-1+ structures . To isolate the lymphatic vessel plexus alone , a filter was applied to retain structures with the largest volume . Thereafter , we generated masks selecting for expression of both PROX1 and LYVE-1 only , generating new channels with either unedited fluorescence intensity ( as in Figure 2A and Figure 4—video 1 ) or binarized for filament tracing analysis . We volume-rendered and segmented lymphatic vessels and clusters in Imaris to generate the 3D models shown in Figure 1B and Figure 4B . Due to the high sensitivity of the two-photon detection system , we generated a custom script in FIJI to create a new fluorescence channel containing segmented and binarized PROX1+/LYVE-1+ lymphatic vessels from Pkd1RC/RC or Pkd1+/+ kidneys . Briefly , PROX1+ nuclei and LYVE-1+ cells were separately segmented to form masks using a rolling ball background subtraction , auto-thresholding and the 3D Simple Segmentation plugin . The nuclear and cellular masks were overlaid to create a new image channel containing only PROX1+/LYVE-1+ cells . These were then exported into IMARIS as TIFF files alongside the original channels and rendered using Imaris as above . Color-coded 3D visualization of vascular parameters ( as in Figure 1—figure supplement 4 or Figure 4C ) , such as vessel volume or diameter were generated using the FilamentTracer tool in Imaris from segmented PROX1+/LYVE-1+ lymphatic vessels . For quantitative analysis , TIFF image stacks were imported into Amira ( ThermoFisher Scientific ) . The Filament Editor tool was used in Amira to generate spatial statistical parameters including vessel branch number , lengths , diameters and volumes from each segmented lymphatic plexus . These were exported as CSV files for graphing and statistical tests . We defined clusters as PROX1+/LYVE-1+ cells with no continuity of PROX1+ nuclei with the lymphatic vessel plexus . To maximize specificity , clusters were quantified manually by rigorously scrolling through serial confocal images of entire mouse embryonic kidneys in FIJI . Where a cluster was observed , it was marked with the Multi-point tool to avoid counting the same clusters twice , and the number of cells within clusters was defined by the number of PROX1+ nuclei they contained . Where tile-scanning was performed , individual tiles were each inspected for the presence of clusters . Total kidney volume was determined in a semi-automated fashion using FIJI . Series confocal images from each kidney had a Gaussian filter applied ( σ = 5 ) . Thresholding was applied using the Li method . A binary mask was then created , and holes filled before 3D Connected Component Analysis . The branches per unit volume of the kidney and total kidney volume occupied by lymphatic vessels were calculated by dividing the total number of vessel branches and the total lymphatic network volume of each kidney , derived from Filament Editor tool in Amira , by the total kidney volume as calculated above . Lymphatic vessel branches per unit kidney volume were expressed per mm , whereas the proportion of the kidney occupied by lymphatic vessels was expressed as a percentage of total kidney volume . We estimated sample sizes based on preliminary experiments and prior quantitative analyses of developing cardiac ( Klotz et al . , 2015 ) , dermal ( Pichol-Thievend et al . , 2018 ) and mesenteric ( Stanczuk et al . , 2015 ) lymphatic vessels . For quantitative analysis of lymphatic development in wildtype mice , we predicted between 4–6 embryonic kidneys per timepoint would be sufficient to power experiments . For all experiments on wildtype mouse embryonic kidneys , 4–6 experimental repeats were randomly selected from kidneys pooled by litter at the required timepoint . In all cases , conclusions were drawn from a minimum of two to three independent litters per timepoint . For analysis of lymphatic abnormalities in the Pkd1RC model at E18 . 5 , we analyzed the right kidney only and the left was taken for histology . All other sample sizes are shown in the relevant figure legends . Storage , statistics and graphing of numerical data was performed in Prism ( v8 , GraphPad ) . A two-tailed p value of less than 0 . 05 was considered statistically significant for all tests . Shapiro-Wilk and Brown-Forsythe tests were used to test Gaussian distribution and equality of variance respectively within all datasets . Where Gaussian distribution and equality of variances were satisfied , we used ANOVA to compare vascular parameters or cluster number and cell content across all timepoints during renal development . Bonferroni tests for multiple comparisons were used to compare individual timepoints after ANOVA . Similarly , we used unpaired Student’s t-test to compare vascular parameters in Pkd1RC/RC and Pkd1+/+ kidney lymphatic vessels and to compare the effect of VEGF-C upon number of lymphatic clusters in mouse embryonic kidney explants . Where Gaussian distribution or equality of variances were violated , non-parametric tests and subsequent multiple comparisons were performed as detailed in the figure legends . All images were compiled into Microsoft PowerPoint where figures were created . All videos , 3D reconstructions , color-coded models and volume renderings were prepared in IMARIS and exported as MP4 or TIFF files . MP4 files were annotated in iMovie ( v10 . 1 , Apple Inc ) . | In most organs in the body , fluid tends to build up in the spaces between cells , especially if the organs become inflamed . Each organ has a ‘waste disposal system’; a set of specialized tubes called lymphatic vessels , to clear away this excess fluid and keep a check on inflammation . Defects in these tubes have been linked to a wide range of diseases including heart attacks , obesity , dementia and cancer . The kidneys are responsible for filtering blood and balancing many of the body’s chemical processes . Polycystic kidney disease ( PKD ) is the most common genetic kidney disorder and it results in cysts filled with fluid building up in the kidney . The growth of cysts in PKD may be due to a problem with the lymphatic vessels . However , compared to other organs , how lymphatic vessels first form within the kidney and what they do is not well understood . Now , Jafree et al . have used three-dimensional imaging to study how lymphatic vessels form in the kidneys of mice and humans . The experiments showed that lymphatic vessels first appear when mouse kidneys are about half developed , and start to grow rapidly when the kidneys are thought to begin filtering blood . Clusters of cells that may help lymphatic vessels to grow were also found hidden deep within the kidneys of mouse embryos . Treating the kidneys with a factor that stimulates the growth of lymphatic vessels increased the numbers of these clusters . Jafree et al . found similar clusters of cells in human kidneys , suggesting that lymphatic vessels in the kidneys of different mammals may develop in the same way . Further experiments showed that the lymphatic vessels of kidneys in mice with PKD become distorted early on in the disease , when cysts are still small and before the mice develop symptoms . In the future , identifying drugs that target kidney lymphatic vessels may lead to more effective treatments for patients with PKD and other kidney diseases . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"short",
"report"
] | 2019 | Spatiotemporal dynamics and heterogeneity of renal lymphatics in mammalian development and cystic kidney disease |
The ‘ribosomal stress ( RS ) -p53 pathway’ is triggered by any stressor or genetic alteration that disrupts ribosomal biogenesis , and mediated by several ribosomal proteins ( RPs ) , such as RPL11 and RPL5 , which inhibit MDM2 and activate p53 . Inosine monophosphate ( IMP ) dehydrogenase 2 ( IMPDH2 ) is a rate-limiting enzyme in de novo guanine nucleotide biosynthesis and crucial for maintaining cellular guanine deoxy- and ribonucleotide pools needed for DNA and RNA synthesis . It is highly expressed in many malignancies . We previously showed that inhibition of IMPDH2 leads to p53 activation by causing RS . Surprisingly , our current study reveals that Inauzhin ( INZ ) , a novel non-genotoxic p53 activator by inhibiting SIRT1 , can also inhibit cellular IMPDH2 activity , and reduce the levels of cellular GTP and GTP-binding nucleostemin that is essential for rRNA processing . Consequently , INZ induces RS and the RPL11/RPL5-MDM2 interaction , activating p53 . These results support the new notion that INZ suppresses cancer cell growth by dually targeting SIRT1 and IMPDH2 .
With ∼22 million people living with cancers that are highly associated with alterations of multiple molecules and pathways , it is important to develop a multiple molecules-targeted therapy that can effectively kill cancer cells . The tumor suppressor p53 pathway is one such a target because nearly all cancers show defects in this pathway . Approximately 50% of human cancers have mutations in the TP53 gene itself , while the rest of them harbor functionally inactive p53 proteins , because active p53 can trigger cell growth arrest , apoptosis , autophagy , and/or senescence , which are detrimental to cancer cells ( Vogelstein et al . , 2000; Vousden and Prives , 2009 ) , and impede cell migration , metabolism , and/or angiogenesis . A major mechanism for functional inactivation of p53 is through overexpression of two chief p53 suppressors , MDM2 and MDMX , which work together to inactivate p53 by directly interacting with p53 , inhibiting its transcriptional activity and mediating its ubiquitin dependent degradation ( Wade et al . , 2010; Huang et al . , 2011; Tollini and Zhang , 2012 ) . This MDM2/MDMX-mediated p53 degradation is also facilitated by SIRT1 , a nicotinamide adenine dinucleotide ( NAD+ ) -dependent deacetylase ( Vaziri et al . , 2001; Cheng et al . , 2003 ) . SIRT1 is highly expressed in human cancers due to down regulation of another p53 target tumor suppressor called hypermethylated in cancer-1 ( HIC-1 ) ( Chen et al . , 2005 ) . Our previous study identified a small molecule named Inauhzin ( INZ ) that effectively inhibits SIRT1 activity and induces p53 acetylation , leading to the increase of p53 level and activity ( Zhang et al . , 2012b ) . Consequently , INZ induces p53-dependent apoptosis and senescence in various p53-wild type human cancer cells , such as H460 , and HCT116 by inducing the expression of p53-dependent transcriptome ( Liao et al . , 2012 ) . INZ markedly inhibits the growth of H460 or HCT116 xenograft tumors , but is not toxic to normal cells and tissues . Also , INZ sensitizes the anti-cancer effect of cisplatin , doxorubicin , or Nutlin-3 ( an MDM2 inhibitor ) as tested in xenograft cancer models ( Zhang et al . , 2012c; Zhang et al , 2013 ) . Thus , this small molecule presents as a promising contender for a molecule-targeted anti-cancer therapy . Since its discovery , we have optimized INZ ( Zhang et al . , 2012a ) and determined additional cellular proteins that INZ might target via a set of biochemical , proteomic , and cell-based analyses . As detailed below , our study unveils inosine monophosphate ( IMP ) dehydrogenase 2 ( IMPDH2 ) as a novel cellular target of INZ .
IMPDH is the key metabolic enzyme supplying guanine nucleotides to a cell as the first and rate-limiting enzyme of de novo GTP biosynthesis by catalyzing NAD+-dependent oxidation of IMP to xanthosine monophosphate ( XMP ) ( Zimmermann et al . , 1995; Zhang et al . , 1999 ) . IMPDH2 is the predominant isoform among its two isoenzymes , and often highly expressed in proliferating cells and neoplastic tissues ( Ishitsuka et al . , 2005; Gu et al . , 2003 ) , correlated to drug resistance , and thus has been used as a validated target for immunosuppressive ( mycophenolic acid [MPA] [Sintchak et al . , 1996] and mizoribine [Gan et al . , 2003] ) , antiviral ( ribavirin [Prosise et al . , 2002] ) , and cancer-chemotherapeutic development [tiazofurin] ( Malek et al . , 2004; Gu et al . , 2005; Chen and Pankiewicz , 2007; Borden and Culjkovic-Kraljacic , 2010 ) . Interestingly , by performing a biotin-INZ avidin affinity purification coupled with mass spectrometry ( MS ) analysis , we identified IMPDH2 as one of the top candidate proteins that INZ specifically targets in cancer cells . Biotinylated INZ analogs ( Figure 1A ) were synthesized for these analyses . Here , Biotin-INZ was as active as INZ ( Zhang et al . , 2012b; Zhang et al . , 2012a ) , while Biotin-INZ ( O ) was inactive and thus used as a negative control . Comparison of the most abundant proteins based on normalized spectral abundance factor ( NSAF ) in the cells treated Biotin-INZ vs DMSO or Biotin-INZ ( O ) revealed the high enrichment of IMPDH2 proteins in the former ( Figure 1B ) with enriched IMPDH2 peptides shown in Figure 1C . This result was firmly validated by immunoblot ( IB ) analysis of the pulled down proteins , as IMPDH2 was specifically brought down with Biotin-INZ , as well as together with our previously identified SIRT1 , but not Biotin-INZ ( O ) or other controls , in both H460 and HCT116 cells ( Figure 1C ) . 10 . 7554/eLife . 03077 . 003Figure 1 . Identification of IMPDH2 as a potential target of INZ . ( A ) Structure of INZ analogs conjugated with Biotin ( Biotin-INZ ) used for INZ target identification , the oxygen-substituent ( Biotin-INZ ( O ) ) as a negative control . ( B–C ) Cells were treated with indicated compounds individually for 18 hr . Cleared cell lysates were incubated with NeutrAvidin beads and washed . The samples from HCT116 cells were then in-beads digested for MS analysis . NSAF: normalized spectral abundance factor . Samples were also resolved by SDS-PAGE and subjected to IB with indicated antibodies . ( D–F ) Knockdown of IMPDH2 alleviates INZ induction of p53 . HCT116 cells were transfected with scrambled siRNA or IMPDH2 siRNA . 18 hr prior to harvesting , cells were treated with 2 µM INZ and harvested for IB with indicated antibodies ( D ) . HCT116 and HCT116−/− cells were exposed to INZ for 72 hr and evaluated by WST cell growth assays ( E and F ) . The IC50 values of INZ in the scrambled siRNA and IMPDH2 siRNA transfected cells are 0 . 87 ± 1 . 08 µM and 10 . 24 ± 2 . 57 µM for HCT116 cells , and 5 . 28 ± 2 . 43 µM and 12 . 34 ± 2 . 02 µM for HCT116−/− cells , respectively ( Mean ± SD , n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03077 . 003 To test if IMPDH2 is required for INZ activation of p53 , we knocked down IMPDH2 in HCT116 cells with specific siRNAs in the presence or absence of INZ . As shown in Figure 1D , knockdown of IMPDH2 impeded INZ-induced p53 activation as indicated by the reduction of INZ-induced p53 , p21 , MDM2 , and Puma levels . Consistently , the growth inhibition by INZ was compromised as the IC50 value for INZ in cell growth analysis decreased by almost ∼10-fold when IMPDH2 was knocked down ( Figure 1E ) . Knockdown of IMPDH2 in HCT116 cells also conveyed much more significant effect on compromising the cytotoxicity of INZ compared to p53 null HCT116 ( HCT116−/− ) cells ( Figure 1F ) , indicating that INZ suppresses cancer cell growth mainly by targeting IMPDH2 in the cells and consequently activating the p53 pathway . Although Biotin-INZ was associated with cellular IMPDH2 ( Figure 1 ) , INZ did not appear to affect the activity of the purified enzyme ( date not shown ) . This discrepancy could be due to differences between the recombinant IMPDH2 in vitro and its native form in cells , as the latter could be regulated via post-translational modifications or partner proteins in cells , or INZ might mimic a nucleoside and be phosphorylated by a kinase in cells to target IMPDH2 . These results also suggest that INZ might not directly bind to the active site of this enzyme . These possibilities remain to be addressed in the future . Our previous study showed that inhibition of IMPDH2 activity by MPA leads to RS and consequent p53 activation by reducing the level of nucleostemin ( NS ) ( Dai et al . , 2008; Lo et al . , 2012 ) , a nucleolar GTP-binding protein important for rRNA processing ( Tsai and McKay , 2005; Lo et al . , 2012 ) . The association of INZ with IMPDH2 suggested that INZ might have a similar effect . As shown in Figure 2A , INZ , but not INZ ( O ) , indeed significantly reduced NS protein levels , which was inversely correlated with the INZ induction of p53 , p21 , MDM2 and cleaved PARP . This result was further confirmed by immunofluorescence staining , as INZ , but not INZ ( O ) , led to apparent decrease of nucleolar NS ( Figure 2B ) . This decrease was due to the reduction of NS's half-life from >10 hr to <6 hr , as shown in Figure 2C , but NS mRNA level did not alter ( data not shown ) . This result , also repeated in HCT116 cell lines ( data not shown ) , demonstrates that INZ can destabilize cellular NS . 10 . 7554/eLife . 03077 . 004Figure 2 . INZ , not its inactive analog INZ ( O ) , treatment , decreases NS expression and destabilizes NS protein . H460 cells were treated with 2 μM INZ or its analogue INZ ( O ) for 20 hr . Cells were harvested and immunoblotted with p53 , NS , MDM2 , p21 , cleaved PARP and β-actin ( A ) , or immunostained with anti-NS ( green ) and anti-p53 ( red ) ( B ) . ( C ) H460 cells were treated with 2 μM INZ for 9 hr before then 50 μg/ml of CHX was added . Cells were harvested at different time points as indicated and assayed for levels of NS . DOI: http://dx . doi . org/10 . 7554/eLife . 03077 . 004 Next , we tested if the depletion of NS by INZ could induce the interaction of the RPL11 and RPL5 with MDM2 , because we previously showed that the reduction of NS by MPA could induce RS and activate p53 by enhancing this interaction ( Sun et al . , 2008 ) . As shown in Figure 3A , INZ , but not INZ ( O ) , indeed enhanced the interaction of MDM2 with RPL5 and RPL11 in H460 cells by immunoprecipitation ( IP ) using anti-MDM2 antibodies followed by IB ( Figure 3A ) . The increased binding of MDM2 to L11 was true in a reciprocal co-IP using anti-L11 antibodies ( Figure 3A ) . This result indicates that INZ-induced p53 activation involves suppression of MDM2 activity by the RPs , further supporting the RS-p53 response of INZ-treated cells . 10 . 7554/eLife . 03077 . 005Figure 3 . INZ treatment enhances the interaction of MDM2 with L5 and L11 by inducing ribosome-free form of RPL5 and RPL11 . ( A ) H460 cells were treated with 2 μM INZ or INZ ( O ) for 18 hr . Cell lysates were used for IP with anti-MDM2 antibodies or anti-L11 antibodies followed by IB using anti-RPL5 , RPL11 or MDM2 antibodies . ( B and C ) HCT116 cells were transfected with siRNAs against RPL5 and RPL11 , or control , for 48–72 hr and treated with 2 µM INZ 18 hr before harvesting , followed by IB using indicated antibodies or subG1 analysis by flow cytometry . ( D ) Ribosomal profile assay . Cytoplasmic extracts containing ribosomes from H460 cells treated with or without 2 µM INZ for 18 hr were subjected to a 10–50% linear sucrose gradient sedimentation centrifugation . Fractions were collected and subjected to IB with anti-RPL11 , anti-RPL5 , anti-p53 , or anti-MDM2 antibodies . The distribution of ribosomes is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 03077 . 005 To determine if RPL5 and RPL11 are required for INZ activation of p53 , we performed a knockdown experiment . As expected , reduction of either RPL5 or RPL11 ( data not shown ) or both levels by siRNA markedly inhibited INZ-induced p53 level , compared to that in scrambled siRNA-transfected cells ( Figure 3B ) . Consistently , knocking down RPL5 and RPL11 abrogated INZ-induced p21 and MDM2 levels ( Figure 3B ) and apoptosis ( Figure 3C ) , indicating that RPL5 and RPL11 are required for INZ-induced p53 activation and apoptosis . Also , ribosome profile analysis followed by IB revealed that INZ significantly increases the levels of ribosome-free RPL5 and RPL11 ( fractions 1–10 ) , whereas it markedly reduces the level of polysomes ( fractions 37–51 ) ( Figure 3D ) , suggesting that INZ could suppress ribosome biogenesis and possibly protein translation . All together , these results demonstrate that INZ can activate p53-dependent apoptosis by interfering with ribosome biogenesis through depletion of NS , causing RS , which then induces the release of ribosome-free RPL11 and RPL5 that bind to MDM2 and consequently inhibit its activity toward p53 . Because Inhibition of IMPDH2 reduces cellular GTP level ( Ji et al . , 2006 ) , and INZ associates with cellular IMPDH2 and reduces nucleolar NS level , consequently causing RS and p53 activation ( Figures 1–3 ) , we then tested if this INZ effect on p53 could be suppressed by supplementing culture media with extra GTP or guanosine . As shown in Figure 4A–B , addition of either GTP or guanosine to cells significantly , though partially , alleviated the INZ induction of p53 level and activity as measured by IB analysis of p53 , Puma and cleaved PARP . This result is well correlated with Figure 4C , showing that INZ markedly reduced the GTP level in H460 cells by 6 . 3-fold and in HCT116 cells by 3 . 7-fold , respectively , as measured by HPLC analysis ( Di Pierro et al . , 1995 ) . These results indicate that INZ can reduce cellular GTP level likely by inhibiting IMPDH2 in cells . Indeed , knockdown of IMPDH2 compromised the GTP depletion by INZ treatment in both p53 wild type and p53 null HCT116 cancer cells ( Figure 4D ) . Since it has been shown that NS is very sensitive to cellular GTP level and low GTP level triggers NS re-localization from the nucleolus to the nucleoplasm , consequently destabilizing it ( Tsai and McKay , 2005; Lo et al . , 2012 ) , these results also suggest that it must be by decreasing cellular GTP level that INZ causes NS degradation and consequent RS , leading to p53 activation ( Figure 4E ) . 10 . 7554/eLife . 03077 . 006Figure 4 . GTP or guanosine lessens INZ activation of p53 in cells . ( A and B ) H460 cells were pretreated with GTP or guanosine for 2 hr before the addition of 2 μM INZ . Cells were harvested and followed by IB with indicated antibodies . ( C ) Effect of INZ on cellular GTP level . The nucleotides were extracted from H460 or HCT116 cells treated with 2 μM INZ by 80% acetonitrile and SPE column . Samples were subjected to GTP analysis by HPLC . Results of quantification of HPLC spectra presented in arbitrary units ( AU ) were presented in this Table . ( D ) HCT116−/− and HCT116 cells transfected IMPDH2 siRNA ( SiIMPDH2 ) or scrambled siRNA ( SiControl ) were exposed to INZ for 18 hr , and cellular GTP was extracted , measured and quantitated by LC-MS/MS . Values represent means ±SD ( n = 2 ) . ( E ) A schematic diagram of the role of IMPDH2 in INZ-induced ribosomal stress ( RS ) and p53 activation . IMPDH2 is a rate-limiting enzyme in the de novo guanine nucleotide biosynthesis . INZ reduced the levels of cellular GTP and NS by targeting IMPDH2 ( or its complex ) , resulting in RS that leads to the enhancement of the RPL11/RPL5-MDM2 interaction , consequently MDM2 inactivation and p53 activation . DOI: http://dx . doi . org/10 . 7554/eLife . 03077 . 006 Cancers are caused by alterations of multiple tumor-associated proteins or genes at the genetic and epigenetic levels ( Hanahan and Weinberg , 2011 ) , including the p53 pathway ( Vousden and Prives , 2009 ) . Thus , targeting multiple proteins of one or more signaling pathways in cancers is necessary for developing a more effective cancer therapy . Several individual SIRT1 or IMPDH2 inhibitors have been reported ( Alcain and Villalba , 2009; Chen et al . , 2010 ) . However , dual targeting SIRT1 and IMPDH2 by INZ to activate p53 would offer the first paradigm for anti-cancer drug development . Our studies ( Figures 1–4 ) together with previously published findings strongly suggest that INZ effectively activates p53 and suppresses tumor growth in a p53-dependent fashion by targeting SIRT1 and IMPDH2 ( Figure 4E , ( Zhang et al . , 2012b ) ) . This dual targeting strategy could also explain why INZ can still partially activate p53 in IMPDH2 knockdown or GTP-supplemented cells ( Figure 1D and Figure 4A , B ) , although the partial impairment of p53 induction could also be due to the inefficiency of completely knockdown IMPDH2 or the possible non-continuous availability of intracellular GTP throughout the experiment .
Human lung carcinoma H460 and human colon cancer HCT116 were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum ( PBS ) , penicillin , and streptomycin . Inauhzin ( INZ ) , Inauhzin inactive analogue INZ ( O ) ( INZ9 in [Zhang et al . , 2012a] ) and Biotinylated INZs were synthesized and characterized by NMR and LC-MS as described ( Zhang et al . , 2012b ) . The purity of the compounds is higher than 90% . Mycophenolic acid ( MPA ) was purchased from Sigma-Aldrich ( St . Louis , Missouri ) . Mouse monoclonal anti-p53 ( DO-1 ) , rabbit anti-p21 ( M19 ) , mouse anti-p21 ( F5 ) , rabbit anti-SIRT1 ( H300 ) and goat anti-RPL11 ( N17 ) were purchased from Santa Cruz Biotechnology , Inc . ( Dallas , Texas ) . for immunoblotting . Cleaved PARP , PARP ( 9542 ) , Puma were from Cell Signaling Technologies . Mouse anti-MDM2 ( 2A10 ) , rabbit anti-RPL11 and anti-RPL5 antibodies were described previously ( Zeng et al . , 1999; Sun et al . , 2008 ) . Antibodies for immunostaining were rabbit polyclonal anti-p53 ( FL-393; Santa Cruz ) and monoclonal nucleostemin antibodies ( Chemicon , Billerica , Massachusetts ) . Cells were plated on 10 cm dishes and treated with compounds at about 60–70% confluence . Cells were harvested and lysed in the PBS buffer with 0 . 1% ( wt/vol ) NP40 ( freshly adding protease inhibitors and 1 mM DTT ) . Incubate the cell lysate with 25 μL of NeutrAvidin Agarose beads ( Thermo Scientific , Waltham , Massachusetts ) ( beads volume ) in for 2 hr at 4°C with end-over-end mixing . Centrifuge at 13 , 000 rpm for 10 s at 4°C in a microcentrifuge . The beads were washed three times with 0 . 5% ( wt/vol ) NP-40 , 0 . 2% ( wt/vol ) Tween20/Tris buffered saline and then subjected to on-beads digestion and mass spectrometry as shown below . On-bead digestions were performed to release the proteins and resulting tryptic peptides from the NeutrAvidin beads . In brief , beads were resuspended in 100 μl of 50 mM ammonium bicarbonate pH 8 . 0 followed by the addition of 1 μg of Trypsin Gold ( Promega , Madison , Wisconsin ) . The samples were then incubated at 37°C for 12 hr with shaking . Following digestion , samples were run through spin columns to remove any trace of the residual purification resin . The digestions were then quenched through the addition of 8 μL of formic acid . Protein samples from cells that were MOCK treated with Biotin-INZ ( O ) or treated with DMSO or Biotin-INZ were pressure loaded onto three-phase MudPIT columns containing Aqua C18 and Luna SCX resins ( Phenomenex , Torrance , California ) as previously described ( Mosley et al . , 2009 , 2011 ) . Ten-step MudPIT was performed using increasing concentrations of ammonium acetate to initiate each step followed by a 100-min gradient of 0–80% acetonitrile . All samples were analyzed on a LTQ Velos mass spectrometer ( Thermo Scientific ) with the dynamic exclusion set to 90 s . The spectra obtained through MudPIT analysis were searched through Proteome Discoverer 1 . 3 ( Thermo Scientific ) using SEQUEST as the peptide-spectrum matching algorithm against the Human NCBI 11-22-10 database containing 29 , 535 protein sequences . In addition to the human proteins , the database also contained ∼140 common contaminant sequences for proteins such as keratins , BSA , and proteolytic enzymes . Using Proteome Discoverer 1 . 3 , all peptides were required to pass a 2% false discovery threshold . The number of spectra obtained for proteins found to interact with Biotin-INZ was compared to the levels of spectra for those same proteins observed in MOCK and DMSO treatments to ensure that the candidate interacting proteins are detected at levels higher than background . Cells were seeded in 6-well plates . All compounds were dissolved in DMSO and diluted directly into the medium to the indicated concentrations , and 0 . 1% DMSO was used as a control . After incubation with the compounds for the indicated times , cells were harvested and lysed in 50 mM Tris-HCl pH 8 . 0 , 150 mM NaCl , 5 mM EDTA , 0 . 5% NP-40 supplemented with 2 mM DTT and 1 mM PMSF . An equal amount of protein samples ( 50 μg ) was subjected to SDS-PAGE and transferred to a PVDF membrane ( PALL Life Science , Port Washington , New York ) . The membranes with transferred proteins were probed with primary antibodies followed by horseradish-peroxidase-conjugated secondary antibody ( 1:10 , 000; Pierce ) . The blots were then developed using an enhanced chemiluminescence detection kit ( Thermo Scientific ) , and signals were visualized by Omega 12iC Molecular Image System ( UltraLUM , Claremont , California ) . H460 cells at 50–70% confluence were treated with 2 µM of Inauhzin ( INZ ) or Inauhzin ( O ) ( INZ ( O ) ) for 16 hr . Cells were fixed in 4% formaldehyde/PBS for 10 min , permeabilized and blocked with 0 . 3% Triton-100 , 8%BSA/PBS . The primary antibodies used were monoclonal nucleostemin antibodies in 1:250 dilution and polyclonal p53 antibodies in 1: 500 dilution according to the manufactural instruction . Images were taken with a Zeiss Axiovert 200M fluorescent microscope ( Germany ) . Control scrambled siRNA ( Santa Cruz ) , or siRNA specific to IMPDH2 ( Santa Cruz and Ambion , Grand Island , New York ) were commercially purchased . These siRNAs ( 60 nM ) were introduced into cells using METAFECTENE SI following the manufacturer's protocol ( Biontex , Germany ) . Cells were treated with INZ for IB , cell viability assays and FACS analysis . To assess cell growth , the cell counting kit ( Dojindo Molecular Technologies Inc . , Rockville , Maryland ) was used according to manufacturer's instructions . Cell suspensions were seeded at 5000 cells per well in 96-well culture plates and incubated overnight at 37°C . Compounds were added into the plates and incubated at 37°C for 72 hr . Cell growth inhibition was determined by adding WST-8 at a final concentration of 10% to each well , and the absorbance of the samples was measured at 450 nm using a Microplate Reader ( Molecular Device , SpectraMax M5e ( Sunnyvale , California ) ) . Cells were harvested , fixed in 70% ethanol overnight and analyzed by propidium iodide ( PI ) staining and flow cytometry ( FACS Calibur , Becton Dickinson , Washington , DC ) as previously described ( Riccardi and Nicoletti , 2006 ) . Cytosolic extractions , sucrose gradient sedimentation of polysomes , and analysis of the polysomes/mRNPs distribution of proteins were carried out as previously described ( Sun et al . , 2007; Dai et al . , 2012; Ingolia et al . , 2012 ) . Briefly , cells were incubated with 100 μg/ml of cycloheximide for 15 min . Cells were homogenized in polysome lysis buffer containing 30 mM Tris–HCl ( pH 7 . 4 ) , 10 mM MgCl2 , 100 mM KCl , 0 . 3% NP-40 , 100 μg/ml of cycloheximide , 30 units/ml RNasin inhibitor , 1 mM DTT , 1 mM phenylmethylsulfonyl fluoride , 0 . 25 μg/ml pepstatin A . After incubation on ice for 5 min , cell lysates were centrifuged at 1300×g at 4°C for 10 min . Supernatants were subjected to sedimentation centrifugation in a 10–50% sucrose gradient solution containing 30 mM Tris–HCl ( pH 7 . 4 ) , 10 mM MgCl2 , 100 mM KCl in a Beckman SW41 rotor at 37 , 000 rpm at 4°C for 2 hr . Fractions were collected and absorbance of RNA at 254 nm was recorded using BR-188 Density Gradient Fractionation System ( Brandel , Gaithersburg , Maryland ) to analyze the distribution of polysomes and monosomes as described ( Esposito et al . , 2010 ) . We adapted previously established whole cell assays using HPLC to determine cellular GTP level ( Di Pierro et al . , 1995; Nakajima et al . , 2010 ) . After 48 hr of growth , cells were treated with 2 μM INZ for 18 hr . Cellular nucleotides were extracted from cell monolayers by addition of ice-cold 80% acetonitrile for 1 hr . The extracts were centrifuged to pellet the cellular debris and the cleared supernatant was loaded to the SPE column ( SAX column , Sigma–Aldrich ) . The elutes were analyzed by Agilent 1100 series liquid chromatograph system with a C18 reversed-phase column ( Agilent Zorbax Extend-C18 , 5 µM , 4 . 6 × 150 mm ) . A gradient elution from 0%B to 50%B in 70 min was used at a flow rate of 1 ml/min ( solvent A: 0 . 05M KH2PO4 , 0 . 005M tetrabutylammonium , pH5 . 5; B: 50% acetonitrile in 0 . 05M KH2PO4 , 0 . 005M tetrabutylammonium , pH 7 . 0 ) . The GTP level was also analyzed by a nanoACQUITY UPLC/Synapt HDMS mass spectrometer ( Waters , Milford , Massachusetts ) using acetonitrile/water ( 0 . 05M NH4Ac buffer solution at pH = 5 . 5 ) as the mobile phase with a flow rate of 0 . 5 μl/min . | Cancer develops when cells lose the ability to control their own growth . About half of cancerous tumors carry a dysfunctional version of a protein called p53 , while the other half have defects in proteins that are important for p53's production and function . When a healthy cell is exposed to damaging chemicals or agents , the p53 protein triggers responses that are aimed at repairing the damage . However , if these attempts fail , p53 causes the damaged cell to essentially destroy itself . As defects in p53-controlled processes cause cells to grow unrestrictedly and can lead to cancer , it is a very attractive target for cancer therapies . Cancer drug developments have focused on both targeting p53 directly and targeting the proteins that work with p53 . Two proteins called Mdm2 and SIRT1 are of particular interest . Mdm2 binds to , inactivates , and leads to the degradation of p53 . SIRT1 can modify p53 and make it more accessible to Mdm2 , and is often found in very high levels in cancer cells . In 2012 , researchers identified Inauhzin as a small molecule that could potentially be used to treat tumors that still have a functional version of the p53 protein . Inauhzin was thought to work by inhibiting SIRT1 , which increases p53 levels—probably through its effects on Mdm2 . This restores the cell's ability to control its growth and to die if it is irreparably damaged . However , not all of this small molecule's effects on cells can be explained by its interaction with SIRT1 . Now Zhang et al . , including some of the researchers involved in the 2012 work , have investigated whether Inauhzin also interacts with other proteins in the cell; and Inauhzin was revealed to bind an enzyme called IMPDH2 . This enzyme is involved in making GTP—a small molecule that is involved in many important processes in living cells . Zhang et al . demonstrated that Inauhzin's effect on the IMPDH enzyme triggered a response that did not involve the SIRT1 protein , and that ultimately led to a decrease in Mdm2 activity and restored p53 activity . Cancer treatments often include a combination of drugs that target different proteins with the goal of reducing the likelihood of a tumor becoming resistant to the treatment . Inauhzin's effect on two different proteins that lead to p53 activation not only increases its potency , but also makes it less likely that drug resistance will develop . | [
"Abstract",
"Introduction",
"Results",
"and",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"cell",
"biology"
] | 2014 | The role of IMP dehydrogenase 2 in Inauhzin-induced ribosomal stress |
Messenger RNA function is controlled by the 3' poly ( A ) tail ( PAT ) and poly ( A ) -binding protein ( PABP ) . La-related protein-4 ( LARP4 ) binds poly ( A ) and PABP . LARP4 mRNA contains a translation-dependent , coding region determinant ( CRD ) of instability that limits its expression . Although the CRD comprises <10% of LARP4 codons , the mRNA levels vary >20 fold with synonymous CRD substitutions that accommodate tRNA dynamics . Separately , overexpression of the most limiting tRNA increases LARP4 levels and reveals its functional activity , net lengthening of the PATs of heterologous mRNAs with concomitant stabilization , including ribosomal protein ( RP ) mRNAs . Genetic deletion of cellular LARP4 decreases PAT length and RPmRNA stability . This LARP4 activity requires its PABP-interaction domain and the RNA-binding module which we show is sensitive to poly ( A ) 3'-termini , consistent with protection from deadenylation . The results indicate that LARP4 is a posttranscriptional regulator of ribosomal protein production in mammalian cells and suggest that this activity can be controlled by tRNA levels .
A key control element of the stability and translatability of eukaryotic mRNA is the 3' poly ( A ) tail ( PAT ) which can vary from ~25 to 250 nucleotides ( Mangus et al . , 2003; Eliseeva et al . , 2013 ) , and accommodate multiple molecules of PABP ( Baer and Kornberg , 1980 ) . PAT length is associated with translation efficiency in early development ( Subtelny et al . , 2014 ) , and for specific mRNAs in somatic cells ( Park et al . , 2016 ) . PABP interacts with multiple different proteins involved in mRNA translation and stability ( Mangus et al . , 2003; Ivanov et al . , 2016 ) . Several of these proteins , among which are key factors involved in mRNA 3' exonucleolytic deadenylation , translation initiation and termination , share a similar peptide sequence termed PAM2 that interacts with the C-terminal domain of PABP ( Xie et al . , 2014 ) . Some proteins with a PAM2 sequence including Paips 1 and 2 , LARPs 1 , 4 and 4B , also harbor other regions that interact with PABP ( Yang et al . , 2011; Tcherkezian et al . , 2014; Xie et al . , 2014; Fonseca et al . , 2015 ) . LARPs 1 , 4 and 4B associate with translating polyribosomes ( Schäffler et al . , 2010; Yang et al . , 2011; Tcherkezian et al . , 2014; Fonseca et al . , 2015 ) . The 'La module' of the eukaryote-ubiquitous nuclear La protein is comprised of a La motif ( LaM ) followed by an RNA recognition motif ( RRM ) that cooperate to form a RNA binding pocket that recognizes the extreme UUU-3'OH termini of RNA polymerase III transcripts and protects them from 3' exonucleases ( Bayfield et al . , 2010 ) . The La module has been highly conserved by a few distinct La-related proteins ( LARPs ) that arose during eukaryotic evolution but diverged in other features of their structure and function ( Bousquet-Antonelli and Deragon , 2009; Maraia et al . , 2017 ) . Of these , LARPs 1 , 4 , 4B and 6 are mostly cytoplasmic and associated with mRNAs ( Maraia et al . , 2017 ) . Yet , except for vertebrate LARP6 which binds a highly conserved stem-bulge-loop found in the 5' UTRs of three α-collagen mRNAs ( Cai et al . , 2010; Martino et al . , 2015; Zhang and Stefanovic , 2016 ) , details of RNA binding by the La modules of LARPs 1 , 4 and 4B , and how such binding may contribute to their activities are largely unknown ( Maraia et al . , 2017 ) . The La module-containing N-terminal domain of LARP4 has been shown to bind homopoly ( A ) whereas its full length homolog , LARP4B has been shown to bind A-rich , U-containing RNA ( Yang et al . , 2011; Küspert et al . , 2015 ) , consistent with key differences in their La motifs ( Bayfield et al . , 2010; Maraia et al . , 2017 ) ; their regulation also differs since LARP4 but not 4B mRNA is destabilized by TTP ( Mattijssen and Maraia , 2015 ) . LARP1 appears to bind poly ( A ) and to stabilize mRNAs containing the 5' terminal oligo pyrimidine ( 5'TOP ) motif ( Aoki et al . , 2013 ) which are comprised of mRNAs encoding ribosomal proteins ( RP ) , translation factors , PABP and other proteins ( Meyuhas and Kahan , 2015 ) . LARP1 is known to have two RNA-binding domains , a La module in its N-terminal half ( Nykamp et al . , 2008 ) and a C-terminal HEAT motif that directly binds the 5' m7G cap and pyrimidine tract of the 5'TOP motif ( Lahr et al . , 2017 ) . LARP1 may regulate a number of transcripts in addition to the 5'TOP mRNAs ( Blagden et al . , 2009; Tcherkezian et al . , 2014; Fonseca et al . , 2015; Mura et al . , 2015 ) . Coding region determinants ( CRD ) of instability have been found in a small number of mRNAs , including Fos and Myc ( Lemm and Ross , 2002; Chang et al . , 2004 ) ( reviewed in Lee and Gorospe , 2011 ) and for Fos involves interactions with PABP , translating ribosomes , and deadenylation ( reviewed in Chen and Shyu , 2011 ) . Accumulating evidence indicate that the overall fraction of optimal vs . suboptimal codons in a mRNA is a major determinant of mRNA decay in yeast ( Presnyak et al . , 2015 ) . A potential link between mRNA codon use and decay appears more complex in higher eukaryotic cells in which the relatively high content of 3′ UTR-destabilizing elements is a confounding issue ( reviewed in Chen and Shyu , 2017 ) . Unresolved issues include whether codon optimality plays a role in higher eukaryotes , and if so to what extent and to what degree do cellular tRNA dynamics including relative abundances as well as codon-anticodon restraints such as wobble decoding which slows translation elongation ( Stadler and Fire , 2011 ) vs . direct Watson:Crick ( W:C ) decoding , contribute ( Chen and Shyu , 2017 ) . We compared expression in HEK293 cells of the open reading frames ( ORFs ) encoding several proteins and found LARP4 to be uniquely low . Detailed characterization revealed a codon-specific , translation-dependent CRD of mRNA instability that comprises <10% of the LARP4 ORF that is a strong determinant of expression . We analyzed the unusual codon characteristics of this CRD and their matches to cellular tRNA levels which we determined for this study . Synonymous substitutions limited only to the CRD including wobble vs . W:C decoding were analyzed for effects on expression of full length LARP4 and correlations with cellular tRNA levels . The synonymous substitutions led to LARP4 expression levels over a >20 fold range with excellent correlation with tRNA levels and codon-anticodon restraints ( R2 = 0 . 9 ) . Furthermore , mild to modest overexpression of the most limiting cellular tRNA cognate to CRD codons increased LARP4 levels in a dose-dependent manner . For some CRD constructs , this tRNA led to increased LARP4 production without increasing mRNA levels , while for other , more codon-optimized constructs , it increased LARP4 protein as well as the mRNA levels . Increases in LARP4 levels by either synonymous codon swaps or tRNA overexpression revealed its dose-dependent activity to promote longer PATs on heterologous mRNAs with associated stabilization . These results and poly ( A ) binding data that indicate 3' end-specific recognition and suggest protection from deadenylation , point to mechanisms by which LARP4 promotes mRNA stability and potential control of RPmRNA by tRNA levels .
We cloned cDNAs for LARPs and La from their second codon to their stop codon into expression vector pFlag-CMV2 ( Figure 1a ) and transfected these into HEK293 cells with a plasmid encoding adenovirus VA1 RNA synthesized by RNA polymerase III as a control . LARP4 accumulated to much lower levels than any other , with LARP4B , a homolog of similar mass ( below ) , as the highest ( Figure 1b ) . Northern blotting showed that LARP4 mRNA accumulated to the lowest levels ( Figure 1c ) . VA1 RNA was increased by La due to direct binding , stabilization ( Rosa et al . , 1981; Francoeur and Mathews , 1982; Mathews and Francoeur , 1984 ) and longer half-life ( not shown ) ; although its levels were more similar with the other LARPs ( Figure 1c ) . LARP4 mRNA was less than LARPs 6 and 4B by ~100 and ~50 fold respectively ( Figure 1d ) , likely reflective of different stabilities of their coding regions . The F-LARP4 constructs ( Figure 1e ) were western blotted after transfection and exhibited the expected mobilities ( Figure 1f , g ) . Fragments 1–286 , 27–286 and 359–724 accumulated to higher levels than full length 1–724 , and fragments 27–724 and 1–430 ( Figure 1f ) . These data suggested a region within codons 287–358 as inhibitory to expression , which was confirmed by the internal deletion construct , Δ287–358 that was expressed at levels nearly as high as 359–724 and much higher than 1–724 and 1–430 ( Figure 1g ) . The differences in protein levels were generally reflected by the mRNAs ( Figure 1f , g lower ) . Thus , codons 287–358 of LARP4 mRNA contain a coding region determinant ( CRD ) that is inhibitory to expression . This comprises a tract of <10% of the ORF length that encodes part of LARP4 protein that is important for interaction with PABP , termed the PABP-binding motif ( PBM ) ( Yang et al . , 2011 ) ( below ) . We used an established β-globin ( βG ) reporter under transcriptional control of a tetracycline/doxycycline-responsive promoter in HeLa tet-off cells to examine mRNA decay ( Gossen and Bujard , 1992; Fialcowitz et al . , 2005 ) . The promoter is turned off upon addition of dox , and RNA is isolated at t = 0 and times thereafter to follow decay . During the 48 hr following transfection until t = 0 , mRNAs transcribed at the same rate but with different half-lives accumulate to different levels , each requiring 3 to 4 half-lives to reach steady state ( Ross , 1995 ) . We inserted the LARP4 CRD into the βG-wt reporter in two contexts , in the ORF preceding the stop codon or following it ( Figure 1h ) . When preceding the stop codon , the CRD led to t = 0 levels that were significantly lower as compared to placement after the stop ( Figure 1i upper panel , lanes 1 and 6 ) . Plotting triplicate time course data normalized to GFP mRNA showed that the CRD produced more instability as part of the ORF as compared to following it ( Figure 1j , WT in CDS vs . WT in 3'UTR ) . The βG mRNA with the WT CRD in the CDS was indeed translated into a longer protein than with no insert ( not shown ) . βG-wt with no insert yielded a half-life as expected ( Fialcowitz et al . , 2005 ) ( not shown ) , similar to WT CRD in the 3’UTR . We examined codon substitutions to the CRD . A synonymous codon-swapped ( CS ) CRD sequence inserted in the βG CDS increased reporter mRNA levels and half-life relative to the WT CRD in the CDS ( Figure 1i , j ) . To further characterize the CRD and distinguish if destabilization might be due to RNA structure , G+C content , or codon-specificity , we inserted the WT CRD in the βG CDS beginning with a +2 frameshift ( +2 FS ) which required mutations to convert premature stops to sense codons . This preserved 97% CRD sequence identity but with only ~10% codon sequence identity relative to the WT CRD . Similar to CS , the +2 FS largely reversed the inhibitory effect of the WT CRD ( Figure 1k ) . LARPs 4 and 4B share most amino acid and nucleotide homology in their La modules but less in other regions including the CRD which is only 52% nucleotide identical in this region ( Figure 2a ) . We replaced the CRD region of LARP4B with the LARP4 WT CRD or a CS CRD and examined LARP4B expression . The WT-CRD decreased LARP4B-CRD-WT levels relative to LARP4B WT ( Figure 2b , lanes 5 , 4 ) . Importantly , the CS CRD rescued the negative effect of WT-CRD ( Figure 2b , lanes 6 , 5 ) ; see quantitation in Figure 2c . Thus , the LARP4 CRD is recognized as inhibitory when transferred to a heterologous mRNA . When the test mRNA is normally expressed at much higher levels than LARP4 , as in this case for LARP4B , the CRD appears to have less effect than on the lower abundance LARP4 mRNA , but quantifications of duplicate experiments revealed that it nonetheless decreased expression of LARP4B mRNA by 2 . 5 to 3 . 5-fold ( Figure 2c ) . The native 3'UTR of LARP4 mRNA is 4 . 2 kb and contains A+U rich elements controlled by TNFα via TTP ( Mattijssen and Maraia , 2015 ) . We replaced the short ( 0 . 48 kb ) 3' UTR of pFlag-CMV2 LARP4 -WT and -CS with the LARP4 4 . 2 kb 3’UTR ( Figure 2d ) . The long UTR lowered expression as expected ( Figure 2d , lanes 2 , 3; normalized to GFP ) . However , the CS CRD rescued the negative effect of the WT CRD ( lanes 3 and 5 , Figure 2d ) . These effects were reflected by differences in accumulation of the corresponding mRNAs containing the WT CRD and CS CRD in the long UTR ( Figure 2e , upper panel , lanes 3 and 5 ) . Quantification of data from duplicate experiments for the long and short mRNAs are shown in Figure 2f . Although the long UTR attenuated the effect as compared to the short UTR , the WT CRD exerted a 5-fold decrease in mRNA levels in the context of its 4 . 2 kb 3'UTR relative to the CS CRD ( Figure 2f ) . In E . coli , yeast and some other organisms , tRNA gene copy number correlates with cellular tRNA abundance and codon use , whereas this correlation is low in humans ( dos Reis et al . , 2004 ) . The tRNA adaptive index ( tAI ) is a measure of tRNA use by mRNAs that is derived from tRNA gene copy number and codon-anticodon base-pairing strength including that which distinguishes wobble vs . direct W:C pairing . Unlike in yeast , significant numbers of tRNA genes are variably inactive in different mammalian cell types and/or under different conditions ( reviewed in Orioli , 2017 ) . tAI scores are correlated with codon use and gene expression in yeast but not in human cells ( dos Reis et al . , 2004 ) . Therefore , for the present study we determined tRNA levels in HEK293 cells by tRNA-HydroSeq ( Arimbasseri et al . , 2015; Gogakos et al . , 2017 ) and the read counts for each tRNA species ( Table 1 ) were used to derive cellular-tAI ( ctAI ) values . These values were incorporated into an algorithm that generated ctAI scores for mRNA ORFs relative to their optimal match to the HEK293 tRNA pool . Based on the tRNA read levels and this algorithm , we designed multiple additional CS constructs with synonymous mutations limited to the CRD region of full length LARP4 , for comparison to WT and our original CS construct , hereafter designated CS-R . Designations and descriptions of the other CS constructs are as follows: CS-B was predicted to be expressed higher than CS-R , CS-W was predicted to be expressed lower than WT , and CS-I was predicted to be expressed at an intermediate level between WT and CS-R; the pattern of relative protein levels of these were generally as predicted ( Figure 3a ) and generally reflective of their mRNA levels ( normalized to VA1 , also see 28S RNA , Figure 3b ) . These results prompted more extensive detailed analyses ( below ) . Additional CS constructs in full length LARP4 were added to the above set and all were analyzed quantitatively for mRNA expression and correlation with the ctAI scores of their CRD regions ( Figure 3c ) . The additional constructs differ from those in Figure 3a in a way that attempted to discern effects of wobble decoding which is prominent in the CRD ( below ) . Constructs , CS-W , CS-I , CS-R and CS-B contain a mixture of synonymous codon swaps , some of which require decoding by a different tRNA anticodon than the original codon and some of which must be wobble decoded by the same tRNA anticodon as the original codon . Construct CSc contains synonymous swaps to Thr , Pro and Ile codons relative to WT , that require decoding by different tRNA anticodons . By contrast , constructs CSa , CSb and CS-Tyr contain synonymous swaps limited to U-to-C substitutions at wobble positions of select codons such that the same tRNA anticodon is used for decoding but with a direct anticodon G match to the wobble base C rather than wobble G:U decoding . CSb differs from WT by 14 synonymous codon swaps . Ten of these are weak codons , composed of all-A + U ( UUU Phe and UAU Tyr ) in the WT CRD . CSa differs from CSb only by five additional U-to-C substitutions at other weak codons in the CRD , at each of the Asn codons ( AAU ) . Construct CS-Tyr differs from CS-B only at the 7 Tyr codons of the CRD; all of which are wobble UAU in CS-B , and in CS-Tyr they are UAC . For the Asn , Phe and Tyr codons , a single tRNA with G34 in the anticodon wobble position must decode both of their codons . Therefore , the U-to-C codon swap provides the only G:C base pair in these otherwise weak codons . The ctAI scores for the CRD regions of the constructs as well as the full length LARP4 constructs and some other proteins are in Table 2 , and the sequences of CRD CS sequences are in Figure 3—figure supplement 1 . As can be seen in Figure 3c , quantitative analysis of mRNA expression levels by all of the LARP4 constructs revealed very good correlation with HEK293 cell tRNA levels and decoding dynamics as represented by the ctAI scores , with R2 = 0 . 886 . The range of expression obtained by the CRD synonymous swaps among all nine constructs tested , CS-W to CS-Tyr , spanned >20 fold ( Figure 3c ) . Thus , although the CRD comprises <10% of LARP4 coding length , it is a significant determinant of its mRNA overall stability and translation . The CRD is a LARP4 mRNA feature that was localized by truncation and deletion constructs ( see Figure 1e ) and functionally analyzed thereafter . We developed a ctAI tool that calculates and plots a 10 nucleotide sliding window average translation proxy score along the length of an ORF . When applied to the human LARP4 sequence , the CRD appeared as a segment of high density low score clusters which otherwise occur relatively infrequently ( Figure 3d ) ( apart from a stretch between residues 12–43 which may represent a conserved initial ramp of low codon optimality common to proteins within their first 50 amino acids [Tuller et al . , 2010; Shah et al . , 2013] ) . Fine mapping revealed that the lowest scoring points of LARP4 which ranged from 0 . 3 to 0 . 21 corresponded to two clusters of codons near the beginning of the CRD , denoted by red brackets at the left of the lower part of Figure 3d . We sorted the HEK293 tRNA read counts into four bins ( Table 1 ) , each containing ~25% of the 45 tRNA anticodon species that decode the standard 61 sense codons ( Novoa et al . , 2012 ) ; bins 1 , 2 , and 4 contain 11 tRNAs and bin 3 contains 12 . This revealed a wide range of tRNA levels; bins 1–4 comprise 5 . 4% , 11% , 21% and 62% of total read counts , respectively ( Table 1 ) . The numbers 1 to 11 above the codons in the lower part of Figure 3d represent the eleven least abundant tRNAs in the HEK293 cells , all in bin-1 ( Table 1 ) . The number 1 to 4 four lowest tRNAs , ThrUGU , ProAGG , ThrCGU , and ProCGG , ranged from 6500 to 10 , 000 reads ( bin-1 ) and the four highest ( bin-4 ) from 160 , 000 to 340 , 000 ( Table 1 ) . To verify a subset of these by another approach , semi-quantitative northern blotting confirmed that tRNAs ThrUGU , ProAGG and PheGAA as well as SerUGA ( bin-3 ) were consistent with tRNA-Seq relative levels whereas TyrGUA appeared lower by northern which our data suggest may be due to base modification-mediated interference with probe hybridization ( not shown ) . Examination of the LARP4 CRD revealed multiple types of codon bias; only 33 of the 61 sense codons are found in the 71-codon long CRD ( Figure 3d ) . 42% of all CRD codons must be decoded by bin-1 tRNAs which comprise only 5 . 4% of total tRNA abundance ( Table 1 ) . By contrast , 80% of codons excluded from the CRD are cognate to tRNAs in bins 2–4 . Thus , the LARP4 CRD shows bias enrichment for codons cognate to low abundance tRNAs and bias for exclusion of codons cognate to high abundance tRNAs . Strikingly , several bin-1 cognate codons are clustered . Moreover , many of these must rely on wobble decoding ( red numbers , Figure 3d ) , which slows translation ( Stadler and Fire , 2011 ) . Some bin-1 clusters are flanked by or include other weak all-A +T codons ( Tyr or Asn ) that also require wobble decoding ( Figure 3d , indicated by γ , λ ) . As alluded to above , the genetic code contains seven sense codons with all-A-or-U nucleotides in the three positions ( the other is a stop codon ) . HEK293 cells contain tRNA anticodon species to decode 4 of these 7 by direct W:C pairing whereas the other 3 must be wobble decoded . For each of these three , Asn AAU , Phe UUU , and Tyr UAU , the corresponding amino acids are encoded by only one other codon , that which ends with C in the wobble position and is W:C decoded by the single tRNA that must decode both codons . The CRD is biased in all 3 of the all-A+U wobble codons , Asn AAU , Phe UUU , and Tyr UAU , relative to their stronger synonymous codons , AAC , UUC and UAC . Specifically , 5 of 7 Tyr codons in the CRD are UAU , all 5 Phe codons are UUU , and all 5 Asn codons are AAU ( Figure 3d ) , comprising NNU:NNC ratios of 5:2 , 5:0 , and 5:0 respectively . As the NNU:NNC ratios for these codon pairs range from 0 . 8:1 to 0 . 87:1 among all human ORFs ( Mauro and Chappell , 2014 ) , it is clear that the CRD is highly biased in its use of each of these three weak wobble codons . The CS constructs that differ from their parent construct only in U-to-C wobble positions CSb and CS-Tyr , have significant effects on ctAI ( Table 2 ) . Therefore , the data show that U wobble codon decoding is suboptimal in the CRD because strengthening these codon:anticodon pairings appear to be functionally relevant . Comparison of the WT and CSb constructs ( 14 U-to-C wobble substitutions ) revealed increase in LARP4 expression by 5-fold and this was further increased by additional Asn AAU codons to AAC in CSa . Separately , conversion of 7 Tyr UAU to UAC codons as reflected by CS-B vs . CS-Tyr , increased expression significantly ( Figure 3c ) . We conclude that the LARP4 CRD is highly enriched in weak wobble codons and other codons cognate to very low level tRNAs , likely to slow ribosomes ( Stadler and Fire , 2011; Shah et al . , 2013 ) ( Discussion ) . Comparison of the ctAI plots of translation proxy scores of the synonymous codon swapped CRDs of the LARP4-CS constructs allowed visualization at near-codon resolution of the influence of wobble vs . direct W:C decoding ( Figure 3—figure supplement 2 ) . For example , higher scores mapped to Asn codons ( asterisks , Figure 3—figure supplement 2b ) in the CSa plot relative to CSb and also to the Tyr codons of the CS-Tyr plot relative to CS-B ( Figure 3—figure supplement 2b ) . pUC plasmids containing human tRNA genes transfected into HEK293 cells led to 3–6 fold increases in the corresponding tRNAs ( Figure 3e , and data not shown ) . When cotransfected with LARP4-WT and GFP , tRNAThrUGU increased LARP4 and GFP levels ( Figure 3f ) while tRNAPheGAA , tRNAProAGG and tRNATyrGUA did not ( Figure 3f ) . Increase in GFP is consistent with a relatively high number of Thr codons in its mRNA and a limiting amount of cellular tRNAThrUGU . Because our data not shown indicated that different tRNA genes compete for expression , confounding the use of combinations thereof , we hereafter focused on overexpression of the single most limiting one , tRNAThrUGU . Figure 3g shows western blots after transfection of HEK293 cells with increasing amounts of tRNAThrUGU plasmid ( 0 , 2 , 4 and 6 ug; empty plasmid , ep = 0 ug ) along with LARP4-WT , CS-W , CS-I or CS-B . The basal levels of each LARP4 construct increased with increasing tRNAThrUGU and this also occurred for GFP ( Figure 3g ) . An increase in GFP also occurred with LARP4-CS-W which is less active than LARP4-WT suggesting that the tRNAThrUGU effect is independent of LARP4 ( corroborated in a later section ) . These data provide strong evidence that the low level of endogenous tRNAThrUGU is functionally limiting in these cells . Quantification of the response of the LARP4 constructs to tRNAThrUGU are shown in Figure 3h using the basal levels with empty plasmid set to 1 . LARP4-WT exhibited the greatest response , up to a six-fold increase , while CS-W , CS-I and CS-B were less responsive ( Figure 3h ) . This pattern suggests that WT LARP4 is programmed to be sensitive to limiting tRNA and that this reflects the unique composition of its CRD . We also examined the effects of tRNAThrUGU on the LARP4 construct and GFP mRNAs as well as endogenous ribosome protein L32 ( Rpl32 ) ( Figure 3i ) . Remarkably , tRNAThrUGU did not increase LARP4-WT and CS-W mRNA levels ( Figure 3i , lanes 5–8 and 13–16 ) despite the increase in their protein products ( Figure 3g , h ) . This suggests that tRNAThrUGU increased the translational efficiency ( protein/mRNA ) of LARP4-WT , perhaps similar to that observed for HIS3 constructs with synonymous codons in yeast ( Presnyak et al . , 2015 ) . By sharp contrast to LARP4-WT and -CS-W mRNAs , the tRNAThrUGU clearly increased the levels of LARP4-CS-B mRNA in a dose-dependent manner ( Figure 3i , lanes 9–12 ) , and to a lesser degree -CS-I ( lanes 17–20 ) . The quantifications are shown in Figure 3j . The data in Figure 3g–j indicate that while overexpression of a single limiting tRNA can increase production of LARP4 protein from LARP4-WT mRNA , it does not lead to increased accumulation of this mRNA . This suggests that overcoming the destabilizing effects of the WT CRD with its intricate codon context ( Figure 3d ) may be too complex for resolution by a single limiting tRNA . Yet , LARP4-CS-B mRNA was increasingly stabilized by tRNAThrUGU and this was observed for CS-I although less so than for CS-B . We propose that because CS-I and CS-B CRDs contain more optimal synonymous codons than -WT and -W , they are more receptive to benefit from the accumulation effects of tRNAThrUGU . These analyses indicate that the LARP4 mRNA CRD can respond to a single limiting cognate tRNA with increased protein production , and moreover , that there may be a separate signal ( s ) in the CRD , apparently more complex than the Thr codons alone , that controls its instability determinant . Features in the patterns of GFP and Rpl32 mRNAs in Figure 3i are noteworthy because as will be shown in the next section they are relevant to LARP4 activity . Figure 3i revealed reproducible upward mobility shift of GFP mRNA , most readily appreciated by comparing lanes 4 and 5 and 12 and 13 . The mobility shift was dependent on cotransfected LARP4 since lanes 1–4 did not show it . In addition , there was gradual but reproducible upward shift in the GFP mRNA band observable in lanes 5–8 , 9–12 and 17–20 in response to increasing tRNAThrUGU . This was specific to cotransfected LARP4 because lanes 1–4 did not reveal it . We note that LARP4 CS-W did not reveal the upward shift of GFP mRNA probably because the levels of its LARP4 protein product were too low ( Figure 3g ) . The GFP mobility shift is not due to electrophoresis or other artefact but as will be documented below results from a specific activity of LARP4 . The same general pattern observed for GFP was apparent for Rpl32 mRNA although less distinctly ( Figure 3i ) . Another feature of Rpl32 mRNA is notable by comparing the vertical distribution of the bands within the lanes . Inspection of lanes 12 and 13 reveals the former more widely distributed than the latter , and similar for lane 5 vs . lane 4 . As will be shown in the next and following sections , this reflects increasing PAT lengths . The robust difference in the levels to which LARP4-WT and LARP4-CS-R accumulate revealed a dose-dependent shift in GFP mRNA relative to empty vector ( Figure 4a ) . This also was observed for endogenous rpRpl32 ( Figure 4a ) as well as Rpl35 , Rps27 , and FAIM mRNAs ( not shown , see below ) . The upward shifts also occurred with LARP4B ( Figure 4a ) . GFP was more shifted than Rpl32 and other mRNAs ( Figure 4a ) . Specifically , little of the shortest length GFP mRNA was present with LARP4 CS and LARP4B whereas short forms of Rpl32 mRNA remained . We believe this reflects that GFP mRNA PATs that were newly synthesized after transfection had not undergone shortening in the presence of ectopic LARP4 , whereas shortened forms of cellular Rpl32 mRNA preexisted upon transfection with LARP4 . RNase H+ oligo ( dT ) treatment of RNA followed by northern blotting can reveal PAT length differences of specific mRNAs ( Shyu et al . , 1991 ) ( Figure 4a , lanes 5–12 ) . The mRNAs were converted to the same faster mobility form after cleavage by RNase H+ oligo ( dT ) ( Figure 4a , lanes 9–12 ) , indicating that their mobility differences were due to differences in the PATs . We next examined the regions of LARP4 necessary for GFP mRNA PAT lengthening ( Figure 4b , upper panel ) . LARP4 constructs that were mutated to debilitate binding to PABP by two motifs , PAM2 and PBM , as well as a mutant designated LARP4-M3 with point mutations to five residues in the LaM and two residues in the RRM of the La module had been described and characterized ( Yang et al . , 2011 ) . Here we created the CS-R versions of those mutants that contained the CRD ( Figure 4b , western ) . The upper panel of Figure 4b shows that LARP4 ΔPAM2 was partially active for PAT lengthening as evident by less GFP shift than LARP4 CS but more active than WT . ΔPBM exhibited less activity than ΔPAM2 whereas ΔPAM2-ΔPBM was similar to ΔPBM . Truncations 1–286 and 359–724 , both lacking the PBM/CRD were qualitatively comparable to ΔPBM in shift mobility . The full length M3 LaM-RRM mutant disabled the shift activity in the low and high expression versions , WT and CS , lanes 9 and 10 respectively , consistent with its documented diminished association with PABP and polysomes ( Yang et al . , 2011 ) . A similar trend was seen for Rpl32 and Rpl35 mRNAs ( Figure 4b , and not shown ) although as noted , their shifts are less distinct . We sometimes observed increased intensity of GFP mRNA signal with some LARP4 constructs without accompanying mobility shift , for example with the ΔPBM overexpressed proteins ( Figure 4b ) . However , because a transcript requires 3–4 half-lives to achieve steady state ( Ross , 1995 ) , the relatively long lived GFP mRNA makes it unsuitable as an accurate reporter of stability for these transfection experiments . Therefore , we examined a β-globin ( βG ) mRNA reporter with a short half-life , βG-TNFα-ARE containing a destabilizing A+U rich ( ARE ) element from tumor necrosis factor ( Fialcowitz et al . , 2005 ) . HeLa tet-off cells were cotransfected with βG-TNFα-ARE , GFP , and the test plasmids that are indicated above Figure 4c . We included LARP6 as a control ( Ysla et al . , 2008 ) . As indicated by the quantification values under the t = 0 lanes 1 , 5 , 9 and 13 of the top panel of Figure 4c , LARP4 CS led to more βG mRNA accumulation than LARP4 WT , LARP6 and empty vector . We also noted that the upper edge of the βG mRNA band shifted down after t = 0 with vector and LARP6 , consistent with PAT shortening ( Ford et al . , 1999; Lai et al . , 2005 ) , but not with LARP4 WT and CS which maintained the longer forms ( Figure 4c ) . The data from the time course of βG mRNA decay were plotted in Figure 4d with the t = 0 values set to 100% . There was ~50% decrease in βG mRNA after 70 min with empty vector , in agreement with previous results ( Fialcowitz et al . , 2005 ) . By contrast , mRNA stability was substantially increased by LARP4 as a decrease of 50% was observed at 240 min with LARP4 WT , and this was extended by LARP4 CS to ~65% remaining at 240 min ( Figure 4d ) . The levels of the test proteins in this experiment are shown in Figure 4e . As the GFP and Rpl32 mRNA mobility differences observed with LARP4-WT and -CS may reflect their relative activity levels for PAT lengthening ( Figure 4a ) , a generally similar pattern was observed for their relative stabilization of βG-TNFα-ARE mRNA ( Figure 4d ) . The t = 0 data indicated that LARP4-CS increased βG-TNFα-ARE mRNA levels ~1 . 6 fold relative to LARP4-WT; consistent with this , extrapolation of the decay data suggested that LARP4-CS extended the mRNA half-life by ~1 . 5 fold relative to LARP4-WT ( not shown ) ( Ross , 1995 ) . We wanted to analyze βG-TNFα-ARE expression in HEK293 cells in which tRNA dynamics were characterized but these cells do not have the tetracycline transactivator that could be used to shut off the promoter . We therefore cloned the βG-TNFα-ARE into a constitutive CMV expression plasmid and analyzed the reporter mRNA 48 hr after transfection , that is , comparable to t = 0 in the previous experiments . This allowed us to examine effects of LARP4 subregions on accumulation of βG-TNFα-ARE mRNA in these cells . This revealed that full length LARP4 WT and CS produced progressively more mRNA than empty vector consistent with the relative amounts of their protein products ( Figure 4f ) . It is notable that the distribution of the βG mRNA in the empty vector was shifted upward by LARP4-WT at both its lower and upper edges ( Figure 4f , compare lanes 1 and 2 ) . The ΔPAM2 protein was expressed at higher levels than CS in this experiment and also led to higher βG mRNA levels ( Figure 4f ) . By contrast , ΔPBM and ΔPAM2-ΔPBM produced significantly less βG mRNA than ΔPAM2 when expressed at comparable levels ( Figure 4f ) ; this is consistent with the PAT lengthening activity of the ΔPAM2 protein observed for GFP mRNA ( Figure 4b ) . Thus , the PBM would appear to contribute more to PAT-mediated mRNA stability than does PAM2 . Finally , the two M3 LaM-RRM mutants accumulated less βG-TNFα-ARE mRNA than their intact-LaM-RRM counterparts including when M3 LaM-RRM CS version was expressed as high as LARP4-CS ( Figure 4f , lanes 3 and 8 ) . We also note that the GFP mRNA appears to report more incremental changes in PAT length among different LARP4 constructs than does the βG-TNFα-ARE mRNA which may be due to the poly ( A ) -destabilizing effect of the ARE ( Ford et al . , 1999; Lai et al . , 1999; Lai et al . , 2014 ) . We also observed LARP4-dependent upward shift of nanoluciferase mRNA , increased mRNA levels and increased nanoluciferase activity ( not shown ) . The cumulative data suggest that LARP4 binds to poly ( A ) and PABP , and protects mRNA from deadenylation , resulting in apparent net increase in PAT length and stabilization . We next examined HEK293 cells for effects of tRNAThrUGU on LARP4-mediated βG-TNFα-ARE mRNA accumulation . Cells were transfected with the βG-TNFα-ARE reporter , pUC19 ( ep ) or tRNAThrUGU plasmid in combination with LARP4-WT , LARP4-M3 LaM-RRM mutant , or empty expression plasmid , and GFP , as indicated above the top panel of Figure 5a . In this experiment , tRNAThrUGU was increased 5–6 fold ( Figure 5a ) . Protein levels are shown in Figure 5b . βG-TNFα-ARE mRNA was quantified in triplicate experiments ( Figure 5c ) . Overexpression of tRNAThrUGU led to an upshift of βG-TNFα-ARE mRNA and increased its levels ( Figure 5a , top ) that was specific to LARP4-WT even though the LARP4-M3 LaM-RRM mutant level was increased by tRNAThrUGU as expected ( Figure 5b ) . tRNAThrUGU stimulated a GFP upshift by LARP4-WT but not by LARP4-M3 LaM-RRM mutant ( Figure 5a ) . The GFP shift was modest here because the increase in LARP4 levels by tRNAThrUGU is not as much as the difference between LARP4-WT and LARP4-CS ( Figure 4a ) . tRNAThrUGU increased GFP protein independent of transfected LARP4 ( Figure 5b , lanes 1 , 2 ) , corroborating that tRNAThrUGU is functionally limiting in these cells . Beyond this , the data validate the ability of LARP4 to promote mRNA PAT length and that this is associated with βG-TNFα-ARE mRNA stabilization . Further , the data show that increases in this novel LARP4 activity can occur in response to elevation of the level of a limiting cellular tRNA . We wanted to examine the effect of the CRD by comparing the distributions of LARP4-WT and LARP4-CS-Tyr mRNAs in HEK293 cell polysome gradient sedimentation profiles in the presence and absence of over-expressed tRNAThrUGU . To achieve similar levels of the LARP4 -WT and -CS-Tyr mRNAs we transfected less of the latter plasmid than the former , maintaining equal amounts of total transfected plasmid and GFP controls ( methods ) . We first examined aliquots of the transfected cell extracts by western blotting ( Figure 6a ) ; F-LARP4 intensities in Figure 6a indicated that tRNAThrUGU stimulated LARP4-WT production more than it stimulated LARP4-CS-Tyr , whereas it stimulated GFP more equally . In addition to the newly synthesized proteins , LARP4 and GFP , endogenous actin and Rps6 were examined , and the blotted membrane was also stained for total protein with Ponceau S to show relative loading ( Figure 6a ) . Polysome sedimentation profiles of the extracts were prepared in parallel and RNAs from each of the fractions were examined by northern blotting ( Figure 6b–e ) . The distribution of a given mRNA species in a polysome sedimentation profile is determined in part by its rates of translation initiation , elongation and termination , as well as its overall length and codon length which limits the number of translating ribosomes . Several comparisons of the data collected are noteworthy . First , the polysome profile distributions revealed relatively more LARP4 -CS-Tyr than -WT mRNA in polysome fractions 8–14 than in fractions 3–7 ( in the absence of tRNAThrUGU , Figure 6b and d ) , consistent with more efficient engagement of ribosomes by LARP4-CS-Tyr than LARP4-WT . The LARP4 mRNAs in these profiles were comparable in overall levels ( Figure 6b , d ) ; quantification is shown as the percentage of total mRNA in each fraction ( Figure 6f ) . This revealed that a larger percentage of LARP4-CS-Tyr mRNA is in fractions 8–14 ( 66% ) as compared to LARP4-WT ( 57% ) , providing evidence to suggest that CS-Tyr mRNA is occupied more densely by ribosomes than the -WT mRNA . This further suggests that synonymous codon substitutions to the CRD not only increase the levels of the LARP4-CS-Tyr mRNA ( Figure 3c ) but also its translational efficacy ( Discussion ) . The next comparison reflects tRNAThrUGU effects on the polysome distributions of LARP4-WT and LARP4-CS-Tyr mRNAs ( Figure 6b–e ) . tRNAThrUGU shifted the LARP4 -WT and -CS-Tyr mRNAs toward heavier polysomes ( Figure 6g , h ) . LARP4 contains 46 Thr codons in addition to those in the CRD; the LARP4 -WT and -CS-Tyr were shifted by tRNAThrUGU to different degrees ( Figure 6g , h ) . We note that tRNAThrUGU led to higher polysome levels and higher levels relative to 80S peaks in the OD254 tracings as compared to the control plasmid pUC19 for both LARP4 -WT and -CS-Tyr ( top panels Figure 6c vs . b and e vs . d ) . This would not appear to be an artefact of polysome dissociation due to mishandling as reflected by comparable GAPDH mRNA profiles . For the mRNAs examined , the tRNAThrUGU and control blots were incubated with probes , washed and imaged together . Thus , the fractions from the tRNAThrUGU gradients ( c and e ) appear to contain more RNA than the pUC19 control gradients ( b and d ) as can be appreciated by comparing the H2A , Rps28 , GAPDG and EtBr panels ( Figure 6b–e ) . A striking shift of GFP mRNA to heavier polysomes was observed in cells transfected with tRNAThrUGU plasmid ( Figure 6b–e , i ) . It is important to note that Figure 5 ( and data not shown ) indicate that tRNAThrUGU is more effective than LARP4-WT at increasing GFP protein levels ( GFP western Figure 5b lanes 4 vs . 2 ) even though LARP4-WT increases GFP mRNA and its PAT length ( Figure 5a ) , suggesting that tRNAThrUGU promotes its translational efficiency . This is consistent with the polysome distribution of GFP mRNA toward heavier polysomes in the presence of tRNAThrUGU ( Figure 6b & c and d & e ) . Unlike LARP4 , the GFP construct is far better codon optimized throughout its length for expression in human cells ( Haas et al . , 1996 ) . Another consideration is that while the UTRs in the GFP and LARP4 constructs are comparably short , their ORF lengths differ at 240 and 724 codons respectively . Thus , there should be more potential homogeneity of ribosome-containing GFP mRNPs than LARP4 mRNPs . The more dramatic shift of GFP mRNA to heavy polysomes as compared to LARP4 mRNA with tRNAThrUGU over expression may reflect both parameters , more benefit from codon context and more efficient ribosome occupancy per ORF length . In any case , the data provide evidence to indicate tRNAThrUGU as limiting for translation of newly transcribed mRNA from transfected plasmids in these cells . We also probed for endogenous steady state mRNAs , graphically shown in Figure 6j–l . Both GAPDH and histone H2A were previously shown to be unaffected by ectopic LARP4 ( Yang et al . , 2011 ) . The non-polyadenylated histone H2A ORF is 130 codons and the peak of its mRNA was mostly in the polysome fractions . GAPDH mRNA ORF is 336 codons and the peak of its mRNA was localized in the heavy polysome fractions reflective of efficient translation in all cases . For both H2A and GAPDH there appeared to be a slight shift to lighter polysomes in Figure 6c as compared to Figure 6b but less so for Figure 6e vs . Figure 6d ( Figure 6j , k ) . The Rps28 ORF is 69 codons; its mRNA peak appeared to be more shifted to heavier polysomes by tRNAThrUGU for 6c vs . b and e vs . d ( Figure 6l ) as compared to H2A and GAPDH . We produced LARP4 gene-deleted embryos from which knock-out ( KO ) and sibling wild type ( WT ) MEFs were made ( Figure 7a ) . PAT length of RPmRNAs can be short ( Park et al . , 2016 ) which facilitates detection of net lengthening ( Figure 4a ) , but may impede detection of further shortening . The mRNA lengths from KO and WT MEFs were analyzed for electrophoretic mobility by northern blots in triplicate ( Figure 7b ) . Rps28 mRNA from KO migrated with focus toward faster mobility as compared to WT MEFs in which it was more widely distributed ( Figure 7b , and right , lane tracings ) . Rpl32 mRNA showed a similar pattern although its ORF + UTR length is greater than Rps28 and not quite as well resolved ( Figure 7b ) . RpmRNAs are stable and accumulate to high levels relative to many other mRNAs . We also probed for protein phosphatase 1 regulatory inhibitor-14A ( PPP1R14A , Figure 7b ) . The nonpolyadenylated histone H2A mRNA was comparable in KO and WT MEFs as expected ( Figure 7b ) . Oligo ( dT ) -directed cleavage of poly ( A ) RNA produced similar fragments from KO and WT MEFs for Rps28 and Rpl32 mRNAs , indicating that the mobility differences in the absence of oligo ( dT ) are due to PAT length ( Figure 7c ) . We treated cells with actinomycin-D to block transcription , and isolated RNA at 0 , 1 , 2 and 4 hr thereafter ( Figure 7d , e ) ; short-lived H2A mRNA showed that act-D was effective . This revealed greater mobility differences for Rps28 and Rpl32 mRNAs in KO relative to WT MEFs , and more concentration in the shorter forms over the time course ( Figure 7d , e ) . These data support a role for LARP4 in protection of mRNA PATs from 3' end shortening in vivo . Figure 7f shows probings of a northern blot of a 12 hr time course after act-D treatment for Rpl32 , Rpl41 and Rps28 mRNAs as well as 18S rRNA in LARP4 WT and KO MEFs . Later times were not examined because evidence of cell death was observed beyond 12 hr in act-D . Quantifications are shown in Figure 7g , using the 18S rRNA for normalization . The data revealed that RPmRNAs decay faster in LARP4 KO than in WT MEFs ( Figure 7g ) . A hallmark feature of the binding pocket of nuclear La protein ( Teplova et al . , 2006 ) is reflected by sensitivity of the RNA 3' terminal ribose 3'OH and 2'OH groups to chemical modifications ( Stefano , 1984; Terns et al . , 1992; Dong et al . , 2004; Teplova et al . , 2006; Kotik-Kogan et al . , 2008 ) . A previous study of LARP4 used a strong-hairpin RNA ( Yang et al . , 2011 ) which recent data suggest can interact with La module proteins via a binding mode that differs from single stranded RNA ( Martino et al . , 2012; Martino et al . , 2015 ) . We examined LARP4 for binding to single stranded A15 RNA with different 3' ends; 3'-OH , 3'-PO4 , and 2'-O-CH3 , and included U15 3'-OH as a sequence-specificity control . LARP4 showed highest avidity for A15 with 3'-OH , and progressively less for 3'-PO4 , 2'-O-CH3 and U15 3'-OH ( Figure 8a ) . Quantification is shown in Figure 8b . A second hallmark feature of the binding pocket of La for 3' end binding is sequence-specific recognition of the penultimate nucleotide which when substituted leads to loss of overall affinity ( Teplova et al . , 2006; Kotik-Kogan et al . , 2008 ) . Figure 8c , d indicate that LARP4 is sensitive to the penultimate A , i . e . , at position minus-2 ( A12AUA ) , significantly more so than at −3 ( A12UAA ) ; quantification in Figure 8e . While other features of RNA binding by La and LARP4 clearly differ in sequence specificity and RNA length requirement ( Yang et al . , 2011 ) and this remains to be understood at the structural level ( reviewed in Maraia et al . , 2017 ) , the data in Figure 8 demonstrate sensitivity of LARP4 to the poly ( A ) RNA 3' end in a manner similar to the La module RNA binding pocket of La protein during RNA 3' end sequestration ( Teplova et al . , 2006 ) and are consistent with a proposed mechanism for LARP4 for mRNA 3' PAT protection from deadenylation . LARP1 is known to bind , stabilize and regulate the translation of RPmRNAs ( Tcherkezian et al . , 2014; Fonseca et al . , 2015; Lahr et al . , 2017 ) . It was reported that LARP1 could recognize the extreme 3' terminus of poly ( A ) with sequence specificity for 3' A , in an extract based system ( Aoki et al . , 2013 ) , and directly bind PABP ( Blagden et al . , 2009; Burrows et al . , 2010; Tcherkezian et al . , 2014; Fonseca et al . , 2015 ) . We found that LARP1 produced a βG-TNFα-ARE length shift ( Figure 9a ) and a GFP mRNA shift comparable to LARP4 -WT and -CS ( Figure 9b , lanes 1–4 ) . RNase H + oligo ( dT ) demonstrated that this was due to net increase in mRNA 3' PAT length ( Figure 9b ) . Histone H2A mRNA exhibited insensitivity to RNase H + oligo ( dT ) as expected ( Figure 9b ) . The proteins expressed in this experiment are shown in Figure 9c .
It is remarkable that LARP4 mRNA levels were ~50 fold lower than LARP4B ( Figure 1d ) and that a substantial part of this could be rescued by synonymous codon substitutions to the CRD which comprises less than 10% of the coding region ( Figure 3c ) . Other data demonstrated that the LARP4 CRD conferred significant instability when transferred to the higher level LARP4B mRNA ( Figure 2b , c ) . It was recently noted that while much understanding of codon optimality has come from studies of yeast , outstanding issues include whether codon optimality plays a major role in mRNA decay in higher eukaryotic cells whose 3′ UTRs tend to contain a plethora of regulatory elements ( Chen and Shyu , 2017 ) . Our data showed that the LARP4 CRD was a significant instability element in the context of the LARP4 mRNA 4 . 2 kb 3' UTR ( Figure 2d–f ) , which was previously documented to harbor negative regulatory elements responsive to TNFα and TTP ( Mattijssen and Maraia , 2015 ) . Results from yeast indicate that codon optimality is a major determinant of mRNA stability as reflected by the overall percentage of optimal versus suboptimal codons of any given mRNA ( Presnyak et al . , 2015 ) . Our data showing that the LARP4 CRD confers major instability despite its length of only 10% of the ORF , together with our analysis of its unusual codon bias including density of clusters of suboptimal codons ( Figure 3d ) argue that the CRD is a potent modular element of instability . These findings suggest that codon control of mRNA stability may reside in distinct regions of higher eukaryotic mRNAs . However , the extent to which this may be so and/or the types of subsets of mRNAs involved , if any , may be determined by future studies . The LARP4 CRD system , approaches and tools developed here should be useful toward addressing these issues . Another outstanding issue in codon optimality is the potential role of the cellular tRNA pool and codon-anticodon dynamics ( Chen and Shyu , 2017 ) . This has been challenging in higher eukaryotes because unlike in yeast , tRNA gene copy numbers do not correlate with codon use by efficiently translated mRNAs ( dos Reis et al . , 2004 ) . This may reflect that significant numbers of tRNA genes of variable identities are inactive in different mammalian cell types ( reviewed in Orioli , 2017 ) . Our data demonstrated that the LARP4 CRD is a determinant of mRNA instability via its codon-specific match to HEK293 cell tRNA levels and their codon-anticodon dynamics . We analyzed several full length LARP4 expression constructs that differ only in the synonymous codon composition of the CRD region . It was informative to consider effects of synonymous CRD substitutions on LARP4 mRNA expression in conjunction with cellular tRNA levels and their codon-anticodon dynamics . This revealed that codons that must be decoded by very low abundance tRNAs are concentrated in the LARP4 CRD along with a heavy bias of weak , all-A+U codons that must be wobble decoded , and clusters of these ( Figure 3d ) . Two relevant findings are noteworthy . First , the CRD codon composition in conjunction with the HEK293 tRNAs predict that certain synonymous swaps cannot significantly improve optimality because all of the tRNAs for that amino acid are very low abundance . Specifically , all the tRNAs Thr , Pro and Ile for all Thr , Pro and Ile codons , are in bin-1 ( Table 1 ) . Therefore , the Thr , Pro and Ile codon synonymous swaps to a different tRNA anticodon would have only limited effect toward increasing LARP4 mRNA levels , as was observed for construct CSc which contains 13 synonymous substitutions and collectively increased levels 2 . 4-fold . The other noteworthy finding resulted from codon swaps of weak , all-A+U wobble codons to their stronger synonymous codons with a C in the third position . In these cases the new codon is decoded by the same tRNA but using a direct match anticodon wobble G34 . These data revealed that wobble decoding of weak codons can be a significant determinant of suboptimality ( Figure 3c ) . This latter point was demonstrated by comparing constructs CS-B and CS-Tyr that differ only in the third positions of all seven Tyr codons ( UAU vs . UAC ) in the CRD ( Figure 3c ) , both of which are decoded by tRNATyrGUA . Also , LARP4-CSb differs from LARP4-WT in 14 U-to-C synonymous codons that are wobble decoded in -WT but decoded by C:G codon:anticodon base pairs in -CSb , and increased expression 5-fold ( Figure 3c ) . Our analysis revealed that tRNAThrUGU led to increased protein production from LARP4-WT without increasing the mRNA levels , whereas some of the CS CRD constructs responded with tRNA dose-dependent increase in mRNA levels . This suggests that translation dependency and the instability component of the CRD can be uncoupled and that reversing the latter may require more than an increase in the levels of a single tRNA . A significant component of this study was quantitative sequencing of HEK293 cell tRNAs and development of ctAIs . This led to the finding that the tRNAs for all Thr , Pro and Ile codons are of very low abundance , in bin-1 ( Table 1 ) . A striking finding was demonstration that over expression of the lowest abundance , tRNAThrUGU , increased production of LARP4 and GFP from transfected plasmids ( Figures 3g–h and 5b ) . As expected , this was accompanied by a shift of the corresponding mRNAs to heavier polysome fractions , presumably reflective of greater ribosome occupancy ( Figure 6g–i ) . This was a robust activity that was specific since over expression of other low ( or high ) abundance tRNAs including tRNAThrCGU or tRNATyrGUA did not increase LARP4 or GFP production ( Figure 3e , f and data not shown ) . The responsiveness of GFP mRNA to tRNAThrUGU was remarkable ( Figure 6b–e , i ) . We note that 14 of the 15 threonines in the codon-optimized GFP construct ( Haas et al . , 1996 ) are encoded by ACC codons and that these require wobble decoding because there is no tRNA with a GGU anticodon ( Table 1 ) ( Gogakos et al . , 2017 ) . While the wobble base modification status of human tRNAThrUGU has not been reported , in yeast it carries the ncm5U modification ( Johansson et al . , 2008 ) . This same ncm5U modified base on tRNAProUGG has been shown to wobble decode the Pro codon with C in the wobble position ( Johansson et al . , 2008 ) . Understanding the determinants of responsiveness of a mRNA to tRNAThrUGU in this system will be a goal of future investigations . While there is a large difference in the instability of the LARP4-WT and LARP4-CS-Tyr mRNAs as the latter accumulates to ~12 fold higher than the former ( Figure 3c ) , we note that we do not know the source , in vivo kinetics , cell biology , specific factors involved nor the mechanism by which the CRD mediates the effect . Although our data on CRD codon composition and cognate tRNA levels would suggest that the mechanism is linked to slow translation ( Radhakrishnan and Green , 2016 ) , some points are nonetheless noteworthy . As was evident from our early analysis of Figure 1e–g , the different LARP4 truncation and deletion construct mRNAs containing or lacking the CRD would appear to have similar apparent translation efficiencies because the relative amounts of the mRNAs more or less match the relative amounts of proteins produced ( Figure 1e–g ) . Moreover , this trend of match between relative amounts of mRNAs and proteins produced was maintained by the synonymous CS constructs analyzed in Figure 3a–c ( and data not shown ) . It is also interesting to consider that although LARP4-WT mRNA accumulates to ~12 fold lower levels than LARP4-CS-Tyr mRNA , the LARP4-WT mRNA that does survive appears to exhibit translational efficiency that is only modestly lower than LARP4-CS-Tyr mRNA based on polysome profile distributions ( Figure 6b–e ) . This is not inconsistent with the observations supporting the generally similar apparent translation efficiencies of CS construct mRNAs made above . The data are consistent with some type of quality control mechanism that leads to decay of a large fraction of the LARP4-WT mRNA because it has suboptimal codons ( Radhakrishnan and Green , 2016 ) ( see Brule and Grayhack , 2017 ) . To gain insight into mechanism , we performed in vitro translation in extracts made from our HEK293 cells programmed with 7mG capped and polyadenylated mRNAs synthesized by T7 RNA polymerase ( Rakotondrafara and Hentze , 2011 ) . Using equal amounts of synthetic transcripts corresponding to LARP4 WT and CS-Tyr mRNAs we sought to observe evidence of ribosome stalling in the CDR in the form of transiently arrested nascent polypeptides . After preliminary experiments revealed no difference in production of the polypeptides from the LARP4-WT and -CS-Tyr mRNAs in standard reactions , we performed time courses to more carefully focus on transition through the CRD region , codons 286–358 . For these experiments we compared LARP4-WT ( 1-358 ) and LARP4-CS-Tyr ( 1-358 ) fragments because in vivo analysis showed that the WT CRD in the LARP4 ( 1–358 ) construct was highly active ( Figure 1e ) and because this approach facilitated the in vitro analysis . However , there was comparably robust translation through the CRD regions of both of the synthetic mRNAs ( Figure 6—figure supplement 1 ) . Additional attempts to elucidate a difference in the in vitro translation of the two mRNAs by decreasing the temperature of the reactions were unsuccessful . Although the polysome distributions of LARP4-WT and LARP4-CS-Tyr ( Figure 6 ) were consistent with lower translational efficiency of WT , we do not know the degree to which such a difference might be expected to manifest as a difference in these in vitro translation reactions . The in vitro translation results suggest among other things the possibility that the mechanism controlling the synonymous codon-specific differential expression/decay of LARP4-WT and -CS-Tyr in cells is coupled to transcription or another nuclear event ( s ) or process that is not faithfully executed during in vitro translation in extract of synthetic mRNAs . Such a possibility would be consistent with a quality control mechanism that is operational in vivo . Prior to this work , a mechanism by which LARP4 or 4B may function in mRNA metabolism was unknown . We also note that sensitivity of single stranded poly ( A ) RNA to 3' end binding by LARP4 had remained untested . We used in vitro RNA binding to reveal that LARP4 exhibits sensitivity to the 3' end of poly ( A ) in a manner similar to that of the RNA binding pocket of La protein during RNA 3' end sequestration ( Teplova et al . , 2006; Kotik-Kogan et al . , 2008 ) . We note that this apparent similarity occurs despite other significant differences in RNA recognition by La and LARP4 including oligo ( U ) vs . oligo ( A ) specificity , RNA length requirements ( Yang et al . , 2011 ) , and that the molecular and structural bases of these for LARP4 are unknown ( reviewed in Maraia et al . , 2017 ) . Nonetheless , this 3' end sensitivity is consistent with a proposed mechanism for LARP4 for mRNA PAT 3' end protection from deadenylation . In any case , future experiments toward understanding how the PABP-interaction domains of LARP4 cooperate with its La module will be necessary to better understand its activity for PAT lengthening . In the current working model , the La module of LARP4 would bind poly ( A ) , and its PBM and PAM2 would bind PABP , the latter of which also binds the PAT . Presumably , the LARP4 PAM2 would serve to compete with or displace from PABP , the PAM2-containing deadenylases which function as 3' exonucleases . MEFs derived from LARP4 gene-deleted KO embryos created for this study were shown to bear RPmRNAs with shorter PAT length and faster decay than in WT MEFs . The mRNA-PAT length maintenance mediated by LARP4 characterized here was mostly for the highly abundant RPmRNAs . However , we wish to emphasize that LARP4 activity to increase PAT length was not limited to these mRNAs . Elevation of cellular LARP4 levels in response to increase in limiting tRNA or other means led to PAT lengthening and stabilization of βG-TNFα-ARE mRNA whose ARE directs deadenylation . Yet LARP4 would appear to differ from other mRNA stabilizing proteins which generally target lower abundance and transiently expressed mRNAs that are relatively unstable in their basal state ( as compared to RPmRNAs ) and regulated through their 3'UTRs , e . g . , via AREs that indirectly modulate PAT metabolism via trans-acting factors ( Chen and Shyu , 2017 ) . According to the model derived from our data , LARP4 differs because it binds directly to poly ( A ) , the RPmRNAs are abundant and contain relatively very short 3' UTRs that are generally believed to be non-regulatory in the conventional sense . Thus , LARP4 would appear to be a general factor that is more directly involved in PAT length maintenance . The RPmRNAs comprise a substantial fraction of cellular mRNA in proliferating cells , and are critical and tightly regulated , including under growth control . Our data demonstrated mRNA PAT net lengthening activity for two other La module proteins , LARP4B and LARP1 ( Figure 4 and 9 ) , the latter of which is a regulator of ribosome biogenesis and a pro-cancer protein ( Aoki et al . , 2013; Tcherkezian et al . , 2014; Fonseca et al . , 2015; Lahr et al . , 2017 ) . By contrast , neither over expression of La protein , a significant fraction of which is cytoplasmic , nor LARPs 6 or 7 , exhibited this activity even when accumulated at higher levels than LARP4-WT and -CS-R ( data not shown ) . LARP4B contains a PAM2 and a separate PBM ( Bayfield et al . , 2010; Schäffler et al . , 2010 ) , and recent data identified a PAM2 candidate in LARP1 ( Fonseca et al . , 2015 ) . LARP1 is a central factor in the regulation of translation of RPmRNAs in response to nutrition-related signals by controlling their repression which is mediated by its DM15 domain which binds their 5'TOP motif ( Lahr et al . , 2017 ) . However , a recognized part of LARP1 activity in regulation is RPmRNA stabilization ( Tcherkezian et al . , 2014; Fonseca et al . , 2015 ) and some data suggested that this may be mediated via binding to the 3′ terminus of the PAT ( Aoki et al . , 2013 ) . Our data showed that LARP1 stabilized/increased accumulation of βG-TNFα-ARE mRNA with accompanying mobility shift and it similarly shifted GFP mRNA which was shown to result from PAT lengthening ( Figure 9 ) . As neither of these mRNAs bear a 5'TOP motif , the data suggest that the PAT-mediated mRNA stability may be a separate activity of LARP1 . This further suggests that PAT-mediated mRNA stability by LARP1 may be uncoupled from translation repression including when LARP1 is over-expressed as occurs in certain cancers and associates with non-TOP mRNAs ( Mura et al . , 2015; Stavraka and Blagden , 2015; Hopkins et al . , 2016 ) . However , while LARP1 contributes to RPmRNA translation as a central negative regulator , the present work suggests that LARP4 may be more of a constitutive positive factor in the control of RPmRNA homeostasis . Finally , an intriguing aspect of this work is that tRNA availability might control or tune LARP4 activity levels . Because this could impact RPmRNA translation , it could theoretically represent signaling of tRNA availability to ribosome biogenesis .
HeLa Tet-Off ( Clontech ) and HEK293 were maintained in DMEM plus Glutamax ( Gibco ) supplemented with 10% heat-inactivated FBS ( Atlanta Biologicals ) in a humidified 37°C , 5% CO2 incubator . Cultures were passed every 2–3 days . The HeLa Tet-Off cells were a gift obtained directly from Gerald Wilson ( U Maryland , Baltimore ) . HEK293 cells were obtained from Tazuko Hirai in Bruce Howard's laboratory at the NIH . Our HEK293 and HeLa cell DNAs were both authenticated by the ATCC via STR profiling . HeLa Tet-Off cells are not commonly misidentified cell lines as listed by the International Cell Line Authentication Committee . Standardized mycoplasma testing ( ATCC ) was performed and tested negative . All mouse studies were performed at the NIH under protocol ASP 10–005 and approved by the IACUC of NICHD . The targeting vector HTGR06019_A_6_DO1 was generated by the trans-NIH Knock-Out Mouse Project ( KOMP ) and obtained from the KOMP Repository ( www . komp . org ) . After exon 5 is floxed out , a termination codon will be encountered 16 codons following the La motif . The vector was linearized with AsiSI and electroporated into mouse embryonic stem cells . Neomycin-resistant colonies were isolated and scored for homologous integration by PCR amplification . Targeted clones were injected into C57BL/6 blastocysts and chimeric founder mice crossed with C57BL/6 females to establish the Larpfloxedfloxed+Neo line . Mice heterozygous for this Larp4floxed allele were crossed with EIIa-cre transgenic females to remove Larp4 sequences between the 2 loxP insertions . No backcrossing was performed . All mice are maintained in microisolator caging within ventilated racks ( Lab Products ) . Caging systems are changed once a week; cage tops and wire lids are changed every other week . The mice are fed NIH Autoclavable Rodent Chow . Mice are given chlorinated water in water bottles . cDNA was generated from HeLa cells and used to amplify the coding regions of human La , LARP4 , LARP4B , LARP6 and LARP7 starting at the second codon . The PCR products were cloned into the HindIII and BamHI sites of the pFLAG-CMV2 vector ( Sigma-Aldrich ) . LARP4 truncation constructs were derived from the full length Flag-LARP4 vector . A deletion construct lacking codons 287–358 was purchased from Genewiz and subcloned into HindIII and BamHI of pFLAG-CMV2 . Each construct was confirmed by sequencing . Full length LARP4 with codons 287–358 of LARP4-WT swapped for synonymous codons ( CS-I , CS-B , CS-W and CS-R , CSb , CSc ) were obtained from Eurofins and subcloned into the HindIII and BamHI sites of the pFLAG-CMV vector . The LARP4 4 . 2 kb 3’-UTR was amplified by PCR from HeLa DNA using primers KpnI-UTR-Fwd 5’TAGGGCGGTACCAAAACAACAAAACTATTC-AAAAACTTCAC and XmaI-UTR-Rev 5’AATGTACCCGGGTTTTTTTTTTTTTTTTTTTCTGCTTTTTAATAATTTTATTTTTTTTCTAATTTTGTTAATTTCCCATAGCACC . The LARP4 WT or CS coding region was amplified using HindIII in the forward primer and KpnI in the reverse primer . After restriction digestion with KpnI , HindIII and XmaI , the CDS and 3’UTR were ligated and inserted into HindIII and XmaI sites of pFlag-CMV2 vector ( Sigma-Aldrich ) . Constructs containing 3 copies per plasmid of each tRNA gene for TyrGUA , PheGAA , ThrUGU or ProAGG , in pUC57-Kan were obtained from Genewiz . For each tRNA copy , 150 nt upstream and 90 nt downstream genome sequence was included . Codons 329 ( V ) through 393 ( K ) of LARP4B were replaced by codons 287–358 of LARP4 ( WT or the CS-B sequence , were obtained from Genewiz and subcloned into HindIII and BamHI sites of pFlag-CMV2 vector ( Sigma-Aldrich ) . The pTRERβ-wt , encoding a rabbit β-globin minigene under control of a tetracycline-responsive promoter ( also known as βG-wt ) and pTRERβ-TNFα-ARE ( containing the 38 nt ARE from TNFα , inserted into the β-globin 3’UTR ) were a gift from G . Wilson ( Fialcowitz et al . , 2005 ) . The pTRERβ-wt contains a unique BglII site located downstream of the stop codon of β-globin . Into this BglII site we cloned the LARP4 CRD sequence ( corresponding to nucleotides for codons 287 to 358 plus TGA ) , to generate construct βG-stop-CRD ( CRD in 3’-UTR ) . To generate βG-CRD-stop ( CRD in CDS ) , we inserted the LARP4 CRD sequence in frame just before the stop codon . Another construct , containing a +2 frameshift in the CRD sequence just before the β-glo stop codon was obtained from Genewiz . In this construct , two As were inserted before the CRD sequence produce the +2 frameshift and nucleotide mutations were introduced downstream to convert premature stop codons to sense codons . β-globin-TNFα−ARE sequence from pTRERβ-TNFα-ARE was subcloned into NheI and PmeI sites of pcDNA3 . 1 ( - ) to be expressed in HEK293 cells from a regular CMV promoter . Used in this study were anti-FLAG ( Sigma , F1804 ) , anti-GFP ( Santa Cruz , sc-8334 ) , anti-actin ( Thermo Scientific , PA1-16890 ) and anti-GAPDH-HRP ( Sigma Aldrich ) . Rabbit anti-LARP4 and anti- polyclonal rabbit antibodies were described ( Yang et al . , 2011; Mattijssen and Maraia , 2015 ) . All plasmids were verified as intact supercoiled and used in parallel at the same concentrations as determined by nanodrop OD260/280 and compared by ethidium bromide staining after gel electrophoresis ( not shown ) . 5 . 5 × 105 HEK293 cells were seeded per well in 6 well plates one day prior to transfection with Lipofectamine 2000 ( Invitrogen ) . Typically , 7 . 5 ul transfection reagent was used per well to transfect 2 . 5 ug of the pCMV constructs , plus 100 ng pcDNATPGFP plasmid ( Hogg and Goff , 2010 ) and 100 ng pVA1 ( Maraia et al . , 1994 ) according to manufacturers’ instructions . 24 hr after transfection , cells were split over multiple plates . To isolate protein samples , cells were washed with ice-cold PBS and cell lysis was directly into RIPA buffer ( Thermo Scientific ) containing protease inhibitors ( Roche ) . For RNA isolation , either the Maxwell 16 simply RNA cells kit ( Promega ) or Tripure ( Roche ) was used . For mRNA and VA1 detection , total RNA was separated in 1 . 8% formaldehyde agarose gel and transferred to a GeneScreen-Plus membrane . For tRNA detection , total RNA was separated on 10% TBE/urea/polyacrylamide gels ( Thermofisher ) before transfer to a GeneScreen-Plus membrane ( PerkinElmer ) using iBlot Dry Blotting System ( Invitrogen ) . The membranes were UV-cross-linked and vacuum-baked at 80°C for 2 hr . The sequences of oligo probes and their hybridization incubation temperatures ( Ti ) can be found in Supplementary file 1 . Membranes were prehybridized in hybridization solution ( 6 x SSC , 2 x Denhardt's , 0 . 5% SDS and 100 ug/ml yeast RNA ) for one hour at Ti . Hybridization of oligo probes was overnight at Ti . by recombinant purified LARP4 ( 1–286 ) was as described ( Yang et al . , 2011 ) . Analysis and quantitation was done using ImageQuant TL ( GE Healthcare ) . For experiments in 6-well plates , Lipofectamine 2000 ( Invitrogen ) was used to transfect HeLa Tet-off cells with 100 ng pTRERβ ( or derivatives , b-globin under a Tet-responsive minimal CMV promoter , see ‘DNA constructs’ ) , 100 ng of a GFP-expression vector pcDNATPGFP containing a conventional CMV promoter ( Hogg and Goff , 2010 ) according to the manufacturer’s instructions . In some experiments , pCMV2 vectors , containing LARPs were co-transfected . Since LARP6 accumulates to relatively high levels , only half the amount of LARP6 plasmid was transfected compared to LARP4 WT and CS-R ( amount adjusted with empty vector ) . The next day , cells were equally divided into multiple wells . 48 hr post transfection , media was replaced by media containing 2 μg/ml doxycycline ( Sigma ) . For total RNA extraction with the Maxwell 16 simply RNA cells kit ( Promega ) , cells were washed with PBS and directly lysed in homogenization buffer containing thioglycerol ( Promega ) . was done by standard methods as described ( Mattijssen and Maraia , 2015 ) using a programmable density gradient fractionation system spectrophotometer ( model Foxy Jr . ; Teledyne Isco , Lincoln , NE ) . 45 × 105 HEK293 cells were seeded in 10 cm culture plates so that they were 80–85% confluent after 16 hr . The cells were transfected with 816 ng GFP plasmid plus 10 . 2 ug pCMV2 plasmid containing either LARP4-WT or -CS-Tyr and 24 . 5 ug empty pUC19 or pUC19 containing 3 copies of the tRNAThrUGU gene . The day after transfection , the cells from each plate were divided into two 15 cm culture plates ( an aliquot was seeded into a 6-well plate for protein isolation the next day ) . One day later , fresh sucrose solutions ( 47% and 7% , wt/vol ) in 10 mM HEPES , pH 7 . 3 , 150 mM KCl , 20 mM MgCl2 , 1 mM DTT were prepared , filter sterilized and used to make the gradients with a Gradient Master ( Biocomp ) . The cell growth medium was replaced 3 hr before addition of cycloheximide ( Chx ) at a final concentration of 100 μg/ml ( from fresh made 10 mg/ml stock in water ) . After 5 min at 37°C , the cells were moved to ice and washed twice with ice-cold PBS plus 100 μg/ml Chx . Five ml of ice-cold PBS with 100 μg/ml Chx was added per plate , the cells were scraped and added to an ice-cold tube . The cell suspension was centrifuged for 3 min at 1 , 200 rpm at 4°C and the pellet taken up in 300 μl lysis buffer ( 10 mM HEPES , pH 7 . 3 , 150 mM KCl , 20 mM MgCl2 , 1 mM DTT , 2% NP-40 , 2x EDTA-free protease inhibitors ( Roche ) , 100 μg/ml Chx , and 40 U/ml RNaseOUT ( Invitrogen ) ) and kept on ice for 2 min with occasional flicking . The lysate was cleared by centrifugation at 13 , 000 rpm for 5 min at 4°C . Four hundred microliters of the gradient was removed from the gradient tubes , and the equivalent amount of 10 OD260 units of each lysate was carefully loaded on top . The gradients were spun in an ultracentrifuge ( Beckman SW41 rotor ) at 33 , 000 rpm for 2 hr and 50 min at 4°C . One ml fractions were collected and RNA was purified from 500 μl of each fraction using the Maxwell 16 LEV simplyRNA kit with the Maxwell 16 instrument , which includes treatment with DNase I ( Promega ) . tRNA-HySeq was as described ( Arimbasseri et al . , 2015; Arimbasseri et al . , 2016 ) . Briefly , cells were cultured under standard conditions ( DMEM with 10% serum , 1X pencillin-strepomycin at 37°C with 5% CO2 ) . Total RNA was isolated from near confluent cells using TriPure reagent ( Roche ) and resolved on a 6% polyacrylamide-urea-TBE gel . tRNA size ( shorter than 5S rRNA were gel purified by incubating the crushed gel pieces in 0 . 3 M NaCl overnight . The tRNA was precipitated , quantified and 300 ng subjected to partial hydrolysis in 10 mM bicarbonate buffer pH 9 . 8 . The hydrolyzed RNA was dephosphorylated using Calf-Intestine Alkaline Phosphatase ( NEB ) and the 5’ termini were phosphorylated using γ−32P-ATP and T4 polynucleotide kinase ( NEB ) . Barcoded pre-adenylated 3’ adapters were ligated to the 3’ ends of the fragments ( for sequences of all adapters and primers used see Hafner et al . , 2012 ) . In a parallel control reaction , two 32P-RNA size markers of 19 and 35 nt were also ligated to the adapters . The ligated fragments were resolved in a 10% polyacrylamide-urea-TBE gel alongside the adapter ligated size markers . tRNA fragments that migrated between the ligated 19 and 35 nt markers were isolated from the gel and subjected to 5’ adapter ligation . The RNA fragments that with both 3’ and 5’ adapters were again size selected and subjected to reverse transcription ( Superscript III , Invitrogen ) at 42°C with the reverse primer . The RT products were subjected to limited PCR amplification and sequenced on an Illumina HiSeq 2500 . The reads were analyzed as described earlier ( Arimbasseri et al . , 2015; Arimbasseri et al . , 2016 ) . Method for calculating ctAI is largely based upon that for calculating tAI ( dos Reis et al . , 2004 ) substituting tRNA read counts for tRNA gene copy numbers when calculating absolute adaptiveness values ( W ) . Wi=∑j=1ni1-sijtRCij Where ni is the number of tRNA isoacceptors which recognize codon i , tRCij is the mapped mature read count for tRNA j which decodes codon i , and sij is a movable constraint on the efficiency by which the decoding of i by j can occur . Values for sij were optimized for expression data of constructs and are as follows for codon3:Anticodon1; U:G = 0 . 6 , C:I = 0 . 3 , G:U = 0 . 8 , A:I = 0 . 5 and A:G = 0 . 9999 . The sij values are used here in the same way as described ( dos Reis et al . , 2004 ) 1-sij . The ctAI approach differs from tAI in notable respects . It attempts to optimize mRNAs for match to the existing tRNAs within the isoacceptor pool . As such , the relative adaptiveness value wi of a codon is weighted not against all codons , but within isoacceptor groups . wi=Wi/WiMax Where i represents a codon being assayed , and iMax represents the maximum isoacceptor W value for the group or 80 k reads ( which is roughly 75% , i . e . , the 3rd quartile ) , whichever is larger . Thus , ctAI is not an absolute overall measurement of translational strength across all possible constructs , but rather a measurement of codon optimization based on tRNA availability . As before for tAI , the ctAI is calculated for a mRNA ( g ) by:ctAIg=∏k=1lgwikg1/lg Where ikg is the kth codon of the mRNA in gene g of codon length lg . As a result , ctAI measures the adaptation of codon selection in a mRNA to the observed tRNA pool . Total RNA was isolated from HEK293 cells 48 hr after transfection . Two ug total RNA was diluted in a total volume of 11 . 5 ul H2O , then 4 . 5 ul of 4X hybridization buffer ( 40 mM Tris pH 7 . 5 , 200 mM NaCl ) and either 2 ul H2O or 2 ul oligo-dT20 ( 50 uM , Invitrogen ) was added . Samples were heated at 85°C for 5 min then put in a 42°C water bath which was allowed to cool to 32°C ( ~1 °C/minute ) . 2 ul of 10X RNase H reaction buffer was added and 10 ul of RNase H ( 0 . 001 U/ul , Thermo Scientific ) and incubated at 37°C for one hour . Reactions were stopped by addition of 1 . 5 ul 0 . 5 M EDTA . To precipitate RNA , 2 ul glycoblue and 13 . 4 ul 3 M NaAc pH 5 . 2 were added and mixed followed by 375 ul EtOH and incubation at −80°C for one hour . Samples were spun for 30 min at 13 , 000 rpm at 4°C and RNA pellets washed with 1 ml 75% EtOH . RNA was analyzed on northern blot after formaldehyde agarose gel electrophoresis . MEFs were generated from E14 . 5 embryos from the same litter by standard methods . Each MEF cell line was derived from a different embryo . MEFs were derived from KO and WT matched siblings , all females . MEFs at passage 3 were transfected with SV40 Large-T antigen-expressing plasmid pBSSVD2005 ( Addgene plasmid 21826 ) using Lipofectamine 2000 ( Invitrogen ) and subcultured at 1:10 for at least 5 passages . Cells were maintained in DMEM plus Glutamax ( Gibco ) supplemented with 10% heat-inactivated FBS ( Atlanta Biologicals ) in a humidified 37°C , 5% CO2 incubator . Cultures were passed every 2–3 days . HEK293 cell lysate was prepared from cells at 70% confluency by adding an equal volume of lysis buffer ( 10 mM HEPES pH 7 . 3 , 10 mM KAc , 0 . 5 mM MgAc , 5 mM DTT and protease inhibitors ( Roche ) to a PBS-washed cell pellet and incubating for 45 mins on ice . The cells were then passed 10 times through a 30 . 5 G needle and checked under a microscope for a lysis of >60% . The lysate was spun at 14 , 000 g for 1 min to remove debris and nuclei . The supernatant was aliquoted and immediately frozen at −80C . DNA templates for T7 RNA polymerase-mediated transcription , were generated by PCR to obtain the following fragments: Flag-LARP4-1-286 ( to mark the start of the CRD ) and 2 versions of Flag-LARP4- 358 ( end of the CRD ) , a WT version and CS-Tyr . Using the mMESSAGE mMACHINE T7 Ultra Kit ( Thermofisher ) , 7mG 5' capped and 3' polyadenylated mRNAs were generated . Polyadenylation after addition of PolyA polymerase was confirmed by denaturing gel electrophoresis ( not shown ) . The in vitro translation reaction contained the following in 10 ul: 40% cell extract , 50 mM KAc , 2 . 5 mM MgAc , 20 U Superasin , 200 ng mRNA template , 1 . 6 mM HEPES pH 7 . 3 , 2 mM creatine phosphate , 0 . 01 ug/ul creatine kinase , 10 uM spermidine , 10 uM amino acid mix ( no methionine ) , 10 . 2 uCi 35S-Methionine ( Perkin Elmer ) ) . Reactions were placed at 37C and after indicated times placed on ice and quenched by addition of 10 ul EDTA ( 25 mM final ) . The reactions were then subjected to immunoprecipitation using Anti-FLAG M2 Magnetic Beads ( Sigma ) according to manufacturer’s protocol . Immunoprecipitated material was eluted from the beads using SDS buffer containing b-mercaptoethanol and heated for 5 mins at 80C . Samples were loaded on an SDS-PAGE gel , then blotted to nitrocellulose and imaged on a phosphorimager screen . | Genes are coded instructions to build proteins and other molecules that make up living organisms . To build a protein , the code within a gene is copied into a molecule called a messenger RNA ( or mRNA short ) . The letters of the genetic code are then read in groups of three , referred to as codons , by a molecular machine called a ribosome . Each codon corresponds to one of the building blocks of all proteins , known as amino acids . However , because there are 64 possible codons but only 20 or so amino acids found in proteins , different codons can encode for the same amino acid . In addition to determining the order of amino acids in a protein , the sequence of codons in an mRNA can have other effects too . Some sequences change how the mRNA binds with other molecules , while others affect how long the mRNA will last within the cell before it is broken down . In fact , the stability of an mRNA is an important way to control a gene’s activity and genes that encode unstable mRNAs typically yield less protein . Understanding the full information potential of the DNA code is a major goal of many biologists . Much research in this field has focused on single-celled organisms such as yeast , yet the regulation of mRNAs in yeast is generally less complex than it is in humans . Mattijssen et al . have now asked how codons within the mRNA for a human protein called LARP4 affect the mRNA’s stability . This protein binds to mRNA molecules , and the experiments uncovered a short segment of codons that made the mRNA of LARP4 very unstable . Replacing specific codons in this segment with other codons for the same amino acid caused the stability of the LARP4 mRNA to increase a lot . This in turn changed how much LARP4 protein was produced . Amino acids are brought to their corresponding codons by molecules called transfer RNAs ( or tRNAs for short ) . Mattijssen et al . found that the codons in the short segment of LARP4 mRNA that caused instability were matched by rare tRNAs . Increasing the levels of these low level tRNAs also increased how much LARP4 protein was produced . The elevated levels of LARP4 revealed a new activity for the protein . Almost all mRNAs have a so-called poly-A tail at one end , and the experiments showed that LARP4 binds to a range of mRNAs to help make these tails longer , which in turn makes the molecules more stable . Deleting the gene for LARP4 from mouse cells lead otherwise stable mRNAs to have shorter poly ( A ) tails and to become less stable . This includes the mRNAs that code for the proteins that make up ribosomes . The regulation of the mRNAs that encode ribosomal proteins has been challenging to understand . These new results may reveal a network of signals that connects the amount of tRNAs in a cell to the production of ribosomes . Since ribosome production is central to controlling cell growth and division , these results may have broad implications in research into areas as varied as human development and cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2017 | LARP4 mRNA codon-tRNA match contributes to LARP4 activity for ribosomal protein mRNA poly(A) tail length protection |
Proliferating cells often have increased glucose consumption and lactate excretion relative to the same cells in the quiescent state , a phenomenon known as the Warburg effect . Despite an increase in glycolysis , however , here we show that non-transformed mouse fibroblasts also increase oxidative phosphorylation ( OXPHOS ) by nearly two-fold and mitochondrial coupling efficiency by ~30% during proliferation . Both increases are supported by mitochondrial fusion . Impairing mitochondrial fusion by knocking down mitofusion-2 ( Mfn2 ) was sufficient to attenuate proliferation , while overexpressing Mfn2 increased proliferation . Interestingly , impairing mitochondrial fusion decreased OXPHOS but did not deplete ATP levels . Instead , inhibition caused cells to transition from excreting aspartate to consuming it . Transforming fibroblasts with the Ras oncogene induced mitochondrial biogenesis , which further elevated OXPHOS . Notably , transformed fibroblasts continued to have elongated mitochondria and their proliferation remained sensitive to inhibition of Mfn2 . Our results suggest that cell proliferation requires increased OXPHOS as supported by mitochondrial fusion .
Depending on cell type and microenvironment , various adaptations in metabolism have been associated with cellular proliferation . The metabolic adaptation that has received the most attention is a phenomenon known as aerobic glycolysis or the Warburg effect , which is characterized by a high level of glucose consumption and a high rate of glucose fermentation to lactate irrespective of oxygen availability ( Liberti and Locasale , 2016; Vander Heiden et al . , 2009 ) . Although the Warburg effect is recognized as a typical feature of dividing cancer cells and is the basis for imaging many tumors in the clinic with fluorodeoxyglucose positron emission tomography ( Hanahan and Weinberg , 2011; Zhu et al . , 2011 ) , it is also found in normal proliferating cells such as non-transformed fibroblasts , lymphocytes , macrophages , thymocytes , endothelial cells , and embryonic stem cells ( Brand , 1985; Hedeskov , 1968; Hume et al . , 1978; Munyon and Merchant , 1959; Wang et al . , 1976 ) . Accordingly , even in non-cancerous contexts , the Warburg effect has been classified as a hallmark of rapid proliferation ( Abdel-Haleem et al . , 2017 ) . Although there is a general consensus that glycolytic flux increases in proliferating cells , the extent to which oxidative metabolism is altered has been historically complicated ( DeBerardinis et al . , 2007; Seyfried , 2015 ) . Warburg originally proposed that cancer cells rely on enhanced glycolysis because of defects in mitochondria ( Warburg , 1956 ) . Some cancers do have defective mitochondrial enzymes ( e . g . succinate dehydrogenase and fumarase ) , but it is now well established that most proliferating cells ( including cancer ) have functional mitochondria ( Ahn and Metallo , 2015; Vyas et al . , 2016 ) . Indeed , functional mitochondria are essential to the proliferation of some cell types . Studies have shown that oxidative phosphorylation ( OXPHOS ) may provide the majority of ATP during proliferation and function to support the synthesis of important molecular building blocks such as aspartate ( Birsoy et al . , 2015; Fan et al . , 2013; Rodríguez-Enríquez et al . , 2010; Sullivan et al . , 2015; Zu and Guppy , 2004 ) . Elevated levels of OXPHOS , however , may not necessarily be required to fulfill such functions . To the contrary , many reports have suggested that proliferating cells suppress mitochondrial respiration and statements that glycolysis is preferred over OXPHOS during proliferation are prevalent in the literature ( Whitaker-Menezes et al . , 2011 ) . Certain cancers of the bladder , breast , and kidney are depleted of mitochondrial DNA and have decreased expression of respiratory genes ( Reznik et al . , 2016 ) . Some cancer cells exhibit high levels of mitochondrial fission and have associated decreases in respiratory capacity as mediated by an imbalance of dynamin-related protein 1 ( DRP1 ) and mitofusin-2 ( Mfn2 ) ( Chen and Chan , 2017; Rehman et al . , 2012; Serasinghe et al . , 2015; Xie et al . , 2015 ) . In other cases , increasing glucose oxidation by inhibiting pyruvate dehydrogenase kinase has been shown to slow the proliferation of transformed cells ( Bonnet et al . , 2007 ) . A challenge of quantitating changes in OXPHOS as a function of proliferation has been the confounding experimental factors that are often associated with cancer studies . Tumors contain non-proliferating cell types that may shift the average of metabolic measurements from bulk tissues . Additionally , cancer cells from tumors often have restricted access to oxygen ( Brahimi-Horn et al . , 2007 ) . Although oxygen limitation can similarly lead to an enhanced glycolytic phenotype , this metabolic program is distinct from the Warburg effect . Finally , many studies have focused on the proliferative state of cancer cells without having an appropriately matched non-proliferating comparison with the same genetic background tested under the same conditions ( Zu and Guppy , 2004 ) . In this study , to directly compare oxidative metabolism in the same cells of the quiescent and proliferative state , we exploited the cell-density-dependent phenotype of non-transformed fibroblasts . We find that even though proliferating fibroblasts exhibit enhanced glycolysis that is consistent with a Warburg phenotype , they also increase OXPHOS by nearly two-fold and increase their mitochondrial coupling efficiency by ~30% . Interestingly , both increases are supported by mitochondrial fusion . Although transformation with the Ras oncogene further elevated OXPHOS , the additional increase was supported by mitochondrial biogenesis rather than changes in mitochondrial dynamics . Blocking mitochondrial fusion slowed proliferation in both non-transformed and transformed cells . Taken together , our results indicate that proliferation of fibroblasts requires an increase in OXPHOS supported by mitochondrial fusion .
Mouse 3T3-L1 fibroblasts are immortalized , non-transformed cells that retain sensitivity to contact inhibition ( Green and Kehinde , 1975 ) . They provide a simple , well-controlled model to compare metabolism in the proliferative and quiescent states , as has been demonstrated previously ( Yao et al . , 2016a ) . The first step in our analysis was to verify that proliferating fibroblasts exhibit the Warburg effect . Relative to quiescent fibroblasts in the contact-inhibited state , proliferating cells had increased glucose consumption and lactate excretion ( Figure 1A ) . As expected , proliferating cells excreted a greater percentage of glucose as lactate ( 47% ) compared to quiescent cells ( 32% ) ( Figure 1—source data 1 ) . Of note , the absolute amount of glucose having a non-lactate fate was also increased by over two-fold in the proliferative state ( 0 . 38 pmol/cell/hr ) relative to the quiescent state ( 0 . 16 pmol/cell/hr ) ( Figure 1—source data 1 ) . Glucose carbon that is not excreted as lactate is potentially available to support an increased rate of oxidative metabolism , which we next aimed to quantify . Strikingly , on a per cell basis , we found that the basal respiration rate was ~81% higher during proliferation compared to quiescence ( Figure 1—figure supplement 1 ) . Given that proliferating cells are larger in size relative to quiescent cells , we also independently normalized the oxygen-consumption data by protein content instead of cell number . Even when normalized by protein content , the respiration rate of proliferating cells was ~59% higher than quiescent cells ( Figure 1B–C ) . Intriguingly , proliferating cells also exhibited a decrease in proton leak and a ~ 112% increase in ATP production ( Figure 1B–C ) . Taken together , the coupling efficiency of proliferating cells was determined to be 34% higher than quiescent cells . We note that the coupling efficiency was calculated as the ratio of the ATP production rate and the basal respiration rate , which is therefore independent of sample normalization method . These data suggest that proliferating fibroblasts with Warburg metabolism not only have increased OXPHOS , but also that they respire more efficiently . We next aimed to investigate which carbon sources fuel mitochondrial respiration by analyzing the utilization of glucose , glutamine , and fatty acids . In addition to increasing glucose consumption and lactate excretion ( Figure 1A ) , proliferating fibroblasts take up two-fold more glutamine compared to quiescent fibroblasts ( Figure 1D ) . Since the rate of glutamate excretion was unchanged , more glutamine carbon was being used for anaplerosis . We also found that the consumption rates of palmitate and oleate were increased in proliferation by 194% and 98% , respectively ( Figure 1E ) . Since proliferating fibroblasts showed increased utilization of all three nutrients we examined , we next applied stable isotope tracing and metabolomics to access the relative contribution of each carbon source to the TCA cycle . In three separate experiments , cells were fed either uniformly 13C-labeled glucose ( U-13C glucose ) , uniformly 13C-labeled glutamine ( U-13C glutamine ) , or uniformly 13C-labeled palmitate ( U-13C palmitate ) for 6 hr . The isotope enrichments in citrate , a representative TCA cycle intermediate , were evaluated to infer the relative contribution of each carbon source to the TCA cycle and mitochondrial respiration . Even though proliferating cells consumed more glucose and more fatty acids , citrate labeling from these two carbon sources was significantly decreased in the proliferative state ( Figure 1F; Figure 1G ) . In contrast , labeling of citrate by glutamine was substantial and significantly increased in proliferating cells relative to quiescent cells , suggesting that glutamine is a major energy source to fuel mitochondrial respiration ( Figure 1H ) . This result is consistent with reports from other cells ( Fan et al . , 2013 ) . Given that previous studies have shown that glutamine dependence is correlated with cystine uptake through the cystine/glutamate antiporter SLC7A11 ( Muir et al . , 2017 ) , we examined whether these metabolite changes were enabled by transporter expression . Even though the level of SLC7A11 protein was unchanged , we observed more than a two-fold increase in cystine comsumption for proliferating fibroblasts compared to quiescent fibroblasts ( Figure 1—figure supplement 2 ) . Increased influx of cystine may drive the export of glutamate , thereby depleting the pool of intracellular glutamate/αKG and promoting glutamine anaplerosis ( Muir et al . , 2017 ) . We next sought to understand how fibroblasts support increased OXPHOS during proliferation . We reasoned that one mechanism might be by increasing mitochondrial mass in the proliferative state . We used RT-PCR to determine the relative copy number of mitochondrial DNA ( mtDNA ) to genomic DNA ( gDNA ) , from which we inferred mitochondrial mass . The mtDNA to gDNA ratio was unchanged between quiescent and proliferating cells ( Figure 2A ) . Consistent with these data , we also observed no change in expression of respiratory enzymes , as determined by immunoblotting of electron transport chain ( ETC ) subunits ( SDHA for complex II , cytochrome c for complex III , COX IV for complex IV , and ATP5A for complex V ) ( Figure 2B ) . Our results indicate that proliferating fibroblasts do not support elevated levels of OXPHOS by increasing mitochondrial mass or by increasing the expression of respiratory enzymes . As an alternative , we then considered the possibility that fibroblasts regulate OXPHOS during proliferation by mitochondrial dynamics . Previous studies have shown that mitochondrial fusion is associated with an increased respiration rate in addition to an improved coupling efficiency ( Legros et al . , 2002; Westermann , 2012 ) . To assess mitochondrial morphology , we applied electron microscopy ( EM ) imaging ( Figure 2C ) and fluorescence imaging ( Figure 2—figure supplement 1 ) . Both techniques showed that proliferating cells have elongated mitochondria , while mitochondria in quiescent cells were relatively short and fragmented . Quantitative analysis of 100 random mitochondria in each condition showed a statistically significant increase in the mitochondrial length of proliferating cells compared to quiescent cells ( Figure 2D ) . In addition , we found that mitochondria in proliferating cells had a significantly higher level of mitochondrial fusion proteins ( Mfn1 , Mfn2 , and OPA1 ) compared to mitochondria in quiescent cells ( Figure 2—figure supplement 2 ) . When we inhibited mitochondrial fusion by knocking down Mfn2 , a protein required for the fusion of the outer mitochondrial membrane , the elongated mitochondrial phenotype in proliferating fibroblasts was suppressed ( Figure 2C–D , Figure 2—figure supplement 3 ) . To determine whether mitochondrial fusion is required for increased OXPHOS during proliferation , we compared the oxygen consumption rates of proliferating scrambled siRNA controls ( SSC ) to Mfn2 knockdowns ( Mfn2KD ) . We point out that these comparisons were conducted when cells were in the proliferating exponential growth phase . Notably , Mfn2 knockdowns had a statistically significant decrease in respiration rate , ATP production , and mitochondrial coupling efficiency ( Figure 2E–F ) . Given that mitochondria in quiescent fibroblasts are already largely fragmented , as expected , the effect of Mfn2 knockdown on basal mitochondrial respiration was minimal in quiescent cells ( Figure 2—figure supplement 4 ) . Independent of contact inhibition , cellular quiescence can also be achieved by serum starvation ( Yao , 2014 ) . By using serum starvation , we sought to extend our comparison of the proliferative and quiescent states to other cells . Consistent with our contact-inhibition results , we found that serum starved 3T3-L1 and HCT116 cells had fragmented mitochondria relative to the same cells in the non-starved state ( Figure 2—figure supplement 5 ) . Mitochondrial elongation occurred as soon as 3 hr after reintroducing serum and continued as cells exited the quiescent state ( Figure 2—figure supplement 5B ) . Having established that mitochondrial fusion increased in dividing cells , we wished to consider its effects on proliferation . Upon Mfn2 knockdown , we observed a ~ 30% decrease in proliferation rate compared to scrambled siRNA controls ( Figure 3A ) . Re-expressing siRNA-resistant Mfn2 ( Mfn2res ) in Mfn2 knockdowns restored Mfn2 protein level , mitochondrial respiration , and cellular proliferation ( Figure 3—figure supplement 1 ) . When we overexpressed Mfn2 in wildtype 3T3-L1 fibroblasts , we observed a significant increase in both mitochondrial respiration and proliferation ( Figure 3—figure supplement 2 ) . These data suggest that promoting mitochondrial fusion is sufficient to drive proliferation . We also observed considerable changes in nutrient utilization between Mfn2 knockdowns and scrambled siRNA controls that were consistent with decreased proliferation and reduced OXPHOS ( Figure 3B ) . Mfn2 knockdowns decreased their glucose uptake by 15% and decreased their lactate excretion by 10% . More notably , knocking down Mfn2 caused cells to decrease glutamine consumption by 40% . Given the reduced rate of OXPHOS in Mfn2 knockdowns , these data are consistent with glutamine serving as a major carbon source for the TCA cycle . Tracing experiments confirmed a significant decrease in labeling of TCA cycle intermediates from U-13C glutamine in Mfn2 knockdowns ( Figure 3—figure supplement 3A–C ) . Consistent with decreased glutamine anaplerosis , Mfn2 knockdowns had a 40% decrease in cystine consumption ( Figure 3—figure supplement 3D ) . Our results show that mitochondrial fusion supports a high level of OXPHOS , which is needed to sustain rapid cellular proliferation . We speculated that the decrease in proliferation rate upon Mfn2 knockdown might be due to a shortage of energy from the decreased rate of OXPHOS . Surprisingly , however , we found that intracellular levels of ATP actually increased slightly in Mfn2 knockdowns relative to scrambled siRNA control cells ( Figure 3C ) . This result suggested that cells could compensate for a reduced energy yield from OXPHOS , but that OXPHOS may serve another indispensable function in our knockdowns . Previous studies have shown that an essential role of OXPHOS in proliferating cells is to regenerate reducing equivalents in support of aspartate synthesis ( Birsoy et al . , 2015; Sullivan et al . , 2015 ) . Through reactions in the malate-aspartate shuttle , increased oxygen consumption may support a higher rate of aspartate generation . Although we did not observe a difference in the intracellular pool of aspartate between scrambled siRNA controls and Mfn2 knockdowns ( Figure 3D ) , we did find a striking change in aspartate uptake . While the control cells excreted aspartate , the Mfn2 knockdown cells consumed aspartate from the media ( Figure 3E ) . Moreover , the proliferation of Mfn2 knockdowns could be partially rescued by supplementing cell-culture media with aspartate ( Figure 3F ) . Although 3T3-L1 fibroblasts are immortalized , unlike transformed cells , they remain sensitive to contact inhibition and retain the ability to differentiate . To evaluate whether transformed cells similarly rely on OXPHOS and mitochondrial fusion , we generated stable transfected 3T3-L1 cells expressing the oncogene H-Ras ( G12V ) , a constitutively active mutant ( Figure 4A ) . H-Ras transfected fibroblasts assumed a transformed morphology and their growth was no longer contact inhibited , with high-density cultures forming multiple layers of cells ( Figure 4B–C ) . It was confirmed that the transformed cells did not undergo oncogene-induced senescence ( Figure 4—figure supplement 1 ) . To study the effect of H-Ras transformation on mitochondrial metabolism , we compared the oxygen consumption rates of proliferating empty vector ( EV ) control cells in the exponential growth phase to H-Ras transformed ( Ras ) cells . Surprisingly , we found that Ras cells had a ~ 73% increase in basal respiration and a ~ 72% increase in ATP production compared to EV cells ( Figure 4D–E ) . It is interesting to note that we found no difference in the mitochondrial coupling efficiencies between conditions ( Figure 4E ) . This result is consistent with the observation that mitochondria are not further fused in Ras cells relative to proliferating EV controls ( Figure 4F ) . Given that Ras cell mitochondria remain elongated to the same extent as EV controls , we next evaluated increased mitochondrial mass as a possible mechanism to support elevated levels of OXPHOS . For Ras cells , we observed a > 20-fold increase in the mRNA level of peroxisome proliferator activated receptor coactivator ( PGC1α ) , a master regulator for mitochondrial biogenesis ( LeBleu et al . , 2014; Scarpulla , 2011 ) ( Figure 4G ) . Ras cells had more mitochondrial mass and increased protein expression levels of ETC subunits ( Figure 4H–I ) , suggesting that increased OXPHOS in Ras cells is driven by mitochondrial biogenesis . Activation of mitochondrial biogenesis upon Ras transformation did not change protein levels of components in the ERK/AMPK pathway or the KSR1 pathway , as has been previously suggested for other cells ( Figure 4—figure supplement 2 ) ( Dard et al . , 2018; Weinberg et al . , 2010 ) . We wish to emphasize that even though Ras cell mitochondria are not further elongated , they remain fused to the same level as proliferating non-transformed fibroblasts . Therefore , when mitochondrial fusion was inhibited by knocking down Mfn2 , the proliferation rates and OXPHOS of both EV cells and Ras cells were significantly attenuated ( Figure 4J , Figure 4—figure supplement 3 ) . In addition , the proliferation of various cancer cell lines also proved sensitive to Mfn2 knockdown ( Figure 4K ) . We conclude that mitochondrial fusion is important to sustain cellular proliferation , independent of oncogenic transformation . We speculated that constitute activation of Ras might enhance metabolic phenotypes we associated with proliferation in non-transformed cells . Indeed , in addition to elevated OXPHOS , Ras cells consumed more glucose , glutamine , and fatty acids relative to EV controls ( Figure 4—figure supplement 4A ) . Given that Ras cells do not proliferate faster than EV cells ( Figure 4B ) , the increase in metabolic activity is unlikely due to proliferative demand but rather associated with Ras signaling activation . Consistent with the notion that energy is not limiting during proliferation ( Locasale and Cantley , 2011 ) , we found that Ras cells had a higher intracellular level of ATP compared to EV controls ( Figure 4—figure supplement 4B ) . In addition to increased ATP production , elevated OXPHOS activity in Ras cells also contributed to higher levels of reactive oxygen species ( ROS ) ( Figure 4—figure supplement 4C ) . The associated oxidative stress could be buffered by treating Ras cells with the antioxidant N-acetyl cysteine ( NAC ) . Alternatively , oxidative stress could be further induced by treating Ras cells with bromodeoxyuridine ( BrdU ) , a thymidine analog ( Figure 4—figure supplement 4C ) . We suspected that the elevated levels of ROS in Ras cells may lead to increased DNA damage . We verified this to be the case by showing that Ras cells had increased phosphorylation on histone H2A . X ( Ser139 ) , which has been suggested as a sensitive marker for DNA damage ( Sharma et al . , 2012 ) ( Figure 4—figure supplement 4D ) . These findings suggest that the increase in OXPHOS upon constitutive Ras activation leads to elevated oxidative stress and DNA damage , while not directly contributing to the anabolic demands of proliferation . Given the pleiotropic effects of Ras and other oncogenes , the generality of these findings to additional cell types will require further investigation .
Most proliferating cells assume a metabolic phenotype known as the Warburg effect ( Lunt and Vander Heiden , 2011 ) . Although the enhanced glycolytic characteristics of the Warburg effect are generally well established , metabolic changes associated with mitochondria have proven more challenging to interrogate . In part , this is because proliferation has largely been studied in the context of cancer where some experimental factors are complicated to control ( e . g . tumor microenvironment , oxygen availability , genetics , proliferation , etc . ) . Here , we applied a well-controlled fibroblast model to quantify changes in mitochondrial respiration that occur in quiescent cells , non-transformed proliferating cells , and transformed proliferating cells . Strikingly , despite the frequent assumption that increased glycolysis in cells exhibiting the Warburg effect is associated with decreased OXPHOS , we found that mitochondrial respiration increased by nearly two-fold in non-transformed proliferating cells relative to quiescent cells . Moreover , mitochondrial respiration increased by nearly another factor of two when the cells were transformed with H-Ras . We wish to point out that the regulatory mechanism for increasing respiration between quiescent and non-transformed proliferating cells was different from that between non-transformed and transformed cells . The quiescent to proliferating transition was supported by mitochondrial fusion without any increase in mitochondrial mass , whereas the non-transformed to transformed transition was supported by mitochondrial biogenesis without any further increase in mitochondrial elongation . Similar increases in mitochondrial biogenesis and OXPHOS upon Ras activation have been reported in other cell lines ( Funes et al . , 2007; Moiseeva et al . , 2009; Telang et al . , 2007 ) . In addition to Ras , various other signaling pathways that regulate cellular proliferation ( e . g . c-Myc and mTOR ) have also been found to activate mitochondrial biogenesis ( Vyas et al . , 2016 ) . Notably , however , our data suggest that the proliferation rate of both non-transformed and transformed fibroblasts is dependent on mitochondrial fusion . It is interesting to consider why OXPHOS is increased by mitochondrial fusion in proliferating cells . Since ATP levels actually increased when OXPHOS was impaired by Mfn2 knockdown , mitochondrial fusion does not seem to be required to support energetic demands . Instead , increased respiration during proliferation seems to be needed to recycle reducing equivalents in support of aspartate synthesis . Only after mitochondrial fusion was inhibited did cells start consuming aspartate from the media . Moreover , proliferation could be partially restored in Mfn2 knockdowns by supplementing cell media with aspartate . Building on previous studies ( Birsoy et al . , 2015; Sullivan et al . , 2015 ) , these data suggest not only that respiration is required to meet aspartate demands , but that a high level of OXPHOS may be necessary to fulfill this role . On the other hand , too much OXPHOS may be detrimental to cells . Increasing OXPHOS beyond the level observed in non-transformed proliferating fibroblasts with the H-Ras oncogene did not increase the rate of proliferation . Instead , the considerably higher levels of OXPHOS induced by mitochondrial biogenesis resulted in oxidative stress and DNA damage . These results suggest that , unlike the mechanisms that increase mitochondrial respiration in normal proliferating cells , Ras activation may promote additional malignant transformations by creating genomic instability ( Tubbs and Nussenzweig , 2017 ) . Despite the association between the Warburg effect and rapid proliferation , a rationalization for the preference of glycolysis over OXPHOS has proven elusive ( Liberti and Locasale , 2016 ) . A challenge has been explaining how the switch to a metabolic program that is less energetically efficient supports the synthetic burden of cell replication . Transformation of glucose to lactate yields only two moles of ATP per mole of glucose , whereas complete oxidation of glucose by the TCA cycle yields ~ 32 moles of ATP per mole of glucose . Hypotheses have emerged that proliferating cells sacrifice ATP yield from glucose for other advantages such as a high rate of ATP production , decreased volume of enzymatic machinery , or increased concentrations of macromolecular precursors ( Lunt and Vander Heiden , 2011; Slavov et al . , 2014; Vazquez et al . , 2010 ) . Yet , to date , the benefits of using glycolysis over OXPHOS for proliferation have remained controversial . In this study , we provide evidence that the Warburg effect does not necessitate decreased OXPHOS . Rather , in the cells we examined here , glycolysis and OXPHOS are both elevated by significant levels during proliferation ( Figure 5 ) . Thus , the need to rationalize a preference for glycolysis over OXPHOS during proliferation may be unnecessary .
3T3-L1 cells were obtained from ATCC . H460 , Hela , BT549 , and MCF7 cells were obtained from Washington University . All cells were found to be negative for mycoplasma contamination . All cells were cultured in high-glucose DMEM ( Life Technologies ) containing 10% fetal bovine serum ( FBS ) ( Life Technologies ) and 1% penicillin/streptomycin ( Life Technologies ) at 37°C with 5% CO2 . To establish a growth curve , cells were collected every 12–24 hr and counted in trypan blue with an automated cell counter ( Nexcelom ) . For assessing proliferation , cells were grown under various experimental conditions for 48–72 hr , and proliferation was determined by manual cell counting or by using a CyQUANT proliferation assay ( Thermo ) according to the manufacturer’s instructions . For serum starvation , cells were cultured in DMEM ( without FBS ) for 48 hr . Proliferation was induced by transferring cells to media containing serum ( 20% FBS ) . The oxygen consumption rate ( OCR ) of whole cells was determined by using the Seahorse XFp Extracellular Flux Analyzer ( Seahorse Bioscience ) . Cells were trypsinized and plated on a miniplate 24 hr prior to the Seahorse assay . For Mfn2 knockdowns , cells were treated with scrambled siRNA as a control or Mfn2 siRNA for 48 hr prior to seeding . The assay medium consisted of 25 mM glucose , 4 mM glutamine , 50 μM palmitate-BSA , and 50 μM oleate-BSA in Seahorse base medium . The OCR was monitored upon serial injections of oligomycin ( oligo , 2 μM ) , FCCP ( 1 μM ) , and a rotenone/antimycin A mixture ( rot/AA , 1 μM ) . A concentration of 1 μM FCCP was determined to be optimal in separate experiments . OCR was normalized to the final cell number or total protein amount as determined by manual cell counting or by using a BCA assay , respectively . Data presented have been corrected for non-mitochondrial respiration . 3T3-L1 fibroblasts were plated at ~20% confluence or 100% confluence to establish the proliferating or quiescent condition , respectively . EV controls and H-Ras transformed fibroblasts were plated at ~40% confluence . Then , 24 hr later , the medium was changed to medium in which natural-abundance glucose was replaced with U-13C glucose or natural-abundance glutamine was replaced with U-13C glutamine . For palmitate labeling , after 24 hr the medium was changed to medium containing 100 μM U-13C palmitate-BSA and 100 μM natural abundance oleate-BSA . After labeling for 6 hr , cells were harvested , extracted , and analyzed as previously described ( Yao et al . , 2018 ) . The polar portion of the extract was separated by using a Luna aminopropyl column ( Phenomenex ) coupled to an Agilent 1260 capillary HPLC system . The Luna column was used with the following buffers and linear gradient: A = 95% water , 5% acetonitrile ( ACN ) , 10 mM ammonium hydroxide , 10 mM ammonium acetate; B = 95% ACN , 5% water . Mass spectrometry detection was carried out on an Agilent 6540 or 6545 Q-TOF coupled with an ESI source operated in negative mode . The identity of each metabolite was confirmed by comparing retention times and MS/MS data with standard compounds . The isotopologue distribution pattern was calculated by normalizing the sum of all isotopologues to 100% . The labeling percentages of tracers at the end of the experiments are presented in Figure 1—source data 2 . Data shown have been corrected for natural abundance and isotope impurity ( see Figure 1—source data 3 for raw data ) . Pool sizes were calculated as the sum of all isotopologues and normalized to dry cell mass ( measured by using an analytical balance ) as well as a D8-phenylalanine internal standard . After incubating cells in fresh media for 24 hr , the spent media were collected and analyzed . Known concentrations of isotope-labeled internal standards ( glucose , lactate , glutamine , glutamate , palmitate , and aspartic acid; Cambridge Isotopes ) were spiked into media samples before extraction . Extraction was performed with glass as previously reported ( Yao et al . , 2016b ) . Samples were measured by LC/MS analysis , with the method described above . The absolute concentration of each compound was determined by calculating the ratio between the fully unlabeled peak from samples and the fully labeled peak from standards . The consumption rates ( x ) were normalized by cell growth over the experimental time period by using the following equation where N0 represents the starting cell number , t represents incubation time , DT represents doubling time , and Y represents nutrient utilization . Y=∫0tx∙N0∙2t/DT∙dt DNA was extracted by using QuickExtract DNA extraction solution ( Epicentre ) according to the manufacturer’s instructions . The ratio of mitochondrial DNA ( mtDNA ) to genomic DNA ( gDNA ) was measured by using a NovaQUANT mouse mitochondrial to nuclear DNA ratio kit ( Millipore ) with RT-PCR ( Applied Biosystems ) . We applied the following expressions: ΔCT = CTMitochondrial CTNucleic and mtDNA/gDNA = 2-ΔCT . For measuring PGC1α−1 expression levels , RNA was extracted using Trizol ( Invitrogen ) . cDNA was synthesized using Super-Script III First-Strand Synthesis SuperMix ( Invitrogen ) . Amplifications were run with RT-PCR by using premade primers ( IDT ) . The results were normalized to a housekeeping gene , RPL27 . The following expressions were applied: ΔΔCT = ΔCTPGC1α-1 – ΔCTRPL27 and fold change = 2-ΔΔCT . Mfn2 silencing was achieved by using a validated pool of siRNA duplexes directed against mouse Mfn2 ( TriFECTa Kit , IDT ) and Lipofectamine RNAiMAX Transfection Reagents ( Invitrogen ) according to the manufacturer's instructions ( see Figure 3—source data 1 for the dicer-substrate short interfering RNA , DsiRNA , sequence ) . Cells given scrambled siRNA were used as a negative control . To rescue siRNA knockdowns , a siRNA-resistant cDNA that expresses wildtype Mfn2 was cloned in the pcDNA3 . 1+vector ( GenScript ) under a constitutive CMV promoter . The codon was optimized to be resistant to the siRNA added ( see Figure 3—source data 1 for cDNA sequence ) . The control vector was pcDNA3 . 1+N eGFP ( GenScript ) , which expresses GFP instead of Mfn2 . For rescue experiments , cells were first knocked down with siRNA for 12 hr and then transfected with plasmids using Lipofectamine 3000 ( Invitrogen ) for 36–48 hr . For overexpression , wildtype cells were transfected with plasmids for 36–48 hr . Cells or isolated mitochondria were lysed with RIPA buffer ( Thermo Fisher Scientific ) in the presence of a protease inhibitor and a phosphatase inhibitor cocktail ( Thermo Fisher Scientific ) . Lysates were separated by SDS–PAGE under reducing conditions , transferred to a nitrocellulose membrane , and analyzed by immunoblotting . For primary and secondary antibodies , please refer to the Key Resources Table . β-tubulin was used as a loading control . Signal was detected with a WesternSure premium chemiluminescent substrate and the C-Digit Blot Scanner ( LI-COR ) according to the manufacturer’s instructions . Samples were fixed in 2% paraformaldehyde/2 . 5% glutaraldehyde ( Polysciences ) in 100 mM sodium cacodylate buffer , pH 7 . 2 for 1 hr at room temperature . Samples were washed in sodium cacodylate buffer and postfixed in 1% osmium tetroxide ( Polysciences ) for 1 hr . Next , samples were rinsed in dH2O prior to en bloc staining with 1% aqueous uranyl acetate ( Ted Pella ) for 1 hr . Following several rinses in dH2O , samples were dehydrated in a graded series of ethanol and embedded in Eponate 12 resin ( Ted Pella ) . Sections of 95 nm were cut with a Leica Ultracut UCT ultramicrotome ( Leica Microsystems ) , stained with uranyl acetate and lead citrate , and viewed on a JEOL 1200 EX transmission electron microscope ( JEOL USA ) equipped with an AMT eight megapixel digital camera and AMT Image Capture Engine V602 software ( Advanced Microscopy Techniques ) . The length of 100 random mitochondria for each condition were measured and plotted . After removing the media , cells were incubated with 100 nM MitoTracker Red CMXRos ( Thermo Fisher Scientific ) dissolved in complete medium at 37°C for 30 min . Nuclei were stained with Hoechst 33342 ( Thermo Fisher Scientific ) . Cells were imaged alive using a Zeiss LSM 880 confocal microscope equipped with Airyscan . Images were acquired with a Zeiss 20x , 40x , 63x/1 . 4 NA objective by using the ZEN Black acquisition software . Samples were excited with 405 and 543 nm laser lines . Images were processed and prepared with the ZEN Black software . To generate lentivirus carrying oncogenic Ras ( HRASV12 ) , HEK293T cells were co-transfected with pCMV-VSV-G ( a gift from Bob Weinberg; Addgene plasmid # 8454 ) , pCMVΔR8 . 2 ( a gift from Didier Trono; Addgene plasmid # 12263 ) , and pLVX-HRasV12-hygromycin ( a gift from David Piwnica-Worms ) constructs with Lipofectamine 2000 reagent ( Invitrogen ) . Cell media were replaced with fresh growth media 24 hr after transfection . Supernatants with lentivirus were collected after a 24-hr incubation period . Collected lentivirus ( 5 mL ) was used to infect 106 3T3-L1 fibroblast cells in the presence of 10 µg/ml protamine sulfate overnight for ~16 hr . Selection of HRASV12-expressing 3T3-L1 cells was achieved by 100 µg/ml hygromycin . Ras expression was verified by immunoblotting . Senescence was tested by using the β-galactosidase staining kit ( Cell Signaling ) according to the manufacturer’s instructions . Intracellular ATP was measured by using an ATP luminescence detection assay kit ( Cayman ) according to the manufacturer’s instructions . The signal was normalized by protein amount as determined by using a BCA assay ( Pierce ) . Cells were treated with 5 mM N-acetyl cysteine ( NAC ) or 5-bromo-2’-deoxyuridine ( BrdU ) for 48 hr . Intracellular reactive oxygen species ( ROS ) were measured by using a DCFDA assay ( Cayman ) according to the manufacturer’s instructions . The signal was normalized by protein amount as determined by using a BCA assay ( Pierce ) . | Most cells in the body contain many small compartments called mitochondria . These tiny powerhouses can use oxygen to break down molecules of glucose ( a type of sugar ) and release the energy that fuels many life processes . Mitochondria can also use oxygen to build certain compounds essential for the cell . Rapidly dividing cells , such as the ones found in tumors , need a lot of energy . Yet , they often ‘choose’ to burn much of their glucose through fermentation , a less efficient process that does not require oxygen or mitochondria . In fact , many theories suggest that cells which divide a lot decrease the quantity of oxygen their mitochondria consume . It is still unclear what role mitochondria have during phases of intense growth , and if they act differently in cancerous and healthy cells . Here , Yao et al . use a cell system where division can be turned on or off , and find that when cells quickly multiply , their mitochondria actually consume more oxygen . Further experiments then reveal that , in both cancerous and healthy cells , the different mitochondria inside a cell merge during periods of intense division . This mechanism allows the compartment to better use oxygen . Yao et al . go on to show that adjusting the fusion process through genetic manipulation helps to control division . When mitochondria cannot combine , cells divide less well; when the compartments can merge more easily , cells multiply faster . If growing cells do not rely on their mitochondria for their energy demands during multiplication , why do these compartments seem to be essential for division ? The reason might be that the mitochondria produce aspartate , a molecule that cells use to replicate . The work by Yao et al . suggests that at least certain cancer cells may increase their consumption of oxygen to sustain their mitochondria; armed with this knowledge , it may be possible to design new diagnostic tests and new treatments to identify , and potentially target these oxygen-dependent tumor cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"cancer",
"biology"
] | 2019 | Mitochondrial fusion supports increased oxidative phosphorylation during cell proliferation |
We present a method for in-focus data acquisition with a phase plate that enables near-atomic resolution single particle reconstructions . Accurate focusing is the determining factor for obtaining high quality data . A double-area focusing strategy was implemented in order to achieve the required precision . With this approach we obtained a 3 . 2 Å resolution reconstruction of the Thermoplasma acidophilum 20S proteasome . The phase plate matches or slightly exceeds the performance of the conventional defocus approach . Spherical aberration becomes a limiting factor for achieving resolutions below 3 Å with in-focus phase plate images . The phase plate could enable single particle analysis of challenging samples in terms of small size , heterogeneity and flexibility that are difficult to solve by the conventional defocus approach .
Phase plates are one of the technologies holding promise for future performance improvements in cryo-electron microscopy ( cryo-EM ) . Direct electron detectors already changed the prospects of the cryo-EM field ( Henderson , 2015 ) with atomic resolution structures becoming almost routine ( Campbell et al . , 2015; Bartesaghi et al . , 2015 ) . Phase plates improve the image contrast and allow in-focus data acquisition which , in theory , will result in increase of the signal-to-noise ratio across the entire frequency spectrum ( Glaeser , 2013 ) . This could enable structural investigations of 'difficult' samples , such as small , heterogeneous and/or flexible molecules or complexes ( Hall et al . , 2011 ) . Until now , however , various practical problems and performance issues have prevented the acquisition of high-resolution datasets with a phase plate ( Danev et al . , 2009 ) . Here we demonstrate that with correct use , in particular accurate focusing , one can achieve near-atomic resolutions . The phase plate is a device that produces phase contrast by introducing a phase shift between the scattered and unscattered waves at a diffraction plane inside the microscope . Phase contrast denotes that phase variations of the electron wave caused by the sample will be transformed into amplitude variations at the camera thus enabling phase observation . Ideally , the phase shift must be π/2 to realize the so-called Zernike phase contrast ( Danev and Nagayama , 2001 ) . In practice , however , a satisfactory phase contrast performance can be obtained within a range of phase shift values ( π/4 ~ 3π/4 ) ( Danev and Nagayama , 2011 ) . There are other ways to produce phase contrast , the most common is acquiring images slightly out of focus , also known as defocus phase contrast . This method is the de facto standard in transmission electron microscopy ( TEM ) but it has the disadvantage of low overall contrast because of poor performance at low spatial frequencies , i . e . large specimen features are not well reproduced in the image . Phase plates for TEM have been in development for more than 15 years with the thin film Zernike phase being one of the most promising candidates ( Glaeser , 2013 ) . It consists of a thin material film , typically amorphous carbon , with thickness selected for π/2 phase shift at the operating voltage of the microscope ( ~22 nm for 100 kV; ~31 nm for 300 kV ) and a small ( ~1 μm ) hole in the center for the central beam of unscattered electrons ( Danev and Nagayama , 2001 ) . Nevertheless , the Zrnike phase plate has some practical issues , such as a short usable lifespan , produces fringe artifacts in the images , is difficult to use and almost impossible to automate because the central diffraction beam must be positioned precisely in the middle of the small central hole ( Danev et al . , 2009 ) . There are a few examples in the literature of single particle analysis ( SPA ) with the Zernike phase plate but in all of them the resolution was limited to about one nanometer ( Danev and Nagayama , 2008; Murata et al . , 2010 ) . The Volta phase plate ( VPP ) was developed recently as a successor to the thin film Zernike phase plate ( Danev et al . , 2014 ) . It solves most of the issues , in particular , the VPP has a virtually unlimited life , does not produce fringe artifacts and does not require precise centering . There are already a few application examples of the VPP in cryo-tomography ( Fukuda et al . , 2015; Asano et al . , 2015; Mahamid et al . , 2016 ) and single particle analysis ( Khoshouei et al . , 2016 ) that demonstrate its benefits . The VPP consists of a thin ( ~10 nm ) continuous amorphous carbon film which is constantly heated to ~200°C to prevent beam-induced contamination . The phase shift is created on-the-fly through the interaction of the central diffraction beam with the carbon film ( Danev et al . , 2014 ) . This mitigates the requirement for precise centering of the phase plate but also gives rise to the only drawback of the VPP – the phase shift is not constant and increases proportionally to the accumulated dose . The evolving phase shift is not a performance limiting factor but rather a parameter that has to be kept in mind when designing data acquisition strategies . Nevertheless , the ability to create phase shift at any given place on the phase plate film is a notable practical advantage . A wide open area of the film provides multiple virtual phase plates by just moving from one position to the next . The distance between the positions must be large enough ( typically ~20 μm apart ) to prevent the previous phase shift spot from interfering with spatial frequencies below the Nyquist frequency of the detector . Having the ability to move arbitrarily on the phase plate and create phase shift opens the VPP to automation . This is an important practical factor in sync with recent trends towards unattended 24/7 data collection . All datasets in this work were collected automatically .
Our initial SPA trials with the VPP were limited to resolutions ( ~8A , unpublished data ) similar to those already reported with the Zernike phase plate . In those trials we used software and data acquisition schemes developed for conventional defocus phase contrast transmission electron microscopy ( CTEM ) . Such schemes place low priority on focusing accuracy because with CTEM the defocus is accurately measured by contrast transfer function ( CTF ) fitting during data processing and some defocus variation is actually desirable and beneficial for the reconstruction ( Cheng et al . , 2015 ) . As shown later , using phase plates in-focus requires very accurate focusing . Therefore , we had to develop and apply a new data acquisition scheme . In general , there are two defocus strategies for SPA data acquisition with a phase plate . The first one involves collecting data in a similar fashion as in CTEM , i . e . with an intentionally applied defocus . The CTF must then be fitted and corrected during data processing . The advantage of this approach is that precise focusing in not required and some defocus variation is actually desirable . Not having to focus accurately also speeds up the acquisition . The disadvantage is that the only benefit from using a phase plate is contract increase for the very low spatial frequencies whereas the remainder of the frequency spectrum has multiple CTF zeroes , like in CTEM . Also , because of the phase plate , fitting the CTF requires an additional phase shift parameter which may reduce the accuracy of the fit . The second and ideal SPA acquisition scheme with a phase plate is to collect data in-focus . The advantage of this approach is that there are no CTF zeroes across the frequency spectrum and there is no need to fit and correct the CTF . However , this approach is more demanding on the experiment in that it requires very accurate focusing . Also , if there are small random defocus errors they cannot be corrected during data processing because fitting the CTF without at least a few detectable CTF zeroes is practically impossible . In our initial SPA trials with the VPP we also tried the defocused approach but without much success ( unpublished data ) , mainly because of the lack of phase shift support in the single particle reconstruction software . Future software developments could change the status quo and make the defocused approach more attractive , but so far we have had more success with the in-focus approach ( Khoshouei et al . , 2016 ) . Figure 1 illustrates the effect of defocus on the expected resolution for in-focus data collection with a phase plate . We use the term 'in-focus' also for small ( <100 nm ) defocus values which are used to counteract the effect of spherical aberration and extend the resolution range by moving the first CTF zero to higher spatial frequencies . Figure 1A shows simulated CTFs for a π/2 phase plate at three defocus values . The figure illustrates what would be an acceptable defocus range for a 4 Å resolution SPA reconstruction . The resolution criterion we used is CTF amplitude dropping below 0 . 5 , i . e . |CTF|=0 . 5 . The zero defocus curve ( dashed line ) illustrates the effect of spherical aberration on the CTF . The CTF crosses the 0 . 5 level at a resolution lower than 4 Å . Small amounts of underfocus ( defocus < 0 ) counter the effect of spherical aberration and extend the region of good information transfer to higher resolutions . The red curve shows the minimum defocus ( −7 nm ) required to reach 4 Å resolution according to the |CTF|=0 . 5 criterion . Increasing the defocus improves the resolution further but at higher defocus values the initially flat part of the CTF starts to develop a dip , as illustrated by the blue curve ( defocus -60 nm ) . This imposes a limit on the maximum defocus value . 10 . 7554/eLife . 13046 . 003Figure 1 . Volta phase plate CTF examples and allowed defocus ranges versus resolution . ( A ) Illustration of CTFs at defocus values that limit the resolution to 4 Å according to a |CTF|=0 . 5 criterion . ( B ) Defocus limits versus resolution according to the |CTF|=0 . 5 criterion for a π/2 phase plate and 2 . 7 mm spherical aberration . The shaded areas are 'prohibited' in a sense that for those defocus values the CTF amplitude drops below 0 . 5 at a resolution lower than the value on the y-axis . ( C ) Same as ( B ) but for three different phase shift values . ( D ) Same as ( B ) but for a Cs-corrected microscope ( 0 mm spherical aberration ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13046 . 003 Based on the |CTF|=0 . 5 criterion we calculated the usable defocus range as a function of the resolution . The result is shown in Figure 1B . The shaded areas are 'prohibited' in terms of CTF performance . The one on the right is caused by the CTF amplitude crossing the 0 . 5 point on its way to the first CTF zero , similar to the red curve in Figure 1A . The shaded area on the left is due to the dip in the flat region of the CTF exhibited by the blue curve in Figure 1A . This dip imposes a limit for all spatial frequencies above the point where it touches the 0 . 5 level which is the reason for the vertical wall-like defocus cutoff for a range of resolutions in Figure 1B . The white region between the shaded areas in Figure 1B is the usable defocus range ( horizontal axis ) for achieving a given resolution ( vertical axis ) . For higher resolutions ( lower numerical values ) the defocus range gets narrower and has a cutoff point at about 2 . 7 Å . Achieving resolutions below that point will require reduction or complete elimination of the spherical aberration ( Cs correction ) . The phase shift of the VPP is not constant and increases with the accumulated dose on the phase plate ( see Figure 3 below ) ( Danev et al . , 2014 , Figure 1 ) . It has a rapid onset in the beginning followed by a gradual increase . For single particle data acquisition this would mean that the first images taken after the phase plate is inserted will have a phase shift below π/2 and later images may have a phase shift above that value . Figure 1C illustrates the effect of phase shift on the usable defocus range and resolution . For lower phase shifts ( blue area ) the defocus ranges are shifted towards more defocus and there is a slight improvement in the absolute resolution cutoff point ( ~2 . 6 Å ) . Higher phase shifts ( red area ) require less defocus but have a worse resolution cutoff ( ~3 . 0 Å ) . Figure 1D shows an ideal case of a π/2 phase plate on a Cs-corrected microscope ( zero spherical aberration ) . There is no resolution cutoff but the allowed defocus range gets progressively narrower for higher resolutions . A variable phase shift will move the shaded areas to the left or right , similar to the behavior in Figure 1C , but in order to avoid clutter the effect is not shown in the figure . The graphs in Figure 1 demonstrate the strict requirements on the focusing accuracy for in-focus data acquisition with a phase plate . In order to achieve the required nanometer-level focusing precision we implemented a double focusing area acquisition scheme . Figure 2A shows an illustration of the scheme superposed on a holey support film . The defocus was measured on opposite sides , areas F1 and F2 , of the data acquisition area D and the average ( linear interpolation ) was used to correct the focus . Such measurement minimizes errors due to local slant of the support film . We measured up to 400 nm defocus difference across the two sides ( 3 . 0 um distance ) of holes on Quantifoil R 1 . 2/1 . 3 holey carbon grids . This would mean that a single focusing area acquisition scheme will have a defocus error of up to 200 nm which is substantial compared to the values in Figure 1 . That provides a plausible explanation for the lower resolution ( ~8 Å ) reconstructions we obtained when using a conventional single focusing area scheme . To improve the accuracy even further the double area focusing was iterated 2 or 3 times depending on how fast it was converging ( we typically aimed at +/− 10 nm from the target defocus ) . In order to prevent errors due to objective lens hysteresis the defocus was measured without applying a defocus offset during the measurement . The phase plate provides enough contrast for beam-tilt focus measurements even close to focus . 10 . 7554/eLife . 13046 . 004Figure 2 . Data acquisition schemes that enable precise focusing superposed on a holey support film . ( A ) Two focusing areas , F1 and F2 , on opposite sides of the data acquisition position D are used to perform a linear interpolation of the measured defocus . ( B ) Multiple focusing areas F1 to F4 around a support film hole can be used to interpolate the defocus and acquire multiple data images ( D1 to D4 ) per hole . DOI: http://dx . doi . org/10 . 7554/eLife . 13046 . 00410 . 7554/eLife . 13046 . 005Figure 3 . Measured phase shift of the VPP as a function of the image number/total dose on the phase plate . The measurement was performed at two consecutive positions on the phase plate . DOI: http://dx . doi . org/10 . 7554/eLife . 13046 . 005 Figure 2B shows another possible acquisition scheme . Holey support films with larger holes permit collection of several data images ( D1 to D4 ) within a single hole . By measuring the defocus at several points around the hole ( F1 to F4 ) it will be possible to calculate a local plane or curvature model of the support film and apply the necessary focus correction for each acquisition area . Spherical aberration influences not only the optimal defocus but also the measured defocus . The most commonly used method to measure defocus is the beam tilt method ( Koster et al . , 1987 ) . It uses the shift between images taken with tilted beams to calculate the defocus . The effect of spherical aberration on the measurement is usually ignored because its contribution is small compared to the typical defocus values used in CTEM . Spherical aberration has an effect equal to a defocus of △z =CSβ2 , where CS is the spherical aberraton coefficient and β is the beam tilt angle ( Koster et al . , 1987 , equation 2 ) . A commonly used tilt angle value is 5 mrad but in our experiments we used 10 mrad in order to improve the sensitivity of the measurement and to move the spots created on the VPP by the tilted beams further away from the central spot . On an FEI Titan Krios microscope ( FEI , Hillsboro , OR ) a 10 mrad beam tilt places the beam spot at a position corresponding to k = β /λ = 0 . 01 rad / 0 . 002 nm = 5 nm-1 spatial frequency or 2 Å resolution ( λ is the electron wavelength ) . This prevents the phase shifting spots created on the phase plate by the tilted beams from disturbing the CTF in the usable frequency range . The spots themselves are beneficial for the focusing because they act as phase plates themselves and increase the contrast of the tilted beam images , thus facilitating the measurement . The offset in the measured defocus caused by spherical aberration with 10 mrad beam tilt on an FEI Titan Krios is △z =CSβ 2 = 2 . 7 mm x 0 . 012 rad2 = 270 nm . This again is a significant amount of defocus compared to the defocus limits in Figure 1 . We used SerialEM software ( Mastronarde , 2005 ) for the data acquisition which does not take into account the effect of spherical aberration on focus measurements . Therefore , we had to adjust the target defocus by the above offset , i . e . to get -20 nm actual defocus we had to set the desired defocus at 270 – 20 = 250 nm in the software . In order to characterize the behavior of the VPP we measured the phase shift under the same experimental conditions as those used to collect the 20S proteasome datasets . Series of images were recorded automatically on carbon film parts of the sample with -1 . 5 μm defocus to facilitate phase shift measurement through CTF fitting . The CTFs of the image series were fitted with the latest version of ctffind4 ( Rohou and Grigorieff , 2015 ) which supports phase shift . The results from two image series recorded at two consecutive positions on the VPP are plotted in Figure 3 . Both series show very similar behavior indicating that there are no significant variations in performance between neighboring positions on the phase plate . The phase shift has a rapid onset in the beginning of the series but the rate of its development slows down as the series progress . The results reproduce well the previously reported VPP behavior ( Danev et al . , 2014 , Figure 1B ) . During single particle data acquisition we try to prevent the phase shift from going much over π/2 by periodically moving to a new position on the phase plate ( depending on the experimental conditions , every 25 to 50 images ) . Changing the phase plate position more often increases the risk of astigmatism due to variations in the phase plate film quality . Therefore , it is desirable to have a phase plate with a slower rate of phase shift development . In practice the rate can be reduced by increasing the phase plate heating temperature ( Danev et al . , 2014 ) but in the current VPP generation such control is quite limited by the specifications of the heater . Using the acquisition scheme shown in Figure 2A we acquired two in-focus VPP datasets of the Thermoplasma acidophilum 20S proteasome , one at −20 nm and one at −50 nm defocus . From the same grid square we also acquired a CTEM dataset with a defocus range of −0 . 8 to −1 . 7 um . The datasets were acquired automatically using SerialEM macros ( Mastronarde , 2005 ) . The phase plate was automatically moved to a new area every ~1 hr ( every ~27 images ) to prevent too much phase shift buildup . Representative images from the datasets are shown in Figure 4 . The VPP image ( Figure 4A ) has higher overall contrast compared to the CTEM image ( Figure 4C ) because of the improved low frequency transfer . The high contrast is very helpful during the particle picking process . The 2D Fourier transforms shown in Figures 4B and C illustrate the CTF differences between in-focus and defocused acquisitions . The VPP ( Figure 4B ) has a uniform transfer without visible zeroes . The first CTF zero is beyond the 3 . 7 Å amorphous ice ring and is not detectable in the transform . The spectrum of the CTEM image ( Figure 4D ) shows characteristic CTF oscillations with multiple zeroes . Both power spectra exhibit a noticeable amplitude decrease in their central region which is a consequence of the relatively high dose rate on the detector ( ~9 e-/pixel/s ) . Higher dose rates increase the coincidence loss during electron counting and lead to amplitude reduction at low spatial frequencies but have little effect on the spectral signal-to-noise ratio ( Li et al . , 2013b ) . 10 . 7554/eLife . 13046 . 006Figure 4 . Representative images from 20S proteasome datasets acquired with and without a phase plate . ( A ) In-focus image acquired with the Volta phase plate . ( B ) Power spectrum of the image in ( A ) . The presence of the amorphous ice ring at 3 . 7 Å indicates that there is good information transfer to at least that spatial frequency . ( C ) Conventional defocus image at −1 . 6 μm defocus . ( D ) Power spectrum of the image in ( C ) showing CTF Thon rings . Scale bar: 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 13046 . 006 An isosurface representation of the reconstructed 3D map from the -20 nm defocus VPP dataset is shown in Figure 5A . The reconstruction is based on 13 , 469 particles selected after a 3D classification step from an initial dataset of 35 , 469 particles . Figure 5B shows a part of the map superposed on an α-helix from the β subunit of a 20S atomic model ( PDB 3J9I ) and demonstrates the presence of side chain densities . Figure 5C contains plots of the Fourier shell correlations ( FSC ) calculated by the internal 'gold standard' procedure in Relion ( Scheres , 2012 ) ( blue curve ) and versus an external 2 . 8 Å resolution density map ( EMD-6287 , Campbell et al . , 2015 ) . Both criteria give an identical resolution estimate of 3 . 2 Å at 0 . 143 level for the 'gold standard' FSC and at 0 . 5 level for the external map FSC . This is the highest resolution phase plate single particle reconstruction reported to date . 10 . 7554/eLife . 13046 . 007Figure 5 . Result from the 3D reconstruction of 20S proteasome from an in-focus dataset acquired with the Volta phase plate . ( A ) An isosurface representation of the density map . ( B ) A ribbon and a stick representations of an α-helical segment from the β subunit docked into the density map demonstrating the presence of sidechain densities . ( C ) Fourier shell correlation ( FSC ) curves from the Relion software’s internal 'gold standard' and from a comparison with an external 2 . 8 Å density map ( EMD-6287 ) . Both criteria indicate a resolution of 3 . 2 Å at a 0 . 143 level for the 'gold standard' FSC and at a 0 . 5 level for the external map . The Nyquist frequency is at 2 . 7 Å ( 0 . 37 1/nm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13046 . 007 Figure 6A contains plots of resolution versus number of particles for various datasets , collected with and without a VPP . The solid lines represent the two main VPP and CTEM datasets collected on two halves of the same grid square . For each point a complete 3D refinement run was performed in Relion using a 60 Å low-pass filtered initial model to avoid model bias . The number of particles was varied by extracting random subsets of particles . The resolutions were calculated based on the 0 . 5 level FSC versus the external map ( EMD-6287 ) . Both the VPP and the CTEM show a gradual improvement of resolution as the number of particles increases with the performance being virtually identical between the two techniques . The conventional dataset contained more micrographs ( 293 vs 158 ) hence it had more particles ( after 3D classification , 35 , 717 vs 13 , 469 ) . With more than twice the number of particles the CTEM dataset reached a resolution of 3 . 1 Å . The measured B-factors in Relion were very similar between the two datasets: 119 for CTEM and 123 for VPP . 10 . 7554/eLife . 13046 . 008Figure 6 . Resolution versus number of particles and B-factor estimations . ( A ) Resolution versus number of particles for several datasets . The resolutions were calculated based on the 0 . 5 FSC level versus an external 2 . 8 Å density map ( EMD-6287 ) . For two of the datasets ( red and blue lines ) the number of particles was varied artificially by using random subsets of particles . ( B ) Logarithm of the number of particles versus the squared reciprocal resolution plots of the two main datasets in ( A ) . The legend contains B-factors estimated by linear fits of two data regions . DOI: http://dx . doi . org/10 . 7554/eLife . 13046 . 008 In order to estimate the B-factors from the resolution versus number of particles data we plotted the logarithm of the number of particles as a function of the squared reciprocal resolution ( Figure 6B ) . The B-factor can then be estimated by a linear fit and is equal to twice the slope of the line ( Rosenthal and Henderson , 2003 , Figure 11 ) . We noticed two approximately linear regions in both the CTEM and the VPP data . The region with particle numbers between 500 and 2000 has a B-factor of ~100 and resolutions above 3 . 6 Å ( solid lines in Figure 6B ) . The higher resolution region with particle numbers above 2000 has a B-factor of ~200 ( dashed lines in Figure 6B ) . Figure 6A also contains individual points for other datasets . A VPP dataset collected at -50 nm defocus and consisting of 6340 particles reached a resolution of 3 . 5 Å which is marginally ( 0 . 2 Å ) worse than the same number of particles from the -20 nm defocus VPP dataset . According to the values of the phase shift in Figure 3 and comparing them with the curves in Figure 1C the -50 nm defocus dataset should have matched or exceeded the performance of the -20 nm defocus dataset . This would suggest that either there was a small systematic focusing error or a variation in the phase plate performance . The black square in Figure 6A is for a VPP dataset collected on an FEI Tecnai F20 microscope equipped with an FEI Falcon II ( FEI , Hillsboro , OR ) direct detector camera . This microscope operates at 200 kV and uses a side-entry cryo-holder . It is inferior to the FEI Titan Krios microscope in both optical performance and specimen stage stability but the obtained resolution of 4 . 5 Å is still respectable for such a system . The orange circle shows the result from Campbell et al . ( 2015 ) , and is at 2 . 8 Å with 49 , 954 particles . In that work a much larger dataset of ~1000 micrographs was collected which was then subjected to CTF screening and only 193 micrographs ( ~19% of the original dataset ) were selected for data processing . We performed only visual pre-screening of the datasets and excluded micrographs that had obvious faults , such as broken ice , large contaminants or duplicate exposures . Consequently , we used ~80% of the original micrographs for data processing . The smaller number of micrographs combined with the fact that our data was not collected in 'super resolution' mode of the K2 camera ( the data in Campbell et al , 2015 , was collected in 'super resolution' mode ) could explain the slightly lower resolution of our reconstructions . To compare the performance for quick sample screening we calculated reconstructions using the first ~10 micrographs from each dataset rather than 'cherry picking' the best micrographs . For CTEM we had to use the first 12 micrographs in order to match the particle number of the first 10 VPP micrographs . The ~2000 picked particles from each dataset were subjected directly to a single 3D refinement run in Relion without initial classification , movie processing or particle 'polishing' . The results are shown with triangles in Figure 6A . The resolution of the VPP reconstruction was better than the CTEM one by ~ 0 . 3 Å ( 3 . 9 Å vs 4 . 2 Å ) . This test shows that VPP may have an advantage over CTEM for initial screening of samples and/or samples containing a mixture of good and bad particles . Although the 3 . 2 Å resolution is a remarkable achievement for a phase plate , the performance is still far from what is theoretically expected both in terms of absolute number of particles required to reach a given resolution and in terms of advantage over the conventional approach ( Chang et al . , 2010 , Table 2 , pol II data ) . Further practical experience with the phase plate will help to improve the performance . The main factors that currently limit the performance are focusing accuracy , spherical aberration and variation in phase shift . We used a constant amount of defocus for the VPP datasets but a better approach would be to adjust the amount of defocus according to the phase shift . As shown in Figure 1C the optimal defocus varies with phase shift . The amount of defocus could be adjusted depending on the number of images already acquired at the current phase plate position . The phase shift versus number of images has to be calibrated in advance , as shown in Figure 3 . Furthermore , the defocused acquisition approach with a phase plate has to be explored . It is much simpler from a data collection point of view because it does not require precise focusing and has a higher data acquisition throughput , equal to that of CTEM . Future tests are needed to determine which approach is more efficient and to establish if either approach has an advantage for difficult samples . The VPP matches or slightly exceeds the performance of CTEM for the same number of particles . Because of the high contrast it provides the VPP could be an ideal tool for quick sample screening and/or initial model building . With the in-focus approach the need for precise focusing reduced the image throughput about ~1 . 7 times ( VPP 27 vs CTEM 45 micrographs per hour ) . Furthermore , setting up VPP experiments is more demanding in terms of user experience . Thus for 'easy' targets , such as relatively big particles and/or such with high symmetry the CTEM can still be more time efficient for achieving high resolutions . In this work we used such a target in order to evaluate the overall performance of the VPP . The full potential of the VPP in terms of improved contrast and uniform spectral coverage will be truly tested and demonstrated by difficult targets , such as small , flexible or heterogeneous samples ( Hall et al . , 2011; Khoshouei et al . , 2016 ) . We expect that for such targets the extra effort for using the VPP will pay off in terms of getting better reconstructions or a reconstruction at all . Therefore , the possibilities offered by the VPP could help to widen the target size overlap between cryo-EM and x-ray crystallography .
Thermoplasma acidophilum 20S proteasomes were recombinantly expressed in Escherichia coli and purified as described in Zwickl et al . ( 1992 ) . Samples were plunge-frozen on an FEI Vitrobot Mark III ( FEI , Hillsboro , OR ) . Quantifoil 200 mesh copper R 1 . 2/1 . 3 ( Quantifoil , Großlöbichau , Germany ) holey carbon grids were first cleaned by placing them on a piece of Whatman No . 1 ( Whatman , Maidstone , UK ) filter paper in a glass Petri dish and then saturating the paper with acetone until the grids were soaked . The Petri dish was left partially open until the acetone evaporated completely . Shortly before plunging the cleaned grids were glow discharged for 30 s in low pressure air in a Harrick plasma cleaner ( Harrick , Ithaca , NY ) . 3 μl of 0 . 5 mg/ml protein solution was applied on a grid in the Vitrobot chamber set to 95% RH at 20°C then blotted for 5 s and plunged into liquid 37% ethane , 63% propane mixture . Excess cryogen was blotted off from the grid with a piece of filter paper held just above the surface of the liquid nitrogen surrounding the cryogen cup before placing the grid in a plastic cryo grid box for storage . Data was collected on an FEI Titan Krios microscope operated at 300 kV and on an FEI Tecnai F20 microscope operated at 200 kV ( FEI , Hillsboro , OR ) . The FEI Titan Krios was equipped with a Gatan GIF Quantum energy filter , a Gatan K2 Summit direct detector ( Gatan , Pleasanton , CA ) and an FEI phase plate ( FEI , Hillsboro , OR ) . The acquisition conditions on the FEI Titan Krios were as follows: EFTEM microprobe mode , magnification 37 , 000x , 50 μm C2 aperture , spot size 6 , 1 . 4 μm beam diameter , zero-loss imaging with 20 eV slit , K2 Summit in counting mode , pixel size 1 . 35 Å , total dose 30 e-/ Å2 , dose rate on the detector 9 . 1 e-/pixel/s , exposure time 6 s , 12 frames 0 . 5 s each . The FEI Tecnai F20 was equipped with an FEI Falcon II direct detector , a Gatan 626 cryo-holder and an FEI phase plate . The acquisition conditions on the FEI Tecnai F20 were as follows: nanoprobe mode , magnification 135 , 000x , 50 μm C2 aperture , spot size 5 , pixel size 1 . 04 Å , total dose 30 e-/ Å2 , exposure time 2 s , 7 frames 0 . 29 s each . On both microscopes we used SerialEM software ( Mastronarde , 2005 ) with custom macros for automated single particle data acquisition . For the VPP datasets we used a macro to realize the dual focusing scheme shown in Figure 2A with focusing parameters in SerialEM set to: beam tilt angle 10 mrad , focus offset 0 , drift protection ON ( three image focusing ) . The focusing mode had identical optical settings as the record mode . The focusing macro measured the defocus on opposite sides of the acquisition area and used the average to correct the defocus . The macro was iterated 3 times to improve the accuracy . To take into account the effect of spherical aberration on the measured defocus ( △z =CSβ 2 ) 270 nm was added to the target defocus on the FEI Titan Krios ( spherical aberration 2 . 7 mm ) , i . e . 250 nm target defocus for the -20 nm VPP dataset and 220 nm for the -50 nm VPP dataset . On the FEI Tecnai F20 ( spherical aberration 2 . 1 mm ) 210 nm was added , i . e . 190 nm target defocus for the -20 nm VPP dataset . The phase plate was automatically moved to a new position with a 20 μm step every one hour . For the CTEM datasets a single focusing was performed on one side of the acquisition area with drift protection set to OFF in SerialEM to provide more defocus spread . The dataset was collected with a 70 μm objective aperture and a defocus range of -0 . 8 to -1 . 7 μm . The acquisition speed was ~27 images/hour for the VPP and ~45 images/hour for the CTEM . The datasets consisted of: Titan Krios VPP −20 nm defocus 200 images , VPP −50 nm defocus 111 images , CTEM 338 images and Tecnai F20 VPP -20 nm defocus 199 images . The phase shift data in Figure 3 was recorded on the FEI Titan Krios microscope . The optical and acquisition parameters were set to the same values as for the 20S datasets . The beam current was measured by the fluorescent screen of the microscope to be 0 . 162 nA . Thus the 6 s exposure per image added 0 . 162 nA x 6 s = 0 . 97 nC dose to the phase plate . Images were recorded using the same SerialEM macros as for the 20S datasets except the record area was set to be on the carbon film part between the holes of a Quantifoil R 1 . 2/1 . 3 plunge-frozen grid . Before each image focusing was performed only once on an adjacent carbon film area with target defocus set to -1 . 5 μm . The images and their Fourier transforms were visually inspected and images with visual faults and/or double exposures were rejected . After the initial screening approximately 80% of the original images remained in the datasets: Titan Krios VPP -20 nm defocus 158 images , VPP −50 nm defocus 86 images , CTEM 293 images and Tecnai F20 VPP −20 nm defocus 180 images . The frame stacks were aligned with a homemade GPU accelerated software based on the algorithm of Li et al . ( 2013a ) . Particles from the Titan Krios datasets were picked reference-free with the e2boxer program from the EMAN2 software package ( Tang et al . , 2007 ) . Particles from the Tecnai F20 dataset were reference-based picked with the Signature software ( Chen et al . , 2007 ) . The total number of picked particles were: Titan Krios VPP −20 nm defocus 35 , 469 particles , VPP −50 nm defocus 16 , 919 particles , CTEM 63 , 474 particles and Tecnai F20 VPP −20 nm defocus 28 , 016 particles . To remove false positives and contaminants from each dataset we performed 3D classification runs in Relion 1 . 4 software ( Scheres , 2012 ) with 6 classes and D7 symmetry . The particles from the best looking class were extracted with the numbers being: Titan Krios VPP -20 nm defocus 13 , 469 particles ( 38% ) , VPP -50 nm defocus 6 , 340 particles ( 38% ) , CTEM 35 , 717 particles ( 56% ) and Tecnai F20 VPP -20 nm defocus 16 , 145 particles ( 58% ) . Those particles were then used to perform a 3D refinement in Relion followed by movie processing , particle 'polishing' and another 3D refinement of the 'polished' particles . The B-factors reported by Relion after post-processing of the final 3D maps were: Titan Krios VPP −20 nm defocus −122 . 6 , VPP -50 nm defocus -124 , CTEM -119 . 3 and FEI Tecnai F20 VPP −20 nm defocus -330 . 4 . In order to calculate the resolution versus number of particles data in Figure 6 random subsets with varying number of particles were extracted from the Titan Krios VPP −20 nm and CTEM datasets . Each subset was subjected to a complete 3D refinement run with a 60 Å low-pass filtered initial model . | One way of investigating how proteins and other biological molecules work is to look at their structure . Light microscopes cannot produce detailed enough images to fully reveal these structures , and so a technique called cryo-electron microscopy is often used instead . In this technique , a biological sample is frozen to the temperature of liquid nitrogen and a beam of electrons is fired at it to create an image . By taking many of these images and then subjecting them to computer processing it is possible to reconstruct the three-dimensional structure of the molecule . Frozen biological samples are essentially transparent to the electron beam used in an electron microscope . To view samples , researchers therefore use a method called phase contrast , which relies on a property of the electron beam ( called its phase ) changing as the beam passes through the sample . The traditional “defocus” method of producing phase contrast from electron microscopy relies on processing a series of slightly out-of-focus images of the sample . Phase plates are add-on devices that are commonly used in light microscopes to produce phase contrast . For many years now , attempts have been made to produce a working phase plate for electron microscopes . However , an effective plate , called the Volta phase plate , has only recently been developed . Danev and Baumeister have now evaluated how well the Volta phase plate performs during the analysis of a single , relatively large protein . This molecule is considered ‘easy’ to analyze using cryo-electron microscopy as relatively few microscopic images need to be recorded to solve the protein’s structure . Danev and Baumeister found that the Volta phase plate matched or slightly exceeded the performance of the traditional defocus method of producing phase contrast , depending on how many images were used to analyze the protein . This is the first time that a phase plate has matched the performance of the defocus method . A future challenge will be to make the experimental procedures and the software involved in using the Volta phase plate more user-friendly . The phase plate also needs to be tested with more ‘difficult’ samples , such as small proteins and samples whose structure could not be established using the defocus method of producing phase contrast . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"tools",
"and",
"resources"
] | 2016 | Cryo-EM single particle analysis with the Volta phase plate |
Antibodies acquired naturally through repeated exposure to Plasmodium falciparum are essential in the control of blood-stage malaria . Antibody-dependent functions may include neutralization of parasite–host interactions , complement activation , and activation of Fc receptor functions . A role of antibody-dependent cellular cytotoxicity ( ADCC ) by natural killer ( NK ) cells in protection from malaria has not been established . Here we show that IgG isolated from adults living in a malaria-endemic region activated ADCC by primary human NK cells , which lysed infected red blood cells ( RBCs ) and inhibited parasite growth in an in vitro assay for ADCC-dependent growth inhibition . RBC lysis by NK cells was highly selective for infected RBCs in a mixed culture with uninfected RBCs . Human antibodies to P . falciparum antigens PfEMP1 and RIFIN were sufficient to promote NK-dependent growth inhibition . As these results implicate acquired immunity through NK-mediated ADCC , antibody-based vaccines that target bloodstream parasites should consider this new mechanism of action .
Plasmodium falciparum ( P . falciparum . ) , the causative agent of malaria , expresses proteins that are displayed at the surface of infected red blood cells ( RBCs ) . Some of these proteins promote sequestration of P . f . -infected RBCs ( iRBCs ) through adhesion to vascular endothelial cells ( Miller et al . , 2002 ) . Humans living in malaria-endemic areas generate , over years of repeated infections , antibodies ( Abs ) to P . f . proteins that contribute to the gradual protection from malaria symptoms ( Boyle et al . , 2015; Bull and Marsh , 2002; Cohen et al . , 1961; Mayor et al . , 2015; Tran et al . , 2013 ) . One of the main objectives in malaria research is to define the mechanisms by which naturally acquired Abs provide protection ( Cohen et al . , 1961; Crompton et al . , 2014 ) . Acquired immunity to malaria is complex as it requires a balance of parasite growth inhibition and control of inflammation ( Zhou et al . , 2015 ) . Neutralizing Abs that prevent P . f . merozoite invasion of RBCs have been described ( Douglas et al . , 2011 ) . However , as merozoites released from late-stage iRBCs rapidly invade uninfected RBCs ( uRBCs ) , high antibody titers are likely needed for inhibition . Abs bound to iRBCs promote phagocytosis by myeloid cells , and Abs bound to merozoites activate the complement pathway ( Bouharoun-Tayoun et al . , 1995; Boyle et al . , 2015; Rowe et al . , 1997 ) . Natural killer ( NK ) cells constitute about 10% of peripheral blood lymphocytes . They kill virus-infected cells and tumor cells through engagement of an array of germ-line encoded co-activation receptors ( Bryceson et al . , 2006a; Cerwenka and Lanier , 2001 ) . In addition to their innate ability to eliminate transformed and infected cells , NK cells perform Ab-dependent cellular cytotoxicity ( ADCC ) through the low-affinity IgG receptor FcγRIIIa ( also known as CD16 ) , thereby killing IgG-coated target cells and secreting pro-inflammatory cytokines such as IFN-γ and TNF-α . A clear role of NK cells in contributing to protection from malaria , and whether iRBCs could be eliminated through ADCC by NK cells , has not been established ( Wolf et al . , 2017 ) . Earlier studies have described direct lysis of iRBCs by NK cells in the absence of Abs or Ab-dependent inhibition of P . f . growth by NK cells ( Mavoungou et al . , 2003; Orago and Facer , 1991 ) . However , other studies have not confirmed such results ( Wolf et al . , 2017 ) . Here , we present a detailed study of the activity of primary , unstimulated human NK cells mixed with RBCs , infected or not by P . f . , and evaluate the NK cell responses using several different quantitative assays . We found that IgG in plasma from subjects living in a malaria-endemic region in Mali bound to iRBCs and induced their rapid lysis through NK-mediated ADCC . Naturally acquired IgG specific for the major P . f . antigen PfEMP1 was sufficient to promote NK-dependent inhibition of P . f . growth in RBCs . Our results demonstrated that primary human NK cells alone are capable of controlling parasite growth in vitro in response to IgG from subjects exposed to malaria . This may represent an important component of Ab-dependent clinical immunity to P . f . blood-stage infection that could be exploited in the development of malaria vaccines .
RBCs infected with P . f . strain 3D7 were enriched for the presence of knobs at the RBC surface ( Figure 1—figure supplement 1A ) . Knobs are protrusions at the surface of iRBCs that appear at the trophozoite stage . iRBC cultures were enriched for the trophozoite stage by percoll-sorbitol gradient . Enrichment was confirmed by Giemsa stain ( Figure 1—figure supplement 1B ) . A pool of plasma from malaria-exposed adults living in a high-transmission region of Mali ( Mali plasma ) was tested for the presence of Abs to the surface of P . f . 3D7-iRBCs at the trophozoite stage by flow cytometry . Adults at the Mali study site are considered ‘semi-immune’ to malaria , as they generally control parasitemia and rarely experience malaria symptoms ( Tran et al . , 2013 ) . Abs in Mali plasma stained iRBCs but not uRBCs ( Figure 1A ) . In contrast , Abs in a pool of serum from malaria-naïve US adults ( US serum ) did not bind to iRBCs any more than they did to uRBCs ( Figure 1A ) . Binding of Abs in Mali plasma to iRBCs was confirmed by immunofluorescence microscopy ( Figure 1B ) . Lower magnification images of mixed uRBCs and iRBCs showed that staining by Mali plasma was selective for iRBCs ( Figure 1—figure supplement 1C ) . We tested the reactivity of primary NK cells , freshly isolated from the blood of healthy malaria-naïve US donors , to iRBCs in the absence of Abs . NK cells did not degranulate during co-incubation with iRBCs , as monitored by staining with anti-LAMP-1 ( CD107a ) Ab ( Figure 1C and D ) . As binding of FcγRIIIA to IgG alone is sufficient to induce activation of resting NK cells ( Bryceson et al . , 2005 ) , IgG bound to RBCs has the potential to induce NK cell degranulation and cytokine production . We first tested stimulation of NK cells in the presence of a polyclonal serum of rabbits that had been immunized with human RBCs . Degranulation by NK cells occurred during incubation with iRBCs in the presence of rabbit anti-RBC Abs ( Figure 1C and D ) . Notably , potent NK cell degranulation occurred with iRBCs in the presence of Mali plasma , whereas US serum induced degranulation in a very small fraction of NK cells ( Figure 1C and D ) . NK cell expression of intracellular interferon ( IFN ) -γ and tumor necrosis factor ( TNF ) -α was also stimulated equally well by rabbit anti-RBC serum and Mali plasma , whereas US serum did not induce cytokine production ( Figure 1E and F and Figure 1—figure supplement 1D ) . These results suggested that Abs from malaria-exposed individuals activate NK cells when bound to iRBCs , which results in NK cell degranulation and production of cytokines . We next investigated whether NK cells could selectively lyse Ab-coated iRBCs without causing bystander lysis of uRBCs . uRBCs and iRBCs were labeled with either eFluor450 or eFluor670 dyes , which bind cellular proteins containing primary amines , and NK cells were labeled with the lipophilic dye PKH67 . The three cell types were incubated together at equal numbers , and examined by live microscopy . Images were acquired in a temperature-controlled chamber every 30 s for several hours ( Videos 1 and 2 ) . Representative images taken at 0 , 2 , and 4 hr are shown in Figure 2A . Quantitative analysis of cell numbers , which were determined every minute , showed that all three cell types remained at a constant ratio when incubated with US serum ( Figure 2B , left panel ) . In contrast , NK cells selectively lysed iRBCs in the presence of Mali plasma , leaving uRBCs intact ( Figure 2B , right panel ) . A compilation of four experiments , each performed with NK cells from a different donor , showed iRBC lysis induced by Mali plasma but not US serum ( Figure 2C , left panel ) , and selective lysis of iRBCs in the presence of Mali plasma ( Figure 2C , right panel ) . The relative change in the frequency of uRBCs and iRBCs over 3 hr in the presence of US serum or Mali plasma is shown in Videos 1 and 2 . We concluded that lysis of P . f . -iRBCs by NK cells in the presence of plasma from malaria-exposed individuals was efficient and specific , causing minimal bystander lysis of uRBCs . The fraction of RBCs invaded by merozoites ( also known as parasitemia ) in P . f . -infected individuals typically ranges from 0 . 005 to 5% ( Gonçalves et al . , 2014 ) . We tested the ability of NK cells to inhibit parasite growth in an RBC culture that was set at 1% parasitemia . As the ratio of NK cells to iRBCs was set at 3:1 and 1:1 , NK cells were outnumbered by a 30 to 100 fold excess of uRBCs during incubation . Cultures were maintained for 48 hr before analysis ( Figure 2—figure supplement 1A ) . Given that iRBC cultures were synchronized at the ring stage and enriched at the trophozoite-stage , a single round of RBC rupture and reinvasion of fresh RBCs by released merozoites occurred in the next ~18 hr . P . f . growth was determined by counting iRBCs in blood smears . In the absence of Abs , growth inhibition was 5 . 69 ± 11 . 53% at an E:T ratio of 3 ( Figure 2D ) . A similar result was obtained in the presence of US serum ( 4 . 33 ± 12 . 15%; Figure 2D ) . In contrast , in the presence of Mali plasma , inhibition of parasite growth was 62 . 56 ± 15 . 41% at an E:T ratio of 3 ( Figure 2D ) . Strong growth inhibition occured also at an E:T ratio of 1 ( Figure 2D ) . A much reduced inhibition occurred with Mali plasma in the absence of NK cells ( 11 . 55 ± 1 . 99% ) , which could be due to Ab-mediated inhibition of merozoite reinvasion . We concluded that NK cells , in the presence of plasma from malaria-exposed individuals , are capable of inhibiting blood-stage P . f . growth even in the presence of a 100-fold excess of uRBCs . The results further suggested that maturation of trophozoites and schizonts into infectious merozoites was inhibited by NK-mediated ADCC toward iRBCs . The standard growth inhibition assay ( GIA ) ( Malkin et al . , 2005 ) was modified to remove Abs that inhibit P . f . growth through neutralization of merozoites . NK cells were first co-incubated with trophozoite-enriched iRBCs for 6 hr , in the presence or absence of Mali plasma . Cultures were then washed to remove unbound Abs and soluble factors prior to addition of a 100-fold excess of fresh uRBCs . Cultures were further incubated for 16 hr to allow for a single round of merozoite release and reinvasion of uRBCs ( Figure 2—figure supplement 1B ) . We refer to this assay for inhibition by NK-mediated ADCC as GIA-ADCC . Parasitemia at the end of the experiment was determined by flow-cytometry ( Figure 2—figure supplement 1C ) . Inhibition of P . f . growth occurred in the presence of Mali plasma but not in the presence of US serum ( Figure 2E ) . These results showed that inhibition of P . f . growth was due to Abs bound to iRBCs prior to the release of merozoites and the addition of uRBCs , confirming the role of NK cell-mediated ADCC . As NK cell-mediated ADCC triggered by FcγRIIIa is dependent on binding to IgG , we tested whether IgG in Mali plasma was sufficient to promote NK-dependent inhibition of P . f . growth . IgG purified from US serum did not bind to uRBCs ( Figure 2—figure supplement 1D ) or to trophozoite-stage iRBCs ( Figure 2F ) , whereas IgG purified from Mali plasma bound to iRBCs ( Figure 2F ) but not uRBCs ( Figure 2—figure supplement 1D ) . In the GIA-ADCC , designed to exclude merozoite neutralization as the basis for inhibition , purified IgG from Mali plasma inhibited P . f . growth ( 37 . 59 ± 12 . 15% inhibition at an IgG concentration of 1 . 8 mg/ml ) ( Figure 2G ) . No inhibition was observed with IgG purified from US serum ( Figure 2G ) . The use of purified IgG eliminated the possibility that the difference observed between Mali and US individuals was due to properties unique to human plasma that were absent in human serum . These results demonstrated that IgG from malaria-exposed individuals promotes inhibition of P . f . growth in RBCs in the presence of NK cells . The P . f . erythrocyte membrane protein 1 ( PfEMP1 ) , which mediates parasite sequestration through binding to vascular endothelial cells , is a major target of host Ab responses ( Bull and Marsh , 2002; Chan et al . , 2012 ) . We used the parasite line DC-J , which lacks PfEMP1 expression ( Dzikowski et al . , 2006 ) , to test the importance of PfEMP1 in promoting Ab-dependent NK cell activation . Staining of P . f . DC-J-iRBCs with Mali plasma gave a positive signal that was approximately one log less than staining of 3D7-iRBCs ( Figure 3A ) , but greater than staining of P . f . DC-J-iRBCs with US serum ( Figure 3B ) . Time-lapse imaging was used to monitor lysis of DC-J-iRBCs by NK cells in the presence of Mali plasma during co-incubation with an equal number of uRBCs ( Figure 3C ) . NK cells did not lyse DC-J-iRBCs in the presence of Mali plasma ( Video 3 ) . Data from four experiments with NK cells from different donors indicated no significant decrease in iRBCs in the presence of Mali plasma compared to US serum over the course of 3 hr ( Figure 3D ) . Therefore , residual Ab-binding in the absence of PfEMP1 ( Figure 3A ) was not sufficient , under the conditions used , to promote NK-mediated ADCC in the presence of Mali plasma . We wanted to test whether the lack of lysis of P . f . DC-J-iRBCs by NK cells could perhaps be due to an intrinsic resistance of DC-J to NK-mediated lysis . To test it we used the rabbit anti-serum raised against human RBCs , which activated degranulation by NK cells in the presence of 3D7-iRBCs ( Figure 1C and D ) . We further developed a quantitative RBC lysis assay based on hemoglobin ( Hb ) release into the supernatant . Maximum Hb release from RBCs was defined as Hb in detergent lysates of RBCs ( Figure 3—figure supplement 1A ) . This control also served to compensate for the loss of Hb during P . f . development in RBCs , as the parasite digests some of the Hb to produce hemozoin . Severe damage to RBCs , as determined by Hb release , occurred at NK cell to RBC ratios of 3:1 and 10:1 , after a 5 hr incubation with 3D7-iRBCs in the presence of rabbit anti-RBC Abs ( Figure 3—figure supplement 1B and Figure 3—figure supplement 1C ) . At an NK cell to iRBC ratio of 5:1 , 47 . 16 ± 8 . 76% of total Hb content was released ( Figure 3E ) . The timing of eukaryotic evolution Does a relaxed molecular clock reconcile proteins and fossils ? A small amount of Hb was released in the absence of NK cells ( 1 . 36 ± 0 . 51% ) and in the absence of rabbit anti-RBC Abs ( 4 . 38 ± 1 . 65% ) ( Figure 3E ) . Similar data were obtained with uRBCs under the same conditions , where 47 . 15 ± 13 . 1% of total Hb content was released ( Figure 3E ) . We concluded that uRBCs and 3D7 P . f . -iRBCs were equally sensitive to NK-mediated ADCC . This approach gave us an opportunity to test whether DC-J-iRBCs were inherently resistant to NK-mediated ADCC . Lysis assays in the presence of rabbit anti-RBC Abs and NK cells were performed . The extent of Hb release ( 52 . 66 ± 11 . 34% ) after incubation at an NK cell to DC-J-iRBC ratio of 5:1 for 5 hr was comparable to that obtained with uRBCs and 3D7-iRBCs ( Figure 3E ) . Hemoglobin release in the absence of rabbit anti-human RBC Abs was minimal . Therefore , we concluded that the lack of lysis of DC-J-iRBCs in the presence of Mali plasma was not due to resistance to NK-dependent ADCC , but rather to the low amount of Abs bound to RBCs infected with this PfEMP1-deficient parasite strain . Together , these data suggested that naturally acquired Abs to PfEMP1 play a critical role in NK cell-mediated destruction of iRBCs . Abs with broad reactivity against certain members of the RIFIN family of P . f . proteins have recently been cloned from memory B cells of malaria-exposed individuals in Kenya , Mali and Tanzania ( Pieper et al . , 2017; Tan et al . , 2016 ) . Similar to PfEMP1 , RIFIN is a type of variant antigen expressed on the surface of iRBCs . Using a P . f . 3D7 strain enriched for expression of RIFIN ( Figure 4A ) , we tested NK-dependent lysis of RIFIN+ iRBCs in the presence of the RIFIN-specific human monoclonal Ab MGD21 . Lysis occurred during incubation with NK cells at an NK cell to iRBC ratio of 5:1 for 6 hr , as measured by Hb release ( Figure 4B ) . Negligible lysis was observed in the absence of NK cells or in the absence of MGD21 . We then tested a variant of monoclonal Ab MGD21 ( MGD21-LALA ) , into which mutations had been introduced in the Fc to reduce binding to Fc receptors ( Tan et al . , 2016 ) . Staining of iRBCs indicated that MGD21 and MGD21-LALA bound similarly to RIFIN+ iRBCs ( Figure 4A ) . However , in the presence of NK cells , only MGD21 , and not MGD21-LALA , induced Hb release ( Figure 4B ) , demonstrating that an intact Fc receptor-binding site was required for NK cell stimulation . In addition , we concluded that P . f . antigens other than PfEMP1 have the potential to induce NK-dependent ADCC in the presence of specific Abs . The timing of eukaryotic evolution Does a relaxed molecular clock reconcile proteins and fossils ? To further define antigenic epitopes with the potential to induce NK-dependent ADCC toward P . f . -iRBCs , we tested polyclonal , affinity-purified IgG from rabbits that had been immunized with the Duffy binding-like 3x ( DBL3X ) domain of the PfEMP1 variant VAR2CSA ( Obiakor et al . , 2013 ) . This rabbit IgG stained VAR2CSA-iRBCs , as measured by flow cytometry ( Figure 4C ) . The VAR2CSA-specific rabbit IgG , but not control rabbit serum IgG , induced Hb release from VAR2CSA-iRBCs after incubation with NK cells ( Figure 4D ) . Uninfected RBCs were not lysed in the presence of VAR2CSA-specific rabbit IgG ( Figure 4—figure supplement 1A ) . These results showed that domain DBL3X was accessible to Abs at the surface of VAR2CSA-iRBCs , and oriented in such a way that bound Abs could engage FcγRIIIA on NK cells . To test the potential of naturally acquired Abs to VAR2CSA PfEMP1 to promote NK-dependent inhibition of P . f . growth in RBCs , we used human IgG isolated from pooled plasma of multigravid women , and affinity-purified on DBL domains of VAR2CSA PfEMP1 ( Doritchamou et al . , 2016 ) . This natural IgG stained VAR2CSA-iRBCs , as measured by flow cytometry , whereas human IgG Abs against another parasite antigen , AMA1 , did not ( Figure 4E ) . NK cells were incubated with trophozoite-stage enriched RBCs infected by P . f . VAR2CSA for 6 hr in the presence of IgG Abs . A 100-fold excess of uRBCs was added and incubation resumed for 42 hr . In the presence of naturally acquired human IgG specific for VAR2CSA PfEMP1 , P . f . growth was inhibited by 49 . 88 ± 8 . 49% , which was similar to inhibition obtained with rabbit anti-DBL3X IgG ( 60 ± 11 . 29% ) ( Figure 4F , Figure 4—figure supplement 1B ) . No inhibition was observed in the absence of NK cells ( Figure 4—figure supplement 1C ) . These results showed that naturally acquired Abs to PfEMP1 induce NK-mediated ADCC , which inhibits parasite growth in RBCs .
The main objective of our study was to test whether NK cells could help control blood-stage malaria by lysing iRBCs through ADCC . Considering the essential role of Abs in conferring clinical immunity to individuals living in areas of high P . f . transmission ( Cohen et al . , 1961 ) , and the limited efficacy of malaria vaccines tested so far , any immune effector function that depends on Abs needs to be evaluated . We provide strong evidence of Ab-dependent NK cell cytotoxicity towards P . f . -iRBCs in the presence of Abs from malaria-exposed individuals in Mali . NK cell responses to iRBCs and their effect on P . f . growth in culture were tested using primary , unstimulated human NK cells . NK cells isolated from the blood of individuals in the US were used . It is possible that NK cells from individuals exposed to malaria may have altered function , or a reduced cell frequency , as reported following a controlled human malaria infection ( Mpina et al . , 2017 ) . Lysis of iRBCs by NK cells , in the presence of Abs to P . f . antigens exposed at the surface of iRBCs , was highly selective , leaving most uRBCs intact . NK cell-mediated ADCC inhibited P . f . growth in RBC cultures . Human Abs specific for a single class of P . f . antigens expressed at the surface of RBCs , such as PfEMP1 and RIFIN , were sufficient to induce NK cell cytotoxicity and P . f . growth inhibition . We propose that NK-dependent ADCC may be an effective mechanism to limit parasite growth , as it combines the powerful cytotoxicity of innate NK cells with the specificity of Abs generated by adaptive immunity . RIFIN antigens may help P . f . evade immune detection by binding to inhibitory receptors on human lymphocytes ( Saito et al . , 2017 ) . However , we observed hemoglobin release after mixing NK cells with RIFIN+ P . f . -infected RBCs in the presence of a human monoclonal Ab to RIFIN . NK cell activation was not due to the masking of RIFIN by the monoclonal Ab ( and release from inhibition ) but to activation of ADCC by the human IgG1 mAb because the same antibody lacking FcR binding did not activate NK cells . The developmental cycle of P . f . in iRBCs provides a window of opportunity for Ab-dependent immune effector responses . Following merozoite invasion of RBCs , P . f . proteins begin to appear at the RBC surface after 16–20 hr and remain exposed until infectious merozoites are released . Once released , merozoites rapidly invade fresh RBCs ( Boyle et al . , 2010 ) . Therefore , it is likely that high Ab titers are needed to neutralize merozoites and block entry into RBCs . In contrast , RBCs harboring non-infectious P . f . , as it progresses through trophozoite and schizont stages , display P . f . antigens at their surface for more than 24 hr , and ADCC responses activated by FcγRIIIa in primary human NK cells are rapid , strong and independent of coactivation signals ( Bryceson et al . , 2005 ) . Evidence of NK cell activation and RBC lysis was obtained with three different assays: ( 1 ) NK cell degranulation and cytokine production by flow cytometry , ( 2 ) loss of intact P . f . -iRBCs by live imaging , and ( 3 ) RBC lysis by measurement of Hb release . NK cell degranulation in a co-culture with P . f . -iRBCs , selectively induced by Abs from malaria-exposed individuals , was just as strong as that obtained with rabbit polyclonal antiserum raised against human RBCs . Furthermore , using Hb-release assays and rabbit anti-RBC serum , it was possible to show that iRBCs are not inherently more resistant or sensitive than uRBCs to NK-mediated lysis . Live imaging of a coculture of NK cells , uRBCs and P . f . -iRBCs , in the presence of plasma from malaria-exposed individuals , revealed selective lysis of iRBCs , with no 'bystander' lysis of uRBCs . Natural cytotoxicity of NK cells towards P . f . -iRBCs was not detected in our assays with resting NK cells . A recent study in humanized mice reconstituted with human lymphocytes and injected with P . f . -infected human RBCs , reported some lysis of P . f . -iRBCs by NK cells ( Chen et al . , 2014 ) . It is possible that under specific stimulatory conditions , including soluble factors and contact with other cells , human NK cells exhibit natural cytotoxic responses towards P . f . -iRBCs . However , considering that clinical immunity to malaria depends in large part on Abs , and that development of an effective vaccine is a high priority , we chose to focus on ADCC by NK cells . Signaling in NK cells by FcγRIIIa alone , independently of other co-activation signals and of integrin-dependent adhesion , is sufficient to induce strong responses , unlike other NK activation receptors , which require synergy through combinations of co-activation receptors ( Bryceson et al . , 2005; 2006b ) . The Ab-mediated activation of NK cell cytotoxicity described here is adding a strong effector mechanism to the other mechanisms by which Abs may confer protection against malaria , including neutralizing Abs and Abs that activate the complement pathway ( Boyle et al . , 2015 ) . We have shown that NK cell-mediated ADCC inhibits the growth of P . f . in RBC cultures in the presence of Abs to P . f . antigens expressed at the surface of iRBCs in a standard growth inhibition assay ( GIA ) by co-incubation of iRBCs and NK cells with a large excess of uRBCs . To distinguish inhibition by NK cells from other Ab-dependent functions , such as merozoite neutralization and activation of complement , we developed a two-step GIA to evaluate inhibition that had occurred prior to iRBC rupture and reinvasion of fresh RBCs . As inhibition of P . f . growth occurred in the presence of purified IgG from plasma of malaria-exposed individuals , other serum components were not required for NK-mediated inhibition . The modified GIA for ADCC could help define P . f . antigens that induce Abs of sufficient titer and quality for FcR activation ( e . g . IgG isotype , glycosylation ) . The GIA-ADCC is well-suited to large screens of plasma from subjects in malaria vaccine trials or in studies of naturally acquired immunity to malaria . Previous work has shown that IL-2 produced by T cells following malaria infection or injection of a malaria vaccine activates IFN-γ production by NK cells ( Wolf et al . , 2017 ) . In addition , P . f . infection activates IL-18 secretion by macrophages . Through IL-18 and direct contact with macrophages NK cells are activated to produce IFN-γ ( Baratin et al . , 2005; Wolf et al . , 2017 ) . In contrast , the effector functions of NK cells we describe here are independent of external signals , since unstimulated primary NK cells respond directly to activation by multivalent IgG Fc binding to FcγRIIIa ( Bryceson et al . , 2005 ) . Experiments performed here used NK cells freshly isolated from human blood , without stimulation prior to incubation with RBCs and Abs . In summary , we have shown that human NK cells have the potential to control P . f . parasitemia through IgG-dependent activation of NK cellular cytotoxicity , and thus contribute to protection from blood-stage malaria .
This study was approved by the Ethics Committee of the Faculty of Medicine , Pharmacy , and Dentistry at the University of Sciences , Techniques , and Technologies of Bamako , in Mali and by the Institutional Review Board of the National Institute of Allergy and Infectious Diseases , National Institutes of Health . Prior to inclusion in this study , written informed consent was received from participants . Plasma samples were collected from adults enrolled in a multi-year malaria study in the rural village Kambila ( Crompton et al . , 2008 ) , by starting with venous blood collected in citrate-containing cell-preparation tubes ( BD , Franklin Lakes , NJ ) . Samples were transported 45 km away to the Malaria Research and Training Centre in Bamako , where peripheral blood mononuclear cells ( PBMCs ) and plasma were isolated . Plasma was frozen in 1 ml aliquots at −80°C . Samples were shipped to the United States on dry ice for analysis . Control serum from US individuals was obtained from Valley Biomedical ( Winchester , VA ) . IgG was purified from plasma or serum by standard affinity chromatography . Briefly , each sample was diluted 1:5 in column equilibration-wash buffer ( 10 mM NaPO4 , 150 mM NaCl , pH 7 . 0 ) . The IgG fraction was isolated on Protein G columns ( GE Healthcare , Amersham-Pharmacia-HiTrap Protein G ) and eluted with 100 mM glycine , pH 2 . 5 . and immediately neutralized to pH 7 . 4 with 4 . 0 M Tris pH 8 . 0 . IgG was concentrated and dialyzed in Pall Macrosep columns ( 30 kDa MW cutoff ) with PBS . Human blood samples from deidentified healthy US donors were drawn for research purposes at the NIH Blood Bank under an NIH IRB approved protocol with informed consent . PBMCs were first isolated using Lymphocyte Separation Medium ( MP Biomedicals , Solon , OH ) , washed with PBS twice , and resuspended in PBS , with 2% FBS and 1 mM EDTA . NK cells were isolated from PBMCs by depletion of non-NK cells using an NK cell isolation kit ( STEMCELL Technologies , Cambridge , MA ) . The manufacturer’s protocol was modified as follows . PBMCs were resuspended at 2 × 108 , per ml and 2 . 5 μl/ml of anti-CD3 biotin ( STEMCELL Technologies ) was added to the 50 μl/ml of the non-NK cocktail to increase NK cell purity . Resting NK cells were resuspended in Iscove's modified Dulbecco’s medium ( IMDM; Invitrogen , Carlsbad , CA ) supplemented with 10% human serum ( Valley Biomedical , Winchester , VA ) , and were used within 1 to 4 days after isolation . Only NK cell preparations that had greater than 95% CD14neg , CD3neg , CD56pos , as determined by flow cytometry , were used in experiments . 3D7 and FCR3 VAR2CSA strains were cultivated at 37°C under 5% O2 , 5% CO2 , 90% N2 at 37°C at <5% hematocrit using O+ human erythrocytes ( Interstate Blood Bank , Inc . ) . The P . f . DC-J strain was cultivated similarly in the presence of Blasticidin ( 2 . 5 μg/ml ) . Parasites were cultured in complete medium , which was RPMI 1640 buffered with 25 mM HEPES and supplemented with 2 . 5% heat-inactivated human AB Serum , 0 . 5% Albumax-II , 28 mM sodium bicarbonate , 25 μg/ml gentamycin , and 50 μg/ml hypoxanthine . Parasite development was monitored by light microscopy using methanol-fixed , Giemsa-stained thin blood films . Parasites were synchronized using sorbitol ( Lambros and Vanderberg , 1979 ) . Parasite-iRBCs were enriched for knobs using Zeptogel ( contains gelatin ) sedimentation routinely . Infected RBCs used in ADCC assays were enriched at the trophozoite stage with percoll-sorbitol gradient and centrifugation ( Aley et al . , 1984; Hill et al . , 2007 ) , washed , and resuspended in complete medium in the absence of human serum . Cultures growing in Albumax-II therefore had no antibodies or complement components . iRBCs , enriched at the trophozoite stage , were resuspended in PBS and 2% FBS , and incubated at 4°C for 30 min with plasma , serum , or purified antibodies at specified dilutions . Cells were washed and incubated at 4°C with appropriate and fluorescently-tagged secondary Abs for an additional 20 min . Cells were washed , and flow cytometry was performed on a FACS LSR-II or a FACS Calibur ( BD Biosciences ) , and data analyzed with FlowJo ( FlowJo , LLC ) . For immunofluorescence analysis , iRBCs , enriched at the trophozoite stage , were mixed with uRBCs at a ratio of 1:1 and incubated with US serum or Mali plasma for 30 min at 4°C . Cells were washed and incubated with fluorescently –tagged secondary Abs for an additional 20 min . Cells were first washed , and then fixed with 1% paraformaldehyde ( PFA ) at room temperature . DAPI was used to visualize the P . f . DNA in iRBCs . Immunofluorescence images were obtained on a LSM 780 confocal laser microscope ( Carl Zeiss , Oberkochen , Germany ) . Images were acquired using the Zen software ( Carl Zeiss ) . uRBCs and iRBCs were washed with 0 . 1 M phosphate buffer ( pH 7 . 4 ) . RBCs were centrifuged at 2500 rpm for 5 min and supernatants were removed . RBCs were again washed with 0 . 1 M phosphate buffer ( pH 7 . 4 ) . RBCs were resuspended in 3 ml of fixative solution ( 3% PFA +0 . 1% glutaraldehyde ) . The cells were stored at 4°C until further processing for imaging . Fixed cells were allowed to settle on silicon chips for 1 hr . Subsequent post-fixation with 1% OsO4 was performed with microwave irradiation ( Pelco 3451 microwave processor , in cycles of 2 min on , 2 min off , 2 min on at 250 W under 15 in . Hg vacuum; Ted Pella , Redding , CA ) . Specimens were dehydrated in a graded ethanol series for 1 min under vacuum . Samples were then dried to a critical point in a Bal-Tec cpd 030 drier ( Balzer , Bal-Tec AG , Balzers , Liechtenstein ) . Cells were then coated with 75 Å of iridium in an IBS ion beam sputter ( South Bay Technology , Inc . , San Clemente , CA . ) Samples were imaged on a Hitachi SU-8000 SEM ( Hitachi , Pleasantown , CA ) . Resting NK cells ( 2 × 105 ) either alone or mixed with enriched iRBCs ( 2 × 105 ) were added to wells of a 96-well plate , in the presence of antibodies ( either US serum diluted 1:10 , Mali plasma diluted 1:10 , or rabbit anti-human RBC antibody at 1 . 25 μg/ml ) . Anti-CD107a Ab–PE diluted 1:20 ( Clone H4A3 , Cat#555801 , BD Biosciences , Franklin Lakes , NJ ) was added at the beginning of the incubation with NK cells . Cells were centrifuged for 3 min at 100 g and incubated for 4 hr at 37°C . Cells were centrifuged , resuspended in PBS containing 2% FBS , and stained with conjugated anti-CD56–PC5 . 5 Ab ( Clone N901 , Beckman Coulter , Brea , CA ) , near-IR fixable Live/Dead dye ( Invitrogen ) , and CD235a–FITC ( Clone H1264 , Biolegend , San Diego , CA ) . Samples were analyzed on a FACS-LSRII flow cytometer ( BD Biosciences ) and data analyzed with FlowJo software ( FlowJo , LLC ) . For intracellular cytokine assays , NK cells were incubated with iRBCs as described above in the presence of Brefeldin A ( 5 μg/ml ) during the 4 hr incubation . Cells were then stained with anti-CD56–PC5 . 5 Ab ( Clone N901 , Beckman Coulter , Brea , CA ) , near-IR fixable Live/Dead dye ( Invitrogen ) , CD235a–FITC ( Clone H1264 , Biolegend , San Diego , CA ) , followed by fixation and permeabilization using the BD Cytofix/Cytoperm Kit ( BD Biosciences ) . IFN-γ was detected using anti-IFN-γ–APC Ab ( Clone B27 , BD Biosciences ) and TNF-α was detected using anti-TNF-α–PE Ab ( Clone 6401 . 1111 , BD Bioscience ) . Samples were resuspended in PBS and analyzed on a FACS LSRII flow cytometer ( BD Biosciences ) . Data analysis was performed with FlowJo software ( FlowJo , LLC ) . NK cells , uRBCs and iRBCs were washed twice with PBS before labeling with different dyes . iRBCs were stained with cell proliferation dye eFluor 670 at 5 μM for 5 min in PBS at 37°C . Similarly , uRBCs were stained with eFluor 450 dye at 2 . 5 μM concentration for 5 min in PBS at 37°C . NK cells were washed , suspended in diluent C and stained with 1 μM PKH67 membrane dye ( PKH67 green fluorescent green linker kit , Sigma-Aldrich ) for 5 min at 37°C . Cells were then washed three times with media containing serum ( e . g . , RPMI with 10% FBS ) . For imaging , cells were resuspended in RPMI 1640 containing 0 . 5% Albumax-II in the absence of Phenol Red . Cells were added in 8-well Lab-Tek I Chambered cover glass ( Nunc ) and allowed to settle for 15 min . Imaging was performed with a Zeiss LSM 780 confocal microscope while maintaining incubation condition at 37°C , 5% CO2 , in a humidified chamber . Images were acquired at 30 s interval for 6 hr . Time-lapse image stacks were imported into the Imaris software . A threshold algorithm eliminated background noise from each channel , and a Gaussian filter was applied to smooth the texture , and to easily segment the regions of interest ( ROIs ) . After filtering , a surface channel was created from each color channel for each cell population , with surface threshold based on intensity . The surface generator was set to run a watershed algorithm . The seed-points diameter was set to 4 . 5 μm for iRBCs and uRBCs , and 6 . 0 μm for NK cells . In order to weed out unwanted particles that passed the intensity threshold , a surface ROI was considered to be one with voxel size greater than 110 voxels . For the tracking algorithm we used autoregressive motion with maximum step distance set to 5 μm and maximum gap size set to two frames . NK cells were incubated with 20 × 103 trophozoite-stage iRBCs at NK to iRBC ratios of 1:1 and 3:1 in the presence of 20 × 105 uRBCs in 96-well plates for 48 hr at 37°C , in complete medium . Thin blood smears were fixed in 100% methanol , stained with 5% Giemsa solution and counted under light microscope . 25 microscope fields , each containing 200 RBCs , were counted . Parasitemia was expressed as [ ( number of iRBCs ÷ total number of RBCs ) ×100] . 2 . 5 × 105 NK cells and 5 × 104 FCR3 VAR2CSA–iRBCs were mixed in 96-well plates and incubated for 6 hr at 37°C in the absence or presence of purified rabbit IgG antibodies to the DBL3X domain ( 0 . 5 mg/ml ) , or purified human IgG antibodies to the DBL domains of PfEMP1 VAR2CSA ( 0 . 5 mg/ml ) , or control rabbit IgG ( 0 . 5 mg/ml ) in a final volume of 25 μl . A 100-fold excess of uRBCs ( 5 × 106 ) was then added , bringing the final volume to 100 μl . Cultures were then maintained for an additional 42 hr at 37°C in standard parasite growth conditions . At the end of incubation , CD45-PE ( Clone H130 , Biolegend , San Diego , CA ) and CD235a-FITC antibodies , and Hoechst were used to stain NK cells , uRBCs and iRBCs . Samples were acquired on FACSLSR-II , and data analyzed with FlowJo ( FlowJo , LLC ) . Parasitemia was determined as the fraction of RBCs ( CD235a+ ) positive for Hoechst . Samples with NK cells but in the absence of antibodies were used as control to calculate growth inhibition . NK cells and iRBCs were resuspended in experimental media ( no human serum ) . 6 × 105 NK cells and 2 × 105 iRBCs were mixed at a 3:1 ratio in 96-well plates and incubated for 6 hr at 37°C in the absence or presence of antibodies . For experiments using plasma or serum , the total amount of plasma or serum in each condition was ( 20 μl plasma/serum into 200 μl media ) to control for the level of plasma . 20 μl of US serum the negative control , then increasing volume of Mali immune plasma was added in ( Example: 2 μl Mali plasma with 18 μl US serum totaling 20 μl plasma/serum ) . After a 6 hr coincubation of iRBCs and NK cells , soluble Abs were removed by a wash . This washing step removed any antibody that would bind to merozoites . A 100-fold excess of uRBCs ( 2 × 107 ) relative to iRBCs was then added and cultures were maintained for an additional 16 hr at 37°C in standard parasite growth conditions . At the end of incubation , CD45- PE ( BD Biosciences ) , CD235a-FITC , and Hoechst were used to stain NK cells , uRBCs and iRBCs ( Figure 2—figure supplement 1C ) . Samples with NK cells but in the absence of antibodies were used as control to calculate growth inhibition . Enriched iRBCs and NK cells were washed with RPMI 1640 , containing 0 . 5% Albumax in the absence of Phenol red . Cells were mixed at defined ratios in 96-well V bottom plates in 150 μl . Antibodies were added as specified . Antibodies tested in the assay are Rabbit anti-human RBC antibodies ( 1 . 25 μg/ml ) , MGD21 and MGD21-LALA antibodies ( Tan et al . , 2016 ) ( 0 . 2 mg/ml ) , Rabbit VAR2CSA and Control IgG antibodies ( 0 . 5 mg/ml ) . Cells were centrifuged at 100 g for 3 min and incubated at 37°C for 5–6 hr as mentioned . Plates were centrifuged at 2000 rpm for 5 min and 50 μl of supernatant was collected . Hemoglobin was measured with QuantiChrom Hemoglobin Assay Kit ( BioAssay , Hayward , CA ) . Hemoglobin absorbance was measured at 405 nm using a 96-well plate reader ( Enspire , Perkin Elmer , MA and SpectraMax plus , Molecular Devices , CA ) . In each experiment , maximum hemoglobin release was determined by lysis of iRBCs in 1% Triton-X-100 . At the end of the 5 hr incubation period , the hemoglobin released in supernatant was quantified as percent fraction of maximum hemoglobin release . Hemoglobin released during the 5 hr incubation period in iRBCs sample was subtracted in each experiment to normalize the background in all experiments . Each graph was generated from at least three independent experiments . For normally distributed data , either mean ± SD or mean ± SEM was used , as specified . Statistical analysis was performed using the software Graphpad Prism v7 . Data was analyzed by either two-tailed Student’s t test , or by one-way analysis of variance ( ANOVA ) . | Malaria is a deadly disease caused by a parasite transmitted by mosquitoes . The parasite infects red blood cells , causing fever with flu-like symptoms . In some people , particularly pregnant women and children , the disease may be very serious and even lead to death . An effective malaria vaccine is urgently needed because malaria parasites are developing resistance to current drugs . People living in areas where malaria is common develop specific proteins called antibodies that protect them from malaria . Learning more about how the antibodies achieve this , could help to develop better vaccines . Scientists already know some antibodies bind to the malaria parasites and prevent them from entering red blood cells . Some vaccines have been based on these antibodies . Other antibodies bind to infected cells flagging them for destruction by cells of the immune system . Immune cells called natural killer cells can eliminate viruses or cancer cells this way , but it was not clear if they could also eliminate malaria parasite-infected red blood cells . Now , Arora et al . show that natural killer cells can selectively destroy malaria-infected red blood cells flagged with antibodies from people who live in areas where malaria is common . In laboratory experiments , natural killer cells from US volunteers , who were never exposed to malaria , did not kill normal or malaria-infected red blood cells . Adding antibodies collected from malaria-resistant volunteers from Africa allowed these natural killer cells from unexposed people to selectively seek out and destroy malaria-infected cells and leave uninfected red blood cells intact . Arora et al . also found that the antibodies from the malaria-resistant volunteers bound to parasite proteins on the surface of infected blood cells . The experiments suggest that vaccines designed to stimulate the production of antibodies to malaria proteins that are displayed on infected red blood cells , could destroy the parasite in infected people and help prevent disease and save lives . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"immunology",
"and",
"inflammation"
] | 2018 | NK cells inhibit Plasmodium falciparum growth in red blood cells via antibody-dependent cellular cytotoxicity |
Biological aging is the gradual , progressive decline in system integrity that occurs with advancing chronological age , causing morbidity and disability . Measurements of the pace of aging are needed as surrogate endpoints in trials of therapies designed to prevent disease by slowing biological aging . We report a blood-DNA-methylation measure that is sensitive to variation in pace of biological aging among individuals born the same year . We first modeled change-over-time in 18 biomarkers tracking organ-system integrity across 12 years of follow-up in n = 954 members of the Dunedin Study born in 1972–1973 . Rates of change in each biomarker over ages 26–38 years were composited to form a measure of aging-related decline , termed Pace-of-Aging . Elastic-net regression was used to develop a DNA-methylation predictor of Pace-of-Aging , called DunedinPoAm for Dunedin ( P ) ace ( o ) f ( A ) ging ( m ) ethylation . Validation analysis in cohort studies and the CALERIE trial provide proof-of-principle for DunedinPoAm as a single-time-point measure of a person’s pace of biological aging .
Aging of the global population is producing forecasts of rising burden of disease and disability ( Harper , 2014 ) . Because this burden arises from multiple age-related diseases , treatments for single diseases will not address the burden challenge ( Goldman et al . , 2013 ) . Geroscience research suggests an appealing alternative: treatments to slow aging itself could prevent or delay the multiple diseases that increase with advancing age , perhaps with a single therapeutic approach ( Gladyshev , 2016; Kaeberlein , 2013 ) . Aging can be understood as a gradual and progressive deterioration in biological system integrity ( Kirkwood , 2005 ) . This deterioration is thought to arise from an accumulation of cellular-level changes . These changes , in turn , increase vulnerability to diseases affecting many different organ systems ( Kennedy et al . , 2014; López-Otín et al . , 2013 ) . Animal studies suggest treatments that slow the accumulation of cellular-level changes can extend healthy lifespan ( Campisi et al . , 2019; Kaeberlein et al . , 2015 ) . However , human trials of these treatments are challenging because humans live much longer than model animals , making it time-consuming and costly to follow up human trial participants to test treatment effects on healthy lifespan . This challenge will be exacerbated in trials that will give treatments to young or middle-aged adults , with the aim to prevent the decline in system integrity that antedates disease onset by years . Involving young and midlife adults in healthspan-extension trials has been approved for development by the National Advisory Council on Aging ( 2019 CTAP report to NACA ) . In midlife trials of treatments to slow aging , called geroprotectors ( Moskalev et al . , 2016 ) , traditional endpoints such as disease diagnosis or death are too far in the future to serve as outcomes . Translation of geroprotector treatments to humans could be aided by measures that quantify the pace of deterioration in biological system integrity in human aging . Such measures could be used as surrogate endpoints for healthy lifespan extension ( Justice et al . , 2016; Justice et al . , 2018; Moskalev et al . , 2016 ) , even with young-to-midlife adult trial participants . A useful measure should be non-invasive , inexpensive , reliable , and highly sensitive to biological change . Recent efforts to develop such measures have focused on blood DNA methylation as a biological substrate highly sensitive to changes in chronological age ( Fahy et al . , 2019; Horvath and Raj , 2018 ) . Methylation-clock algorithms have been developed to identify methylation patterns that characterize individuals of different chronological ages . However , a limitation is that individuals born in different years have grown up under different historical conditions ( Schaie , 1967 ) . For example , people born 70 years ago experienced more exposure to childhood diseases , tobacco smoke , airborne lead , and less exposure to antibiotics and other medications , and lower quality nutrition , all of which leave signatures on DNA methylation ( Bell et al . , 2019 ) . As a result , the clocks confound methylation patterns arising from early-life exposures to methylation-altering factors with methylation patterns related to biological aging during adulthood . An alternative approach is to study individuals who were all born the same year , and find methylation patterns that differentiate those who have been aging biologically faster or slower than their same-age peers . The current article reports four steps in our work toward developing a blood DNA methylation measure to represent individual variation in the pace of biological aging . In Step 1 , which we previously reported ( Belsky et al . , 2015 ) , we collected a panel of 18 blood-chemistry and organ-system-function biomarkers at three successive waves of the Dunedin Study , which follows a 1972–73 population-representative one-year birth cohort ( N = 1037 ) . We used repeated-measures data collected when Study members were aged 26 , 32 , and 38 years old to quantify rates of biological change . We modelled the rate of change in each biomarker and calculated how each Study member’s personal rate-of-change on that biomarker differed from the cohort norm . We then combined the 18 personal rates of change across the panel of biomarkers to compute a composite for each Study member that we called the Pace of Aging . Pace of Aging represents a personal rate of multi-organ-system decline over a dozen years . Pace of Aging was normally distributed , and showed marked variation among Study members who were all the same chronological age , confirming that individual differences in biological aging do emerge already by age 38 , years before chronic disease onset . In Step 2 , which we previously reported , we validated the Pace of Aging against known criteria . As compared to other Study members who were the same chronological age but had slower Pace of Aging , Study members with faster Pace of Aging performed more poorly on tests of physical function; showed signs of cognitive decline on a panel of dementia-relevant neuropsychological tests from an early-life baseline; were rated as looking older based on facial photographs; and reported themselves to be in worse health ( Belsky et al . , 2015 ) . Subsequently , we reported that faster Pace of Aging is associated with early-life factors important for aging: familial longevity , low childhood social class , and adverse childhood experiences ( Belsky et al . , 2017 ) , and that faster Pace of Aging is associated with older scores on Brain Age , a machine-learning-derived measure of structural MRI differences characteristic of different age groups ( Elliott et al . , 2019 ) . Notably , Pace of Aging was not well-correlated with published epigenetic age clocks , which were designed to measure how old a person is rather than how fast they are aging biologically ( Belsky et al . , 2018b ) . In Step 3 , which we report here , we distill the Pace of Aging into a measurement that can be obtained from a single blood sample . Here we focused on blood DNA methylation as an accessible molecular measurement that is sensitive to changes in physiology occurring in multiple organ systems ( Birney et al . , 2016; Bolund et al . , 2017; Chambers et al . , 2015; Chu et al . , 2017; Hedman et al . , 2017; Ma et al . , 2019; Mill and Heijmans , 2013; Morris et al . , 2017; Wahl et al . , 2017 ) . We used data about the Pace of Aging from age 26 to 38 years in the Dunedin Study along with whole-genome methylation data at age 38 years . Elastic-net regression was applied to derive an algorithm that captured DNA methylation patterns linked with variation among individuals in their Pace of Aging . The algorithm is hereafter termed ‘DunedinPoAm’ . DunedinPoAm is qualitatively different from previously published DNA methylation measures of aging that were developed by comparing older individuals to younger ones . Those measures , often referred to as ‘clocks , ’ are state measures . They estimate how much aging has occurred in an individual up to the point of measurement . DunedinPoAm is a rate measure . It is based on comparison of longitudinal change over time in 18 biomarkers of organ-system integrity among individuals who are all the same chronological age . DunedinPoAm estimates how fast aging is occurring during the years leading up to the time of measurement . Rather than a clock that records how much time has passed , DunedinPoAm is designed to function as a speedometer , recording how fast the subject is aging . In Step 4 , which we report here , we validated the DunedinPoAm in five ways . First , using the Dunedin Study , we tested if Study member’s DunedinPoAm measured when they were aged 38 years could predict deficits in physical and cognitive functioning seven years later , when the cohort was aged 45 years . Second , we applied the DunedinPoAm algorithm to DNA methylation data from a second , cross-sectional , study of adults to evaluate patterning of DunedinPoAm by chronological age and sex and to test correlations of DunedinPoAm with self-reported health and proposed measures of biological age , including three epigenetic clocks . Third , we applied the DunedinPoAm algorithm to DNA methylation data from a third , longitudinal study of older men to test associations with chronic-disease morbidity and mortality . Fourth , we applied the DunedinPoAm algorithm to DNA methylation data from a fourth , longitudinal , study of young people to test if DunedinPoAm was accelerated by exposure to poverty and victimization , factors which are known to shorten healthy lifespan . Finally , to ascertain the potential usefulness of DunedinPoAm as a measure for trials of geroprotector treatments , we applied the algorithm to DNA methylation data from a randomized trial of caloric restriction , CALERIE ( Ravussin et al . , 2015 ) . Earlier we reported from this trial that the intervention ( two years of prescribed 25% caloric restriction ) slowed the rate of biological aging as measured by a blood-chemistry biological-age composite measure ( Belsky et al . , 2018a ) . Here , using newly generated methylation data from blood drawn at the CALERIE baseline assessment , we tested if ( a ) DunedinPoAm from blood drawn before caloric restriction could predict the future rate of biological aging of participants during the two-year trial , and ( b ) if this prediction was disrupted in participants who underwent caloric restriction , but not among control participants . We report promising results from this four-step research program , while appreciating that additional measurement development will be needed to support applied use of DunedinPoAm . A graphical illustration of our study design is presented in Figure 1 .
We derived the DunedinPoAm algorithm using data from Dunedin Study members for whom age-38 DNA methylation data were available ( N = 810 ) . We applied elastic-net regression ( Zou and Hastie , 2005 ) using Pace of Aging between ages 26 to 38 years as the criterion . We included all methylation probes that appear on both the Illumina 450 k and EPIC arrays as potential predictor variables . We selected this overlapping set of probes for our analysis to facilitate application of the derived algorithm by other research teams using either array . We fixed the alpha parameter to 0 . 5 , following the approach reported by Horvath ( 2013 ) . This analysis selected a set of 46 CpG sites ( Supplementary file 1A ) . The 46-CpG elastic-net-derived DunedinPoAm algorithm , applied in the age-38 Dunedin DNA methylation data , was associated with the longitudinal 26–38 Pace of Aging measure ( Pearson r = 0 . 56 , Figure 1—figure supplement 1 ) . This is likely an overestimate of the true out-of-sample correlation because the analysis is based on the same data used to develop the DunedinPoAm algorithm; bootstrap-cross-validation analysis estimated the out-of-sample correlation to be r = 0 . 33 ( Figure 1—figure supplement 2 ) . To test variation in DunedinPoAm and to compare it with published methylation measures of biological aging , we conducted analysis using data on N = 1175 participants aged 28–95 years ( M = 58 , SD = 15; 42% male ) in the UK Understanding Society Study . In this mixed-age sample , the mean DunedinPoAm was 1 . 03 years of biological aging per each calendar year ( SD = 0 . 07 ) . We first tested if higher DunedinPoAm levels , which indicate faster aging , were correlated with older chronological age . Mortality rates increase with advancing chronological age , although there may be some slowing at the oldest ages ( Barbi et al . , 2018 ) . This suggests the hypothesis that the rate of aging increases across much of the adult lifespan . Consistent with this hypothesis , Understanding Society participants who were of older chronological age tended to have faster DunedinPoAm ( r = 0 . 11 , [0 . 06–0 . 17] , p<0 . 001; Figure 3 Panel A ) . We also compared DunedinPoAm with three methylation measures of biological age: the epigenetic clocks proposed by Horvath , Hannum , and Levine ( Hannum et al . , 2013; Horvath , 2013; Levine et al . , 2018 ) . These epigenetic clocks were highly correlated with chronological age in the Understanding Society sample ( Horvath Clock r = 0 . 91 , Hannum Clock r = 0 . 92 , Levine Clock r = 0 . 88 ) . Next , to test if DunedinPoAm captured similar information about aging to published epigenetic clocks , we regressed each of the published clocks on chronological age and predicted residual values , following the procedure used by the developers of the clocks . These residuals are referred to in the literature as measures of ‘epigenetic age acceleration . ’ None of the 46 CpGs included in the DunedinPoAm algorithm overlapped with CpGs in these epigenetic clocks . Nevertheless , DunedinPoAm was moderately correlated with epigenetic age acceleration measured from the clocks proposed by Hannum ( r = 0 . 24 ) and Levine ( r = 0 . 30 ) . DunedinPoAm was less-well correlated with acceleration measured from the Horvath clock ( r = 0 . 06 ) . Associations among DunedinPoAm and the epigenetic clocks in the Understanding Society sample are shown in Figure 3 Panel B . Finally , we tested correlations of DunedinPoAm with ( a ) a measure of biological age derived from blood chemistry and blood pressure data , and ( b ) a measure of self-rated health . We computed biological age from Understanding Society blood chemistry and blood pressure data following the Klemera and Doubal method ( KDM ) ( Klemera and Doubal , 2006 ) and the procedure described by Levine ( 2013 ) . KDM Biological Age details are reported in the Materials and Methods . Participants with faster DunedinPoAm had more advanced KDM Biological Age ( r = 0 . 20 95% CI [0 . 15–0 . 26] , p<0 . 001; ) and worse self-rated health ( r = −0 . 22 [-0 . 28 , –0 . 16] , p<0 . 001; ) . Covariate adjustment to models for estimated cell counts ( Houseman et al . , 2012 ) and smoking status did not change results . Results for all models are reported in Supplementary file 1D . In comparison to DunedinPoAm , effect-sizes for associations with self-rated health and KDM Biological Age were smaller for the epigenetic clocks and , in the cases of the Horvath and Hannum clocks , were not statistically different from zero at the alpha = 0 . 05 threshold ( Figure 3 Panels C and D ) . Effect-sizes are reported in Supplementary file 1D and plotted in Figure 3—figure supplement 1 . To test if faster DunedinPoAm was associated with morbidity and mortality , we analyzed data from N = 771 older men in the Veterans Health Administration Normative Aging Study ( NAS; at baseline , mean chronological age = 77 , SD = 7 ) . We first tested if higher DunedinPoAm levels , which indicate faster aging , were associated with increased risk of mortality . During follow-up from 1999 to 2013 , 46% of NAS participants died over a mean follow-up of 7 years ( SD = 7 ) . Those with faster DunedinPoAm at baseline were at increased risk of death ( Hazard Ratio ( HR ) = 1 . 29 [1 . 16–1 . 45] , p<0 . 001; Figure 4 ) . We next tested if NAS participants with faster DunedinPoAm experienced higher levels of chronic disease morbidity , measured as the count of diagnosed diseases ( hypertension , type-2 diabetes , cardiovascular disease , chronic obstructive pulmonary disease , chronic kidney disease , and cancer ) . During follow-up across 4 assessments during 1999–2013 ( n = 1448 observations of the N = 771 participants ) , n = 175 NAS participants were diagnosed with a new chronic disease . Those with faster baseline DunedinPoAm were at increased risk of new diagnosis ( HR = 1 . 19 [1 . 03–1 . 38] , p<0 . 019 ) . In repeated-measures analysis of prevalent chronic disease , faster DunedinPoAm was associated with having a higher level of chronic disease morbidity ( IRR = 1 . 15 [1 . 11–1 . 20] , p<0 . 001 ) . Finally , we utilized the repeated-measures data to test if NAS participants’ DunedinPoAms increased as they aged . We tested within-person change in DunedinPoAm over time ( n = 1253 observations of N = 536 participants with 2–4 timepoints of DNA methylation data ) . Consistent with Understanding Society analysis showing faster DunedinPoAm in older as compared to younger adults , NAS participants’ DunedinPoAm values increased across repeated assessments . For every five years of follow-up , participants’ DunedinPoAms increased by 0 . 012 ( SE = 0 . 002 , p<0 . 001 ) units , or about 0 . 2 standard deviations . Covariate adjustment to models for estimated cell counts ( Houseman et al . , 2012 ) and smoking status did not change results , with the exception that the effect-size for DunedinPoAm was attenuated below the alpha = 0 . 05 threshold of statistical significance in smoking-adjusted analysis of chronic disease incidence . Results for all models are reported in Supplementary file 1E . In comparison to DunedinPoAm , effect-sizes for associations with mortality and chronic disease were smaller for the epigenetic clocks and were not statistically different from zero in many of the models ( Supplementary file 1E and Figure 4—figure supplement 1 ) . To test if DunedinPoAm indicated faster aging in young people with histories of exposure thought to shorten healthy lifespan , we analyzed data from N = 1658 members of the E-Risk Longitudinal Study . The E-Risk Study follows a 1994–95 birth cohort of same-sex twins . Blood DNA methylation data were collected when participants were aged 18 years . We analyzed two exposures associated with shorter healthy lifespan , childhood low socioeconomic status and childhood victimization . Socioeconomic status was measured from data on parents’ education , occupation , and income ( Trzesniewski et al . , 2006 ) . Victimization was measured from exposure dossiers compiled from interviews with the children’s mothers and home-visit assessments conducted when the children were aged 5 , 7 , 10 , and 12 ( Fisher et al . , 2015 ) . The dossiers recorded children’s exposure to domestic violence , peer bullying , physical and sexual harm by an adult , and neglect . 72% of the analysis sample had no victimization exposure , 21% had one type of victimization exposure , 4% had two types of exposure , and 2% had three or more types of exposure . E-Risk adolescents who grew up in lower socioeconomic-status families exhibited faster DunedinPoAm ( Cohen’s d for comparison of low to moderate SES = 0 . 21 [0 . 06–0 . 35]; Cohen’s d for comparison of low to high SES = 0 . 44 [0 . 31–0 . 56]; Pearson r = 0 . 19 [0 . 13–0 . 24] ) . In parallel , E-Risk adolescents with exposure to more types of victimization exhibited faster DunedinPoAm ( Cohen’s d for comparison of never victimized to one type of victimization = 0 . 28 [0 . 15–0 . 41]; Cohen’s d for comparison of never victimized to two types of victimization = 0 . 48 [0 . 23–0 . 72]; Cohen’s d for comparison of never victimized to three or more types of victimization = 0 . 53 [0 . 25–0 . 81]; Pearson r = 0 . 15 [0 . 10–0 . 20] ) . Covariate adjustment to models for estimated cell counts ( Houseman et al . , 2012 ) did not change results . Adjustment for smoking status attenuated effect-sizes by about half , but most associations remained statistically different from zero at the alpha = 0 . 05 level . Results for all models are reported in Supplementary file 1F . Differences in DunedinPoAm across strata of childhood socioeconomic status and victimization are graphed in Figure 5 . In comparison to DunedinPoAm , effect-sizes for associations with childhood socioeconomic circumstances and victimization were smaller for the epigenetic clocks and , in the cases of the Horvath and Hannum clocks , were not statistically different from zero at the alpha = 0 . 05 threshold . Effect-sizes are reported in Supplementary file 1F and plotted in Figure 5—figure supplement 1 . The CALERIE Trial is the first randomized trial of long-term caloric restriction in non-obese adult humans . CALERIE randomized N = 220 adults on a 2:1 ratio to treatment of 25% caloric restriction ( CR-treatment ) or control ad-libitum ( AL-control , as usual ) diet for two years ( Ravussin et al . , 2015 ) . We previously reported that CALERIE participants who were randomized to CR-treatment experienced a slower rate of biological aging as compared to participants in the AL-control arm based on longitudinal change analysis of clinical-biomarker data from the baseline , 12 month , and 24 month follow-up assessments ( Belsky et al . , 2018a ) . Among control participants , the rate of increase in biological age measured using the Klemera-Doubal method ( KDM ) Biological Age algorithm was 0 . 71 years of biological age per 12 month follow-up interval . ( This slower-than-expected rate of aging could reflect differences between CALERIE Trial participants , who were selected for being in good health , and the nationally representative NHANES sample in which the KDM algorithm was developed [Belsky et al . , 2018a] . ) In contrast , among treatment participants , the rate of increase was only 0 . 11 years of biological age per 12 month follow-up interval ( difference b = −0 . 60 [-0 . 99 , –0 . 21] ) . We subsequently generated DNA methylation data from blood DNA that was collected at the baseline assessment of the CALERIE trial for a sub-sample ( N = 68 AL-control participants and 118 CR-treatment participants ) . We used these methylation data to calculate participants’ DunedinPoAm values at study baseline . We then tested if baseline DunedinPoAm could predict participants’ future rate of biological aging as they progressed through the trial . We first replicated our original analysis within the methylation sub-sample . Results were the same as in the full sample ( Supplementary file 1G ) . Next , we compared DunedinPoAm between CR-treatment and AL-control participants . As expected , there was no group difference at baseline ( AL M = 1 . 00 , SD = 0 . 05; CR M = 1 . 01 , SD = 0 . 06 , p-value for difference = 0 . 440 ) . Finally , we tested if participants’ baseline DunedinPoAm was associated with their rate of biological aging over the 24 months of follow-up , and if this association was modified by randomization to caloric restriction as compared to ad libitum diet . For AL-control participants , faster baseline DunedinPoAm predicted faster biological aging over the 24 month follow-up , although in this small group this association was not statistically significant at the alpha = 0 . 05 level ( b = 0 . 22 [-0 . 05 , 0 . 49] , p=0 . 104 ) . For CR-treatment participants , the association of baseline DunedinPoAm with future rate of aging was sharply reduced , ( b = −0 . 08 [-0 . 24 , 0 . 09] , p=0 . 351 ) , although the difference between the rate of aging in the AL-control and CR-treatment groups did not reach the alpha = 0 . 05 threshold for statistical significance ( interaction-term testing difference in slopes b = −0 . 30 [-0 . 61 , 0 . 01] , p-value=0 . 060 ) . Slopes of change in KDM Biological Age for participants in the AL-control and CR-treatment groups are plotted for fast baseline DunedinPoAm ( 1 SD above the mean ) and slow baseline DunedinPoAm ( 1 SD below the mean ) in Figure 6 . CALERIE DNA methylation data are not yet available to test if the intervention altered post-treatment DunedinPoAm .
Breakthrough discoveries in the new field of geroscience suggest opportunities to extend healthy lifespan through interventions that slow biological processes of aging ( Campisi et al . , 2019 ) . To advance translation of these interventions , measures are needed that can detect changes in a person’s rate of biological aging ( Moffitt et al . , 2017 ) . We previously showed that the rate of biological aging can be measured by tracking change over time in multiple indicators of organ-system integrity ( Belsky et al . , 2015 ) . Here , we report data illustrating the potential to streamline measurement of Pace of Aging to an exportable , inexpensive and non-invasive blood test , and thereby ease implementation of Pace of Aging measurement in studies of interventions to slow processes of biological aging . We conducted machine-learning analysis of the original Pace of Aging measure using elastic-net regression and whole-genome blood DNA methylation data . We trained the algorithm to predict how fast a person was aging . We called the resulting algorithm ‘DunedinPoAm’ for ‘ ( P ) ace ( o ) f ( A ) ging ( m ) ethylation’ . There were four overall findings: First , while DunedinPoAm was not a perfect proxy of Pace of Aging , it nevertheless captured critical information about Dunedin Study members’ healthspan-related characteristics . Across the domains of physical function , cognitive function , and subjective signs of aging , Study members with faster DunedinPoAm at age 38 were worse off seven years later at age 45 and , in repeated-measures analysis of change , they showed signs of more rapid decline . Effect-sizes were equal to or greater than those for the 18-biomarker 3-time point measure of Pace of Aging . This result suggests that the DNA-methylation elastic-net regression used to develop DunedinPoAm may have distilled the aging signal from the original Pace of Aging measure and excluded some noise . In sum , DunedinPoAm showed promise as an easy-to-implement alternative to Pace of Aging . Emerging technologies for deep-learning analysis ( Zhavoronkov et al . , 2019 ) may improve methylation measurement of Pace of Aging . Alternatively , integration of methylation data with additional molecular datasets ( Hasin et al . , 2017; Zierer et al . , 2015 ) may be needed to achieve precise measurement of Pace of Aging from a single time-point blood sample . Second , DunedinPoAm analysis of the Understanding Society and NAS samples provided proof-of-concept for using DunedinPoAm to quantify biological aging . Age differences in DunedinPoAm parallel population demographic patterns of mortality risk . In the Understanding Society sample , older adults had faster DunedinPoAm as compared to younger ones . In the NAS sample , participants’ DunedinPoAm values increased as they aged . These observations are consistent with the well-documented acceleration of mortality risk with advancing chronological age ( Robine , 2011 ) . However , it sets DunedinPoAm apart from other indices of biological aging , which are not known to register this acceleration ( Finch and Crimmins , 2016; Li et al . , 2020 ) . DunedinPoAm may therefore provide a novel tool for testing how the rate of aging changes across the life course and whether , as demographic data documenting so-called ‘mortality plateaus’ suggest , processes of aging slow down at the oldest chronological ages ( Barbi et al . , 2018 ) . DunedinPoAm is related to but distinct from alternative approaches to quantification of biological aging . DunedinPoAm was moderately correlated with aging rates measured by the epigenetic clocks proposed by Hannum et al . ( 2013 ) ; Levine et al . ( 2018 ) as well as KDM Biological Age derived from clinical biomarker data ( Klemera and Doubal , 2006; Levine , 2013 ) , and with self-rated health . Consistent with findings for the measured Pace of Aging ( Belsky et al . , 2018b ) , DunedinPoAm was only weakly correlated with the multi-tissue clock proposed by Horvath . DunedinPoAm was more strongly correlated with a clinical-biomarker measure of biological age , with self-rated health , with functional test-performance and decline , and with morbidity and mortality as compared to the epigenetic clocks . Third , DunedinPoAm is already variable by young adulthood and is accelerated in young people at risk for eventual shortened healthspan . E-Risk young adults who grew up in socioeconomically disadvantaged families or who were exposed to victimization early in life already showed accelerated DunedinPoAm by age 18 , consistent with epidemiological observations of shorter healthy lifespan for individuals with these exposures ( Adler and Rehkopf , 2008; Danese and McEwen , 2012 ) . We previously found that Dunedin Study members with histories of early-life adversity showed accelerated Pace of Aging in their 30 s ( Belsky et al . , 2017 ) . DunedinPoAm analysis of the E-Risk cohort suggests effects may be already manifest at least a decade earlier . DunedinPoAm may therefore provide a useful index that can be applied to evaluate prevention programs to buffer at-risk youth against health damaging effects of challenging circumstances . Fourth , DunedinPoAm analysis of the CALERIE trial provided proof-of-concept for using DunedinPoAm to quantify biological aging in geroprotector intervention studies . DunedinPoAm measures the rate of aging over the recent past . Control-arm participants’ baseline DunedinPoAm correlated positively with their clinical-biomarker pace of aging over the two years of the trial , consistent with the hypothesis that their rate of aging was not altered . In contrast , there was no relationship between DunedinPoAm and clinical-biomarker pace of aging for caloric-restriction-arm participants , consistent with the hypothesis that caloric restriction altered participants’ rate of aging . Ultimately , data on DunedinPoAm for all CALERIE participants ( and participants in other geroprotector trails ) at trial baseline and follow-up will be needed to establish utility of DunedinPoAm as a surrogate endpoint . In the mean-time , these data establish potential to use DunedinPoAm as a pre-treatment covariate in geroprotector trials to boost statistical power ( Kahan et al . , 2014 ) or to screen participants for enrollment , for example to identify those who are aging more rapidly and may therefore show larger effects of treatment . We acknowledge limitations . Foremost , DunedinPoAm is a first step toward a single-assay cross-sectional measurement of Pace of Aging . The relatively modest size of the Dunedin cohort and the lack of other cohorts that have the requisite three or more waves of repeated biomarkers to measure the Pace of Aging limited sample size for our machine-learning analysis to develop methylation algorithms . As Pace of Aging is measured in additional cohorts , more refined analysis to develop DunedinPoAm-type algorithms will become possible . A related issue is scaling of DunedinPoAm . The original Pace of Aging measure from which DunedinPoAm was developed is denominated in ‘years’ of physiological decline occurring per 12 months of calendar time . Units of DunedinPoAm can , in principle , be interpreted in the same way . But replication in additional cohorts is needed . In addition , our work thus far has not addressed population diversity in biological aging . The Dunedin cohort in which DunedinPoAm was developed and the Understanding Society , NAS , and E-Risk cohorts and CALERIE trial sample in which it was tested were mostly of white European descent . Follow-up of DunedinPoAm in more diverse samples is needed to establish cross-population validity . Finally , because methylation data are not yet available from CALERIE follow-up assessments , we could not test if intervention modified DunedinPoAm at outcome . Ultimately , to establish DunedinPoAm as a surrogate endpoint for healthspan , it will be necessary to establish not only robust association with healthy lifespan phenotypes and modifiability by intervention , but also the extent to which changes in DunedinPoAm induced by intervention correspond to changes in healthy-lifespan phenotypes ( Prentice , 1989 ) . Within the bounds of these limitations , our analysis establishes proof-of-concept for DunedinPoAm as a single-time-point measure that quantifies Pace of Aging from a blood test . It can be implemented in Illumina 450 k and EPIC array data , making it immediately available for testing in a wide range of existing datasets as a complement to existing methylation measures of aging . Critically , DunedinPoAm offers a unique measurement for intervention trials and natural experiment studies investigating how the rate of aging may be changed by behavioral or drug therapy , or by environmental modification . DunedinPoAm may be especially valuable to studies that collect data outside of clinical settings and lack blood chemistry , hematology , and other data needed to measure aging-related changes to physiology .
Data were used from five studies: the Dunedin Study , the Understanding Society Study , the Normative Aging Study ( NAS ) , the Environmental Risk ( E-Risk ) Longitudinal Twin Study , and the CALERIE Trial . The five datasets and measures analyzed within each of them are described below . The Dunedin Study is a longitudinal investigation of health and behavior in a complete birth cohort . Study members ( N = 1 , 037; 91% of eligible births; 52% male ) were all individuals born between April 1972 and March 1973 in Dunedin , New Zealand ( NZ ) , who were eligible based on residence in the province and who participated in the first assessment at age 3 . The cohort represents the full range of socioeconomic status on NZ’s South Island and matches the NZ National Health and Nutrition Survey on key health indicators ( e . g . , BMI , smoking , GP visits ) ( Poulton et al . , 2015 ) . The cohort is primarily white ( 93% ) ( Poulton et al . , 2015 ) . Assessments were carried out at birth and ages 3 , 5 , 7 , 9 , 11 , 13 , 15 , 18 , 21 , 26 , 32 , 38 and , most recently , 45 years , when 94% of the 997 study members still alive took part . At each assessment , each study member is brought to the research unit for a full day of interviews and examinations . Study data may be accessed through agreement with the Study investigators ( https://moffittcaspi . trinity . duke . edu/research-topics/dunedin ) . Dunedin Study measures of physical and cognitive functioning and subjective signs of aging are described in detail in Supplementary file 1H . Understanding Society is an ongoing panel study of the United Kingdom population ( https://www . understandingsociety . ac . uk/ ) . During 2010–12 , participants were invited to take part in a nurse’s exam involving a blood draw . Of the roughly 20 , 000 participants who provided clinical data in this exam , methylation data have been generated for just under 1200 . We analyzed data from 1175 participants with available methylation and blood chemistry data . Documentation of the methylation ( University of Essex , 2012 ) and blood chemistry ( University of Essex , 2017 ) data resource is available online ( https://www . understandingsociety . ac . uk/sites/default/files/downloads/documentation/health/user-guides/7251-UnderstandingSociety-Biomarker-UserGuide-2014 . pdf ) . Klemera-Doubal method ( KDM ) Biological Age . We measured KDM Biological age from blood chemistry , systolic blood pressure , and lung-function data using the algorithm proposed by Klemera and Doubal ( 2006 ) trained in data from the NHANES following the method originally described by Levine ( 2013 ) and using the dataset compiled by Hastings ( Hastings et al . , 2019 ) . We included 8 of Levine’s original 10 biomarkers in the algorithm: albumin , alkaline phosphatase ( log ) , blood urea nitrogen , creatinine ( log ) , C-reactive protein ( log ) , HbA1C , systolic blood pressure , and forced expiratory volume in 1 s ( FEV1 ) . We omitted total cholesterol because of evidence this biomarker shows different directions of association with aging in younger and older adults ( Arbeev et al . , 2016 ) . Cytomegalovirus optical density was not available in the Understanding Society database . Self Rated Health . Understanding Society participants rated their health as excellent , very-good , good , fair , or poor . We standardized this measure to have Mean = 0 , Standard Deviation = 1 for analysis . The Normative Aging Study ( NAS ) is an ongoing longitudinal study on aging established by the US Department of Veterans Affairs in 1963 . Details of the study have been published previously ( Bell et al . , 1972 ) . Briefly , the NAS is a closed cohort of 2280 male veterans from the Greater Boston area enrolled after an initial health screening to determine that they were free of known chronic medical conditions . Participants have been re-evaluated every 3–5 years on a continuous rolling basis using detailed on-site physical examinations and questionnaires . DNA from blood samples was collected from 771 participants beginning in 1999 . We analyzed blood DNA methylation data from up to four repeated assessments conducted through 2013 ( Gao et al . , 2019b; Panni et al . , 2016 ) . Of the 771 participants with DNA methylation data , n = 536 ( 46% ) had data from two repeated assessments and n = 178 ( 23% ) had data from three or four repeated assessments . We restricted the current analysis to participants with at least one DNA methylation data point . The NAS was approved by the Department of Veterans Affairs Boston Healthcare System and written informed consent was obtained from each subject before participation . Mortality . Regular mailings to study participants have been used to acquire vital-status information and official death certificates were obtained from the appropriate state health department to be reviewed by a physician . Participant deaths are routinely updated by the research team and the last available update was on 31 December 2013 . During follow-up , n = 355 ( 46% ) of the 771 NAS participants died . Chronic Disease Morbidity . We measured chronic disease morbidity from participants medical histories and prior diagnoses ( Gao et al . , 2019a; Gao et al . , 2019c; Lepeule et al . , 2018; Nyhan et al . , 2018 ) . We counted the number of chronic diseases to compose an ordinal index with categories of 0 , 1 , 2 , 3 , or 4+ of the following comorbidities: hypertension , type-2 diabetes , cardiovascular disease , chronic obstructive pulmonary disease , chronic kidney disease , and cancer . The Environmental Risk ( E-Risk ) Longitudinal Twin Study tracks the development of a birth cohort of 2 , 232 British participants . The sample was drawn from a larger birth register of twins born in England and Wales in 1994–1995 . Full details about the sample are reported elsewhere ( Moffitt and E-Risk Study Team , 2002 ) . Briefly , the E-Risk sample was constructed in 1999–2000 , when 1116 families ( 93% of those eligible ) with same-sex 5-year-old twins participated in home-visit assessments . This sample comprised 56% monozygotic ( MZ ) and 44% dizygotic ( DZ ) twin pairs; sex was evenly distributed within zygosity ( 49% male ) . Families were recruited to represent the UK population of families with newborns in the 1990 s , on the basis of residential location throughout England and Wales and mother’s age . Teenaged mothers with twins were over-selected to replace high-risk families who were selectively lost to the register through non-response . Older mothers having twins via assisted reproduction were under-selected to avoid an excess of well-educated older mothers . The study sample represents the full range of socioeconomic conditions in the UK , as reflected in the families’ distribution on a neighborhood-level socioeconomic index ( ACORN [A Classification of Residential Neighborhoods] , developed by CACI Inc for commercial use ) : 25 . 6% of E-Risk families lived in ‘wealthy achiever’ neighborhoods compared to 25 . 3% nationwide; 5 . 3% vs . 11 . 6% lived in ‘urban prosperity’ neighborhoods; 29 . 6% vs . 26 . 9% lived in ‘comfortably off’ neighborhoods; 13 . 4% vs . 13 . 9% lived in ‘moderate means’ neighborhoods , and 26 . 1% vs . 20 . 7% lived in ‘hard-pressed’ neighborhoods . E-Risk underrepresents ‘urban prosperity’ neighborhoods because such households are likely to be childless . Home-visits assessments took place when participants were aged 5 , 7 , 10 , 12 and , most recently , 18 years , when 93% of the participants took part . At ages 5 , 7 , 10 , and 12 years , assessments were carried out with participants as well as their mothers ( or primary caretakers ) ; the home visit at age 18 included interviews only with participants . Each twin was assessed by a different interviewer . These data are supplemented by searches of official records and by questionnaires that are mailed , as developmentally appropriate , to teachers , and co-informants nominated by participants themselves . The Joint South London and Maudsley and the Institute of Psychiatry Research Ethics Committee approved each phase of the study . Parents gave informed consent and twins gave assent between 5–12 years and then informed consent at age 18 . Study data may be accessed through agreement with the Study investigators ( https://moffittcaspi . trinity . duke . edu/research-topics/erisk ) . Childhood Socioeconomic Status ( SES ) . Childhood SES was defined through a standardized composite of parental income , education , and occupation ( Trzesniewski et al . , 2006 ) . The three SES indicators were highly correlated ( r = 0 . 57–0 . 67 ) and loaded significantly onto one factor . The population-wide distribution of the resulting factor was divided in tertiles for analyses . Childhood Victimization . As previously described ( Danese et al . , 2017 ) , we assessed exposure to six types of childhood victimization between birth to age 12: exposure to domestic violence between the mother and her partner , frequent bullying by peers , physical and sexual harm by an adult , and neglect . The CALERIE Trial is described in detail elsewhere ( Ravussin et al . , 2015 ) . Briefly , N = 220 normal-weight ( 22 . 0 ≤ BMI < 28 kg/m2 ) participants ( 70% female , 77% white ) aged 21–50 years at baseline were randomized to caloric restriction or ad libitum conditions with a 2:1 ratio ( n = 145 to caloric restriction , n = 75 to ad libitum ) . ‘Ad libitum’ ( normal ) caloric intake was determined from two consecutive 14 day assessments of total daily energy expenditure using doubly labeled water ( Redman et al . , 2014 ) . Average percent caloric restriction over six-month intervals was retrospectively calculated by the intake-balance method with simultaneous measurements of total daily energy expenditure using doubly labeled water and changes in body composition ( Racette et al . , 2012; Wong et al . , 2014 ) . Over the course of the trial , participants in the caloric-restriction arm averaged 12% reduction in caloric intake ( about half the prescribed reduction ) . Participants in the ad libitum condition reduced caloric intake by <2% ( Ravussin et al . , 2015 ) . CALERIE data are available at https://calerie . duke . edu/samples-data-access-and-analysis . Klemera-Doubal method ( KDM ) Biological Age . KDM Biological age was measured according to the procedure described in our previous article ( Belsky et al . , 2018a ) . Briefly , we computed KDM Biological Age from CALERIE blood chemistry and blood pressure data using the algorithm proposed by Klemera and Doubal ( 2006 ) trained in data from the NHANES following the method originally described by Levine ( 2013 ) and NHANES data from years matched to the timing of the CALERIE Trial . We included 8 of Levine’s original 10 biomarkers in the algorithm: albumin , alkaline phosphatase ( log ) , blood urea nitrogen , creatinine ( log ) , C-reactive protein ( log ) , HbA1C , systolic blood pressure , and total cholesterol . Cytomegalovirus optical density and lung function were not measured in CALERIE . We supplemented the algorithm with data on uric acid and white blood cell count . DNA methylation was measured from Illumina 450 k Arrays in the Dunedin Study , NAS , and E-Risk Study and from Illumina EPIC 850 k Arrays in the Understanding Society study and the CALERIE Trial . DNA was derived from whole blood samples in all studies . Dunedin Study blood draws were conducted at the cohort’s age-38 assessment during 2010–12 . Understanding Society blood draws were conducted in 2012 . NAS blood draws were conducted during 1999–2013 . E-Risk blood draws were conducted at the cohort’s age-18 assessment during 2012–13 . CALERIE blood draws were conducted at the trial baseline assessment in 2007 . Dunedin and CALERIE methylation assays were run by the Molecular Genomics Shared Resource at Duke Molecular Physiology Institute , Duke University ( USA ) . Understanding Society and E-Risk assays were run by the Complex Disease Epigenetics Group at the University of Exeter Medical School ( UK ) ( www . epigenomicslab . com ) . NAS methylation assays were run by the Genome Research Core of the University of Illinois at Chicago . Processing protocols for the methylation data from all studies have been described previously ( Dai et al . , 2017; Hannon et al . , 2018; Marzi et al . , 2018; Panni et al . , 2016 ) . ( CALERIE data were processed according to the same protocols used for the Dunedin Study . ) We conducted analysis of Dunedin , Understanding Society , NAS , E-Risk , and CALERIE data using regression models . We analyzed continuous outcome data using linear regression . We analyzed count outcome data using Poisson regression . We analyzed time-to-event outcome data using Cox proportional hazard regression . For analysis of repeated-measures longitudinal DNA methylation data in the NAS , we used generalized estimating equations to account for non-independence of repeated observations of individuals ( Ballinger , 2004 ) , following the method in previous analysis of those data ( Gao et al . , 2018 ) , and econometric fixed-effects regression ( Wooldridge , 2012 ) to test within-person change over time . For analysis in E-Risk , which include data on twin siblings , we clustered standard errors at the family level to account for non-independence of data . For analysis of longitudinal change in clinical-biomarker biological age in CALERIE , we used mixed-effects growth models ( Singer and Willett , 2003 ) following the method in our original analysis of those data ( Belsky et al . , 2018a ) . For regression analysis , methylation measures were adjusted for batch effects by regressing the measure on batch controls and predicting residual values . Dunedin Study , Understanding Society , E-Risk , and CALERIE analyses included covariate adjustment for sex ( the NAS included only men ) . Understanding Society , NAS , and CALERIE analyses included covariate adjustment for chronological age . ( Dunedin and E-Risk are birth-cohort studies and participants are all the same chronological age . ) Sensitivity analyses testing covariate adjustment for estimated leukocyte distributions and smoking are reported in Supplementary file 1C-G . | People’s bodies age at different rates . Age-related biological changes that increase the risk of disease and disability progress rapidly in some people . In others , these processes occur at a slower pace , allowing those individuals to live longer , healthier lives . This observation has led scientists to try to develop therapies that slow aging . The hope is that such treatments could prevent or delay diseases like heart disease or dementia , for which older age is the leading risk factor . Studies in animals have identified treatments that extend the creatures’ lives and slow age-related disease . But testing these treatments in humans is challenging . Our lives are much longer than the worms , flies or mice used in the experiments . Scientists would have to follow human study participants for decades to detect delays in disease onset or an extension of their lives . An alternative approach is to try to develop a test that measures the pace of aging , or essentially “a speedometer for aging” . This would allow scientists to more quickly determine if treatments slow the aging process . Now , Belsky et al . show a blood test designed to measure the pace of aging predicts which people are at increased risk of poor health , chronic disease and an earlier death . First , data about chemical changes to an individual’s DNA , called DNA methylation , were analyzed from white blood cell samples collected from 954 people in a long-term health study known as “The Dunedin Study” . Using the data , Belsky et al . then developed an algorithm – named “DunedinPoAm” – that identified people with an accelerated or slowed pace of aging based on a single blood test . Next , they used the algorithm on samples from participants in three other long-term studies . This verified that those people the algorithm identified as aging faster had a greater risk of poor health , developing chronic diseases or dying earlier . Similarly , those identified as aging more slowly performed better on tests of balance , strength , walking speed and mental ability , and they also looked younger to trained raters . Additionally , Belsky et al . used the test on participants in a randomized trial testing whether restricting calories had potential to extend healthy lifespan . The results suggested that the calorie restriction could counter the effects of an accelerated pace of aging . The test developed by Belsky et al . may provide an alternate way of measuring whether age-slowing treatments work . This would allow faster testing of treatments that can extend the healthy lifespan of humans . The test may also help identify individuals with accelerated aging . This might help public health officials test whether policies or programs can help people lead longer , healthier lives . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2020 | Quantification of the pace of biological aging in humans through a blood test, the DunedinPoAm DNA methylation algorithm |
Processes that define immunoglobulin repertoires are commonly presumed to be the same for all murine B cells . However , studies here that couple high-dimensional FACS sorting with large-scale quantitative IgH deep-sequencing demonstrate that B-1a IgH repertoire differs dramatically from the follicular and marginal zone B cells repertoires and is defined by distinct mechanisms . We track B-1a cells from their early appearance in neonatal spleen to their long-term residence in adult peritoneum and spleen . We show that de novo B-1a IgH rearrangement mainly occurs during the first few weeks of life , after which their repertoire continues to evolve profoundly , including convergent selection of certain V ( D ) J rearrangements encoding specific CDR3 peptides in all adults and progressive introduction of hypermutation and class-switching as animals age . This V ( D ) J selection and AID-mediated diversification operate comparably in germ-free and conventional mice , indicating these unique B-1a repertoire-defining mechanisms are driven by antigens that are not derived from microbiota .
Follicular B ( FOB ) , marginal zone B ( MZB ) and B-1a cells are the major mature B cell populations in the mouse . Although these B cell subsets all produce functionally important antibodies , they differ profoundly in function and developmental origin ( Kantor and Herzenberg , 1993; Hardy and Hayakawa , 2001; Baumgarth , 2011 ) . Previous studies have shown that B-1a cells are efficiently generated during fetal and neonatal life , and are maintained by self-replenishment in adult animals ( Hayakawa et al . , 1985; Montecino-Rodriguez et al . , 2006; Kantor et al . , 1992 ) . In contrast , both FOB and MZB populations emerge later and are replenished throughout life by de novo development from bone marrow ( BM ) hematopoietic stem cells ( HSC ) . Our recent studies show that BM HSC reconstitute FOB and MZB , but fail to reconstitute B-1a cells ( Ghosn et al . , 2012 ) , which are derived from distinct progenitors at embryonic day 9 yolk sac ( Yoshimoto et al . , 2011 ) . For each B cell subset , their antibody responses are enabled by the basic processes that generate the immunoglobulin ( Ig ) structure . Multiple mechanisms contribute to creating the primary Ig heavy ( IgH ) and light chain ( IgL ) diversity . For IgH , these include combinatorial assortment of individual variable ( V ) , diversity ( D ) and joining ( J ) gene segments , nucleotide ( s ) trimming in the D-J and V-DJ joining site , and , template-dependent ( P-addition ) and independent ( N-addition ) nucleotide ( s ) insertion at the joined junctions ( Yancopoulos and Alt , 1986; Kirkham and Schroeder , 1994 ) . The V ( D ) J joining processes define the third IgH complementarity-determining region ( CDR3 ) , which often lies at the center of antigen binding site and plays a crucial role in defining antibody specificity and affinity ( Xu and Davis , 2000 ) . After encountering antigen , “‘naïve”’ B cells are activated and can further diversify their primary antibody repertoire by activation-induced cytidine deaminase ( AID ) –mediated somatic hypermutation ( SHM ) , which introduces single or multiple mutations into the IgV regions ( Muramatsu et al . , 2000; Wagner and Neuberger , 1996 ) . SHM commonly occurs in germinal centers ( GC ) ( Victora and Nussenzweig , 2012 ) , where memory B cells expressing high affinity antibodies are selected ( Rajewsky , 1996; Gitlin et al . , 2014 ) . Since the antigen-driven SHM-mediated secondary Ig diversification is viewed as a crucial adaptation to the environmental needs , the IgH repertoire ( s ) expressed by FOB , MZB and B-1a cells from non-immunized animals are thought to be free of SHM . Our studies here , however , introduce a previously unrecognized SHM mechanism that increasingly diversifies the B-1a pre-immune IgH repertoire as animals age . Importantly , the SHM operates equally in the presence or absence of microbiota influence . The B-1a antibody repertoire is commonly thought to be ‘restricted’ with expressing germline genes , largely because the hybridomas generated from fetal and neonatal B cells , which are mainly B-1a , have few N-insertions ( Carlsson and Holmberg , 1990 ) and preferentially express the proximal 7183 , Q52 VH family genes ( Perlmutter et al . , 1985 ) . The N diversity deficit is ascribed to the absence of expression of terminal deoxynucleotidyl transferase ( Tdt ) , which adds the N nucleotides to the CDR3 junction ( Gilfillan et al . , 1993 ) , during fetal life ( Feeney , 1990 ) . These early studies left the impression that the proximal VH gene usage predominates and that there is little N-addition in the B-1a IgH repertoire . Later studies by the Rajewsky group , however , showed that although neonatal ( 4 day ) splenic B-1a cells contain very few N-insertions , N addition is readily detected in substantial numbers of peritoneal B-1a cells from adult animals ( Gu et al . , 1990 ) , indicating that B-1a cells are continuously generated after Tdt is expressed . Holmberg lab similarly found the low N-region diversity in the adult peritoneal B-1a repertoire ( Tornberg and Holmberg , 1995 ) . Our early studies confirm and extend these findings by showing that roughly two thirds of the IgH sequences from individually sorted peritoneal B-1a cells have N additions ( Kantor et al . 1997 ) . Furthermore , recent studies have shown that B-1a progenitors from both fetal liver and adult BM sources generate peritoneal B-1a cells with substantial N-addition ( Holodick et al . , 2014 ) . Collectively , these findings demonstrate that the peritoneal B-1a IgH repertoire diversity is greater than previously thought . However , these studies mainly characterized the repertories of B cells in the peritoneal cavity ( PerC ) and leave the questions open as to whether and how the repertoire changes throughout ontogeny in B cells at various sites of development and function . Studies here address these issues . We show that the B-1a IgH repertoire differs drastically from the repertories expressed by splenic FOB , MZB and peritoneal B-2 cells . In addition , we track the development of B-1a cells from their early appearance in neonatal spleen to their long-term residence in adult peritoneum and spleen , and elucidate the previous unrecognized somatic mechanisms that select and diversify the B-1a IgH repertoire over time . Most importantly , the potent mechanisms that uniquely act in B-1a ( not in FOB and MZB cells ) operate comparably in germ-free ( GF ) and conventional mice reared under specific pathogen free ( SPF ) condition , indicating that these repertoire-defining mechanisms are not driven by microbiota-derived antigens . The dearth of these advanced understandings in the previous studies is largely due to technical difficulties that limited both their scope and depth . Studies analyzing Ig sequences from immortalized cell lines ( e . g . , hybridomas ) or LPS-stimulated B cells had obvious sampling biases . In addition , earlier studies mainly focused on particular VH families ( e . g . , J558 , 7183 ) , even though the mouse IgH locus contains over 100 functional VH genes ( Kirkham and Schroeder , 1994 ) . The introduction of single cell analyses enabled higher precision and lower bias than the bulk measurements . However , they were constrained profoundly by sequencing costs and technical challenges . Indeed , our previous single cell analysis reported only 184 IgH sequences derived from 85% recovered sorted single cells representative of three types of peritoneal B subsets ( Kantor et al . , 1997 ) . Thus , while the data yielded key insights , hundreds or thousands of single cells would need to be analyzed to obtain a more comprehensive view for a single B subset repertoire . Finally , difficulties in defining and cleanly sorting rare B subsets ( e . g . , splenic B-1a ) further compromise the attempt to develop a thorough view of repertoire ( s ) expressed by various B cell subsets at the different anatomic location and ontogenic stage . To overcome these obstacles , we have coupled high-dimensional ( Hi-D ) FACS sorting with unique IgH multiplex PCR technologies , which allow inclusive amplification of IgH transcripts for each sorted B subset and ultimate sequencing of these sequences . Using barcoded sample multiplexing , we have performed a large-scale quantitative and comparative study of the ‘pre-immune’ IgH repertoires expressed by various functionally and developmentally distinct mature B subsets ( splenic FOB , MZB and B-1a; peritoneal B-2 and B-1a ) from non-immune C57BL/6J mice . In addition , since microbiota are often thought to influence the Ig repertoire , we have compared the B-1a IgH repertoires in GF or conventional mice .
We sorted splenic and peritoneal B-1a ( dump- CD19+ CD93-IgMhi IgDlo/- CD21-/lo CD23- CD43+ CD5+ ) ; splenic FOB and peritoneal B-2 ( dump- CD19+ CD93- IgMlo IgDhi CD23+ CD43- CD5- ) ; and splenic MZB ( dump- CD19+ CD93- IgMhi IgDlo/- CD21hi CD23lo/- CD43- CD5- ) from non-immune C57BL/6 mice ( Figure 1 ) . We generated and amplified IgH cDNA libraries from each subset . We then pooled the libraries , which are distinguishable by barcode , and sequenced them ( Illumina MiSeq ) . In all , we sequenced 60 separately prepared libraries , each derived from 1-2 x104 B cells of a given subset sorted from mice at the same or different ages ( from 2 days to 6 months , > 30 mice ) ( Table 1 ) . Overall 18 million total clean nucleotide sequences ( CNT ) and about half million unique clean nucleotide sequences ( CNU ) were analyzed in the study ( Table 1 ) . 10 . 7554/eLife . 09083 . 003Figure 1 . The B-1a IgH CDR3 sequences are much less diverse and recur more frequently than the CDR3 sequences expressed by FOB and MZB B subsets . IgH CDR3 tree-map plots illustrating the IgH CDR3 nucleotide sequences expressed by indicated B cell subsets sorted from one 2-month old C57Bl/6 mouse . Each rectangle in a given tree-map represents a unique CDR3 nucleotide sequence and the size of each rectangle denotes the relative frequency of an individual sequence . The colors for the individual CDR3 sequences in each tree-map plot are chosen randomly thus do not match between plots . The numbers shown in the CDR3 tree-map plots highlight the highly reoccurring CDR3 sequences including PtC-binding CDR3 sequences . 1 , ARFYYYGSSYAMDY , V1-55D1-1J4; 2 , MRYGNYWYFDV , V11-2D2-8J1; 3 , MRYSNYWYFDV , V11-2D2-6J1; 4 , MRYGSSYWYFDV , V11-2D1-1J1 . Lower middle panel: FACS plots showing the gating strategy used to sort the phenotypically defined each B cell subset from spleen ( s ) or peritoneal cavity ( p ) . Note: peritoneal B-1a cells are well known to express CD11b , a marker expressed on many myeloid cells including macrophage and neutrophils . The level of CD11b expressed on peritoneal B-1a cells , however , is roughly 100 fold lower than the level of CD11b expressed on the myeloid cells . This drastic difference is sufficient to separate the CD11b+ B-1a cells from the myeloid cells if monoclonal anti-CD11b reagent is included in the dump channel ( Figure 1—figure supplement 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 00310 . 7554/eLife . 09083 . 004Figure 1—figure supplement 1 . FACS plots showing CD43+ CD5+ IgM+ B-1a cells in E19 fetal liver . Live dump- ( CD11b- CD11c- Gr-1- F4/80- CD3- TCRαβ- ) CD45+ CD19+ cells from E19 fetal liver of C57Bl/6 mouse were gated to show IgM and IgD expression . The boundary for IgM expression was determined from fluorescence-minus-one ( FMO ) control in which fluorescently labeled anti-mouse IgM antibodies are omitted from the staining sets ( right plot ) . IgM+ IgD- cells were further gated to reveal CD43+ CD5+ B-1a cells . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 00410 . 7554/eLife . 09083 . 005Figure 1—figure supplement 2 . Recurrent VH11-encoded PtC-binding V ( D ) J sequences . ( A-C ) lists three VH11-encoded PtC-binding V ( D ) J sequences . In each plot , the first line of nucleotides is the obtained sequence read while the second line refers the germline reference sequence . The underlined nucleotides are CDR2 and CDR3 . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 00510 . 7554/eLife . 09083 . 006Figure 1—figure supplement 3 . CD11b expression on peritoneal B-1a ( CD5+ ) and B-1b ( CD5- ) is roughly 100-fold lower than the CD11b expression on myeloid cells . Live cells from C57Bl/6 peritoneal cavity were gated to show CD19 and CD11b expression . The CD19 + B cells and CD11bhi myeloid cells were shown . The CD19+ B cells were gated to reveal CD5 and CD11b expression . CD11b+ B-1a and CD11b + B-1b cells were gated based on FMO control staining where anti-CD11b antibody was omitted in the staining . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 00610 . 7554/eLife . 09083 . 007Table 1 . Summary of the sequences for 60 separately sorted B cell populations analyzed in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 007SampleIdSubsetStrainAgeConditionMiceRNT*RNU*RPU*CNT*CNU*CPU*17631FOBWT2MSPFsingle1006030151210658719034002124020470213966FOBWT3 . 5MSPFsingle1508123100324801130652149111467838706FOBWT4MSPFsingle1803655357727817159710169011656848702FOBWT5MSPFsingle15668154195277281361011695116649513967FOBAID KO5MSPFsingle3596714623132032772671877133611161MZBWT1MSPFsingle3354819628127442567465846471710658MZBWT2MSPFsingle71458269781827861258115121117087630MZBWT2MSPFsingle103238113983262520932353207801979298701MZBWT4MSPFsingle21423855075264581910651546115021108700MZBWT5MSPFsingle118863423102279410289414517141801113338MZBWT4MGFsingle162754399302361114164612939126051213343MZBWT4MGFsingle595780854974582053607219266184801311163pB-1aWT1MSPFpool of 3 mice4588211290559641368323730071410660pB-1aWT2MSPFsingle222324173118630207749389136491513018pB-1aWT2MSPFsingle808879360311481775386847694374167628pB-1aWT2MSPFsingle178467759458221051706235660158481711160pB-1aWT2WSPFpool of 8 mice6531714700702558034424037041810655pB-1aWT3WSPFpool of 5 mice628751216266225755841803694198705pB-1aWT4MSPFsingle310077284411188628769550634707209870pB-1aWT4MSPFsingle2291002629910469211514474544802111165pB-1aWT5MSPFsingle1054101952889269599444354162228707pB-1aWT5MSPFsingle320252297861242329694647224384239861pB-1aWT6MSPFsingle26613568332352354215211461248704pB-1aAID KO4MSPFsingle2643403374514519245941664862942510657pB-2WT2MSPFsingle53953230591688344986100849923267629pB-2WT2MSPFsingle131566312347247337123822516925160652713969pB-2WT3 . 5MSPFsingle186817243041768917076890898925289862pB-2WT4MSPFsingle2259113377873717343438243572913973pB-2AID KO5MSPFsingle617893623194116556682617536169653013000sB-1aWT2dSPFpool of 8 mice294399542492525369314827583110651sB-1aWT5dSPFsingle1233602247210838113161745359763210659sB-1aWT5dSPFsingle210055281401241119266273075812339866sB-1aWT5dSPFsingle5298615600686446580459538373410652sB-1aWT6dSPFsingle172875264371254515930476836365359865sB-1aWT7dSPFsingle713091844687756424154824941369868sB-1aWT7dSPFsingle2018133506914473186227784768433710656sB-1aWT2MSPFsingle3697323960319759342914948990483813004sB-1aWT2MSPFsingle185948279521387516852273137022397632sB-1aWT2MSPFsingle182521810279743190171924612428111444011168sB-1aWT2WSPFsingle536603702012882949667111948109134113005sB-1aWT2WSPFsingle98017283311500185489882082074210654sB-1aWT3WSPFsingle146560338141969713109111995114514313970sB-1aWT3 . 5MSPFsingle170925138099289160480451342734413335sB-1aWT4MSPFsingle221754822344918683113110904513342sB-1aWT4MSPFsingle283072236681294726274453575032468699sB-1aWT4MSPFsingle14283819151993813091543704086479863sB-1aWT4MSPFsingle7367616599871365571423340924811167sB-1aWT5MSPFsingle501367389121733646386375737163498708sB-1aWT5MSPFsingle577114527232227253150891468441509867sB-1aWT6MSPFsingle1134922061210625101791456343435113965sB-1aAID KO4MSPFsingle1777821641912281164189653962935213971sB-1aAID KO4MSPFsingle5171413415922031482543896683955313968sB-1aAID KO5MSPFsingle4276713083920510396974916285455413972sB-1aAID KO5MSPFsingle7061163621723255660874929487445513001sB-1aWT4MGFsingle435078734485538947231822495613002sB-1aWT4MGFsingle472038683482042279205319655713003sB-1aWT4MGFsingle2133472224611068197769470544495813017sB-1aWT4MGFsingle5322504049717375501908701963985913337sB-1aWT4MGFsingle285596322441724047154414866013341sB-1aWT4MGFsingle388208289421483736072756745144Id is a unique identifier for the sequence runRNT* , total raw nucleotide sequencesRNU* , unique raw nucleotide sequencesRPU* , unique raw peptide sequencesCNT* , total clean nucleotide sequencesCNU* , unique clean nucleotide sequencesCPU* , unique clean peptide sequencesSequence statisticsRNT*RNU*RPU*CNT*CNU*CPU*Total1 . 9E + 072 . 1E + 061 . 1E + 061 . 8E + 074 . 9E + 054 . 7E + 05Mean319865356101784829517482337762% CV12286741256163 We also attempted to analyze the B-1a repertoire in fetal liver but found that there were too few B-1a cells to reliably sequence with our method . In essence , FACS analysis of embryonic day 19 ( E19 ) fetal liver cells shows that IgM+ B cells represent only 0 . 6% of CD19+ total B cells and that only around 20% of these IgM+ B cells express the B-1a CD43+ CD5+ phenotype ( Figure 1—figure supplement 1 ) . The frequencies of IgM+ B cell in E18 fetal liver are even lower ( 0 . 2% of CD19+ B cells ) . These numbers are too low for us to recover enough material for sequencing from a feasible number of embryos . The IgH CDR3 tree maps for each B cell subset show that splenic FOB and peritoneal B-2 cells express highly diversified IgH CDR3 nucleotide sequences , as do MZB cells ( Figure 1 ) . In contrast , CDR3 nucleotide sequences expressed by B-1a cells from either spleen or PerC are far less diverse and recur much more frequently ( Figure 1 ) . The recurrent CDR3 sequences include the well-studied VH11-encoded sequences specific for phosphatidylcholine ( PtC ) ( Figure 1—figure supplement 2 ) and known to occur frequently in B-1a cells ( Mercolino et al . , 1988; Hardy et al . , 1989; Seidl et al . , 1997 ) . D50 metric analysis quantifying the IgH CDR3 nucleotide sequence diversity shows that the IgH CDR3 nucleotide sequences expressed by the FOB and MZB subsets are significantly more diverse than those expressed by splenic and peritoneal B-1a cells ( p = 0 . 0002 , Mann-Whitney-Wilcoxon Test ) ( Figure 2A ) . Consistent with this finding , IgH CDR3 peptide pairwise sharing analysis , which measures the similarity of IgH CDR3 peptide expression for each B cell subset sorted from different mice , shows that the same CDR3 peptide sequences frequently appear in both splenic and peritoneal B-1a cells from different mice whereas the common CDR3 peptides are rare in FOB and MZB subsets ( Figure 2B ) . Taken together , these data demonstrate that the B-1a pre-immune IgH repertoire is far more restricted and repetitive than IgH repertoires expressed by FOB and MZB subsets . 10 . 7554/eLife . 09083 . 008Figure 2 . The B-1a pre-immune IgH repertoire is far more restricted than the pre-immune IgH repertoires expressed by splenic FOB , MZB and peritoneal B-2 cells . ( A ) D50 metric analysis quantifying the IgH CDR3 diversity for B cell subsets from mice at the indicated age . Low D50 values are associated with less diversity . Each dot represents the data for a B cell sample from an individual mouse except for the 2 day splenic B-1a data , which are derived from sorted cells pooled from 8 mice . B-1a samples are labeled with red; B-2 samples include FOB ( green , n = 4 ) , pB-2 ( purple , n = 4 ) and MZB ( yellow , n = 4 ) . The data for germ-free ( GF ) animals is discussed at the end of the Result section . ( B ) CDR3 peptide pair-wise sharing analysis of IgH repertoire similarity among multiple samples for each B cell group ( n = 5-9 ) . Each dot represents the percentage of common CDR3 peptides in one sample that are also found in another sample within a given group . For example , to compute the similarity between sample A and B , the percentage of CDR3 peptides in sample A that are also found in sample B ( pA → B ) , together with the percentage of CDR3s in sample B that are also in sample A ( pB→A ) are used as an indicator . For comparison of 6 splenic B-1a samples in 5-7 day group , there are 30 comparisons . Right upper: p values showing the statistical significance between two groups . Box plots represent the 10th , 25th , 50th , 75th and 90th percentiles here and in other figures . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 008 We quantified the frequency of IgH sequences expressing individual VH gene for each sorted B cell sample and then compared the VH gene usage between two B cell subsets . B-1a cells are well-known to undergo self-replenishing in adult ( Kantor et al . , 1995 ) . To minimize the impact of clonal expansion on the VH gene usage profile , we collected normalized data , in which we scored each distinct IgH CDR3 nucleotide sequence expressing a given VH gene as one , no matter how many times this sequence was detected . Our approach enables detection of Ig transcripts expressing about 100 different VH genes that belong to 14 VH families ( Figure 3 ) . B-1a cells express all of these detected VH genes ( Figure 3A ) , contrasting with earlier impressions , based largely on hybridomas sequences from fetal and neonatal mice ( Malynn et al . , 1990 ) , that VH usage in the B-1a repertoire is very restricted . However , despite the broad VH usage , certain VH genes , notably V10-1 ( DNA4 ) , V6-6 ( J606 ) , V11-2 ( VH11 ) and V2-6-8 ( Q52 ) , are expressed at a significantly higher frequency in splenic B-1a than MZB cells ( p<0 . 05 , Welch's t-test , Figure 3B ) . 10 . 7554/eLife . 09083 . 009Figure 3 . Comparison of VH gene usage by splenic B-1a vs MZB B cells . ( A ) VH gene usage profile shown as the percentage of IgH sequences expressing the listed individual VH genes for individual B cell samples . The profiles are shown for adult splenic B-1a samples ( n = 9 , red ) and for MZB samples ( n = 5 , green ) . VH genes ( from left to right ) are ordered in 5’- to 3’-direction bases on chromosome location; the IMGT VH gene nomenclature is used ( Lefranc , 2003 ) . ( B ) VH genes showing the statistically significant differences ( Welch’s t-test p<0 . 05 ) between two groups are listed and also highlighted with asterisks in the plot . To minimize the impact of the clonal expansion on the VH gene usage profile , data are presented as the normalized distribution that counts each distinct CDR3 nucleotide sequence expressing a given VH gene as one , no matter how many times the sequence was detected . Note: VH12-3 encoded IgH sequences are not detected in this study due to the technical limitations that exclude the VH12-3 primer from the set of primers designed about three years ago and used for studies presented here . We have since corrected this problem so that VH12-3 primer is now part of our new set of primers . Comparison of sequence data obtained with old vs . the new set of primers shows that , aside from now detecting VH12-3 sequences with the new set of primers , the sequences obtained with both primer sets are highly similar ( Figure 3—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 00910 . 7554/eLife . 09083 . 010Figure 3—figure supplement 1 . VH gene usage profile pair-wise comparison of B cell groups . The colors shown at the bottom right distinguish the B cell groups ( n = 4-9 ) . VH genes showing the statistically significant differences ( Welch’s t-test p<0 . 05 ) between two groups are listed on the bottom ( A' to F' ) and also highlighted with asterisks in each plot . The data for germ-free ( GF ) animals is discussed at the end of the Result section . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 01010 . 7554/eLife . 09083 . 011Figure 3—figure supplement 2 . Almost identical top 10 highly recurring CDR3 sequences are detected for splenic B-1a IgH libraries obtained either with the old or new primer set . We sorted two splenic B-1a populations individually from two 4 month old C57BL/6J mice . We extracted RNA from each population and divided each RNA into two parts . For one part , we prepared an amplified library using the old primer set; and for the other , we prepared an amplified library using the new primer set . We then sequenced these amplified IgH libraries . Analysis of the resultant sequences showed that the sequences obtained from the IgH libraries are highly similar , regardless of the primers used ( old or new ) . In essence , the top 10 highly recurring CDR3 sequences ( both peptide and V ( D ) J recombination ) are almost identical and show similar representation order between each pair of libraries . As expected , we detected VH12-3 encoded sequences from the splenic B-1a IgH libraries prepared with the new primer set , and these VH12-3 encoded sequences included several published PtC-binding VH12-3 encode sequences , i . e . , AGDYDGYWYFDV ( VH12-3D2-4J1 ) , AGDRDGYWYFDV ( VH12-3D3-2J1 ) , AGDRYGYWYFDV ( VH12-3 D2-9 J1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 011 Similar to MZB cells , splenic FOB and peritoneal B-2 cells show lower frequency in expressing these B-1a favored VH genes , i . e . , V6-6 ( J606 ) , V11-2 ( VH11 ) and V2-6-8 ( Q52 ) ( Figure 3—figure supplement 1B–C ) . Conversely , these B subsets tend to preferentially use the largest VH family , V1 ( J558 ) , located distal to DH and JH gene segments ( Yancopoulos and Alt , 1986 ) . MZB cells , in particular , have a higher tendency to express certain V1 ( J558 ) family genes including V1-82 , V1-72 , V1-71 , V1-42 , V1-18 and V1-5 ( Figure 3B ) . The VH usage in the peritoneal B-1a cells is further biased toward V6-6 ( J606 ) , V9-3 ( Vgam3 . 8 ) , V2-9 ( Q52 ) and V2-6-8 ( Q52 ) genes , which are already favored in the splenic B-1a cells ( Figure 3—figure supplement 1A ) . This finding indicates that the splenic and peritoneal B-1a populations are not in equilibrium and the latter is further enriched for cells expressing certain VH genes . Unlike FOB and MZB subsets , de novo B-1a development initiates prior to birth and decreases to a minimum in adult animals ( Lalor et al . , 1989; Barber et al . , 2011 ) . B-1a cells persist thereafter as a self-replenishing population ( Kantor et al . , 1995 ) . To minimize the impact of self-replenishment on the N-addition distribution profile , and hence to weight the repertoire for de novo generated IgH sequences for B-1a cells , we collected normalized data that counts each distinct IgH sequence containing indicated N nucleotide insertions as a single sequence , regardless how many times this sequence was detected . Consistent with Tdt expression , which is absent during the fetal life and initiates shortly after birth ( Feeney , 1990; Bogue et al . , 1992 ) , N nucleotide insertion analysis of the splenic B-1a IgH repertoires demonstrate that roughly 60% of IgH sequences expressed by splenic B-1a cells from 2-–6 day mice do not contain N insertions at IgH CDR3 junction ( D-J and V-DJ ) ; about 30% contain 1–2 insertions; and , <15% contain 3–4 N-nucleotide insertions ( Figure 4A , B ) . After 6 days , however , the frequency of sequences containing >3 N-additions progressively increases until the animals are weaned ( roughly 3 weeks ) ( Figure 4A , B ) . After weaning , the N-addition pattern stabilizes , i . e . , about 50% IgH sequences contain 3–7 N nucleotide insertions and about 30% have more than 8 N nucleotide insertions at IgH CDR3 junctions , and remains stable at this level for at least 5 months ( Figure 4A , B ) . 10 . 7554/eLife . 09083 . 012Figure 4 . N nucleotide insertion distribution patterns for the B-1a pre-immune IgH repertoires during ontogeny . ( A ) Percentage of IgH sequences containing the indicated number of N nucleotide insertions at the IgH CDR3 junctions ( V-DJ + D-J ) is shown for each spleen B-1a sample from mice at indicated ages ( shown at the right ) . To minimize the impact of self-renewal on the N-addition profile , normalized data are presented . Thus , each distinct IgH sequence containing indicated N nucleotide insertions is counted as one regardless how many times this sequence was detected . Note that the N insertion pattern changes as animals age . Colors distinguish three age-related patterns: green , D2 to D6; blue , D7 to 3W; red , 2M to 6M . ( B ) Percentages of IgH sequences containing the indicated N-nucleotide insertions ( shown at the top ) for splenic B-1a samples at the indicated ages are shown . Each dot represents data from an individual mouse , except for day 2 sample , n = 5-7 . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 012 In essence , splenic B-1a cells from 2-6 day mice largely originate from fetal and early neonatal wave ( s ) of B-1a development when Tdt is poorly expressed . As newborns progress to maturity , B-1a cells , which are originated in the earlier wave ( s ) , are ‘diluted’ by B-1a cells that emerge during later development . The high frequency of N nucleotide additions in the adult splenic B-1a IgH repertoire indicates that a higher proportion of B-1a cells are actually generated postnatally after Tdt is expressed . Cohering with the increased N diversity in the adulthood , CDR3 peptide pairwise sharing analysis shows that the expression of common IgH CDR3 peptides is significantly more frequent in neonatal splenic B-1a cells than in adult splenic B-1a cells ( p<2e-16 , Mann-Whitney-Wilcoxon Test , Figure 2B ) . VH usage also shifts as animals mature . Splenic B-1a cells from neonatal mice ( 2-–7 days ) preferentially express the V3 ( 36–60 ) , V5 ( 7183 ) and V2 ( Q52 ) families that are largely located proximal to D and J gene segments ( Figure 3—figure supplement 1D ) , consistent with previous findings that hybridomas derived from fetal/neonatal B cells are bias in expressing proximal V5 ( 7183 ) and V2 ( Q52 ) family genes ( Perlmutter et al . , 1985 ) . In contrast , the splenic B-1a cells from adult animal ( 2–6 months ) show higher frequencies in expressing distal V1 ( J558 ) family genes including V1-75 , V1-64 , V1-55 and V1-53 ( Figure 3—figure supplement 1D ) . Collectively , we conclude that the B-1a IgH repertoire integrates rearrangements from sequential waves of de novo B-1a development that mainly occur during the first few weeks of life . The IgH repertoires defined during these waves are distinguishable both by N-additions at CDR3 junctions and by VH gene usage . Certain V ( D ) J nucleotide sequences become progressively more dominant with age in the B-1a repertoire . Thus , only a lower proportion of V ( D ) J sequences are detected at relative higher frequency in the splenic B-1a IgH repertoire before 3 weeks , after which , both the number of recurrent sequences and the frequency at which each is represented increase progressively until the animals reach 4–6 month of age ( Figure 5A , Table 2 ) . Consequently , the distribution of the splenic B-1a IgH CDR3 nucleotide sequences diversity is much less random in adults ( 2–6 months ) than in neonates ( 2–7 days ) ( Figure 2A ) . 10 . 7554/eLife . 09083 . 013Figure 5 . Certain V ( D ) J sequences increase progressively with age in the B-1a pre-immune IgH repertoire . ( A ) IgH CDR3 tree map plots for splenic B-1a samples from mice at different ages are shown . Each plot represents data for an individual mouse , except for the day 2 sample . Recurrent sequences are visualized as larger contiguously-colored rectangles in each plot . ( B ) Relative frequencies of three PtC-binding IgH CDR3 sequences in indicated splenic B-1a sample groups ( n = 5–8 for each group ) are plotted with mouse age . Sequence information ( peptide and V ( D ) J recombination ) is shown at the top . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 01310 . 7554/eLife . 09083 . 014Figure 5—figure supplement 1 . The peritoneal B-1a IgH repertoire is increasingly restricted during ontogeny . IgH CDR3 tree map plots for peritoneal B-1a samples from different ontogenic stages . Each plot represents the data for a sample from an age-defined individual mouse , except for the 2 week , 3 week and 1 month samples , which are obtained from cells pooled from several mice . Recurrent sequences are visualized as larger contiguously-colored rectangles in each plot . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 01410 . 7554/eLife . 09083 . 015Table 2 . Top 10 highly recurring CDR3 sequences ( peptide and V ( D ) J recombination ) detected in each of the listed splenic B-1a samples . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 015sB-1a samplesTop 10 IgH CDR3 sequencesIdAgePeptideV ( D ) J111682 weeks1ANDYV1-53 J22AKHGYDAMDYV2-9 D2-9 J43ARRYYGSSYWYFDVV1-55 D1-1 J14ANWDYV1-53 D4-1 J25MRYSNYWYFDVV11-2 D2-6 J16ARDAYYWYFDVV7-1 J17ATDYYAMDYV1-26 J48ARFYYYGSSYAMDYV1-55 D1-1 J49AIYYLDYV1-53 D2-8 J210ARHYGSSYWYFDVV2-6-2 D1-1 J1106543 weeks1ARRYYGSSYWYFDVV1-55 D1-1 J12ARSYSNYVMDYV1-76 D2-6 J43ARYYGSNYFDYV7-3 D1-1 J24ARGASYYSNWFAYV1-55 D2-6 J35ALTGTAYV1-53 D4-1 J36ARAGAGWYFDVV5-9 D4-1 J17TYSNYV6-6 D2-6 J28ARTGTYYFDYV1-53 D4-1 J29AMVDYV1-64 D2-9 J210ARWGTTVVGYV1-7 D1-1 J276322 months1MRYGNYWYFDVV11-2 D2-8 J12MRYSNYWYFDVV11-2 D2-6 J13MRYGSSYWYFDVV11-2 D1-1 J14ATFSYV1-55 J25ARFYYYGSSYAMDYV1-55 D1-1 J46ARIPNWVWYFDVV1-55 D4-1 J17ARWDTTVVAPYYFDYV1-7 D1-1 J28ARDYYGSSWYFDVV1-26 D1-1 J19TYYDYDLYAMDYV14-4 D2-4 J410ARFITTVVATRYWYFDVV1-9 D1-1 J186994 months1ARSADYGGYFDVV1-64 D2-4 J12ARGAYV1-80 J23ARSYYDYPWFAYV1-76 D2-4 J34ARRWLLNAMDYV1-9 D2-9 J45ARPYYYGSSPWFAYV1-69 D1-1 J36ARNDYPYWYFDVV1-4 D2-4 J17ARSGDYV1-64 J28ARVIGDYV1-53 D2-14 J49ARANYV1-55 J310AVNWDYAMDYV1-84 D4-1 J487085 months1ASLTYV1-55 J22TCNYHV14-4 D2-8 J43LIGRNYV1-55 D2-14 J24MRYSNYWYFDVV11-2 D2-6 J15AKQPYYGSSYWYFDVV2-3 D1-1 J16AGSSYAYYFDYV1-66 D1-1 J27ARRGIDLLWYHYYAMDYV1-26 D2-8 J48ARKSSGSRAMDYV7-3 D3-2 J49ASYAMDYV7-3 J410ARLYYGNSYWYFDVV1-55 D2-8 J198676 months1ARKYYPSWYFDVV1-55 D1-1 J12AREGGKFYV1-7 J23AKSSGYAMDYV1-55 D3-2 J44ARWVITTVARYFDVV1-85 D1-1 J15ARGFYV1-80 J26AKEGGYYVRAMDYV1-55 D1-2 J47ARSMDYV1-80 J48ASAMDYV1-64 J49TKGGYHDYDDGAWFVYV1-53 D2-4 J310ARKFYPSWYFDVV1-55 J3Table lists the top 10 highly recurring CDR3 sequences ( peptide and V ( D ) J recombination ) shown in the individual CDR3 tree-map plot of the splenic B-1a samples from 2 week to 6 month old mice ( Figure 5A ) . For each splenic B-1a sample , the Id number and mouse age are shown in column 1 and column 2 respectively . The recurrent V ( D ) J sequences include VH11-encoded PtC-binding V ( D ) J sequences , which are initially present at very low frequencies ( 2–6 days ) but increase aggressively as animals mature to middle age ( 6 months ) ( Figure 5B ) . Since de novo B-1a development is minimum at adulthood , the progressive increase in the representation of the recurrent V ( D ) J sequences as animals reach adulthood suggests that B-1a cells are self-replenishing . To determine to what extent the IgH CDR3 sequences ( amino acid and nucleotide ) expressed by each B cell subset are shared across different individuals , we carried out CDR3 sharing analysis . In the B-1a IgH repertoire , overall , we found 30 such highly shared IgH CDR3 peptides , each of which is expressed in over 80% of the splenic B-1a samples taken from more than 20 animals with nine different ages ( from 2 days to 6 months ) ( Table 3 ) . Each of the shared CDR3 peptides would be expected to be encoded by several convergent V ( D ) J recombinations , i . e . , distinct V ( D ) J rearrangements encode the same CDR3 amino acid sequence ( Venturi et al . , 2008 ) . Strikingly , we found that each of the shared CDR3 peptides is encoded by an identical V ( D ) J nucleotide sequence in over 70% of splenic B-1a samples from adult animals ( 2-6 months , 9 mice ) ( Table 3 ) . 10 . 7554/eLife . 09083 . 016Table 3 . Certain V ( D ) J sequences are positively selected and conserved in adult B-1a pre-immune IgH repertoires . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 016CDR3 peptidePredominant V ( D ) JCDR3 junction diversityRepresentation in indicated repertoiresplenic B-1a ( 2d-6M ) splenic B-1a ( 2-6M ) additiondeletionPerC B-1a ( 2W-6M ) splenic B-1a ( 4M germ free ) FOB ( 2-5M ) MZB ( 1-5M ) 1TRWDY17/20V6-6 J28/9TGGJ2 ( 8 ) 11/115/61/80/72MRYSNYWYFDV17/20V11-2 D2-6 J19/90011/116/61/81/73MRYGNYWYFDV18/20V11-2 D2-8 J19/90011/116/61/81/74MRYGSSYWYFDV17/20V11-2 D1-1 J19/90011/116/61/81/75VRHYGSSYFDY15/20V10-1 D1-1 J25/90J2 ( 1 ) 11/113/60/80/76ARHYYGSSYYFDY19/20V5-6-1 D1-1 J29/90011/116/62/80/77ARLDY20/20V1-53 J27/9CTg/aJ2 ( 8 ) 10/114/60/81/78ARDYYGSSYWYFDV19/20V7-1 D1-1 J16/90V7-1 ( 3 ) 9/115/61/81/79ARDYYGSSWYFDV19/20V1-26 D1-1 J17/9GJ1 ( 3 ) 2/114/60/81/710ANWDY19/20V14-3 D4-1 J26/90V14-3 ( 2 ) J2 ( 8 ) 5/112/60/80/711ATGTWFAY18/20V1-19 D4-1 J35/90V1-19 ( 2 ) 6/112/60/81/712ARYYYGSSYAMDY19/20V7-3 D1-1 J48/90V7-3 ( 1 ) J4 ( 4 ) 10/113/63/83/713ARYSNYYAMDY18/20V1-39 D2-6 J46/90J4 ( 2 ) 8/111/60/80/714ARDFDY19/20V1-64 J26/9GJ2 ( 3 ) 1/113/61/81/715ARYYSNYWYFDV17/20V1-9 D2-6 J16/9004/111/60/80/716ARYDYDYAMDY17/20V1-39 D2-4 J46/90J4 ( 3 ) 7/111/60/80/717ARHYYGSSYWYFDV18/20V2-6-2 D1-1 J16/9006/112/61/83/718ARFYYYGSSYAMDY19/20V1-55 D1-1 J46/9TJ4 ( 4 ) 8/113/61/81/719ARWDFDY19/20V1-7 J26/9TGGGJ2 ( 3 ) 1/113/61/81/720ARGAY19/20V1-80 J35/9GGGJ3 ( 8 ) 7/116/61/81/721ARRFAY18/20V1-26 J37/9C/AJ3 ( 8 ) 9/113/61/81/722ARRDY18/20V1-55 J25/9AGg/aJ2 ( 8 ) 6/113/61/81/723ASYDGYYWYFDV18/20V1-55 D2-9 J18/9CTATGV1-55 ( 1 ) 9/115/60/80/724ASYAMDY16/20V7-3 J48/90V7-3 ( 5 ) J4 ( 4 ) 9/116/60/81/725ARRYYFDY17/20V1-78 J27/9CGg/cT08/112/60/80/726ARNYYYFDY15/20V1-53 D1-2 J28/9t/a010/112/60/80/727ARYYGNYWYFDV15/20V3-8 D2-8 J15/9005/112/60/80/728ARRYYGSSYWYFDV15/20V1-55 D1-1 J17/9CGG010/115/61/81/729ARRLDY13/20V1-22 J27/9CGACJ2 ( 6 ) 8/112/60/81/730ARFAY18/20V1-80 J34/90J3 ( 4 ) 2/113/60/80/7Column 1: CDR3 peptide sequences identified to be shared in >80% of splenic B-1a samples ( 20 samples from mice ranging from 2 day to 6 month old ) ; Column 2: for each shared CDR3 peptide , a single V ( D ) Jrearrangement sequence is selected and conserved in over 70% of adult B-1a samples ( 9 samples , 2-6 month old ) ; Columns 3 and 4: nucleotides added or deleted in CDR3 junctions; Columns 5-8: the representation of each selected V ( D ) J sequence within the indicate repertoires ( age and number of samples are shown for each group ) . Rows 2-4 are PtC-binding CDR3 sequences; Row 8 is CDR3 sequence for T15 Id+ anti-PC antibody . The data for germ-free animals is discussed at the end of the Result section . These V ( D ) J nucleotide sequences represent the IgH structures that are positively selected into the shared adult B-1a IgH repertoire among C57BL/6 mice . Although the specificities of the majority of these selected V ( D ) J sequences remain to be defined , they include sequences that are specific for PtC and sequence for the T15 idiotype B-1a anti-PC antibodies ( Masmoudi et al . , 1990 ) . Of note , most of these V ( D ) J sequences have nucleotide additions and/or deletions in the CDR3 junction ( Table 3 ) , indicating that the driving force for the selection may include , but is certainly not restricted to the germline rearrangement . The majority of the V ( D ) J nucleotide sequences that are conserved in the splenic B-1a IgH repertoire are also conserved in the peritoneal B-1a IgH repertoires ( 2W-6M , 11 samples ) ( Table 3 ) . Such V ( D ) J nucleotide sequences , however , are rarely detectable in FOB and MZB IgH repertoires ( 1-5M , 7-8 samples ) , either because these cells do not express these CDR3 peptides or because they use different V ( D ) J recombination sequences to encode them ( Table 3 ) . For example , although MZB cells express antibodies encoding the same CDR3 peptide as B-1a T15-id+ , they use different V ( D ) J recombinations and no single V ( D ) J recombination dominates within the MZB IgH repertoire ( Table 4 ) . In essence , the selection of a predominant V ( D ) J nucleotide sequence encoding a given CDR3 peptide is unique for the B-1a IgH repertoire . 10 . 7554/eLife . 09083 . 017Table 4 . MZB IgH repertoires use different V ( D ) J recombination sequences to encode the same CDR3 peptide as that of B-1a anti-PC T15Id+ . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 017MZB sample IdAge ( Months ) V ( D ) J recombination76302V1-76 D1-1 J1 and V1-39 D1-1 J1106582V1-76 D1-1 J187004V1-72 D1-1 J1 and V8-12 D1-1 J187015V1-58 D1-1 J1 and V1-61 D1-1 J1133384V1-61 D1-1 J1 and V5-16 D1-1 J1Column 1: individual MZB samples tested; column 2: age of mouse for each MZB sample; column 3: for each MZB sample , V ( D ) J recombination events that encode ARDYYGSSYWYFDV , which is the CDR3 peptide associated with B-1a anti-PC T15Id+ . In 2–7 day animals , a few selected V ( D ) J nucleotide sequences , such as PtC-binding sequences , have already emerged as the predominant V ( D ) J recombination for their corresponding CDR3 peptide ( Figure 6A , pattern II ) . However , most of the selected V ( D ) J nucleotide sequences , including T15Id+ , do not initially represent the predominant recombination for their corresponding CDR3 peptide . In particular , some CDR3 peptides are each encoded by multiple different V ( D ) J recombinations with similar frequencies in neonate mice . However , after weaning , a particular V ( D ) J recombination gradually increases its representation until it dominates in the adult B-1a IgH repertoire ( Figure 6A , pattern I ) . In essence , although multiple distinctive V ( D ) J recombinations encoding the same CDR3 peptide exist in the neonatal/young B-1a IgH repertoire , a single identical V ( D ) J recombination sequence is selected to encode the particular CDR3 peptide in adult repertoire of almost all individuals . 10 . 7554/eLife . 09083 . 018Figure 6 . The level of convergent recombination in the B-1a IgH repertoire declines with age . ( A ) Entropy heat map showing the diversity of V ( D ) J recombination events for each indicated CDR3 peptide ( shown at the left ) in splenic B-1a samples at different ages ( shown at the bottom ) . The higher the entropy value , the more diverse the V ( D ) J recombinations for a given CDR3 peptide . CDR3 peptide sequences for T15 Id+ anti-PC ( pattern I ) and anti-PtC ( pattern II ) antibody are shown in bold . ( B ) The diversities of the V ( D ) J recombination for each CDR3 peptide for the indicated splenic B-1a samples ( shown at the bottom ) are quantified as entropy values ( see Methods and materials ) , which are ranked into 4 ranges ( shown at the right ) . For each sample , the frequencies of CDR3 peptide sequences belonging to each entropy range are shown as stacks . ( C ) Splenic B-1a samples are grouped based on age . For each group ( n = 5–7 ) , the frequencies of CDR3 peptide sequences belonging to each of four entropy ranges are shown . *p<0 . 05 , Welch’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 01810 . 7554/eLife . 09083 . 019Figure 6—figure supplement 1 . Distinct V ( D ) J sequences encoding the same CDR3 peptide differ in VH usage . Plots showing an example , in which four different V ( D ) J sequences expressed by the 5 day splenic B-1a sample all encode the same CDR3 ( ANWDY ) . Red line denotes the V ( D ) J recombination . CDR2 sequences are highlighted with the blue doted lined box . The V ( D ) J recombination ( V14-3 D4-1 J2 ) shown in the bottom plot is the predominant V ( D ) J for ANWDY identified in adult splenic B-1a IgH repertoire . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 019 In accordance with this finding , quantification of the diversity of V ( D ) J recombination events for each CDR3 peptide reveals the profound convergent recombination in the neonatal B-1a IgH repertoire . Thus , about 30% of CDR3 peptide sequences in splenic B-1a IgH repertoire at 2–6 day are encoded by more than one V ( D ) J recombination ( entropy >0 . 5 , Figure 6B , C ) , and about 10% of CDR3 peptide sequences show the highest level of convergent recombination ( entropy >1 . 5 , Figure 6B , C , the higher the entropy value , the more diverse the V ( D ) J recombinations ) . However , the frequency of CDR3 peptides showing convergent recombinations steadily decrease until the animals reach adulthood ( 2 months ) , after which very few ( <1% ) CDR3 peptide sequences show the multiple V ( D ) J recombinations ( entropy >1 . 5 , Figure 6B , C ) . The step-wise decreases in the level of convergent recombination as animals age indicate the potent selection that over-time shapes the B-1a IgH repertoire . In most cases , the related V ( D ) J sequences that ‘converge’ to encode the same CDR3 peptide share the same D and J segments but use distinct VH genes ( Figure 6—figure supplement 1 ) . Therefore , despite encoding the same CDR3 peptide sequence , these related V ( D ) J sequences differ in their upstream regions including the CDR2 ( Figure 6—figure supplement 1 ) . These upstream differences , which can contribute to ligand binding , may be central to the selection of the predominant V ( D ) J sequence for the corresponding CDR3 peptide . Greater than 25% of splenic B-1a IgH sequences in 4–6 month old mice have at least one nucleotide change ( Figure 7A ) . Such mutations are principally mediated by AID because they are rare ( <2% ) in splenic B-1a cells from age-matched AID-deficient mice ( Figure 7A ) . The SHM even targets V ( D ) J sequences that are positively selected into the shared B-1a IgH repertoire in wild type mice ( but not in AID-deficient mice ) ( Figure 7B , D ) . The observed mutations , most of which result in amino acid changes , are largely targeted AID hotspots , i . e . , DGYW ( D = A/G/T; Y = C/T; W = A/T ) or WRCH ( R = A/G , H = T/C/A ) ( Di Noia and Neuberger , 2007 ) ( Figure 7B , C ) . 10 . 7554/eLife . 09083 . 020Figure 7 . AID-mediated SHM accumulates on splenic B-1a IgVH with age . ( A ) Percentages of sequences containing > = 1 ( red ) or > = 2 ( green ) nucleotide changes for B cell samples from mice at the indicated ages are shown ( n = 3-8 ) . Seven B cell samples from 4-5 month old AID knockout mice include sB-1a ( n = 4 ) , pB-1a ( n = 1 ) , FOB ( n = 1 ) and pB-2 ( n = 1 ) . Sequences with the identical V ( D ) J recombination encoding ARGAY CDR3 peptide obtained from splenic B-1a sample from 4 month old specific pathogen free mouse , ( B ) germ-free mouse ( C ) and AID knockout mouse ( D ) are listed . The nucleotide substitution is analyzed at the VH region stretching from the start of CDR2 ( red box ) to the beginning of CDR3 ( yellow box ) . Obtained sequence ( upper line ) is aligned with the reference ( lower line ) for V1-80 ( red ) , J3 ( blue ) and constant region of IgM isotype ( orange ) . Mutations are highlighted with triangles; asterisks indicate mutations resulting in an amino acid change; red and blue triangles denote mutations in DGYW and WRCH motifs , respectively . ( E ) Numbers of mutations per 104 base pairs for indicated B cell group are shown . Each dot represents data from an individual sample ( n = 3–8 ) . The data for germ-free ( GF ) animals is discussed at the end of the Result section . Note: The mutation profiles for the splenic B-1a IgH libraries prepared by using either old ( VH12-3 deficient ) or new primer set ( VH12-3 included ) are highly similar ( Figure 7—figure supplement 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 02010 . 7554/eLife . 09083 . 021Figure 7—figure supplement 1 . Splenic B-1a cells do not contain cells expressing GC phenotype . FACS analysis showing of live dump- CD19 + CD93 ( AA41 ) - IgMHi IgD-/lo CD23- CD21-/lo B cells from spleen of 5 month old C57BL6/J mouse were gated to reveal CD43 + CD5 + B-1a cells , which were further gated to reveal GL7 , CD38 and CD95 expression . GC B cells are GL7 + CD38-/lo CD95hi . The boundary for CD5 ( rightmost middle plot ) and GL7 ( rightmost bottom plot ) expression were determined from FMO controls in which fluorescently labeled anti-mouse CD5 or anti-mouse GL7 antibodies are omitted from the staining sets . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 02110 . 7554/eLife . 09083 . 022Figure 7—figure supplement 2 . Percentage of sequences containing > = 4 nucleotides changes for each B cell group . A , sB-1a ( 2-7d ) ; B , sB-1a ( 2W-1M ) ; C , sB-1a ( 2M ) ; D , sB-1a ( 4-6M ) ; E , sB-1a ( GF , 4M ) ; F , B cells ( AIDKO , 4-5M ) ; G , pB-1a ( 2-6M ) ; H , FOB , pB-2 ( 2-5M ) ; I , MZB ( 1-5M ) . Each dot represents the data for an individual B cell sample , n = 3-8 . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 02210 . 7554/eLife . 09083 . 023Figure 7—figure supplement 3 . Identical V ( D ) J recombination sequences containing identical mutated nucleotides are detected in sequence data sets for IgH libraries obtained by using either old or new primer set . We sorted two splenic B-1a populations individually from two 4 month old C57BL/6J mice . We extracted RNA from each population and divided each RNA sample into two parts . For one part , we prepared an amplified library using the old primer set; and for the other , we prepared an amplified library using the new primer set . We then sequenced two pair of amplified IgH libraries . In two separate comparisons , we detected identical IgH sequences containing identical nucleotides substitutions in each library . One example is shown from comparing one pair of sequence data sets . Red nucleotides are the mutated bases . Upper line of sequence is the obtained sequence reads and the lower line of sequences is the V , D and J reference sequences . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 023 In contrast , mutations are minimal in IgVH of splenic FOB , MZB and peritoneal B-2 cells from adult mice ( Figure 7A ) . Interestingly , the frequency of mutated IgH sequences in peritoneal B-1a cells in 4-6 month old mice is substantially lower than that in age-matched splenic B-1a cells and mutations are mainly single nucleotide change ( Figure 7A ) . SHM in splenic B-1a IgVH initiates after weaning and the frequency of mutated IgH transcripts increases with age . Thus , mutations are minimally detectable in the IgVH of splenic B-1a cells from neonates ( 2–7 days ) and young mice ( 2–3 weeks ) , are at lower frequencies in 2 month old mice , and are at substantially higher frequencies in 4–6 month old animals ( Figure 7A ) . This age-dependent increase in splenic B-1a IgVH mutation argues that the detected SHM is not due to contamination with co-sorted B cells of other subsets , including GC cells , i . e . , cells with the germinal center phenotype ( GL7+ CD38lo CD95hi ) are not detectable in the splenic B-1a population ( Figure 7—figure supplement 1 ) . Furthermore , SHM is cumulative , becoming more pronounced with age . Thus , roughly 25% of IgH sequences from 4–6 month old splenic B-1a samples contain > = 1 nucleotide change , 19% contain > = 2 changes , and 9% contain > = 4 changes ( Figure 7A and Figure 7—figure supplement 2 ) . This translates to an average SHM rate of roughly 5 per 103 base pairs ( bp ) ( Figure 7E ) , the similar range as that for SHM in GC responses , i . e . , 10-3 bp per generation ( Wagner and Neuberger , 1996 ) . Both the frequency of mutated sequences and the mutation rate for splenic B-1a samples from 2 month old mice are substantially lower than those in 4–6 month old mice ( Figure 7A , E ) , further supporting that the SHM in the splenic B-1a IgVH is an accumulative process . Class switch recombination ( CSR ) is another genetic alteration process that somatically diversifies rearranged IgH genes . Both SHM and CSR are triggered by AID , which targets and introduces lesions in the IgV region for SHM and the switch regions for CSR ( Muramatsu et al . , 2000; Chaudhuri and Alt , 2004 ) . Although both events require AID , SHM and CSR employ different enzymes and thus can occur independently ( Li et al . , 2004 ) . Nevertheless , since they usually occur at the same differentiation stage and both are initiated by AID , the question arises as to whether the detected SHM in B-1a IgH is associated with CSR . Our method allows detection of all different Ig isotypes . For each B cell sample , we quantified the frequency of IgH sequences expressing a given isotype and examined the relationship between the isotype profiles to the mutation status . Consistent with the close relationship between CSR and SHM , wild-type B cell samples that have minimal IgVH mutations , including the splenic FOB , MZB , peritoneal B-2 , neonate splenic B-1a ( 2–7 days ) , young splenic and peritoneal B-1a ( 2–3 weeks ) , rarely express class-switched transcripts ( Table 5 ) . Similarly , for B cell populations that show lower levels of mutation , e . g . , splenic B-1a from 2 month old animals and peritoneal B-1a from 2–6 month old animals , the majority of both mutated and non-mutated sequences are either IgM or IgD and thus rarely class-switched ( Figure 8A , Table 5 ) . 10 . 7554/eLife . 09083 . 024Figure 8 . Progressive increase in the splenic B-1a IgVH mutation frequency with age is accompanied by increased class-switching . ( A ) Left panel: The frequencies of non-mutated or mutated ( > = 1 nucleotide substitution ) IgH sequences obtained from indicated B cell samples are shown; Right panel: The frequencies of sequences expressing class-switched isotypes ( neither IgM nor IgD ) among non-mutated or mutated sequences are shown . Each dot represents data from an individual sample ( n = 5–6 ) . p values are calculated based on the Nonparametric Wilcoxon test . ( B ) In each plot , the IgH sequences obtained from each splenic B-1a sample from 3 . 5–6 month old mice are divided into five categories , based on the number of mutated nucleotides ( 0 , 1 , 2 , 3 , 4 , > = 5 ) per read . In each plot , the values shown at the top are the frequencies of sequences in each category . For each category of sequences , frequencies of the distinct isotype sequences are shown as stacks . A = IgA; D = IgD; G = IgG1 + IgG3 + IgG2c + IgG2b . Each plot represents the data for a splenic B-1a sample from an individual mouse reared under either specific pathogen free ( SPF ) ( upper plots ) or germ-free ( GF ) ( lower plots ) conditions . The data for germ-free ( GF ) animals is discussed at the end of the Result section . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 02410 . 7554/eLife . 09083 . 025Table 5 . B cell samples that show minimal or low level mutations in IgVH rarely express class-switched transcripts . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 025Sample Idsubsetagestrainnon-mutated or mutated sequences ( % ) IgM ( % ) IgD ( % ) IgG1 ( % ) IgG3 ( % ) IgG2c ( % ) IgG2b ( % ) IgE ( % ) IgA ( % ) 13965sB-1a4MAIDKOnon-mutated98 . 299 . 50 . 5mutated1 . 810013968sB-1a5MAIDKOnon-mutated99100mutated110013971sB-1a4MAIDKOnon-mutated98 . 499 . 60 . 4mutated1 . 610013972sB-1a4MAIDKOnon-mutated10099 . 90 . 18704pB-1a4MAIDKOnon-mutated98 . 299 . 90 . 1mutated1 . 810013973pB-25MAIDKOnon-mutated10091 . 68 . 48700MZB5MWTnon-mutated99 . 899 . 90 . 18701MZB4MWTnon-mutated98 . 699 . 90 . 17630MZB2MWTnon-mutated99 . 599 . 90 . 110658MZB2MWTnon-mutated1001008702FOB5MWTnon-mutated99 . 898 . 61 . 413966FOB3 . 5MWTnon-mutated99 . 599 . 70 . 37631FOB2MWTnon-mutated99 . 372 . 927 . 17629pB-24MWTnon-mutated98 . 488 . 411 . 5mutated1 . 6891113969pB-23 . 5MWTnon-mutated99 . 590 . 19 . 913974sB-1aday 2WTnon-mutated99 . 299 . 80 . 213000sB-1aday 2WTnon-mutated99 . 210010659sB-1aday 5WTnon-mutated99 . 21009866sB-1aday 5WTnon-mutated10010010651sB-1aday 5WTnon-mutated99 . 710010652sB-1aday 6WTnon-mutated99 . 41009868sB-1aday 7WTnon-mutated99 . 399 . 90 . 19865sB-1aday 7WTnon-mutated99 . 599 . 60 . 411168sB-1a2WWTnon-mutated99 . 199 . 90 . 113005sB-1a2WWTnon-mutated99 . 510010654sB-1a3WWTnon-mutated95 . 8100mutated4 . 210011160pB-1a2WWTnon-mutated9910010655pB-1a3WWTnon-mutated99 . 210011163pB-1a1MWTnon-mutated99 . 199 . 97632sB-1a2MWTnon-mutated88 . 199 . 10 . 9mutated11 . 999 . 90 . 110656sB-1a2MWTnon-mutated88 . 199 . 80 . 2mutated11 . 910013004sB-1a2MWTnon-mutated97 . 799 . 9mutated2 . 394613018pB-1a2MWTnon-mutated91 . 8100mutated810013660pB-1a2MWTnon-mutated92 . 2100mutated7 . 61007628pB-1a2MWTnon-mutated92 . 399 . 50 . 40 . 1mutated7 . 199 . 60 . 20 . 28705pB-1a4MWTnon-mutated93 . 899 . 70 . 3mutated4 . 499 . 99870pB-1a4MWTnon-mutated86 . 499 . 9mutated12 . 699 . 911165pB-1a5MWTnon-mutated98 . 199 . 9mutated1 . 51008707pB-1a5MWTnon-mutated91 . 697 . 20 . 120 . 50 . 10 . 1mutated6 . 297 . 90 . 129861pB-1a6MWTnon-mutated82 . 499 . 60 . 4mutated17 . 5100Table lists each individual B cell sample ( labeled as distinct Id number ) from wild-type ( WT ) or AID-deficient ( AIDKO ) mice . The mouse age and sample subset information are also shown . For each sample , the sequences are divided into non-mutated or mutated ( > = 1 nucleotide change ) categories , the frequencies of each category are shown . For each category , the frequencies of sequences with each isotype are also shown . In contrast , both the mutated and non-mutated IgH sequences from splenic B-1a in 4–6 month old animals contain class-switched Ig ( Figure 8 ) . Importantly , the class-switched Ig ( mainly IgG3 , IgG2b , IgG2c and IgA ) represents a significantly higher proportion of the mutated sequences than of the non-mutated sequences ( Figure 8A , Table 6 ) , indicating that the increased SHM with age in the splenic B-1a IgH repertoire is accompanied by increased class-switching . However , despite the increased class switching among mutated sequences , the frequency of class-switched sequences appears not to correlate with the increased number of mutations ( Figure 8B ) . Consistent with the class-switching dependence on AID , we did not detect isotypes other than IgM and IgD in splenic B-1a cells from 4–5 month old AID-deficient mice ( Table 5 ) . 10 . 7554/eLife . 09083 . 026Table 6 . Both the mutated and non-mutated IgH sequences obtained from splenic B-1a cells in 4-6 month old animals contain class-switched Ig . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 026sample Idsubsetageconditionnon-mutated or mutated sequences ( % ) IgM ( % ) IgD ( % ) IgG1 ( % ) IgG3 ( % ) IgG2c ( % ) IgG2b ( % ) IgE ( % ) IgA ( % ) 9867sB-1a6MSPFnon-mutated5095 . 82 . 70 . 60 . 80 . 1mutated50650 . 0316 . 84 . 34 . 89 . 18699sB-1a4MSPFnon-mutated5699 . 50 . 30 . 10 . 1mutated4489 . 93 . 421 . 33 . 49863sB-1a4MSPFnon-mutated74 . 193 . 90 . 23 . 611 . 3mutated25 . 944 . 2419 . 73 . 71 . 413970sB-1a3 . 5MSPFnon-mutated74 . 192 . 70 . 53 . 71 . 71 . 3mutated25 . 992 . 12 . 50 . 54 . 80 . 113342sB-1a4MSPFnon-mutated88 . 997 . 60 . 50 . 80 . 60 . 20 . 3mutated11 . 185 . 20 . 30 . 27 . 3713337sB-1a4MGFnon-mutated69 . 798 . 50 . 10 . 80 . 40 . 1mutated30 . 37915 . 350 . 713003sB-1a4MGFnon-mutated74 . 897 . 20 . 30 . 20 . 50 . 21 . 6mutated25 . 289 . 81 . 12 . 31 . 25 . 613341sB-1a4MGFnon-mutated78 . 2990 . 10 . 20 . 10 . 6mutated21 . 872 . 29 . 65 . 512 . 60 . 113017sB-1a4MGFnon-mutated80 . 995 . 60 . 4211mutated19 . 1790 . 27 . 93 . 98 . 90 . 113002sB-1a4MGFnon-mutated88 . 597 . 40 . 50 . 60 . 21 . 3mutated11 . 563 . 814 . 88 . 413Table lists individual splenic B-1a cell sample sorted from 4-6 month old C57BL6/J mice reared under either specific pathogen free ( SPF ) or germ-free ( GF ) condition . For each sample , the sequences are divided into non-mutated or mutated ( > = 1 nucleotide change ) categories , the frequencies of each category are shown . For each category , the frequencies of sequences expressing each isotype are shown . The data for germ-free animals is discussed at the end of the result section . The splenic B-1a cells that express class-switched Ig still express IgM on the surface , since cells were sorted as IgMhi IgDlo/- dump- CD19+ CD93- CD21-/lo CD23- CD43+ CD5+ . In addition , IgM+ cells described here barely co-express other surface isotypes . Thus the class-switched transcripts are derived from IgM+ cells that apparently have already undergone class switching but have yet to turn off IgM surface protein expression . Since all of the cell preparation , staining and sorting were performed equivalently for all samples , our finding that the class-switched transcripts were selectively and predominantly detected in splenic B-1a cells from 4–6 month old mice argues that the detection of these transcripts is not due to contamination or other technical problems . The microbiota are often thought to participate in shaping the repertoire of ‘natural’ antibodies , which is largely produced by B-1a ( Baumgarth et al . , 2005 ) . Nevertheless , we find that germ-free ( GF ) animals have normal numbers of B-1a cells in spleen ( Figure 9—figure supplement 1 ) . Notably , the splenic B-1a IgH repertoires in age-matched ( 4-5 month old ) specific pathogen free ( SPF ) and GF mice are very similar: 1 ) their IgH repertoires are comparably less diversified and enriched in the recurrent V ( D ) J sequences ( Figures 2A , 9A , Table 7 ) ; 2 ) their VH usage patterns show no significant differences ( Figure 3—figure supplement 1E ) ; 3 ) their CDR3 peptide expressions show a comparable extent of sharing between SPF and GF mice ( Figure 9B ) ; and 4 ) a substantial proportion of V ( D ) J sequences selected in the B-1a IgH repertoire in adult SPF mice are similarly selected in the B-1a IgH repertoire in GF mice ( Table 3 ) . 10 . 7554/eLife . 09083 . 027Figure 9 . The B-1a IgH repertoires from mice raised in specific pathogen free condition are comparable to the B-1a IgH repertoire from age-matched germ-free mice . ( A ) IgH CDR3 tree map plots for splenic B-1a cells from GF mice ( upper panel ) , or SPF mice in Caltech animal facility ( middle panel ) , or SPF mice in Stanford animal facility ( bottom panel ) . Each plot represents the data for a sample from a 4 month old mouse . Recurrent CDR3 ( nucleotide ) sequences are visualized as larger contiguously-colored rectangles in each plot . ( B ) CDR3 peptide pair-wise sharing analysis of IgH repertoire similarity between multiple splenic B-1a samples from age-matched GF and SPF mice . GF mice ( n = 6 ) ; SPF mice ( n = 6 ) . CDR3 peptide pair-wise analysis was conducted between GF mice ( GF/GF ) , SPF mice ( SPF/SPF ) and GF vs . SPF mice ( GF/SPF ) . Each dot represents the percentage of shared CDR3 peptide sequences between two mice . There was no statistical difference between each comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 02710 . 7554/eLife . 09083 . 028Figure 9—figure supplement 1 . Normal splenic B-1a compartment in GF mice . ( A ) FACS plot showing the B-1a population in spleen from SPF or GF mouse . Live dump- CD19 + CD93- IgMhi IgDlo CD23lo/- CD21- cells were gated to reveal CD5 + CD43 + B-1a cells . ( B ) Absolute number of splenic B-1a cells in GF and SPF mice . Each dot represents the data from an individual mouse . There is no significant difference shown between two groups . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 02810 . 7554/eLife . 09083 . 029Table 7 . Top 10 highly recurring CDR3 sequences ( peptide and V ( D ) J recombination ) detected in listed splenic B-1a samples from age-matched SPF and GF mice . DOI: http://dx . doi . org/10 . 7554/eLife . 09083 . 029sB-1a samples ( 4 months ) Top 10 IgH CDR3 sequencesPeptideV ( D ) Jgerm-free #11MRYGSSYWYFDVV11-2 D1-1 J12ARGAYV1-80 J23ARNPDGYYTYYYAMDYV2-2 D2-9 J44ARDPFYYYGSSYWYFDVV5-16 D1-1J15MRYSNYWYFDVV11-2 D2-6 J16AITRAYV1-55 J37ARRYYGSSYWYFDVV1-55 D1-1 J18ARSDYYGSSSLSYV1-26 D1-1 J29ASGGNYFDYV1-75 J210ARSLYNV1-9 J2germ-free #21ARNYGSSYDYV1-53 D1-1 J22TRPSYYGSDYV14-4 D1-1 J23TRESYDGYYVWYAMDYV5-9-1 D2-9 J44ARGDYV14-3 J25ASNWAYV1-53 D4-1 J26MRYSNYWYFDVV11-2 D2-6 J17AKGDYYGSSYYFDYV1-9 D1-1 J28VRHGPRAFDYV10-1 D3-2 J29ARLNGDYV1-69 J210MRYGNYWYFDVV11-2 D2-8 J1specific pathogen free #1 ( from Caltech ) 1ASYSNSDVV3-6 D2-6 J12ARVSYSRAMDYV14-3 D2-6 J43ARSGNYGAMDYV1-7 D2-8 J44ASRLRSTFAYV2-6-8 D1-1 J35ARVTTVHAMDYV1-55 D1-1 J46ARNYGSSYWYFDVV1-53 D1-1 J17ARTPNWEARDYV1-55 D4-1 J48ARRYYGSSYWYFDVV1-55 D1-1 J19ARPLLYRYYFDYV1-75 D2-6 J210ARNYGSSYDWYFDVV1-9 D1-1 J1specific pathogen free #2 ( from Caltech ) 1ARGGIYYDYDEVYYYAMDYV1-55 D2-4 J42MRYSNYWYFDVV11-2 D2-6 J13ARDYYGSSWYFDVV1-26 D1-1 J14MRYGNYWYFDVV11-2 D2-8 J15MRYGSSYWYFDVV11-2 D1-1 J16ARYYDGYYGYYAMDYV1-26 D2-4 J47ALITTWYFDVV1-78 D1-2 J18ARHYYGSSWGYV1-53 D1-1 J29ARSFSPYYFDYV1-26 J210ARSHGYYPFDYV1-54 D2-9 J2specific pathogen free #1 ( from Stanford ) 1ARSADYGGYFDVV1-64 D2-4 J12ARGAYV1-80 J23ARSYYDYPWFAYV1-76 D2-4 J34ARRWLLNAMDYV1-9 D2-9 J45ARPYYYGSSPWFAYV1-69 D1-1 J36ARNDYPYWYFDVV1-4 D2-4 J17ARSGDYV1-64 J28ARVIGDYV1-53 D2-14 J49ARANYV1-55 J310AVNWDYAMDYV1-84 D4-1 J4specific pathogen free #2 ( from Stanford ) 1ARGNYV1-80 J22ARWVYYGSSSYWYFDVV1-54 D1-1 J13ARSSNYAMDYV1-78 D2-11 J44ARYYYGSNYAMDYV7-3 D1-1 J45ARGAYV1-80 J26ARRYYGSSYWYFDVV1-55 D1-1 J17ARSPYYSNYEGYFDVV1-72 D2-6 J18ARKNYGSSYWYFDVV1-55 D1-1 J19ARLEIYYGNYGRVFDVV1-80 D2-8 J210ARRDYYGSSYVLAYV1-9 D1-1 J3Table lists the top 10 highly recurring CDR3 sequences ( peptide and V ( D ) J recombination ) shown in each of CDR3 tree-map plot ( Figure 9A ) . Further , hypermutation occurs equally in the splenic B-1a IgVH in 4–6 month old SPF and GF mice , i . e . , the frequency of mutated sequences and the mutation rate are comparable under two conditions ( Figure 7A , E ) . Indeed , AID introduces mutations into the identical V ( D ) J sequences expressed by splenic B-1a cells from either SPF or GF mice ( Figure 7B , C ) . Finally , similar to SPF mice , AID-mediated class-switch occurs comparably in splenic B-1a cells from GF mice ( Figure 8A ) . Since the V ( D ) J selection , hypermutation and class-switching operate comparably in splenic B-1a from GF and SPF mice , we conclude that the somatic mechanisms that select and diversify B-1a IgH repertoire over time are not driven by microbiota-derived antigens . Nevertheless , the environment has a strong impact on the isotype representation . IgA transcripts are readily detected in splenic B-1a from 4–6 month old SPF mice; however , these transcripts are minimally detected in the splenic B-1a from age-matched GF mice ( Figure 8B , Table 6 ) . This finding is consistent with the recognition that class-switching to IgA is strongly associated with the presence of gut microbiota ( Kroese et al . , 1989; Macpherson et al . , 2000 ) . Studies presented here open a new perspective on the origin and breadth of humoral immunity that protect against invading pathogens and regulate autoimmunity . Recent studies have already shown B-1a develops prior to and independent from BM HSC , which fail to generate B-1a but fully constitute FOB and MZB compartment ( Ghosn et al . , 2012; Yoshimoto et al . , 2011 ) . Cohering the fundamental difference in their development origin , our studies reveal two distinct IgH repertoires that develop at different times and are shaped by distinct functional mechanisms . The first of these repertoires is expressed in B-1a cells . The de novo IgH rearrangements in this repertoire occur mainly during the first few weeks of age and largely cease thereafter . Then B-1a cells persist as a self-replenishing population . The B-1a repertoire , however , continues to evolve under stringent selection . Thus , certain V ( D ) J sequences increase with age , and certain V ( D ) J nucleotide sequences gradually emerge as the predominant recombinations encoding the specific CDR3 peptides in all adults . Furthermore , the age-dependent V ( D ) J selection coincides with the progressive introduction of IgVH mutation and increased class-switch . Importantly , the V ( D ) J selection and AID-mediated diversification occur comparably in germ-free and conventional mice , indicating that these unique repertoire-defining mechanisms are not driven by microbiota-derived antigens . In contrast , MZB , FOB and peritoneal B-2 cells develop later , and continuously develop de novo from BM HSC throughout life and express drastically different IgH repertoire ( s ) . Their IgH repertoires tend to preferentially utilize V1 ( J558 ) family , are far more diverse and less repetitive and , unlike B-1a cells , show no apparent selection for particular V ( D ) J recombination sequences and do not show IgVH mutation and class-switch . In essence , AID introduces SHM and CSR in these B cell subsets only when they respond to their cognate antigens that are largely exogenous in nature . These findings were enabled by employing the amplicon-rescued multiplex PCR technology , which allows the capture and amplification of Ig transcripts from a given B cell population in an inclusive and quantitative fashion . Specifically , the first RT-PCR reaction , which uses an array of gene-specific primers for almost all VH families and all constant ( CH ) genes , is carried out only for a few cycles . The second round of PCR is then carried out with communal primers that recognize the unique sequences tagged into each of the VH and CH primers . Since these ‘tag sequences’ were already introduced during the initial cycles , the use of the communal primers assures that all of the targets are amplified with reduced bias during the following exponential phase of amplification . Coupled with the next generation sequencing , our method is quite robust and allows detection of diverse Ig transcripts that collectively carry about 100 VH genes associated with different isotypes . As with other bulk RNA sequence measurement , our methods cannot determine the absolute number of each Ig transcript in a given B cell population . Hence the actual number of cells expressing a certain Ig sequence is unknown . In addition , our methods do not allow determination of whether certain sequences associated with distinct isotypes belong to the same cell . Further , since the Ig transcript copy number variation among cells is unknown , the frequency of a given Ig transcript is roughly viewed as the relative index of the frequency of cells expressing this Ig transcript . This assumption is generally valid since our studies exclude plasmablast and plasma cells , which do not express surface CD5 . Since B-1a cells are well-known to undergo self-replenishment in adult , the dramatic increase in certain V ( D ) J sequences in the B-1a IgH repertoire over time likely reflects the expansion of cells expressing this particular V ( D ) J sequence . Single cell sequencing analysis has advantages in reducing technical bias and in enabling paired IgH/IgL sequencing . Nevertheless , sequencing costs are still a big hurdle to the large-scale single cell analysis , which , as our studies demonstrate , is necessary to develop a comprehensive view of the various B cell subset repertoires . Therefore , at least for the present , our approaches are more efficient and practical . B-1a produce ‘natural’ antibodies , many of which recognize endogenous ( self ) antigens ( Baumgarth et al . , 2005 ) and play house-keeping roles in clearing the cellular debris or metabolic wastes ( Shaw et al . , 2000; Binder and Silverman , 2005 ) . Since the natural antibodies can also react/cross-react with microorganism-derived antigens , they also participate in the first line of immune defense ( Ochsenbein et al . , 1999; Baumgarth et al . , 2000 ) . Germ-free mice have normal levels of circulating ‘natural’ IgM ( Bos et al . , 1989 ) . Earlier immunologists have postulated that the natural antibody repertoire is selected by endogenous ( self ) antigens ( Jerne , 1971; Coutinho et al . , 1995 ) . Our studies , which demonstrate that B-1a IgH repertoire ( hence the re-activities of natural antibodies ) is highly similar between individual adult C57BL/6 mice , regardless of whether the animals are reared in conventional or germ-free facilities , introduce the solid evidence supporting this argument . Our studies also demonstrate that the B-1a IgH repertoire is selected over time . Thus , recurrent V ( D ) J sequences appear later , and most of the V ( D ) J sequences that are selected to be conserved in all individuals do not emerge until the animals reach the adulthood . As a result , the sequence composition of B-1a IgH repertoire in adult mice becomes much less random than that expressed in neonate and younger mice . Furthermore , the convergent selection of a particular V ( D ) J recombination sequence encoding a specific CDR3 peptide indicates that the selection is strikingly precise and occurs at both the protein and the nucleotide level . Unexpectedly , our studies find that both SHM and CSR participate in diversifying the B-1a IgH repertoire . However , unlike GC response SHM , which occurs within a few days following antigenic stimulation , SHM in B-1a IgVH starts after weaning and is cumulative with age . The progressive increase in the SHM is also associated with increased class switching . Most importantly , SHM and CRS occur comparably in germ-free and conventional mice , indicating that SHM and CSR in the B-1a primary IgH repertoire are not driven by microbiota-derived antigen . Since B-1a cells are well-known to produce anti-self antibodies , stimulation by endogenous antigens is likely the major driving force for the AID-mediated diversification processes . Ongoing SHM in the absence of external antigens influence have been reported in sheep B cells ( Reynaud et al . , 1995 ) . The accumulation of SHM in B-1a IgVH over time likely represents a similar strategy to further diversify their restricted Ig repertoire as animal age . Such diversification may potentiate defenses against newly encountered pathogens . However , the age-dependent accumulative SHM , which is likely driven by self-antigens , may also increase the risk of autoimmune disease due to pathogenic high affinity auto-reactive antibodies . Indeed , deregulated B-1a growth have been reported in NZB/W mice , where autoantibody-associated autoimmune disease develops as animal age ( Hayakawa et al . , 1984 ) . AID-mediated mutagenesis in B-1a IgVH may occasionally introduce mutations elsewhere in the genome that facilitate dysregulated growth and neoplastic transformation , e . g . , B-chronic lymphocytic leukemia ( B-CLL ) ( Stall et al . , 1988; Kipps et al . , 1992; Phillips and Raveche , 1992 ) . Although the mechanism by which the IgM+ splenic B-1a cells from older mice express class-switched Ig transcripts remains elusive , this finding suggests that certain cells are undergoing vigorous genetic alteration that may share the similar mechanisms that underlie the malignant transformation . In fact , cells with simultaneous expression IgM and class-switched Ig transcripts have been reported in B-CLL and other B cell tumors ( Oppezzo et al . , 2002; Kinashi et al . , 1987 ) . The splenic and peritoneal B-1a IgH repertoires show similar characteristics . Both repertoires become more restricted with age with increased recurrent V ( D ) J sequences ( Figure 5—figure supplement 1 ) and retain the positive selected V ( D ) J sequences in adult animals . However , our studies reveal the key repertoire differences between B-1a cells at their two native locations . Although both repertoires show extensive CDR3 sharing among individual mice , the peritoneal B-1a IgH repertoire is more similar to neonatal splenic B-1a repertoire and shows a significantly higher level of CDR3 peptide sharing among individual mice than the splenic B-1a repertoire ( Figure 2B ) . In addition , the peritoneal B-1a IgH repertoire is more biased in using V6-6 ( J606 ) , V9-3 ( Vgam3 . 8 ) , V2-9 ( Q52 ) and V2-6-8 ( Q52 ) , which are preferentially expressed in splenic B-1a from neonate and younger mice . These findings suggest that peritoneal B-1a cells are enriched for cells that are generated during neonatal and young age of life , thus are largely consist of cells migrated from spleen into PerC when the animals were younger . This argument is further supported by the findings that the frequencies of mutated sequences in the peritoneal B-1a cells from 4-6 month old mice are substantially lower and the mutations are mainly single nucleotide changes whereas a proportion of IgH sequences with multiple mutations is detected in splenic B-1a cells from the same aged mice ( Figure 7 ) . MZB and B-1a share many phenotypic and functional characteristics ( Martin and Kearney , 2001 ) . Our studies show that the MZB IgH repertoire differs drastically from the B-1a IgH repertoire , but is very similar to the repertoires expressed by splenic FOB and peritoneal B-2 . Since MZB and FOB cells are mainly derived from BM HSC ( Ghosn et al . , 2012 ) , there findings collectively support the idea that these B cells belong to the same ( i . e . , B-2 ) developmental lineage . Nevertheless , the MZB repertoires from individual mice contain substantially higher levels of common CDR3 sequences ( peptides ) than the splenic FOB and peritoneal B-2 repertoires ( Figure 2B ) . Years ago , we postulated that B-1a and B-2 B cells belong to distinct developmental lineages that are evolved sequentially to play complementary roles in immunity ( Herzenberg and Herzenberg , 1989 ) . The sequence data presented here , which reveal the key distinctions in the repertoires as well as the repertoire-defining mechanisms between B-1a and B-2 subsets , support this argument and greatly extend our earlier version . These key distinctions provide the genetic bases for their well-known fundamental functional difference between B-1a and other B subsets . In particular , they are central to vaccine development , where the recognition that the B cells have distinct targeting antibody repertoires clearly invites attention . In addition , our findings offer insights in understanding the origins and behaviors of B cell neoplasms , particularly B-CLL , and the autoimmune diseases in which over production of autoantibodies is implicated in the pathology .
C57BL/6J mice were purchased from the Jackson Laboratory . AID-deficient C57BL6/J mice were kindly provided by Dr . Michel Nussenzweig ( Rockefeller University ) . Mice were breed and kept in the Herzenberg laboratory colony under SPF conditions at the Stanford Veterinary Service Center ( VSC ) . Spleens from germ-free C57BL6/J mice were provided by Dr . Sarkis Mazmanian ( Caltech ) . Germ-free mice were maintained in sterile Trexler isolators and fed autoclaved food and water . Germ-free status was assayed monthly by aerobic and anaerobic plating; and by 16s rRNA PCR . Study protocols were approved by the Stanford VSC . FACS staining has been previously described ( Yang et al . , 2012 ) . Briefly , cell suspensions were incubated with LIVE/DEAD Aqua ( Life Technologies , San Diego , CA ) , washed , and incubated with unconjugated anti-CD16/CD32 ( FcRII/III ) mAb to block Fc-receptors . Cells were then stained on ice for 20 min . with a ‘cocktail’ of fluorochrome-conjugated antibodies including: anti-CD21-FITC ( Becton Dickenson , San Jose , CA ) , anti-CD43-PE ( BD ) , anti-CD5-PE-Cy5 ( BD ) , anti-CD19-PE-Cy5 . 5 ( Life Technologies ) , anti-CD93 ( AA41 ) -PE-Cy7 ( eBioscience , San Diego , CA ) , anti-B220-APC ( BD ) , anti-IgM-Alexa700 ( Herzenberg lab ) , anti-IgD-APC-Cy7 ( BioLegend , San Diego , CA ) , anti-CD23-Biotin ( BD ) , anti-CD11b-PB ( Life Technologies ) , anti-Gr-1-PB ( Life Technologies ) , anti-TCRαβ-PB ( Life Technologies ) , anti-CD11c-PB ( Life Technologies ) , and anti-CD3 ? -PB ( Life Technologies ) . After washing , cells were stained with Streptavidin-Qdot 605 ( Life Technologies ) . Cells were sorted on FACS Aria ( BD ) at the Stanford Shared FACS Facility . Sorting purity was greater than 99% . Five types of B cell populations were sorted based on tissue and phenotype: splenic and peritoneal B-1a cells ( dump- CD19+ CD93- IgMhi IgD-/lo CD21-/lo CD23- CD43+ CD5+ ) ; splenic FOB and peritoneal B-2 cells ( dump- CD19+ CD93- IgMlo IgDhi CD23+ CD43- CD5- ) ; splenic MZB cells ( dump- CD19+ CD93- IgMhi IgD-/lo CD21hi CD23lo/- CD43- CD5- ) . 1-2 x 104 cells for each cell population were sorted directly into 0 . 5 mL Trizol LS ( Life Technologies ) . RNA was extracted according to the protocol provided by Trizol LS ( Life Technologies ) . RT-PCR reactions were conducted using a set of sequence specific primers covering almost all of mouse VH genes ( forward primers ) and constant CH primers covering all of isotypes ( reverse primers ) . Illumina paired-end sequencing communal primer B is linked to each forward VH primer . Illumina paired-end sequencing communal primer A and a barcode sequence of 6 nucleotides are linked to each reverse CH primers . In brief , cDNA was reverse transcribed from total RNA sample using mixture of forward VH and reverse CH primers and reagents from the OneStep RT-PCR kit ( Qiagen , Valencia , CA ) . The first round of PCR was performed at: 50°C , 40 minutes; 95°C , 15 min; 94°C , 30 s , 58°C , 2 min , 72°C , 30 s , for 15 cycles; 94°C , 30 s , 72°C , 2 min , for 10 cycles; 72°C , 10 min . After the first round of PCR , primers were removed by Exonuclease I digestion at 37°C for 30 min ( New England Biolabs , lpswich , MA ) . Then 2 μL of the first-round PCR products were used as templates for the second round of amplification using communal A and B primers and reagents from the Multiplex PCR kit ( Qiagen ) . The second round PCR was performed as: 95°C , 15 min; 94°C , 30 s , 55°C , 30 s , 72°C , 30 s , for 40 cycles; 72°C , 5 min . About 400bp long PCR products were run on 2% agarose gels and purified using a gel extraction kit ( Qiagen ) . The IgH libraries were pooled and sequenced with Illumina MiSeq pair-end read-length platform . The output of IgH sequence covers CDR2 , CDR3 and the beginning of the constant region . The sequence information for all primers used for the library preparation can be found in US Patent Office ( US9012148 ) . Sequence reads were de-multiplexed according to barcode sequences at the 5’ end of reads from the IgH constant region . Reads were then trimmed according to their base qualities with a 2-base sliding window , if either quality value in this window is lower than 20 , this sequence stretches from the window to 3’ end were trimmed out from the original read . Trimmed pair-end reads were joined together through overlapping alignment with a modified Needleman-Wunsch algorithm . If paired forward and reverse reads in the overlapping region were not perfectly matched , both forward and reverse reads were thrown out without further consideration . The merged reads were mapped using a Smith-Waterman algorithm to germline V , D , J and C reference sequences downloaded from the IMGT web site ( Lefranc , 2003 ) . To define the CDR3 region , the position of CDR3 boundaries of reference sequences from the IMGT database were migrated onto reads through mapping results and the resulting CDR3 regions were extracted and translated into amino acids . C57BL/6J mouse VH reference sequences were pair-wise aligned with a Smith-Waterman algorithm . Two VH reference sequences are considered related if the aligned region between them is > 200bp matched and < 6 mismatches . Two sequence reads were considered related if the best mapped VH sequences are related and the CDR3 segments have less than 1 mismatch . If two sequences are related and the frequency of the minor one is less than 5% of the dominant one , the minor one is removed from further consideration . In addition , single copy CDR3s are removed from further consideration . To allow multiplexing of multiple samples in a single sequence run , CH primers were linked with barcodes containing 6 different nucleotides . The barcode CH primers were used in a first round RT-PCR . To compensate for potential in chemical synthetic , PCR and/or sequencing error , barcodes were designed with a Hamming distance ≥3 . Given that the chemical synthetic error is roughly 5% per position , there is about a 1/8000 chance that one barcode is mistakenly synthesized as another barcode . For a CDR3 with n occurrences in one sample and the same CDR3 ( nucleotide sequence ) with N occurrences in another sample in the same sequencing run , we calculated the probability that such a CDR3 would occur n or more times if it were due to cross-contamination , using the following formula P = 1 - ∑k=1n-1 e-λ · λkk ! where λ is the expected number of errors given N reads and is computed by λ = N · μ and μ is the cross-contamination rate which is preset as 1/8000 . CDR3s that yielded p<0 . 001 were considered highly unlikely to be due to cross-contamination . Sequences were obtained for 60 separately sorted cell populations ( details for each population are in Table 1 ) . To draw the IgH CDR3 tree-map for each sequence sample , the entire rectangle was divided: 1st into a set of rectangles with each rectangle corresponding to a distinct VH gene segment; 2nd into a set of V-J rectangles with each rectangle corresponding to a distinct V-J; and 3rd into a set of V-J-CDR3 rectangles with each rectangle representing a distinct V-J-CDR3 combination . The rectangles are ordered based on area from largest at the bottom right to smallest at the top left . The size of an individual rectangle is proportional to the relative frequency for each V-J-CDR3 combination sequence . In order to distinguish neighboring rectangles , corners of each rectangle are rounded and each rectangles are colored randomly . Therefore , each rectangle drawn in the map represents an individual CDR3 nucleotide sequence . D50 is a measurement of the diversity of an immune repertoire of J individuals ( total number of CDR3s ) composed of S distinct CDR3s in a ranked dominance configuration , where r1 is the abundance of the most abundant CDR3 , r2 is the abundance of the second most abundant CDR3 , and so on . C is the minimum number of distinct CDR3s with > = 50% total sequencing reads . D50 is given byAssume that r1≥r2 ⋯ ≥ri ⋯ ≥ri+1 ⋯ ≥rs⏟s , ∑i=1s ri=Jif ∑i=1cri ≥ J/2 and ∑i=1C-1ri< J/2D50 = CS × 100 The forward VH primers used to amplify expressed IgH genes are located at the IgH framework region 2 . To avoid primers interfering with the mutation analysis , the variable region stretching from the beginning of the CDR2 to the beginning to the CDR3 was examined for mismatches between the sequence read and the best-aligned germline reference sequence . To eliminate the impact of sequencing error on this calculation , only sequence reads with more than 4 copies were included in the mutation calculation . For this measurement , we introduce an entropy value as the index of diversity level . Assuming a distinct CDR3 peptide sequence X in a sample is derived from n number of distinct V ( D ) J recombinations ( nucleotide ) with each frequency as P1 , P2 , … Pn respectively , the entropy for X ( Ex ) is then calculated as: Ex = -∑i=1n Pi log2 Pi . For a sample , after computing entropy values for each distinct peptide CDR3 fragments , the E values for distinct peptide CDR3 fragments are categorized into four ranges: [0 , 0 . 5 ) , [0 . 5 , 1 . 5 ) , [1 . 5 , 2 . 5 ) and [2 . 5 , + ∞ ) . The higher the entropy value , the more diverse the V ( D ) J recombinations for a given CDR3 peptide . | Our immune system protects us by recognizing and destroying invading viruses , bacteria and other microbes . B cells are immune cells that produce protective proteins called antibodies to stop infections . These cells are activated by ‘antigens’ , which are fragments of molecules from the microbes or from our own cells . When an antigen binds to a B cell , the cell matures , multiplies and produces proteins called antibodies . These antibodies can bind to the antigen , which marks the microbe for attack and removal by other cells in the immune system . Each antibody consists of two ‘heavy chain’ and two ‘light chain’ proteins . B cells are able to produce a large variety of different antibodies due to the rearrangement of the gene segments that encode the heavy and light chains . In mice , there are two kinds of B cells – known as B-1a and B-2 cells – that play different roles in immune responses . B-1a cells have long been known to produce the ‘natural’ antibodies that are present in the blood prior to an infection . On the other hand , B-2 cells produce antibodies that are specifically stimulated by an infection and are better adapted to fighting it . Previous studies have shown that both types of antibodies are required to allow animals to successfully fight the flu virus . Here , Yang , Wang et al . used a technique called fluorescence-activated cell sorting ( or FACS ) and carried out extensive genomic sequencing to study how the B-1a and B-2 populations rearrange their genes to produce heavy chains . This approach made it possible to separate the different types of B cells and then sequence the gene for the heavy chain within the individual cells . The experiments show that the “repertoire” of heavy chains in the antibodies of the B-1a cells is much less random and more repetitive than that of B-2 populations . Furthermore , Yang , Wang et al . show that B-1a cells produce and maintain their repertoire of heavy chains in a different way to other B-2 populations . B-1a cells develop earlier and the major genetic rearrangements in the gene that encodes the heavy chain occur within the first few weeks of life . Although the gene rearrangements have mostly stopped by adulthood , the B-1a antibody repertoire continues to evolve profoundly as the B-1a cells divide over the life of the animal . On the other hand , the gene rearrangements that make the heavy chains in the B-2 cells continue throughout the life of the animal to produce the wider repertoire of antibodies found in these cells . In addition , the processes that continue to change the antibody reperotire in the B-1a cells during adulthood do not occur in the B-2 populations . Importantly , the these reperotire-changing processes in B-1a cells also occur in mice that have been raised in germ-free conditions , which demonstrates that – unlike other B cells – the repertoire of heavy chains in B-1a cells is not influenced by antigens from microbes . Instead , it is mainly driven by antigens that are expressed by normal cells in the body . These findings open the way to future work aimed at understanding how B-1a cells help to protect us against infection , and their role in autoimmune diseases , where immune cells attack the body’s own healthy cells . | [
"Abstract",
"Introduction",
"Results",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2015 | Distinct mechanisms define murine B cell lineage immunoglobulin heavy chain (IgH) repertoires |
Cut-and-paste DNA transposons of the mariner/Tc1 family are useful tools for genome engineering and are inserted specifically at TA target sites . A crystal structure of the mariner transposase Mos1 ( derived from Drosophila mauritiana ) , in complex with transposon ends covalently joined to target DNA , portrays the transposition machinery after DNA integration . It reveals severe distortion of target DNA and flipping of the target adenines into extra-helical positions . Fluorescence experiments confirm dynamic base flipping in solution . Transposase residues W159 , R186 , F187 and K190 stabilise the target DNA distortions and are required for efficient transposon integration and transposition in vitro . Transposase recognises the flipped target adenines via base-specific interactions with backbone atoms , offering a molecular basis for TA target sequence selection . Our results will provide a template for re-designing mariner/Tc1 transposases with modified target specificities .
Transposable elements are ubiquitous in most genomes and promote evolution by generating genetic diversity ( Biémont and Vieira , 2006 ) . Invading transposons can alter genes , affect gene expression or spread drug resistance in bacteria . As shuffling of DNA by transposition can be deleterious , transposons often become inactivated or transcriptionally silenced . Conversely , transposons can give rise to new , useful cell functions . For example , domestication of a Transib-type DNA transposon led to V ( D ) J recombination and adaptive immunity in jawed vertebrates ( Kapitonov and Jurka , 2005 ) . Similarly , the Cas1 integrase component of prokaryotic CRISPR-Cas systems of adaptive immunity originated from DNA transposons named Casposons ( Krupovic et al . , 2014 ) . Integration of spacer sequences into the CRISPR locus by the Cas1-Cas2 complex has similarities with transposon and retroviral DNA integration ( Nuñez et al . , 2015 ) . DNA transposons move from one genomic location to another using transposon-encoded recombinases , often by a DNA cut-and-paste mechanism . Many DNA transposases ( e . g . Mos1 , Tn5 and bacteriophage MuA ) share a conserved RNase H-like catalytic domain , along with retroviral integrases ( e . g . HIV-1 ) and RAG recombinases . These DDE/D enzymes use common active site chemistry to perform similar DNA cleavage and DNA integration reactions . By contrast , there is wide diversity in their preferred target integration sites . Most DDE/D recombinases show only limited preference for a consensus target DNA sequence , which is usually palindromic ( Goryshin et al . , 1998; Halling and Kleckner , 1982 ) . The number of base pairs separating the integration sites on complementary DNA strands also varies , from 2 to 9 . Some retroviral integrases ( e . g . prototype foamy virus ( PFV ) and HIV-1 ) preferentially insert their viral genome into nucleosomal DNA ( Pruss et al . , 1994; Maskell et al . , 2015 ) . Similarly , some transposases ( e . g . Tn10 ) favour bent target DNA structures ( Pribil and Haniford , 2003 ) . In other transposition systems ( e . g . IS21 , Mu ) , an element-encoded accessory ATPase facilitates strand transfer ( Mizuno et al . , 2013; Arias-Palomo and Berger , 2015 ) ; and can prevent self-destructive insertion of the transposon into its own sequence ( target immunity ) , ( Mizuno et al . , 2013 ) . Despite this biochemical knowledge , the molecular and structural origins for transposon target specificities remain unknown . Mariner/Tc1/IS630 family transposases are unusual as they integrate their transposons , with a 2 bp stagger , strictly at TA target sequences ( Tellier et al . , 2015 ) . They are widespread in nature and are used as tools for genome engineering and therapeutic applications . For example , the reconstructed Tc1 transposase Sleeping Beauty ( Ivics et al . , 1997 ) is being used in human clinical trials to treat B-cell lymphoma by genetic engineering of T cells ( Maiti et al . , 2013 ) and in pre-clinical studies to reduce age-related macular degeneration ( Johnen et al . , 2012 ) . Up to 45 kb of DNA can be inserted into the C . elegans genome using a transposition system engineered from the mariner transposon Mos1 from Drosophila mauritiana ( Frøkjær-Jensen et al . , 2014 ) . The ability to pre-select specific sites for integration , beyond the requisite TA , may be desirable for certain genome engineering applications , e . g . controlled genomic integration of a therapeutic gene . Such targeted transposition has been achieved for the bacterial transposase ISY100 using a C-terminal Zif268 DNA-binding domain fusion ( Feng et al . , 2010 ) ; and for Sleeping Beauty transposase either by fusing it with a targeting domain ( Yant et al . , 2007 ) or by exploiting interactions with a targeting protein ( Ivics et al . , 2007 ) . Conversely , it may be useful to randomise mariner/Tc1 integrations; for example in whole genome sequencing applications as an alternative to Tn5 ( Amini et al . , 2014 ) . Understanding in molecular detail how mariner/Tc1 transposons are integrated at TA target sites will aid development of these elements as genome engineering tools . The wealth of structural and biochemical data for the naturally active , eukaryotic transposon Mos1 offers a paradigm for determining the molecular mechanism of mariner/Tc1 transposon integration . The 1286 bp transposon is framed by 28 bp imperfect inverted repeats ( IR ) ( Jacobson et al . , 1986 ) and encodes a 345 amino acid transposase that can perform cut-and-paste DNA transposition in vitro ( Lampe et al . , 1996 ) , as shown in Figure 1a . The Mos1 transposase homodimer binds to the IR at one transposon end ( Cuypers et al . , 2013 ) and then captures the other IR , forming a paired-end complex ( PEC ) . The trans architecture of the PEC regulates coordinated excision of the transposon ends ( Richardson et al . , 2009 ) and cross-talk between transposase sub-units ( Bouuaert et al . , 2014; Dornan et al . , 2015 ) . After excision , the Mos1 transpososome locates a TA target integration site ( Pflieger et al . , 2014 ) and , upon binding , forms a target capture complex ( TCC ) ( Figure 1a ) . Attack by the 3'-OH at each transposon end on the phosphodiester 5' of the TA dinucleotide joins the excised transposon to the target site , in the DNA strand transfer reaction ( Figure 1a , b ) . The DNA product of transposition , which contains a gap at each transposon end , is bound to the transposase in a strand transfer complex ( STC ) . 10 . 7554/eLife . 15537 . 003Figure 1 . Mos1 transposition . ( a ) Schematic of pathway and complexes formed . Each transposon end has a 28 bp IR sequence ( triangle ) flanked by the TA target site duplication . First and second strand cleavages ( scissors ) are staggered by three bp and generate a 5' phosphate ( filled circle ) on the non-transferred strand ( NTS ) , 3 bases within the IR , and a 3'OH ( arrow ) at the transferred strand ( TS ) end , respectively . After target DNA capture , the transposon 3' ends integrate at a symmetrical TA sequence , resulting in a 5 nt gap . Gap repair duplicates the TA . ( b ) Mos1 strand transfer . The transposon 3'OHs attack the phosphodiester bond between T0 and C-1 on both the top ( t , black ) and bottom ( b , magenta ) target DNA strands , joining each TS to target DNA , separating the TA base pairs , and leaving a 3'OH at C-1 . ( c ) Sequence and numbering of the DNA used to crystallise the STC; see also Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 003 To establish how mariner/Tc1 transposases recognize and integrate transposon DNA at a TA dinucleotide , we determined a crystal structure of the Mos1 STC . In this snapshot of the transposition machinery after DNA transposon integration , the target DNA is distorted from B-form and the target adenines are flipped into extra-helical positions . Base-flipping is confirmed in solution by time-resolved fluorescence of strand transfer complexes in which the target adenines are substituted by 2-aminopurine . Adenine-specific interactions , between the flipped adenine bases and transposase backbone atoms , provide a molecular basis for recognition of the TA target sequence . Interactions with Mos1 transposase residues W159 , R186 , F187 and K190 , which are essential for transposon integration in vitro , stabilise distortions in the target DNA . Conservation of key residues involved in stabilising the target DNA distortions suggests this mechanism may also occur with other mariner/Tc1 family transposons .
To assemble the Mos1 STC , full length T216A Mos1 transposase was combined , in a 1:1 molar ratio , with DNA representing the product of transposon integration ( Figure 1c ) . This DNA contains the transposon IR joined at its 3' end to an unpaired TA dinucleotide and target DNA ( Table 1 ) . The bottom target DNA strand ( strand b , magenta , Figure 1c ) has a cohesive 4 nt 5' overhang ( sequence GGCC ) to promote interactions between adjacent complexes in the crystal lattice . This approach , of assembling the STC using the strand transfer product , and bypassing catalysis of integration , proved successful for the preparation of bona fide PFV strand transfer complexes ( Yin et al . , 2012 ) . 10 . 7554/eLife . 15537 . 004Table 1 . Sequences of oligonucleotides used in the crystallisation , target integration and fluorescence experiments . The target TA dinucleotide ( and its variants ) are highlighted in bold . The adenine analogue 2-aminopurine is denoted P and 2 , 6-diaminopurine is D; the thymine analogue 2-thio-thymine , is indicated by S . IR700 indicates the 5' addition of the infrared fluorescent dye 700 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 004NameSequenceLength ( nt ) Crystallisation of STCTS5' AAA CGA CAT TTC ATA CTT GTA CAC CTG ATA GCA GTG36NTS5' GGT GTA CAA GTA TGA AAT GTC GTT T25target DNA5' GGC CCA CTG C10Target Integration AssaysTS IR DNA5' AAA CGA CAT TTC ATA CTT GTA CAC CTG A28TS 5' labelled IR DNA5' IR700 / AAA CGA CAT TTC ATA CTT GTA CAC CTG A28NTS IR DNA5' GGT GTA CAA GTA TGA AAT GTC GTT T25TA target DNA ( top strand ) 5' AGC AGT GCA CTA GTG CAC GAC CGT TCA AAG CTT CGG AAC GGG ACA CTG TT50TA target DNA ( bottom strand ) 5' AAC AGT GTC CCG TTC CGA AGC TTT GAA CGG TCG TGC ACT AGT GCA CTG CT50TP target DNA ( top strand ) 5' AGC AGT GCA CTP GTG CAC GAC CGT TCA AAG CTT CGG AAC GGG ACA CTG TT50TP target DNA ( bottom strand ) 5' AAC AGT GTC CCG TTC CGA AGC TTT GAA CGG TCG TGC ACT PGT GCA CTG CT50TD target DNA ( top strand ) 5' AGC AGT GCA CTD GTG CAC GAC CGT TCA AAG CTT CGG AAC GGG ACA CTG TT50TD target DNA ( bottom strand ) 5' AAC AGT GTC CCG TTC CGA AGC TTT GAA CGG TCG TGC ACT DGT GCA CTG CT50SD target DNA ( top strand ) 5' AGC AGT GCA CSD GTG CAC GAC CGT TCA AAG CTT CGG AAC GGG ACA CTG TT50SD target DNA ( bottom strand ) 5' AAC AGT GTC CCG TTC CGA AGC TTT GAA CGG TCG TGC ACS DGT GCA CTG CT50Fluorescence experimentsTS_P15' AAA CGA CAT TTC ATA CTT GTA CAC CTG AtP gca gtg gac gta ggc c46TS_P135' AAA CGA CAT TTC ATA CTT GTA CAC CTG Ata gca gtg gac gtP ggc c46TS_A15' AAA CGA CAT TTC ATA CTT GTA CAC CTG Ata gca gtg gac gta ggc c46NTS5' GGT GTA CAA GTA TGA AAT GTC GTT T25Target_165' g gcc tac gtc cac tgc16 Mos1 STC crystals diffracted X-rays to a maximum resolution of 3 . 3 Å . Crystallographic phases were determined by molecular replacement ( Materials and methods ) . The difference electron density after molecular replacement and before model building is shown in Figure 2—figure supplement 1 . The crystallographic asymmetric unit contains one Mos1 STC and , as predicted , base pairing of the 4 nt overhangs in adjacent complexes facilitates crystal packing ( Figure 2—figure supplement 2 ) . The refined model has an R ( free ) of 27 . 9% and good stereochemistry . The X-ray diffraction and refinement statistics are shown in Table 2 . 10 . 7554/eLife . 15537 . 005Table 2 . X-ray diffraction and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 005CrystalMos1 Strand transfer complexPDB ID5HOOSpace groupC121Cell dimensionsa = 256 . 3 Å b = 58 . 9 Å c = 110 . 2 Å α = 90 . 0° , β = 94 . 9° , γ = 90 . 0°Wavelength ( Å ) 0 . 9795Average mosaicity0 . 22OverallOuter shellResolution ( Å ) 86 . 99–3 . 293 . 52–3 . 29Rsymm0 . 0770 . 152Total observations7835814630Unique observations252014479< I>/σ<I>8 . 13 . 3Correlation CC0 . 9270 . 996Completeness ( % ) 99 . 699 . 5Multiplicity3 . 13 . 3Rwork0 . 243Rfree ( 5 . 21% of reflections ) 0 . 279R . m . s . deviations: Bond Length ( Å ) Bond Angle ( deg ) Chiral volume ( Å ) 0 . 0077 1 . 2072 0 . 0785Average B factor ( Å2 ) 74 . 0Ramachandran plot: Core ( % ) Allowed ( % ) Outliers ( % ) 90 . 8 9 . 2 0 The refined Mos1 STC crystal structure ( Figure 2a ) contains a transposase homodimer bound to two DNA duplexes representing the products of transposon integration . Target DNA binds in a channel between the two catalytic domains and the active sites contain the strand transfer products . As the TCC also contains a transposase dimer ( Pflieger et al . , 2014 ) , our new STC structure indicates that Mos1 strand transfer , like transposon excision , is catalysed by a transposase dimer . 10 . 7554/eLife . 15537 . 006Figure 2 . Architecture of the Mos1 strand transfer complex . ( a ) Structure of the STC , with transposase subunits ( orange and blue ) , IR DNA ( orange and green ) and target DNA ( magenta and black ) . Figure 2—figure supplement 1 shows the crystal packing arrangement . ( b ) Schematic of the Mos1 STC . See Figure 2—figure supplement 2 for details of transposase DNA interactions . ( c ) DNA components of the STC: target DNA is bent and each IR TS connects at the 3' end to a target DNA strand . ( d ) The active site of catalytic domain B , showing the product of strand transfer into the bottom target strand ( magenta ) . The simulated annealing composite omit 2Fo-Fc electron density map ( grey mesh ) is contoured at 1 . 2σ . The single Mg2+ is coordinated by D249 , D156 and the 3'OH of C-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 00610 . 7554/eLife . 15537 . 007Figure 2—figure supplement 1 . Stereo views of the difference electron density after molecular replacement . The Fo-Fc electron density ( at 2 . 3σ ) is plotted as a pink mesh with the molecular replacement model: the Mos1 PEC structure ( 3HOS , chains A to F ) . ( a ) Full view of one Mos1 PEC molecule , and ( b ) close-up view of the catalytic domains and the TSs . Transposase subunits ( chains A and B ) are shown as ribbons and the IR DNA duplexes ( chains C to F ) as ladders . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 00710 . 7554/eLife . 15537 . 008Figure 2—figure supplement 2 . Packing arrangement and DNA interactions in the Mos1 STC crystal lattice . Four copies of the Mos1 STC are shown . The 5' end of each target DNA strand has a 4 nt overhang , with the self-complementary sequence GGCC , which base pairs with a symmetry related overhang in an adjacent STC molecule . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 00810 . 7554/eLife . 15537 . 009Figure 2—figure supplement 3 . Schematic depiction of the interactions between transposase and DNA in the Mos1 STC structure . The IR DNA is shown in blue , and the target DNA shown in black ( top strand ) and magenta ( bottom strand ) . Lines connect the detailed description of the interaction to either a circle ( denoting a backbone phosphate ) or the central rectangle ( representing a base ) . Transposase interactions with the backbone phosphates of nucleotides surrounding the TA target sequence support the target DNA conformation , and include contacts between the backbone amides of Y276 and N250 and C-1; Y276 OH and G-2; T213 HG1 and A4; A216 NH and G5 and R257 NH and T6 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 009 The transposase subunits adopt a crossed ( or trans ) arrangement with IR DNA in the Mos1 STC , similar to the pre- and post-TS cleavage Mos1 PECs ( Dornan et al . , 2015; Richardson et al . , 2009 ) : each IR is recognised by the DNA-binding domain of one transposase subunit and by the catalytic domain of the other subunit ( Figure 2b ) , and vice versa . This brings the cleaved transposon ends together , and ensures suitable spacing for their integration into TA target DNA . Transposase interactions with IR DNA in the STC ( Figure 2—figure supplement 3 ) are similar to those in the post-TS cleavage PEC , suggesting that , after transposon excision , the transpososome is poised for target capture . Thus , rather than rearrange the transposase and IR DNA , target DNA is deformed to enable strand transfer . The target DNA is severely distorted from B-form conformation ( Figure 2c ) : the backbone is bent by 147° , with the apex of the bend at the TA target dinucleotide . DNA unwinds most readily at TA sequences due to the inherent bendability of a pyrimidine-purine step ( which has minimal base-to-base overlap and low twist angles ) and the lower stability of a TA base pair , compared to CG . The strand transfer reaction creates a nick in both target DNA strands 5' of the target thymine T0 , which likely relieves steric constraints and allows the extreme bend across the TA di-nucleotide . Transposase interactions with the backbone phosphates of target nucleotides surrounding the TA sequence support this conformation ( Figure 2—figure supplement 3 ) . The transposase performs three nucleophilic substitution reactions at each transposon end: sequential hydrolysis of both DNA strands to excise the transposon , followed by strand transfer to join the IR to target DNA ( Figure 1a ) . One IR is transferred to the top strand ( t , black , Figure 1b ) , and the other to the bottom strand ( b , magenta , Figure 1b ) . In-line SN2 attack by each transposon 3'-OH on the scissile T0 target DNA phosphate ( Figure 1c ) creates a new bond between the transposon end and the target thymine ( T0 ) . At the same time , the phosphodiester linking C-1 and T0 is broken , leaving a 3'-OH on C-1 and inverting the stereochemistry of the scissile T0 phosphate . Each Mos1 transposase active site comprises the carboxylate side-chains of three conserved aspartates ( D156 , D249 and D284 ) from the same catalytic domain , which coordinate the metal ions ( Mg2+ or Mn2+ ) required for catalysis . One Mg2+ was observed in each active site in the Mos1 STC , coordinated by the D156 and D249 carboxylates , the 3'-OH of C-1 and a water molecule . The phosphodiester joining each transposon 3' end ( A56 ) to a T0 passes close to an active site ( Figure 2b , c , d ) . The T0 phosphate oxygens are 4 . 4 Å and 7 Å from the Mg2+ , precluding chelation . Moreover , the C-1 3'-OH is not in-line with the T0–A56 phosphodiester bond , consistent with repositioning of the nascent transposon-target DNA junction , away from the active site Mg2+ after strand transfer . Similar to the PFV STC ( Maertens et al . , 2010 ) , this likely prevents self-destructive disintegration and drives transposition forwards . Each transposon–target thymine junction is clearly defined in the electron density map ( Figure 2d ) . There is also clear density for the nucleobase of T0 on strand b ( magenta ) in active site B . However , we observed no clear density for the T0 base on strand t ( active site A ) , indicating some disorder in its position . Therefore this nucleotide was built as abasic . In active site B the T0 nucleobase π–π stacks with the base of C-1 to which T0 would have been connected before strand transfer . The T0 base is unpaired and O4 is 3 . 4 Å from the H122 imidazole NH , suggesting a possible base-specific hydrogen bond ( Figure 3a ) . However , the mutation H122A had no effect on the strand transfer efficiency ( Figure 3—figure supplement 1 ) , and we conclude that the putative thymine-specific hydrogen bond is not required for target integration and may be transient , due to T0 base mobility . 10 . 7554/eLife . 15537 . 010Figure 3 . Dynamic base flipping of the target adenines . ( a ) Target DNA binding in the Mos1 STC , showing the flipped A1 conformation . The unpaired T0 base stacks with the C-1 base of the same strand . See Figure 3—figure supplement 1 for the effect on strand transfer activity of the mutation H122A . ( b ) Schematic of the TA1 DNA duplex and gel filtration chromatograms of Mos1 transposase ( red ) , TA1 ( blue ) and the STC ( black ) . UV absorbance at 280 nm ( solid line ) and 260 nm ( dotted line ) . ( c and d ) Fluorescence spectroscopy of the 2AP-labelled DNA oligonucleotides TP13 and TP1 , shown schematically in ( c ) and ( d ) respectively . The A-factor ( fractional population ) and lifetime of each of the four fluorescence decay components are plotted for TP13 and TP1 alone ( black circles and lines ) and in the presence of Mos1 transposase ( red triangles and lines ) ; and tabulated in Figure 3—source data 1 . The steady-state fluorescence emission spectra are inset in each case . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 01010 . 7554/eLife . 15537 . 011Figure 3—source data 1 . Fluorescence decay parameters for 2AP-containing duplexes , TP13 and TP1 , in the absence and presence of Mos1 transposase . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 01110 . 7554/eLife . 15537 . 012Figure 3—figure supplement 1 . Strand transfer assay comparing the activity of T216A and H122A/T216A Mos1 transposases . ( a ) Denaturing PAGE of the strand transfer reaction products . Lanes 1 and 6 contain markers; lane 2 is without transposase; lane 3 has no target DNA , but integration occurs at the two TA dinucleotides within the IR DNA sequence . ( b ) Quantification of the 40 nt and 68 nt strand transfer products for each mutant transposase , as a percentage of total DNA . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 012 The most striking feature of the Mos1 STC structure is flipping of both target adenine ( A1 ) bases of the symmetrical TA sequence into extra-helical positions ( Figure 3a ) . The phosphate backbone atoms of A1 and G2 are rotated by ~180° , with respect to the adjacent nucleotides , so that the A1 bases cannot pair with their complementary T0 . Instead each unpaired A1 is wedged against the ribose face of the complementary target strand , at an oblique angle to bases G2 and C3 ( Figure 3a ) . The aromatic ring of each F187 occupies the space vacated by a flipped A1 , forming a π–π stack with the adjacent G2 nucleobase , stabilising this conformation ( Figure 3a ) . To confirm A1 flipping in solution , and to investigate the extent and dynamics of this distortion , we performed fluorescence experiments with DNA containing the adenine analogue 2-aminopurine ( 2AP ) . 2AP is an exquisitely sensitive probe of chemical environment: its fluorescence is highly quenched by stacking with the DNA bases and , hence , is sensitive to local duplex structure and enzyme-induced distortion of that structure ( Jones and Neely , 2015 ) . We designed three DNA duplexes , each mimicking the strand transfer product: TA1 , an unlabelled control analogous to the oligonucleotide used for STC crystallisation ( Figure 3b ) ; TP13 , a labelled control , with 2AP in place of A13 on the top target strand ( black , Figure 3c ) , where it is base-paired and stacked in duplex DNA; TP1 , with 2AP in place of A1 ( Figure 3d ) , where the unpaired 2AP is stacked with the adjacent T0 and G2 bases . Upon addition of Mos1 transposase to TA1 , we observed complete formation of a nucleoprotein complex by gel-filtration chromatography ( Figure 3b ) . When Mos1 transposase was added to TP13 , there was no measurable change in the ( very low ) steady-state fluorescence intensity at 367 nm ( inset , Figure 3c ) , consistent with no change in the 2AP environment upon STC formation . In contrast , there was a ten fold increase in fluorescence intensity when Mos1 transposase was added to TP1 ( inset , Figure 3d ) , consistent with 2AP at the target site becoming unstacked by flipping into an unquenched , extra-helical environment in the Mos1 STC . A dynamic picture of DNA conformations in solution can be gained from the interpretation of the fluorescence decay of 2AP , measured by time-resolved fluorescence spectroscopy . In duplex DNA the exponential decay of 2AP fluorescence is typically described by four lifetime components , each reporting on different quenching environments that 2AP experiences as a result of the conformational dynamics of the duplex . The lifetime indicates the degree of quenching ( stacking ) in a particular conformation and the corresponding A–factor indicates the fractional occupancy of that conformation . The shortest lifetime ( τ1 ≅ 50 ps ) is due to a highly stacked conformation , which typically accounts for >70% of the population . The longest lifetime ( τ4 ≅ 9–10 ns ) corresponds to an unstacked conformation in which 2AP is extra-helical and solvent-exposed; this conformation is typically <5% of the population . The intermediate lifetimes ( τ2 ≅ 500 ps and τ3≅ 2 ns ) are due to conformations in which 2AP is intra-helical but imperfectly or partially stacked . We measured the fluorescence decays of the 2AP-containing DNA duplexes TP13 and TP1 in the absence and presence of Mos1 transposase ( Figure 3c , d ) . In the absence of transposase , 90% of the 2AP population of TP13 has the shortest lifetime ( τ1=30ps ) , indicating a tightly stacked duplex structure . Upon addition of Mos1 transposase , the decay parameters are essentially unchanged showing that the local duplex structure is unaffected , confirming the steady-state fluorescence results . TP1 fluorescence decay , in the absence of protein , is also dominated by the shortest lifetime , stacked component ( 76% , τ1= 50 ps , Figure 3d ) , with only 6% of the population in the unstacked state ( τ4 = 7 . 5 ns ) . ( The differences in the decay parameters between TP13 and TP1 are consistent with a less tightly stacked environment in the latter , where 2AP is unpaired ) . However , upon addition of transposase to TP1 , the decay parameters change markedly ( Figure 3d ) . Most notably , there is a large transfer of population from the highly stacked state ( τ1 = 80 ps ) to the unstacked , unquenched state ( τ4 9 . 7 ns ) ; the population of the former falls to 38% and that of the latter increases concomitantly to 31% . This clearly confirms that , in solution , 2AP at the position of the target adenine A1 in the Mos1 STC experiences base-flipping into an extra-helical environment . Moreover , flipping of this 2AP is dynamic: a number of conformational states are sampled , including base-flipped and base-stacked environments . Base flipping of each A1 severely distorts the surrounding target DNA . Side-chain atoms of transposase residues R186 , K190 and W159 stabilise these distortions ( Figure 4a ) by forming salt bridges or hydrogen bonds with the A1 and G2 phosphates . The DNA backbone rotations bring the G2 phosphates on both target DNA strands within 6 . 7 Å of each other and close to the guanidinium group of R186 in subunit A ( R186A ) ; each NηH2 group hydrogen bonds with a G2 phosphate oxygen on one strand ( Figure 4a ) . In both subunits , the K190 side-chain NζH2 forms a salt bridge with the other G2 phosphate oxygen on one strand . Furthermore , the W159 indole N1H interacts with an A1 phosphate oxygen . Additionally , the NηH2 group of R186B ( which has a different conformation to R186A ) interacts with the other A1 phosphate oxygen on the top strand ( black , Figure 4a ) . The NζH2 of K190 is 5 . 2 Å from the W159 indole ring and forms a cation-π stack , further enhancing stability . Together these extensive transposase–DNA backbone phosphate interactions stabilise the distorted conformation of the strand transfer product . 10 . 7554/eLife . 15537 . 013Figure 4 . Transposase interactions with rotated backbone phosphates stabilise the target DNA . ( a ) Target DNA phosphate interactions with catalytic domain residues . The side-chains of R186 , W159 and K190 can form hydrogen bonds ( dotted lines ) with backbone phosphate oxygens of A1 and G2 ( distances in Å ) . ( b ) Schematic of the in vitro Mos1 strand transfer assay . Integration of the 28 nt TS into the top target strand , yields a 68 nt product , whereas integration into the bottom strand gives a 40 nt product . ( c ) Denaturing PAGE of the strand transfer reaction products . Lanes 1 and 12 contain markers; lanes 2 and 13 , reactions without transposase; lanes 3 and 14 , reactions without target DNA . ( Integration occurs at the two TA dinucleotides in the IR sequence ) . ( d ) Quantification of the 40 nt and 68 nt products ( as a percentage of total DNA ) for each mutant transposase; error bars represent the standard deviation and were calculated from 3 experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 013 Consistent with the structural roles of W159 , R186 , F187 and K190 in the STC , individual substitution of each of these residues with alanine severely reduced the in vitro strand transfer activity of transposase ( Figure 4b ) . We detected <0 . 03% integration of fluorescently labelled Mos1 IR DNA into a target DNA duplex with a sole TA , using transposases containing the mutation W159A , R186A , K190A or F187A ( Figure 4c , d ) . By contrast , the F187W substitution resulted in 9 . 5% strand transfer , compared to 9 . 1% with T216A Mos1 transposase . Thus , an indole ring , like a phenyl ring , can occupy the space vacated by the flipped A1 base and stabilise the strand transfer product by stacking with the G2 base . The individual substitutions W159A , K190A or F187A also reduced the in vitro transposition efficiency to <20% that of T216A Mos1 transposase ( Figure 5 ) . 10 . 7554/eLife . 15537 . 014Figure 5 . Residues that stabilise the transposition product are required for efficient Mos1 transposition in vitro . Efficiencies of an in vitro Mos1 hop assay , performed using Mos1 transposase mutants and donor plasmids containing a kanamycin resistance gene flanked by Mos1 inverted repeats , as described previously ( Trubitsyna et al . , 2014 ) . Excision of the IR-flanked gene from a circular plasmid by transposase , and its integration into a supercoiled target plasmid , results in transfer of the kanamycin resistance to the target plasmid . Each mutant transposase also contained the mutation T216A , which allows soluble protein expression . Sequencing of the transposition products revealed that each mutant transposases retained faithful integration at TA sites . Error bars represent the standard deviation , calculated from three repeats of two experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 014 To test if W159 , R186 , F187 and K190 are also required for transposon excision , we performed a plasmid-based transposon cleavage assay ( Figure 6a ) . Transposon excision , and concomitant plasmid backbone release , was not affected by the transposase mutations R186A , F187A , F187W , K190A , F161A or F161W ( Figure 6b , c ) . However , the W159A mutant transposase excised only 3 . 9% of the plasmid after 24 hr , compared to 55 . 6% for the T216A transposase . Thus , Mos1 transposase residues F187 , R186 and K190 are required for target DNA integration , but are not essential for earlier cleavage steps , whereas W159 is required for both excision and strand transfer . 10 . 7554/eLife . 15537 . 015Figure 6 . Plasmid-based transposon cleavage assays . ( a ) Schematic of the in vitro plasmid-based Mos1 cleavage assay . ( b ) Agarose gel showing the products of plasmid-based transposon cleavage assays , for each mutant transposase ( Tnp ) after 2 hr and 24 hr . Control experiments show linearization of the plasmid with Sac1 ( lane 2 ) , excision of the transposon by Xba1 digestion ( lane 3 ) and reaction with no transposase ( lane 4 ) . ( c ) Quantification of the transposon and plasmid backbone released ( as a percentage of total DNA ) after 2 hr and 24 hr . Error bars represent the standard deviation calculated from 2 experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 015 Alignment of the Mos1 transposase sequence , with other mariner and Tc1-like transposases ( Figure 7 ) , reveals that K190 and W159 , which form a cation-π stack and interact with target DNA phosphates , are strictly conserved among mariner transposases . Despite the crucial role of Mos1 R186 for strand transfer in vitro , this residue is not conserved in all mariner transposases . However , the aromatic nature of F187 is conserved as either F or H in most other mariner transposases . Thus , many of the target-stabilising interactions observed in the Mos1 STC may also exist in other mariner transposases . 10 . 7554/eLife . 15537 . 016Figure 7 . Alignment of the amino acid sequence of Mos1 with six other mariner transposases and five Tc1-family transposases . The secondary structure elements of Mos1 transposase in the Mos1 STC are shown above the alignment . A red star below the alignment denotes the position of each of the catalytic acidic residues of the DDE/D triad . The third residue of this triad is typically D in the mariner sub-family and E in the Tc sub-family . The key residues involved in target DNA stabilisation in the Mos1 STC are highlighted in blue and marked by a blue dot . The figure was created with ESPript 3 . 0 ( http://espript . ibcp . fr/ESPript/cgi-bin/ESPript . cgi ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 016 The Tc1-like sequences have a conserved lysine at the position equivalent to W159 in Mos1 and there are two , proximal conserved tryptophans – aligned with Mos1 residues 149 and 194 – which could fulfil the role of W159 in Mos1 ( Figure 7 ) . Furthermore , the Tc1-like transposases contain either K or R one amino acid upstream of R186 in Mos1 , followed by an aromatic residue: F , H or Y . These residue pairs could stabilise target DNA in a similar way to R186 and F187 in Mos1 . Thus , there may be common features in the target DNA integration mechanisms of the two branches of the mariner/Tc1 family . The target-stabilising interactions described above are non-specific . In contrast , in the flipped conformation , the Watson-Crick face of each unpaired A1 base makes two adenine-specific hydrogen bonds with V214 backbone atoms ( Figure 8a ) : the exocyclic 6-amine interacts with the carbonyl oxygen , and N1 interacts with the backbone amide . 10 . 7554/eLife . 15537 . 017Figure 8 . Base-specific recognition of the flipped adenine . ( a ) Close up view of one of the flipped target adenines in the Mos1 STC crystal structure showing the hydrogen bond interactions ( dotted cyan lines , distance in Å ) with the V214 backbone atoms and the 2 and 6 positions of the adenine ring . The simulated annealing composite omit 2Fo-Fc electron density map ( grey mesh ) is contoured at 1 . 2σ . ( b ) Chemical structures and base-pairing of adenine , A , and its analogues 2-aminopurine , P , and 2 , 6-diaminopurine , D , with thymine , T or 2-thio-thymine , S . A steric clash between the 2-thio group of S and the 2-amino group of D tilts the bases relative to each other , and thus only one H-bond forms . ( c ) Denaturing PAGE of the products of strand transfer reactions with target DNA containing adenine and/or thymine analogues , as indicated above lanes 4 to 11 . ( d ) Quantification of the 40 nt and 68 nt strand transfer products for each target DNA duplex , as a percentage of total DNA . Error bars represent the standard deviation , calculated from 2 experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 017 To test if these adenine-specific hydrogen bonds are important for transposon integration specifically at a TA , we performed in vitro strand transfer assays with 2AP-containing target DNA . The arrangement of H-bond donors differs between adenine and 2AP ( Figure 8b ) . Therefore , by replacing each A1 with 2AP we expect to lose the H-bond between the A1 6-amino and V214 CO , and introduce a steric clash between the 2-amino of 2AP and T213 Cα . Since 2AP , like adenine , can make two hydrogen bonds in a base pair with thymine ( Figure 8b ) , replacing A1 with 2AP is unlikely to alter the stability and bendability of duplex target DNA . We found that replacing both A1s with 2AP ( Figure 8c , d ) resulted in a dramatic loss of specific integration at each T0 5' of the 2AP , consistent with the predicted loss of adenine-specific hydrogen-bonds with transposase . Our fluorescence experiments show that 2AP at position 1 in target DNA undergoes dynamic base flipping in the Mos1 STC ( Figure 3d ) , whereas our crystallographic snapshot with adenine at the equivalent position suggests a static flipped conformation . This may reflect different experimental conditions: the fluorescence experiments were performed in solution at room temperature , whereas the crystal structure was obtained at cryogenic temperatures . However , it is also consistent with a lack of specific interactions between 2AP and transposase , leading to an inability to trap the flipped 2AP conformation . We conclude that Mos1 integration at TA requires adenine-specific interactions with transposase to trap the flipped A1 conformation . Next we asked which A1 of the symmetrical TA sequence is essential for integration at T0: the adjacent A1 on the same strand or the complementary A1 . We replaced each A1 individually with 2AP , and efficient Mos1 integration occurred at a T0–A1 step when the T0 was base-pairedwith 2AP , but was reduced at a T0–2AP1 step ( Figure 8c , d ) . We conclude that specific Mos1 integration at a T0 requires trapping of the flipped A1 adjacent to it on the same strand . Finally we asked if the lower stability of a T:A base-pair , compared to G:C , favours mariner/Tc1 transposon integration at TA sites . We predicted that A1 flipping , and therefore strand transfer , would be hindered if the T:A base-pairing was strengthened by a third hydrogen bond , but enhanced with weakened base pairs . We replaced both A1s with 2 , 6-diaminopurine ( 2-amino-dA , or D ) , which forms three hydrogen bonds with dT ( thereby increasing base-pair stability ) but only one hydrogen bond when paired with 2-thio-dT ( or S ) ( Kutyavin et al . , 1996 ) ( Figure 8b ) . 2-amino-dA can form the adenine-specific interactions with V214 seen in the Mos1 STC structure , however the 2-amino group adds a potential clash with transposase that could lead to reduced specificity . We compared Mos1 strand transfer into TA and the altered target sequences TD and SD , with strengthened and weakened base pairing respectively ( Figure 8c , lanes 10 and 11 ) . We measured 6 . 99% integration into TA , but only 0 . 24% integration into the TD sequence ( Figure 8c , d ) and 0 . 86% integration into SD ( Figure 8c ) ; in the latter experiment many other , non-specific integration products were also observed . Thus , the weakness of the T:A base pair promotes integration at the TA sequence , and the pattern of H-bond donors and acceptors on the Watson-Crick face of adenine is important for specificity .
The Mos1 STC structure provides a snapshot of Mos1 transposition in the post-integration state . The severe target DNA bend ( ~147° ) is consistent with a bias for mariner/Tc1 integration at highly bendable , palindromic AT-rich sequences ( Vigdal et al . , 2002; Yant et al . , 2005 ) . Studies by Pflieger et al . suggested that target DNA also bends before Mos1 strand transfer ( Pflieger et al . , 2014 ) . Comparison of the Mos1 STC structure with our previous TCC model ( containing straight target DNA ) and both the pre- and post-TS cleavage PECs ( Dornan et al . , 2015; Richardson et al . , 2009 ) supports this conclusion . Our previous TCC model ( Richardson et al . , 2009 ) of straight B-form target DNA binding highlighted clashes with some transposase loop residues , indicating conformational changes in the target DNA and/or the transposase would be required for target capture . The similar architectures and interactions of the IR DNA and transposase in the STC and both PEC structures ( Figure 9a and Figure 9—figure supplement 1 ) suggest that target DNA is likely deformed . Changes to the transposase conformation are subtle and include closing-in of the catalytic domain towards the target DNA after strand transfer ( Videos 1 and 2 ) . The largest displacement ( 5 . 7 Å ) is at P210 in the turn between β7 and α8 and around helices α8 and α10 , which cradle the target DNA ( Figure 9b ) . T0 in the Mos1 STC is in a different orientation to the thymine ( T57 ) of the flanking target site duplication in the pre-TS cleavage PEC ( Figure 9c ) , which is recognised by base-specific interactions with the WVPHEL motif ( Dornan et al . , 2015 ) . By contrast , T0 closely aligns with T54 of the additional DNA duplex in the post-TS cleavage PEC ( Figure 9d ) , which may represent the target strand before integration . 10 . 7554/eLife . 15537 . 018Figure 9 . Structural comparison of the Mos1 STC with the pre- and post-TS cleavage Mos1 paired-end complexes . ( a ) Orthogonal views of the Mos1 STC ( orange ) superimposed on the pre-TS cleavage PEC ( PDB ID: 4U7B , green ) : r . m . s . d . over all transposase backbone atoms , 1 . 2 Å . Video 1 and video 2 show the transposase morphing from the pre- and post-cleavage PEC structures to the STC , respectively . ( b ) Close-up view of part of the catalytic domain , boxed in ( a ) . Mos1 STC target DNA and the pre-TS cleavage PEC flanking DNA are shown as sticks ( pink and black ) and a green cartoon , respectively . Dotted lines indicate the displacement between the two structures , with distances in Å . ( c ) and ( d ) Close-up view of the Mos1 STC ( orange ) active site superimposed on ( c ) the pre-TS cleavage PEC ( green ) and ( d ) the post-TS cleavage PEC ( PDB ID: 3HOS ) : T54 in the additional DNA duplex ( lavender sticks ) may represent T0 of target DNA before strand transfer . A full view of the Mos1 STC superposed on the post-TS cleavage PEC structures is shown in Figure 9—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 01810 . 7554/eLife . 15537 . 019Figure 9—figure supplement 1 . Structural comparison of the Mos1 STC with the post-TS cleavage Mos1 paired-end complex . The Mos1 STC ( orange ) is superimposed on the post-TS cleavage PEC ( PDB ID: 3HOS , lavender ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 01910 . 7554/eLife . 15537 . 020Video 1 . Morphing of the Mos1 transposase conformation in the pre-TS cleavage PEC ( PDB ID: 4U7B ) into the Mos1 STC conformation . Related to Figure 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 02010 . 7554/eLife . 15537 . 021Video 2 . Morphing of the Mos1 transposase conformation in the post-cleavage PEC ( PDB ID: 3HOS ) into the Mos1 STC conformation . Related to Figure 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 021 Strain created by target DNA bending during target capture likely drives the phosphate backbone rotations that flip the target adenines into extra helical positions ( Figure 10 ) . Subsequent trapping of the flipped adenines may correctly position the scissile target phosphates for in-line attack by the cleaved transposon ends . Breaking of the target DNA strands by strand transfer would allow displacement of the new transposon-target DNA junction from the active site , preventing reversal of the reaction . Structural and biochemical characterisation of the target capture complex will illuminate this sequence of events . 10 . 7554/eLife . 15537 . 022Figure 10 . A proposed mechanism for Mos1 transposon integration incorporates target DNA bending and trapping of flipped target adenines . Schematic representation of key features of the STC ( right ) and the proposed target capture complex ( left ) , with transposase subunits ( orange and blue ) . Filled circles represent residues W159 ( W ) , R186 ( R ) , F187 ( F ) and K190 ( K ) and the encircled DDD depicts each active site . Arrows indicate rotation of the DNA backbone of each target DNA strand ( magenta and black ) . Dotted lines represent hydrogen bonds between TA base pairs in the TCC and between flipped target adenines and transposase backbone atoms in the STC . DOI: http://dx . doi . org/10 . 7554/eLife . 15537 . 022 Many DNA-metabolising enzymes use base flipping to expose bases normally embedded within a double helix; enabling base methylation ( Klimasauskas et al . , 1994 ) , removal of damaged or mismatched bases or for DNA sequence recognition ( Bochtler et al . , 2006; Neely et al . , 2009 ) . During Tn5 transposon excision , formation and resolution of DNA hairpins at the transposon ends requires base flipping: rotation of a base close to the cleavage site , into a protein pocket , relieves strain in the tight hairpin bend and aligns the transposon ends for cleavage ( Ason and Reznikoff , 2002 ) . Similarly , DNA hairpin stabilisation by base flipping has been proposed for V ( D ) J recombination and transposition of Hermes and Tn10 ( Lu et al . , 2006; Bischerour and Chalmers , 2009 ) . Active intrusion of a probe amino acid residue can drive base flipping . During Mos1 integration , the F187 aromatic ring may actively force the A1s from the DNA helix , similar to the methionine probe in Tn10 transposon excision ( Bischerour and Chalmers , 2009 ) . Alternatively , F187 may passively fill the gap left by A1 after flipping , to enhance the stability of the distorted target DNA conformation . In this scenario , the conserved transposase residues K190 and W159 may be alternative drivers of the base-flipping rotations . Target adenine–specific interactions with V214 backbone atoms suggest a molecular basis for TA target sequence recognition in the post-integration state . The structured loop containing V214 has the consensus sequence T- ( V/I ) - ( N/T ) in mariner transposases ( Figure 7 ) , suggesting that the role of transposase backbone atoms in TA recognition may be conserved among mariner-family transposases . The Tc1-family transposases , including Sleeping Beauty , also display sequence conservation in this region ( Figure 7 ) , suggesting similar recognition mechanisms exist in that closely-related family . Structures of Sleeping Beauty transposition intermediates will reveal if this is the case . Target DNA bending is a recurring theme in DNA transposition by DDE/D recombinases . The severe target DNA bend ( ~147° ) observed in our Mos1 STC structure is similar to the ~140° target DNA distortion in the bacterial MuA transpososome , which was proposed to drive the isoenergetic strand transfer reaction forward ( Montaño et al . , 2012 ) . Mu employs a helper protein ( MuB ) in its transposition , which may facilitate target DNA bending by forming helical filaments on DNA , prior to capture by the MuA transpososome ( Mizuno et al . , 2013 ) . Similarly , the bacterial insertion sequence IS21 requires IstB for efficient transposition . In the presence of ATP , IstB self-assembles into decamers that can bend ~50 bp DNA by 180° ( Arias-Palomo and Berger , 2015 ) . In the PFV integrase target capture and strand transfer complexes ( Maertens et al . , 2010 ) naked target DNA is bent by 55° . Nucleosomal DNA is peeled from the histone octamer and similarly deformed by interactions with the PFV intasome , providing a structural basis for retroviral integration at nucleosomes ( Maskell et al . , 2015 ) . By contrast , recent evidence indicates that mariner/Tc1 transposons preferentially integrate at linker regions between nucleosomes ( Gogol-Döring et al . , 2016 ) . Our results provide a structural basis for this preference: severe target DNA bending ( ~147° ) by the transpososome can be more easily achieved on flexible linker DNA than on DNA tightly engaged with the histone octamer in a nucleosome structure . Taken together our structural and biochemical data support a dynamic bend , flip and trap mechanism for Mos1 transposon integration at TA target sites ( Figure 10 ) that may be a conserved feature of mariner/Tc1 transposition . As such , our results provide a framework for designing mariner/Tc1 transposases with modified target specificities .
Expression constructs encoding Mos1 transposase mutants H122A , W159A , F161A , F161W , R186A , F187A , F187W , K190A were generated by site-directed mutagenesis ( Quikchange , Stratagene ) of the codon-optimised Mos1 gene ( Trubitsyna et al . , 2014 ) , according to the manufacturer’s protocol . Each plasmid also incorporated the T216A mutation allowing soluble expression of Mos1 transposase in E . coli ( Richardson et al . , 2004 ) . Each mutant transposase was expressed and purified as described previously ( Richardson et al . , 2004 ) , exchanged into buffer containing 25 mM PIPES pH 7 . 5 , 250 mM NaCl , 0 . 5 mM DTT and 50% ( v/v ) glycerol and concentrated to between 10–20 mg mL-1 . The sequences of all DNA oligonucleotides are shown in Table 1 . HPLC purified oligonucleotides for crystallisation of the STC were purchased from IDT ( Belgium ) , PAGE purified and dissolved to 1 mM in TEN buffer ( 10 mM Tris pH 8 , 1 mM EDTA , 50 mM NaCl ) . The 36 nt TS incorporates the 28 nt IR and target DNA ( as shown in Figure 1c ) . The 25 nt NTS is complementary to the TS IR DNA sequence and represents the authentic product the first cleavage . The 10 nt target DNA sequence , includes six nucleotides complementary to the 3' TS target sequence and four self-complementary nucleotides ( cohesive 5' ends ) . The three oligonucleotides were mixed in a 1:1:1 molar ratio and annealed by heating to 363 K for 3 min and cooling to room temperature over ~2 hr . For time-resolved fluorescence experiments , DNA oligonucleotides were synthesised and HPLC purified by ATDBio ( Southampton , UK ) . Three TS sequences , extended at the 3' end to 46 nt , were synthesised: TS_A1 , an unlabelled control; TS_P1 , with 2AP in place of the target adenine; TS_P13 with 2AP at position 13 , another control . Each TS was annealed with the 25 nt NTS and the 16 nt target_16 sequence complementary to the TS 3' end . This yielded the three duplexes – TA1 , TP1 and TP13 – which mimic the Mos1 strand transfer product . For the strand transfer assays , the IR DNA was prepared by annealing the 28 nt 5'-IRDye 700 labelled TS with the 25 nt complementary NTS . The 50-mer TA target DNA , was prepared by annealing complementary top and bottom strands ( Table 1 ) . Five target DNA variants were similarly prepared: three had 2-aminopurine ( P ) in place of the target adenine on the top and/or bottom strand . A fourth had 2 , 6-diaminopurine ( D ) in place of the target adenine on both strands , and the fifth also had 2-thio-thymine ( S ) in place of the target thymine on both strands . The annealed IR and target oligonucleotides were purified by HPLC . The STC was formed by adding T216A Mos1 transposase ( 438 μM ) and STC ds DNA ( 229 μM ) together to final concentrations of 50 μM each in a solution of 25 mM PIPES-NaOH pH 7 . 5 , 250 mM NaCl , 20 mM MgCl2 and 1 mM DTT . The final concentration of the STC was 25 μM . Crystals were grown by sitting drop vapour-diffusion . Drops contained 2 μL of STC ( 25 μM ) and 1 μL of well solution comprising 30% ( v/v ) MPD , 0 . 1 M sodium cacodylate pH 6 . 5 and 0 . 2 M magnesium acetate tetrahydrate . The crystals were cooled in liquid nitrogen for X-ray diffraction experiments . X-ray diffraction data were collected on beam line I02 at the Diamond Light Source . Crystals displayed C-centred ( C121 ) symmetry and diffracted X-rays to a maximum resolution of 3 . 3 Å . The X-ray diffraction data were processed with iMosflm , scaled and merged with AIMLESS and the statistics are shown in Table 2 . Initial phases were determined by molecular replacement , using our structure of the Mos1 PEC ( PDB ID: 3HOS , chains A to F , comprising the transposase dimer and two cleaved IR DNA molecules ) as the search model in PHASER . The difference electron density after molecular replacement is shown in Figure 2—figure supplement 1 . The remaining structure was built manually . Restrained refinement was performed with Refmac and Coot and included automatic non-crystallographic symmetry restraints on the protein and DNA chains . The refinement statistics are shown in Table 2 . All structural diagrams were prepared using PyMOL ( http://www . pymol . org/ ) and Adobe Illustrator . Target integration assays were performed as described previously ( Wolkowicz et al . , 2014 ) . 20 μL reactions containing 15 nM of a 50-mer target DNA , 1 . 5 nM IR DNA and 15 nM Mos1 transposase in buffer containing 25 mM HEPES pH 7 . 5 , 50 mM Potassium Acetate , 10% ( v/v ) glycerol , 0 . 25 mM EDTA , 1mM DTT , 10 mM MgCl2 , 50 μg/mL BSA and 20% ( v/v ) DMSO were incubated for two hours at 30°C and the products separated on an 8% denaturing polyacrylamide gel . To visualise the products , the IRDye700 was excited at 680 nm and detected on a LI-COR Odyssey system . The fluorescence intensities of the product bands were quantified using Image Studio software . Plasmid-based transposon cleavage assays were performed as described previously ( Trubitsyna et al . , 2014 ) . Measurements were acquired , in photon counting mode , on a Fluoromax–3 spectrofluorimeter ( Jobin Yvon , Stanmore , UK ) , on samples of the 2AP-containing duplexes TP13 or TP1 ( 10 μM ) , alone or mixed with 11 μM Mos1 transposase , in buffer composed of 25 mM PIPES-NaOH pH 7 . 5 , 250 mM NaCl , 20 mM CaCl2 , 1 mM DTT . A circulating water bath maintained sample temperatures at 25°C . Emission spectra were recorded in the range 325–550 nm , with an excitation wavelength of 317 nm and excitation and emission bandwidths of 2 . 5 nm . Measurements were performed using time-correlated single photon counting , on an Edinburgh Instruments spectrometer equipped with TCC900 photon counting electronics , as described previously ( Neely et al . , 2005 ) . The excitation source was the third harmonic of the pulse-picked output of a Ti-sapphire femtosecond laser system ( Coherent , 10 W Verdi and Mira Ti-Sapphire ) , consisting of ~200 fs pulses at a repetition rate of 4 . 75 MHz and a wavelength of 317 nm . The instrument response of the system was ~80 ps full-width at half-maximum . Fluorescence decay curves were analysed by iterative re-convolution , assuming a multi-exponential decay function , given in Equation ( 1 ) ( 1 ) I ( t ) = ∑i=14 Aiexp ( −tτi ) where I is the fluorescence intensity as a function of time ( t ) ; τi is the fluorescence lifetime of the ith decay component and Ai is the fractional amplitude ( A-factor ) of that component . Decays were collected at two emission wavelengths ( 375 nm and 390 nm ) and were analysed globally , with τi as the common parameter , using Edinburgh Instruments software FAST . | The complete set of DNA in a cell is referred to as its genome . Most genomes contain short fragments of DNA called transposons that can jump from one place to another . Transposons carry sections of DNA with them when they move , which creates diversity and can influence the evolution of a species . Transposons are also being exploited to develop tools for biotechnology and medical applications . One family of transposons – the Mariner/Tc1 family – has proved particularly useful in these endeavours because it is widespread in nature and can jump around the genomes of a broad range of species , including mammals . DNA transposons are cut out of their position and then pasted at a new site by an enzyme called transposase , which is encoded by some of the DNA within the transposon . DNA is made up of strings of molecules called bases and Mariner/Tc1-family transposons can only insert into a new position in the genome at sites that have a specific sequence of two bases . However , it was not known how this target sequence is chosen and how the transposon inserts into it . Morris et al . have now used a technique called X-ray crystallography to build a three-dimensional model of a Mariner/Tc1-family transposon as it inserts into a new position . The model shows that , as the transposon is pasted into its new site , the surrounding DNA bends . This causes two DNA bases in the surrounding DNA to flip out from their normal position in the DNA molecule , which enables them to be recognised by the transposase . Further experiments showed that this base-flipping is dynamic , that is , the two bases continuously flip in and out of position . Furthermore , Morris et al . identified which parts of the transposase enzyme are required for the transposon to be efficiently pasted into the genome . Together these findings may help researchers to alter the transposase so that it can insert the transposon into different locations in a genome . This will hopefully lead to new tools for biotechnology and medical applications . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | A bend, flip and trap mechanism for transposon integration |
The integrity of the nuclear membranes coupled to the selective barrier of nuclear pore complexes ( NPCs ) are essential for the segregation of nucleoplasm and cytoplasm . Mechanical membrane disruption or perturbation to NPC assembly triggers an ESCRT-dependent surveillance system that seals nuclear pores: how these pores are sensed and sealed is ill defined . Using a budding yeast model , we show that the ESCRT Chm7 and the integral inner nuclear membrane ( INM ) protein Heh1 are spatially segregated by nuclear transport , with Chm7 being actively exported by Xpo1/Crm1 . Thus , the exposure of the INM triggers surveillance with Heh1 locally activating Chm7 . Sites of Chm7 hyperactivation show fenestrated sheets at the INM and potential membrane delivery at sites of nuclear envelope herniation . Our data suggest that perturbation to the nuclear envelope barrier would lead to local nuclear membrane remodeling to promote membrane sealing . Our findings have implications for disease mechanisms linked to NPC assembly and nuclear envelope integrity .
The molecular machinery that biochemically segregates the nucleus and the cytoplasm has been extensively investigated . The foundational components of this selective barrier include the double-membrane nuclear envelope with embedded nuclear pore complexes ( NPCs ) . NPCs impose a ‘soft’ diffusion barrier to macromolecules larger than ~40 kD ( Popken et al . , 2015; Timney et al . , 2016 ) while providing binding sites for the rapid and selective transport of signal-bearing ( nuclear localization and nuclear export signals; NLSs and NESs ) macromolecules , which are ferried through the NPC by shuttling nuclear transport receptors ( NTRs; a . k . a . karyopherins/importins/exportins; Schmidt and Görlich , 2016 ) . Directionality and energy for NTR-selective transport is imparted by the spatial segregation of the Ran-GTPase whose nuclear GTP-bound form destabilizes and stabilizes import and export complexes , respectively ( Floch et al . , 2014 ) . Interestingly , the robustness of the nuclear envelope barrier has been shown to be compromised in several different contexts , including in diverse human diseases ( Hatch and Hetzer , 2014; Lusk and King , 2017 ) . For example , there is an emerging body of work linking the function of NTRs and NPCs with neurodegenerative diseases like ALS and FTD ( Nousiainen et al . , 2008; Freibaum et al . , 2015; Jovičić et al . , 2015; Kaneb et al . , 2015; Zhang et al . , 2015; Kim and Taylor , 2017; Shi et al . , 2017 ) . These studies , coupled to the observations of age-related declines in NPC function in both post-mitotic multicellular systems ( D'Angelo et al . , 2009; Savas et al . , 2012; Toyama et al . , 2013 ) and also in replicative aging models like budding yeast ( Janssens et al . , 2015; Lord et al . , 2015 ) , support a theme in which the function of the nuclear envelope could be mitigatory of age-related disease progression ( Schreiber and Kennedy , 2013; Jevtić et al . , 2014; Serebryannyy and Misteli , 2018 ) . Of similar interest , the hallmark cellular pathophysiology of early-onset dystonia are nuclear envelope herniations ( Goodchild et al . , 2005 ) that emanate from NPC-like structures ( Laudermilch et al . , 2016 ) . As analogous herniations have been observed in many genetic backgrounds associated with defects in NPC biogenesis in yeast over several decades ( Thaller and Lusk , 2018 ) , this has contributed to the idea that the herniations might be the result of either defective NPC assembly events ( Scarcelli et al . , 2007; Onischenko et al . , 2017; Zhang et al . , 2018 ) and/or the triggering of a NPC ( Wente and Blobel , 1993 ) or NPC assembly quality control pathway ( Webster et al . , 2014; Webster et al . , 2016 ) . The latter could depend on the function of the endosomal sorting complexes required for transport ( ESCRT ) , a membrane scission machinery that has been proposed to seal-off malforming NPCs ( Webster et al . , 2016 ) . That there could be mechanisms to surveil the assembly of NPCs makes considerable sense as there are hundreds of NPCs , each containing hundreds of nucleoporins/nups ( Kosinski et al . , 2016; Kim et al . , 2018 ) , that are assembled during interphase in mammalian cells ( Maul et al . , 1972; Doucet et al . , 2010; Dultz and Ellenberg , 2010 ) . There are approximately 100 NPCs formed during a budding yeast cell cycle , which includes a closed mitosis ( Winey et al . , 1997 ) . As interphase NPC assembly likely occurs through an inside-out evagination of the inner nuclear membrane ( INM ) followed by membrane fusion with the outer nuclear membrane ( ONM ) ( Otsuka et al . , 2016 ) , holes are constantly being formed in the nuclear envelope . Without mechanisms to surveil this process , de novo NPC biogenesis might pose a threat to nuclear-cytoplasmic compartmentalization ( Webster et al . , 2014 ) . Consistent with this idea , malformed or damaged NPCs are not passed on to daughter cells in budding yeast ( Colombi et al . , 2013; Makio et al . , 2013; Webster et al . , 2014 ) . Further , deletion of the ESCRT machinery in the context of genetic backgrounds where nuclear envelope herniations have been observed for example nup116Δ ( Wente and Blobel , 1993 ) or apq12Δ ( Scarcelli et al . , 2007 ) cells require a nuclear envelope-specific ESCRT , Chm7 ( the orthologue of mammalian CHMP7 ) , for viability ( Bauer et al . , 2015; Webster et al . , 2016 ) . While we have previously proposed that a biochemical signature of malforming NPCs is surveilled by integral inner nuclear membrane proteins of the Lap2-emerin-MAN1 ( LEM ) domain family , specifically Heh2 , it remains to be formally established what the signal that leads to ESCRT recruitment to the nuclear envelope actually comprises ( Webster et al . , 2014 ) . Evidence that the ESCRT machinery acts at holes in the nuclear envelope is further exemplified by their critical role in performing annular fusion events during the final stages of nuclear envelope reformation at the end of mitosis in mammalian cells ( Olmos et al . , 2015; Olmos et al . , 2016; Vietri et al . , 2015; Gu et al . , 2017; Ventimiglia et al . , 2018 ) . Moreover , ESCRTs are also required for the efficient repair of nuclear ruptures that arise during the migration of cells through tight constrictions ( Denais et al . , 2016; Raab et al . , 2016 ) . And , it is most likely that they also act to repair nuclear envelope ruptures that are induced by intracellular mechanical stresses from either the actin cytoskeleton ( Hatch and Hetzer , 2016; Robijns et al . , 2016 ) , or from those observed during telomere crisis ( Maciejowski et al . , 2015 ) . Lastly , recent work also suggests a role for ESCRTs in the context of turning over NPCs in quiescent cells ( Toyama et al . , 2019 ) . It remains an open question , however , whether the mechanisms that repair nuclear ruptures , seal the nuclear envelope at the end of mitosis , and protect against defective NPC assembly respond to an identical upstream signal and proceed through the same membrane-sealing mechanism . Clues to what might constitute the upstream signal that leads to nuclear envelope-recruitment of ESCRTs could be drawn from other contexts where ESCRTs protect membrane compartments including endolysosomes ( Radulovic et al . , 2018; Skowyra et al . , 2018 ) and the plasma membrane ( Jimenez et al . , 2014; Scheffer et al . , 2014; Gong et al . , 2017 ) . In both of these cases , there is evidence to suggest that the local release of Ca2+ is a trigger for ESCRT recruitment , through ( at least at the plasma membrane ) a Ca2+ binding protein , ALG-2 ( Jimenez et al . , 2014; Gong et al . , 2017 ) . Whether Ca2+ plays a role at the nuclear envelope remains unaddressed . More generally , there are two , often redundant , recruitment mechanisms seeded by either an ESCRT-I , II complex and/or ESCRT-II and ALIX ( Bro1 in yeast ) that bind and activate ESCRT-III subunit polymerization ( Wemmer et al . , 2011; Henne et al . , 2012; Tang et al . , 2015; Tang et al . , 2016; Christ et al . , 2016 ) on specific membranes throughout the cell ( reviewed in Schöneberg et al . , 2017; McCullough et al . , 2018 ) . ESCRT-III polymers predominantly made up of the most abundant ESCRT-III ( Snf7/CHMP4B ) scaffold negative but also in at least one case , positive membrane curvature ( McCullough et al . , 2015 ) , and directly contribute to membrane scission ( Adell et al . , 2014; Adell et al . , 2017; Schöneberg et al . , 2018 ) . The AAA +ATPase Vps4 disassembles ESCRT-III filaments by directly interacting with MIM ( MIT interacting motif ) domains present on a subset of ESCRT-III subunits ( Obita et al . , 2007; Stuchell-Brereton et al . , 2007; Agromayor et al . , 2009; Xiao et al . , 2009; Han et al . , 2015 ) by threading the ESCRT-III filaments through the central cavity of a hexameric ring ( Yang et al . , 2015; Han et al . , 2017; Monroe et al . , 2017; Su et al . , 2017 ) . It is likely that ESCRT-III disassembly by Vps4 directly contributes force to promote membrane scission ( Schöneberg et al . , 2018 ) . Whether the membrane scission reaction is different in distinct subcellular contexts like at the nuclear envelope remains to be understood . Consistent with the idea that there might be unique ESCRT membrane remodeling mechanisms at play in distinct compartments , a step-wise recruitment and activation mechanism requiring the ESCRT-II Vps25 and the ESCRT-III Vps20 is thought to be required at budding yeast endosomes ( Saksena et al . , 2009; Teis et al . , 2010; Tang et al . , 2015; Tang et al . , 2016 ) , but both of these proteins are absent from the genetic and biochemical analyses of the nuclear envelope arm of the ESCRT pathway ( Webster et al . , 2014; Webster et al . , 2016 ) . These data suggest that other proteins likely contribute to ESCRT-III activation at the nuclear envelope . Key candidates are Chm7 and the inner nuclear membrane ( INM ) proteins , Heh1/Src1 ( orthologue of LEM2 ) and Heh2 ( orthologue of MAN1 or other LEM-domain proteins ) . These proteins have collectively been shown to interact biochemically and genetically with Snf7 ( Webster et al . , 2014; Webster et al . , 2016 ) and Heh1 is required for the focal accumulation of Chm7 at the nuclear envelope in genetic backgrounds where NPC assembly is inhibited ( Webster et al . , 2016 ) . Remarkably , the interactions between Heh1 and Chm7 are well conserved in both fission yeast ( Gu et al . , 2017 ) but also in mammalian cells , where LEM2 is required to recruit CHMP7 to the reforming nuclear envelope at the end of mitosis ( Gu et al . , 2017 ) . It remains to be understood , however , whether LEM proteins ( or Chm7 ) directly contribute to ESCRT-III activation at the nuclear envelope or whether additional proteins are involved . Heh1 and Heh2 contain an N-terminal helix-extension-helix ( heh ) motif ( the LEM domain ) , followed by an INM targeting sequence that ( at least in the case of Heh2 but likely also Heh1 [King et al . , 2006; Lokareddy et al . , 2015] ) includes an NLS and a ~ 200 amino acid unstructured region that are both required for INM targeting ( Meinema et al . , 2011 ) . They both also contain a second nuclear-oriented domain , which likely folds into a winged helix ( WH ) ( also called MAN1/Src1-C-terminal homology domain or MSC; Caputo et al . , 2006 ) ; this domain is also well conserved through evolution ( Mans et al . , 2004; Mekhail et al . , 2008 ) . The LEM domain proteins as a family have been ascribed diverse roles in gene expression either through binding to transcription factors , BAF , or the lamins ( Barton et al . , 2015 ) . While yeasts lack BAF and lamins , the LEM domain proteins nonetheless directly interface with chromatin ( Grund et al . , 2008; Barton et al . , 2015 ) , contribute to rDNA repeat stability ( Mekhail et al . , 2008 ) and the mechanical robustness of the nucleus ( Schreiner et al . , 2015 ) . This latter function is likely revealed by the observation in many diverse yeast species that loss of Heh1 leads to nuclear envelope disruption ( Yewdell et al . , 2011; Gonzalez et al . , 2012; Yam et al . , 2013 ) , however , it might also reflect Heh1’s role in recruiting ESCRTs to the nuclear envelope . Thus , the relationship between how the LEM proteins contribute to nuclear integrity and ESCRT-mediated surveillance remains to be clearly defined . In the following , we further explore the molecular determinants of Chm7 recruitment to the budding yeast nuclear envelope by Heh1 . We determine that the spatial segregation of Chm7 and Heh1 on either side of the nuclear envelope is driven by NTRs and the robustness of the nuclear transport system . Perturbations of this system , or the exposure of Heh1 to the cytosol leads to the local recruitment and Heh1 WH-dependent activation of Chm7 . At sites of Chm7 hyperactivation , we observe remarkable alterations to nuclear envelope morphology including nuclear envelope herniations and intranuclear INM invaginations suggesting a role for membrane expansion and remodeling during nuclear envelope repair .
It was our previous experience that visualizing endogenous levels of Chm7-GFP was challenging due to its low level of expression ( Webster et al . , 2016 ) , thus , to gain further insight into the localization determinants of Chm7 in budding yeast , we overexpressed Chm7-GFP behind the control of a galactose ( GAL1 ) inducible promoter . As shown in Figure 1C , culturing of cells in galactose for a short time ( ~45 min ) led to the appearance of Chm7-GFP fluorescence throughout the cytosol . Unexpectedly , we also observed that Chm7-GFP was excluded from the nuclear interior ( orange asterisks ) . These data raised the possibility that Chm7-GFP is unable to cross the diffusion barrier imposed by NPCs , or , there is an active nuclear export pathway that prevents Chm7 from accessing the nucleus . Consistent with the possibility that Chm7 might be recognized by an export NTR , we identified two putative leucine-rich NESs at the C-terminus of Chm7 using the NES prediction algorithm , LocNES ( Xu et al . , 2015; Figure 1A , B ) . Interestingly , the higher scoring predicted NES ( NES2 ) overlapped with the potential MIM1 motif ( Bauer et al . , 2015 ) ( Figure 1A , Figure 1—figure supplement 1A ) . Indeed , the Chm7 putative MIM1 motif stands out from that of other budding yeast ESCRT-IIIs because of an additional isoleucine that contributes the fourth hydrophobic amino acid required for an effective class 1a NES ( Figure 1—figure supplement 1A ) . A second leucine in the middle of this region also aligns with the predictive spacing of residues in class 1b NESs ( Figure 1—figure supplement 1A ) . Moreover , these putative NESs were conserved in Chm7 orthologues in other species of yeast , mice , flies and humans , although are curiously absent from C . elegans ( Figure 1—figure supplement 1B ) . As LocNES predicts NESs specific for the major export NTR , Xpo1/Crm1 , we tested whether the inhibition of Xpo1 reversed the nuclear exclusion of Chm7-GFP . For these experiments , we took advantage of the xpo1-T539C allele , which sensitizes budding yeast Xpo1 to the Xpo1 inhibitor Leptomycin B ( LMB ) ( Neville and Rosbash , 1999 ) ( Figure 1E ) . As shown in Figure 1C , 45 min LMB treatment of cells expressing Chm7-GFP led to the striking accumulation of Chm7-GFP at the periphery of the nucleus most often in highly fluorescent foci in over 60% of cells ( Figure 1D ) , which is an underestimate as only midplanes were quantified . The focal accumulation of Chm7-GFP at the nuclear periphery suggested that Chm7 was able to enter the nucleus upon Xpo1 inhibition . Therefore , to distinguish whether nuclear entry was driven by an active NTR-mediated nuclear import pathway , or , whether it was the result of passive diffusion across the NPC , we generated a Chm7-GFP fused to five maltose-binding proteins ( Chm7-MGM4 ) ; this ~280 kD protein would be extremely inefficient at transiting through the NPC unless it contained an NLS . Consistent with the idea that Chm7-GFP’s entry into the nucleus was governed by diffusion and not active NLS-mediated transport , the distribution of Chm7-MGM4 was indistinguishable from Chm7-GFP but , in contrast , incubation with LMB had no effect on its localization ( Figure 1C–E ) . As inhibition of Xpo1 led to Chm7-GFP accumulation at the nuclear envelope , we reasoned that it was likely that the prediction of NESs in Chm7 was likely accurate . However , as both NLS and NES prediction is of limited utility , we directly tested whether the predicted NESs were indeed sufficient to prevent an inert GFP reporter from localizing in the nucleus at steady state . We tested the localization of GFP-fusions to each predicted Chm7 NES alone , and in combination ( Figure 1A ) . To help demark the nuclear boundary , these constructs were expressed alongside dsRED-HDEL , which localizes throughout the continuous nuclear envelope/ER lumen ( Madrid et al . , 2006 ) . As shown in Figure 1F , when compared to GFP-alone , all three NESCHM7-GFP constructs showed a deficit of nuclear accumulation that could be reversed by treatment with LMB . Interestingly , when we compared line profiles of GFP fluorescence drawn from the cytosol and bisecting the nucleus , NES1CHM7-GFP was more obviously excluded from the nucleus than NES2CHM7-GFP with NES1-2CHM7-GFP showing the most striking absence of nuclear signal ( compare the depth of the ‘valleys’ of the red lines; Figure 1G ) . Therefore , it is likely that both predicted NESs contribute to the efficient export of Chm7 . Consistent with this , the examination of a truncation of Chm7 ( chm7OPEN; Figure 1A ) lacking both NESs dramatically accumulates in one or two foci on the nuclear envelope in a way that is not impacted by LMB ( Figure 1C , D and see Webster et al . , 2016 ) . Thus , the steady state nuclear exclusion of Chm7 in wildtype cells is determined by its passive diffusion into the nucleus and the Xpo1-mediated recognition of NESs in Chm7 . That , having entered the nucleus , Chm7 accumulates in a focus along the nuclear periphery is consistent with its binding and activation at the INM . The latter being reflected in its focal accumulation , which would be consistent with a polymerization event . That Chm7 can be recruited and activated at the INM without any perturbation to the nuclear envelope raises the possibility that there are no other upstream signals that are necessary to trigger Chm7 recruitment . While this is difficult to conclusively prove , we nonetheless tested whether Ca2+ could reflect an additional signal because of its role in other ESCRT-mediated membrane repair processes ( Jimenez et al . , 2014; Scheffer et al . , 2014; Gong et al . , 2017; Skowyra et al . , 2018 ) . As Chm7 is recruited to the nuclear envelope in apq12Δ cells when grown at elevated ( 37°C ) temperatures ( Webster et al . , 2016 ) , we evaluated whether this recruitment was influenced by chelating Ca2+ using BAPTA-AM . As shown in Figure 1H , there was no obvious change to the number of Chm7-GFP foci that appear during the temperature shift in the presence or absence of Ca2+ ( Figure 1H , I ) . Thus , it is unlikely that a Ca2+ signal is a major contributor to this pathway . We hypothesized that Chm7 was excluded from the nucleus in order to prevent its untimely or inappropriate ‘activation’ in the absence of a perturbation of the nuclear envelope barrier . Such a model predicts that there must be a nuclear-binding partner that itself might be ‘hidden’ from cytosolic Chm7; based on our and others’ prior work ( Webster et al . , 2014; Webster et al . , 2016; Gu et al . , 2017 ) the most obvious candidate was Heh1 . To test this hypothesis , we generated deletion constructs of Heh1 coupled to the Red Fluorescent Protein ( RFP ) expressed behind the GAL1 promoter ( note that there is vacuolar autofluorescence even under repressed glucose conditions , see asterisks in Figure 2A ) . Unlike many other INM proteins that tend to back up into the ER upon overexpression ( Lusk et al . , 2007 ) , Heh1-RFP continues to accumulate at the INM even at high levels due to its use of an active NTR-dependent INM targeting pathway ( See Figure 2A , B and King et al . , 2006 ) . Thus , even when overexpressed at levels that we estimate to be an order of magnitude higher than endogenous levels , the majority of Heh1 is localized to the INM and would be predicted to be inaccessible to cytosolic Chm7 ( Figure 2A , galactose , middle panels ) . Consistent with this , we observed no change to the steady state distribution of endogenously-expressed Chm7-GFP , which includes a minor fraction within a nuclear envelope focus in ~30% of cells ( Webster et al . , 2016; Figure 2A ) . Deletion of the LEM domain of Heh1 also had no effect on Chm7-GFP distribution as heh1 ( 51-834 ) -RFP was also exclusively localized at the INM ( Figure 2A ) . We next tested deletions that encompassed the putative NLSs in Heh1 including heh1 ( 303-834 ) , and heh1 ( 442-834 ) , which resulted in the accumulation of these truncations throughout the cortical ER . Strikingly , we observed a concurrent re-distribution of Chm7-GFP into foci that colocalized with the RFP signal ( Figure 2A ) . In the case of heh1 ( 442-834 ) , only the WH domain is available for Chm7 binding . Consistent with this , there was a complete lack of Chm7-GFP at the ER in cells expressing heh1 ( 442-735 ) , where the WH is removed ( Figure 2A ) . Thus , exposure of the Heh1 WH domain to the cytosol is both necessary and sufficient to recruit Chm7-GFP to ER membranes . We next assessed the functional importance of the Heh1-WH domain to apq12Δ cells , which require both CHM7 and HEH1 for full viability ( Yewdell et al . , 2011; Bauer et al . , 2015; Webster et al . , 2016 ) ( Figure 2C ) . Interestingly , the loss of fitness observed in heh1Δapq12Δ cells could only be rescued by the gene encoding full length Heh1 or the heh1 ( 51-834 ) allele . In contrast , deletions that resulted in Heh1 mistargeting or those that are unable to recruit Chm7 ( e . g . heh1 ( 1-735 ) , which lacks the coding sequence for the WH domain ) were unable to fully complement growth . Thus , while the WH domain is important , the N-terminal INM targeting domain is also a critical component of Heh1 functionality in the context of apq12Δ cells . The localization data clearly pointed to a direct interaction between the WH domain of Heh1 and Chm7 . Unfortunately , we were unable to detect a stable interaction in vitro with purified recombinant proteins ( one example shown in Figure 3—figure supplement 1A ) , although we note such an interaction has been shown with the human versions of these proteins ( Gu et al . , 2017 ) . While there are many potential reasons for these negative data , one possibility is that there are additional proteins ( or lipids ) that contribute to the interaction in vivo . To test this idea , we affinity purified Chm7-GFP from whole cell extracts using anti-GFP nanobody-coupled beads ( Figure 3—figure supplement 1B ) and subjected protein eluates to MS/MS peptide identification . Consistent with the idea that Chm7 is localized throughout the cytosol in a potentially inactive form , we detected few specific peptide spectra with the exception of Chm7 itself when compared to proteins derived from wildtype cell extracts that bind non-specifically to the anti-GFP beads ( Figure 3A ) . To facilitate visualization , we directly relate the average spectral counts ( two experiments ) from bound fractions of affinity purifications of Chm7-GFP and no-GFP controls in Figure 3A . We therefore turned to examining the interactome of Chm7-GFP under conditions in which it accumulates at the nuclear envelope , for example in vps4Δpom152Δ cells , which we had previously shown leads to Chm7-GFP accumulation within a nuclear envelope domain enriched for malformed NPCs ( Webster et al . , 2014; Webster et al . , 2016 ) . Shot-gun MS identification of peptides derived from bound proteins to Chm7-GFP now revealed specific interactions with several ESCRTs including Snf7 and Vps36 ( Figure 3B ) . Most interestingly , dozens of spectra specific for Heh1 were identified . Considering Heh1 is a low abundant integral membrane protein ( measured to be as low as 428 molecules/cell; Kulak et al . , 2014 ) , this result was particularly striking . In addition , another low abundant ( ~354 molecules/cell; Kulak et al . , 2014 ) integral INM protein , Nur1 was also detected . As Nur1 is known to interact with Heh1 within the CLIP ( chromosome linkage INM proteins ) complex ( Mekhail et al . , 2008 ) , these data suggest that Chm7 engages Heh1 within a broader INM platform , at least in the context of cells lacking VPS4 . Of note , no components of the NPC were specifically detected , nor was Heh2 . We next tested binding partners of chm7OPEN-GFP ( Figure 3—figure supplement 1B ) , which also provides a potential mimic of the physiological circumstances when Chm7 is recruited to the nuclear envelope . In this case , Heh1 was the top hit ( Figure 3C ) . In addition to Nur1 , other members of CLIP were also specifically identified including Lrs4 and Csm1 . Curiously , Gsp1 ( budding yeast Ran ) was also found ( Figure 3C ) . Further , alongside Snf7 , other ESCRT-IIIs including Vps2 , Vps24 and Did2 were detected ( Figure 3C ) . In contrast to bound proteins purified with Chm7-GFP in the vps4Δpom152Δ cells , we did not detect any specific peptides for ESCRT-II subunits . We surmise this is likely because Chm7-GFP can be seen in cytosolic foci in vps4Δ cells ( See Webster et al . , 2016 and Figure 5A ) , whereas chm7OPEN-GFP exclusively localizes to the nuclear envelope . As a further test of the specificity of the interactions between chm7OPEN-GFP and integral INM proteins , we observed the near-quantitative accumulation of both Heh1 and Nur1 fluorescent fusion proteins ( produced at endogenous levels ) at the chm7OPEN focus , while the distribution of Heh2 was unaltered ( Figure 3D ) . We noted that although we detected several additional ESCRT-III proteins in the affinity purifications of chm7OPEN-GFP , we did not detect any peptides for Vps4 . Thus , we investigated whether a functional Vps4-GFP fusion ( Adell et al . , 2017 ) could also be specifically recruited to the chm7OPEN focus at the nuclear envelope , which it was in virtually all cells ( Figure 3E ) . The fluorescence intensity of Vps4-GFP could be correlated to that of the chm7OPEN-mCherry focus , suggesting a close relationship between the number of Chm7 and Vps4 molecules recruited to this nuclear envelope site ( Figure 3—figure supplement 1C ) . We therefore next assessed the molecular determinants of Vps4-GFP recruitment to the chm7OPEN site by first focusing on deleting the genes encoding ESCRT-III subunits found in the chm7OPEN-GFP affinity purifications including SNF7 , VPS24 , VPS2 , and DID2 ( Figure 3C ) . In all these deletion strains , Vps4-GFP recruitment to chm7OPEN-mCherry foci was reduced or , in the case of snf7Δ cells completely eliminated , with minimal impact on the accumulation of chm7OPEN-mCherry itself ( Figure 3E; Figure 3—figure supplement 1D , E ) ; although we noted that the average area encompassed by the chm7OPEN-mcherry focus was increased in snf7Δ cells ( Figure 3E; Figure 3—figure supplement 1F ) . These data support a model in which Vps4 may be recruited to the nuclear envelope using a similar cohort of ESCRT-III interactions as those observed at endosomes ( Babst et al . , 2002 ) . Consistent with this idea , we also tested ist1Δ and vps60Δ cells , which encode proteins that impact Vps4 recruitment ( Dimaano et al . , 2008; Rue et al . , 2008 ) and ATPase activation ( Azmi et al . , 2008; Yang et al . , 2012 ) at endosomes . In both of these strains , there was a modest decrease of Vps4-GFP fluorescence at the chm7OPEN-mCherry foci ( Figure 3E; Figure 3—figure supplement 1D ) . In marked contrast to the disruption of ESCRT-III components that act downstream of Snf7 , deletion of VPS20 , which acts upstream of Snf7 at endosomes ( Teis et al . , 2008 ) and is absent from the nuclear envelope ( Webster et al . , 2014; Webster et al . , 2016 ) , we observed a remarkable ~3 fold increase of Vps4-GFP at the nuclear chm7OPEN focus ( Figure 3E , Figure 3—figure supplement 1D ) . These data suggest that in the absence of the endosome ESCRT arm , there is a larger pool of Vps4 that is available to interact with the chm7OPEN network at the nuclear envelope . Importantly , the total Vps4-GFP protein levels were not notably altered in any of the ESCRT deletion backgrounds tested ( Figure 3—figure supplement 1G ) . Together , these data reinforce that there are unique components of the endosome and nuclear envelope ESCRT pathways , but Vps4 can nonetheless be recruited to the INM alongside or downstream of Snf7 . Interestingly , as shown in Figure 3D , chm7OPEN is able to shift the distribution of both Heh1 and Nur1 from an evenly-distributed nuclear peripheral localization to one that is co-localized with the chm7OPEN focus . This raised the formal possibility that Chm7 recruitment to the nuclear envelope might in fact be independent of and precede binding to Heh1; in such a model Heh1 would be required for its focal accumulation , which we interpret to be ‘activation’ . Thus , this result illustrated the need to develop a better controlled experimental system where the mechanism of Chm7 recruitment and activation can be decoupled . We thus conceived of an experimental approach that we termed the Fluorescent ESCRT Targeting and Activation Assay ( FETA; Figure 4A ) . FETA exerts both temporal control over the expression of Chm7-GFP ( through the GAL1 promoter ) in addition to spatial control over its recruitment to the nuclear envelope by binding to a GFP-nanobody ( GFP-binding protein/GBP ) appended to Heh2 . Thus , without any perturbations to the nuclear envelope barrier , we can monitor the recruitment and activation of Chm7-GFP at the nuclear envelope , the latter of which we interpret as the local clustering of Chm7 as the most logical visual outcome of Chm7 polymerization at the level of fluorescence microscopy . By shifting cells to medium containing galactose , we induced the expression of Chm7-GFP and monitored its distribution by timelapse microscopy . As shown in Figure 4B , Chm7-GFP was first observed in the cytosol but accumulated at the nuclear envelope within 20 min . Importantly , this nuclear envelope binding was due to its interaction with Heh2-GBP-mCherry as overexpression of Chm7-GFP in strains lacking Heh2-GBP-mCherry ( Figure 1C ) or lacking GBP ( Figure 4—figure supplement 1A , B ) did not lead to nuclear envelope accumulation or clustering . In contrast , Heh1-mCherry was incorporated into the Chm7-GFP-Heh2-GBP foci ( Figure 4—figure supplement 1C ) . Interestingly , nearly simultaneously with the broader nuclear envelope-localization , Chm7-GFP and Heh2-GBP-mCherry accumulated in multiple foci throughout the nuclear envelope ( see arrowheads at 40 min ) . These foci coalesced into one or two foci/cell over the length of the timecourse ( 90 min; Video 1 ) . As a means to quantify this focal accumulation , we calculated a coefficient of variation ( CV ) of the mCherry fluorescence along the nuclear envelope in a mid-plane , which we plotted over time ( Figure 4C ) . This approach faithfully represented the observed clustering , which reached a maximum value between 50 and 60 min ( Figure 4B , C ) . With the ability to temporally resolve recruitment and ‘activation , ’ we next interrogated how Heh1 impacted these steps . Strikingly , the induction of Chm7-GFP expression in heh1Δ cells led to its accumulation at the nuclear envelope at a similar timepoint as in wildtype cells , however , we did not observe any focal accumulation with a CV remaining at ~1 over the 90 min timecourse ( Figure 4B , C , and Video 1 ) . Thus , Chm7 recruitment to the INM is not sufficient to lead to Chm7-GFP clustering . Instead , activation requires Heh1 . We next investigated whether other Chm7-interacting partners influenced Chm7-GFP clustering in the FETA assay beginning with Nur1 . Interestingly , deletion of NUR1 led to a statistically-significant drop in the CV of Heh2-GBP-mCherry at the end point of the FETA assay suggesting it could also contribute to Chm7 activation ( Figure 5A , B ) . Consistent with this idea , we also observed considerably less chm7OPEN-GFP accumulation at the nuclear envelope in nur1Δ cells ( Figure 5—figure supplement 1A , B ) . However , we also noted that the total levels of Heh1 are reduced in nur1Δ cells ( Figure 5—figure supplement 1C ) , suggesting that this effect may be indirect and ultimately mediated through Heh1 . We further investigated the impact of deleting both SNF7 and VPS4 on the extent of Heh2-GBP-mCherry clustering in the FETA assay . In both cases , the CV of Heh2-GBP-mCherry was unaltered ( Figure 5B ) . Yet , qualitatively , we observed that there were more discrete foci at the nuclear envelope in both of these genetic backgrounds , in addition to some foci ( lacking Heh2-GBP-mCherry ) in the cytoplasm ( Figure 5A ) . Thus , these downstream components could impact other events that elude measurement and interpretation with this approach . The lack of Heh2-GBP-mCherry clustering in heh1Δ cells also provided a genetic background to more fully vet the mechanism of Chm7 activation . First , introduction of HEH1 on a plasmid rescued clustering of Chm7-GFP and Heh2-GBP-mCherry in the heh1Δ strains , confirming that lack of clustering was indeed due to the absence of Heh1 ( Figure 5C , D ) . In contrast , expression of the HEH1 paralogue , HEH2 , failed to rescue clustering supporting that there are unique sequence elements in Heh1 that interface with Chm7 ( Figure 5—figure supplement 1D ) . Consistent with this , expression of heh1 ( 1-735 ) , which lacks the C-terminal WH domain , failed to restore Chm7 focal accumulation and Heh2-GBP-mCherry clustering suggesting that the WH domain was required for Chm7 activation ( Figure 5C , D ) . Interestingly , however , the WH domain alone was insufficient to rescue clustering but instead it required a membrane anchor through a Heh1-transmembrane domain ( Figure 5C , bottom panel ) . In the latter case , we also observed Chm7-GFP foci in the cytoplasm , likely in ER membranes , as the heh1 ( 703-834 ) construct does not contain INM targeting sequences , consistent with data presented in Figure 2A . Thus , there is a clear coupling between the Heh1 WH domain and the membrane required for Chm7 activation . Our data support a model in which Chm7 and Heh1 are spatially segregated , but upon binding , Chm7 is locally activated at a membrane interface . To investigate how Chm7 activation could translate into a mechanism capable of sealing a nuclear envelope hole , we turned to a correlative light EM ( CLEM ) approach to investigate nuclear envelope morphology at sites of Chm7 activation . While our initial attempts focused on examining sites of Chm7-GFP localization in NPC assembly-defective strains that have nuclear envelope herniations like in apq12Δ and nup116Δ cells , the combination of the low abundance ( and transience ) of Chm7-GFP at these nuclear envelope foci coupled to the loss of GFP fluorescence through the freeze-substitution process precluded this as a viable approach . Thus , again , we turned to chm7OPEN-GFP ( and Chm7-GFP in vps4Δpom152Δ cells ) as proxies to interrogate the membrane morphology at sites of surveillance ( hyper ) activation . As shown in panel i of Figure 6A and B , we could effectively correlate fluorescence images and electron tomograms ( Kukulski et al . , 2012 ) , which showed the chm7OPEN-GFP foci apposed to the INM . Remarkably , this fluorescence demarked extensions of the INM that invade the nucleus and form a fenestrated network of INM cisterna – the lumen of the cisterna is continuous with the lumen of the nuclear envelope and is colored teal or purple to facilitate visualization . In Figure 6A and B , the panels ii-iv represent slices along the Z axis of the tomograms; 3D models were generated by isosurface rendering , which can be visualized as still frames ( Figure 6A , B , v–viii ) and in movies ( Videos 2 and 3 ) . These 3D views facilitate the observation of nuclear pores ( denoted by stars ) in addition to a perspective of the extent of the network of fenestrated cisternal membranes at the INM . Nearby , fenestrated membranes that we interpret to be ER also appeared with intriguing frequency ( Figure 6A , vii , viii; Figure 6B , iii; 6 out of 14 tomograms ) . The INM-associated cisternal network was often ( 11 out of 14 tomograms ) found underneath balloon-like herniations of the nuclear envelope ( i . e . both the INM and ONM ) that extended several hundred nanometers into the cytosol ( Figure 6A , B ) . The lumen of the herniations were open to the nucleoplasm and tapered into ~45 nm-diameter membrane ‘necks’: in the image shown in Figure 6A , ii , two ‘necks’ can be observed ( black arrowheads ) . Often ( in 10 out of 14 tomograms ) vesicles appear nearby the sites of herniations some of which can be seen either fusing with , or fissioning from , the ONM ( Figure 6 , white arrowheads ) . Additional INM evaginations ( extending into the nuclear envelope lumen , black arrowheads ) are observed in tandem arrays that suggest they may be precursors of the herniations ( Figure 6A , iii ) . Indeed , often several nuclear envelope herniations can be observed in a single tomogram , as indicated by white arrows in Figure 6B . Interestingly , these membrane deformations could also be formed in the absence of SNF7 ( Figure 6—figure supplement 1A , B , Video 4 ) . Similar membrane morphology was also observed when CLEM was applied to the focal accumulation of Chm7-GFP in vps4Δpom152Δ cells ( Figure 7A , i , ) . Consistent with the idea that the INM expansion and nuclear envelope herniation need not occur simultaneously , in Figure 7A an example of a single herniation ( with two necks; see black arrow heads ) is shown . A nuclear-perspective ( bottom-up , Figure 7 , v ) view allows a direct comparison between the herniation neck at the Chm7-GFP signal and nuclear pores that would be filled with NPCs ( stars ) . Lastly , a more dramatic example of a vps4Δpom152Δ nuclear envelope where connections between the INM and a cisternal , lamellar membrane with multiple deformations is presented in Figure 7B . Interestingly , in this thick section , no nuclear envelope herniation is observed suggesting that there is no implicit link between the INM network and nuclear envelope herniation; these two morphologies might arise stochastically and are not necessarily directed in one , or the other , direction . Taken together , these data support a model in which the interaction of Chm7 and Heh1 , and Chm7 activation , can lead to expansion of the INM and the formation of nuclear envelope herniations . It was tempting to speculate that the nuclear envelope herniations that we observed under conditions of Chm7 activation were directly analogous to those observed in genetic backgrounds where NPC assembly is perturbed , like in apq12Δ and nup116Δ cells . To perform a direct comparison , we first confirmed that , as in nup116Δ cells ( Wente and Blobel , 1993 ) , NPC-like structures were found at the bases of the nuclear envelope herniations seen in cells lacking APQ12 ( Scarcelli et al . , 2007 ) by staining thin sections with the MAb414 antibody that recognizes several FG-nups . As shown in Figure 8A , gold particles that label the MAb414 antibody were specifically found at the bases of these herniations confirming that they emanate from structures with nups . Furthermore , the diameter of the bases of these herniations averaged 78 nm , which while statistically similar to those found at mature NPCs ( mean of 87 nm ) , were considerably larger than the ~45 nm diameter openings found at the necks of herniations caused by Chm7 ( Figure 8B , Video 5 ) . Lastly , the lumen of the herniations associated with both apq12Δ ( Figure 8A , C ) and nup116Δ cells ( Figure 8—figure supplement 1A , Video 6 and Wente and Blobel , 1993 ) are filled with electron density , whereas those associated with Chm7 appear to be empty ( Figure 6A , B ) . Thus , we suggest that the herniations associated with overactive Chm7 and those associated with NPC assembly are morphologically distinct . That activated Chm7 might drive membrane expansion and nuclear envelope herniations with unique characteristics to those found in NPC assembly mutants does not exclude the possibility that Chm7 might nonetheless contribute to the formation of both of these herniation types . We therefore next investigated whether deletion of CHM7 impacted the prevalence of herniations in apq12Δ cells . As shown in Figure 8B and D , deletion of CHM7 had little impact on the number of herniations observed in thin sections of apq12Δ cells , which were only modestly reduced ( 23% versus 31% of nuclei; Figure 8B ) . Most strikingly , however , we observed that 35% of the nuclei in the thin sections of apq12Δchm7Δ cells ( Figure 8B , E and Figure 8—figure supplement 2 ) had large ( >500 nm ) discontinuities in their nuclear membranes , which suggests that these nuclei were unstable and could rupture ( labeled as nuclear envelope rupture/NER ) . Similar nuclear envelope discontinuities were observed in apq12Δsnf7Δ strains ( Figure 8—figure supplement 3 ) . Indeed , in some cases we could observe nucleoplasm escaping into the cytosol ( Figure 8E , left panel ) . This result provides an explanation for the striking loss of NLS-GFP reporter accumulation in the nucleus that was observed in only ~35% of apq12Δchm7Δ cells ( Webster et al . , 2016 ) . Thus , while these NPC-assembly-associated nuclear envelope herniations might not require CHM7 or SNF7 for their biogenesis , ESCRTs are nonetheless required to maintain the integrity of the nuclear membranes in the context of these herniations .
While there has been considerable focus over the last few decades on mechanisms that control the targeting of proteins and lipids to distinct intracellular compartments , it is equally important to understand the protective mechanisms that maintain this compartmentalization in the face of challenges to membrane integrity and/or the specific biochemical identity of organelles . Here , we further explore the mechanism of ESCRT surveillance of the nuclear envelope . We interpret our data in a model where the nuclear envelope is surveilled by two principle components , the ESCRT Chm7 and the integral INM protein , Heh1 . This surveillance system appears to be set up to respond directly to perturbations in the nuclear envelope barrier in a way that we suggest is agnostic as to whether the perturbation is a result of defectively formed NPCs or a mechanical ( or other ) disruption of the nuclear membranes . The rationale behind this assertion is that the nuclear envelope surveillance system is itself directly established by a functioning nuclear transport system , which physically segregates Chm7 and Heh1 on either side of the nuclear envelope . For example , prior work has shown that Heh1 requires the function of the NTRs Kap-α and Kap-β1 in addition to the Ran-GTPase in order to be actively targeted to the INM through NPCs ( King et al . , 2006 ) . Here , we establish that Chm7 , while small enough to passively leak through the NPC diffusion barrier , is actively exported by the major export NTR , Xpo1 ( Figure 1 ) . Therefore , any perturbations that impact active nuclear transport or the diffusion barrier across the nuclear envelope would lead to increased Chm7 diffusion into the nucleus and/or a deficit in its nuclear export increasing the likelihood that it meets Heh1 . While similar perturbations could also lead to Heh1 mistargeting and/or its diffusion into the ONM , we suspect that this would be kinetically slower than Chm7 diffusion into the nucleus as Heh1 is bound to chromatin at the INM ( Grund et al . , 2008; Mekhail et al . , 2008; Gonzalez et al . , 2012; Yam et al . , 2013; Barton et al . , 2015; Schreiner et al . , 2015 ) . Indeed , it is probable that Heh1 functions in two major roles with respect to nuclear integrity: first , it provides mechanical stability to the nucleus by binding chromatin , and second , provides a binding site for Chm7 through its C-terminal WH domain . These two inter-related roles could also help to explain why both the N- and C-terminal domains of Heh1 are required to maintain viability of apq12Δ cells ( Figure 2C ) , and why HEH1 is generally more essential in budding and fission yeasts compared to CHM7 . This likely holds true in mammalian models as well as LEM2 is essential whereas CHMP7 is dispensable for viability ( in cell culture ) ( Hart et al . , 2015 ) . Once a perturbation occurs to the nuclear envelope barrier through defects in NPC assembly , loss of function of NPCs , or mechanical disruption of the nuclear membranes , Chm7 and Heh1 are able to come together . While we have so far been unsuccessful in reconstituting a direct biochemical interaction between Chm7 and Heh1 ( although others have between LEM2 and CHMP7 [Gu et al . , 2017] ) , we have nonetheless provided evidence that this interaction likely leads to Chm7 activation , thus tightly coupling recruitment and activation ( Figure 4 ) . Our data support a model in which the C-terminal WH domain of Heh1 in addition to a membrane anchor are necessary and sufficient for this activation event . As our prior work ( Webster et al . , 2016 ) and that from Olmos et al . ( 2016 ) , support that it is the N-terminal ESCRT-II-like domain of Chm7 ( which , interestingly , is also predicted to be made up of tandem WH domains [Horii et al . , 2006; Bauer et al . , 2015; Figure 1A] ) that is necessary for recruitment to the nuclear envelope , it seems likely that the binding between the Heh1-WH domain and the N-terminus of Chm7 could trigger activation by removing some form of autoinhibition , which is a common theme among ESCRT-III proteins ( Zamborlini et al . , 2006; Shim et al . , 2007; Lata et al . , 2008; Bajorek et al . , 2009; Henne et al . , 2012; Tang et al . , 2015; Tang et al . , 2016 ) that ensures polymerization in the correct compartment . Clearly , the precise molecular mechanism of Chm7 activation by Heh1 will require structural insight , which is no doubt on the horizon . Even with structural information , there are many additional factors that need to incorporated into a nuclear envelope surveillance mechanism . For example , future work must directly address how Heh2 fits into this pathway . While Heh2 has a similar domain architecture as Heh1 including a WH domain ( King et al . , 2006 ) , it does not appear to impact Chm7 activation within the FETA assay ( Figure 5—figure supplement 1D ) , nor was it detected in any of the affinity purifications ( Figure 3 ) . This was surprising , as we had previously established that the N-terminus of Heh2 ( that lacks the WH domain ) , can directly bind to both Chm7 and Snf7 , at least in their ‘open’ forms ( Webster et al . , 2016 ) . Moreover , we reported physical interactions between Chm7 and Snf7 with Heh2 in vivo ( Webster et al . , 2014; Webster et al . , 2016 ) . While there are many possible explanations for these results , we favor models that consider Heh2 in a regulatory role that might modulate Chm7 and/or Snf7 function at the nuclear envelope , although this remains speculative and awaits direct experimentation . Additional downstream ( of Chm7 and Heh1 ) components also need to be figured into any nuclear envelope surveillance mechanism including Snf7 , but also Did2 , Vps24 and Vps2 ( Figure 3B , C ) . That these ESCRT-IIIs impact the recruitment of Vps4 to the nuclear envelope ( Figure 3E ) raises the possibility that the sequence of their recruitment might be similar to that found at ILVs during MVB formation , which begins with Snf7 and ends with Did2 , Vps24 , and Vps2 ( Babst et al . , 2002; Teis et al . , 2008 ) , the latter of which play important roles in Vps4 recruitment and ATPase activation required for membrane scission ( Azmi et al . , 2008; Shestakova et al . , 2010; Schöneberg et al . , 2018 ) . Interestingly , however , while we observe Vps4 recruitment to sites of chm7OPEN at the INM ( Figure 3E ) , we do not have any explicit evidence for membrane scission per se . Indeed , the INM evaginations and nuclear envelope herniations appear to be stabilized by ~45 nm diameter necks ( Figures 6 , 7 and 8B ) . That these necks have similar diameters to those observed at the plasma membrane or endosomes when HIV budding ( von Schwedler et al . , 2003; Morita et al . , 2011; Cashikar et al . , 2014; Jackson et al . , 2017 ) or ILV formation ( Adell et al . , 2014; Buono et al . , 2017; Frankel et al . , 2017; Wenzel et al . , 2018 ) , respectively , are stalled , is suggestive that they are stabilized by a similar ESCRT-III-like polymer . We suggest that the most likely explanation is that Chm7 itself might play a unique role in a nuclear membrane-scission step , perhaps through its potential MIM1 domain that might directly recruit Vps4 but is absent from the chm7OPEN construct . Such a conclusion is supported by the presence of similar INM evaginations and nuclear envelope herniations observed in the presence of full length Chm7 but lacking Vps4 ( Figure 7 ) . Any scission mechanism that requires a direct interaction between the MIM1 domain of Chm7 and Vps4 will also need to consider the overlap of this sequence with NES2 ( Figure 1 ) . This clearly predicts a competition between Xpo1 and Vps4 for Chm7 , which provides a potential entryway for an additional level of regulation for any polymer formation or scission reaction by this NTR ( that would also be modulated by Ran-GTP ) . Indeed , NTRs might regulate this pathway at the level of Heh1 as well . Heh1 is synthesized and inserted into ER membranes so is exposed to the cytosol and would thus be capable of binding to Chm7 before it is targeted to the INM . Such an assumption is supported by the observation that the exposure of the WH domain of Heh1 is sufficient to recruit Chm7 to ER membranes ( Figure 2 ) . Does NTR-binding to the NLSs in the N-terminus of Heh1 inhibit Chm7 recruitment in this compartment ? Answering this question will also be relevant in mammalian systems where LEM2 and CHMP7 are found in the same compartment during mitotic nuclear envelope breakdown suggesting the need for mechanisms to prevent their binding in this phase of the cell cycle . Interestingly , recent work is suggesting that proteins like Lgd/CC2D1B help control the spatiotemporal timing of CHMP4B and CHMP7 activity in mammalian cells ( Ventimiglia et al . , 2018 ) , supporting the existence of regulatory mechanisms of this pathway . Lgd/CC2D18 acts during nuclear envelope reformation when ESCRTs play critical roles in sealing the nuclear envelope at the end of mitosis , often at sites where spindle microtubules perforate the nascent nuclear envelope ( Olmos et al . , 2015; Olmos et al . , 2016; Vietri et al . , 2015 ) . It will be interesting to understand the similarities and differences between how ESCRTs function at the end of mitosis ( in an open mitosis ) versus events that are triggered upon acute disruption of nuclear envelope integrity or NPC function during interphase . For example , two of the more interesting observations of the EM tomographic analyses are the generation of intranuclear membrane invaginations and the proximity of ER sheets and vesicles at the ONM sites of nuclear envelope herniations ( Figure 6 , Figure 6—figure supplement 1 , Figure 7 ) . Both of these observations suggest that nuclear envelope surveillance might be coupled to the delivery of new membrane at sites of rupture . It seems plausible that sealing holes in the nuclear envelope ( particularly those larger than a nuclear pore ) would require the local delivery of membranes . Similarly , NPC quality control mechanisms that suggest the sealing of defective NPCs would likely require some form of expansion of the pore membrane , as has been proposed ( Wente and Blobel , 1993 ) . Whether such membrane is derived from new synthesis or from the mobilization of existing stores remains to be explored . Of note , recent work supports that even the INM may be metabolically active and have the capacity to generate new lipid locally ( Romanauska and Köhler , 2018 ) . Further , close examination of sites of chm7OPEN accumulation reveals that the morphology of the INM extensions are both sheet-like and tubular in nature and resemble the ‘normal’ continuities between the ONM and the broader ER network ( West et al . , 2011 ) ; together these data suggest that ER-shaping proteins and lipid synthesis pathways might also play an important role in contributing to nuclear envelope sealing - topics for the future . Both ER remodeling proteins and lipid synthesis pathways are required for de novo NPC biogenesis ( Schneiter et al . , 1996; Dawson et al . , 2009; Hodge et al . , 2010 ) . As NPC biogenesis proceeds through an INM evagination step , we also must consider whether the INM evaginations observed in the electron tomograms reflect a function for Chm7 in this process . However , while this remains a compelling hypothesis , our data do not yet support this idea . For example , the herniations observed in the context of inhibiting NPC assembly are morphologically distinct from those found in the chm7OPEN or vps4Δpom152Δ scenarios with wider necks that contain nups ( Figure 8A , B ) . Further , they arise in the absence of CHM7 and are thus more likely to be formed due to a defect in NPC biogenesis ( perhaps because of an inhibition of INM-ONM fusion ) , or , through the triggering of a NPC assembly surveillance mechanism ( Webster et al . , 2016; Thaller and Lusk , 2018 ) . Should Chm7 be a critical component of the latter mechanism , it would be predicted that the nuclear envelope herniations associated with nups would be unsealed without CHM7 . Indeed , that we observe often-dramatic openings in the nuclear envelope in apq12Δchm7Δ cells ( Figure 8D and Figure 8—figure supplement 2 ) , suggest that the herniations themselves might be prone to rupture with Chm7 required for their repair . It follows then that like in mammalian cells where much larger nuclear envelope herniations are precursors to nuclear rupture ( De Vos et al . , 2011; Vargas et al . , 2012; Hatch et al . , 2013; Denais et al . , 2016; Hatch and Hetzer , 2016; Raab et al . , 2016 ) , these smaller NPC-assembly associated herniations might also impact nuclear envelope integrity through mechanisms that remain to be fully defined . In either case , it reinforces the concept that the assembly of NPCs can be perilous , and it will be important to consider this possibility when interpreting the underlying pathology of human diseases that are associated with defects in NPC function or assembly , for example , DYT1 early-onset dystonia ( Laudermilch et al . , 2016; Pappas et al . , 2018 ) or Steroid Resistant Nephrotic Syndrome ( Miyake et al . , 2015; Braun et al . , 2016; Braun et al . , 2018 ) . Lastly , while we acknowledge then that the INM evaginations that we observe in cases of Chm7 hyperactivation might not necessarily be a physiological event in a nuclear envelope sealing process , it remains tempting to speculate that they might be in the context of proposed mechanisms of nuclear egress be it Mega-RNPs ( Speese et al . , 2012; Jokhi et al . , 2013 ) , viruses ( Lee et al . , 2012; Lee et al . , 2016; Arii et al . , 2018 ) , or nucleophagy ( Roberts et al . , 2003; Dou et al . , 2015; Mochida et al . , 2015; Mostofa et al . , 2018 ) that so-far remain obscure but would nonetheless require a membrane scission step . Interestingly , recent work suggests that herpes virus nuclear egress requires ESCRTs ( Arii et al . , 2018 ) , including a role in controlling INM extensions into the nucleus; such intranuclear membrane might also be relevant in cell types that have so-called ‘nucleoplasmic reticulum’ ( Malhas et al . , 2011 ) . The biogenesis and function of nucleoplasmic reticulum remain enigmatic but our observations of intranuclear fenestrated membrane emanating from the INM might suggest a yet-to-be discovered role for Chm7 and Heh1 in forming such structures as well .
All strains used in this study are from a W303 parent; their derivation and genotypes are listed in Supplementary file 1 . Fluorescent protein tagging and gene deletions were generated using a PCR-based integration approach using the pFA6a plasmid series ( Supplementary file 2 ) as templates ( Longtine et al . , 1998; Van Driessche et al . , 2005 ) . Standard yeast protocols for transformation , mating , sporulation , chromosomal DNA isolation and tetrad dissection were followed ( Amberg et al . , 2005 ) . Cells were grown to mid-log phase in YPA ( 1% Bacto yeast extract ( BD ) , 2% Bacto peptone ( BD ) , 0 . 025% adenine hemi-sulfate ( Sigma ) ) or complete synthetic medium ( CSM ) supplemented with 2% raffinose ( R; BD ) , 2% D-galactose ( G; Alfa Aesar ) or 2% D-glucose ( D; Sigma ) as indicated . To compare relative growth rates of heh1Δapq12Δ strains expressing HEH1 alleles ( DTCPL1498 , DTCPL1517 , DTCPL1581 , DTCPL1519 , DTCPL1520 ) roughly equivalent cell numbers from overnight cultures grown in in YPAR were spotted in 10-fold serial dilutions onto YPG to induce expression of Heh1 or indicated truncations and imaged after 36 hr at 30oC . To test the impact of inhibiting nuclear export on the steady state distribution Chm7-GFP , Chm7-MGM4 , chm7OPEN-GFP , GFP , NES1CHM7-GFP , NES2CHM7-GFP , or NES1-2CHM7-GFP , we used KWY175 ( a gift from B . Montpetit and Karsten Weis ) in which the genomic deletion of XPO1 is covered with pRS413 expressing the xpo1-T539C allele that confers sensitivity to Leptomycin B ( LMB ) . These strains were grown in YPAR and galactose ( final concentration of 1% ) was added to the growth medium to induce the expression of the GFP fusion proteins for 2 hr before the addition of 2% D-glucose to repress protein production . Cultures were then treated with 50 ng/mL LMB dissolved in 7:3 MeOH:H2O solution ( Sigma ) for 45 min alongside a control of the equivalent volume of MeOH before imaging . To test if Ca2+ plays a role in the physiological recruitment of Chm7 to the nuclear envelope , apq12Δ cells expressing Chm7-GFP ( DTCPL567 ) were cultured overnight at 30°C , diluted to an OD600 of 0 . 2 and grown for an additional 2 hr at RT . Cells were treated with either 25 µM ( Li et al . , 2011 ) of the cell permeable calcium chelater BAPTA-AM ( Tocris Bioscience ) dissolved in DMSO or DMSO alone for 30 min followed by a 45 min incubation at either RT or 37°C before imaging . All plasmids are listed in Supplementary file 2 . To generate pRS406-ADH1-GFP , the GFP coding sequence was amplified by PCR and inserted into pRS406-ADH1 ( p406ADH1 was a gift from Nicolas Buchler and Fred Cross - Addgene plasmid # 15974; http://n2t . net/addgene:15974; RRID:Addgene_15974 ) using EcoRI and HindIII restriction sites . To generate pCHM7-MGM4 , the CHM7 ORF was amplified by PCR with ClaI restriction sites and subcloned in pPP004 linearized with ClaI ( gift of L Veenhoff; Popken et al . , 2015 ) , placing CHM7 in between the first MBP gene and the GFP . To generate pRS406-ADH1-NES1-GFP a Geneblock ( IDT ) was synthesized with the coding sequence for amino acids 370–390 of Chm7 with homology arms containing 40 base pairs flanking sequences outside of the multiple cloning site in pRS406-ADH1-GFP . The geneblock was assembled into pRS406-ADH1-GFP using the Gibson Assembly reaction ( New England BioLabs ) . To generate pRS406-ADH1-NES2-GFP , complimentary 4 nmol Ultramers ( IDT ) were designed to code for amino acids 409–424 of Chm7 with overhangs that would be generated with XbaI or BamHI . Ultramers were annealed by heating to 95°C for 5 min in Taq polymerase PCR buffer ( Invitrogen ) and allowing them to slowly cool to RT . Annealed primers were then ligated using T4 ligase ( Invitrogen ) into pRS406-ADH1-GFP linearized with XbaI/BamHI ( New England BioLabs ) . To generate pRS406-ADH1-NES1-2CHM7-GFP , the sequence encoding amino acids 370–429 of Chm7 was amplified with oligonucleotide primers containing the XbaI or BamHI restriction sites . The PCR product was digested with XbaI/BamHI and gel purified ( Qiagen ) before ligation with T4 ligase ( Invitrogen ) into pRS406-ADH1-GFP linearized with XbaI/BamHI . Gibson Assembly ( New England BioLabs ) was used to generate pFA6a-3xHA-GFP-his3MX6 for functional tagging of Vps4 ( Adell et al . , 2017 ) . The 3xHA epitope was PCR-amplified from a pFA6a-3xHA-his3MX6 ( Longtine et al . , 1998 ) plasmid using Q5 DNA polymerase ( New England BioLabs ) and assembled into pFA6a-GFP-his3MX6 ( Longtine et al . , 1998 ) digested with SalI and PacI ( New England BioLabs ) . For whole cell protein extracts , approximately 2 OD600 of cells in mid log phase were collected by centrifugation , washed in 1 mM EDTA , pelleted again and resuspended/lysed in 2 M NaOH for 10 min on ice . Proteins were precipitated by the addition of 50% Trichloroacetic acid for 20 min on ice and then collected by centrifugation . The precipitated proteins were washed in ice-cold acetone , air dried and then resuspended in SDS-PAGE sample buffer . Samples were then denatured at 95°C for 5 min . Denatured proteins were separated on precast SDS-PAGE , 4–20% gradient gels ( BioRad ) and transferred to 0 . 2 µm nitrocellulose membranes ( BioRad ) . Relative protein loading was visualized using Ponceau S Solution ( Sigma ) . Membranes were subsequently washed in TBST and blocked for 1 hr in 5% skim milk in TBST at RT . Membranes were then incubated with HRP-conjugated anti-HA ( Roche 3F10 ) , or anti-actin ( mAbcam 8224 ) diluted in TBST . Primary antibodies were detected directly with ECL ( ThermoFisher ) or with anti-rabbit HRP-conjugated secondary antibodies , followed by ECL and visualized using a VersaDoc Imaging System ( Bio-Rad ) . With the exception of the correlative light electron microscopy experiments described below , all fluorescence micrographs were acquired using a DeltaVision microsope ( Applied Precision/GE Healthcare ) fitted with a 100x , 1 . 4 NA objective ( Olympus ) . Images were taken using a CoolSnapHQ2 CCD camera ( Photometrics ) , with the exception of those in Figure 2 and Figure 4B , which were acquired using a Evolve EMCCD camera ( Photometrics ) . For timecourse assessment of Heh2-GBP-mCherry clustering in the FETA assay in Figure 4B , cells were imaged in microfluidic plates ( Y04C/CellASIC ) in the ONIX microfluidic platform ( CellASIC ) . Cells were loaded into the microfluidic chamber in CSM with 2% raffinose . CSM with 2% galactose was perfused into the microfluidic chamber at 0 . 25 psi for the course of the experiment . Z-stacks ( 0 . 4 µm sections ) were acquired for 90 min at 10 min intervals . All presented fluorescent micrographs were deconvolved using an iterative algorithm in softWoRx ( 6 . 5 . 1; Applied Precision GE Healthcare ) . Unprocessed images after background subtraction were used for quantification of fluorescence intensities . Assessment of Heh2-GBP-mCherry or Heh2-mCherry clustering was quantified by calculating the coefficient of variation ( SD/mean x 100 ) of fluorescence of individual nuclear envelopes ( Fernandez-Martinez et al . , 2012 ) . A four pixel wide , freehand line was traced over the entire nuclear envelope in a mid-plane section using FIJI/ImageJ ( Schindelin et al . , 2012 ) and the mean fluorescence contained in the traced area was measured . Cells with vacuolar autofluorescence that obscured fluorescence at the nuclear envelope were excluded from quantification . Further , in some heh1Δ cells Heh2-GBP-mCherry was found in plaque-like clusters before Chm7-GFP induction . Thus , as this phenotype was not triggered by expression of Chm7-GFP , they were excluded from the clustering analysis . To correlate the fluorescence intensity of co-localized Vps4-GFP and chm7OPEN-mCherry , the integrated density of Vps4-GFP and chm7OPEN-mCherry was measured and plotted on a correlation curve . The linear correlation coefficient ( Pearson coefficient , r ) was calculated in Prism ( GraphPad 8 . 0 ) . Similarly , quantification of the integrated density and average fluorescence intensity of Vps4-GFP and chm7OPEN-mCherry were measured by selecting a region of interest ( ROI ) around the chm7OPEN-mCherry signal and measuring average fluorescence intensity in both mCherry and GFP channels . To measure relative nuclear exclusion of GFP , NES1CHM7-GFP , NES2CHM7-GFP , and NES1-2CHM7-GFP constructs at steady state , a 3 . 75 µm line was traced across each nucleus ( encompassing cytoplasm both times the line crosses the nuclear envelope border ) as determined from the dsRed-HDEL localization . GFP fluorescence was measured using the Plot Profile function in FIJI/ImageJ ( Schindelin et al . , 2012 ) . Traces were normalized to the maximum value measured within each trace before averaging . Graphs and statistical analyses were generated using Prism ( GraphPad 8 . 0 ) . P-values in all graphs were generated with tests as indicated in figure legends and are represented as follows: ns , p>0 . 05; *p≤0 . 05; **p≤0 . 01 ***p≤0 . 001 , ****p≤0 . 0001 . All error bars represent the standard deviation from the mean . Scatter plots of spectral counts from MS/MS analysis for Figure 3A , B and C were generated using Excel ( Microsoft ) . Xpo1/Crm1 NES sequences were predicted using LocNES ( Xu et al . , 2015 ) with default threshold settings . GST , GST-Chm7 and GST-heh1 ( 735-834 ) ( the Heh1 WH domain ) proteins were recombinantly produced and purified as previously described ( Webster et al . , 2016 ) in lysis buffer ( 50 mM Tris pH 7 . 4 , 500 mM NaCl , 2 mM MgCl2 , 2 mM CaCl2 , 10% glycerol , 0 . 5% NP-40 , 1 mM DTT , complete protease inhibitors ( Roche ) ) . The soluble fraction was incubated with glutathione sepharose ( GT ) beads for 1 hr at 4°C for binding . The GT beads were collected by centrifugation and washed thrice with lysis buffer . GST and GST-Chm7 proteins were eluted from GT beads by 10 mM reduced glutathione and dialyzed in lysis buffer . The Heh1 WH was cleaved from GT beads by incubating with HRV3C protease ( Thermo Scientific ) at 4°C overnight . For binding experiment , Heh1 WH was incubated with GST and GST-Chm7 in a dialysis cassette ( 3 . 5K Slide-A-Lyzer , Thermo Scientific ) and the binding reaction was dialyzed overnight in a binding buffer ( 50 mM Tris pH 7 . 4 , 150 mM NaCl , 2 mM MgCl2 , 2 mM CaCl2 , 10% glycerol , 0 . 5% NP-40 , 1 mM DTT ) . The reaction was collected , incubated with GT beads for 1 hr at 4°C , washed thrice with binding buffer and eluted in 2X Laemmli sample-buffer . Proteins were resolved on a SDS-PAGE gel and visualized by SimplyBlue Safe-Stain ( Invitrogen ) . S . cerevisiae cells expressing Chm7-GFP ( DTCPL81 ) , Chm7-GFP vps4Δpom152Δ ( DTCPL133 ) , chm7OPEN-GFP ( DTCPL413 ) were grown to log phase in YPD at 30°C and collected by centrifugation . Cells were washed once with ice-cold water , collected by centrifugation and resuspended in a small volume of freezing buffer ( 20 mM HEPES , pH 7 . 4 , 1 . 2% polyvinylpyrrolidone and protease inhibitor [Sigma; Oeffinger et al . , 2007] ) and flash frozen in liquid nitrogen . The frozen yeast pellets were pulverized in a Retsch MM400 mixer mill for 6 times at 30 Hz for 3 min . For immunoaffinity purification , 200 mg of frozen , ground yeast powder was solubilized in 4 volumes of homogenization buffer ( 400 mM trisodium citrate , pH 8 , 0 . 5% n-Dodecyl β-D-maltoside ) and protease inhibitor cocktail ( Roche ) . The soluble fraction was incubated with 10 µl magnetic beads ( Dynabeads , M-270 Epoxy , Invitrogen ) slurry coated with GFP-nanobody for 1 hr at 4°C ( Cristea et al . , 2005; LaCava et al . , 2016 ) . The beads were collected on a magnetic rack and washed three times with 500 µl homogenization buffer . Bound proteins were eluted by incubating the beads in 20 µl 1X NuPAGE LDS ( lithium dodecyl sulfate ) sample buffer ( Invitrogen ) at 70°C for 10 min . Eluates were separated on a magnetic rack and further incubated with 50 mM DTT at 70°C for 10 min . The eluates were run on a 4–12% NuPAGE gel ( Novex ) until the dye front just entered the gel . The gels were stained with Imperial protein stain ( Thermo Scientific ) and a protein band ( consisting of all eluted proteins ) were excised for MS analysis . MS/MS was performed at the Yale Keck Proteomics facility . Excised bands described above were transferred to clean 1 . 5 mL Eppendorf tubes and digested with trypsin . Subsequently , chromatographic separation of peptides was done using a Waters nanoACQUITY ultra high pressure liquid chromatograph ( UPLC ) , and peptides were detected on a Waters/Micromass AB QSTAR Elite . Analysis of MS/MS peptide results was completed using Scaffold 4 . 8 . 7 ( Proteome Software Inc ) . Peptides were identified by SEQUEST and Mascot using X ! Tandem ( Craig and Beavis , 2003; Searle et al . , 2008 ) and validated using PeptideProphet ( Keller et al . , 2002; Nesvizhskii et al . , 2003 ) within Scaffold software ( Proteome Software Inc ) . Proteins were identified by comparison with SwissProt database where peptide identifications required ≥2 peptides from each replicate and ≥95 . 0% probability of correct identification to be included in analysis . Quantitative analysis to determine significance of enrichment between samples was done with total spectral counts from two replicates using Fischer’s exact test with a significance threshold p<0 . 05 ( Figure 3A , C ) , or on presence/absence from one replicate ( Figure 3B ) . Correlated fluorescence and electron microscopy were conducted as previously described ( Kukulski et al . , 2012; Curwin et al . , 2016 ) . In brief , yeast cells were high pressure frozen ( HPM010 , AbraFluid ) , freeze substituted ( EM-AFS2 , Leica ) with 0 . 1% uranyl acetate in acetone and infiltrated with Lowicryl . 300 nm sections were cut with a microtome ( EM UC7 , Leica ) and picked up on carbon coated 200 mesh copper grids . 50 nm TretraSpeck fluorescent microspheres ( fluorescence and electron dense fiducials , Life technologies , Carlsbad , CA ) were added to the grid for correlation . Grids used for Figure 6A , B , Figure 6—figure supplement 1A , B , Figure 8C , Figure 8—figure supplement 1A , and Figure 8—figure supplement 3A , were poststained with lead citrate to increase contrast . In all cases , 15 nm protein A-coupled gold beads were adsorbed on both sides of each grid and used as fiducial markers for overlaying high and low magnification tomograms . 60° to −60° tilt series were acquired on a Technai F30 ( Thermofisher , FEI ) at 300 kV with Serial-EM ( Mastronarde , 2005 ) at 20000x and either 3900x or 4700x to facilitate ease of correlation with TetraSpeck fiducials . To perform CLEM , fluorescence images were acquired of the EM grids on images a Nikon TI-E ( Figure 6A , B ) with sCMOS PCO edge 4 . 2 CL camera and solid state illumination , or an Olympus IX81 with MT20 ( Olympus ) lamp and CCD ( Orca-ER; Hamamatsu Photonics ) ( Figure 7A , B , Figure 6—figure supplement 1A , B ) . To distinguish protein fluorescence signal from fluorescent fiducials , for each field of view/grid four channels were acquired ( GFP , mCherry/RFP , Cy5 , and brightfield ) . Acquired images were further processed in FIJI using the Extended Depth of Field Plugin ( Forster et al . , 2004 ) . Correlation of fluorescence and reconstructed electron tomograms was performed using the ec-CLEM Plugin ( Paul-Gilloteaux et al . , 2017 ) in ICY ( de Chaumont et al . , 2012 ) . Alignment was determined by clicking on corresponding pairs of TetraSpeck fiducials in the two imaging modalities . Tomograms were reconstructed using the IMOD package ( Windows Version 6 . 2 ) and Etomo ( Version 4 . 9 . 8 , Kremer et al . , 1996 ) . Patch tracking function was used to perform a fiducial-less image alignment for reconstruction . 3DMOD software was used for manual segmentation of the tomograms . Further editing and annotation were done in Adobe Illustrator ( Adobe ) . Video sequences were compiled in 3DMOD and exported with further editing in ImageJ/FIJI ( Schindelin et al . , 2012 ) . Video frames were compressed as JPGs to reduce file size . To examine the ultrastructure of apq12Δ ( CPL1326 ) and apq12Δchm7Δ ( CPL1327 ) strains , unfixed cells were high-pressure frozen using a Leica HMP100 at 2 , 000 psi and freeze-substituted using a Leica Freeze AFS unit using 1% osmium tetroxide and 1% glutaraldehyde . Samples were infiltrated with durcupan resin ( Electron Microscopy Science ) and cut in 100 nm thick sections using a Leica UltraCut UC7 . Sections were collected on formvar/carbon coated nickel grids and stained with 2% uranyl acetate and lead citrate . Grids were imaged in a FEI Tecnai Biotwin TEM at 80 kV with a Morada CCD camera and iTEM ( Olympus ) Software . For immunogold labeling of nucleoporins , 70 nm Lowicryl sections generated as described above for correlative light and electron tomography were cut using Leica UltraCut UC7 onto 200 mesh copper grids ( Quantifoil Micro Tools GmbH ) . Immunolabeling was carried out with the MAb414 antibody diluted 1:100 in 1% BSA , followed by washes in PBS , and probing with a secondary 10 nm gold-conjugated antibody . After further washes , the grids were fixed in 1% glutaraldehyde in PBS . Lastly , grids were post-stained with 1% uranyl acetate , washed in water and viewed with a Biotwin CM120 Philips equipped with a 1K CCD Camera ( Keen View , SIS ) . | With the exception of bacteria , living cells contain most of their DNA inside a structure called the nucleus . The membranes of the nucleus form a protective wall around the DNA , while pores within this wall act as entry check-points , controlling what can and cannot get inside . Maintaining the structure of this wall is critical for cell survival . Problems can occur if the nuclear wall or its pores become disrupted , as in the case of cancer and neurodegenerative diseases . Thankfully cells have developed a protective surveillance system that can rapidly identify and repair any damage made to the nuclear wall . However , how this damage is found and what activates its repair is poorly understood . Now , Thaller et al . have investigated two key proteins that they suspected were involved in the surveillance of the nuclear border in budding yeast: Chm7 and Heh1 . Chm7 is part of a complex group of proteins that can cut and sculpt the shape of membranes , while Heh1 is normally embedded on the inside of the nucleus . Thaller et al . discovered that , when the nuclear wall is disrupted , Heh1 recruits Chm7 to the site of damage and activates it . Once activated Chm7 can repair the damage to the nuclear wall , by sealing over defective nuclear pores and closing gaps caused by breakages . Thaller et al . showed that the transport system that normally moves molecules into and out of the nucleus also imports Heh1 and actively excludes Chm7 , physically segregating them to opposite sides of the nuclear border . If the nuclear wall becomes damaged this leads to the local meeting of Heh1 and Chm7 at these sites . Heh1 will then activate the membrane shaping mechanisms of Chm7 , rapidly repairing the nuclear border in response to the damage . It is possible cell structures other than the nucleus use a similar surveillance system to protect their borders . Manipulating the border surveillance system of the nucleus could be used to treat the detrimental impacts of damage caused to the nuclear wall by disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2019 | An ESCRT-LEM protein surveillance system is poised to directly monitor the nuclear envelope and nuclear transport system |
Under chronic or severe liver injury , liver progenitor cells ( LPCs ) of biliary origin are known to expand and contribute to the regeneration of hepatocytes and cholangiocytes . This regeneration process is called ductular reaction ( DR ) , which is accompanied by dynamic remodeling of biliary tissue . Although the DR shows apparently distinct mode of biliary extension depending on the type of liver injury , the key regulatory mechanism remains poorly understood . Here , we show that Lutheran ( Lu ) /Basal cell adhesion molecule ( BCAM ) regulates the morphogenesis of DR depending on liver disease models . Lu+ and Lu- biliary cells isolated from injured liver exhibit opposite phenotypes in cell motility and duct formation capacities in vitro . By overexpression of Lu , Lu- biliary cells acquire the phenotype of Lu+ biliary cells . Lu-deficient mice showed severe defects in DR . Our findings reveal a critical role of Lu in the control of phenotypic heterogeneity of DR in distinct liver disease models .
The liver is known to possess high capacity for regeneration upon injury . In acutely injured or surgically resected livers , regeneration is usually achieved by proliferation and hypertrophy of residual hepatocytes ( Fausto and Campbell , 2003; Miyaoka et al . , 2012 ) . By contrast , under chronic or severe liver injury that impairs the proliferation of hepatocytes , liver progenitor cell ( LPC ) has been postulated to contribute to liver regeneration by differentiating into hepatocytes and biliary epithelial cells ( BECs ) , also known as cholangiocytes ( Thorgeirsson , 1996; Fausto , 2004; Miyajima et al . , 2014 ) . This response is known as ductular reaction ( DR ) , in which LPC/biliary cell with BEC marker expression proliferates from the portal areas of injured livers , forming pseudo-ductular structures . DRs are frequently observed in human chronic liver diseases and rodent models including fatty liver disease and cholangiopathy ( Shafritz and Dabeva , 2002; Roskams et al . , 2003; Gouw et al . , 2011; Wood et al . , 2014 ) . In zebrafish models , biliary cells have been reported to contribute to regenerating hepatocytes after substantial loss of hepatocytes ( Choi et al . , 2014 ) . In mouse models , accumulating evidence by in vitro assay or transplantation experiments of biliary cells supports the presence of potential LPC with clonogenicity and bi-lineage differentiation capacity in the biliary compartment ( Suzuki et al . , 2008a; Okabe et al . , 2009; Dorrell et al . , 2011; Lu et al . , 2015 ) . In addition , a recent study using in vivo genetic lineage tracing experiment demonstrated that biliary cells can regenerate hepatocytes as facultative LPC under impaired hepatocyte regeneration in mice ( Raven et al . , 2017 ) . Thus , DR is considered as a process of liver regeneration in chronically injured liver . In fact , genetically manipulated mice with defects in DR have been reported to show impaired recovery from chronic liver injury ( Ishikawa et al . , 2012; Takase et al . , 2013; Shin et al . , 2015 ) . To explore the nature of LPC in DR , several mouse injury models have been developed previously . In particular , two dietary models using 3 , 5-Diethoxycarbonyl-1 , 4-dihydrocollidine supplemented ( DDC ) diet and choline-deficient , ethionine-supplemented ( CDE ) diet have been extensively utilized to characterize LPC in mice for many years ( Preisegger et al . , 1999; Akhurst et al . , 2001 ) . Although both models induce massive DR , the pathological features resulting from these two methods are quite distinct; CDE-induced injury is thought to be a mouse model of non-alcoholic fatty liver disease with extensive hepatic damage ( Knight et al . , 2005; Aharoni-Simon et al . , 2011 ) , while DDC-induced injury is considered as a model of chronic cholangiopathy with portal biliary damage and severe cholestasis ( Fickert et al . , 2007 ) . Considering the different pathological features of these two models , DR is assumed to be regulated depending on the severity and type of liver injury . In fact , it has been reported that morphological and functional heterogeneity of ductal cells is evident in DRs originating from many different pathological conditions in human patients and rodent models ( Sell , 1998; Alvaro et al . , 2007; Priester et al . , 2010; Kaneko et al . , 2015 ) . Similar to such previous observations , CDE- and DDC-induced DR exhibited quite distinct morphology; the former extends outwards away from the portal vein , showing primitive ductules with spindle-like shape , while the latter remains around the portal vein , forming obvious bile duct-like structures . However , very little is known about the molecular mechanisms accounting for the phenotypic difference of DR among liver disease models . In the present study , we identified Lutheran blood group glycoprotein ( Lu ) as a crucial molecule to control the morphological heterogeneity of DR . Lu , also known as Basal Cell Adhesion Molecule ( BCAM ) or CD239 is a member of the immunoglobulin superfamily , which is composed of five Ig-like domains on the extracellular site , a single transmembrane domain and a short C-terminal cytoplasmic tail ( Parsons et al . , 1995 ) . Lu is known as a laminin receptor , which has been studied in context of sickle cell disease ( Udani et al . , 1998; El Nemer et al . , 1999 ) , and is capable of binding to laminin-511/521 via laminin alpha5 ( Lama5 ) chain ( Parsons et al . , 2001 ) . It has been reported that Lu/BCAM is widely expressed in cells and tissues including hematopoietic cells , placenta and kidney , and is developmentally regulated in human liver ( Parsons et al . , 1995 ) . However , the expression profile of Lu in mouse LPCs remains to be investigated . Here , we show that Lu is a robust marker to discriminate between CDE-induced and DDC-induced DR; Lu is highly expressed in proliferating biliary cells in the CDE model , but downregulated in the DDC model . By using fluorescence activated cell sorting ( FACS ) , we have isolated Lu-positive and Lu-negative biliary cells from injured liver . By comparison of both biliary cells in scratch assay and cyst formation assay , we revealed a role for Lu in regulating the morphological heterogeneity of DR . Further analysis using Lu-deficient mice demonstrated that the extension of biliary tree was significantly suppressed during DR in the CDE-induced liver injury . Collectively , our findings demonstrate that Lu functionally regulates the remodeling of biliary tissue during liver regeneration , and provide new insights into the heterogeneity of LPCs among liver disease models .
We have previously identified Epithelial cell adhesion molecule ( EpCAM ) as a marker for murine LPC/BEC ( Okabe et al . , 2009 ) . To investigate the difference in DRs between the CDE and DDC models , immunohistochemical analysis of EpCAM was performed . Both protocols of feeding CDE and DDC diet for 3 weeks induced robust DR accompanied by biliary cell expansion . However , the appearance of propagating ducts exhibited strikingly distinct features between two models; CDE-induced biliary cells displayed primitive capillary-like morphology , invading into the parenchymal area extensively , while DDC-induced biliary cells exhibited remarkable duct-like structure , remaining around the periportal area ( Figure 1A ) . Similarly , immunostaining of CK19 , another LPC marker , showed a pattern similar to that of EpCAM ( Figure 1—figure supplement 1 ) . Because most known LPC markers are uniformly expressed in both types of DR , these molecules may not account for the heterogeneity of LPC . Therefore , we examined the expression profile of Lu in both injury models , because we have identified Lu as a marker for hepatoblasts , a fetal type LPC during liver development . Co-staining of liver sections using anti-EpCAM and anti-Lu antibodies revealed that Lu was detected in extending biliary cells of the CDE-fed liver , whereas we could not find such signals in biliary cells of DDC-fed liver except intense signal in EpCAM- ducts ( Figure 1B ) . As reported by the previous paper that Lu is stained in hepatic arteries and portal vein of adult human liver ( Parsons et al . , 1995 ) , co-staining of Lu and platelet endothelial cell adhesion molecule ( PECAM ) , an endothelial marker , in the DDC-fed liver revealed that the EpCAM- duct with strong fluorescence is hepatic artery ( Figure 1—figure supplement 2 ) . Next , to compare the expression level of Lu in biliary cells , we performed flow cytometric ( FCM ) analysis using anti-EpCAM and anti-Lu antibodies . In untreated normal liver , FCM and immunohistochemical analyses demonstrated that most EpCAM+ biliary cells showed moderate expression of Lu ( Figure 1C and Figure 1—figure supplement 3 ) . By contrast , approximately 35% of EpCAM+ cells in CDE-fed liver showed high expression of Lu , whereas nearly half of EpCAM+ cells in DDC-fed liver exhibited low or negative expression of Lu ( Figure 1C ) . These opposing expression profiles of Lu in the CDE and DDC models led us to hypothesize that Lu might regulate the morphological heterogeneity of expanding biliary cells between these distinct injury models . To uncover the role of Lu in regulating biliary cell morphology , we isolated Luhigh and Lu-/low biliary cell fractions from CDE-injured livers by FACS , and then cultured each population . Although Lu is highly expressed in endothelial cells , we confirmed that a contamination of such cells in EpCAM+ gating is highly unlikely as shown in Figure 2—figure supplement 1 . We refer to these cultured biliary cells as Lu+ BC and Lu- BC in the following studies . Interestingly , similar to the morphological characteristics of DR in vivo , Lu+ BC displayed spindle-like shape while Lu- BC showed epithelial-like morphology even after several passages ( Figure 2A ) . Fluorescent staining using Phalloidin demonstrated that the actin fibers and pseudopods are formed in the atypical cell body of Lu+ BC while F-actin accumulates in the periphery of round-shaped cell body of Lu- BC ( Figure 2B ) . By contrast , when freshly isolated EpCAM+ cholangiocytes from untreated normal liver were cultured on the dish , the attached cells exhibited a mixture of round and indefinite morphology ( Figure 2—figure supplement 2 ) . Furthermore , FCM analysis of cultured Lu+ BC and Lu- BC revealed that the expression profile of Lu was maintained in each cell; Luhigh-derived biliary cells continued to express Lu at a high level , while Lu-/low-derived biliary cells hardly expressed Lu ( Figure 2D ) . By contrast , several stem cell markers such as CD24 and CD44 were expressed similarly between Lu+ BC and Lu- BC , suggesting that these cells are closely related , but distinct homogeneous populations . Considering that biliary cell with high expression of Lu in the CDE model exhibits an invasive phenotype in vivo , Lu+ BC is expected to have a higher capacity for cell motility than Lu- BC . To compare the capacity for cell motility between Lu+ BC and Lu- BC , we performed an in vitro scratch assay , by which the moving distance was evaluated after creating a scratch on the dish . As expected , Lu+ BC showed significantly higher motility capacity than Lu- BC ( Figure 3A ) , suggesting that there is a causal link between Lu expression level and cell motility . On the other hand , biliary cells with low or no expression of Lu in the DDC model exhibited an obvious luminal structure around the portal vein in vivo , suggesting that Lu- BC has higher capacity for duct formation than Lu+ BC . As previously reported , LPCs are able to form cyst-like luminal structures with biliary epithelial polarity in a three-dimensional ( 3D ) organoid culture ( Tanimizu et al . , 2007 ) . We thereby compared the cyst-forming capacity between Lu- BC and Lu+ BC in the 3D culture system ( Figure 3B ) . After 6 days of culture , Lu- BC formed a large number of cystic structures , while Lu+ BC formed only small cell aggregates , demonstrating that Lu- BC has higher duct-forming capacity than Lu+ BC ( Figure 3C ) . Thus , Lu+ BC and Lu- BC showed distinct features of cell motility and cyst formation in vitro . These results suggest that Lu may play a crucial role in the cell kinetics of biliary cell . To investigate whether Lu expression is responsible for the characteristics of biliary cells , cDNA encoding either mouse Lu ( mLu ) and green fluorescent protein ( GFP ) or only GFP as a control was transduced into Lu- BC by retroviral vector . The mLu-transduced Lu- BC ( Lu- BC-mLu ) showed morphological change into a spindle-like shape , while GFP-transduced control ( Lu- BC-GFP ) retained a round shape ( Figure 4A ) . In the scratch assay , Lu- BC-mLu showed higher mobility than Lu- BC-GFP ( Figure 4B ) . In contrast , the cyst formation assay demonstrated the reduced cyst-forming capacity of Lu- BC-mLu , resulting in the formation of numerous small aggregates ( Figure 4C ) . These results strongly suggested that Lu- BC acquired the Lu+ BC-like phenotype after Lu expression , and that Lu might endow biliary cells with characteristics of spindle shape morphology and enhanced motility . Although the differential expression profile of Lu is likely to be relevant to the phenotypic heterogeneity of DR , the molecular mechanism by Lu remains unclear . Because Lu is known to be a receptor for Lama5 and bind to Laminin-511 and -521 , we next investigated the expression of Lama5 in CDE- and DDC-injured livers . Intriguingly , double staining of EpCAM and Lama5 revealed that most expanding biliary cells are fully surrounded by Lama5 in both liver injury models ( Figure 5A ) . Considering the accumulation of Lama5 in the vicinity of biliary cells , it is plausible that Lama5 may be secreted from biliary cell itself rather than the environmental niche cell . Indeed , the expression of Lama5 mRNA was verified in both EpCAM+ biliary cells isolated from CDE- and DDC-injured livers ( Figure 5B ) , implying the involvement of Laminins in Lu-driven regulation . While Lu is capable of binding to Laminin-511/521 via Lama5 , these laminins are also known as a ligand for Integrinα3β1/α6β1 ( Kikkawa et al . , 2007 ) . It has been reported that Lu binds to Lama5 competitively with Integrinα3β1/α6β1 and promotes tumor cell migration by modulating Integrin-mediated cell attachment to Laminin-511 protein ( Kikkawa et al . , 2013 ) . Taking these evidences into account , Lu may regulate the morphogenesis of DR via an Integrin-mediated manner . Given that Lu plays a role in the competitive inhibition against Laminin-511/521 and Integrinα3β1/6β1 axis in biliary cell as shown in Figure 5—figure supplement 1 , high expression of Lu would be reproduced by inhibition of integrinβ1 ( Itgb1 ) signaling . To address this possibility , we first investigated the expression of Integrinα3 ( Itga3 ) , Integrinα6 ( Itga6 ) and Itgb1 in Lu- BC and Lu+ BC . As shown in Figure 5—figure supplement 2 , all integrin components were expressed in Lu- BC and Lu+ BC , indicating that Lu- BC and Lu+ BC are potentially competent to cell signaling via Integrinα3β1/α6β1-Laminin-511/521 axis . We next examined the effect of neutralizing antibody against Itgb1 on the motility and duct formation capacity of Lu- BC in vitro . Although the inhibition of Itgb1 signaling did not affect the expression of Lu ( Figure 5—figure supplement 3 ) , it dramatically changed Lu- BC to Lu+ BC-like phenotype in both scratch assay and cyst formation assay ( Figure 5C and D ) . Conversely , we investigated the effect of Itgb1 activation on Lu+ BC . Because TS2/16 antibody has been reported to activate Itgb1 signaling ( Rozo et al . , 2016 ) , we added it to the 3D culture of Lu+ BC . As a result , Lu+ BC acquired cyst formation capacity by the activation of Itgb1 ( Figure 5—figure supplement 4 ) . These data strongly suggested that Lu regulates the characteristic of DR by modulating the Itgb1 signaling . To verify the role of Lu/Bcam in DR in vivo , we generated Bcam knockout ( KO ) mice , in which the 11 bp deletion within a signal sequence-coding region of exon1 resulted in the frame shift of the Bcam gene ( Figure 6A and Figure 6—figure supplement 1 ) . The Bcam KO mice were healthy , showing no obvious developmental abnormality as reported previously ( Rahuel et al . , 2008 ) . To compare the phenotype of DR between Wild-type ( WT ) and Bcam KO mice , mice were fed a CDE or DDC diet for 3 weeks and then their livers were analyzed by immunohistochemistry for EpCAM . When mice were fed the CDE diet , the proliferation of biliary cells occurred in both WT and Bcam KO mouse livers ( Figure 6B ) . The loss of Lu protein in KO mouse was confirmed by immunostaining for Lu ( Figure 6C ) . To our surprise , Lu-deficient biliary cells failed to spread outwards from the portal vein while WT biliary cells extended into the parenchymal area . The migrating distance of biliary cells from the portal vein was significantly shorter in Bcam KO mice than WT mice ( Figure 6D ) . By contrast , in the DDC model , the migrating distance of biliary cells from the portal vein showed no significant difference between WT and Bcam KO mice ( Figure 6D and E ) . However , a slight hyperplasia of duct-like structures was observed in Bcam KO mice in both DDC and CDE models ( Figure 6F ) . Moreover , freshly isolated EpCAM+ biliary cells from WT CDE-fed mouse livers showed spindle-like shape and pseudopods on the dish , while those from Bcam KO mice predominantly exhibited rounder morphology resembling Lu- BC ( Figure 6G ) . These results suggested that Lu plays a critical role in the definition of morphological heterogeneity of DR in vivo . To investigate the expression of Lu/BCAM ( CD239 ) in human liver disease , we performed immunostaining of CD239 for a few resected samples obtained from the surgery to remove liver cancer . Intriguingly , CD239 was stained in DR in patients of chronic liver disease including non-alcoholic steatohepatitis , hepatitis B virus and hepatitis C virus ( Figure 7 ) , suggesting that the expression profile of Lu is conserved in humans .
DR is often observed in various situations of chronic liver injury or submassive liver cell loss . A number of anatomical and histological analyses of human and rodent liver tissues have supported the notion that DR represents the expansion of LPC for supplying transit-amplifying cells to replenish the damaged hepatic cells . However , the origin and the role of LPC in liver regeneration is still under intensive debate . Recent lineage tracing experiments in mice have revealed that the hepatocytic differentiation of LPCs derived from the biliary compartment is negligible in DDC-induced liver injury ( Malato et al . , 2011; Español-Suñer et al . , 2012; Tarlow et al . , 2014; Rodrigo-Torres et al . , 2014 ) . This is reasonable because DDC-injury is a model of chronic cholangiopathy , which requires replenishment of cholangiocytes or bile ducts to be recovered . In fact , it has been reported that mice with impaired DR causes severe jaundice in the DDC model ( Takase et al . , 2013 ) . Consistently , we observed many bile ducts with an obvious luminal structure in the DR of DDC-fed liver . Therefore , the downregulation of Lu in LPC may represent a process of cholangiocytic differentiation and reinforcement of duct formation for bile excretion in the DDC model . In contrast to the DDC-model , it is still controversial whether LPC may contribute to hepatocyte regeneration in other chronic liver injury models . The contribution of LPC derived from biliary component to hepatocyte regeneration has been reported in the CDE model using two different lineage tracing approaches based on BEC marker genes , Osteopontin ( Spp1 ) and Hnf1β ( Español-Suñer et al . , 2012; Rodrigo-Torres et al . , 2014 ) , although to a much lesser extent . The valid but low contribution of LPC of biliary origin to hepatocytic differentiation may be explained by robust proliferation of mature hepatocytes , because hepatocyte-mediated regeneration is not inhibited in the CDE model . This notion is strongly supported by a more recent report from Forbes’s group suggesting that a combination of CDE-injury and inhibition of hepatocyte proliferation causes physiologically significant and higher contribution of biliary cells to hepatic regeneration ( Raven et al . , 2017 ) . Therefore , Lu-mediated high motility of LPC in the CDE model may contribute to the rapid delivery of hepatic progenitor cells to the damaged parenchymal area far from the portal vein . In the DR , environmental factors play crucial roles in the regulation of LPC proliferation and differentiation . Several immune cell-derived cytokines such as TNF-related WEAK inducer of apoptosis ( TWEAK ) , interleukin-6 and interleukin-22 have been shown to be pro-mitotic for LPCs ( Jakubowski et al . , 2005; Yeoh et al . , 2007; Feng et al . , 2012 ) . Cell signaling pathways including Wnt , Notch , HGF and EGF are reportedly responsible for fate decisions of LPCs ( Boulter et al . , 2012; Fiorotto et al . , 2013; Kitade et al . , 2013 ) . Although the regulatory mechanism of Lu expression in LPCs is unclear , these signaling pathways may be worthy of investigating . Extracellular matrix ( ECM ) is also known to serve a niche for LPC regulation in chronic liver injury as well as liver development . Of note , laminin is an important component of LPC niche affecting cell fate decision . It has been reported that the escape of LPCs from the laminin basement favors their hepatocytic differentiation ( Español-Suñer et al . , 2012; Paku et al . , 2004 ) , while laminin aids maintenance of LPC and biliary cell phenotype ( Lorenzini et al . , 2010; Boulter et al . , 2013 ) . Consistently , it has been reported that Lama5 KO mice show defects in bile duct formation during liver development and that Itgb1 signaling is required for cholangiocytic differentiation from hepatoblast , a fetal-type LPC ( Tanimizu et al . , 2012 ) . It is therefore highly probable that Itgb1 signaling in the context of laminin/integrin axis is crucial for biliary regeneration and duct formation from LPC . Our in vitro data also supported the idea that Itgb1 signaling modulated by Lu expression may govern DR , depending on liver injury type . It remains undetermined whether high expression of Lu in biliary cells is an essential step for hepatocytic differentiation . Alternatively , the up-regulation of Lu may provide LPC with a cue to escape from ECM-rich periportal area , as evidenced by Bcam KO phenotype in the CDE model . We showed that the downregulation of Lu in ductular reactive cells facilitated the cystogenic phenotype . However , the loss of Lu would be also expected for the hepatocytic differentiation from LPC because mature hepatocytes do not express Lu . In line with the idea , intriguingly , the expression of Lu seemed to be downregulated in EpCAM+ cells at the tip of DR in the CDE model ( Figure 1B ) . The discrepancy may be explained by the microenvironment surrounding LPCs; the periportal region is rich in ECM including laminin whereas the parenchymal region is poor in it . Taking such microenvironment into consideration , the loss of Lu in parenchymal area may be a sign of hepatocytic differentiation by the escape of LPC from laminin deposition . Further expression analysis of Lu and Laminin with a combination of lineage tracing experiment will uncover the mechanism underlying LPC-mediated liver regeneration depending on the microenvironment . In addition to the signaling molecules and ECM , several non-parenchymal cells have been shown to be involved in the microenvironment for LPC . Intriguingly , Hul et al . reported that prolonged Kupffer cell ( KC ) depletion did not influence the proliferation of LPCs but reduced their invasive behavior in the CDE model ( Van Hul et al . , 2011 ) . The LPCs of KC-depleted mice exhibit phenotype resembling cells of biliary lineage with rounder morphology . More interestingly , the LPCs remain closer to the portal area , in places delineating a pseudo-lumen . This phenotypic change of LPCs by KC depletion closely resembles that of Bcam KO mice in the CDE diet . On the other hand , Boulter et al . reported that hepatic macrophage played a role in promoting LPC specification to hepatocytes by expressing Wnt3a in the CDE model ( Boulter et al . , 2012 ) . These evidences suggest that hepatic macrophages may be a key regulator of Lu expression in a process of LPC specification . Further studies using a lineage tracing experiment in Bcam KO mice under various types of chronic liver injuries will provide a clue to better understanding the molecular mechanisms underlying the phenotypic heterogeneity , as well as fate determination of LPCs during liver regeneration . In conclusion , the present study demonstrates that the expression profile of Lu in biliary cells dramatically changes during DR depending on the type of liver injury , which in turn dictates the morphological characteristics of biliary cells such as cell motility and duct formation ( Figure 8 ) . This molecular mechanism would be expected to be conserved in human liver diseases . Thus , Lu is a novel marker for classification of DR and an interesting functional molecule for investigating the nature of LPCs . Our findings will provide new insights into the significance of biliary cell heterogeneity in liver regeneration .
C57BL/6J mice were purchased from Clea-Japan , Inc . ( Tokyo , Japan ) . All animals were maintained in a standard Specific-Pathogen-Free ( SPF ) room at the institutional animal facility . All animal experiments were performed according to institutional guidelines and approved by the Animal Care and Use committee of the Institute of Molecular and Cellular Biosciences , The University of Tokyo ( approval numbers 2501 , 2501–1 , 2609 , 2706 , and 3004 ) , Kumatomo University ( approval number A27-092 ) , Hyogo College of Medicine ( approval number 16–043 , 16–046 ) , and National Center for Global Health and Medicine Research Institute ( approval numbers 15080 , 16023 , 17086 and 18069 ) . To induce liver injury , a diet containing 0 . 1% DDC ( Clea-Japan Inc . Tokyo , Japan ) or the choline-deficient , ethionine-supplemented diet ( MP Biomedicals , CA , USA ) was fed to 6-week-old mice for 3 weeks . The study using human samples was approved by the Kanazawa University Ethics Committee ( approval number 305–4 ) , and all of the analyzed samples are derived from patients who provided informed written consent for the use of their tissue samples in research . The information about antibodies used for FACS and immunohistochemistry is described in Key resources table . The rat anti-EpCAM monoclonal antibody was generated as described previously ( Okabe et al . , 2009 ) . The rabbit anti-CK19 polyclonal antibody was generated as described previously ( Tanimizu et al . , 2003 ) . The rat anti-Lutheran monoclonal antibody used in this study was generated by immunization of a rat with mouse fetal hepatic cells as described previously ( Suzuki et al . , 2008b ) , and biotinylated for FACS using ECL Protein Biotinylation Module ( GE Healthcare UK Ltd , UK ) . The specific reactivity against mouse Lu was validated by flow cytometric ( FCM ) analysis of Ba/F3 cells transfected with Bcam cDNA by a retroviral vector , pMxs/IRES-GFP ( Kitamura et al . , 2003 ) ( Figure 1—figure supplement 4 ) . The anti-Lutheran monoclonal antibody ( D295-3 ) and anti-EpCAM monoclonal antibody ( D269-3 ) are commercially available from MBL International Corporation , MA , USA . Cells were isolated from murine livers as described previously ( Okabe et al . , 2009 ) . Briefly , liver cells were dissociated by perfusion of collagenase solution . Non-parenchymal cells ( NPCs ) were prepared by removal of hepatocytes with repeated centrifugation at 100 g for 2 min . Then , NPCs were incubated with anti-FcR antibody for blocking non-specific binding , followed by with fluorescein isothiocyanate ( FITC ) -conjugated anti-EpCAM monoclonal antibody for 30 min on ice . After incubation with anti-FITC microbeads ( 1:10–100 dilution , Miltenyi Biotec , Bergisch Gladbach , Germany ) , EpCAM+ cells were enriched by autoMACS pro ( Miltenyi Biotec ) . After MACS , cells were incubated with biotin-conjugated anti-Lutheran monoclonal antibody for 30 min on ice . After wash , cells were incubated with allophycocyanin ( APC ) -conjugated streptavidin ( 1:100–500 dilution , BD bioscience , NJ , USA ) for 20 min on ice and analyzed or purified by fluorescence-activated cell sorting ( FACS ) using Moflo XDP ( Beckman-Coulter , CA , USA ) and BD FACSCanto II ( BD bioscience ) . Dead cells were excluded by propidium iodide ( Sigma-Aldrich , MO , USA ) staining . Sorted EpCAM+ cells from CDE-fed mouse liver were cultured with modified William’s-E medium as previously reported ( Okabe et al . , 2009 ) . For in vitro assay , the cells expanded at 3 to 6 passages were used . For cytoskeleton staining , Alexa Fluor 488 Phalloidin ( 1:500 dilution , Thermo Fisher Scientific , MA , USA ) was used . The cells were seeded in a 24-well plate at a confluency of 90–95% the day before scratch assay . After 12 to 24 hr of culture , the medium was removed and fresh medium with 10 μg/mL Mitomycin C ( Wako Pure Chemical Industries ) was added to fully confluent cells to inhibit further proliferation . The treated cells were incubated continuously for 150 min at 37°C . After incubation , the cell layer was scratched crosswise with a micropipette tip . Each well was washed twice with PBS to prevent detached cells from re-adhering . After creating the scratch , the cells were continuously cultured without cytokines . The moving distance was calculated by subtracting the half of gap length at Day 1 from that at Day 0 . For experiments of antibody administration , Hamster anti-rat CD29 ( 555002 , BD Pharmingen ) or Hamster IgM ( 553957 , BD Pharmingen ) was added in the culture at a final concentration of 1 μg/mL . The resected left lobe of liver was embedded into OCT compound ( Sakura Finetek Japan ) , and frozen by liquid nitrogen . The frozen block was cut into 8 μm slices by Microtome Cryostat HM 525 ( Thermo Fisher Scientific ) . Fixation was performed by using 4% paraformaldehyde ( Wako Pure Chemical Industries , Osaka , Japan ) or cold acetone ( Wako Pure Chemical Industries ) . For blocking buffer , 5–10% skim milk ( BD bioscience ) or 3% FBS ( Thermo Fisher Scientific ) was used . Primary antibodies used for immunohistochemistry were rabbit anti-CK19 polyclonal antibody , rat anti-EpCAM monoclonal antibody , rabbit anti-EpCAM polyclonal antibody , rabbit polyclonal anti-Laminin α5 antibody ( a kind gift from Dr . Jeffrey H . Miner ) , rabbit anti-PECAM polyclonal antibody , rat anti-Ki67 monoclonal antibody and rat anti-Lutheran monoclonal antibody . The information about antibodies is described in Key resources table . All images were captured using KEYENCE BZ-X710:BZ-X Viewer , Zeiss Axio observer z1: AxioCamHR3 or Olympus FV3000 . The ratio of Ki67+ cell per EpCAM+ cell was calculated using Hybrid Cell Count function in the Dual Signal Extraction mode of BZ-X Analyzer . An average value of three random images per mouse was treated as a representative value for the mouse . Quantification of the distance of biliary cell cluster/cell from the center of the portal vein was performed using a previously reported method ( Best et al . , 2016 ) . Briefly , the distance from the center of the portal vein to the most distal EpCAM-stained cell was measured , and then the mean diameter of the portal vein was subtracted from this value to eliminate the influence of the size of the portal vein . An average value of six to fourteen random images of portal region per mouse was treated as a representative value for the mouse . For three-dimensional culture , Cellmatrix Type I-A ( Nitta Gelatin , Osaka , Japan ) and Matrigel with Growth Factor Reduced ( Corning , MA , USA ) were used for gel components . Chilled cellmatrix and Matrigel were mixed at 1:9 ratio and used to coat the surface of the culture dish . After solidification of the coating layer by incubating at 37°C , the mixture of cell suspension and gel at 1:1 ratio was added . After 30 min of incubation at 37°C in 5% CO2 chamber , culture medium was loaded on top of the double-layered gel . The top layer of medium was changed twice a week . All images were captured using DS-Fi2-L3 ( Nikon Corporation , Tokyo , Japan ) under a phase-contrast microscope ( Nikon ECLIPSE TS100 , Nikon Corporation ) after 6–12 days of culture . For the experiments using neutralizing and activating antibodies against Itgb1 , Hamster anti-rat CD29 ( 555002 , BD Biosciences ) and TS2/16 ( 303010 , Biolegend ) were added in the culture at a final concentration of 5 μg/mL and 50 μg/mL , respectively . For each control , Hamster IgM , λ1 isotype control ( 553957 , BD Pharmingen ) or Mouse IgG1 , κ isotype control ( 401404 , Biolegend ) were used . For quantification of the size and formation efficiency of cyst , 50 cells were cultured in individual wells of a 96-well plate . After 6 days of culture , the image was captured by a phase-contrast microscope . All visible cell clusters were counted according to the diameter of lumen . The cell cluster devoid of luminal structure was counted as ‘Cell aggregate’ . For overexpression of mouse Lu ( mLu ) in Lu- BC , Bcam cDNA was amplified with two primers 3’- CTCGAGTCACTGCCGCCACTGCAG −5’ and 3’- GTCGACTTACATTCCCTGGAGGAAG −5’ by RT-PCR and inserted into the EcoRI and XhoI restriction enzyme sites of pMxs-IG plasmid vector ( kindly provided by Dr . Kitamura ) was used . For the production of retrovirus , pMxs-mLu-IG or pMxs-IG was transfected into the packaging cell line Platinum-E ( Morita et al . , 2000 ) by using lipofectamine 2000 ( Invitrogen ) . The culture supernatant was centrifuged at 6000 g at 4°C overnight to recover virus particles . The precipitated virus particles were dissolved in culture medium and used to infect Lu- BC . After 16–24 hr , the culture media was replaced with fresh media and the culture was continued overnight . The cells expressing both mLu and GFP or only GFP were sorted by FACS and named as Lu- BC-mLu or Lu- BC-GFP , respectively . RNA extraction was performed using ISOSPIN Cell and Tissue RNA ( NIPPON GENE , Toyama , Japan ) according to the manufacturer’s instruction . For tissue homogenization , FastPrep-24 ( MP Biomedicals ) was used . Reverse transcription from RNA to cDNA was performed by PrimeScript RT Master Mix ( Takara-bio , Shiga , Japan ) . Quantitative RT-PCR was performed using LightCycler480 ( Roche , Basel , Switzerland ) with the Universal Probe Library system . The sequence of used primers is ( 5’ to 3’ ) EpCAM-Forward: AGAATACTGTCATTTGCTCCAAACT , EpCAM-Reverse: GTTCTGGATCGCCCCTTC , Lama5-Forward: GGCCTGGAGTACAATGAGGT , Lama5-Reverse: CACATAGGCCACATGGAACA , ITGB1-Forward: TCAACATGGAGAACAAGACCA , ITGB1-Reverse: CCAACCACAGCTCAATCTCA , ITGA3-Forward: TCAACATGGAGAACAAGACCA , ITGA3-Reverse: CCAACCACAGCTCAATCTCA , ITGA6-Forward: GCGGCTACTTTCACTAAGGACT , and ITGA6-Reverse: TTCTTTTGTTCTACACGGACGA . pT7-sgRNA and pT7-hCas9 plasmid were kindly provided from Dr . Ikawa ( Osaka University , Japan ) ( Mashiko et al . , 2013 ) . After digestion with EcoRI , Bcam mRNA synthesis was performed using an in vitro RNA transcription kit ( mMESSAGE mMACHINE T7 Ultra Kit , Thermo Fisher Scientific ) , according to the manufacturer’s instructions . A pair of oligos targeting Bcam gene was annealed and inserted into the BbsI site of the pT7-sgRNA vector . The sequences of the oligos were as follows: Bcam/Lu ( 5’- AAC CCC CTG ACG CCC GCG CA −3’ ) , which is located at exon 1 of Bcam/Lu gene . After digestion with XbaI , gRNAs were synthesized using the MEGAshortscript Kit ( Thermo Fisher Scientific ) . The precipitated RNA was dissolved in Opti-MEM I ( Thermo Fisher Scientific ) at 0 . 4 μg/μL . C57BL/6N female mice ( Clea-Japan Inc . ) were used in this study . IVF was performed according to the Center for Animal Resources and Development’s ( at Kumamoto University , Japan ) protocol ( http://card . medic . kumamoto-u . ac . jp/card/english/sigen/manual/onlinemanual . html ) . Electroporated embryos were cultured in KSOM medium , and transferred the next day to the oviducts of pseudo-pregnant females on the day of vaginal plug detection . Genome Editor electroporator and LF501PT1-10 platinum plate electrode ( BEX Co . Ltd . , Tokyo , Japan ) were used for electroporation . 50 embryos prepared were subjected to electroporation . The collected embryos cultured in KSOM medium were placed in the electrode gap filled with 5 μl of Opti-MEM I containing sgRNA and hCas9 mRNA . The electroporation conditions were 25V , five times . The eggs were then cultured in KSOM medium at 37°C and 5% CO2 in an incubator until the two-cell stage . Three individual surgical specimens of cirrhotic liver were obtained from the patients with hepatocellular carcinoma . The deparaffinized and rehydrated sections were microwaved in EDTA buffer ( pH 9 . 0 ) for 20 min in a microwave oven . Following endogenous peroxidase blocking , these sections were incubated at 4˚C overnight with rabbit anti-CD239 monoclonal antibody against human Lutheran/BCAM ( 1:100 dilution , Epitomics , CA , USA ) and then at RT for 1 hr with goat anti-rabbit immunoglobulins conjugated to peroxidase labeled-dextran polymer ( K4003 , Envision , Dako , Tokyo , Japan ) . After benzidine reaction , sections were lightly counterstained with hematoxylin . Statistical analyses and the determination of p value were performed using GraphPad Prism software . Statistical significance between two groups was evaluated using Mann-Whitney U test and considered for p<0 . 05 . For comparison of four groups , one-way analysis of variance ( ANOVA ) was applied , and once F-test was significant , multiple comparisons between each group were conducted by Tukey’s multiple comparisons . Statistical significance was set at two-tailed p values < 0 . 05 . Values derived from at least four biological replicates were plotted in a graph with mean and standard deviation . The exact number of biological samples was described in each figure legend and source data . There was no exclusion of outliers in all experiments . Group allocation was performed without any bias . A statistical method of sample size calculation was not used during study design . | Bile is a green to yellow liquid that the body uses to break down and digest fatty molecules . The substance is produced by the liver , and then it is collected and transported to the small bowel by a series of tubes known as the bile duct . When the liver is damaged , the ‘biliary’ cells that line the duct orchestrate the repair of the organ . In fact , the duct often reorganizes itself differently depending on the type of disease the liver is experiencing . For example , the biliary cells can form thin tube-like structures that deeply invade liver tissues , or they can grow into several robust pipes near the existing bile duct . However , it remains largely unknown which protein – or proteins – drive these different types of remodeling . Miura et al . find that , in mice , the biliary cells which invade an injured liver have a large amount of a protein called Lutheran at their surface , but that the cells that form robust ducts do not . This protein helps a cell attach to its surroundings . In addition , the biliary cells can adopt different types of repairing behaviors depending on the amount of Lutheran in their environment . Further experiments show that it is difficult for genetically modified mice without the protein to reshape their bile duct after liver injury . Finally , Miura et al . also detect Lutheran in the remodeling livers of patients with liver disease . Taken together , these results suggest that Lutheran plays an important role in tailoring the repairing roles of the biliary cells to a particular disease . The next step would be to clarify how different liver conditions coordinate the amount of Lutheran in biliary cells to create the right type of remodeling . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"developmental",
"biology"
] | 2018 | Differential expression of Lutheran/BCAM regulates biliary tissue remodeling in ductular reaction during liver regeneration |
Intestinal goblet cells maintain the protective epithelial barrier through mucus secretion and yet sample lumenal substances for immune processing through formation of goblet cell associated antigen passages ( GAPs ) . The cellular biology of GAPs and how these divergent processes are balanced and regulated by goblet cells remains unknown . Using high-resolution light and electron microscopy , we found that in mice , GAPs were formed by an acetylcholine ( ACh ) -dependent endocytic event remarkable for delivery of fluid-phase cargo retrograde into the trans-golgi network and across the cell by transcytosis – in addition to the expected transport of fluid-phase cargo by endosomes to multi-vesicular bodies and lysosomes . While ACh also induced goblet cells to secrete mucins , ACh-induced GAP formation and mucin secretion were functionally independent and mediated by different receptors and signaling pathways , enabling goblet cells to differentially regulate these processes to accommodate the dynamically changing demands of the mucosal environment for barrier maintenance and sampling of lumenal substances .
The simple columnar epithelium lining the gastrointestinal tract is an expansive surface exposed to lumenal contents containing innocuous substances from the diet and potentially harmful microbes and their products . Underneath and within this epithelium lies the largest collection of immune cells in the body . It has long been appreciated that the gastrointestinal immune system is not ignorant of the lumenal contents , and in the healthy state gut lumenal contents are sampled to induce adaptive immune responses , characterized by T cell-mediated antigen-specific tolerance . How substances from the lumen are encountered by , or delivered to , the immune compartment is a central concept in mucosal immunity that is incompletely understood , but fundamentally underlies how the immune system can mount tolerogenic responses to substances encountered in this potentially hostile environment . Goblet cells ( GCs ) are specialized intestinal epithelial cells that have a well-established role in innate immunity through secretion of mucins and maintenance of the mucus layer . The mucus layer provides a first line of defense against physical and chemical injury and protects against pathogen invasion ( Kim and Ho , 2010 ) . GCs also produce and secrete biologically active products that contribute to innate immunity by promoting epithelial restitution , inhibiting intestinal nematode chemotaxis , and stabilizing the mucus layer ( Herbert et al . , 2009; Johansson et al . , 2009; Taupin and Podolsky , 2003 ) . Recently , a new role for GCs in intestinal immunity was identified; the ability of GCs to form GC-associated antigen passages ( GAPs ) , which deliver lumenal substances to lamina propria antigen presenting cells to generate antigen-specific T cell responses ( Knoop et al . , 2017a; Knoop et al . , 2017b; Knoop et al . , 2015; McDole et al . , 2012; Knoop et al . , 2017b ) . The potent GC secretagogue acetylcholine ( ACh ) induces GAP formation in the homeostatic state ( Knoop et al . , 2015; Miller et al . , 2014 ) , indicating that GAP formation and mucus secretion can be induced by the same stimulus . However , GAP formation and the delivery of lumenal substances to the immune compartment are closely regulated to prevent inflammatory responses to gut bacteria and dietary antigens in settings where the secretion of mucus to maintain the barrier is critical ( Knoop et al . , 2017a; Knoop et al . , 2016; Kulkarni et al . , 2018 ) . Thus , in response to the same stimulus , GCs perform two , but apparently opposing , central roles in intestinal immunity . Here , we demonstrate that GAPs form by an endocytic event capturing fluid-phase cargo and often follow secretory granule exocytosis . But in contrast to the fate of fluid-phase endocytosis seen in adjacent enterocytes , GAPs traffic a substantial portion of fluid-phase cargo across the epithelial barrier by transcytosis for capture by underlying phagocytic cells . Further , we find that while ACh induces both mucus secretion and GAP formation in GCs , these processes are not functionally linked and can be performed independently or in parallel to meet the needs of the changing lumenal environment in the small intestine and colon . Together these observations suggest that GAP formation may have evolved as a cell-type specific and specialized endocytic pathway emerging from the basic exocytic machinery of GCs to sample and deliver luminal substances to the mucosal immune system .
To elucidate the cellular mechanism underlying GAP formation and function , we studied the intestines of mice 1 hr after administration of fluorescent dextran to the gut lumen . The appearance of GAPs on two-photon in vivo imaging ( Figure 1A ) and on wide-field fluorescence microscopy ( Figure 1B ) of the small intestine ( SI ) suggested that the GC cytoplasm was filled with lumenal substances ( marked by fluorescently labeled dextran ) , consistent with passive diffusion of dextran into the cytosol through apical membrane tears following ACh-induced mucus granule secretion . To better understand this process , we examined GAPs using super-resolution microscopy approaches . The appearance of SI GAPs by structured illumination microscopy ( SIM ) revealed that the cytoplasm of GCs forming GAPs was not completely filled with the fluorescent luminal cargo , rather the luminal solute tracer was contained within a network of vesicular appearing structures located predominantly at the periphery of the cell and extending from apical to basolateral cell surfaces , but largely excluded from the region where mucin granules are contained within the theca . Notably , in places , the luminally administered ovalbumin ( OVA ) was also observed in vesicular appearing structures contained within ItgaxYFP+ mononuclear phagocytes ( MNPs ) located beneath the epithelial barrier ( Figure 1C ) . The vesicular patterns observed suggested that GAPs were formed by an active endocytic process that delivered fluid-phase cargo across the epithelial barrier to the lamina propria . Indeed , blockade of endocytosis by the dynamin inhibitors Dyngo 4a and dynasore strongly attenuated GAP formation ( Figure 1D and Figure 1—figure supplement 1A–C ) . Inhibition of actin and microtubule polymerization using cytochalasin D and colchicine , respectively , and inhibition of phosphoinositide three kinase ( PI3K ) using LY294002 also resulted in a significant reduction in GAP formation ( Figure 1D and Figure 1—figure supplement 1D–F ) , as did inhibition of the microtubule motor proteins dynein and kinesin using ciliobrevin D and dimethylenastron ( DMEA ) respectively ( Figure 1D and Figure 1—figure supplement 1G , H ) . These treatments did not adversely affect epithelial integrity ( Figure 1—figure supplement 1A–H ) . Thus , GAPs appear to represent a fluid-phase endocytic pathway capable of delivering lumenal solutes across the cell by transcytosis ( Clague et al . , 1995; Gottlieb et al . , 1993; Thyberg and Stenseth , 1981 ) . The endocytic capacity of secretory cells such as neurons and enterocromaffin cells has been extensively studied using lipophilic styryl dyes such as FM 1–43FX ( Pyle et al . , 1999; Raghupathi et al . , 2016 ) . These dyes become fluorescent when incorporated in membranes , and the endocytic capacity of a cell and the intracellular fate of the endocytic vesicles can be studied by evaluating the subcellular localization of FM 1–43FX stained membranes . Lumenal administration of FM 1–43FX resulted in intracellular staining of a subset of GCs , with a staining pattern similar to that of endocytosed dextran ( Figure 1E compare with Figure 1B ) . To evaluate whether endocytosed dextran localized to newly endocytosed membrane structures within the same cell , FM 1–43FX and 10 kDa dextran-Alexa647 were co-administered into the SI lumen and the intracellular staining pattern was evaluated by confocal microscopy . We observed a similar staining pattern of dextran-Alexa647 and FM 1–43FX within the same cell having GC morphology ( Figure 1F ) , further supporting that GAP formation represents an endocytic process . In addition to observing intracellular FM 1–43FX staining and dextran uptake by GCs , we also observed uptake of lumenal dextran by surrounding intestinal epithelial cells . However , the pattern and morphology of dextran-containing endosomes in absorptive cells differed from those in GCs . In GCs ( asterisk; Figure 1G and upper panel insets ) , dextran-containing structures were observed throughout the cell , whereas in adjacent epithelial cells , likely enterocytes ( arrows; Figure 1G and lower panel insets ) , a punctuated pattern of FM 1–43FX staining and dextran uptake was observed localized to the apical cytoplasm . In adjacent epithelial cells , we also observed co-localization of dextran-containing vesicles and the early endosome marker EEA1 ( arrows; Figure 1H ) , but EEA1 staining was not apparent in the endocytic vesicles of adjacent GCs . Thus , the vesicles and trafficking pathways of fluid-phase endocytosis differ between GCs forming GAPs and adjacent enterocytes . To understand the trafficking and fates of the vesicles-containing lumenal cargo in GAPs , we performed focused ion beam scanning electron microscopy ( FIB-SEM ) using intestinal tissue obtained from mice 1 hr after administration of 10 kDa lysine fixable dextran biotin into the gut lumen . Full thickness tissues were incubated with heavy metal labeled streptavidin and processed for FIB-SEM and imaged at a voxel resolution of 10 nm per slice with >1000 slices encompassing the thickness of a GC . Datasets of four GCs endocytosing lumenal dextran were imaged with this technique . A dataset containing 1199 images encompassing the volume of a GC that had taken up dextran was used to create the 3D model of a GC forming a GAP illustrated in Figure 2A and Video 1 . The remaining datasets as well as transmission electron microscopy ( TEM ) of dextran incubated tissues were used as to confirm the model . The endocytosed dextran was seen in apically located endosomes , multi-vesicular bodies ( MVBs ) , lysosomes , the TGN , and in vesicles close to basolateral membrane consistent with delivery of dextran into the transcytotic pathway ( Figure 2A and B and Video 2 ) . Shuttling of endocytosed cargo to endo-lysosomal structures as well as the TGN is consistent with models of fluid-phase endocytosis and their trafficking in most cell types ( Maxfield and McGraw , 2004 ) , and the specialized need for compensatory endocytosis required by secretory cell types like GCs to recycle large amounts of membrane back into the secretory pathway ( i . e . , TGN ) after exocytosis ( Engisch and Nowycky , 1998 ) . Notably the endocytic vesicles in GCs forming GAPs did not stain positive for EEA1 ( Figure 1H ) , a marker of early endosomes , or the late endosome marker Rab7 ( Figure 2—figure supplement 1A ) , but did in some cells stain positive for the lysosome marker LAMP-1 and the TGN marker TGN46 ( Figure 2—figure supplement 1B , C ) . To explore if the endosomes of GAPs could be related to membrane retrieval following secretory granule exocytosis , we examined the location of the secretory granule protein Rab3D that has been shown to be retrieved and redistributed to the TGN following regulated secretion in GCs and pancreatic acinar cells ( Jena et al . , 1994; Valentijn et al . , 2007 ) . We found that in GCs forming GAPs , Rab3D co-localized with areas of dextran uptake , while in GCs not forming GAPs , Rab3D was localized to the secretory granules within the GC theca ( Figure 2C and compare with Figure 2D ) . A similar staining pattern was observed when evaluating the localization of the secretory granule protein VAMP8 in GCs forming GAPs ( Figure 2—figure supplement 1D ) . These results could be consistent with GAP endosomes forming as a result of membrane recycling after secretory granule exocytosis , which was supported by findings of dextran localizing to wheat germ agglutinin ( WGA ) -positive vesicular structures in GCs forming GAPs ( Figure 2E ) . Furthermore , unlike absorptive intestinal epithelial cells where transcytosis is largely restricted to receptor-mediated endocytosis ( Fung et al . , 2018 ) , the GCs forming GAPs appeared to deliver fluid-phase cargo into the transcytotic pathway , as evidenced by dextran-containing vesicles closely adjacent to the basolateral membranes ( Figure 2B ) . In some cases , dextran-containing cargo was observed at a fusion point of GCs and adjacent cells in the lamina propria ( Figure 2F and G ) , which is consistent with the transfer of luminal cargo taken up by GAPs to lamina propria MNPs we observed by SIM ( Figure 1C ) . These results suggest that GAPs represent a fluid-phase transcellular endocytic process capable of efficiently delivering lumenal substances across the epithelial barrier to be captured by phagocytic cells of the innate immune system . We have previously shown that in the homeostatic state GAPs are present in the distal colon and SI , preferentially occurring in villus GCs , but largely absent in the proximal colon due to GC intrinsic sensing of the microbiota inhibiting muscarinic ACh receptor 4 ( mAChR4 ) -driven GAP formation in adult mice ( Knoop et al . , 2015; Kulkarni et al . , 2020 ) . In the SI , steady-state GAP formation is mediated by ACh acting on mAChR4 expressed by GCs ( Knoop et al . , 2015 ) . ACh is a potent GC secretagogue known to induce secretion of large quantities of mucus by activation of mAChRs when added to intestinal explants . However , baseline mucus secretion in the intestine has not been linked to muscarinic receptor activation ( Specian and Neutra , 1980 ) , suggesting that during steady state , activation of mAChR4 triggers endocytic retrieval of secretory granule membrane without being the main driver of secretory granule exocytosis . To explore this further we treated mice with the mAChR4 antagonist tropicamide and evaluated the effect on GAP formation and the mucus remaining within the GCs as an indication of inhibition of baseline mucus secretion . Our results showed that inhibition of mAChR4 signaling by tropicamide reduced GAP formation in the SI villus and crypt as expected ( Figure 3A and B , image E ) , but had no measurable effect on mucus secretion , quantified as WGA+ mucus area per villus or crypt cross section ( Figure 3C and D , image E ) . We have previously shown that GAP formation in the distal colon is driven by ACh acting on muscarinic receptors , but independent of mAChR4 signaling ( Kulkarni et al . , 2020 ) . However , colonic GCs have been shown to express mAChR3 and mAChR1 ( Tabula Muris Consortium et al . , 2018 ) . 1 , 1-Dimethyl-4-diphenylacetoxypiperidinium iodide ( 4-DAMP ) is an mAChR3 preferring antagonist with some capacity to bind mAChR1 . Treatment of mice with 4-DAMP significantly decreased distal colon GAP formation similar to that observed with the pan-muscarinic receptor antagonist atropine , while the mAChR1 antagonist telenzepine had no measurable effect on GAP formation ( Figure 3F , image G ) , indicating that distal colonic GAPs are driven by mAChR3 signaling . Similar to the SI , inhibition of mAChR3 signaling using 4-DAMP had no measurable effect on mucus secretion in the distal colon as evaluated by Ulex europaeus agglutinin 1 ( UEA1+ ) mucus area per crypt cross section ( Figure 3H ) . Thus , mAChR4 and mAChR3 drive GAP formation in the SI and distal colon , respectively , but do not play a major role in baseline mucus secretion . Our results indicate that in the steady state , ACh acting on mAChR4 in the SI and mAChR3 in the distal colon triggers a response resulting in GAP formation with negligible effects on the degree of mucus secretion ( Figure 3A–H ) . In contrast , exposing intestinal explants to micro-molar concentrations of ACh induces rapid expulsion of large quantities of mucus , primarily from the intestinal crypts ( Ermund et al . , 2013; Gustafsson et al . , 2012; Phillips , 1992; Specian and Neutra , 1980 ) . To explore how GAP formation correlates with mucus secretion induced by the same stimulus , ACh , we treated mice with the stable ACh analogue carbamylcholine ( CCh ) and measured GAP formation and mucus secretion . Mucus secretion was assayed as a decrease in mucus content within the tissue . Exposure to CCh resulted in a significant increase in GAP formation in the SI villus and crypts and the distal colon crypts ( Figure 4A , images C , F , and G ) and was paralleled by an accelerated mucus secretory response defined as loss of WGA+ ( SI ) or UEA1+ ( distal colon ) mucus area , at all three locations ( Figure 4B , images C , F , G ) . Despite all three locations responding with a mucus secretory response , the degree and type of response differed between compartments . In the crypts of the SI and distal colon , but not the SI villus , exposure to CCh resulted in a significant reduction in the number of WGA+ ( SI ) and UEA1+ ( distal colon ) GCs ( Figure 4D ) , indicative of complete emptying of a subset of GCs in these regions of the alimentary tract – the crypts of the SI and colon . However , the remaining SI crypt GCs that did not undergo complete emptying and SI villus GCs demonstrated a reduction in the GC theca area ( Figure 4E ) consistent with partial emptying of the mucin granules in response to CCh in this compartment . The remaining GCs in the distal colon crypts staining positive for mucins demonstrated no change in the size of the GC theca ( Figure 4E ) , suggesting that GCs in the distal colon predominantly responded to CCh by complete emptying of the theca . Further correlation of the CCh-induced mucus secretory response with the GAP response between compartments revealed that the SI crypts had the largest reduction of total mucus content per cross section ( crypt SI: –47 . 2% ± 2 . 2% , villus SI: –34 . 1% ± 3 . 1% , crypt DC: –32 . 8% ± 4 . 9% , crypt SI vs . villus SI: p < 0 . 05 , crypt SI vs . crypt DC: p < 0 . 05 ) , whereas the SI villus had the largest GAP response to CCh quantified as the percentage of GCs forming GAPs in response to CCh ( villus SI: 58 . 8% ± 2 . 5% , crypt SI: 21 . 38% ± 1 . 2% , crypt DC: 17 . 4% ± 1 . 2% , villus SI vs . crypt SI , p < 0 . 001 , villus SI vs . crypt DC: p < 0 . 001 , crypt SI vs . crypt DC , p > 0 . 05 ) . The SI and distal colon crypts responded to CCh with complete emptying of an average of 3 and 4 GCs per crypt cross section , respectively ( Figure 4D ) , and formed ~1 additional GAP per cross section ( Figure 4A ) , demonstrating that the degree of CCh-induced mucus secretion does not directly correlate with the extent of GAP formation . We have previously observed Paneth cells acquiring lumenal administered dextran and ovalbumin ( Kulkarni et al . , 2020; Noah et al . , 2019 ) and observed Paneth cells acquiring dextran in the present study ( Figure 4H ) , therefore the low frequency of GAP formation in the intestinal crypts cannot be due to restricted access of dextran to the SI crypt lumen . Thus , in response to CCh , GCs in all the examined locations of the intestine produced a mucus secretory response and formed GAPs , however SI and distal colon crypt GCs were more likely to completely empty their mucus content and villus GCs were more likely to form GAPs , further suggesting that GAP formation and mucus secretion may occur independently of each other , even in response to the same stimulus . To evaluate if the observed regional and spatial differences in ACh-induced mucus secretion and GAP formation were due to the two processes being mediated by different mAChRs , we treated mice with mAChR antagonists and measured the effect on CCh-induced GAP formation and mucus secretion . Our results showed that the mAChR4 antagonist tropicamide had no effect on the CCh-induced mucus secretory response in either the SI villus or crypt compartment ( Figure 5A and C ) , but it blocked the CCh-induced GAP response at both locations ( Figure 5B and D ) , similar to the effects of mAChR4 blockade in the basal state ( Figure 3A and B ) . In the distal colon , the mAChR3 inhibitor 4-DAMP inhibited the CCh-induced GAP response similar to the effects of mAChR3 inhibition in the basal state ( Figure 5F ) , but it did not reverse the CCh-induced mucus secretory response ( Figure 5E ) – indicating that the CCh-induced mucus secretory response was mediated by other mAChRs . As previously noted , intestinal GCs also express mAChR1 ( Haber et al . , 2017; Tabula Muris Consortium et al . , 2018 ) . Evaluation of the role of mAChR1 in regulating the CCh response showed that pretreatment of mice with the mAChR1 antagonist telenzepine reversed both the CCh-induced mucus secretory response ( Figure 5A , C and E ) and the CCh-induced GAP response ( Figure 5B , D and F ) at all three locations . Thus , the CCh-induced mucus secretory response is mediated by activation of mAChR1 while the CCh-induced increase in GAPs involves activation of both mAChR1 and mAChR4 in the SI , and mAChR1 and mAChR3 in the distal colon . We evaluated the expression of mAChRs in the SI and distal colon to correlate the CCh-induced mucus secretory response and the GAP response with expression of the respective receptors . In the SI , the mAChR4 expression increased along the crypt – villus axis , correlating with the higher prevalence of mAChR4-dependent GAP formation in the villi as compared to the crypts ( Figure 5—figure supplement 1A ) . In contrast , the mAChR1 expression was more evenly distributed along the crypt – villus axis , correlating with mAChR1’s involvement in both the mucus secretory response and the GAP response in the SI villi and crypts ( Figure 5—figure supplement 1B ) . In the distal colon , epithelial expression of mAChR3 was primarily located to the the lower part of the crypts , correlating with the location of distal colon GAPs ( Figure 5—figure supplement 2A ) . In addition , mAChR3 expression was also observed in the muscle layers ( Figure 5—figure supplement 2A ) . Similar to mAChR3 , mAChR1 expression was primarily observed in the lower part of the crypt correlating with the location of the CCh-induced mucus secretory response ( Figure 5—figure supplement 2B ) . Based on previous studies demonstrating that ACh-induced mucus secretion is mediated by elevated levels of intracellular Ca2+ and endocytosis being a Ca2+-dependent process ( Barbieri et al . , 1984; Seidler and Sewing , 1989 ) , we explored the role of Ca2+ in regulating ACh-induced mucus secretion and GAP formation . Elevated levels of intracellular Ca2+ can either be obtained by Ca2+ release from intracellular stores or activation of Ca2+ channels in the plasma membrane triggering influx of extracellular Ca2+ . Therefore , we explored the source of Ca2+ needed to drive CCh-induced increase in GAP formation and mucus secretion in the SI and distal colon . Our results showed that the CCh-induced mucus secretory response was reversed by chelation of extracellular Ca2+ ( EGTA ) as well as by chelation of intracellular Ca2+ ( BAPTA-AM ) in the SI ( Figure 6A , B and E–H ) and in the distal colon ( Figure 7A and C–F ) , indicating that the CCh-induced mucus secretory response is dependent upon influx of extracellular Ca2+ and possibly also dependent on Ca2+ release from intracellular stores . In contrast , the CCh-induced GAP response in the SI ( Figure 6C–H ) and distal colon ( Figure 7B–F ) was independent of extracellular Ca2+ , but dependent on intracellular Ca2+ , and by extension dependent upon the release of Ca2+ from intracellular stores . Notably , in the SI villus , chelation of intracellular Ca2+ by BAPTA-AM reduced GAP numbers below baseline levels ( vehicle: 3 . 54 ± 0 . 34 , BAPTA-AM+ CCh , 0 . 64 ± 0 . 10 , p < 0 . 001 ) demonstrating that both baseline and CCh-induced GAP formation are dependent on intracellular Ca2+ . Intracellular stores of Ca2+ that are released in response to mAChR activation include the endoplasmatic reticulum ( ER ) and acidic organelles such as endosomes/lysosomes , and in the case of GCs , mucin granules that represent a large acidic compartment with high Ca2+ content ( Wu et al . , 2001 ) . Ca2+ release from these intracellular stores occurs via three different pathways mediated by the second messengers inositol trisphosphate ( IP3 ) , cyclic ADP ribose ( cADPr ) , and nicotinic acid adenine dinucleotide phosphate ( NAADP ) , each acting on IP3 receptors ( IP3R ) , ryanodine receptors ( RyR ) , and two pore channels ( TPC ) , respectively ( Calcraft et al . , 2009 ) . To explore the role of Ca2+ release from intracellular stores in the CCh-induced mucus secretory response and GAP response , we inhibited these signaling pathways; IP3R using xestospongin C ( Xesto C ) , cADPr using 8-Br-cADPr , and NAADP using trans-Ned-19 ( T-Ned-19 ) , and evaluated the effects on CCh-induced mucus secretion and GAP formation . In the SI villus , despite the CCh secretory response being reversed by chelation of intracellular Ca2+ , none of the respective inhibitors reversed the CCh-induced secretory response , suggesting possible redundancy in the signaling pathways inducing this response ( Figure 6A , F and I-K ) . The CCh-induced mucus secretion was dependent on NAADP , but independent of IP3R or cADPr in the SI crypts ( Figure 6B , F and I-K ) and distal colon crypts ( Figure 7A , D and G-I ) . The GAP response was on the other hand dependent on NAADP and cADPr but independent of IP3R in the SI ( Figure 6C , D , F and I-K ) and distal colon ( Figure 7B , D and G-I ) . Since the GAP response was shown to be dependent on intracellular Ca2+ , we evaluated the location of the ER within GCs forming GAPs . Immunostaining of tissue sections using the ER marker Calnexin showed positive staining throughout the GC , with the exeption of the theca . The ER was observed in close proximity to the dextran at the apical , lateral , and basal sides of the cell ( Figure 6—figure supplement 1 ) . Thus , the CCh-induced mucus secretory response involves release of intracellular Ca2+ via NAADP-mediated pathways and influx of extracellular Ca2+ , while the GAP response involves release of intracellular Ca2+ via NAADP- and cADPr-mediated pathways but is not dependent upon influx of extracellular Ca2+ . Furthermore , the finding that CCh-induced GAP formation occurred in the absence of CCh-induced mucus secretion in tissues treated with the extracellular Ca2+ chelator EGTA , suggests that the role of mAChR1 in driving the CCh-induced increase in GAPs occurs independently of its role in inducing mucus secretion . The findings that CCh-induced GAP formation can occur in the absence of CCh-induced mucus secretion , and that CCh-induced mucus secretion can occur in the absence of GAP formation could either be interpreted as the two processes being induced in separate populations of GCs or that they can occur in parallel in the same GCs but are not functionally linked or dependent upon each other . To explore this question , we evaluated the effect of CCh on the theca area , as a surrogate for mucus secretion , in GCs forming GAPs ( epithelial cells staining positive for WGA/UEA1 and tetramethylrodamine [TRITC]-dextran ) and in those not forming GAPs ( epithelial cells staining positive for WGA/UEA1 but negative for TRITC-dextran ) . We observed that CCh induced a significant decrease in the average theca area of GCs forming GAPs at all three locations ( Figure 8A ) , indicating that GCs forming GAPs also respond to CCh with accelerated mucus secretion . The secretory response was also seen as a shift in the size distribution of the theca area in all three locations ( Figure 8B–D ) . The CCh-induced decrease in theca area was reversed by EGTA at all three locations ( Figure 8A–D ) , showing that GCs forming GAPs respond to CCh with a mucus secretory response mediated by influx of extracellular Ca2+ . Similar to these findings , GCs in the SI villus and crypt that did not form GAPs in response to CCh responded with a decrease in the average theca area ( Figure 8E ) and a shift in the size distribution of the GCs theca , which was reversed by EGTA ( Figure 8F and G ) , consistent with these GCs also responding to CCh with a mucus secretory response mediated by influx of extracellular Ca2+ . In contrast , distal colon GCs not forming GAPs did not respond to CCh with a reduction in the average theca area or a shift in the theca size distribution , and EGTA had no effect on either average size or the size distribution ( Figure 8E and H ) . Given our prior observation that GCs in the distal colon are more likely to completely empty in response to CCh , this suggests that GCs in the distal colon that did not form GAPs , but responded to CCh with accelerated mucus secretion , completely emptied their theca and were lost to our analysis of the theca size . Notably , in vehicle treated mice , particularly in the SI villus , the theca area of GCs forming GAPs was generally smaller as compared to GCs not forming GAPs ( compare Figure 8A and E ) , consistent with GAP formation occurring in GCs with high baseline mucus secretion . However , the observation that inhibition of CCh-induced mucus secretion did not affect the ability of GCs to form GAPs suggests that the large exocytic event driving CCh-induced mucus secretion is different from the exocytic event preceding GAP formation . In aggregate , these results demonstrate that activation of select muscarinic receptor subtypes and their respective downstream signaling pathways allows GCs to respond to ACh with either GAP formation and antigen uptake , mucus secretion , or with both processes in parallel .
The intestinal epithelium is faced with the complex task of acting as a semi-permeable barrier allowing efficient nutrient absorption and controlled exposure of the immune system to lumenal antigens to sustain immune tolerance , while at the same time limiting contact with harmful agents that pass through the gastrointestinal tract . GCs have long been appreciated for their role in barrier function via production and secretion of the mucus layer that covers the intestinal surfaces ( Johansson et al . , 2008; Van der Sluis et al . , 2006 ) . Recently , the role of GCs in intestinal homeostasis has expanded to include participation in adaptive immune responses to lumenal substances via sampling and delivery of lumenal antigens to lamina propria antigen presenting cells ( Knoop et al . , 2015; McDole et al . , 2012 ) . How GCs balance these seemingly opposing tasks and the cellular mechanisms underlying lumenal antigen sampling were largely unknown . In the present study , we demonstrate that GCs acquire lumenal substances via an endocytic event that efficiently delivers fluid-phase cargo not only to lysosomes , MVBs , and the TGN , but also into the transcytotic pathway allowing the capture of lumenal substances by underlying cells ( Figure 9 ) . Such transcellular transport of lumenal solutes was not observed in adjacent enterocytes , and appears to be a feature specific to intestinal secretory cells and potentially linked to retrieval of secretory granule membrane . We cannot exclude that transcellular transport of luminal substances captured by fluid-phase endocytosis in enterocytes occurs to some degree , but if so , the efficiency of such transport must be low as expected for fluid-phase endocytic cargo in enterocytes , and below the level of detection in our assays . Endocytic retrieval of secretory granule membranes has been shown to occur in various secretory cells including neurons , enterochromaffin cells , pancreatic acinar cells , and endothelial cells ( Henkel et al . , 2001; Stevenson et al . , 2017; Wen et al . , 2012 ) and is linked to membrane recycling and regulation of the secretory capacity of the cell ( Stevenson et al . , 2017; Wen et al . , 2012 ) . The general consensus regarding exocytosis – endocytosis coupling in secretory cells is that during primary exocytosis , the process where one granule at the time is inserted into the plasma membrane , inserted membranes have to be retrieved to enable subsequent rounds of exocytosis and to restore plasma membrane structure . However , during compound exocytosis , the process where multiple secretory granules fuse with one another and release their content through a common secretion pore , large volumes of membrane retrieval are not necessarily required to sustain secretion as individual secretory granules are not inserted directly into the plasma membrane ( Liang et al . , 2017 ) . Our previous studies demonstrate that not all GCs form GAPs ( Knoop et al . , 2015; McDole et al . , 2012 ) , that GAP formation is not required to maintain the mucus barrier ( Kulkarni et al . , 2020 ) , and our current findings that blockade of GAP formation does not inhibit mucus secretion , indicate that GAP formation cannot be required to maintain mucus secretion or the mucus barrier . Though compensatory endocytosis must exist following secretion in GCs , to balance membrane trafficking , the fate of these endocytic vesicles does not necessarily intersect with the GAP pathway . Rather , we suggest that GAPs represent a specialized endocytic pathway enabling enhanced and regulated transcellular transport of gut lumenal solutes across the epithelial barrier by transcytosis . We consider it likely the GAP pathway evolved from adaptations to the secretory and membrane recovery machinery specific to this highly secretory cell type . Despite the associations between GAP formation and mucus secretion , our results demonstrate that ACh-induced GAP formation and mucus secretion are not functionally linked or dependent upon one another . Rather we found that they are regulated by separate muscarinic receptors and intracellular signaling pathways – allowing them to be performed independently or in parallel in the same cell . These observations provide further evidence that the exocytic events driving ACh-induced mucus secretion can be different from that preceding GAP formation . In spinal cord neurons , exposure to high concentrations of K+ triggers neurotransmitter release paralleled by endocytic retrieval of secretory granule membrane and uptake of high molecular weight substances from the extracellular environment . Similar to our observations demonstrating that inhibition of CCh-induced mucus secretion did not affect CCh-induced GAP formation , inhibition of K+ evoked exocytosis did not affect K+-induced endocytic retrieval of secretory granule membrane . The intact endocytic response in response to elevated K+ levels was shown to be mediated by retrieval of previously fused secretory vesicles , thus , although elevated K+ triggered secretory granule exocytosis , retrieval of previously fused granules was enough to sustain the uptake process ( Neale et al . , 1999 ) . Applied to our data this would suggest that in situations when ACh-induced mucus secretion does not occur; an intact GAP response can be maintained via retrieval of previously fused secretory granules inserted into the plasma membrane during a previous exocytotic event . In addition to secreting mucus in response to secretagogues such as ACh , GCs secrete mucus constitutively resulting in continuous insertion of secretory granule membrane into the apical membrane to be retrieved in response to ACh ( Oliver and Specian , 1990 ) . Further support in favor of the two processes being functionally separate , ACh-induced mucus secretion is initiated at the center of the GC theca ( Specian and Neutra , 1980 ) , while GAPs initiate along the lateral apical surface of the cell and progress basally . This spatial separation of the two processes may imply that GAPs primarily form by retrieval of secretory granules released during steady state , which in contrast to ACh-induced exocytosis are supplied by granules transported along the outer borders of the theca and released at the apical lateral sides ( Oliver and Specian , 1990 ) . Alternatively , ACh may induce multiple endocytic events , some originating from retrieval of secretory granule membrane , and others originating from endocytosis of the plasma membrane . Despite our findings of muscarinic receptor subtype-specific regulation of ACh-induced GAP formation and mucus secretion , muscarinic receptor antagonists are known for their receptor subtype promiscuity , and difficulties in assessing differences in in vivo metabolism of antagonists warrant caution when assessing the relative contribution of the respective receptor subtypes ( Bozkurt and Sahin-Erdemli , 2009; Erosa-Rivero et al . , 2014; Lazareno and Birdsall , 1993 ) . Nonetheless , in support of these results , we also find that the CCh-induced GAP response and the mucus secretory response use different pools of Ca2+ and different second messenger systems . Furthermore , the fact that the two processes can be performed separately supports a model of regulation by separate mechanisms . In lacrimal gland acinar cells , exposure to low concentrations of ACh was shown to increase fluid-phase endocytosis by 80% while protein release was increased only by 40% . In contrast , exposure to high concentrations of ACh was shown to increase protein release by 80% , while endocytic retrieval was increased by 40% ( Gierow et al . , 1995 ) , demonstrating that , similar to what we observe in intestinal GCs , in lacrimal acinar cells , ACh-induced exocytosis and endocytosis can , at least to some degree , occur separately . The intracellular Ca2+ concentration needed to trigger endocytosis has been shown to be lower than that needed to trigger exocytosis ( Marks and McMahon , 1998 ) . This may explain why in the steady state , ACh acting on mAChR4 in the SI and mAChR3 in the distal colon induces GAP formation ( endocytosis ) without an apparent effect on mucus secretion ( exocytosis ) ; and yet in response to exogenous administration of high concentrations of ACh , which could trigger a larger increase in intracellular Ca2+ , GCs respond with both mucus secretion and GAP formation . When evaluating the frequency of GAPs along the SI crypt villus axis , we observed that villus GCs formed GAPs at a higher frequency when compared to crypt GCs both during steady state and in response to CCh . Crypt GCs on the other hand responded to CCh with a stronger mucus secretory response resulting in complete emptying of a subset of GCs and partial emptying of the remaining GCs , which is in accordance with previously published data ( Phillips , 1992 ) . These results indicate that in the SI , in response to ACh , villus GCs preferentially perform antigen uptake and delivery to the immune system , while crypt GCs have the capacity to secrete large volumes of mucus when needed . Baseline mucus secretion has been shown to be higher in the SI villus as compared to the crypts which may explain why baseline GAP formation is more prevalent in SI villus GCs as there would be more membrane to retrieve to form GAPs ( Schneider et al . , 2018 ) . In GCs , mAChR4 signaling ( the pathway regulating GAP formation in the SI and proximal colon ) is inhibited by activation of the epidermal growth factor receptor ( EGFR ) , which in the proximal colon can occur by GC intrinsic sensing of gut microbial products via Toll-like receptors or by the presence of EGFR ligands in the SI and colon lumen ( Knoop et al . , 2015 ) . The benefit of mucus secretion not being functionally linked to GAP formation is apparent in the proximal colon where both steady-state and CCh-induced GAP formation are inhibited , while baseline and CCh-induced mucus secretion remain intact ( Ermund et al . , 2013; Knoop et al . , 2015 ) . Overriding the inhibition of mAChR4 signaling to allow GAPs to form in the proximal colon has the undesired outcome of bacterial translocation across GAPs and induction of inflammatory responses ( Knoop et al . , 2016 ) . A further example of the importance of controlling GAP formation is that during enteric infection with Salmonella typhimurium , mAChR4 signaling and GAP formation in the SI is inhibited to prevent lumenal antigen delivery to the immune system and limit inflammatory T cell responses to dietary antigens ( Kulkarni et al . , 2018 ) . Thus , functional uncoupling of ACh induced mucus secretion from GAP formation , the usage of different mAChRs to perform mucus secretion and GAP formation , and the ability to regulate mAChR4 signaling in GCs , underlies the ability to control when and where GAPs are formed and are essential in maintaining intestinal health . Although the focus of the present study was to explore the cellular basis of ACh-induced GAP formation , the ability of GCs to acquire lumenal substances is not restricted to activation of muscarinic receptor signaling and additional ligands and receptors activating cADPr production , such as IL13 , have been demonstrated to induce GCs to take up lumenal substances in some settings ( Noah et al . , 2019 ) . This raises the possibility that other ligands , whose receptors are expressed by GCs and induce cADPr production , may likewise induce GCs to acquire lumenal substances ( Deshpande et al . , 2004; Tliba et al . , 2004 ) . While it is becoming apparent that physiologic ACh-induced GAP formation supports antigen-specific T cell responses and tolerance to the lumenal content , the functional consequences and downstream events resulting from GC uptake of lumenal substances induced by other stimuli remain to be explored . In summary , our results define the basis by which GCs sample lumenal antigens for delivery to the immune system and how this process is balanced with mucus secretion within the same cell . Our observations indicate that GCs have evolved a pathway of fluid-phase endocytosis that efficiently serves the transcytotic pathway for non-specific uptake of lumenal substances that are sampled by sub-epithelial phagocytic cells . Compared to neighboring enterocytes , the fluid-phase endocytic GAP pathway of GCs is remarkable for its efficient transport of lumenal substances across the epithelial barrier . Perhaps GAPs evolved from an endocytic system originally used in secretory cell types to recycle secretory granule membranes , but adapted to ensure that during situations when the lumenal environment is toxic for antigen sampling , the GC can suppress GAP formation while retaining its ability to secrete mucus for maintenance of the mucus barrier . In total , our observations reveal the basis by which GCs perform the function of lumenal antigen delivery to the immune system , and provide mechanistic insights into how the critical roles of barrier maintenance and antigen delivery are achieved within the same cell .
Data is presented as mean ± standard error or the mean . Statistical analysis was performed using GraphPad Prism 7 ( GraphPad Software Inc , San Diego , CA ) . Analyses between two groups were performed using the Student’s t-test . Comparisons between three groups or more were performed using a one-way ANOVA with Dunnett’s post hoc test for correction of multiple comparisons . A cut-off of p < 0 . 05 was used for statistical significance . Details regarding the statistical test used in the respective experiments are indicated in the figure legends together with populations size . All experimental groups consist of three or more technical replicates . Results of statistical tests are indicated in the figures and in the text . | Cells in the gut need to be protected against the many harmful microbes which inhabit this environment . Yet the immune system also needs to ‘keep an eye’ on intestinal contents to maintain tolerance to innocuous substances , such as those from the diet . The ‘goblet cells’ that are part of the gut lining do both: they create a mucus barrier that stops germs from invading the body , but they also can pass on molecules from the intestine to immune cells deep in the tissue to promote tolerance . This is achieved through a ‘GAP’ mechanism . A chemical messenger called acetylcholine can trigger both mucus release and the GAP process in goblet cells . Gustafsson et al . investigated how the cells could take on these two seemingly opposing roles in response to the same signal . A fluorescent molecule was introduced into the intestines of mice , and monitored as it pass through the goblet cells . This revealed how the GAP process took place: the cells were able to capture molecules from the intestines , wrap them in internal sack-like vesicles and then transport them across the entire cell . To explore the role of acetylcholine , Gustafsson et al . blocked the receptors that detect the messenger at the surface of goblet cells . Different receptors and therefore different cascades of molecular events were found to control mucus secretion and GAP formation; this explains how the two processes can be performed in parallel and independently from each other . Understanding how cells relay molecules to the immune system is relevant to other tissues in contact with the environment , such as the eyes , the airways , or the inside of the genital and urinary tracts . Understanding , and then ultimately harnessing this mechanism could help design of new ways to deliver drugs to the immune system and alter immune outcomes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2021 | Intestinal goblet cells sample and deliver lumenal antigens by regulated endocytic uptake and transcytosis |
Many cellular processes are driven by cytoskeletal assemblies . It remains unclear how cytoskeletal filaments and motor proteins organize into cellular scale structures and how molecular properties of cytoskeletal components affect the large-scale behaviors of these systems . Here , we investigate the self-organization of stabilized microtubules in Xenopus oocyte extracts and find that they can form macroscopic networks that spontaneously contract . We propose that these contractions are driven by the clustering of microtubule minus ends by dynein . Based on this idea , we construct an active fluid theory of network contractions , which predicts a dependence of the timescale of contraction on initial network geometry , a development of density inhomogeneities during contraction , a constant final network density , and a strong influence of dynein inhibition on the rate of contraction , all in quantitative agreement with experiments . These results demonstrate that the motor-driven clustering of filament ends is a generic mechanism leading to contraction .
The mechanics , motions , and internal organization of eukaryotic cells are largely determined by the cytoskeleton . The cytoskeleton consists of filaments , such as actin and microtubules , and molecular motors , which consume chemical energy to exert forces on and arrange the filaments into large-scale networks . Motor proteins , including dynein and roughly 14 different families of kinesin ( Wordeman , 2010 ) , organize microtubules to form the spindle , which segregates chromosomes during cell division . The motor protein myosin organizes actin filaments into networks which drive cell motility , polarity , cytokinesis , and left-right symmetry breakage ( Mitchinson and Cramer , 1996; Mayer et al . , 2010; Naganathan et al . , 2014 ) . The non-equilibrium nature of motor activity is essential for the organization of the cytoskeleton into these diverse sub-cellular structures , but it remains unclear how the interactions between filaments , different motor proteins , and other biomolecules influence the behaviors of the networks they form . In particular , it is difficult to extrapolate from the biochemical properties of motors characterized in reconstituted systems to the biological function of those motors in vivo . To address this question , we study self-organization of cytoskeletal filaments in Xenopus extracts , which recapitulate the biochemical complexity of the in vivo system . The self-organization of cytoskeletal filaments has been extensively studied in cell extracts and in reconstituted systems of purified components . Actin can form macroscopic networks that exhibit a myosin-dependent bulk contraction ( Murrell and Gardel , 2012; Bendix et al . , 2008; Köhler and Bausch , 2012; Alvarado et al . , 2013; Szent-Györgyi , 1943 ) . Microtubule networks purified from neuronal extracts have also been observed to undergo bulk contraction ( Weisenberg and Cianci , 1984 ) , while microtubules in mitotic and meiotic extracts are found to assemble into asters ( Gaglio et al . , 1995; Mountain et al . , 1999; Verde et al . , 1991 ) . Aster formation in meiotic Xenopus egg extracts is dynein-dependent , and has been proposed to be driven by the clustering of microtubule minus ends by dynein ( Verde et al . , 1991 ) . It has also been suggested that dynein binds to the minus ends of microtubules in spindles and clusters the minus ends of microtubules to form spindle poles ( Heald et al . , 1996; Burbank et al . , 2007; Khodjakov et al . , 2003; Goshima et al . , 2005; Elting et al . , 2014 ) and dynein has been shown to accumulate on microtubule minus ends in a purified system ( McKenney et al . , 2014 ) . Purified solutions of microtubules and kinesin can also form asters ( Nédélec et al . , 1997; Hentrich and Surrey , 2010; Urrutia et al . , 1991 ) , or under other conditions , dynamic liquid crystalline networks ( Sanchez et al . , 2012 ) . Hydrodynamic theories have been proposed to describe the behaviors of cytoskeletal networks on length scales that are much greater than the size of individual filaments and motor proteins ( Prost et al . , 2015 , Marchetti et al . , 2013 ) . These phenomenological theories are based on symmetries and general principles of non-equilibrium physics , with the details of the microscopic process captured by a small number of effective parameters . As hydrodynamic theories are formulated at the continuum level , they cannot be used to derive the values of their associated parameters , which must be obtained from more microscopic theories ( Prost et al . , 2015 , Marchetti et al . , 2013 ) or by comparison to experiments ( Mayer et al . , 2010; Brugués and Needleman , 2014 ) . A key feature of networks of cytoskeletal filaments and motor proteins that enters hydrodynamic theories , and differentiates these non-equilibrium systems from passive polymer networks , is the presence of additional , active stresses ( Prost et al . , 2015 , Marchetti et al . , 2013 ) . These active stresses can be contractile or extensile , with profound implications for the large-scale behavior of cytoskeletal networks . Contractile stresses can result from a preferred association of motors with filament ends ( Kruse and Jülicher , 2000; Hyman and Karsenti , 1996 ) , nonlinear elasticity of the network ( Liverpool et al . , 2009 ) , or the buckling of individual filaments ( Murrell and Gardel , 2012; Lenz , 2014; Soares e Silva et al . , 2011 ) . Extensile active stresses can arise from polarity sorting or result from the mechanical properties of individual molecular motors ( Gao et al . , 2015; Blackwell et al . 2015 ) . In networks with dynamically growing and shrinking filaments , polymerization dynamics can also contribute to the active stress . Experimentally , acto-myosin systems ( Murrell and Gardel , 2012; Bendix et al . , 2008; Köhler and Bausch , 2012; Alvarado et al . , 2013; Szent-Györgyi , 1943 ) and microtubule networks from neuronal extracts ( Weisenberg and Cianci , 1984 ) are observed to be contractile , while purified solutions of microtubules and kinesin can form extensile liquid crystalline networks ( Sanchez et al . , 2012 ) . It is unclear which microscopic properties of filaments and motor proteins dictate if the active stress is contractile or extensile in these different systems . Here , we investigate the motor-driven self-organization of stabilized microtubules in Xenopus meiotic egg extracts . These extracts are nearly undiluted cytoplasm and recapitulate a range of cell biological processes , including spindle assembly and chromosome segregation ( Hannak and Heald , 2006 ) . We have discovered that , in addition to microtubules forming asters in this system as previously reported ( Verde et al . , 1991 ) , the asters assemble themselves into a macroscopic network that undergoes a bulk contraction . We quantitatively characterized these contractions and found that their detailed behavior can be well understood using a simple coarse-grained model of a microtubule network in which dynein drives the clustering of microtubule minus ends . This end clustering mechanism leads to a novel form of active stress , which drives the system to a preferred microtubule density . Our results suggest that the dynein-driven clustering of microtubule minus ends causes both aster formation and network contraction , and have strong implications for understanding the role of dynein in spindle assembly and pole formation . Furthermore , the close agreement we find between experiments and theory demonstrates that simple continuum models can accurately describe the behavior of the cytoskeleton , even in complex biological systems .
To further study the motor-induced organization of microtubules , we added 2 . 5 μM Taxol to Xenopus egg extracts and loaded them into microfluidic channels ( Figure 1A ) . Taxol causes microtubules to rapidly assemble and stabilize ( Mitchison et al . , 2013 ) , which allowed us to decouple the effects of motor-driven self-assembly from the complicating effects of polymerization-depolymerization dynamics . In some regions of the channel , microtubules organized into asters ( Figure 1B ) as observed previously ( Verde et al . , 1991 ) . A NUMA antibody was used to locate microtubule minus ends ( Mitchison et al . , 2013 ) , and was found to localize to the aster core , confirming the polarity of the aster ( Gaglio et al . , 1995 ) . Isolated asters were found to interact and coalesce ( Figure 1C , Video 1 ) . In other regions of the channel , microtubules formed networks of aster-like structures ( Figure 1D ) , which were highly dynamic and exhibited large-scale motion that persisted for several tens of seconds ( Figure 1E , Video 2 ) . NUMA was found to localize to the interior of these structures , confirming their aster-like nature ( Figure 1F , G ) . 10 . 7554/eLife . 10837 . 003Figure 1 . Stabilized microtubules form asters in Xenopus egg extracts . ( A ) Experiments were performed in thin rectangular channels of width W0 , height H0 , and length L0 . ( B ) In some regions of the channel , microtubules organize into asters , with minus ends localized in the aster core ( Scale bar , 5 μm ) . ( C ) Isolated asters fuse together over minute timescales ( Scale bar , 5 μm ) . ( D ) Aster-like structures form in other regions of the channel ( Scale bar , 10 μm ) ( E ) Aster-like structures show large scale movement on minute timescales . ( Scale bar , 25 μm ) . ( F ) NUMA localizes to the network interior ( Scale bar , 20 μm ) . ( G ) Closeup of aster-like structure showing NUMA localized on the interior ( Scale bar , 10 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 00310 . 7554/eLife . 10837 . 004Video 1 . Isolated asters undergo coalescence . Taxol stabilized microtubules in Xenopus oocyte extracts self-organize into asters that can then coalesce . The mageneta channel depicts microtubules while the green channel depicts NUMA localization , here used as a proxy for microtubule minus ends . Time is shown in minutes : seconds . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 00410 . 7554/eLife . 10837 . 005Video 2 . Microtubules organize into dynamic aster-like structures . In other regions of the channel , microtubules organize into aster-like structures that exhibit large-scale movement on the minute timescale . Time is shown in minutes : seconds . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 005 To characterize these large-scale motions , we next imaged networks at lower magnification , obtaining a field of view spanning the entire channel width . The networks , which initially filled the entire channel ( width W0 = 1 . 4 mm ) , underwent a strong contraction , which was uniform along the length of the channel ( Figure 2A , Video 3 ) . The contractile behavior of these microtubule networks is highly reminiscent of the contractions of actin networks in these extracts ( Bendix et al . , 2008 ) , but in our experiments actin filaments are not present due to the addition of 10 μgmL Cytochalasin D . We characterized the dynamics of microtubule network contractions by measuring the width , W ( t ) , of the network as a function of time ( Figure 2B ) . Occasionally , we observed networks tearing along their length ( Video 4 ) , yet these tears seemed to have little impact on the contraction dynamics far from the tearing site , arguing that the Poisson ratio of the network is ≈ 0 . We then calculated the fraction contracted of the network: ( 1 ) ϵ ( t ) =W0-W ( t ) W0 , 10 . 7554/eLife . 10837 . 006Figure 2 . Stabilized microtubules form a contractile network in Xenopus egg extracts . ( A ) Low magnification imaging shows that microtubules form a contractile network ( Scale bar , 500 μm ) . ( B ) The width of the microtubule network decreases with time ( n = 6 experiments ) . ( Inset ) Representative plot of ϵ ( t ) ( Blue line ) and fit from ( Equation 2 ) ( Pink line ) , with ϵ∞=0 . 81 , τ=3 . 49 min , Tc=1 . 06 min . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 00610 . 7554/eLife . 10837 . 007Figure 2—figure supplement 1 . Plots of ϵ ( t ) from data in Figure 1F ( Blue lines ) along with fits from ( Equation 2 ) ( Pink lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 00710 . 7554/eLife . 10837 . 008Video 3 . Microtubule networks undergo a spontaneous bulk contraction . Low magnification imaging of the channels reveals that microtubules organize into a macroscopic network that spontaneously contracts on the millimeter length scale . Time is shown in minutes : seconds . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 00810 . 7554/eLife . 10837 . 009Video 4 . Microtubule networks can undergo tearing . During contraction , tears can develop in the microtubule network , causing the network to break . Time is shown in minutes : seconds . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 009 The time course of ϵ ( t ) was found to be well fit by an exponential relaxation: ( 2 ) ϵ ( t ) ≃ϵ∞1-e- ( t-Tc ) τ , where ϵ∞ is the final fraction contracted , τ is the characteristic time of contraction , and Tc is a lag time before contraction begins ( Figure 2B , inset , Figure 2—figure supplement 1 ) . We next sought to investigate which processes determine the timescale of contraction and the extent that the network contracts . For this , we exploited the fact that different mechanisms predict different dependence of the timescale τ on the channel dimensions . For instance , in a viscoelastic Kelvin-Voight material driven to contract by a constant applied stress , τ = η/E depends solely on the viscosity η and the Young’s modulus E and is independent of the size of the channel ( Oswald , 2009 ) . In contrast , in a poroelastic material driven by a constant stress , τ ∝ W02 ( Coussy , 2004 ) , where W0 is the width of the channel . Thus , studying how τ varies with channel width provides a means to test the validity of these models . We fabricated microfluidic channels of varying width , W0 = 1 . 4 , 0 . 9 , 0 . 44 , and 0 . 16 mm , all with height H0 = 125 μm , loaded the channels with extracts supplemented with 2 . 5 μM Taxol , and imaged the networks at low magnification ( Figure 3A , Video 5 ) . Results for each channel width were averaged together to produce master curves of the width , W ( t ) ( Figure 3B ) , and fraction contracted , ϵ ( t ) ( Figure 3C ) , of the networks in each channel . Visual inspection of the fraction-contracted curves , ϵ ( t ) , reveals that networks in smaller channels contract faster , but all reach a similar final fraction contracted ( Figure 3C ) . To quantify these trends , we fit the ϵ ( t ) curves using ( Equation 2 ) and extracted the characteristic time to contract , τ , and the final fraction contracted , ϵ∞ , for each channel width . We find that the dependence of τ on channel width is inconsistent with the time of contraction resulting from either viscoelastic or poroelastic timescales , which would predict constant and quadratic scalings respectively ( Figure 3D ) . We next explored the influence of channel height H0 ( H0 = 75 , 125 , 150 μm , all with width W0 = 1 . 4 mm ) and found that τ does not significantly vary in these channels ( Figure 3E ) . 10 . 7554/eLife . 10837 . 010Figure 3 . Contraction dynamics in channels of different width provide a means to test potential contraction mechanisms . ( A ) Microtubules form contractile networks in channels with various widths ( Scale bar , 500 μm , t=10 min ) . ( B ) Width of the networks as a function of time in channels with various widths . ( C ) Fraction contracted as a function of time , ϵ ( t ) , calculated from the data in B . The networks all contract to a similar final fraction , while the timescale of contraction differs . ( D ) The scaling of the characteristic time , τ , with channel width does not vary as W02 , as would result for a poroelastic timescale , and is not a constant , independent of width , as would result from a viscoelastic timescale . The scaling is well described by an active fluid model ( green line analytic scaling , fit to ( Equation 6 ) ; green dots numerical solution ) . ( E ) The characteristic time , τ , is found to be independent of channel height . The dashed line is the mean value of τ . ( F ) ϵ∞ is constant for all channel widths and heights , indicating that the network contracts to a constant final density . The dashed line is the mean value of ϵ∞ . All panels display mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 01010 . 7554/eLife . 10837 . 011Video 5 . Network contraction in channels of varying width . Devices were fabricated with different widths . Each video panel depicts a representative experiment using channels of the given width . Time is shown in minutes : seconds . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 011 In all cases , the networks contracted to a similar final fraction , ϵ∞ , of ≈ 0 . 77 , irrespective of channel geometry ( Figure 3F ) . Since the Taxol concentration was held constant , all experiments started with the same initial density of microtubules , regardless of the dimensions of the channel . Thus , all networks contracted to the same final density . By using fluorescence intensity as a proxy for tubulin concentration ( see Materials and methods ) , we estimate the final concentration of tubulin in the network to be ρ0≈ 30 μM . Remarkably , this is comparable to the concentration of microtubules in reconstituted meiotic spindles in Xenopus extracts ( Needleman et al . , 2010 ) , which is ≈ 60 μM . As neither the simple viscoelastic nor poroelastic models are consistent with these results , we sought to construct an alternative model of the contraction process . Since Taxol stabilizes microtubules in these experiments , the density of microtubules ρ is conserved throughout the contraction process , implying ( 3 ) ∂tρ=-∇→⋅ ( ρv→ ) , where v→ is the local velocity of the microtubule network . The velocity v→ is set by force balance . If the relevant timescales are long enough that the microtubule network can be considered to be purely viscous , and if the network’s motion results in drag , then the equation for force balance is ( 4 ) η∇2v→-γv→=∇→⋅σ , where η and γ are the viscosity and drag coefficients , respectively , and σ is an active stress caused by motor proteins which drive the contraction of the microtubule network . The observation that the timescale of contraction , τ , is independent of channel height ( Figure 3E ) shows that the drag does not significantly vary with channel height , and thus could arise from weak interactions between the microtubule network and the device wall . We obtain an expression for the active stress , σ , by considering the microscopic behaviors of microtubules and motor proteins . As the contracting networks consist of microtubule asters ( Figure 1D , E ) , and microtubule asters in meiotic extracts are thought to assemble by the dynein-induced clustering of microtubule minus ends ( Verde et al . , 1991 ) , we hypothesize that the contraction process is also driven by dynein pulling microtubule minus ends towards each other ( Figure 4A ) . 10 . 7554/eLife . 10837 . 012Figure 4 . Cartoon of the microscopic model underlying the active fluid theory of network contractions by minus end clustering . ( A ) Microtubule sliding by dynein drives microtubule minus ends together . ( B ) Minus end clustering leads to the formation of aster-like structures . Due to steric interactions between microtubules , there is an upper limit to the local microtubule density . ( C ) The microtubule network is composed of interacting asters . Motor activity driving aster cores together leads to bulk contraction of the network . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 012 In an orientationally disordered suspension of microtubules , we expect dynein mediated collection of microtubule minus ends to drive a contractile stress which is proportional to the number of motor molecules m and the local density of microtubules ρ , ( see Appendix ) . As only a finite number of microtubules can fit near the core of an aster , steric collisions will counteract the contractile stress at high densities ( Figure 4B ) . Since most motion in the suspension is motor driven , thermal collisions can be ignored , and the extensile stress driven by steric interactions will be be proportional to the number of motor molecules m and quadratic in the local density of microtubules ρ ( see Appendix ) . Taken together , these two effects lead to the active stress ( 5 ) σ=sρ ( ρ-ρ0 ) 𝕀 , where s is the strength of the active stress , ρ0 is the final density at which the effects of dynein mediated clustering and steric repulsion between microtubules balance , and 𝕀 is a unit tensor ( see Appendix ) . Importantly , since the contractile and extensile parts of the active stress both depend linearly on the number of motor molecules , the prefered density ρ0 that the suspension will reach after contraction depends only on the interaction geometry between microtubules and motors and not on the actual number of active motors . Only the strength s of the active stress will be affected if the number of active motors could be changed . Taken together , Equations ( 3 , 4 , 5 ) constitute an active fluid theory of microtubule network contraction by minus end clustering . We note that this theory could be reformulated , essentially without change , as the clustering of aster cores , again driven by dynein mediated clustering of minus ends . Isotropy of interactions remains a fundamental assumption . We first investigated if this active fluid theory can explain the dependence of the timescale of contraction on sample geometry . An analysis of the equations of motion , Equations ( 3 , 4 , 5 ) , near equilibrium predicts that the timescale of contractions obeys ( 6 ) τ ( W0 ) =αηsρ02+βγsρ02W02 , where α = 2 . 2 ± 0 . 05 and β = 0 . 085 ± 0 . 006 are dimensionless constants , which we determined numerically ( see Appendix ) . This predicted scaling is both consistent with the experimental data and simulations of the full theory ( Figure 3D ) . Fitting the scaling relationship to the data allows combinations of the parameters to be determined , giving η∕ ( sρ02 ) = 0 . 82 ± 0 . 20 min and γ∕ ( sρ02 ) = 1 . 0 ×10-5±0 . 7×10-5 min∕ ( μm2 ) ( mean ± standard error ) . Combining this measurement with an estimate for the network viscosity taken from measurements in spindles of η ≈ 2×102Pa⋅s ( Shimamoto et al . , 2011 ) , we can estimate the dynein-generated active stress to be sρ02 ≈ 4Pa which is consistant with having ≈ 0 . 4 dynein per microtubule minus end each exerting an average force of 1 pN ( Nicholas et al . , 2015 ) . To further explore the validity of the active fluid theory of contraction by microtubule minus end clustering , we explored other testable predictions of the theory . This theory predicts that: ( i ) the preferred density of the network ρ0 is constant and does not depend on the initial conditions . This is consistent with the constant ϵ∞ measured experimentally ( Figure 3F ) ; ( ii ) since contractions are driven by stress gradients ( Equation 4 ) and stress depends on microtubule density ( Equation 5 ) the density discontinuity at the edge of the network should produce large-stress gradients , leading to an inhomogeneous density profile in the network during contraction; ( iii ) the magnitude of the active stress , s , is proportional to the number of active motors , but the final density of the network , ρ0 , is independent of the number of molecular motors ( see Appendix ) . Thus , reducing the number of motors should lead to slower contractions , but still yield the same final density . We first examined prediction ( ii ) , that the stress discontinuity at the edge of the network should lead to a material buildup in the film . To test this , we averaged the fluorescence intensity along the length of the channel ( see Materials and Methods ) and found that the microtubule density does indeed increase at the network’s edge during contraction ( Figure 5A ) . We next explored if the inhomogeneous density profile could be quantitatively explained by our active fluid theory . We numerically solved Equations ( 3 , 4 , 5 ) and used least squares fitting to determine the simulation parameters which most closely matched the experimentally measured profiles ( Figure 5B ) , yielding η∕ ( sρ02 ) = 0 . 82 ± 0 . 03 min , γ∕ ( sρ02 ) = 6 . 1 ± 0 . 1×10-6 min/ ( μm2 ) , and ρinitial∕ρ0 = 0 . 32 ± 0 . 01 ( mean ± s . e . m . , n=4 experiments ) . Within error , these values are the same as those determined from the dependence of the timescale of contraction on channel width ( Figure 3D ) . The simulated profiles closely match the experimental ones for most of the contraction ( Figure 5B ) , but at late times the simulated inhomogeneities dissipate in contrast to the experiments ( Figure 5—figure supplement 1 ) . This might be caused by a long-term aging of the network that is not incorporated into our simple model . To confirm that the density buildup was due to an increased velocity near the network’s edge , we measured the velocity throughout the network using Particle Image Velocimetry ( PIV , see Materials and Methods ) ( Figure 5C ) and found that the velocities increase superlinearly with distance from the network’s center , as predicted ( Figure 5D ) . 10 . 7554/eLife . 10837 . 013Figure 5 . Microtubule density increases at the network’s edges during contraction . ( A ) Time series of contraction showing intensity averaged along the length of the channel . The average intensity peaks at the network’s edges due to increased local microtubule density . ( Scale bars , 500 μm ) ( B ) Comparison of measured density profiles ( solid lines ) with density profiles from simulation ( dashed lines ) . Data are plotted at 1 min intervals starting at t = 40 s . ( C ) Representative frame from PIV showing the network’s local velocity component along the network’s width . ( D ) Comparison between measured ( solid red line ) and simulated ( dashed red line ) velocity along the width of the channel at t = 80 s . The measured and simulated velocities increase superlinearly with distance from the center of the network , as can be seen by comparison to a linear velocity profile ( dashed black line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 01310 . 7554/eLife . 10837 . 014Figure 5—figure supplement 1 . Comparison between measured ( solid lines ) and simulated ( dashed lines ) density profiles . Data are plotted at 2 min intervals starting at t = 40 s . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 014 Finally , we sought to determine the molecular basis of the contraction process , and check prediction ( iii ) , that the number of motors driving the contraction affects the rate of contraction , but not the final density the network contracts to . Aster assembly is dynein-dependent in Xenopus egg extracts ( Gaglio et al . , 1995; Verde et al . , 1991 ) , and dynein ( Heald et al . , 1996 ) and Kinesin-5 ( Sawin et al . , 1992 ) are two of the most dominant motors in spindle assembly in this system . We inhibited these motors to test their involvement in the contraction process . Extracts supplemented with STLC for Kinesin-5 inhibition or p150-CC1 for dynein inhibition were loaded into channels with a width , W0 , of 0 . 9 mm and imaged at low magnification . Inhibiting Kinesin-5 had little effect on the contraction process ( Figure 6—figure supplement 1 ) . In contrast , inhibiting dynein caused a dose-dependent slowdown of the contraction ( Figure 6A ) . In spindle assembly , inhibiting Kinesin-5 suppresses the morphological changes caused by dynein inhibition ( Mitchison et al . , 2005 ) . We , therefore , tested how simultaneously inhibiting both motors influences the contraction process , but found that the effects of dynein inhibition were not rescued by the simultaneous inhibition of Kinesin-5 ( Figure 6—figure supplement 1 ) , suggesting that in this context , Kinesin-5 is not generating a counteracting extensile stress . This further suggests the possibility that in the spindle , the role of Kinesin-5 may be in orienting , polarity sorting , and sliding microtubules as opposed to active stress generation . Curves of ϵ ( t ) were fit using Equation ( 2 ) to extract the final fraction contracted , ϵ∞ , and the characteristic time of contraction , τ . By varying the concentration of p150-CC1 , the characteristic time , τ , could be tuned over a wide range from ≈ 3 min to ≈ 75 min ( Figure 6B ) . Fitting a sigmoid function to the τ vs . p150-CC1 concentration curve yields an EC50 value of 0 . 22 ± . 02 μM ( mean ± standard error ) , similar to the value of ≈ 0 . 3 μM reported for the effect of p150-CC1 on spindle length in Xenopus extracts ( Gaetz and Kapoor , 2004 ) , which is consistent with active stress generated by dynein being required for pole focusing . Despite this large change in the contraction timescale , we found no apparent differences in ϵ∞ ( Figure 6C ) . Thus , the microtubule networks contract to approximately the same final density irrespective of the concentration of p150-CC1 . The observation that inhibiting dynein affects the timescale of contraction but not the final density to which the network contracts is consistent with the predictions of our model . We note that even at the highest p150-CC1 concentrations used , the network still undergoes a bulk contraction . This could possibly be due to incomplete inhibition of dynein by p150-CC1 , or by another motor protein present in the extract that also contributes to the contraction process . As the characteristic time , τ∝1s , by comparing the characteristic times in the uninhibited and 2 μM p150-CC1 cases , we can estimate that the strength of the active stress , s , in the 2 μM p150-CC1 condition is only ≈ 4% of the strength of the active stress in the uninhibited case , arguing that even if another motor is involved in the contraction , dynein contributes ≈ 96% of the active stress . 10 . 7554/eLife . 10837 . 015Figure 6 . Network contraction is a dynein-dependent process . ( A ) Fraction contracted as a function of time , ϵ ( t ) , when dynein is inhibited using p150-CC1 . ( B ) The characteristic time of contraction , τ , increases with increasing p150-CC1 concentration . Solid green line indicates fit of sigmoid function . ( C ) ϵ∞ has no apparent variation with p150-CC1 concentration . Solid green line indicates the mean value of ϵ∞ . All panels display mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 01510 . 7554/eLife . 10837 . 016Figure 6—figure supplement 1 . Inhibition of Kinesin-5 has little effect on the contraction process . ( A ) Comparison of ϵ ( t ) curves for samples where Kinesin-5 was inhibited using STLC and control where no STLC was added . ( B ) Simultaneous inhibition of dynein with p150-CC1 and Kinesin-5 with STLC does not rescue the effects of dynein inhibition alone . All panels display mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 01610 . 7554/eLife . 10837 . 017Figure 6—figure supplement 2 . Plots of ϵ ( t ) from experiments with 2 μM p150-CC1 ( blue lines ) along with fits from Equation ( 2 ) ( pink lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10837 . 017
Here , we have shown that networks of stabilized microtubules in Xenopus egg extracts undergo a bulk contraction . By systematically varying the width of the microfluidic channel in which the network forms , we demonstrated that the timescale of contraction is not a poroelastic or viscoelastic timescale . A simple active fluid model of network contraction by dynein-driven clustering of microtubule minus ends correctly predicts the dependence of the contraction timescale on channel width , the nonuniform density profile in the network during contraction , and that inhibiting dynein affects the timescale of contraction but not the final density that the network contracts to . Parameters of this model can be measured by the scaling of the contraction timescale with channel width and by a detailed analysis of the inhomogeneities in the network that develop during contraction . Both methods give similar values . Our results demonstrate that the behaviors of a complex biological system can be quantitatively described by a simple active matter continuum theory . These active matter theories aim to describe the behavior of cytoskeletal systems at large-length scales and long-timescales by effectively averaging all of the molecular complexity into a small set of coarse-grained parameters . Previously , these theories have been predominately applied to describe biological systems near non-equilibrium steady states ( Prost et al . , 2015; Brugués et al . , 2014 ) . In the present work , we augment previous theories with a nonlinear active stress term derived from microscopic considerations to capture the far from steady state dynamics of the contraction process . This approach allows us to quantitatively explain our experimental results using a theory with only four parameters , while a complete microscopic model would require understanding the behavior of the thousands of different proteins present in Xenopus egg extracts . Furthermore , the considerations of the model are general , and it will be interesting to consider whether the end clustering mechanism proposed here could contribute to contraction in actin networks as well . In our model , the active stress which drives network contraction results from the motor-induced clustering of microtubule minus ends , the same process thought to be responsible for aster formation and spindle pole focusing ( Gaglio et al . , 1995; Mountain et al . , 1999; Verde et al . , 1991 , Elting et al . , 2014; Heald et al . , 1996; Burbank et al . , 2007; Khodjakov et al . , 2003; Goshima et al . , 2005 ) . Our results , and previous data ( Verde et al . , 1991; Heald et al . , 1996; Burbank et al . , 2007 ) , are consistent with minus end clustering in Xenopus egg extracts primarily arising from the activity of dynein . The ability of dynein to cluster microtubule minus ends could result from dynein being able to accumulate on the minus end of one microtubule , while simultaneously walking towards the minus end of another ( Hyman and Karsenti , 1996; McKenney et al . 2014; Figure 4A ) . There is indication that such behaviors may indeed occur in spindles ( Elting et al . , 2014 ) , and pursuing a better understanding of those processes is an exciting future direction that will help to clarify the function of dynein in spindles . The observation that microtubule networks contract in Xenopus egg extracts suggests that motor-induced stresses in spindles are net contractile and not extensile as previously assumed ( Brugués and Needleman , 2014 ) . The contribution of dynein to spindle pole focusing may ultimately be due to these contractile stresses . The presence of contractile stresses from dynein might also explain both the observation that the fusion of spindles is dynein-dependent ( Gatlin et al . , 2009 ) , and the apparently greater cohesion between microtubules at spindle poles , ( where dynein is localized [Gatlin et al . , 2010] ) . It is unclear what processes set the density of microtubules in the spindle , and the finding that the active stress generated from minus end clustering saturates at a preferred microtubule density could play an important role .
CSF-arrested extracts were prepared from Xenopus llaevis oocytes as previously described ( Hannak and Heald , 2006 ) . Crude extracts were sequentially filtered through 2 . 0 , 1 . 2 , and 0 . 2 micron filters , frozen in liquid nitrogen , and stored at −80°C until use . Channel negatives were designed using AutoCAD 360 ( Autodesk ) and Silhouette Studio ( Silhouette America ) software , cut from 125-micron-thick tape ( 3M Scotchcal , St . Paul , MN ) using a Silhouette Cameo die cutter , and a master was made by adhering channel negatives to the bottom of a petri dish . PDMS ( Sylgard 184 , Dow Corning , Midland , MI; 10:1 mixing ratio ) was cast onto the masters and cured overnight at 60°C . Devices and coverslips were each corona treated with air plasma for 1 min before bonding . Channels containing a degassed solution of 5 mg/mL BSA ( J . T . Baker , Center Valley , PA ) supplemented with 2 . 5% w/w Pluronic F127 ( Sigma , St . Louis , MO ) were incubated overnight at 12°C . Unless stated otherwise , the microfluidic devices had a length of 18 mm , a height of 0 . 125 mm , and a width of 1 . 4 mm . GST-tagged p150-CC1 plasmid was a gift from Thomas Surrey ( Uteng et al . , 2008 ) . GST-p150-CC1 was expressed in E . coli BL21 ( DE3 ) -T1R ( Sigma ) for 4 hr at 37°C . The culture was shifted to 18°C for 1 hr before adding 0 . 2 mM IPTG and the culture was grown overnight at 18°C . Cells were centrifuged , resuspended in PBS supplemented with Halt Protease Inhibitor Cocktail ( Thermo Scientific , Rockford , IL ) and benzonase ( Novagen , San Diego , CA ) before lysis by sonication . GST-p150-CC1 was purified from clarified lysate using a GSTrap column FF ( G . E . Healthcare , Sweden ) as per the manufacturer’s instructions . GST-p150-CC1 was dialyzed overnight into 20 mM Tris-HCl , 150 mM KCl , and 1 mM DTT . The GST tag was cleaved using Prescission Protease ( overnight incubation at 4°C ) . After removing free GST and Prescission Protease using a GSTrap FF column , p150-CC1 was concentrated , frozen in liquid nitrogen , and stored at -80°C until use . 20 μL aliquots of filtered extract were supplemented with ∼1 μM Alexa-647 labeled tubulin and 2 . 5 μM Taxol before being loaded into channels . For dynein inhibition experiments , 1 μL of either p150-CC1 or buffer alone was added to the extract immediately before Taxol addition . For Kinesin-5 inhibition experiments , 100 μM STLC ( Sigma Aldrich ) was added concurrently with Taxol . Channels were sealed with vacuum grease and imaged using a spinning disk confocal microscope ( Nikon Ti2000 , Yokugawa CSU-X1 ) , an EMCCD camera ( Hamamatsu ) , and a 2x objective using Metamorph acquisition software ( Molecular Devices ) . t=0 is defined when the imaging begins , ≈ 1 min after Taxol addition to the extract . After a brief lag time , the microtubule networks spontaneously begin contraction . Images were analyzed using ImageJ and custom build MATLAB and Python software ( available at https://github . com/peterjfoster/eLife ) . Parameters were fit to contraction data using timepoints where ϵ ( t ) > 0 . 1 . The final density was estimated using contraction experiments with 2 . 5 μM Taxol in 0 . 9 mm channels . For each experiment , the frame closest to t = τ + Tc was isolated , where τ and Tc are the timescale of contraction and the offset time respectively , obtained from fits of the time course of contraction to Equation 2 of the main text . After correcting for the camera offset and inhomogeneous laser illumination , the average fluorescence intensity of the network , ρN and the average fluorescence intensity in the channel outside the network , ρM were calculated . The fluorescence intensity in the channel but outside the network comes from monomeric fluorescently labeled tubulin and was assumed to be constant throughout the channel . The fractional concentration was then estimated as ρ ( τ+Tc ) =ρN-ρMρN+ρM . Using this measurement along with the fit curves for ϵ ( t ) and under the assumption that the network contracts in the z direction such that ϵ ( t ) in the z direction is the same as along the width , the inferred fractional concentration at t = ∞ was calculated as ρ ( t=∞ ) =ρ ( τ+Tc ) ( 1-ϵ∞ ) 2 ( 1-ϵ∞ ( 1-e-1 ) ) 2 Assuming the fluorescently labeled tubulin incorporates into microtubules at the same rate as endogenous tubulin , we can multiply the derived fractional density ρ ( t = ∞ ) by the tubulin concentration in extract , ≈18 μM ( Parsons and Salmon , 1997 ) to arrive a final network tubulin concentration of ≈30 μM . Images from contraction experiments were corrected for the camera offset and inhomogeneous laser illumination before being thresholded in order to segment the microtubule network from background fluorescence . Rotations of the channel relative to the CCD were detected by fitting linear equations to edges of the microtubule network . If the average of the slopes from the top and bottom of the network was greater than 1/ ( the number of pixels in the length of the image ) , a rotated , interpolated frame was constructed where pixels were assigned based on the intensity of the pixel in the original frame weighted by their area fraction in the interpolated pixel . Frames were averaged along the length of the channel before background signal subtraction . For density profiles compared with simulations , the edge peaks of the density profile were identified and pixels between the two peaks were retained . Profiles were normalized such that the integral of the profile was set to 1 . Particle Imaging Velocimetry was performed using PIVLab software ( Thielicke and Stamhuis , 2014 ) using the FFT window deformation algorithm with a 16-pixel interrogation area and 8 pixel step for the first pass and an 8 pixel interrogation area with a 4-pixel step for the second pass . After PIV was performed , intensity images were thresholded to segment the microtubule network from the background , and only velocity vectors within the microtubule network that were > 8 pixels from the network’s edges were retained . | The ability of cells to move , divide , and carry out other processes depends on networks of protein filaments and motor proteins collectively known as the cytoskeleton . The motor proteins can move along the filaments to transport molecules and larger structures around the cell , or to rearrange the filaments themselves . The cytoskeleton of animal , plant , and other eukaryotic cells contains two main types of filaments , known as actin filaments and microtubules . Both types of filament have distinct ends , known as the plus and minus ends . Previous studies have revealed that networks of actin filaments can rapidly contract to drive the movement of muscles and other processes . However , it is not known whether networks of microtubules can also contract . Foster et al . studied the microtubules in extracts made from the eggs of a frog called Xenopus laevis . The experiments show that these microtubules form networks that can spontaneously contract . Foster et al . propose that this contraction is caused by the minus ends of the microtubules clustering together due to the activities of a motor protein called dynein . To test this idea , Foster et al . developed a mathematical model based on an 'active fluid' theory . This model makes predictions that agree very well with the experimental data . The next step in this work is to find out if this model of microtubule contraction applies to other networks of microtubules . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology"
] | 2015 | Active contraction of microtubule networks |
Dysfunction or death of pancreatic β cells underlies both types of diabetes . This functional decline begins with β cell stress and de-differentiation . Current drugs for type 2 diabetes ( T2D ) lower blood glucose levels but they do not directly alleviate β cell stress nor prevent , let alone reverse , β cell de-differentiation . We show here that Urocortin 3 ( Ucn3 ) , a marker for mature β cells , is down-regulated in the early stages of T2D in mice and when β cells are stressed in vitro . Using an insulin expression-coupled lineage tracer , with Ucn3 as a reporter for the mature β cell state , we screen for factors that reverse β cell de-differentiation . We find that a small molecule inhibitor of TGFβ receptor I ( Alk5 ) protects cells from the loss of key β cell transcription factors and restores a mature β cell identity even after exposure to prolonged and severe diabetes .
Dysfunction or death of pancreatic β cells underlies all types of diabetes . In the case of Type 1 diabetes , it is unknown whether the initiating cause of β cell destruction is an immune attack or a β cell pathology that instigates autoimmunity . β cell failure in type 2 diabetes ( T2D ) is thought to begin as a compensatory response to peripheral insulin resistance and eventually results in the loss of a mature β cell phenotype , without necessarily leading to β cell death ( Weir and Bonner-Weir , 2004; Weir et al . , 2013 ) . The loss of a mature β cell phenotype , sometimes called de-differentiation , can result from exposure to high levels of glucose , lipids , and inflammatory cytokines ( Accili et al . , 2010 ) . De-differentiation of β cells in the context of diabetes has been shown in vivo with the genetic disruption of key transcription factors , including FoxO1 ( Talchai et al . , 2012 ) and NeuroD ( Gu et al . , 2010 ) , and is also seen in isolated islets cultured in vitro on an adherent substrate ( Gershengorn et al . , 2004; Weinberg et al . , 2007; Russ et al . , 2008; Bar-Nur et al . , 2011; Bar et al . , 2012; Negi et al . , 2012 ) . In both the FoxO1 knockout mice and obese diabetic ( LeprDb/Db ) mice , de-differentiating β cells gradually lose insulin expression and begin to express progenitor-cell markers including Ngn3 and Sox9 ( Talchai et al . , 2012 ) . Oxidative stress , also associated with T2D , inactivates the β cell specific transcription factors MafA , Nkx6 . 1 , and Pdx1 , again leading to the loss of mature β cell identity ( Guo et al . , 2013 ) . β cell de-differentiation may represent a reversal of the normal ontogeny of β cells , or follow a different pathway , but it is clear that de-differentiation depletes the pool of functionally mature β cells in T2D patients ( Weir and Bonner-Weir , 2004; Weir et al . , 2013 ) . It is not known whether there are stages of de-differentiation at which the cells can recover or re-differentiate back into fully mature β cells . The commonly used T2D drugs act by suppressing glucose production in the liver ( e . g . , Metformin ) , by enhancing peripheral insulin sensitivity ( e . g . , Rosiglitazone and other thiazolidinediones ) , or by forcing the secretion of more insulin from the already-stressed β cells ( e . g . , sulfonylureas such as Glyburide ) . There is no evidence that any of these drugs reverse β cell de-differentiation or restore the functionally mature β cell mass after β cell de-differentiation has occurred ( Kahn et al . , 2006; Accili et al . , 2010 ) . The availability of markers for early β cell stress ( Akirav et al . , 2011; Mahdi et al . , 2012; Erener et al . , 2013 ) allows one to test whether dysfunctional , stressed β cells can be revived or re-differentiated . The gene Urocortin 3 ( Ucn3 ) is a marker for functionally mature β cells , cells capable of glucose stimulated insulin secretion ( Blum et al . , 2012 ) . Ucn3 expression appears relatively late in postnatal mouse development and its expression levels correlates with functional β cell maturation in mice , and with the maturation of human pluripotent stem cell-derived β cells after transplantation ( Blum et al . , 2012; van der Meulen et al . , 2012; Hua et al . , 2013; van der Meulen and Huising , 2014 ) . We hypothesized that Ucn3 expression may be lost or reduced early during β cell de-differentiation in T2D and if so , could be used to investigate the first steps of stress-induced β cell de-differentiation .
Ucn3 and insulin expression in β cells of T2D mice were examined by immunostaining on pancreata of obese diabetic ( LepOb/Ob and LeprDb/Db ) mice and from insulin-dependent diabetic mice ( Ins2Akita ) , and compared to pancreata of age matched non-diabetic ( C57BL/6 ) mice . The intensity of insulin staining in diabetic mice is indistinguishable from non-diabetic controls , but the immunoreactivity of Ucn3 is almost completely abolished in islets of diabetic mice ( Figure 1A ) . Quantitative real-time PCR ( qRT-PCR ) showed that the expression of Ucn3 mRNA levels is significantly ( p > 0 . 001 ) reduced in islets of mice from all three diabetic models ( Figure 1B ) . Statistically significant reduction in Ins1 levels was only seen in the Ins2Akita mice , which also showed the highest fed blood glucose levels ( Figure 1B ) . The disappearance of Ucn3 from β cells that still express high levels of insulin suggests that the loss Ucn3 is an early marker of β cell stress in diabetes , occurring before the reduction in insulin expression ( Talchai et al . , 2012; Guo et al . , 2013 ) . 10 . 7554/eLife . 02809 . 003Figure 1 . Loss of Ucn3 expression is an early marker for β cell de-differentiation in diabetes . ( A ) Immunostaining with antibodies against insulin ( red ) and Ucn3 ( green ) in pancreata from T2D ( LepOb/Ob and LeprDb/Db ) , insulin-dependent diabetic ( Ins2Akita ) , and healthy control ( C57BL/6 ) mice . Ucn3 protein but not insulin protein is down regulated in diabetic pancreata compared to the healthy control . ( B ) Quantitative Real-Time PCR analysis of Ins1 and Ucn3 gene expression in islets from C57BL/6 ( n = 10 ) , LepOb/Ob ( n = 9 ) , LeprDb/Db ( n = 8 ) , and Ins2Akita ( n = 11 ) mice . Ucn3 mRNA is significantly reduced in all diabetes models , while insulin mRNA is significantly reduced only in the most diabetic model ( Ins2Akita ) . ( C ) Quantitative Real-Time PCR analysis of Ins1 and Ucn3 gene expression in islets from non-diabetic control mice ( n = 10; average blood glucose 167 ± 5 mg/dl ) , mildly diabetic ( n = 16; average blood glucose 381 ± 17 mg/dl ) and severely diabetic mice ( n = 11; average blood glucose 588 ± 8 mg/dl ) . Error bars represent ±SEM . ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 003 Insulin expression has been previously reported to be diminished in β cells of severely diabetic mice , those with blood glucose levels exceeding 500 mg/dl ( Guo et al . , 2013 ) . To confirm that loss of Ucn3 is an early marker of diabetes , we divided the diabetic mice from all three models ( LepOb/Ob , LeprDb/Db , and Ins2Akita ) into groups according to the severity of their diabetes , regardless of the genetic cause . Thus , the expression levels of Ins1 and Ucn3 mRNAs in the mildly diabetic ( blood glucose levels between 200–500 mg/dl ) and the severely diabetic ( blood glucose levels >500 mg/dl ) groups was compared to that of age-matched non-diabetic controls ( C57BL/6 , blood glucose levels <200 mg/dl ) . The average ( non-fasting ) blood glucose level was 381 ± 18 mg/dl in mildly diabetic mice , 588 ± 8 mg/dl in the severely diabetic mice , and 167 ± 5 mg/dl in the non-diabetic control mice . The expression level of Ins1 mRNA was slightly , but not significantly , higher in islets of mildly diabetic mice as compared to non-diabetic controls , but was reduced to 28% of control levels in islets of the severely diabetic group ( p < 0 . 001 ) . In contrast to the late reduction in insulin expression , the levels of Ucn3 mRNA in the mildly diabetic group were already reduced threefold , to 34% of the level in the healthy control group ( p < 0 . 001 ) , and by 10-fold , to approximately 10% of the control levels , in the severely diabetic group ( p < 0 . 001 ) ( Figure 1C ) . We conclude that the loss of Ucn3 mRNA is an early event in β cell de-differentiation . Because Ucn3 expression is reduced early during β cell de-differentiation , its expression could be used to test whether β cells at early or late stages of de-differentiation are able to regain a fully mature state . The hypothesis is that while late-stage de-differentiated β cells ( negative for both insulin and Ucn3 ) may not be able to re-differentiate into fully mature β cells , cells at an earlier stage ( negative for Ucn3 , but still expressing insulin ) may be able to recover from their de-differentiation if the stress inducing factor ( i . e . , the diabetes ) is removed . To test this hypothesis , we induced transient insulin resistance in healthy , lean wild-type mice with the insulin-receptor antagonist S961 ( Vikram and Jena , 2010; Yi et al . , 2013 ) . Mice treated with S961 develop acute insulin resistance and severe diabetes within 1 week , with non-fasting glucose levels of ≥500 mg/dl . Removal of S961 relieves the diabetes , and the mice restore their glucose control within 1 week . We thus induced transient hyperglycemia in wild-type mice with S961 for 1 week; control animals were similarly treated with PBS . At the end of the first week , half of the animals were sacrificed for analysis , and half were taken off S961 treatment and allowed to recover from diabetes for another week by which time their blood glucose levels returned to normal ( ≤200 mg/dl ) . Immunostaining of pancreata from all groups shows the levels of Ucn3 and insulin proteins ( Figure 2A ) . As expected , animals treated with S961 developed diabetes ( reaching blood glucose levels ≥460 mg/dl ) and show an increase in insulin staining , while Ucn3 staining was almost completely abolished . In the diabetic animals that recovered and showed normoglycemia following withdrawal of S961 for 1 week , there was a complete recovery of Ucn3 staining , with a staining intensity comparable to that of the PBS-treated controls ( Figure 2A ) . Quantitative RT-PCR analyses on islets at different time point during the development of S961-induced diabetes and its subsequent recovery showed that the levels of Ucn3 mRNA are significantly ( p > 0 . 005 ) reduced to about half of the levels in control mice as early as 4 days after S961 induction and are down to about a third by day 7 ( Figure 2B ) . A small reduction of the Ins1 mRNA was also seen , but this was not statistically significant . The expression of both Ins1 and Ucn3 increases to its normal levels ( and even slightly higher ) 3 days after pump removal , followed by a non-statistically significant decline 7 days after the withdrawal of the S961 pumps , which is not observed at the protein level ( Figure 2B , A , respectively ) . A similar trend was seen with the expression of MafA , Nkx6 . 1 , and Pdx1 , confirming the loss and re-gain of the mature β cell state in this model ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 02809 . 004Figure 2 . Insulin resistance-induced β cell de-differentiation is reversible . ( A ) Immunostaining with antibodies against insulin ( red ) and Ucn3 ( green ) in pancreata from wild-type C57BL/6 mice treated with either vehicle ( PBS ) or S961 ( insulin receptor antagonist ) for 7 days ( upper and middle panels ) or treated with S961 for 7 days followed by a 7-day-recovery period in the absence of S961 ( lower panel ) . Ucn3 protein expression is down regulated in β cells following 7 days S961 treatment but returns to normal expression levels upon remission to normoglycaemia ( see text ) . Nuclei are stained with DAPI ( blue ) . ( B ) Quantitative Real-Time PCR analysis of Ins1 and Ucn3 gene expression in islets from ICR lean mice taken at different time points during S961-induced de-differentiation and post S961 withdrawal recovery ( n = 3 mice for each stage ) . S961 osmotic pumps are transplanted on day 0 and removed on day 7 . Control designates mice not treated with S961 . Error bars represent ±SEM . *p < 0 . 05; ***p < 0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 00410 . 7554/eLife . 02809 . 005Figure 2—figure supplement 1 . Expression of β cell genes during S961-induced de-differentiation and subsequent recovery . Quantitative Real-Time PCR analysis of MafA , Nkx6 . 1 , and Pdx1 gene expression in islets from ICR lean mice taken at different time points during S961-induced de-differentiation and post S961 withdrawal recovery ( n = 3 mice for each stage ) . S961 osmotic pumps are transplanted on day 0 and removed on day 7 . Control designates mice not treated with S961 . Error bars represent ±SEM . *p < 0 . 05; ***p < 0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 005 We next tested whether more severely de-differentiated β cells can also return to a mature state after removal of the de-differentiation inducing stress . It has previously been reported that substantial β cell de-differentiation occurs when islets are cultured in vitro on an adherent substrate ( Gershengorn et al . , 2004; Weinberg et al . , 2007; Russ et al . , 2008; Negi et al . , 2012 ) . Cells de-differentiated using this method can be analyzed for the loss of their functional character and can be transplanted back into non-diabetic mice to test their differentiation state after being returned to a healthy environment ( Bar-Nur et al . , 2011; Bar et al . , 2012 ) . In order to follow de-differentiated β cells , even after they cease to express insulin , we developed a lineage tracing system that marks cells that have expressed insulin ( transcribed the insulin gene ) in the past . Insulin2-Cre transgenic mice were crossed with mice carrying a floxed reporter of histone H2B fused to mCherry ( R26H2BCherry ) , such that cells that had expressed insulin are marked with nuclear mCherry . These mice also contained a transgene driving cytoplasmic EGFP protein under the control of the Ucn3 promoter ( Figure 3A ) . The consequence of this genetic system is that cells with nuclear mCherry have , at some time , transcribed the insulin gene , but need not be actively producing insulin protein , and the ( reversible ) expression of cytoplasmic GFP indicates whether the β cell is fully mature ( GFP positive ) or de-differentiated ( GFP negative ) . We labeled this genetic system ‘RCU’ , for R26H2BmCherry; Ins2-Cre; Ucn3-GFP ( Figure 3A ) . 10 . 7554/eLife . 02809 . 006Figure 3 . Adherent culture-induced β cell de-differentiation is reversible . ( A ) RCU reporter mice are made by crossing mice homozygous for the Insulin2-Cre transgene with mice doubly-homozygous for Rosa26-lox-stop-lox-H2BmCherry and Ucn3-GFP . Insulin expression in RCU progeny is permanently marked by red nuclear fluorescence , and Ucn3 expression is marked by green cytoplasmic fluorescence . ( B ) Pancreas sections of PBS-treated control and S961-treated diabetic RCU mice . Ucn3-GFP is reduced in diabetic mice , but not in controls , and Ucn3 expression returns after remission from diabetes . All images show live ( unstained ) reporter fluorescence . ( C ) De-differentiation and re-differentiation of RCU islets cultured in vitro . Islets from adult RCU mice were isolated and plated on 804G matrix for 1 week ( left and middle panel ) . Note islet spreading and loss of Ucn3-GFP in the de-differentiated islets ( middle panel ) . After 7 days , the de-differentiated islets were transplanted into euglycemic SCID mice for 3 weeks ( right panel ) after which time the transplants show the return of Ucn3 expression in β cells . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 00610 . 7554/eLife . 02809 . 007Figure 3—figure supplement 1 . RCU mice show nuclear insulin expression-coupled mCherry and Ucn3-derived cytoplasmic GFP . Shown are confocal images of an islet from an adult RCU mouse . Note co-localization of nuclear H2BmCherry ( red ) and cytoplasmic GFP ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 00710 . 7554/eLife . 02809 . 008Figure 3—figure supplement 2 . Ucn3 and insulin expression are down regulated in islets grown in adherent culture . Shown are quantitative Real-Time PCR analyses of Ins1 and Ucn3 from islets of wild-type C57BL/6 mice grown on 804G matrix for 1 week . Each bar represents average gene expression in three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 00810 . 7554/eLife . 02809 . 009Figure 3—figure supplement 3 . β cells lose glucose-stimulated insulin secretion upon de-differentiation in culture . Shown are static GSIS analyses of adult islets de-differentiated on 804G for 1 week . Each bar represents average insulin secretion of three biological repeats . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 009 Triple hemizygous RCU progeny are healthy and euglycemic ( data not shown ) . The frequency of cytoplasmic Ucn3-derived GFP staining in all Ins2-Cre-derived H2BCherry-labeled cells was determined by FACS to be 57 ± 16% in both male and female mice , between 1 month and 4 months of age ( data not shown ) . Confocal imaging of β cells from triple hemizygous progeny RCU mice shows red nuclear fluorescence in β cells that is easily distinguished from the cytoplasmic green fluorescence emitted by the Ucn3-GFP reporter ( Figure 3–figure supplement 1 ) . T2D-like symptoms were induced in RCU mice using the insulin antagonist S961 as described above . Ucn3-GFP levels are down-regulated in diabetic mice , treated with S961 for 6 days , but not in PBS-infused controls ( Figure 3B , left and middle panels ) . After removal of S961 , the expression level of Ucn3-GFP was up-regulated , returning to levels comparable to control animals ( Figure 3B , right panel ) , corresponding to the remission of hyperglycemia ( Figure 3B , right panel ) . These data show that loss of Ucn3 expression is not permanent and that β cells can return to a mature Ucn3-positive state after a 7-day period of hyperglycemia . When RCU islets are plated on an adherent 804G matrix ( a laminin-rich extracellular matrix produced by the 804G rat epithelial cell line [Lefebvre et al . , 1998] ) and are cultured on the adherent matrix for 7 days , the islets flatten , cells spread out , and β cells lose Ucn3-GFP expression ( Figure 3C , left and middle panels ) . The levels of both Ins1 and Ucn3 in such adherent cultures of islets from wild-type mice were reduced to 4% and 29% of the levels in freshly harvested islets , respectively ( Figure 3—figure supplement 2 ) . Consistent with the loss of the Ucn3 marker , these islets completely lose their ability for glucose-stimulated insulin secretion ( GSIS , Figure 3—figure supplement 3 ) . Most notably , the β cells re-express Ucn3-GFP 3 weeks after transplantation into the kidney capsule of euglycemic SCID mice ( Figure 3C , right panel ) . These data suggest that the β cell de-differentiation caused by culturing cells ex vivo on adherent culture is reversible . The reversion of de-differentiated β cells to a mature state after transplantation to a healthy in vivo environment prompted us to look for factors that can recapitulate this phenomenon , as these factors could be candidates for drug development aimed at reversing β cell de-differentiation in T2D . We used the RCU platform to screen an array of 114 growth factors representing most major signaling pathways ( Supplementary file 1A ) . The experimental design , outlined in Figure 4A , employs healthy islets from adult RCU mice , isolated on day 1 and plated on an adherent 804G matrix in a 384-well plate format . The islets were first cultured for 1 week to achieve adequate de-differentiation ( see Figure 3C and Figure 3—figure supplement 2 ) . Test compounds were then added on day 7 for another week . Each compound was tested in duplicate at two or three concentrations ( listed in Supplementary file 1A , B ) . Fresh un-manipulated RCU islets were used as a positive control , and DMSO- or non-treated cultures were used as a negative control . The islets were fixed on day 11 for automated imaging and subsequent analysis . Percentages of mCherry positive cells that co-express GFP were calculated for each well and used to identify conditions that significantly increased the number of GFP positive cells over negative ( DMSO- or non-treated ) controls ( Figure 4A ) . Positive hits were selected according to their statistical significance ( p value ) over the negative control . Of the 114 tested factors , three growth factors restored Ucn3-GFP expression with a high statistical significance ( p < 0 . 01 , Figure 4B ) . These factors are BMP9 , soluble TGFβ receptor 3 ( TGFβ sRIII , also known as betaglycan ) , and the GDNF-family member Artemin . 10 . 7554/eLife . 02809 . 010Figure 4 . TGFβ pathway inhibitors and Artemin signaling reverse β cell de-differentiation . ( A ) Islets from adult RCU mice are isolated and plated on 804G matrix for 1 week in a 384-well plate format during which time the β cells de-differentiate . A compound library is added on day 7 , and islets are cultured for an additional week in the presence of compounds . Each compound is tested in duplicates of two or three concentrations . Fresh un-manipulated RCU islets are used as a positive control , and DMSO- or untreated islets are used as negative controls . Islets are fixed on day 11 for automated imaging and subsequent analysis . Percentages of mCherry positive cells that co-express GFP are calculated for each well and used to identify conditions that significantly increase the number of GFP positive cells over negative ( DMSO- or non-treated ) controls . Positive hits are selected according to their statistical significance ( p value ) over the negative control . ( B ) Results of screen with 114 growth factor proteins . Factors are ordered from left to right based on the p-value of their Ucn3-GFP fluorescence over the negative ( non-treated ) control . For convenience , values on the Y axis are presented as 1/p-value . Red bar represents the threshold for statistical significance ( p < 0 . 01 ) . ( C ) Results of screen with 19 TGFβ pathway inhibitors , 18 RET/GFRα3 inhibitors , and 42 known T2D drugs . Factors are ordered from left to right based on the statistical p-value of their Ucn3-GFP fluorescence over the negative ( DMSO-treated ) control as above . For convenience , values on the Y axis are presented as 1/p-value . Red bar represents the threshold for statistical significance ( p < 0 . 01 ) . A full list of the factors tested is presented in the Supplementary file 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 010 Both BMP9 and TGFβ sRIII signal through receptors of the TGFβ receptor family ( Massague and Chen , 2000; David et al . , 2007 ) , whereas Artemin signals through RET and GFRα3 receptors ( Airaksinen and Saarma , 2002 ) . To delve deeper into the effects of BMP9 , TGFβ sRIII , and Artemin on β cell re-differentiation , a second screen was performed using a library of 19 small molecule kinase inhibitors of TGFβ signaling and 18 small molecules effectors of RET/GFRα3 signaling ( Figure 4C and Supplementary file 1B ) . In addition , we included a library of 42 known drugs for T2D ( Figure 4C and Supplementary file 1B ) . Among the 19 small molecules tested in the TGFβ receptor inhibitors group , Alk5 inhibitor I , Alk5 inhibitor II , and a SMAD3 inhibitor , restored Ucn3-GFP expression in de-differentiated β cells ( p > 0 . 01; Figure 5C ) . Of the 18 RET/GFRα3 inhibitors two molecules with relatively low specificity to the RET kinase , namely PHA-739358 and VEGFR inhibitor V , induced Un3-GFP in the cells with p-values below 0 . 01 , and of the 42 known T2D drugs , only the two potassium-channels blockers , Repaglinide and Tolbutamide , gave marginal results . 10 . 7554/eLife . 02809 . 011Figure 5 . Alk5 inhibitor II restores β cell maturation in 804G-induced β cell de-differentiation . ( A ) Islets from adult RCU mice were isolated and plated on 804G matrix for 14 days with or without the addition of Alk5i at day 7 ( right and middle panels , respectively ) . Live fluorescence images of H2BmCherry and Ucn3-GFP were taken on day 14 , and compared to fresh RCU islets cultured on 804G for 24 hr . ( B ) H2BmCherry-positive cells from the above cultured were sorted by FACS and subjected to qRT-PCR analysis for the expression of various mature β cell genes . Statistical significance relates to the difference between Alk5i-treated and non-treated islets for each gene . Expression levels are normalized to the levels of freshly isolated islets ( dashed line ) . Error bars represent ±SEM of three biological repeats . ***p < 0 . 001 . B . G . ( C ) Immunostaining with antibodies against insulin ( red ) and Ucn3 ( green ) in islets from ICR mice treated as above . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 01110 . 7554/eLife . 02809 . 012Figure 5—figure supplement 1 . Alk5 inhibitor II induces Ucn3-GFP in RCU islets in a dose-dependent manner . Shown is a dose-curve for the induction of Ucn3-GFP in RCU islets de-differentiated on 804G matrix for 1 week , following by 1 week treatment with the indicated concentration of ALk5 inhibitor II . Bars represent ±S . D . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 012 Alk5 inhibitor II showed the strongest effect among all molecules tested , both by its reproducibility ( as measured by its statistical p value over DMSO-treated controls ) and on the levels of Ucn3-GFP expression , restoring Ucn3-GFP fluorescence of de-differentiated RCU islets to levels comparable to that of fresh islets ( Figure 5A ) . A dose–response test showed that its effect on Ucn3-GFP expression in de-differentiated RCU β cells begins at nanomolar concentrations ( Figure 5—figure supplement 1 ) . To confirm the reviving effect of Alk5 inhibitor II , we performed qRT-PCR analyses on FACS-sorted mCherry-positive β cell from islets de-differentiated on 804G matrix for 1 week , followed by another week of culture on 804G matrix supplemented with Alk5 inhibitor II , and compared those cultures grown on 804G matrix without added Alk5 inhibitor II and those of freshly-isolated islets ( Figure 5B ) . The expression levels of Ucn3 , MafA , Nkx6 . 1 , Pdx1 , and Ins1 were severely reduced in cultures de-differentiated for 2 weeks on 804G matrix . Strikingly , addition of Alk5 inhibitor II for 1 week after the initial first week of de-differentiation significantly ( p < 0 . 001 ) induced the expression levels of Ucn3 , MafA , Nkx6 . 1 and Pdx1 , and in the case on MafA and Pdx1 , to levels greater than those of freshly isolated islets ( Figure 5B ) . The recovery of the expression of Ins1 mRNA is also statistically significant , but its expression levels were still lower than those of fresh islets , perhaps because the experiment was done in low glucose medium . Immunostaining on islets from WT mice cultured as above confirmed the recovery of insulin and Ucn3 proteins in Alk5 inhibitor II re-differentiated β cells ( Figure 5C ) . Nevertheless , the addition of Alk5 inhibitor II to 804G matrix-induced de-differentiated β cells was not sufficient to restore GSIS in vitro to a statistically significant level ( data not shown ) , indicating that there is more to functionally mature GSIS than mature β cell gene expression . It has recently been shown that under diabetes-related stress , the expression and activity of key β cell transcription factors , including MafA , Nkx6 . 1 , and Pdx1 , are compromised ( Guo et al . , 2013 ) . We thus tested whether Alk5 inhibitor II is capable of preventing the down regulation in expression of these transcription factors . Islets harvested from lean , non-diabetic mice , were exposed to a diabetes-related cytokine challenge for 24 hr , with or without the presence of Alk5 inhibitor II , and the expression of several β cell genes was measured by qRT-PCR and compared to islets not treated with cytokines ( Figure 6 ) . 10 . 7554/eLife . 02809 . 013Figure 6 . Alk5 inhibitor II induces expression of mature β cell transcription factors and prevents their reduction under cytokine stress . Quantitative Real-Time PCR analysis of gene expression in wild-type islets treated with cytokines as shown ( A ) IL-β , ( B ) TNFα , ( C ) INFγ . Each bar represents average gene expression in three independent experiments . Expression levels are normalized to the levels of control islets not treated with any cytokine ( dashed line ) . Statistical significance relates to the difference between Alk5i-treated and DMSO-treated islets for each gene . Error bars represent ±SEM . *p < 0 . 05; ***p < 0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 01310 . 7554/eLife . 02809 . 014Figure 6—figure supplement 1 . β cells lose glucose-stimulated insulin secretion upon cytokine treatment . Shown are static GSIS analyses of adult islets treated with a combination of IL-1β , TNFα , and INFγ ( 10 ng/ml each ) . Each bar represents insulin secretion in three biological repeats . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 014 Islets exposed to 10 ng/ml of either IL-1β , TNFα or IFNγ showed abrogated GSIS response ( Figure 6—figure supplement 1 ) and reduced expression of Ucn3 , MafA , Nkx6 . 1 , and Pdx1 mRNAs , whereas expression of Ins1 and FoxO1 was less affected ( Figure 6 ) . Addition of 1 µM Alk5 inhibitor II with any of the cytokines prevented the diminution of expression levels for Ucn3 , MafA , Nkx6 . 1 , and Pdx1 . In fact , the expression levels of the latter three genes remained at levels comparable to , and in some cases higher than , that found in control islets ( those not exposed to cytokines ) ( Figure 6 ) . However , as seen with 804G de-differentiation , the addition of Alk5 inhibitor II to cytokine-treated islets was not sufficient to restore fully functional GSIS ( data not shown ) . We asked whether Alk5 inhibitor II can restore the expression levels of specific β cell genes from severely diabetic mice , β cells that were exposed to an extreme diabetic environment for several months . To answer this question , gene expression analyses were performed on islets from lean non-diabetic C57BL/6 mice and from mice with advanced to severe diabetes ( LeprDb/Db , LepOb/Ob , and Ins2Akita; blood glucose levels of 406 ± 39 mg/dl , 527 ± 48 mg/dl , and >600 mg/dl , respectively ) . The islets isolated from these diabetic animals , and controls , were cultured in vitro for 24 hr with or without Alk5 inhibitor II ( Figure 7 ) . Culturing control healthy islets for 24 hr with Alk5 inhibitor II results in a 1 . 5- to2 . 5-fold higher expression of Ucn3 , MafA , Nkx6 . 1 , Pdx1 , and FoxO1 compared to DMSO-treated controls ( Figure 7A ) . We also observe an unexplained twofold decrease in Ins1 expression ( Figure 7 ) . Similarly , islets from mice with advanced diabetes ( Figure 7B ) and islets from mice with severe diabetes ( Figure 7C , D ) responded to the Alk5 inhibitor II . The increase in Ucn3 , Nkx6 . 1 , and Pdx1 gene expression caused by Alk5 inhibitor II in severely diabetic mice was 1 . 5- to 2 . 5-fold , similar to the effect on non-diabetic islets . The induction of MafA expression by Alk5 inhibitor II in the severely diabetic islets increased to fivefold and sixfold over DMSO-treated controls ( Figure 7C , D ) . This may reflect the early role of MafA disappearance in β cell stress ( Guo et al . , 2013 ) . We conclude that Alk5 inhibitor II can induce mature gene expression in β cells that were exposed to extreme diabetic conditions for several months . 10 . 7554/eLife . 02809 . 015Figure 7 . Alk5 inhibitor II induces expression of mature β cell transcription factors even in β cells that were exposed to extreme diabetic conditions for several months . ( A–D ) Alk5 inhibitor II ( Alk5i ) induces expression of specific β cell genes in islets from healthy and severely diabetic mice . Shown are quantitative Real-Time PCR analysis of gene expression in islets of healthy control ( C57BL/6 ) and diabetic mice ( LeprDb/Db , LepOb/Ob , and Ins2Akita ) . Each bar represents average gene expression in three independent experiments for each group . Statistical significance relates to the difference between Alk5i-treated and DMSO-treated islets for each gene . Expression levels are normalized to the levels of C57BL/6 islets treated with DMSO ( dashed line ) . Error bars represent ±SEM . *p < 0 . 05; ***p < 0 . 005 . B . G . = Blood glucose level at time of sacrifice . ( E ) Alk5 inhibitor II ( Alk5i ) induces expression of specific β cell transcription factors in human islets . Shown are quantitative Real-Time PCR analyses of gene expression . Error bars represent three technical repeats on islets from a single donor . Error bars represent ±SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02809 . 015 We also tested whether the Alk5 inhibitor II is effective in restoring specific β cell gene expression in human islets . Primary human islets were treated with Alk5 inhibitor II for 24 hr and subjected to gene transcript analyses ( Figure 7E ) . Similar to the results with mouse islets , human islets treated with Alk5 inhibitor II show an increase in mRNA expression for Insulin , MafA , Nkx6 . 1 , and Pdx1 mRNAs , but not for FoxO1 ( Figure 7E ) .
The response of β cells to the progression of T2D begins with an adaptive stage , in which the cells compensate for insulin resistance by over-production and over-secretion of insulin , as well as increasing β cell replication ( Weir and Bonner-Weir , 2004; Guo et al . , 2013; Yi et al . , 2013 ) . This adaptation is reversible , as can be seen when β cell function returns with the remission from T2D after bariatric surgery ( Bradley et al . , 2012 ) . However , if the metabolic stress persists , β cells succumb to the metabolic overload and de-differentiation occurs . De-differentiation begins with translocation of the transcription factor FoxO1 to the nucleus , and continues with an inactivation of β cell-specific transcription factors including MafA , Nkx6 . 1 , and Pdx1 and consequently , a reduction in insulin production and secretion . All together , these changes result in the escalation of the disease and eventually to a non-recoverable loss of a functionally mature β cell mass ( Weir and Bonner-Weir , 2004; Talchai et al . , 2012; Guo et al . , 2013 ) . Our results put the loss of Ucn3 expression as an early event in β cell stress , occurring at the compensatory stage , before reduction in insulin expression and deterioration to frank diabetes . Ucn3 protein and mRNA were dramatically down regulated even in mildly diabetic mice , some of which had blood glucose levels that were just slightly above normal . This is further demonstrated by the loss of Ucn3 in S961-treated mice , exposed to insulin resistance and hyperglycemia for only a week . In severely diabetic mice , we observed a dramatic reduction in insulin expression indicative of advanced β cell de-differentiation , and the expression of Ucn3 mRNA was almost completely abolished . Ucn3 is a small neuropeptide hormone , expressed mainly in the Islets of Langerhans ( where , in the mouse , it is restricted to β cells ) , the small intestine , the skin , and specific brain regions such as the hypothalamus , amygdala , and brainstem ( Lewis et al . , 2001; Li et al . , 2003; Benner et al . , 2014; van der Meulen and Huising , 2014 ) . Ucn3 has been suggested to regulate GSIS in response to high blood glucose ( Li et al . , 2007 ) and was linked to peripheral glucose homeostasis and food intake behavior ( Kuperman and Chen , 2008; Kuperman et al . , 2010; Jamieson et al . , 2011 ) . The precise role of Ucn3 in the pancreatic islets is not yet clear , but the expression of its receptors in mouse and human islets suggests an islet-autonomous autocrine and/or paracrine action ( Huising et al . , 2011 ) . It is noteworthy , however , that loss of expression of Ucn3 per se is not a driver of diabetes , but is rather caused by it , as mice homozygous for a Ucn3-null allele are not diabetic , and even show slightly better glucose tolerance under high-fat feeding and aging ( Li et al . , 2007 ) . To utilize the finding that Ucn3 is an early marker of β cell de-differentiation , we developed triple-transgenic mice , in which a sensitive Ucn3-regulated GFP reporter is combined with β cell lineage tracing . This genetic system allows one to trace β cells even after profound de-differentiation . Using this system , we show that β cell de-differentiation can be reversed after 1 week of S961 treatment in vivo or after 1 week of adherent culture in vitro . A screen for pathways that can rescue β cells from de-differentiation identified three growth factors that restored Ucn3-GFP expression , namely BMP9 , TGFβ sRIII , and Artemin , all belonging to the TGFβ superfamily . Of those , only BMP9 had previously been identified as having an active role in glucose homeostasis ( Chen et al . , 2003 ) . The gene encoding TGFβ receptor III has been shown to be up regulated in pancreata from obese human patients compared to lean subjects ( Muharram et al . , 2005 ) , while Artemin and its receptor GFRα3 have , to the best of our knowledge , not been described in pancreatic islet function . Tests on small molecule mediators of BMP/TGFβ and Artemin signaling identified Alk5 inhibitor II as a potent compound able to restore mature β cell identity even in islets from severely diabetic mice . This inhibitor also blocked the loss of specific β cell gene expression under cytokine-induced stress . Alk5 inhibitor II , identified using mouse β cells , can induce the expression of key β cell transcription factors in human islets . While human UCN3 is a marker of the functional maturation for both β and α cells ( van der Meulen et al . , 2012; Benner et al . , 2014 ) , and despite evidence that UCN3 is not a faithful marker for functional β cell maturation in human islets during human pancreas development ( Hrvatin et al . , 2014 ) , the signals that reverse β cell de-differentiation ( i . e . , inhibition of Alk5 signaling ) may be conserved between mouse and human . Alk5 inhibitor II has been previously identified by Rezania et al . in an independent screen aimed at inducing functionally mature endocrine cells from human embryonic stem cells ( Rezania et al . , 2011 ) . Ichida et al . showed that this inhibitor can replace Sox2 in cell reprogramming ( Ichida et al . , 2009 ) . Interestingly , it was recently reported that β cells of mice carrying a conditional deletion of both Alk5 ( referred to as TGFβ receptor I ) and TGFβ receptor II do not proliferate in response to inflammatory cytokines ( Xiao et al . , 2013 ) . In our results , Alk5 inhibitor II restored specific β cell gene expression in de-differentiated β cells , blocked cytokine-induced β cell stress , and stimulated over-expression of these genes in β cells from healthy , non-diabetic mice , and humans . It is noteworthy that the up regulation of SMAD7 , a downstream mediator of TGFβ signaling , promotes β cell proliferation ( Xiao et al . , 2014 ) . Inhibition of Alk5 would inhibit SMAD7-induced proliferation . This may hint on the intriguing idea that β cell proliferation , at least under inflammatory stress , may require a phase of de-differentiation . Taken together , these results suggest that Alk5 signaling may be constitutively active in β cells , that sustaining mature β cell phenotype depends on constant inhibition of this signal , and that the inhibition of Alk5 signaling may confer its effect by inducing expression of β cell transcription factors including MafA , Nkx6 . 1 , and Pdx1 . This postulated inhibition of Alk5 signaling in mature β cells develops during the first postnatal weeks , when the cells reach their fully mature state ( Blum et al . , 2012; Szabat et al . , 2010; Szabat et al . , 2011 ) , and is reduced under diabetic stress or when the cells are taken out of their proper niche and grown in vitro . If inhibition of Alk5 signaling is not restored , the β cells will evidently de-differentiate and disappear . Alk5 is a broadly expressed protein and its activity is required in many other tissues besides β cells . Indeed , our preliminary attempts to inject high dosage of Alk5 inhibitor II to diabetic mice resulted in overall poor health without significant reduction of blood glucose . We therefore suggest that screening for compounds that inhibit Alk5 signaling specifically in β cells may yield compounds that in combination with traditional blood-glucose lowering medicines will delay , prevent , or perhaps restore the loss of healthy , mature β cell function in T2D patients .
Animal experiments were performed in compliance with the Harvard University International Animal Care and Use Committee ( IACUC ) guidelines . Mouse strains used were C57BL/6 , LepOb/Ob ( Zhang et al . , 1994 ) , LeprDb/Db ( Chen et al . , 1996 ) , Ins2Akita ( Wang et al . , 1999 ) , Insulin2-Cre transgenic mice ( Postic et al . , 1999 ) , Ucn3-GFP transgenic mice ( Gong et al . , 2003 ) , SCID-beige mice , R26H2BCherry mice , and RCU mice . R26H2BCherry mice ( carrying a floxed nuclear-labeling reporter composed of histone H2B fused mCherry ) were generated by genetic targeting of the Rosa26 locus of V6 . 5 mouse ES cells with the construct Rosa26-Puro-p ( A ) -CAGS-lox-PGK:neo-p ( A ) -lox-H2BCherry-p ( A ) . Targeted ES cells were injected into BDF1xB6 blastocysts , and germline transmission was detected through breeding of chimeras with C57BL/6 females . To generate RCU mice , mice homozygous for both R26H2BCherry and Ucn3-GFP were crossed with homozygous Insulin2-Cre mice . All RCU progeny are triple hemizygous at all three alleles . Induction of transient insulin resistance by S961 was done with an osmotic pump as previously described ( Yi et al . , 2013 ) . Blood glucose levels were measured in non-fasted animals using OneTouch Ultra2 glucometer ( LifeScan , Milpitas , CA ) . For islet isolation , adult pancreata were perfused through the common bile duct with 0 . 8 mM Collagenase P ( Roche ) , and fetal and neonatal pancreata were dissected wholly without perfusion . Pancreata were digested with 0 . 8 mM Collagenase P ( Roche ) and purified by centrifugation in Histopaque gradient ( Sigma ) . Pancreata were fixed by immersion in 4% paraformaldehyde overnight at 4°C . Samples were washed with PBS , incubated in 30% sucrose solution overnight , and embedded with optimal cutting temperature compound ( Tissue-Tek ) . 10-µm sections were blocked with 10% donkey serum ( Jackson Immunoresearch ) in PBS/0 . 1% Triton X and incubated with primary antibodies overnight at 4°C . Secondary antibodies were incubated for 1 hr at room temperature . Antibodies and dilutions used include rabbit anti-mouse Ucn3 ( 1:600-1:800 , Phoenix Pharmaceuticals ) , Guinea Pig anti-insulin ( 1:800 , DAKO ) , Alexa Fluor 488 donkey anti-rabbit ( 1:400 , Invitrogen ) , and DyLight 649 donkey anti-guinea pig ( 1:400 , Jackson Immunoresearch ) . Nuclei were visualized with DAPI . Images were taken using an Olympus IX51 Microscope or Zeiss LSC 700 confocal microscope . Islets from adult RCU mice were isolated and plated on 804G matrix ( Lefebvre et al . , 1998 ) for 1 week in a 384-well plate format . Compound libraries were added on day 7 , and islets were cultured for an additional week in the presence of compounds . Each compound was tested in duplicates of two or three concentrations . A list of all compounds and concentrations appears in Supplementary files 1 . Fresh un-manipulated RCU islets were used as a positive control , and DMSO- or untreated islets were used as a negative control . The islets were fixed on day 11 for automated image acquisition and analysis using a Cellomics ArrayScanVTI . Cell nuclei of target cells were identified by nuclear mCherry expression , and a 2 pixel cytoplasmic mask was drawn around each nucleus . The GFP fluorescence in the cytoplasmic mask of freshly isolated islets was used as a control to identify fluorescence intensity thresholds that enabled automated calls on each individual cell . Cells that displayed GFP fluorescence equal or greater than found in control cells were identified as being positive for the Ucn3-GFP reporter . Percentages of mCherry positive cells that co-express GFP were calculated for each well and used to identify conditions that significantly increased the number of GFP positive cells over negative controls . Positive hits are selected according to their statistical significance ( p value by t test ) over the negative control . Total RNA from fresh or cytokine-treated whole islets was isolated using RNeasy Plus Mini Kit ( Qiagen ) . cDNA was prepared with random primers using SuperScript III reverse transcriptase ( Life Technologies ) . For cytokine treatment , isolated islets were recovered overnight in islet media ( DMEM containing 1gr/l glucose , 10% vol/vol FBS , 0 . 1% vol/vol Penicillin/Streptomycin ) , followed by 24-hr incubation with 10 ng/ml of either mouse IL-1β , mouse TNFα or mouse INFγ ( R&D Systems ) , with the addition of Alk5 inhibitor II ( 1 µM , Axxora ) or vehicle ( DMSO ) at the same dilution . Relative expression of Ucn3 , Ins1 , Nkx6 . 1 , Pdx1 , and FoxO1 was determined using gene-specific TaqMan probes with TaqMan Fast Universal PCR Master Mix ( Life Technologies ) on an ABI 7900 Real-Time PCR machine . Relative expression of mouse MafA was determined using Brilliant III Ultra-Fast SYBR Green QPCR Master Mix ( Agilent ) on the same machine . Primers for mouse MafA were 5′-AGCGGCACATTCTGGAGAG-3′ forward and 5′-TTGTACAGGTCCCGCTCCTT-3′ reverse . Levels of gene expression were normalized to the expression of Ubc or Eif2A genes . Institutional review board approval for research use of human tissue was obtained from the Harvard University Faculty of Arts and Sciences . Human islets were obtained from NDRI ( The National Disease Research Interchange ) . Donor anonymity was preserved , and the human tissue was collected under applicable regulations and guidelines regarding consent , protection of human subjects and donor confidentiality . Human islets were grown in CMRL 1066 Supplemental medium ( Mediatech ) , 10% vol/vol HyClone FBS ( Thermo Scientific ) , 1% vol/vol Penicillin/Streptomycin ( Corning Cellgro ) for 4 days before treatment for 24 hr with Alk5 inhibitor II ( Axxora ) . | Diabetes is a condition that develops when the body does not produce or use a hormone called insulin effectively . Insulin helps fat and muscle cells absorb glucose from the blood , and so diabetes can result in high levels of blood glucose , which can cause strokes , blindness , and heart disease . In healthy individuals , beta cells in the pancreas ( a large gland located behind the stomach ) produce insulin . The beta cells develop from endocrine progenitor cells , which are unspecialised cells that can either duplicate themselves or ‘differentiate’ to form one of the specialised cell types found in the pancreas . In diabetic patients , however , certain stresses ( such as an immune attack in type-1 diabetics or insulin-resistance due to obesity , pregnancy , or ageing in type-2 diabetics ) can cause mature beta cells to lose their identity in a process known as ‘de-differentiation’ . This means that beta cells either revert back to an earlier stage in their development or adopt a new dysfunctional identity . When this occurs , the body loses beta cells and is unable to produce insulin . It was not known whether de-differentiated beta cells in diabetic patients can recover to form mature beta cells that are capable of producing insulin . Additionally , the drugs currently used to treat diabetes are able to lower blood glucose levels , but these drugs do not replace the lost beta cells . Blum et al . now show that mice stop expressing a gene called Urocortin 3 when beta cells first start to de-differentiate . Only functional beta cells express Urocortin 3 , so this gene is a useful ‘marker’ that can be used to tell if a cell is a mature , functional beta cell or not . Using this system , Blum et al . found that if de-differentiated cells are transplanted into a non-diabetic mouse , they are able to revert back into mature beta cells that can produce insulin . This happens even if the cells have been de-differentiated for a long time . Blum et al . then used this system to investigate ways of protecting against or reversing beta cell de-differentiation . Using small molecules to block the activity of a protein called TGF beta receptor 1 was found to protect against beta cell de-differentiation and to restore the identity of mature beta cells . The findings of Blum et al . represent a first step towards the possible development of new drugs to prevent or even restore the loss of healthy , mature beta cells in diabetic patients . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2014 | Reversal of β cell de-differentiation by a small molecule inhibitor of the TGFβ pathway |
Many plant and animal viruses are spread by insect vectors . Cauliflower mosaic virus ( CaMV ) is aphid-transmitted , with the virus being taken up from specialized transmission bodies ( TB ) formed within infected plant cells . However , the precise events during TB-mediated virus acquisition by aphids are unknown . Here , we show that TBs react instantly to the presence of the vector by ultra-rapid and reversible redistribution of their key components onto microtubules throughout the cell . Enhancing or inhibiting this TB reaction pharmacologically or by using a mutant virus enhanced or inhibited transmission , respectively , confirming its requirement for efficient virus-acquisition . Our results suggest that CaMV can perceive aphid vectors , either directly or indirectly by sharing the host perception . This novel concept in virology , where viruses respond directly or via the host to the outside world , opens new research horizons , that is , investigating the impact of ‘perceptive behaviors’ on other steps of the infection cycle .
Transmission is a pivotal step in the infection cycle of viruses: it controls the passage from one host to another and is thus essential for dissemination . This step can represent a significant bottleneck for the infection cycle , since it is common for only a small proportion of the countless viral genomes produced to be passed on to a new host in a transmission event; for example , only one to three of the many transmissible genomes initiate a new infection after Potato virus Y transmission ( Moury et al . , 2007 ) . It is thus expected that viruses have adapted their life cycle and developed sophisticated strategies to optimize their transmission . Whereas non-viral pathogens are known to allocate resources for the production of transmission-specific morphs ( discussed in Matthews , 2011 ) , surprisingly little is known for this mechanism regarding viruses ( for review see Blanc et al . , 2011 ) . Some viruses are transmitted vertically to host offspring and others are transmitted by contact between hosts ( e . g . , by wind , water or physical contact ) , but most viruses rely on vectors for rapid proliferation within host populations ( for review see Kuno and Chang , 2005; Blanc et al . , 2011; Bak et al . , 2012 ) . The most important vectors are found among the arthropods . Those with a piercing-sucking feeding behavior such as mosquitoes ( or other blood-feeding dipterans ) and ticks are especially significant for vertebrate viruses , and likewise aphids , white flies and other sap-feeding bugs are consequential for plant viruses . These vectors are ideal , as their variety of mouth parts can puncture cells , blood vessels , and plant sap vessels with great precision , thus enabling efficient uptake and injection of pathogens without killing the host . Vector transmission can be classified into two main transmission modes . In circulative transmission , the virus is taken up by the vector together with the nutrients ( e . g . , blood , plant sap , cell contents ) , where it actively crosses from the intestine into the vector interior . Then it cycles through the hemocoel ( the internal body cavity awash in hemolymph ) to the salivary glands , where the virus can be secreted together with the saliva into a new host . The second transmission mode is alternatively referred to as mechanical transmission ( for human and animal viruses ) or non-circulative transmission ( for plant viruses ) . In this transmission mode , the arthropod vector only briefly comes into contact with the virus , in which it transiently attaches to the vector mouth parts and is subsequently released; an internalization step does not occur . The viral proteins involved in this seemingly simple process have been well-described in the literature , often down to the molecular level ( for review see Ng and Falk , 2006 ) . On the other hand , their precise roles during virus-acquisition by the vector remains largely unexplored ( for review see Blanc et al . , 2011 ) . Cauliflower mosaic virus ( CaMV ) , the virus studied here , is a non-circulative virus transmitted by aphids . CaMV binds to a receptor protein located at the tip of the aphid's needle-like mouth parts , the stylets ( Uzest et al . , 2007 , 2010 ) . The CaMV transmissible complex is composed of the icosahedral viral particle ( containing the viral genome enclosed by a shell of capsid protein P4 ) , the virus-associated protein P3 , and finally the aphid-transmission factor or helper component , the viral protein P2 ( Blanc et al . , 1993a; Leh et al . , 1999; Plisson et al . , 2005 ) . P2 is central to the virus's transmission , as it links the virus particle to the aphid stylets through the interaction of its C-terminus with virus-associated P3 , as well as the linking of its N-terminus with the stylet receptor ( Figure 1A ) . Interestingly , although P2 deletion mutants of CaMV are not transmissible by aphids , they are perfectly infectious when inoculated artificially to host plants . This shows that the only role for P2 in the CaMV life cycle is virus–vector interaction . P2 localizes exclusively to a specific cytoplasmic inclusion in infected plant cells , the transmission body ( TB , Figure 1B ) . There , P2 co-aggregates with the viral protein P3 to form a matrix in which some virus particles are embedded; the existence of any cellular components within this matrix remains elusive ( Espinoza et al . , 1991; Drucker et al . , 2002 ) . The TB-contained P3 is most likely dissimilar in conformation to the P3 associated with the virus particle , and has been suggested to play a role in TB structure and maintenance , but any details are yet unknown ( Drucker et al . , 2002; Hoh et al . , 2010 ) . TBs are indispensable to this transmission , as it has previously been shown that aphids are unable to acquire the virus in their absence ( i . e . , P2 deletion mutants , Woolston et al . , 1983 ) , as well as when TBs are malformed . The P2 mutant , P2G94R ( described in Khelifa et al . , 2007 ) , assists efficiently in the transmission of purified CaMV particles associated with P3 , when aphids are allowed to acquire all three components in vitro from suspensions across Parafilm membranes . However , when the P2G94R mutant is expressed in planta in the context of CaMV infection , it induces the formation of a misshaped TB ( for details see Khelifa et al . , 2007 ) , preventing plant-to-plant transmission by the aphid vector . 10 . 7554/eLife . 00183 . 003Figure 1 . The CaMV transmissible complex and the transmission body . ( A ) Left: the CaMV transmissible complex comprises the virus particle , composed of capsid protein P4 ( yellow ) , virus-associated protein P3 ( blue ) and the helper component P2 ( red ) . P2 binds via its C-terminus to P3 and via its N-terminus to a protein receptor localized in the stylet tips of the aphid vector ( middle and right ) . ( B ) Infected cells contain many cytoplasmic virus factories ( VF ) , where most virus particles ( blue-yellow circles ) accumulate in a matrix composed of virus protein P6 ( grey ) , and a single transmission body ( TB ) . The TB ( also cytoplasmic ) is composed of P2 ( red ) and P3 ( blue ) as well as scattered virus particles . P3 in TBs is most likely in a conformation that differs from virus-associated P3 . The spatial separation of the components of the transmissible complex ( P2 in the TB and most virus particles in VFs ) lead us to propose that they unite only at the moment of vector acquisition ( Drucker et al . , 2002 ) . Cortical microtubules are designated in green and the cell wall in dark green . Cell organelles are not shown , for clarity . The CaMV model is from Plisson et al . ( 2005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 003 To understand CaMV acquisition , some knowledge of the unique feeding behavior of aphid vectors is required . Aphids feeding on plant leaves insert their stylets into the middle lamella that separates adjacent cells , and subsequently perform a series of brief test punctures into the epidermis and parenchyma cells; this continues until eventually reaching the phloem where they can feed for long periods , provided the plant is a suitable host ( for review see Fereres and Moreno , 2009 ) . Most plant cells survive the initial test punctures , during which only minute amounts of cytoplasm are ingested . CaMV and hundreds of other viral species can be acquired during this feeding behavior , but detailed events occurring within the punctured cell at the precise moment of stylet entry and how they result in virus acquisition are largely unknown . In the example of CaMV , it is known that microtubules are involved in the generation of TBs at the onset of infection ( Martinière et al . , 2009 ) and that the microtubule depolymerizing drug oryzalin inhibits virus acquisition by aphids ( Martinière et al . , 2011a ) , but any details of the mode of action of oryzalin on TBs and how TBs function in virus transmission are still unclear . Here , we have analyzed the three-way interaction between CaMV , host plant cell , and aphid at the precise moment of the intracellular penetration of the stylets and imminent virus acquisition . Our study reveals an unforeseen capability of CaMV in that it senses—probably by using the host cell machinery—the aphid feeding , and then instantly produces a transmissible form for uptake by the insect .
At the beginning of this study was the observation that several different TB phenotypes could be discerned in infected tissues , as viewed by double-labeling experiments using antibodies against the TB marker P2 and the microtubule protein α-tubulin . Thus , typical TBs were detected , and these were rather large ( 2–5 μm in diameter ) and mostly ovoid single cytoplasmic inclusions , having a cortex heavily labeled by P2 antibody and a less intensely labeled interior . Most importantly , little if any tubulin was detected in this regular form of the TB ( Figure 2A ) . Interestingly , we also observed a second class of TBs that was phenotypically nearly identical , with the exception that tubulin had greatly accumulated in their centers ( Figure 2B ) . Even more surprisingly , we were unable to detect TBs in some cells; at best , small P2 foci without the typical TB structure were visible ( hereafter referred to as fragmented TBs ) . Instead , most P2 decorated the microtubule network in these cells ( a P2 distribution pattern hereafter designated as ‘mixed-networks’ and referring to mixed P2-tubulin networks; Figure 2C ) . A common point among these three TB phenotypes was that their occurrence varied greatly from one experiment to the next . Depending on the tissue preparation , anywhere from almost none to practically all of the TBs contained tubulin; the proportion of cells containing mixed-networks also varied from one experiment to another . To account for these observations , we hypothesized that , among the different TB morphologies observed ( hereafter referred to as ‘morphs’ ) , the tubulin-loaded ( Tub+|TB ) and mixed-network phenotypes were induced by unidentified stresses during leaf handling , whereas the tubulin-less phenotype ( Tub−|TBs ) corresponded to unstressed ‘standby’ TBs found under normal conditions . As the only known role for TBs is in transmission , this further raised the question of whether and how the presence of tubulin within TBs impacts transmission by aphids . To investigate this phenomenon , we first aimed to identify artificial stresses that could trigger specific transformation of standby TBs into Tub+|TBs and mixed-networks . Heat-shock , wounding and CO2 exposure all induced TB transformation ( Figure 2D–F and Figure 2—figure supplement 1 ) . Quantification of the various TB morphs ( Figure 3 ) established that all three treatments induced Tub+|TBs . However , mixed-networks ( co-existing with Tub+|TBs and fragmented TBs ) were significantly increased only in wounded or CO2-exposed leaves , and were rarely observed in heat-shocked or untreated leaves . This indicates that the different stresses had different and specific effects on TBs . 10 . 7554/eLife . 00183 . 004Figure 2 . Stress induces different TB morphs . ( A–C ) The three TB morphs . Immunofluorescence of infected leaves against P2 ( red ) and α-tubulin ( green ) , with co-labeling appearing as yellow/orange , reveals the different TB forms: ( A ) a tubulin-less TB ( arrow ) , ( B ) a Tub+|TB ( arrow ) and ( C ) mixed-networks . Images show confocal projections; insets show optical single sections from the TBs indicated by the arrows in ( A ) and ( B ) , and of the enclosed zone in ( C ) . The orange arrows in the insets mark the line scans and the direction used to create the profiles of P2 ( red ) and tubulin ( green ) label intensity , shown to the right of the insets . The line scans show that the TB in ( A ) contains hardly any tubulin , whereas the TB in ( B ) is heavily tubulin-labeled , revealing stronger tubulin labeling in the center of the TB than at the cortex . Finally , the distributions of P2 and tubulin labels colocalize in the mixed-networks shown in ( C ) . The intensities are indicated in arbitrary units ( AU ) since the acquisition conditions were not identical for the different samples . ( D–F ) Stress induces TB transformation . Immunofluorescence labeling ( P2 in red , tubulin in green , DAPI nucleic acid stain in blue ) of infected leaves after the indicated stress treatment shows that heat shock ( D ) induces only Tub+|TBs , whereas wounding stress ( E ) and exposure to CO2 ( F ) additionally induce TB fragmentation ( as revealed by the small red or orange foci in E and F ) and mixed-networks . The upper panels of ( D–F ) show confocal projections , and the lower panels show selected optical single sections . For heat shock ( D ) , two individual sections representing a median section through each of the two encircled TBs are shown . In ( E–F ) , the arrows indicate filamentous P2 labeling that is continuous with microtubule labeling , and the arrowheads point to small P2 aggregates in the vicinity of microtubules . Scale bars: 5 μm . The confocal single sections used to create the projections shown here can be found in Figure 2—source data 1–6 . See also Figure 2—figure supplement 1 that shows in vivo stress response of GFP-labeled tubulin in infected plants . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 00410 . 7554/eLife . 00183 . 005Figure 2—source data 1 . Confocal single sections and acquisition parameters used for Figure 2A . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 00510 . 7554/eLife . 00183 . 006Figure 2—source data 2 . Confocal single sections and acquisition parameters used for Figure 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 00610 . 7554/eLife . 00183 . 007Figure 2—source data 3 . Confocal single sections and acquisition parameters for Figure 2C . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 00710 . 7554/eLife . 00183 . 008Figure 2—source data 4 . Confocal single sections and acquisition parameters for Figure 2D . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 00810 . 7554/eLife . 00183 . 009Figure 2—source data 5 . Confocal single sections and acquisition parameters for Figure 2E . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 00910 . 7554/eLife . 00183 . 010Figure 2—source data 6 . Confocal single sections and acquisition parameters for Figure 2F . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01010 . 7554/eLife . 00183 . 011Figure 2—figure supplement 1 . Tubulin accumulation in large inclusions after different stresses is specific to TBs . Confocal projections of Arabidopsis GFP-TUA6 leaves subjected to the stress indicated and examined by in vivo confocal fluorescence microscopy . ( A ) Wild type CaMV-infected leaf epidermis , ( B ) Leaf epidermis infected with the CaMV-ΔP2 mutant that does not form TBs , and ( C ) Healthy lead epidermis . Arrows indicate tubulin-containing TBs , visualized by live GFP-tubulin fluorescence ( green ) . Chloroplast autofluorescence is shown in magenta . Note , however , that heat-shock induced some small GFP-tubulin aggregates . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01110 . 7554/eLife . 00183 . 012Figure 3 . Quantitative analysis of the different TB morphs induced under stress conditions . Leaf samples were either left untreated ( Control ) , exposed for 2 h at 37°C ( Heat shock ) , exposed for 15 min to CO2 atmosphere ( CO2 exposure ) , cut with a razor blade and then fixed within 10 s ( Wounding ) , or fixed first and then cut with a razor blade ( Wounding after fixation ) . All leaf samples were then processed in parallel for immunostaining against P2 and α-tubulin and scored for the occurrence of the different TB morphs: ‘standby’ Tub−|TBs ( turquoise ) , ‘activated’ Tub+|TBs ( yellow ) or mixed-networks ( red ) . Results are from three independent experiments and the total number of TBs and networks counted for each condition were control ( 284 ) , heat shock ( 293 ) , CO2 exposure ( 282 ) , wounding stress ( 313 ) , and 288 in tissues wounded after fixation . See Figure 3—source data 1 for details . SD: standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01210 . 7554/eLife . 00183 . 013Figure 3—source data 1 . Source data for Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 013 Having access to an experimental system to specifically induce TB transformation , we next examined what role ( s ) the different TB morphs might play in CaMV aphid-transmission . Here , we pursued three complementary objectives: i ) to characterize the dynamics of TB transformation; ii ) to investigate whether aphid feeding activity can trigger this transformation; and iii ) to test whether TB transformation is required for successful CaMV transmission . For TB changes to have any biological relevance in CaMV acquisition by aphid vectors , they must happen fast enough to be compatible with the duration of the intracellular penetration of the aphid stylets during test probes , that is , within ∼10 s . To estimate the speed of TB transformation , we first tested wounding stress by inflicting cuts with a razor blade on infected turnip leaves . The tissue was then fixed immediately ( ∼10 s ) and the TB phenotype analyzed by immunofluorescence . Tub+|TBs and mixed-networks were detected readily at the wounding sites; by contrast , the cells of control tissue , or tissue wounded after fixation , predominantly harbored standby TBs ( Figure 3 ) . We next evaluated the kinetics of tubulin entry into TBs . The surface of a CaMV-infected Arabidopsis leaf expressing genetically tagged GFP-tubulin ( GFP-TUA6; Ueda et al . , 1999 ) was touched with a microelectrode , and the response of TBs in epidermis cells was recorded using confocal time-lapse macroscopy . In the event of a fast entry of tubulin into TB , we should expect to observe a rapid appearance of fluorescent foci , corresponding to Tub+|TBs within these cells . Figure 4A and Movie 1 reveal the detection of GFP-tubulin in TBs as early as ∼5 s after microelectrode impact; fluorescence in these inclusions reached a maximum and stabilized within ∼10 s . Similar results were obtained in ∼50% of all infected cells tested ( Table 1 ) . Contrarily , GFP-tubulin formed a diffuse fluorescent cloud at the impact site in healthy control cells , in line with previous reports ( Hardham et al . , 2008 ) ; fast appearance of tubulin inclusions as in infected cells was never observed . Subsequently , we examined whether the tubulin within TBs is exchanged with that of the cytoplasm , by measuring GFP-tubulin turnover in Tub+|TBs in fluorescence recovery after photobleaching experiments ( FRAP ) . Photo-bleached GFP-tubulin in TBs was rapidly replaced by fresh cytoplasmic GFP-tubulin ( Figure 4B–D ) , suggesting that this protein circulates continuously between Tub+|TB and the cytoplasm . Taken together , these results indicate that the appearance of both Tub+|TBs and mixed-networks is fast enough to occur during an aphid puncture , and that there is a dynamic equilibrium between cytosolic and TB-contained tubulin . 10 . 7554/eLife . 00183 . 014Figure 4 . Tubulin influx into TBs occurs on a rapid time scale . ( A ) Kinetics of tubulin entry into TBs . The epidermis of CaMV-infected Arabidopsis leaves expressing GFP-tagged α-tubulin ( Arabidopsis GFP-TUA6 ) was touched with a microelectrode tip ( yellow ) , and the effect of the impact recorded by time-lapse confocal macroscopy . GFP-tubulin fluorescence is shown in green , chloroplast fluorescence in orange/red . Negative and positive time points are before and after the microelectrode-epidermis contact , respectively . The red circle denotes the impact zone , and the three arrows point to newly formed GFP-tubulin inclusions . The blue asterisk indicates a reference epidermis cell that did not change its z-position during the time lapse recording and can be used as a landmark for orientation . ( B–D ) Tubulin cycles between TBs and the cytoplasm . Arabidopsis GFP-TUA6 plants were infected with CaMV and the epidermis was screened for rare spontaneously occurring Tub+|TBs ( no deliberate stress treatment was inflicted on the leaf ) ; these were identified by the characteristic shape of the fluorescent tubulin-containing inclusions ( see Figure 2—figure supplement 1 ) . The GFP-tubulin in these Tub+|TBs was photobleached , and the recovery of the GFP fluorescence ( due to replacement by fresh cytoplasmic GFP-tubulin ) was recorded by time lapse microscopy . ( B ) Microscopic images of a typical FRAP experiment . The first picture shows a GFP-tubulin-containing Tub+|TB before photobleaching . The dashed circle in the second picture indicates the photobleached zone at t = 0 s , and the following pictures show recovery of the GFP-fluorescence at indicated time points after photobleaching . ( C–D ) The graphs show quantifications of fluorescence recovery: after photobleaching of TBs ( C ) , and after photobleaching of a cytoplasmic zone as a control of free tubulin diffusion ( D ) . The fluorescence levels were normalized ( 100% = fluorescence before bleaching , 0% = fluorescence just after bleaching ) . For the two quantification graphs , FRAP trend lines ( red ) were calculated from seven FRAP experiments on GFP-tubulin-containing TBs , or from 18 FRAP experiments on cytoplasmic zones . The difference in t ( 1/2 ) for fluorescence recovery between TBs and the cytoplasm was highly significant ( p<0 . 0001 , t-test with n = 18 for TBs and n = 21 for cytoplasm ) . In contrast , the difference in the mobile fractions , that is , the percentage of exchangeable GFP-tubulin , the so-called mobile fraction , was not significant ( p=0 . 504 , t-test with n = 18 for TBs and n = 21 for cytoplasm ) . These results indicate that tubulin cycles between the cytoplasm and TBs , albeit at much slower rates than free diffusion in the cytoplasm . See Figure 4—source data 1 and 2 for details . Scale bars: 10 μm; MF: mobile fraction; IF: immobile fraction . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01410 . 7554/eLife . 00183 . 015Figure 4—source data 1 . Source data for Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01510 . 7554/eLife . 00183 . 016Figure 4—source data 2 . Source data for Figure 4D . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01610 . 7554/eLife . 00183 . 017Movie 1 . Time lapse confocal macroscopy of a leaf epidermis touched with a microelectrode tip . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 01710 . 7554/eLife . 00183 . 018Table 1 . Speed of influx of GFP-TUA6 into TBs after touching of epidermis cells with a microelectrodeDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 018Tubulin reactionIn infected cellsIn healthy cellsNo reaction14 ( 27% ) 3 ( 20% ) Appearance of large fluorescent inclusions within 30 s24 ( 47% ) 0 ( 0% ) Appearance of large fluorescent inclusions after >1 min1 ( 2% ) 0 ( 0% ) Formation of diffuse tubulin cloud12 ( 24% ) 12 ( 80% ) Number of experiments5115Epidermal cells of infected or healthy tissues were touched with a microelectrode and the appearance of tubulin fluorescence in inclusions was observed by time-lapse microscopy . A large fluorescent inclusion detected within 30 s or less was considered to be a wound-related tubulin entry into TB . In roughly half the experiments using infected tissue , rapid formation of large fluorescent inclusions was observed; in the other experiments , either no reaction occurred or formation of diffuse fluorescent clouds prevailed . In healthy controls , most cells responded with the appearance of diffuse tubulin clouds , as previously reported ( Hardham et al . , 2008 ) ; rapid appearance of tubulin inclusions was never observed . The above experiments demonstrating the existence of distinct TB morphs , we next turned our attention to the possible transformation of one form into another , and aimed to establish a chronology of TB morphological changes . To facilitate tracking , we used infected protoplasts to screen for various conditions that could induce TB transformation ( Table 2 ) , including various physical , chemical and biological stresses . These wide-ranging cell treatments showed that of all the tested stresses , only heat , the chemical sodium azide , compacting of cells by sedimentation and carbon dioxide ( CO2 ) induced TB transformation . As in leaves , heat treatment of protoplasts only induced Tub+|TBs ( data not shown ) , whereas CO2 and azide stimulated Tub+|TBs as well as mixed-networks ( Figure 5A , B ) . The kinetics of TB transformations followed a precise order: first , standby TBs are loaded with tubulin , and then mixed-networks and tubulin-containing TB fragments appear , at the expense of Tub−|TBs . After 5 min ( CO2 treatment ) or 40 min ( azide treatment ) most cells displayed mixed-networks ( Figure 5C , D ) . A most remarkable property of the TBs was their rapid reversion from the mixed-network phenotype back to tubulin-less TBs; this was provoked either after substituting normal air for CO2 ( Figure 5E ) , or after removing azide from the culture medium ( Figure 5F ) . Moreover , it was possible to induce several consecutive rounds of TB transformation in the mixed-networks , as well as reversion to the same cell suspension , by continually relieving and resubmitting cells to stress ( Figure 5E ) . 10 . 7554/eLife . 00183 . 019Table 2 . Effect of various treatments on TB phenotype in infected protoplastsDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 019Type of treatmentTreatmentEffect on TBHormonesAbscisic acid [5 μM]–Jasmonate [40 μM]–Auxin [5 μM]–Salicylic acid [1 mM]–Mechanical/physical stressCompacting by sedimentationTub+|TB , mixed-networksElectroporation–Heat shockTub+|TBLight/dark cycle–Membrane depolarization–Membrane hyperpolarization–Microwaves–Music–Ultrasonication–Vortexing–ElicitorsArabinogalactan [1 mg/ml]–Chitosan [40 μg/ml]–Cryptogein [1 μM]–OthersCO2Tub+|TB , mixed networksSodium azide [0 . 02%]Tub+|TB , mixed networkspH–Infected protoplasts were treated/incubated under the conditions indicated , and the TB phenotype was then analyzed by immunofluorescence against P2 and α-tubulin . Protoplasts were incubated with hormones and elicitors , at the indicated final concentrations , for 60 min . Compaction of protoplasts by sedimentation was achieved by exposing them for 2 h at 9 . 81 m/s2 on a bench-top . Electroporation conditions were 400 Ω , 0 . 25 μFD and 0 . 5 or 1 kV . Heat shock was for 1 h at 37°C . Daylight/dark cycle was for 2 h each condition . Membrane depolarization and hyperpolarization were induced with 100 and 0 . 1 mM KCl in protoplast buffer , respectively . Microwave exposure was 3 s at 750 W . For the music treatment ( inspired by Braam and Davis , 1990 ) , Vanessa Paradis's ‘Joe le taxi’ song was played at moderate volume ( ∼60 db ) for 3 . 5 min with protoplasts ‘listening’ from opened Eppendorf tubes . Ultrasonication consisted of a 2 s pulse at 80% power using a Bioblock Vibracell 72434 apparatus; vortexing was for 5 s at maximal power using a Vortex Genie 2 machine . Conditions for CO2 and sodium azide treatments are described in 'Materials and methods' . For pH treatment , cells were incubated for 5 min with 10 mM K2HPO4/KH2PO4 titrated to pH 3 . 0 , 5 . 6 , 6 . 9 or 8 . 2 . Lower and higher pH values proved lethal to the cells and were not considered for analysis . In all cases , the survival of cells was verified as described by Widholm ( 1972 ) and only treatments sustaining viability of the cells were used for analysis . 10 . 7554/eLife . 00183 . 020Figure 5 . TB transformations have a precise temporal order and are reversible . ( A–B ) Kinetics of TB transformation . Protoplasts were treated with ( A ) CO2 or ( B ) azide , and then processed for immunofluorescence against P2 ( red ) and α-tubulin ( green ) ; nuclei were counterstained with DAPI ( blue ) . In both ( A ) and ( B ) , untreated protoplasts display tubulin-less TBs , and the three subsequent images show representative treated protoplasts , respectively displaying a Tub+|TB , a disintegrating TB and mixed-networks . All images are confocal projections , with the exception of the dissociating TB after azide treatment , which is a single section; each inset shows a single optical section from the enclosed zone . The orange arrows show the line scans and the scanning direction used to create the profiles of P2 ( red ) and α-tubulin ( green ) labeling intensity ( in arbitrary units , AU ) , which are displayed in the graphs to the right of the single sections . They reveal as in Figure 2A–C , that unstressed TBs display little to no tubulin label , whereas stressed TBs contain large amounts of tubulin in their centers . P2 colocalizes with microtubules in mixed-networks . The confocal stacks used to generate the image projections can be found in Figure 5—source data 1 and 2 . ( C–D ) Quantification of TB kinetics . The histograms show the kinetics of CO2- ( C ) and azide-triggered ( D ) TB transformation in protoplasts . Results from one out of three independent experiments are displayed . 1235 TBs ( Tub−|TBs , Tub+|TBs , and mixed networks ) were evaluated for the CO2 experiments , and 1662 TBs were evaluated for the azide experiments . See Figure 5—source data 3 and 4 for details . ( E–F ) Reversion of mixed-networks through two CO2/air cycles ( E ) , and after azide treatment ( F ) . Infected protoplasts were treated with CO2 or azide for the duration indicated . CO2 was subsequently removed by ventilation of the suspension with air; azide was removed by resuspending the protoplasts in fresh medium . Shown are data from one of three independent experiments . For the three repetitions , a total of 1339 TB morphs were analyzed for CO2 reversion , and 2262 TB morphs were analyzed for the azide reversion experiments . See Figure 5—source data 5 and 6 for details . SD: standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02010 . 7554/eLife . 00183 . 021Figure 5—source data 1 . Confocal single sections and acquisition parameters for Figure 5ADOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02110 . 7554/eLife . 00183 . 022Figure 5—source data 2 . Confocal single sections and acquisition parameters for Figure 5BDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02210 . 7554/eLife . 00183 . 023Figure 5—source data 3 . Source data for Figure 5CDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02310 . 7554/eLife . 00183 . 024Figure 5—source data 4 . Source data for Figure 5DDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02410 . 7554/eLife . 00183 . 025Figure 5—source data 5 . Source data for Figure 5EDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02510 . 7554/eLife . 00183 . 026Figure 5—source data 6 . Source data for Figure 5FDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 026 These time series experiments reveal P2 relocalization , from TBs to microtubules , following the application of different stresses . Are virus particles , the other major component of the CaMV transmissible complex , also distributed onto the mixed-networks ? To answer this question , we carried out immunofluorescence of the mixed-networks using a CaMV capsid protein antibody . Figure 6A shows that capsid protein localizes to cytoplasmic inclusions ( which are most likely virus factories ) in unstressed cells . Under conditions that trigger the formation of the mixed-networks , that is , CO2 ( Figure 6B ) or azide treatment ( Figure 6C ) , P4 label colocalized with microtubules . Quantification of this observation ( Figure 6D ) indicates that almost all cells displayed P4 networks after stress treatment . Nearly the same proportion of cells contained P2 or P4 networks after treatment with CO2 or azide ( compare Figure 6D with Figure 5C , D ) , which suggested that mixed-networks are also associated with virus particles . We confirmed this by electron- and immunogold microscopy , which showed ( in treated cells ) that cortical microtubules displaying P2 were indeed also decorated heavily with CaMV particles ( Figure 7A–C ) . Finally , we also examined the ultrastructure of standby and Tub+|TBs . The latter were induced by heat shock and then the tissues prepared for electron microscopy . Figure 7D–G shows that TBs consist of an electron-lucent matrix in which some virus particles are embedded as previously reported ( Espinoza et al . , 1991; Drucker et al . , 2002 ) . In control TBs , the virus particles seemed either to be distributed evenly throughout the TB matrix ( Figure 7D ) or to be more concentrated at their cortex ( Figure 7F ) . Heat-shocked TBs displayed similar TB phenotypes ( Figure 7E , G ) and we could not observe any flagrant differences in TB phenotype between heat-shock-induced Tub+|TBs and control TBs . This corresponded to the results obtained by fluorescence microscopy ( Figure 2D , E ) where likewise no obvious differences between the two TB morphs were observed . 10 . 7554/eLife . 00183 . 027Figure 6 . TB transformation mobilizes virus particles onto microtubules . ( A–C ) Viral capsid protein P4 colocalizes with mixed-networks . Protoplasts were either left unstressed ( A ) , incubated with CO2 for 15 min ( B ) or treated with azide for 40 min ( C ) , and then fixed and labeled to detect capsid protein P4 ( red ) and α-tubulin ( αTUA , green ) . The split channel representations and merges as well as the P4 and α-tubulin profiles obtained by scanning the lines indicated by the orange arrows show that the two stress treatments induced relocalization of capsid protein P4 from inclusions onto microtubules . As in Figure 2A–C , the intensity of the P4 and α-tubulin label is indicated in arbitrary units ( AU ) because different acquisition settings were used to record the images . ( A ) is a confocal projection , ( B–C ) are confocal single sections . Refer to Figure 6—source data 1–3 for image details . ( D ) Quantification of the effect of azide and CO2 on the localization of P4 . Cells were treated as indicated , processed for immunofluorescence against P4 and α-tubulin and scored for the presence of P4 in inclusions only , or in inclusions and on microtubules . The histogram shows that almost all cells display P4 networks that colocalize with microtubules after stress treatment . Data are from one of three independent experiments , in which a total of 524 cells were analyzed . Refer to Figure 6—source data 4 for details . Scale bars: 5 μm . SD: standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02710 . 7554/eLife . 00183 . 028Figure 6—source data 1 . Confocal projection and acquisition parameters for Figure 6A . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02810 . 7554/eLife . 00183 . 029Figure 6—source data 2 . Confocal single section and acquisition parameters for Figure 6B . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 02910 . 7554/eLife . 00183 . 030Figure 6—source data 3 . Confocal single section and acquisition parameters for Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03010 . 7554/eLife . 00183 . 031Figure 6—source data 4 . Source data for Figure 6D . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03110 . 7554/eLife . 00183 . 032Figure 7 . Electron microscopy of the different TB morphs . ( A–C ) Mixed-networks display virus particles on microtubules . The images show typical spherical CaMV virus particles ( arrowheads ) that decorate microtubules ( black arrows ) in cortical regions of ( A ) a CO2-treated or ( B ) an azide-treated protoplast . ( C ) Positive immunogold labeling against P2 ( the gray arrow points to an exemplary nanogold particle ) identifies the virus-decorated microtubules as mixed-networks , in which all components of the CaMV transmissible complex are present . ( D–G ) TBs in unstressed ( D , F ) and heat-shocked ( E , G ) tissue display the same TB phenotype . Infected Arabidopsis TUA6-GFP leaves were exposed for 1 h at 37°C . The presence of tubulin in TBs was then verified by fluorescence microscopy and the same leaf samples were processed for transmission electron microscopy . The arrowheads point to virus particles . CW: cell wall . For scale bars , ( A–C ) : 100 nm; ( E–G ) : 250 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 032 Taken together , these results demonstrate that mixed-network formation is always preceded by a massive entry of tubulin into the standby TB , which in itself does not induce a major ultrastructure change . However , under specific stress conditions such as azide or CO2 ( but not heat-shock ) , this tubulin entry can be followed by a total disruption of the TB . This includes an even dispersion of the components of the CaMV transmissible complex ( i . e . , P2 and virus particles ) onto microtubules throughout the cortical cytoplasm . This redistribution might render the CaMV transmissible complexes more readily accessible to aphid vectors . The results in Figures 3–5 show that standby TBs can transform reversibly into Tub+|TBs , and then again into mixed-networks , on a time scale that is fully compatible with aphid intracellular probing ( Fereres and Moreno , 2009 ) . However , in all of the above cases , TB transformations were induced by artificial stresses and might not reflect a ‘natural’ induction of TB changes in aphid-infested plants . We therefore aimed to determine whether aphid feeding activity by itself triggers TB transformation . Accordingly , we developed a protocol that simultaneously allows observation of the TB phenotype in infected cells and the discrimination of cells , based on whether or not they had been in direct contact with aphid stylets . For this procedure , aphids were allowed to feed on infected leaves for 15 min , before the leaf was fixed and subsequently screened by confocal microscopy . We then identified , by their auto-fluorescence , the salivary sheaths that remain in the tissue after aphid removal , and which precisely document the path followed by the stylets ( Miles , 1968 ) . In addition , we identified by immunofluorescence the different TB morphs in plant cells that were either in close contact or farther away from these sheaths . Standby TBs were predominantly detected in uninfested tissues , or in infested tissues greater than 15 μm from the stylet track ( Figure 8A ) . In contrast , cells of infested leaves within a 15 μm perimeter of salivary sheaths often displayed typical mixed-networks , fragmented TBs and Tub+|TBs ( Figure 8B , C ) . Mixed-networks in cells close to the stylet track were also loaded with virus particles , as indicated by positive P4 capsid protein label of microtubules ( Figure 8D ) . Quantification of the aphid-induced effect showed that 30–40% of TBs in cells in contact with a stylet track displayed a modified TB phenotype , whereas 99% of TBs in cells found more than one cell layer away from this track remained in the standby state ( Figure 8E ) . In parallel experiments , aphids were removed after the 15-min feeding period and the leaves were allowed to recover for 2 h before analysis . This intriguingly provoked the aphid-induced mixed-networks to revert back to standby TBs , as demonstrated by the strong decrease in the number of modified TBs close to the stylet tracks ( Figure 8E ) . This resembled the reversion observed in protoplasts , upon relief from either azide or CO2 treatment ( Figure 5E , F ) . Taken together , these results show that the probing activity of aphid stylets is a robust trigger of TB transformation , and that these aphid-induced TB changes are completely reversible . 10 . 7554/eLife . 00183 . 033Figure 8 . Mixed-networks appear in tissue zones pierced by aphid stylets . Unstressed CaMV-infected leaves ( A ) or leaves infested by aphids for 15 min ( B–D ) were analyzed by immunofluorescence microscopy . ( A ) Cells in leaf regions that were not foraged by aphids display standby Tub−|TBs , as shown by confocal projections of tissue sections labeled against P2 ( red ) and α-tubulin ( green ) . The optical single sections used for this projection are deposited in Figure 8—source data 1 . ( B ) In contrast , a cell close to a salivary sheath ( blue autofluorescence , digitally enhanced ) displays mixed-networks , in aphid-infested tissue . An enlargement of the zone enclosed in ( B ) is shown in ( C ) . ( B–C ) show confocal projections , please refer to Figure 8—source data 2 for the corresponding image stack . ( D ) Immunofluorescence microscopy against capsid protein P4 ( red ) and α-tubulin ( green ) shows that virus particles also localize to mixed-networks in cells close to salivary sheaths ( blue , digitally enhanced ) . Chloroplast autofluorescence appears in magenta . The confocal single sections used to produce this projection can be found in Figure 8—source data 3 . ( E ) Aphids trigger TB transformation , and this transformation is reversible . Aphids were placed for 15 min on infected leaves . Following this , the leaves were fixed immediately and processed for immunofluorescence ( 15-min aphid infestation ) , or the aphids were removed and the leaves were processed 2 h later ( Aphids removed ) . The TB phenotype ( standby Tub−|TBs , Tub+|TBs and mixed-networks ) was scored next to salivary sheaths ( 0–15 μm ) and in surrounding tissue ( 15–100 μm ) . Tub+|TBs and mixed-networks were predominantly observed close to salivary sheaths in freshly aphid-infested tissue . The effect was highly significant ( p<0 . 0001 , GLM , df = 1 , χ2 = 194 . 59 , n = 3 ) . Tub+|TBs and mixed-networks reverted back to ‘stand-by’ Tub−|TBs 2 h after aphid removal , indicating that TB activation is reversible . This effect was also highly significant ( p<0 . 0001 , GLM , df = 1 , χ2 = 17 . 98 , n = 3 ) . SD in ( E ) : standard deviation from three independent experiments . A total of 969 TBs surrounding 42 sheaths were counted from freshly aphid-infested tissue , and 194 TBs surrounding eight sheaths were counted in the ‘aphids removed’ experiments . Original data can be found in Figure 8—source data 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03310 . 7554/eLife . 00183 . 034Figure 8—source data 1 . Confocal single sections and acquisition parameters for Figure 8A . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03410 . 7554/eLife . 00183 . 035Figure 8—source data 2 . Confocal single sections and acquisition parameters for Figure 8B . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03510 . 7554/eLife . 00183 . 036Figure 8—source data 3 . Confocal single sections and acquisition parameters for Figure 8D . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03610 . 7554/eLife . 00183 . 037Figure 8—source data 4 . Source data for Figure 8E . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 037 In order to be biologically relevant , aphid-induced TB morphs ( Tub+|TBs and/or mixed-networks ) should display a significant impact on CaMV transmission . We therefore investigated whether TB transformation is required for aphid-transmission of CaMV . Assessing the effect of TB transformation on transmission was however not straightforward , since aphids themselves provoke TB transformation . As a possible means to circumvent this problem , we observed while quantifying TB morphs ( Figure 8E ) that only 30–40% of the TBs had transformed after contact with the stylets . We thus reasoned that a pre-treatment of infected cells that would substantially increase the proportion of modified TBs prior to aphid feeding should ‘prime’ these cells for virus acquisition and enhance the CaMV transmission rate . To investigate this , we first induced mixed-networks in protoplasts with azide , and then used the cells in aphid transmission experiments . We observed significantly elevated transmission rates compared to transmission from control cells that displayed mainly standby TBs ( Figure 9A ) . The effect was not caused by altered aphid behavior resulting from the presence of azide in the protoplast medium , because the chemical had no effect in transmission experiments in which aphids were allowed to acquire CaMV instead from cells from suspensions containing purified virus , recombinant P2 and P3 ( Figure 9B ) . Azide also had no effect on protoplast viability under the conditions used ( Figure 9C ) . Taken together , these results rule out a possible confounding effect of azide , and clearly indicate that it was the presence of mixed-networks that lead to increased transmission rates . In order to compare this situation with that of intact plant tissues , we examined infected CO2-treated or heat-shocked leaves in transmission assays . Under heat shock treatment , TB transformation appeared incomplete and arrested at the Tub+|TB stage , as reported above ( Figure 2D and Figure 2—figure supplement 1 ) . The heat-shocked leaves did not perform any better than controls in aphid-transmission tests ( Figure 9D ) , indicating that tubulin entry into TBs alone is not sufficient to enhance transmission . In contrast , CO2 induced complete TB transformation into mixed-networks ( Figure 2E ) . Moreover , significantly enhanced transmission rates were recorded when CO2-treated leaves were used in aphid transmission experiments ( Figure 9E ) . 10 . 7554/eLife . 00183 . 038Figure 9 . TB transformation correlates with enhanced transmission efficiency . ( A ) Azide enhances transmission from protoplasts . Aphids were allowed to acquire CaMV from infected protoplasts that displayed mixed-networks induced by azide . They were then transferred to healthy test plants for inoculation , and infected plants were counted 3 weeks later . The difference in transmission was highly significant ( p<0 . 0001 , hierarchical GLM model , Table 3 and Figure 9—source data 1 ) . ( B ) Azide does not affect aphid behavior . To rule out an unwanted effect of azide on aphid viability and behavior , aphids were membrane-fed solutions containing purified virus particles , recombinant P2 and P3 , in the presence or absence of azide , and then transferred to healthy test plants for inoculation . Transmission rates were determined 3 weeks later by scoring infected plants . Data from one experiment are shown , using three different virus preparations as a virus source for each condition . See Figure 9—source data 2 for details . ( C ) Azide does not affect protoplast viability . Protoplasts were incubated for 1 h ( the duration of a transmission test ) in the presence or absence of 0 . 02% azide , and then protoplast viability was determined with the FDA test ( Widholm , 1972 ) . Data from one out of two experiments are shown . The difference in viability was insignificant in this experiment ( p=0 . 0658 , n = 6 , Mann–Whitney test ) and also in the second experiment . See Figure 9—source data 3 for all data . ( D ) Heat shock does not enhance CaMV transmission . Leaves from GFP-TUA6 Arabidopsis either received heat shock ( + ) or did not ( − ) . The presence of Tub+|TBs was verified by fluorescence microscopy and the leaves were then used in aphid transmission assays . No significant difference in transmission was observed in either of two independent experiments ( p=0 . 73 and p=0 . 08 , respectively , hierarchical GLM model , see Table 4 and Figure 9—source data 4 ) . ( E ) CO2 enhances CaMV transmission . Leaves with mixed-networks induced by CO2 were used in plant-to-plant aphid transmission experiments . CO2-treated leaves performed significantly better in transmission tests than controls ( p=0 . 0025 , hierarchical GLM model , see Table 5 and Figure 9—source data 5 ) . ( F ) Oryzalin induces Tub+|TBs . Immunofluorescence of oryzalin-treated protoplasts shows that α-tubulin ( green ) accumulates with P2 ( red ) in TBs . The nucleus is stained with DAPI ( blue ) . The image is a confocal projection . The insets show a separate channel presentation of a representative optical single section of the TB , for details refer to the image stack in Figure 9—source data 6 that was used for this projection . Scale bar = 10 μm . ( G ) Kinetics of Tub+|TB formation in protoplasts that were treated with oryzalin for the duration indicated . Most TBs transformed to the Tub+-state within 15 min . Mixed-networks were not observed and thus are not indicated in the histogram . Data is from one of three independent experiments , where a total of 1556 TBs were analyzed . See Figure 9—source data 7 for details . SD: standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03810 . 7554/eLife . 00183 . 039Figure 9—source data 1 . Source data for Figure 9A . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 03910 . 7554/eLife . 00183 . 040Figure 9—source data 2 . Source data for Figure 9B . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04010 . 7554/eLife . 00183 . 041Figure 9—source data 3 . Source data for Figure 9C . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04110 . 7554/eLife . 00183 . 042Figure 9—source data 4 . Source data for Figure 9D . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04210 . 7554/eLife . 00183 . 043Figure 9—source data 5 . Source data for Figure 9E . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04310 . 7554/eLife . 00183 . 044Figure 9—source data 6 . Confocal single sections and acquisition parameters for Figure 9F . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04410 . 7554/eLife . 00183 . 045Figure 9—source data 7 . Source data for Figure 9G . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04510 . 7554/eLife . 00183 . 046Table 3 . Statistical analysis of transmission experiments using azide-treated protoplasts as virus sourceDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 046ExperimentnTransmission frequencyLCIUCIGlobal−Azide160 . 240 . 180 . 30+Azide160 . 400 . 330 . 47Experiment 1−Azide30 . 320 . 220 . 43+Azide30 . 380 . 270 . 50Experiment 2−Azide30 . 440 . 320 . 55+Azide30 . 660 . 540 . 76Experiment 3−Azide30 . 190 . 110 . 29+Azide30 . 410 . 300 . 52Experiment 4−Azide30 . 330 . 220 . 44+Azide30 . 460 . 350 . 58Experiment 5−Azide40 . 100 . 030 . 19+Azide40 . 380 . 290 . 48We measured the transmission rate by aphids for each condition of treatment ( −azide and +azide ) for the five experiments . Since a non-significant interaction between experiments and treatments was found ( GLM , df = 4 , χ2 = 6 . 69 , p=0 . 15 ) , the data was pooled in the line named ‘global’ . Azide induced a highly significant increase of the transmission rate compared to the control without azide ( hierarchical GLM model using Firth's penalized likelihood , df = 1 , χ2 = 35 . 29 , p<0 . 0001 ) with a transmission rate of 23 . 7% ( 95% CI: 17 . 6–29 . 7% ) for control and 39 . 9% ( 33 . 2–46 . 6% ) for azide treatment . CI: confidence interval; n: number of repetitions per experiment; LCI , UCI: lower and upper limits of confidence intervals , respectively . 10 . 7554/eLife . 00183 . 047Table 4 . Statistical analysis of transmission experiments using heat-shocked leaves as virus sourceDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 047ExperimentnTransmission frequencyLCIUCIExperiment 1Control60 . 620 . 540 . 7037°C60 . 640 . 400 . 72Experiment 2Control60 . 600 . 510 . 6837°C60 . 490 . 400 . 57Data for experiments 1 and 2 were analyzed independently , as heat-shock slightly increased and decreased transmission in these experiments , respectively . The heat shock treatment ( 90 min at 37°C ) induced no significant difference in transmission rate ( compared to controls ) for experiment 1 ( hierarchical GLM , df = 1 , χ2 = 0 . 12 , p=0 . 73 ) , with 62% ( 95% CI: 53 . 6–70% ) for control and 64% ( 55 . 8–71 . 7% ) for heat-shock treatment , or for experiment 2 ( hierarchical GLM , df = 1 , χ2 = 3 . 12 , p=0 . 08 ) , with 59 . 5% for control ( 50 . 6–67 . 9% ) and 48 . 6 % ( 40 . 4–56 . 9% ) for heat-shock treatment . n: number of repetitions per experiment; LCI , UCI: lower and upper limits of confidence intervals , respectively . 10 . 7554/eLife . 00183 . 048Table 5 . Statistical analysis of transmission experiments using CO2-treated leaves as virus sourceDOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 048ExperimentnTransmission frequencyLCIUCIGlobalControl160 . 130 . 090 . 17CO2160 . 220 . 180 . 27Experiment 1Control40 . 110 . 050 . 19CO240 . 150 . 070 . 25Experiment 2Control60 . 140 . 080 . 21CO260 . 210 . 140 . 28Experiment 3Control60 . 130 . 070 . 20CO260 . 300 . 220 . 39We measured the transmission rate by aphids for each treatment condition ( control or CO2 ) for the three experiments . As we found a non-significant interaction between experiments and treatments ( GLM , df = 2 , χ2 = 2 . 92 , p=0 . 23 ) , the three experiments were pooled for analysis ( line named ‘global’ ) . CO2 induced a significant increase in the transmission rate compared to the controls without CO2 ( hierarchical GLM model , df = 1 , χ2 = 9 . 12 , p=0 . 0025 ) , with 13 . 0% ( 95% CI: 9 . 4–17 . 0% ) for control in ambient atmosphere and 22 . 4% ( 17 . 8–27 . 2% ) for CO2-treatments . n: number of repetitions per experiment; LCI , UCI: lower and upper limits of confidence intervals , respectively . We conclude from these results that the sole entry of tubulin into TBs is insufficient to explain increased transmission . Furthermore , we hypothesize that the higher accessibility of CaMV to its aphid vector can be explained by microtubules serving as a scaffold for the rapid redistribution of P2 and virus particles over the entire cell . If this were true , then depolymerization of microtubules by oryzalin should both prevent the formation of mixed-networks , and significantly decrease the transmission rate . We have previously shown that oryzalin diminishes transmission from infected protoplasts ( Martinière et al . , 2011a ) . Consistent with these results , we show here that oryzalin induces Tub+|TBs within 15 min , but prevents mixed-network formation ( Figure 9F , G ) . To further confirm the role of TB changes in aphid-transmission of CaMV , we pursued an independent approach and examined a CaMV mutant impaired in transmission . This mutant , CaMV P2-TC , harbors a 7-amino acid insertion , including a tetracysteine tag ( Griffin et al . , 1998 ) at position 100 of the P2 protein . The mutant virus was fully infectious as compared to wild type virus; furthermore , the P2-TC protein as well as viral proteins P3 , P4 and P6 accumulated to similar levels in infected plants ( Figure 10A ) . However , in comparison to wild type TBs , this mutant induced TBs ( TB-TCs ) that seemed to be smaller , with a more regular rounded shape , and a more pronounced P2-rich cortex , as revealed by immunofluorescence ( Figure 10B ) . Heat shock induced an influx of tubulin into TB-TCs , and photobleached GFP-tubulin contained in TB-TC was observed to be exchanged with cytoplasmic tubulin in FRAP experiments ( Figure 10C ) , although the kinetics differed from tubulin replacement in wild type TBs ( compare Figure 10C with Figure 4B ) . Thus the TB of the P2-TC mutant bears some similarity to wild type TB . Nevertheless , CaMV P2-TC was completely non-transmissible in plant-to-plant transmission experiments ( Figure 10D ) . This could be due to a defect of TB-TC in undergoing correct transformation upon aphid puncture; an alternative is that this is due to a lack of interaction between the mutant P2-TC protein and either virus particles or aphid stylets . To distinguish between these possibilities , we first tested whether P2-TC protein can mediate the binding of virus particles to the aphid stylets . One way to test this is to allow aphids to feed on suspensions containing recombinant P2-TC , P3 and purified virus particles through membranes , before they are transferred to test plants for inoculation . Binding of transmissible complexes acquired by the aphids from the feeding solution is then scored by counting the number of successful transmission events , that is , the number of infected test plants . Although the transmission rates were significantly lower than that obtained with wild type P2 , the P2-TC mutant protein was indeed active in such assays , retaining around 50% of the wild P2 activity ( Figure 10E ) . We thus reasoned that defects in P2-TC binding to virus particles or stylets can only partially explain the complete failure in plant-to-plant transmission of the P2-TC mutant , and that a failure in TB-TC transformation may also be involved . To investigate this possibility , we allowed aphids to infest CaMV-P2TC-infected leaves for 15 min . The leaves were then processed for immunofluorescence against P2 and α-tubulin , and scored for the TB phenotype in cells found close to and farther away from the salivary sheaths . Figure 10F , G demonstrates that only standby TBs were detected , and not Tub+|TBs or mixed-networks . This result is further evidence for a strong positive correlation between the appearance of mixed-networks ( i . e . , TB transformation ) and successful aphid transmission , and presents direct biological evidence in support of our transmission hypothesis . 10 . 7554/eLife . 00183 . 049Figure 10 . The P2-TC mutant of CaMV is inactive in plant-to-plant transmission . ( A ) Accumulation of viral proteins in CaMV P2-TC-infected plants . Western blot analysis of total leaf extracts shows that virus factory protein P6 , the three forms of capsid protein P4 , as well as P2 and P3 accumulate to similar levels in plants infected with wild type CaMV ( B-JI ) or the P2-TC mutant ( TC ) . Rub = Rubisco loading control stained with Ponceau Red . ( B ) CaMV P2-TC-infected plants display TBs . Confocal projection of infected leaf sections labeled for P2 ( red ) and α-tubulin ( green ) shows that the CaMV mutant P2-TC forms TBs ( arrows ) that are smaller and more regular than wild type TBs . Nuclei are counterstained with DAPI ( blue ) . The optical single sections used for the projection are presented in Figure 10—source data 1 . ( C ) Tubulin turnover in P2-TC mutant TBs . Arabidopsis plants constitutively expressing GFP-tubulin were infected with the CaMV P2-TC mutant . Leaf epidermis was screened by fluorescence microscopy for GFP-tubulin inclusions that were identified as TBs based on their typical shape . The GFP-tubulin was photobleached in these TBs , and the recovery of the GFP-fluorescence ( due to replacement of the photobleached GFP-tubulin by fresh tubulin ) was recorded in FRAP experiments . The graph shows recovery kinetics from t = 0 s ( time point of photobleach ) onwards . The fluorescence levels were normalized ( 100% = fluorescence before bleaching , 0% = fluorescence just after bleaching ) . The red trend line was calculated from nine experiments . Data points from the nine experiments are indicated as blue dots . These results indicate that tubulin cycles between the cytoplasm and mutant TBs , albeit with different kinetics than for wild type TBs ( compare with Figure 4C ) . Compared to wild type TBs , the t ( 1/2 ) for fluorescence recovery was significantly slower ( p<0 . 0001 , t-test with n = 21 for wild type TBs and n = 23 for TC-TBs ) and the proportion of the mobile fraction was significantly higher in P2-TC TBs ( p=0 . 0001 , t-test with n = 21 for wild type TBs and n = 23 for TC-TBs ) . Refer to Figure 10—source data 2 and Figure 4—source data 1 ( wild type TBs ) for data sets . ( D ) The mutant P2-TC does not support plant-to-plant transmission . Aphids were placed for 15 min on CaMV wild type-infected ( B-JI ) or P2-TC-infected ( TC ) leaves and then transferred to healthy test plants for inoculation . Infected plants were scored 3 weeks later . Pooled data are shown from two independent experiments using 12 different leaves for each condition . No statistical analysis was performed , as the effect of the P2-TC mutant on plant-to-plant transmission was total ( no transmission from P2-TC-infected plants was observed ) . See Figure 10—source data 3 for the data sets . ( E ) The P2-TC protein itself is active in transmission . Recombinant wild type P2 ( B-JI ) or mutant P2-TC ( TC ) were mixed together with recombinant P3 protein and purified CaMV particles . Aphids were allowed to feed on the suspensions across membranes for 15 min and were then transferred to healthy test plants for inoculation . Infected plants were counted 3 weeks later . The histogram shows that P2-TC supported aphid transmission of CaMV under these conditions , although this was significantly reduced as compared to the wild type P2 ( p=0 . 02 , n = 8 from two independent experiments , Mann–Whitney test ) . See Figure 10—source data 4 for data . ( F ) Aphid stylet activity does not trigger TB transformation in CaMV P2-TC-infected leaves . Aphids were allowed to feed on CaMV-P2-TC-infected leaves for 15 min . The tissue was then processed for immunofluorescence against P2 ( red ) and α-tubulin ( green ) ; nuclei were stained with DAPI ( blue ) . The confocal projection in ( F ) indicates that cells in contact with a salivary sheath ( Sheath ) display tubulin-less TBs ( arrows ) . Chloroplasts are displayed in magenta to better distinguish the cells . Please see Figure 10—source data 5 for the confocal single sections used to create the projection . ( G ) Quantitative analysis of the TB forms of CaMV-P2-TC in aphid-infested tissue reveals the absence of Tub+|TBs and mixed-networks , both close to salivary sheaths ( 0–15 μm ) and farther away ( 15–100 μm ) . Data shown are from three independent experiments where a total of 510 TBs were analyzed ( see Figure 10—source data 6 for details ) . As the effect was total , no statistical analysis was performed . SD: standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 04910 . 7554/eLife . 00183 . 050Figure 10—source data 1 . Confocal single sections and acquisition parameters for Figure 10B . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 05010 . 7554/eLife . 00183 . 051Figure 10—source data 2 . Source data for Figure 10C . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 05110 . 7554/eLife . 00183 . 052Figure 10—source data 3 . Source data for Figure 10D . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 05210 . 7554/eLife . 00183 . 053Figure 10—source data 4 . Source data for Figure 10E . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 05310 . 7554/eLife . 00183 . 054Figure 10—source data 5 . Confocal single sections and acquisition parameters for Figure 10F . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 05410 . 7554/eLife . 00183 . 055Figure 10—source data 6 . Source data for Figure 10G . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 055
Combined , our results establish that the TB is a dynamic structure , which in the presence of an aphid vector can react immediately to promote CaMV transmission ( Figure 11 ) . Transmission is controlled by the different TB morphs . The tubulin-less TB found in normal ‘unstressed’ cells has been well-described over the past several decades ( Shalla et al . , 1980; Rodriguez et al . , 1987; Espinoza et al . , 1991; Blanc et al . , 1993b; Drucker et al . , 2002 ) , and here functions as a standby TB that ‘anticipates’ the vector arrival . Then , when aphids land on the plant and insert their stylets into tissues , CaMV uses the plant's response during stylet entry to its advantage at a very early stage of the plant–aphid interaction . This results in TBs that undergo dramatic , short-lived changes leading to the temporary redistribution of P2 and virus onto microtubules . These hitherto overlooked TB alterations result in a reversible TB ‘activation’ that optimizes virus acquisition . The activation or increased transmission efficiency is probably due to the mixed-networks distributing P2 and virus homogeneously on microtubules throughout the cell periphery . This could in turn facilitate virus acquisition: in this new configuration , P2 and the virus are more accessible to the vector during its random punctures , as compared to the remote localization of P2 in isolated TBs . This hypothesis is supported by the observation that inhibition of mixed-network formation , either in the CaMV-P2TC mutant or pharmacologically by oryzalin , resulted in decreased transmission . The converse situation also applies , as artificial induction of mixed-networks by CO2 and azide were correlated strongly with increased transmission rates . These results additionally indicate that the TB reaction is required for transmission . 10 . 7554/eLife . 00183 . 056Figure 11 . Model of CaMV acquisition . ( A ) In an infected cell in the ‘standby’ state , there are numerous virus factories ( VF ) containing most of the replicated virus particles ( yellow-blue circles ) , enclosed within a matrix of viral protein P6 ( grey ) . This is accompanied in the cell by a mostly single transmission body ( TB ) , composed of a matrix containing all of the cell's P2 ( red ) , co-aggregated with P3 ( blue ) and some virus particles . Microtubules are represented in green . ( B ) An aphid landing on an infected plant inserts its stylets into a cell to test the plant . This causes a mechanical stress ( stylet movement ) and/or a chemical stress ( e . g . , elicited by saliva components ) . This stress , symbolized by the yellow flashes , is immediately perceived by the plant and can induce subsequent defense responses . The initial aphid recognition signal is transduced simultaneously in a TB response , characterized by an influx of tubulin ( green ) into the TB . ( C ) In a second step , the TB disintegrates rapidly ( within seconds ) , and all the P2 as well as some virus particles relocalize on the cortical microtubules as mixed-networks . Whether or not the virus particles originate at the VFs , as presented here , is unknown . Transmissible P2-virus complexes are now homogeneously distributed throughout the cell periphery , which significantly increases the chances of successful binding of P2 and virus to the stylets and thus transmission . ( D ) After departure of the aphid vector ( here loaded with P2 and virus ) , a new TB is reformed from the mixed-networks and is ready for another round of transmission . DOI: http://dx . doi . org/10 . 7554/eLife . 00183 . 056 One enduring question that deserves further attention is precisely where the virus particles come from that are recruited onto the microtubules . Whereas the origin of microtubule-associated P2 is clearly the TB ( since the TB is the only source of P2 ) , the origin of the virus particles aligning on the microtubules is less clear . They could derive either from the TB , which contains some virus particles , or from the many virus factories dispersed throughout the cytoplasm ( Espinoza et al . , 1991; Drucker et al . , 2002 ) . We were also intrigued to observe that only certain stresses—notably aphid feeding activity , wounding , azide and CO2—could trigger TB transformation . This reveals a certain level of specificity in TB activation , although it is more broad than the classical pattern recognition receptor-mediated defense responses of plants against pathogens ( including insects ) that are often species or even isolate-specific ( Hogenhout and Bos , 2011 ) . This broad specificity is not surprising , since CaMV is transmitted by at least 30 different aphid species . Importantly , this suggests that the plant responses against aphids ( that are probably exploited by CaMV for the TB reaction ) are triggered by an elicitor common to all aphids . Whether this elicitor triggers an innate plant immunity pathway or a separate perception/reaction cascade remains an open question . The significance of this remarkable phenomenon described here extends beyond CaMV transmission to broader fields of research . First , this opens up a fascinating new direction within virology , to explore whether other viruses form transmission morphs in response to vector-sensing by the host . Second , the transient accumulation of tubulin in TBs , followed by redistribution of TB contents on microtubules uncovers unforeseen capacities of tubulin/microtubule dynamics and raises further questions pertinent to cell biology . For example: How can apparently soluble tubulin concentrate in the TB and what is its function there ? Does this serve as the source for the mixed-networks ? Most strikingly , our work reveals that a virus can detect external stresses ( probably by using its host's perception system ) and respond in a way that is somewhat independent of the host's response . We propose naming this phenomenon: ‘virus perceptive behavior’ . This concept is nicely illustrated by three compelling observations made in this study . First , TB transformation occurs while the host plant is still in the process of transducing the triggering signal . This shows that the virus appropriates the host's perception machinery itself , rather than relying upon downstream reactions that take tens of minutes ( or hours ) to manifest and establish local and systemic defense responses ( Kuśnierczyk et al . , 2008; de Vos and Jander , 2010 ) . Second , after transformation of the TB into mixed-networks , the fate of CaMV appears disconnected from the final host response . Indeed , within the time frame required for the host plant to respond to an aphid attack , mixed-networks have already served as a robust virus source for aphids , and have long since reverted back to standby TBs . Third , the response mechanisms themselves are also entirely different from the reported plant physiological responses to aphid attack . These include callose deposition near the salivary sheaths and within sieve tubes ( Villada et al . , 2009 ) , changes in gene expression patterns ( Kuśnierczyk et al . , 2008 ) , altered emission of volatile compounds ( de Vos and Jander , 2010 ) ; and the initiation of salicylic , abscisic , and jasmonic acid systemic defense pathways ( Giovanini et al . , 2007; Kuśnierczyk et al . , 2008; de Vos and Jander , 2009 ) . In contrast , the TB response seems to be restricted to a CaMV-specific and immediate diversion of tubulin/microtubules ( plus putative unknown associated partners ) for virus transmission , in a manner unlike anything described before . Whether such viral perceptive behaviors play a role in the vector-transmission of other viruses is entirely unknown , and will thus be a question of great priority in the field of research on virus transmission . Viruses tightly regulate all the different steps of their life cycle , from intracellular replication and short- and long-distance intra-host movement , to inter-host spread . In this sense , the ability to specifically trigger the ‘transmission-mode’ at the right time and the right place seems like a valuable adaptation for avoiding the deleterious interference between these various functions . On a more broad scope , our results highlight many unexpected research horizons to explore in the biology of these fascinating pathogens . The possible instances in which viruses could react directly to cues from the host environment , the diversity of sensorial pathways that could be exploited in both animal and plant hosts , and the number of key life cycle steps that could be optimized accordingly all inspire questions that will shape future research directions in this field . Finally , aside from being an academic challenge , this phenomenon also represents a potential Achilles heel in viral transmission that could lead to novel virus control strategies .
Turnip plants ( Brassica rapa cv . ‘Just Right’ ) and transgenic Arabidopsis thaliana Col0 plants with a gl1 marker expressing GFP-TUA6 under control of the 35S promoter ( Ueda et al . , 1999 ) were alternatively used as CaMV hosts , depending on the experiment . Two-week-old plants were mechanically inoculated with wild-type CaMV strain Cabb B-JI ( Delseny and Hull , 1983 ) or Cabb B-JI ΔP2 as described in Martinière et al . ( 2009 ) , and processed as indicated at 14 days post infection ( dpi ) . In order to obtain the mutant virus Cabb B-JI P2-TC ( referred to as P2-TC in the text ) , the oligonucleotides 5′-TCGAGTTGCTGTCCAGGATGTTGC-3′ and 5′-TCGAGCAACATCCTGGACAGCAAC-3′ were annealed . This created XhoI-compatible restriction sites at the two extremities of the then double-stranded oligonucleotide that were used for insertion into the unique XhoI site in the Cabb B-JI genome cloned into the pCa24 plasmid ( Delseny and Hull , 1983 ) . Positive clones were identified by PCR and verified by sequencing . They contained a seven-amino-acid insertion at amino acid position 100 of the P2 open reading frame , coding for a tetracysteine tag ( CCPGCC [Griffin et al . , 1998] ) as well as an additional serine . A non-viruliferous clonal Myzus persicae population was reared under controlled conditions ( 22/18°C day/night with a photoperiod of 14/10 h day/night ) on eggplant and cultivated by G . Labonne ( INRA , Montpellier ) . The population was started from a single virginiparous female . To produce recombinant P2-TC using the Sf9/baculovirus system , the tetracysteine sequence and an additional serine were introduced into the unique XhoI site in the P2 coding region of plasmid p119-P2 , using the same strategy as for cloning CaMV-P2TC described above . Recombinant baculovirus was obtained by homologous recombination as described in Blanc et al . ( 1993b ) . Infected Sf9 cells were harvested 48 h after inoculation and total cell extracts were prepared in SES buffer and stored at −20°C until use . Protoplasts were prepared from healthy or infected ( 14 dpi ) leaves of turnip plants as described in Martinière et al . ( 2009 ) . Briefly , leaves were sterilized by submerging them in 20-fold diluted Domestos solution ( http://www . unilever . com ) for 3 min . The leaves were then washed three times with water , prior to overnight incubation in protoplast medium M ( 0 . 5 M mannitol , 1 mM CaCl2 , 10 mM MES , pH 5 . 8 ) containing freshly added 0 . 5% cellulose ‘Onozuka’ R10 and 0 . 05% macerozyme R10 ( both enzymes obtained from Yakult , http://www . yakult . co . jp/ypi/ ) . Protoplasts were separated from undigested tissue by filtration through Miracloth ( http://www . merckmillipore . com ) , and washed three times with buffer M by centrifugation at 80×g for 5 min in a swing-out rotor . Prior to treatments , protoplasts were maintained fourfold diluted in buffer M at room temperature with slow agitation ( 5 rpm ) for 2 h . The different drug treatments with respective experimental times were: 10 μM oryzalin ( 1 h ) , 0 . 02% azide ( 40 min ) , or CO2 atmosphere ( 15 min ) . Azide ( 100× concentration ) or oryzalin ( 1000× concentration ) stock solutions were added to water or DMSO , respectively . Pure solvent was used as a control . For CO2 treatment , leaves or protoplasts were placed in a plastic box filled with CO2 that was generated by sublimation of dry ice in water , contained in a small beaker in the box . We visually confirmed the displacement of the water cloud initially created by the subliming CO2 to assure that the heavier CO2 had replaced the air ( no more water vapor visible ) ; only then was the plant material placed in the box . We also verified that the dry ice did not lower the temperature of the atmosphere in the box . For TB reversion , protoplasts were cycled every 15 min between the plastic box and standard bench-top conditions for the CO2 treatment or the azide was removed by replacing the protoplast medium with fresh medium after centrifugation of the protoplasts for 5 min at 80×g in a swing-out rotor . Protoplast viability was verified by the fluorescein diacetate test ( Widholm , 1972 ) . For heat shock treatment , protoplasts or plants were placed in an incubator at 37°C . To inflict wounding , leaves were cut with a new razor blade . Leaf segments ( 5- to 7-mm-long ) of turnip or Arabidopsis were fixed in 1% glutaraldehyde prepared in stabilizing buffer ( 50 mM HEPES , pH 8 ) . The tissue was then embedded in either Steedman's wax , as described in Vitha et al . ( 2000 ) , or in 5% agarose . For the Steedman's wax method , the leaf segments were rinsed twice for 10 min each with 50 mM HEPES pH 8 , followed by two 10-min washes with PBS . After dehydration through an ethanol series , the samples were infiltrated at 40°C with Steedman's wax by using a graded ethanol/wax series . Finally , the segments were embedded in pure Steedman's wax . After polymerization of blocks at room temperature , a microtome was used to cut slices with a 14-μm thickness . For the agarose method , the leaf segments were rinsed twice for 5 min in PBS and then embedded in 5% low melting temperature agarose . Finally , 50 μm sections were cut with a vibratome . Samples for electron microscopy and immunoelectron microscopy were processed as described in Drucker et al . ( 2002 ) . For transmission electron microscopy , infected leaves or agarose-embedded protoplasts were fixed with 4% glutaraldehyde , postfixed with 2% OsO4 , and embedded in Epon resin ( http://www . emsdiasum . com ) . For immunoelectron microscopy , protoplasts were fixed with 0 . 5% glutaraldehyde and 2% paraformaldehyde and embedded in LR Gold resin ( http://www . emsdiasum . com ) . All primary antisera and secondary antibodies were used at 1:25 dilution . The grids were observed in a Jeol JEM 100CX II electron microscope ( http://www . jeol . com ) operated at 60–80 kV . The following antibodies or antisera were used: rabbit anti-P2 ( Blanc et al . , 1993b ) , anti-P3 ( Drucker et al . , 2002 ) , anti-P6 ( Khelifa et al . , 2007 ) , monoclonal mouse anti-α-tubulin DM1A ( http://www . sigmaaldrich . com; Blose et al . , 1984 ) , and rabbit anti-P4 ( http://plant . neogeneurope . com ) . For secondary antibodies , we used Alexa 488 and Alexa 594 conjugates ( http://www . lifetechnologies . com ) or 10 nm colloidal gold conjugates ( http://www . bbigold . com ) . Plant tissue was ground in liquid nitrogen , and the powder was resuspended in 2× Laemmli buffer ( Laemmli , 1970 ) and boiled for 5 min . After brief centrifugation in a tabletop centrifuge ( 5 min at 16 , 000×g ) , aliquots were separated by SDS/PAGE using 12% gels . Proteins were transferred onto nitrocellulose membranes and antigens were revealed by the NBT-BCIP reaction as described in Drucker et al . ( 2002 ) . After treatments , protoplasts were fixed for 20 min at room temperature with 1% glutaraldehyde in 0 . 5 M mannitol and 50 mM HEPES pH 8 , and washed with TS buffer ( 50 mM Tris , 150 mM NaCl , pH 7 . 4 ) . Protoplasts were immobilized on polylysine-coated slides , incubated for 15 min with 0 . 2% NaBH4 , washed with TS and blocked with TS containing 5% dry milk powder ( TS-M ) for 30 min . The slides were incubated with primary antisera ( all diluted 1:250 in TS-M ) for at least 2 h . After two rinses with TS , slides were incubated for at least 2 h with secondary antibodies diluted 1:300 . After two rinses with TS , slides were mounted in antifading medium , which optionally included DAPI ( 50 ng/ml ) . The 14-μm microtome sections were immobilized on polylysine-coated slides , incubated for 1 h with 0 . 2% NaBH4 , washed with TS and incubated for 90 min in an enzyme solution ( containing 2% cellulose ‘Onozuka’ R10 , 1% macerozyme R10 ) and 2% driselase ( http://www . sigmaaldrich . com , prepared in 10 mM MES pH 5 . 6 ) . The 50-μm vibratome sections were incubated for 1 h with 0 . 2% NaBH4 in a 24-well plate . All sections were then blocked with 3% BSA or 5% BSA in TS supplemented with 0 . 01% Tween20 for 30 min and incubated with primary antisera in 1% BSA/0 . 01% Tween20 in TS-M for at least 12 h at the following dilutions: 1:200 for rabbit anti-P2 and rabbit anti-P4 and 1:100 for mouse anti-α-tubulin . After two rinses with TS or PBS , slides were incubated for at least 12 h with Alexa Fluor conjugates at a 1:200 dilution . After two rinses with TS or PBS , slides were mounted as described above . Slides were observed with Zeiss LSM510 or LSM700 ( http://www . zeiss . com ) or Leica SP2 ( http://www . leica . com ) confocal microscopes operated in sequential mode to avoid crosstalk . Raw images were processed using LSM , ZEN or LAS software and final figures were prepared using GIMP 2 . 6 . 11 ( http://www . gimp . org ) and OpenOffice 3 . 4 ( http://www . openoffice . org ) . Quantification of TB phenotype was performed by counting 200–300 cells in 10 different , randomly chosen microscopy fields per treatment . Groups of about 500 aphids were placed inside copper rings covered with stretched Parafilm M membranes ( http://www . parafilm . com ) for a 1 h pre-acquisition period in a humid chamber . Then , either protoplasts or suspensions containing purified virus particles , P2 and P3 in SES buffer ( Blanc et al . , 1993a ) were placed on the Parafilm M and covered with a cover slip . Aphid were then allowed a 15 min acquisition feed through the Parafilm membranes on the suspensions . For plant-to-plant transmission experiments , aphids were transferred to an infected detached leaf for 1–5 min acquisition feeding . Afterwards , either 1 aphid ( fed on a leaf ) or 10 aphids ( fed on protoplasts or virus particles ) were transferred onto each turnip test plantlet for a 4 ± 1 h inoculation period; 24 plants were inoculated per plant tray and 12 trays were used in a typical assay . Aphids were killed with 0 . 2% Pirimor G ( http://www . certiseurope . fr ) as described in Martinière et al . ( 2011a ) . Finally , the fraction of symptomatic plants was scored by visual inspection 3 weeks later . FRAP experiments were performed according to Martinière et al . ( 2011b ) on infected or healthy Arabidopsis GFP-TU6 leaves using a Zeiss LSM700 confocal microscope with a 63× NA 1 . 4 oil-immersion objective . Leaf samples were mounted in 1% low melting point agar to prevent focus shift . Twenty scans of the entire field of view were made at pre-bleach intensity , and then a circular 20 μm2 region of interest ( ROI ) , which included a TB , was photobleached . Three iterations of the 488-nm laser at 100% intensity were used for the bleaching . For recovery of the fluorescence in TBs , images were recorded for 110 s , with a 512 × 512 px picture size , a scan speed of 167 ms/frame and a delay between frames of 0 . 2 s . For controls , to account for fast tubulin diffusion in the cytosol , recovery was recorded in identical ROIs for only 5 . 9 s , with a 100 × 50 px picture size , a scan speed of 16 ms/frame and a delay between frames of 23 ms . We verified that the energy of the 488-nm laser used for the post-bleach added no bleaching effect by recording a control region outside the bleaching ROI . The experiment was repeated 21 times for wild type TB bleaching , 23 times for TB-TC bleaching and 18 times in the case of cytoplasmic soluble tubulin . Average intensities in all ROIs including the background signal were measured using ImageJ 1 . 44p software ( http://imagej . nih . gov/ij ) , before exporting data into Microsoft Excel 2007 ( http://www . microsoft . com ) . Fluorescence recovery data was normalized as follows:In= ( ( It−Imin ) / ( Imax−Imin ) ) ×100 , where In is normalized intensity , It is intensity at any time t , Imin is the minimum intensity post bleach and Imax is the mean intensity pre-bleach . Non-linear regression was used to model FRAP data . In this case , a one-phase exponential curve was used:Y ( t ) =A Exp ( −k ) ( t ) +B , where A , B and k are parameters of the curve and t is time . From this curve , the half time of recovery was calculated as t ( 1/2 ) = 0 . 69/k . Finally , t ( 1/2 ) was used to calculate the diffusion rate as D = ( 0 . 88 R2 ) / ( 4 t ( 1/2 ) ) , where D is the diffusion rate and R is the radius of the bleaching area . A pulled glass microelectrode was fitted on a micromanipulator ( http://www . prioruk . com ) and placed either on the stage of a Leica confocal LSI macroscope equipped with a 0 . 56–16× zoom and a 5× objective or a Zeiss LSM700 confocal microscope with a 10× objective . Either a whole potted plant ( visualized with the macroscope ) or a detached leaf taped to a slide and with a water-soaked paper wrapped around its stalk ( visualized with the microscope ) was placed under the objective . The microelectrode was carefully brought up to the leaf , and the epidermis was touched or pierced . The approach , the mechanical stress and the reaction of the plant leaf epidermis cell were all recorded by time lapse fluorescence microscopy using acquisition settings as described above for the LSM700 microscope , or excitation with a 488-nm diode laser and an emission bandwidth from 505–550 nm for the LSI macroscope . The pinholes were opened to record sections approximately with 20-μm thickness , and the microscope settings were selected for minimal acquisition times at the expense of image quality . For FRAP , D values were compared with a two-tailed t-test . TB activation states close to and distant from salivary sheaths , as well as transmission rates , were analyzed using GLM and hierarchical GLM models with a binomial distribution . For P2-TC transmission experiments , the Mann–Whitney test was used . To test for differences in TB states between wild type-infected and P2-TC-infected tissue , a nominal logistic model was used since three parameters were analyzed . Statistical analyses were carried out using JMP 10 ( http://www . jmp . com ) , R 2 . 9 . 2 ( http://www . r-project . org ) and Vassarstats ( http://vassarstats . net/ ) software . The p values <0 . 05 were regarded as statistically significant . | Viruses are infectious agents that can replicate only inside a living host cell . When a virus infects an animal or plant , it introduces its own genetic material and tricks the host cells into producing viral proteins that can be used to assemble new viruses . An essential step in the life cycle of any virus is transmission to a new host: understanding this process can be crucial in the fight against viral epidemics . Many viruses use living organisms , or vectors , to move between hosts . In the case of plant viruses such as cauliflower mosaic virus , the vectors are often aphids . When an aphid sucks sap out of a leaf , virus particles already present in the leaf become attached to its mouth , and these viruses can be transferred to the next plant that the insect feeds on . However , in order for cauliflower mosaic virus particles to become attached to the aphid , structures called transmission bodies must form beforehand in the infected plant cells . These structures are known to contain helper proteins that bind the viruses to the mouth of the aphid , but the precise role of the transmission body has remained obscure . Now Martinière et al . show that the transmission body is in fact a dynamic structure that reacts to the presence of aphids and , in so doing , boosts the efficiency of viral transmission . In particular , they show that the action of an aphid feeding on an infected leaf triggers a rapid and massive influx of a protein called tubulin into the transmission body . The transmission body then bursts open , dispersing helper protein-virus particle complexes throughout the cell , where they become more accessible to aphids . This series of events increases viral transmission rates twofold to threefold . The results show that a virus can detect insect vectors , likely by using the sensory system of its host , and trigger a response that boosts viral uptake and thus transmission . This is a novel concept in virology . It will be important to discover whether similar mechanisms are used by other viruses , including those that infect animals and humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2013 | A virus responds instantly to the presence of the vector on the host and forms transmission morphs |
Sophisticated tool use is a defining characteristic of the primate species but how is it supported by the brain , particularly the human brain ? Here we show , using functional MRI and pattern classification methods , that tool use is subserved by multiple distributed action-centred neural representations that are both shared with and distinct from those of the hand . In areas of frontoparietal cortex we found a common representation for planned hand- and tool-related actions . In contrast , in parietal and occipitotemporal regions implicated in hand actions and body perception we found that coding remained selectively linked to upcoming actions of the hand whereas in parietal and occipitotemporal regions implicated in tool-related processing the coding remained selectively linked to upcoming actions of the tool . The highly specialized and hierarchical nature of this coding suggests that hand- and tool-related actions are represented separately at earlier levels of sensorimotor processing before becoming integrated in frontoparietal cortex .
Tool use , whether using a stone , stick , rake , or pliers , provides an extension of the body ( Van Lawick-Goodall , 1970 ) and involves , among other things , the transfer of a proximal movement goal for the hand into a more distal goal for the tool ( Johnson and Grafton , 2003; Arbib et al . , 2009 ) . A compelling demonstration that this transfer might actually occur at the cortical level comes from neural recordings of grasping neurons in the ventral premotor cortex ( PMv ) and motor cortex ( M1 ) of macaque monkeys trained to use pliers ( Umilta et al . , 2008 ) . In both these areas , many neurons that encoded the specifics of hand grasping subsequently encoded tool grasping , even when use of the specific tool ( reverse pliers that close as the hand grip opens ) required hand kinematics opposite to those required when grasping with the hand alone . These findings suggest that tool use is supported by an effector-independent level of representation , in which the overall goal of the motor act is coded separately from the precise hand kinematics required to operate the tool . In further support of this notion , findings from human neuropsychology ( Berti and Frassinetti , 2000; Maravita and Iriki , 2004 ) , human behavior ( Gentilucci et al . , 2004; Cardinali et al . , 2009 , 2012 ) , and macaque monkey neurophysiology ( Iriki et al . , 1996 ) suggest that following training , a tool may actually become incorporated into the body schema of the actor and coded as an extension of the hand/limb . While provocative , how well does this single mechanism explain the neural substrates of tool use in humans , particularly within established networks that have been identified for tools ( Lewis , 2006 ) , hand actions ( Culham et al . , 2006 ) , and body perception ( Peelen and Downing , 2007 ) ? Although considerable research has been done on the brain networks specialized for visual processing of tools and bodies and the visual-motor processing of hand actions , these topics have largely been studied in isolation . Increasing evidence from functional magnetic resonance imaging ( fMRI ) suggests that human frontoparietal and occipitotemporal cortex contain specialized regions that selectively represent tools and bodies ( Downing et al . , 2001; Lewis , 2006; Frey , 2007; Peelen and Downing , 2007; Peeters et al . , 2009; Valyear and Culham , 2010; Bracci et al . , 2012 ) . For instance , when individuals view , imagine , or pantomime tool use actions , the supramarginal gyrus ( SMG ) , posterior middle temporal gyrus ( pMTG ) , and dorsal premotor cortex ( PMd ) —areas that have shown some of the greatest evolutionary expansion in humans ( Van Essen and Dierker , 2007 ) —are often co-activated ( Lewis , 2006; Frey , 2007 ) . fMRI studies further suggest that human occipitotemporal cortex also contains body-selective regions for perception , such as the extrastriate body area ( EBA ) , which preferentially respond to viewing of the body and its parts ( Astafiev et al . , 2004; David et al . , 2007; Peelen and Downing , 2007 ) . The frontoparietal regions activated by tools are spatially close to ( and perhaps overlapping with ) brain areas implicated in hand actions , particularly the grip component of reach-to-grasp actions . Specifically , SMG lies very near the grasp-selective anterior intraparietal sulcus ( aIPS , Chao and Martin , 2000; Valyear et al . , 2007 ) and PMd shows grasp-selective as well as tool-selective responses ( Grezes and Decety , 2002; Gallivan et al . , 2011 ) . In addition , real hand actions activate other frontoparietal regions including the superior parieto-occipital cortex ( SPOC ) region , PMd , and additional areas along the IPS ( Culham et al . , 2006; Filimon , 2010 ) , but the specific role of these areas in tool use remains unexplored . Moreover , almost all of the human neuroimaging studies of tools to date have used proxies for real tool use ( reviewed in Lewis , 2006 ) , including visual stimuli such as images or movies ( e . g . , Beauchamp et al . , 2002 ) , semantic tasks ( e . g . , Martin et al . , 1995 ) , or simulated tool actions like pantomiming , imitating or imagining tool use ( e . g . , Johnson-Frey et al . , 2005; Rumiati et al . , 2004 ) or making perceptual judgments about how one would use a tool ( e . g . , Jacobs et al . , 2010 ) . It remains unclear whether the highly specialized brain areas within these tool- , body- , and action-related networks in humans also play important roles in planning real movements with a tool or with the body ( hand ) alone . The purpose of the current study was to examine exactly how and where in the human brain tool-specific , hand-specific , and effector-independent ( shared hand and tool ) representations are coded . To this aim we used fMRI to examine neural activity while human subjects performed a delayed-movement task that required grasp or reach actions towards a single target object . Critically , subjects performed these two different movements using either their hand or reverse tongs , which required opposite operating kinematics compared to when the hand was used alone . This manipulation allowed us to maintain a common set of actions throughout the experiment ( grasping vs reaching ) while at the same time varying the movement kinematics required to achieve those actions ( i . e . , depending on whether the hand vs tool effector was used ) . Using multi-voxel pattern analysis ( MVPA ) to decode preparatory ( pre-movement ) signals , we then probed exactly where in frontoparietal cortex and in tool- and body-selective areas in occipitotemporal cortex movement plans ( grasping vs reaching ) for the hand and tool were distinct ( effector-specific ) vs where signals related to upcoming actions of the hand could be used to predict the same actions performed with the tool ( effector-independent ) . Consistent with an effector-specific coding of hand- and tool-related movements we found that preparatory signals in SPOC and EBA differentiated upcoming movements of the hand only ( i . e . , hand-specific ) whereas in SMG and pMTG they discriminated upcoming movements of the tool only ( i . e . , tool-specific ) . In addition , in anterior parietal regions ( e . g . , aIPS ) and motor cortex we found that pre-movement activity patterns discriminated planned actions of ‘both’ the hand and tool but , importantly , could not be used to predict upcoming actions of the other effector . Instead , we found that this effector-independent type of coding was constrained to the preparatory signals of a subset of frontoparietal areas ( posterior IPS and premotor cortex ) , suggesting that in these regions neural representations are more tightly linked to the goal of the action ( grasping vs reaching ) rather than the specific hand movements required to implement those goals .
To pinpoint when predictive movement information was available in the spatial voxel patterns , we ran a single-trial decoding analysis for each point in time over the course of a trial ( Soon et al . , 2008; Harrison and Tong , 2009 ) . This time-resolved decoding analysis revealed a full range of decoding profiles during movement planning across the network of specified regions . For instance , preparatory voxel patterns in SPOC and EBA accurately predicted upcoming grasping vs reaching actions with the hand only whereas preparatory voxel patterns in SMG and pMTG successfully predicted grasping vs reaching actions with the tool only ( Figures 3 and 6 , red and blue decoding traces ) . Notably , in nearly all the remaining regions , we were able to successfully use the activity patterns to predict the action performed ( grasping vs reaching ) for both the hand and tool effector ( Figures 3–5 , red and blue decoding traces; purple traces will be discussed in the next section entitled ‘Separate and shared representations for the hand and tool’ ) . For instance , in parietal cortex , preparatory activity in pIPS , midIPS , post . aIPS , aIPS , and t-aIPS could be used to accurately discriminate which object-directed hand or tool movement was to be performed moments later ( for overlap between t-aIPS and the post . aIPS and aIPS regions , see Figure 5 ) . Likewise , in frontal cortex , predictive movement activity for hand and tool actions was also found in motor cortex , PMd , and PMv . Importantly , consistent with expectations and previous investigations ( Gallivan et al . , 2011a , 2011b ) , our sensory control region , SS-cortex , failed to decode any planned movements and only discriminated the different actions upon execution ( Figure 3 ) . To verify the observations obtained from the time-resolved decoding analysis , we also averaged the spatial activity patterns generated over a 4-s ( 2-volume ) window of time immediately prior to the cue for subjects to perform the movement ( denoted by gray shaded bars in Figures 3–6 ) . In line with our previous fMRI investigations ( Gallivan et al . , 2011a , 2011b ) , this plan-epoch decoding approach supports the notion that discriminatory predictive signals for movement can arise moments prior to action execution . 10 . 7554/eLife . 00425 . 006Figure 3 . Separate movement plans for the hand and tool decoded from frontoparietal cortex . Decoding accuracies are shown for each time point in the trial ( time-resolved decoding ) and for the Plan-epoch only , the latter based on a windowed average of the spatial activity patterns denoted by the gray shaded bars in the time-resolved decoding plots . In the time-resolved decoding plots , vertical lines correspond to the onset of the Plan and Execute phases of each trial ( from left to right ) . For decoding accuracies discriminating grasp vs reach actions with the Hand ( in red ) and Tool ( in blue ) classifier training and testing was done using a single trial N-1 cross-validation approach . Across-effector decoding accuracies ( in purple ) were computed using all the available data and from training classifiers on Hand-G vs Hand-R trials and testing on Tool-G vs Tool-R trials and then averaging these values with the opposite train-and-test ordering , within each subject . ( A ) Areas of frontoparietal cortex that could decode movement plans with the hand and/or with the tool but not between hand and tool ( i . e . , no Across-effector decoding ) . ( B ) Decoding accuracies from the sensory control region , SS-cortex . Note that SS-cortex significantly decodes movements only following action onset ( and not during planning ) . Error bars represent standard error of the mean ( SEM ) across subjects . Solid black horizontal lines are chance accuracy level ( 50% ) . Asterisks assess statistical significance with two-tailed t-tests across subjects with respect to 50% . Four-pointed stars assess statistical significance based on a false discovery rate ( FDR ) correction of q ≤ 0 . 05 . Note also that in the time-resolved decoding plots , the color of each asterisk/star denotes which specific pair-wise discrimination is significant at each point in time . G: grasp; R: reach . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 00610 . 7554/eLife . 00425 . 007Figure 3—figure supplement 1 . Classifier decoding accuracies in non-brain control regions . ( Left ) Non-brain control ROIs defined in each subject ( denoted in light yellow; example subject shown ) . ( Right ) Linked to each ROI is the % SC time-course activity and the time-resolved and plan-epoch decoding accuracies ( computed and plotted the same as in Figure 3 ) . Error bars represent standard error of the mean ( SEM ) across subjects . Solid black lines are chance accuracy level ( 50% ) . Note that no significant differences at any point in the trial were found with respect to 50% chance . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 00710 . 7554/eLife . 00425 . 008Figure 4 . Shared movement plans for the hand and tool decoded from frontoparietal cortex . Decoding accuracies are plotted and computed the same as in Figure 3 . Significant across-effector decoding ( purple traces ) shows where and when the movement action ( Grasp vs Reach ) is being represented with some invariance to the acting effector ( Hand vs Tool ) . See Figure 3 caption for format . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 00810 . 7554/eLife . 00425 . 009Figure 4—figure supplement 1 . Time-resolved and plan-epoch decoding accuracies for across-effector classification , separated according to the direction of classifier training and testing . Pink traces and bars denote accuracies that were computed by training classifiers to discriminate hand trials ( Hand-G vs Hand-R ) and testing on tool trials ( Tool-G vs Tool-R ) . Light blue traces and bars denote accuracies that were computed by training classifiers to discriminate tool trials and testing on hand trials . As in Figures 3–6 , across-effector accuracies were computed using all the available data . Error bars represent standard error of the mean ( SEM ) across subjects . Solid black horizontal lines are chance accuracy level ( 50% ) . Asterisks assess statistical significance with two-tailed t-tests across subjects with respect to 50% . Four-pointed stars assess statistical significance based on a FDR correction of q ≤ 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 00910 . 7554/eLife . 00425 . 010Figure 4—figure supplement 2 . Voxel weight analyses for the plan-epoch activity in the cross-decoding ROIs ( L-pIPS , L-midIPS , L-PMd , and L-PMv ) , shown for two representative subjects ( in A and B ) . ( Top of A and B ) ROIs ( in yellow ) overlayed on the transverse slices of subjects 1 ( A ) and 2 ( B ) . ( Below ) Voxel weights for the trained SVM classifier for the plan-epoch ( 2 imaging volumes prior to movement initiation ) . Voxel weights are based on the single-shot train iterations using all the available data . Each column of boxes corresponds to one of the two pair-wise comparisons and each row of boxes corresponds to a transverse slice ( 3 mm thickness ) through the ROI ( see expanded box at bottom in ( C ) for legend; voxel size = 3 mm × 3 mm × 3 mm ) . The color of each voxel in each box denotes its relationship ( weight ) with the class label ( as determined by the trained SVM discriminant function; see scale at bottom in ( C ) for voxel weight color coding ) . Positive and negative values ( red and blue colors , respectively ) denote a stronger weighting of a particular voxel towards one planned action vs the other ( red = grasp-selective voxels , blue = reach-selective voxels ) . Gray patches denote the borders of the ROI . Accuracies below each column denotes the test accuracy for that specific pair-wise comparison in the subject ( when averaged across the N train-and-test iterations ) and shown at the very bottom , the test accuracies for the specific cross-decoding case ( based on all the available data ) . The spatial arrangement of grasp- and reach-selective voxels indicates considerable local variability . ( Bottom of A and B ) Voxel weight fingerprints for the 10 most discriminative voxels within a ROI . For each ROI , the raw voxel weights across the two pair-wise comparisons were ordered and the top 10 voxels were selected ( i . e . , the same 10 discriminative voxels are shown in each plot for each ROI ) . Within each pair-wise comparison , voxel weights for the common voxel set were normalized to 1 and their magnitudes were plotted around the polar axis ( each axis of the polar plot represents a single voxel ) . The direction of the voxel weights is encoded by the line color: Positive ( grasp-specific ) voxel weights are plotted in red and the negative ( reach-specific ) weights are plotted in blue , congruent with the ROI voxel weight maps . L: left; R: right; A: anterior; P: posterior; S: superior; I=inferior . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 01010 . 7554/eLife . 00425 . 011Figure 4—figure supplement 3 . Movement instructions decoded from transient ( but not sustained ) responses in auditory cortex . ( Top ) Left auditory cortex activity localized by the same contrast used to identify the frontoparietal sensorimotor network [ ( Plan + Execute ) > 2* ( Preview ) ] . Results calculated across all subjects ( Random Effects GLM ) are displayed on one representative subject's inflated left hemisphere . The general location of Heschl's gyrus is outlined in a blue circle ( actual ROIs were anatomically defined separately in each subject according to stringent anatomical criteria , see main manuscript text ) . ( Below ) % SC time-course activity and the time-resolved and plan-epoch decoding accuracies from left auditory cortex . Error bars represent standard error of the mean ( SEM ) across subjects . Solid black lines are chance accuracy level ( 50% ) . Asterisks assess statistical significance with two-tailed t-tests across subjects with respect to 50% . Four-pointed stars assess statistical significance based on a FDR correction of q ≤ 0 . 05 . Note that above chance auditory cue decoding transiently arises halfway through the Plan-phase ( consistent with a discrimination of the ‘Grasp’ and ‘Touch’ auditory commands delivered to subjects via headphones at the onset of the Plan-phase ) but , importantly , not during the pre-defined plan-epoch ( denoted by gray bar ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 01110 . 7554/eLife . 00425 . 012Figure 5 . Hand and Tool movement plans decoded from the localizer-defined t-aIPS . ( A ) Block-design protocol and experimental timing of the Bodies , Objects , and Tools ( BOT ) localizer . ( B ) Overlay of tool and anterior parietal ROIs . The Motor experiment-defined anterior parietal ROIs ( post . aIPS and aIPS; defined by the [ ( Plan + Execute ) > 2* ( Preview ) ] contrast ) and the Localizer experiment-defined anterior parietal ROI ( t-aIPS; defined by the [ ( Tools > Scrambled ) AND ( Tools > Bodies ) AND ( Tools > Objects ) ] conjunction contrast ) are superimposed on the transverse anatomical slices of three representative subjects . Across all subjects we found a reasonable degree of overlap between the Motor and Localizer experiment-defined anterior parietal ROIs . ( C ) % SC time-course activity and time-resolved and plan-epoch decoding accuracies from t-aIPS . See Figure 3 caption for format . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 01210 . 7554/eLife . 00425 . 013Figure 6 . Tool and hand movement plans decoded from the localizer-defined pMTG and EBA , respectively . ( Top ) The pMTG ( in red ) and EBA ( in green ) are shown in the same three representative subjects as in Figure 5 . pMTG was defined using the conjunction contrast of [ ( Tools > Scrambled ) AND ( Tools > Bodies ) AND ( Tools > Objects ) ] in each subject . EBA was defined using the conjunction contrast of [ ( Bodies > Scrambled ) AND ( Bodies > Tools ) AND ( Bodies > Objects ) ] . ( Below ) % SC time-course activity and time-resolved and plan-epoch decoding accuracies shown for pMTG ( bordered in red ) and EBA ( bordered in green ) . See Figure 3 caption for format . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 013 To ensure our decoding accuracies could not result from spurious factors ( e . g . , task-correlated head or arm movements ) , we ran the same classification analyses in two non-brain ROIs where decoding should not be expected: the right ventricle and outside the brain . Consistent with our previous work ( Gallivan et al . , 2011a , 2011b; Gallivan et al . , 2013 ) , MVPA in these two areas showed no accurate decoding for any phase of the trial ( Figure 3—figure supplement 1 ) . Three general observations can be made based on the results of these decoding analyses . First , predictive movement information , if it is to emerge , generally arises in the two time points prior to initiation of the movement ( although note that in a few brain areas , such as L-pIPS and L-PMd , this information is also available prior to these two time points ) . Second , in support of the notion that this predictive motor information is directly related to the ‘intention’ to make a movement , accurate classification never arises prior to the subject being aware of which action to execute ( i . e . , prior to the auditory instruction delivered at the initiation of the Plan phase ) . Finally , decoding related to the planning of a movement can be fully disentangled from decoding related to movement execution , which generally arises several imaging volumes later . Expanding on these MVPA results—and perhaps more important to the overall interpretations of our findings—we next examined in which brain areas the final action ( grasping vs reaching ) was being represented with some invariance to the effector to be used . To do this , we trained pattern classifiers to discriminate Hand-G vs Hand-R trials and then tested their performance in discriminating Tool-G vs Tool-R trials ( the opposite train-and-test process—train set: Tool-G vs Tool-R → test set: Hand-G vs Hand-R—was also performed , and then we averaged the accuracies from both approaches ) ( for this technique , see also Formisano et al . , 2008; Harrison and Tong , 2009; Gallivan et al . , 2011a ) . If successful , this cross-classification would suggest that the object-directed action plans being decoded are to some extent independent of the acting effector ( at least to the extent that accurate across-effector classification can be achieved ) . When we performed this analysis , we found accurate across-effector classification in four regions during planning: two areas in posterior parietal cortex ( PPC ) , pIPS and midIPS , and two areas in premotor cortex , PMd and PMv ( see purple decoding traces and bars in Figure 4 ) . ( Note that separating these tests , Train set: Hand →Test set: Tool and Train set: Tool → Test set: Hand , revealed no major asymmetries in classification , see Figure 4—figure supplement 1 ) . Importantly , recall that because the object location was changed ( with respect to fixation ) between hand and tool experimental runs coupled with the fact that the reverse tool required operating mechanics opposite from those required when the hand was used alone , accurate across-effector classification cannot be attributed to low-level visual , haptic , or kinematic similarities between hand and tool trials . Furthermore , note that accurate across-effector classification does not simply arise in ‘any’ area where the pattern classifiers are able to successfully discriminate grasp vs reach movements for both the hand and tool . Indeed , although several other areas accurately differentiated the two upcoming movements for both effectors ( e . g . , post . aIPS , aIPS , t-aIPS , and motor cortex ) , the preparatory spatial patterns of activity in these areas did not allow for accurate cross-classification . This finding is in itself notable , as it suggests that these latter areas may contain separate coding schemes for the hand and tool . One obvious interpretation of this result is that these latter areas separately code the kinematics used to operate the hand vs tool , providing a neural instantiation of the effector-specific representations thought to be critical for complex tool use . These findings are summarized in Figure 7 . 10 . 7554/eLife . 00425 . 014Figure 7 . Summary of action plan decoding in the human brain for hand and tool movements . Pattern classification revealed a wide range of activity profiles across motor and sensory cortices within networks implicated in hand actions , tool understanding , and perception . Some regions ( SPOC and EBA ) coded planned actions with the hand but not the tool ( areas in red ) . Some regions ( SMG and MTG ) coded planned actions with the tool but not the hand ( areas in blue ) . Other regions ( aIPS and M1 ) coded planned actions with both effectors ( areas in pink ) but did so using different neural representations . A final set of brain areas ( pIPS , PMd and PMv ) instead coded the final type of action to be performed with invariance as to whether the hand or tool was to be used ( areas in purple ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00425 . 014 We examined whether the qualitative differences in decoding accuracies between the three pairwise comparisons within each region ( i . e . , within-hand decoding , within-tool decoding and across-effector decoding ) reached statistical significance . We reasoned that a brain area involved in coding the hand , for example , might show significantly higher decoding accuracies for actions planned with the hand vs tool . A 13 ( number of ROIs ) × 3 ( number of pairwise comparisons per ROI ) repeated-measures ANOVA ( rmANOVA ) of the plan-epoch decoding accuracies revealed a strong trend towards a significant interaction even in a relatively low-powered omnibus test ( F5 . 701 = 2 . 170 , p=0 . 069 , Greenhouse-Geisser [GG] corrected ) , suggesting differences in the patterns of decoding across regions . Further investigation of the decoding accuracies within each area ( using GG-corrected rmANOVAs and False Discovery Rate [FDR]-corrected follow-up paired sample t-tests ) revealed only a few significant effects: in L-SPOC , decoding accuracies for the hand were significantly higher than for the tool and for across-effector decoding ( both at p<0 . 05; F1 . 538 = 6 . 084 , p=0 . 014 ) ; in L-SMG , decoding accuracies for the tool were significantly higher than for the hand and for across-effector decoding ( both at p<0 . 05; F1 . 959 = 10 . 016 , p<0 . 001 ) ; in L-motor cortex , decoding accuracies for the hand were significantly higher than across-effector decoding ( p<0 . 05; F1 . 398 = 6 . 239 , p=0 . 016 ) ; and lastly , in L-pMTG , decoding accuracies for the tool were significantly higher than for the hand ( p=0 . 049; F1 . 968 = 4 . 171 , p=0 . 037 ) ( note that in L-EBA , although decoding accuracies for the hand showed a trend to be higher than for the tool , this did not reach significance; p=0 . 106; F1 . 370 = 3 . 635 , p=0 . 078 ) . Taken together , these analyses suggest tool-specific decoding in SMG and pMTG and hand-specific decoding in SPOC and EBA . To further examine the underlying patterns of activity that led to accurate decoding and cross-decoding we investigated the voxel weights assigned by the classifier ( where the direction of the weight indicates the relationship of the voxel with the class label , as learned by the classifier; see also the caption for Figure 4—figure supplement 2 ) . In particular , we looked for correspondence in the voxel weights across pair-wise comparisons within single subjects as a potential explanation for why the spatial activity patterns in certain areas might show across-effector decoding ( the data from two representative subjects is shown in Figure 4—figure supplement 2; see also Formisano et al . , 2008 for a similar approach ) . That is , if the exact same population of voxels were responsible for driving the observed across-effector classification effects than this same voxel set might be consistently biased towards coding one type of action vs the other ( i . e . , grasping or reaching ) for both effectors ( hand and tool ) . ( Note that because our pattern classification analysis was performed on non-Talairached data [MVPA was in fact performed on single-subject ACPC-aligned data] , comparing the weights across subjects on a single cortical surface was not feasible ) . Visual inspection of the voxel weightings failed to reveal any structured or consistent topography within or across subjects ( for similar results , see also Harrison and Tong , 2009; Gallivan et al . , 2011a ) . That is , while the weightings of some voxels appeared to be consistent across pair-wise comparisons within a subject , others appeared to shift their weighting depending on the effector to be used in the movement . ( Note that the only consistency observed was that voxels coding for one particular type of action [as indicated by the positive or negative direction of the weight] tended to spatially cluster [which is sensible given the spatial blurring of the hemodynamic response; see Gallivan et al . , 2011a for a further discussion of this issue] ) . One possible explanation for the anisotropies observed in the voxel weight distributions across pair-wise comparisons is that they relate to the fact that the decoding accuracies reported here , while statistically significant , are generally quite low ( means across participants ∼55% ) . This indicates some appreciable level of noise in the measured planning-related signals , which , given the highly cognitive nature of planning and related processes , likely reflects a wide range of endogenous factors that can vary throughout the course of an entire experiment ( e . g . , focus , motivation , mood , etc . ) . Indeed , even when considering the planning-related activity of several frontoparietal structures at the single-neuron level , responses from trial to trial can show considerable variability ( e . g . , Snyder et al . , 1997; Hoshi and Tanji , 2006 ) . When extrapolating these neurophysiological characteristics to the far coarser spatial resolution measured with fMRI , it is therefore perhaps to be expected that this type of variability should also be reflected in the decoding accuracies generated from single-trial classification . With regards to the resulting voxel weights assigned by the trained SVM pattern classifiers , it should be noted that even in cases where brain decoding is quite robust ( e . g . , ∼80% for orientation gratings in V1–V4 ) , the spatial arrangement of voxel weights still tends to show considerable local variability both within and across subjects ( e . g . , Kamitani and Tong , 2005; Harrison and Tong , 2009 ) . One alternative explanation to account for the accurate across-effector classification findings reported may be that our frontoparietal cortex results arise not because of the coding of effector-invariant movement goals ( grasp vs reach actions ) but instead simply because grasp vs reach movements for both hand and tool trials are cued according to the same ‘Grasp’ and ‘Reach’ auditory instructions . In other words , the cross-decoding observed in PPC and premotor cortex regions might only reflect the selective processing of the auditory commands common to Hand-G and Tool-G ( ‘Grasp’ ) and Hand-R and Tool-R ( ‘Touch’ ) trials and actually have nothing to do with the mutual upcoming goals of the object-directed movement . If this were the case , then we would expect to observe significant across-effector classification in primary auditory cortex ( Heschl’s gyrus ) for the same time-points as that found for PPC ( pIPS and midIPS ) and premotor ( PMd and PMv ) cortex . We directly tested for this possibility in our data by separately localizing left Heschl’s gyrus in each subject with the same contrast used to define the sensorimotor frontoparietal network , [Plan & Execute > 2*Preview] ( recall that auditory cues initiate the onset of the Plan and Execute phases of the trial and so this was a robust contrast for localizing primary auditory cortex ) . We found that although accurate across-effector classification does indeed arise in Heschl’s gyrus during the trial , it does so distinctly earlier in the Plan-phase compared to that of the frontoparietal areas ( Figure 4—figure supplement 3 ) . This observation is consistent with the noticeably transient percentage signal change response that accompanies the auditory instructions delivered to participants at the beginning of the Plan-phase ( see time-course in Figure 4—figure supplement 3 ) , as compared to the more sustained planning-related responses that emerge throughout the entire frontoparietal network ( Figure 2 ) . The temporal disconnect between the cross-decoding found in Heschl’s gyrus ( which emerges in the fourth volume of the Plan-phase ) and frontoparietal cortex ( which generally emerges in the fifth-sixth volumes of the Plan-phase ) makes it unlikely that the effector-invariant nature of the responses revealed in PPC and premotor cortex can be fully attributable to simple auditory commonalities in the planning cues . It is worth emphasizing that while accurate decoding in a region points to underlying differences in the neural representations associated with different experimental conditions ( e . g . , for reviews see Haynes and Rees , 2006; Kriegeskorte , 2011; Naselaris et al . , 2011; Norman et al . , 2006 ) , a lack of decoding or ‘null effect’ ( i . e . , 50% chance classification ) can either reflect that the region 1 ) is not recruited for the conditions being compared , 2 ) contains neural/pattern differences between the conditions but which cannot be discriminated by the pattern classification algorithm employed ( i . e . , a limit of methodology , see Pereira et al . , 2009; Pereira and Botvinick , 2011 ) , or 3 ) is similarly ( but non-discriminately ) engaged in those conditions . With respect to the first possibility , given that we selected frontoparietal cortex ROIs based on their involvement in the motor task at the single-subject level ( using the contrast of [Plan & Execute > Preview] across all conditions ) , it is reasonable to assume that all the localized areas are in some way engaged in movement generation . ( Note that this general assumption is confirmed by the higher-than-baseline levels of activity observed in the signal amplitude responses during the Plan- and Execute-phases of the trial in areas of frontoparietal cortex [Figures 2 and 5] and that this even appears to be the case in the independently localizer-defined lateral occipitotemporal areas , EBA and pMTG [Figure 6] ) . Although it is understandably difficult to rule out the second possibility ( i . e . , that voxel pattern differences exist but are not detected with the SVM classifiers ) , it is worth noting that we do in fact observe null-effects with the classifiers in several regions where they are to be expected . For instance , SS-cortex is widely considered to be a lower-level sensory structure and thus anticipated to only show discrimination related to the motor task once the hand’s mechanoreceptors have been stimulated at object contact ( either through the hand directly or through the tool , indirectly ) . Accordingly , here we find that SS-cortex activity only discriminates between grasp vs reach movements following movement onset ( i . e . , during the Execute phase of the trial ) . Likewise , in motor cortex we show decoding for upcoming hand- and tool-related actions but , importantly , find no resulting across-effector classification . This latter result is highly consistent with the coding of differences in the hand kinematics required to operate the tool vs hand alone and accords with the presumed role of motor cortex in generating muscle-related activity ( Kalaska , 2009; Churchland et al . , 2012; Lillicrap and Scott , 2013 ) . These findings in SS-cortex and motor cortex , when combined with the wide-range of decoding profiles found in other areas ( i . e . , from the hand-selective activity patterns in SPOC and EBA at one extreme , to the tool-selective activity patterns in SMG and pMTG at the other , for summary see Figure 7 ) , suggest that the failure of some areas to decode information related to either hand- or tool-related trials ( but not those of the other effector ) is closely linked to an invariance in the representations of those particular conditions . ( To the extent that in cases where the activity of an area fails to discriminate between experimental conditions it can be said that the area is therefore not involved in coding [or invariant to] those particular conditions , we further expand upon interpretations related to these types of null effects in the ‘Discussion’ section . )
Hierarchical theories of motor control have existed for more than a century ( Jackson , 1889; Sherrington , 1906; Hebb , 1949 ) , distinguishing between the various levels of abstraction required for action planning—for example , at the level of muscles , joints , motor kinematics , and movement goals . The present findings provide insights into where different brain regions might be situated within such a hierarchy . For instance , at some lower level along this hierarchy we likely have hand-selective regions like SPOC and EBA and tool-selective regions like SMG and pMTG . Although typically associated with visual-perceptual processing , EBA , like SPOC , has been implicated in coding movements of the hand/arm ( Astafiev et al . , 2004; Orlov et al . , 2010 , although see Peelen and Downing , 2005 ) and the fact that we were unable to decode tool movement plans from these regions suggests that they fail to incorporate tools into the body schema ( see also Gallivan et al . , 2009 ) . SMG and pMTG , in contrast , are typically activated when human subjects view ( Lewis , 2006; Peeters et al . , 2009 ) or pantomime ( Johnson-Frey et al . , 2005 ) tool-related actions , and damage to these areas creates difficulty in pantomiming or performing tool use actions ( Haaland et al . , 2000 ) . That planning-related signals in SMG and pMTG are able to ‘predict’ real tool actions , as shown here , provides an important extension of these previous findings , demonstrating that these areas also play an important and selective role in generating object-directed tool actions . We also found several parietal and frontal brain regions ( post . aIPS , aIPS , t-aIPS and motor cortex ) that , although able to predict upcoming grasp vs reach movements with both the hand and the tool , did not generalize across the effector ( i . e . , no across-effector classification ) . When considering the particular tool used here—where the operating mechanics of the tool were opposite to those of the hand alone—this effector-specific level of action planning is imperative . It provides a coding for the kinematic properties and/or dynamics associated with each effector ( Umilta et al . , 2008; Jacobs et al . , 2010 ) as well as the other low-level differences that exist between hand and tool trials ( e . g . , spatial location of target ) . These features match the known properties of motor cortex; it provides the largest source of descending motor commands to the spinal neurons that produce hand kinematics ( Porter and Lemon , 1993 ) and correspondingly , much of its activity can be accounted for in muscle control terms ( see Kalaska , 2009; Churchland et al . , 2012; Lillicrap and Scott , 2013 ) . In parietal cortex , aIPS has been strongly implicated in grasp planning and execution ( e . g . , Murata et al . , 2000; Culham et al . , 2003 ) . Notably , it has also been implicated in tool use ( Gallivan et al . , 2009; Jacobs et al . , 2010 ) , but to date , its precise role in tool-related behaviour has remained unclear . The current findings provide two important clarifications with respect to this previous work . First , the anterior IPS is recruited in the planning of tool actions in addition to those of the hand , suggestive of an important role in preparing actions with both effectors . Second , this pattern of findings on its own does not demonstrate that hand and tool actions rely on the same underlying representations , as previously interpreted ( e . g . , Rijntjes et al . , 1999; Castiello et al . , 2000 ) . Rather , as indicated by our cross-classification findings , the representations may differ , perhaps depending on the specifics of the kinematics or object-effector interactions . At higher-levels within this hierarchy , we also found several areas ( pIPS , midIPS , PMd and PMv ) that not only discriminated movement plans for the hand and tool , but moreover , did so using a shared neural code . In the human and macaque monkey , the posterior IPS appears to serve a variety of high-level visual-motor- and cognitive-related functions , such as integrating target- and effector-related information for movement ( Andersen and Buneo , 2002 ) and encoding 3D features of objects for hand actions ( Sakata et al . , 1998 ) . One possibility , in line with this previous work , is that effector-independent responses in these areas emerge due to a common coding of object features that are more relevant for grasping than reaching . That is , the same set of object features pertinent for grasping with the hand ( object contact points , orientation , distribution of mass , etc . ) are pertinent for grasping with the tool and a coding of these features may explain why pattern classifiers trained on hand trials can decode actions performed on tool trials ( and vice versa ) . We also found evidence for these same types of effector-independent representations in premotor areas , PMd and PMv . Each area is engaged in hand actions in both the monkey ( Rizzolatti and Luppino , 2001; Raos et al . , 2004 , 2006 ) and human ( Davare et al . , 2006; Gallivan et al . , 2011 ) and their implication in higher-level goal-related processing ( Rizzolatti and Luppino , 2001; Cisek et al . , 2003 ) , particularly in the case of tool use with PMv ( Umilta et al . , 2008 ) , strongly resonates with the findings reported here . The focus of the present work was to reveal , at the level of the actor , how tool use is planned and implemented in the human brain . In addition to providing insights into how action-centred behavior is cortically represented ( discussed above ) these findings offer a new lens through which to view findings reported from previous observation-based fMRI studies . To date , nearly all fMRI studies examining action-centred coding have done so by adopting tasks that require the observation of others’ actions ( Lewis , 2006; Grafton and Hamilton , 2007; Peeters et al . , 2009; Valyear and Culham , 2010 ) , in which most commonly , 2D static images or movies of action-related behaviors or tool use are passively viewed by participants ( Lewis , 2006; Grafton and Hamilton , 2007; Peeters et al . , 2009; Valyear and Culham , 2010 ) . Notably , the aim of many of these previous investigations has not necessarily been to reveal how the brain plans and executes different actions per se , but instead , to reveal how the brain understands the goals and intentions of an observed actor . This particular line of research has been primarily motivated by the discovery of ‘mirror-neurons’ in the monkey ( Rizzolatti et al . , 2001 ) , located in inferior parietal and ventral premotor cortex ( Fogassi et al . , 2005; Umilta et al . , 2008 ) , which discharge both when the monkey performs a motor act and when the monkey views the same act performed by another individual . When embedded within this larger context , however , it becomes important to not just understand how the actions of other individuals are represented but also how these perceptual representations may relate to the coding of self-generated motor actions . With respect to the latter , past fMRI research has largely left open the question of how goal-directed movements , particularly in the case of tool use , are cortically represented . Here we provide compelling evidence for a strong coupling between the categorical-selectivity of a brain region , as defined through visual-perceptual processing , and its specific role in behavior , as defined through movement planning . For instance , in occipitotemporal cortex we found that the preparatory activity in the independently localized body-selective EBA and the tool-selective pMTG decoded movement plans for hand and tool actions , respectively . This indicates that , similar to the highly modular nature of visual-perceptual processing in occipitotemporal cortex ( Downing et al . , 2006; Kanwisher , 2010 ) , hand- and tool-related actions at certain cortical processing levels may also recruit distinct neural populations . As an interesting departure from these occipitotemporal cortex results , we found that we could decode upcoming movements for ‘both’ the hand and tool from the independently defined tool-selective t-aIPS . At the functional level , the decoding of tool actions in t-aIPS is entirely congruent with its activation in observation-based tool-related tasks ( for reviews , see Lewis , 2006; Frey , 2007 ) and , at the anatomical level , the decoding of hand actions in t-aIPS accords with its close proximity to parietal areas involved in hand preshaping and manipulation ( Culham et al . , 2003; Valyear et al . , 2007; Gallivan et al . , 2011 ) . When compared to the findings in occipitotemporal cortex , this result indicates that hand and tool movement planning may only begin recruiting similar neural structures at the level of parietal cortex . The decoding of planned hand- and tool-related actions in EBA and pMTG , respectively , raises important questions as to what exactly is being represented in these two occipitotemporal cortex regions . Although others have shown that hand/arm movements can activate different regions in occipitotemporal cortex ( Astafiev et al . , 2004; Filimon et al . , 2009; Cavina-Pratesi et al . , 2010; Oosterhof et al . , 2010; Orlov et al . , 2010 ) , here we demonstrate that these signals reflect the ‘intention’ to perform a motor act rather than the sensory feedback responses ( visual , proprioceptive , tactile ) that accompany it . This distinction is important because it indicates that these occipitotemporal cortex areas may play a significant role in action planning and control , possibly by predicting the sensory consequences of actions/movement even before those consequences unfold . Given the delay of incoming sensory signals , this type of forward-state estimation is featured prominently in models of action control ( Wolpert and Ghahramani , 2000; Wolpert and Flanagan , 2001 ) and , from the standpoint of perception , predicting the sensory consequences of movement can be used to disambiguate movements of the body ( self ) vs movements of the world ( others ) ( von Helmohltz , 1866 ) . The current findings would suggest that such forward-state estimations , at least at the level of occipitotemporal cortex , remain linked to the type of effector ( hand vs tool ) to be used in an upcoming movement . Updating the considerably simpler notion that action planning , particularly in the case of tool use , merely involves ‘access’ to ventral stream resources ( Milner and Goodale , 1995; Valyear and Culham , 2010 ) , these findings show that hand- and tool-related action plans can actually be decoded from preparatory signals in body- and tool-selective occipitotemporal cortex areas . In addition to suggesting a role for OTC in visual-motor planning , these findings might also shed light on the organizing principles of the ventral visual stream . Several theories have been proposed to account for the categorical-selectivity of responses throughout OTC ( e . g . , for faces , scenes , bodies , tools , etc ) , with the majority arguing that this modular arrangement arises due to similarities/differences in the visual structure of the world and/or how it is experienced ( Kanwisher , 2010 ) . For example , according to one prominent view , faces and scenes activate different regions of OTC due to underlying visual field preferences ( i . e . , faces activate areas with stronger foveal representations , like FFA , whereas scenes activate areas with stronger peripheral representations , like PPA; Levy et al . , 2001 ) . According to another well-known view , it is instead similarity in visual shape/form that is mapped onto ventral temporal cortex ( Haxby et al . , 2001 ) . One particularly compelling alternative view , however , argues that the organization of OTC may be largely invariant to bottom-up visual properties and that it instead emerges as a by-product of the distinct connectivity patterns of OTC areas with the rest of the brain , particularly the downstream motor structures that use the visual information processed in OTC to plan movements of the body ( Mahon et al . , 2007; Mahon and Caramazza , 2009 , 2011 ) . Under this view , the neural specificity frequently observed for the visual presentation of body parts and/or tools in particular regions of OTC may reflect , to a certain extent , their anatomical connectivity with frontoparietal areas involved in generating movements of the body and/or interacting with and manipulating tools , respectively—a notion that garners some empirical support from the ‘downstream’ functional connectivity patterns of areas involved in body part- and tool-related processing ( Mahon et al . , 2007; Bracci et al . , 2012 ) . Assuming the sharing of action-related information within functionally interconnected circuits , this conceptual framework might help explain the matching object-selective and planning-related responses observed here within both EBA and pMTG . This compatibility of visual- and motor-related responses within single brain areas resonates with neurophysiological findings in macaque parietal cortex showing that the visual-response selectivity of neurons in AIP ( for size , shape , orientation , etc . ) are often matched to their motor-response selectivity during action ( e . g . , Murata et al . , 2000 ) . This coupling is thought to mediate the transformation of visual information regarding physical object properties into corresponding motor programs for grasping or use ( Jeannerod et al . , 1995; Rizzolatti and Luppino , 2001 ) and resonates with the broader concept of motor affordances , whereby the properties of objects linked to action are automatically represented in movement-related areas of the brain ( Cisek , 2007; Cisek and Kalaska , 2010 ) . Where exactly the current findings fit within the context of these broader frameworks remains unclear , nevertheless , our results provide novel evidence suggesting that the specificity of visual object categorical responses in OTC are in some way linked to a specific role in preparing related motor behaviors .
Thirteen right-handed volunteers participated in the Motor experiment ( seven females; mean age: 25 . 7 years , age range: 20–33 years ) and were recruited from the University of Western Ontario ( London , Ontario , Canada ) . Eight of these same participants ( four females ) participated in a second Localizer experiment . All subjects had normal or corrected-to-normal vision and were financially compensated for their participation . Informed consent and consent to publish was obtained in accordance with ethical standards set out by the Declaration of Helsinki ( 1964 ) and with procedures approved by the University of Western Ontario’s Health Sciences Research Ethics Board ( ethics review number: 13507 ) . Subjects were naive with respect to hypothesis testing . To extract the visual-motor planning response for the hand and tool from the simple visual and motor execution responses , we used a slow event-related planning paradigm with 34 s trials , each consisting of three distinct phases: ‘Preview’ , ‘Plan’ and ‘Execute’ ( Figure 1C ) . We adapted this paradigm from previous fMRI work with eye- and arm-movements that have successfully isolated delay period activity from the transient neural responses following the onset of visual input and movement execution ( Curtis et al . , 2004; Beurze et al . , 2007 , 2009; Pertzov et al . , 2011 ) and from other previous studies from our lab in which we successfully used the spatial voxel patterns of delay period responses in order to show that different upcoming movements can be accurately predicted ( Gallivan et al . , 2011a; 2011b ) . In our task , each trial began with the Preview phase , where the subject’s workspace was illuminated revealing the centrally located target object . After 6 s of the Preview phase , subjects were given an auditory cue ( 0 . 5 s ) , either ‘Grasp’ or ‘Touch’ , informing them of the upcoming movement required; this cue marked the onset of the Plan phase . Although there were no visual differences between the Preview and Plan phase portions of the trial ( i . e . , the single object was always visually present ) , only in the Plan phase did participants have the necessary motor information in order to prepare the upcoming movement . After 12 s of the Plan phase , a 0 . 5-s auditory beep cued participants to immediately execute the planned action , initiating the Execute phase of the trial . 2 s following the beginning of this Go cue , the illuminator was turned off , providing the cue for subjects ( during both hand and tool runs ) to return the hand to its peripheral starting position . After the illuminator was extinguished , subjects then waited in the dark while maintaining fixation for 14 s , allowing the BOLD response to return to baseline prior to the next trial ( ITI phase ) . The two trial types ( grasp or reach ) , with ten repetitions per condition ( 20 trials total ) were randomized within a run and balanced across all runs ( that required the same effector ) so that each trial type was preceded and followed equally often by every other trial type across the entire experiment . Separate practice sessions were carried out before the actual experiment to familiarize participants with both the mechanics of the reverse tool and the timing of the paradigm , where in particular , the delay timing required the cued action to be performed only at the beep ( Go ) cue . These sessions were carried out before the subjects entered the scanner as well as during the anatomical scan ( collected at the beginning of every experiment ) . A testing session for each participant included set-up time ( ∼45 min ) , 8 functional runs ( although two subjects participated in 6 and 10 functional runs , respectively ) and 1 anatomical scan , and lasted approximately 3 hr . Throughout the experiment , the subject’s fixation and hand movements were monitored using an MR-compatible infrared-sensitive camera optimally positioned directly below the fixation point and facing towards the subject . The videos captured during the experiment were analyzed off-line to verify that the subjects were indeed performing the task as instructed . A more rigorous tracking of the eyes was not performed because our eye-tracking system does not work while the head is tilted due to a partial occlusion from the eyelids . | The use of tools is a key characteristic of primates . Chimpanzees—our closest living relatives—use sticks to probe for termites as well as stones to crack open nuts , and have even been seen using specially sharpened sticks as spear-like tools for hunting . However , despite its importance in human evolution , relatively little is known about how tool use is supported by the brain . One possibility is that the brain areas involved in controlling hand movements may also begin to incorporate the use of tools . Another is that distinct brain areas evolved to enable tool use . To test these ideas , Gallivan et al . scanned the brains of human subjects as they reached towards and grasped an object using either their right hand or a set of tongs . The tongs had been designed so that they opened whenever the subjects closed their grip , thereby requiring subjects to perform a different set of movements to use the tongs as opposed to their hand alone . Three distinct patterns of brain activity were observed . First , areas previously linked to the processing of hand movements and the human body were found to represent actions of the hand alone ( and not those of the tool ) , whereas areas previously linked to the processing of tools and tool-related actions represented actions of the tool alone ( and not those of the hand ) . Second , areas of motor cortex implicated in the generation of movement represented actions performed with both the hand and the tool , but showed distinct activity patterns according to which of these was to be used . Lastly , areas associated with high-level cognitive and action-related processing showed similar patterns of activity regardless of whether the subjects were about to use the tongs or just their hand . Given that use of the hand and tool required distinct patterns of muscle contractions , this suggests that these higher-level brain regions must be encoding the action itself rather than the movements needed to achieve it . This study is one of the first to use functional neuroimaging to examine real as opposed to simulated tool use , and increases our understanding of the neural basis of tool use in humans . This knowledge could ultimately have applications for the development of brain-machine interfaces , in which electrodes implanted in motor regions of the brain are used to control prosthetic limbs . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Decoding the neural mechanisms of human tool use |
Here , we demonstrate that Arabidopsis thaliana Formin 2 ( AtFH2 ) localizes to plasmodesmata ( PD ) through its transmembrane domain and is required for normal intercellular trafficking . Although loss-of-function atfh2 mutants have no overt developmental defect , PD’s permeability and sensitivity to virus infection are increased in atfh2 plants . Interestingly , AtFH2 functions in a partially redundant manner with its closest homolog AtFH1 , which also contains a PD localization signal . Strikingly , targeting of Class I formins to PD was also confirmed in rice , suggesting that the involvement of Class I formins in regulating actin dynamics at PD may be evolutionarily conserved in plants . In vitro biochemical analysis showed that AtFH2 fails to nucleate actin assembly but caps and stabilizes actin filaments . We also demonstrate that the interaction between AtFH2 and actin filaments is crucial for its function in vivo . These data allow us to propose that AtFH2 regulates PD's permeability by anchoring actin filaments to PD .
Plasmodesmata ( PD ) function as the intercellular channels in plants and are important for the growth and development of plants , as well as during their interaction with the surrounding environment ( Cheval and Faulkner , 2018; Lee , 2015; Lucas and Lee , 2004; Maule , 2008; Maule et al . , 2011 ) . The density and size of PD are developmentally controlled ( Xu and Jackson , 2010 ) , which allows the formation of spatial symplastic domains in order to establish tissue-specific developmental programs . The permeability of PD must be precisely regulated at specific times during plant growth and development . However , the mechanisms that tightly regulate the permeability of PD are largely unknown . The actin cytoskeleton has been implicated in intercellular communication via PD ( Aaziz et al . , 2001; Chen et al . , 2010; Pitzalis and Heinlein , 2017; White and Barton , 2011 ) . In support of this notion , actin and some actin-associated proteins were demonstrated to localize to PD ( Deeks et al . , 2012; Faulkner et al . , 2009; Radford and White , 1998; Van Gestel et al . , 2003 ) . However , due to technical problems , the existence of filamentous actin in PD remains controversial , as does the nature of its organization . This has been a barrier to our understanding of the function of the actin cytoskeleton in the regulation of cell-to-cell trafficking via PD . In addition , to date , most of the data regarding the function of actin at PD have been obtained from experiments using actin-based pharmacological treatments , which showed that destabilization of actin filaments increases the permeability of PD , whereas stabilization of actin filaments decreases it ( Ding et al . , 1996; Su et al . , 2010 ) . Given that these actin drugs non-selectively target the actin cytoskeleton within cells , it is hard to assess whether and to what extent the changes in PD permeability depend on the alteration in PD actin dynamics . In view of this point , specific manipulation of the actin dynamics at PD via genetic means is an ideal approach , although it requires the identification of PD-localized actin or actin-associated proteins that can serve as the target for the genetic manipulation . Nevertheless , previous observations suggest that the presence and dynamics of the actin cytoskeleton are indeed vital for intercellular trafficking via PD . For example , an interesting study showed that Cucumber mosaic virus ( CMV ) Movement Protein ( MP ) behaves like an actin-binding protein ( ABP ) that severs and caps actin filaments in vitro and its filament severing activity is required for its function in increasing the size exclusion limit ( SEL ) of PD ( Su et al . , 2010 ) . The authors therefore argued that actin filaments at PD function as a filter to control the permeability of PD ( Chen et al . , 2010 ) . The study of Chen et al . also suggests that some endogenous ABPs may be involved in regulating actin dynamics at PD and consequently in controlling the permeability of PD . Formin ( formin homology protein ) is one type of actin nucleation factor that has been implicated in the generation of linear actin bundles ( Blanchoin and Staiger , 2010; Chesarone et al . , 2010; Kovar , 2006 ) . The formin proteins are characterized by the presence of formin homology domain1 ( FH1 ) and FH2 , which are capable of nucleating actin assembly from actin or actin-profilin complexes ( Blanchoin and Staiger , 2010; Chesarone et al . , 2010; Kovar , 2006 ) . There are numerous formin genes in plants ( Blanchoin and Staiger , 2010; Cvrcková et al . , 2004 ) . For instance , the Arabidopsis genome contains 21 formin genes that are divided into two classes: 11 members in Class I , which encode proteins that contain the characteristic transmembrane ( TM ) domain except Arabidopsis formin 7; and 10 members in Class II , which encode proteins without a TM domain but carry an N-terminal phosphatase and tensin-related ( PTEN ) -like domain ( Blanchoin and Staiger , 2010; Deeks et al . , 2002 ) . The Formin proteins have been implicated in numerous actin-based cellular processes in plants , such as polarized pollen tube growth and root hair growth , cell division , cytokinesis and cell morphogenesis , and plant defense ( Cheung et al . , 2010; Favery et al . , 2004; Ingouff et al . , 2005; Li et al . , 2010; Yang et al . , 2011; Ye et al . , 2009; Zhang et al . , 2011 ) . However , it remains largely unexplored how different formin proteins evolve and adapt to various actin-based cellular functions in plants . Here we demonstrate that AtFH2 , along with several Class I formins , localizes to PD and is involved in the regulation of cell-to-cell trafficking via PD . The PD-localization pattern of AtFH2 is determined by its TM domain . Interestingly , we found that several rice Class I formins also target to PD , suggesting that the involvement of Class I formins in regulating actin dynamics at PD may be evolutionarily conserved in plants . We showed that AtFH2 caps and stabilizes actin filaments in vitro , which allows us to propose that AtFH2 regulates actin dynamics and cell-to-cell trafficking by anchoring and stabilizing actin filaments at PD . Loss of function of AtFH2 increases PD permeability , which supports the notion that the stability and/or the amount of actin filaments at PD is crucial for their role in regulating intercellular trafficking . Our study thus adds formin as an important component of the actin-mediated machinery that regulates PD permeability .
To determine the physiological functions of AtFH2 ( At2g43800 ) , two T-DNA insertion mutants of AtFH2 were obtained and analyzed ( Figure 1—figure supplement 1A ) . They were shown to be knockout alleles since the full-length AtFH2 transcript is absent in both atfh2 mutants ( Figure 1—figure supplement 1B , C ) . However , no overt developmental defects were observed in young seedlings and mature plants of atfh2 mutants ( Figure 1—figure supplement 1D ) . This might be due to functional redundancy among Arabidopsis formin genes . However , considering that Arabidopsis formin genes exhibit distinct tissue expression patterns and their encoded proteins are quite diverged ( Cvrcková et al . , 2004 ) , they likely perform some distinct and specialized functions that cannot be uncovered by simply examining the morphology of seedlings and adult plants . Determination of the intracellular localization of AtFH2 may provide hints regarding whether AtFH2 performs specialized cellular functions . We therefore generated an eGFP fusion construct of AtFH2 driven by its own promoter ( AtFH2p:AtFH2-eGFP ) and demonstrated that it is fully functional ( see below ) , which suggests that AtFH2-eGFP will faithfully indicate the intracellular localization of AtFH2 . Initial observations revealed that AtFH2-eGFP decorates dot-like structures along the borders of cells ( Figure 1—figure supplement 2A ) . Given that AtFH2 contains a TM domain , we speculated that the AtFH2-decorated dot-like structures may be located on the plasma membrane . In partial support of this speculation , the AtFH2-eGFP-decorated structures overlap with the staining of propidium iodide ( PI ) ( Figure 1—figure supplement 2B ) , a nucleic acid binding dye that cannot cross the membrane of living cells and consequently stays outside the membrane , thus revealing the contours of cells . Interestingly , the AtFH2-eGFP-positive dot-like structures are reminiscent of PD . To demonstrate this , we performed callose staining , which has been used previously to detect PD ( Bell and Oparka , 2011; Currier , 1957; Fitzgibbon et al . , 2010 ) . We found that the AtFH2-eGFP-decorated structures overlap with dot-like structures revealed by callose staining ( Figure 1A ) . This result was extended by showing that AtFH2 colocalized with two PD markers , PDLP1-GFP ( Figure 1B; [Thomas et al . , 2008] ) and YFP-PDCB1 ( Figure 1C; [Simpson et al . , 2009] ) . Thus , the data demonstrate unambiguously that AtFH2 localizes to PD . Next , we asked how AtFH2 targets to PD . We initially found that the localization pattern of AtFH2 does not depend on the presence of an intact actin cytoskeleton ( Figure 1—figure supplement 2C , D ) . This suggests that the PD localization pattern of AtFH2 is not controlled by the C-terminal FH1 and FH2 domains , which together occupy most of the AtFH2 protein and mediate its interaction with the actin cytoskeleton . We speculated that the PD localization pattern of AtFH2 might be solely determined by its TM domain , which is located near the N-terminus . To quickly test this hypothesis , we adopted the strategy of exogenous expression in tobacco leaves . The feasibility of using this strategy was demonstrated by showing that full-length AtFH2 targets to PD in tobacco leaf epidermal cells ( Figure 1D ) . We generated several fusion constructs in which eGFP was fused with truncated fragments of AtFH2 , starting from the N-terminus ( Figure 1E ) . We subsequently demonstrated that eGFP fusion proteins of truncated AtFH2 localize to PD as long as they contain the TM domain ( Figure 1F–H ) . The eGFP fusion protein without the TM domain failed to target to PD ( Figure 1I ) . Thus , the data suggest that PD-localization of AtFH2 is determined by its TM domain . To determine whether cell-to-cell trafficking was altered in atfh2 , we performed an eGFP diffusion assay as reported previously ( Levy et al . , 2007; Liarzi and Epel , 2005 ) . Leaf epidermal cells were transformed by DNA bombardment with an eGFP-expressing construct . After bombardment by plasmids expressing non-mobile HDEL-mCherry along with plasmids expressing eGFP , the HDEL-mCherry signal mainly appeared in one cell ( Figure 2—figure supplement 1A ) , which suggests that the eGFP signal outside this cell results from diffusion rather than local expression . The efficiency of eGFP diffusion from bombarded cells into surrounding epidermal layers was assessed by counting the number of cell layers with a diffuse eGFP signal ( Figure 2A ) . Our results showed that eGFP diffusion across cells is significantly higher in atfh2 than in WT ( Figure 2B; Figure 2—figure supplement 1B ) . The eGFP diffusion phenotype in atfh2 is rescued by transforming AtFH2p:AtFH2-eGFP into atfh2 ( Figure 2C ) , suggesting that increased eGFP diffusion is indeed caused by loss of function of AtFH2 . Consistent with this notion , the number of eGFP-expressing cells within a diffusion cluster is significantly higher in atfh2 mutant plants ( Figure 2D ) . We found that loss of function of AtFH2 does not affect cytoplasmic streaming ( Figure 2—figure supplement 2 ) , which suggests that loss of function of AtFH2 specifically alters cell-to-cell trafficking . However , we found that overexpression of AtFH2 does not affect the overall organization of actin filaments or cell-to-cell trafficking in leaf epidermal cells ( Figure 2—figure supplement 3 ) . Next , we asked whether loss of function of AtFH2 also affects specific movement of macromolecules across cells by examining the cell-to-cell movement of an eGFP fusion of the Cucumber mosaic virus ( CMV ) protein MP ( MP-eGFP ) . We found that the cell-to-cell movement of MP-eGFP is also increased in atfh2 ( Figure 2E ) . This finding also motivated us to speculate that atfh2 plants might become more sensitive to virus infection . To test this speculation , we challenged Arabidopsis plants with CMV strain Fny ( Fny-CMV ) , which is known to infect Arabidopsis ( Du et al . , 2014 ) . We found that Arabidopsis leaves became curved and distorted after inoculation with Fny-CMV when compared to the control ( Figure 2—figure supplement 4A , B ) . The growth of leaves was inhibited ( Figure 2—figure supplement 4C , D ) and the proportion of symptomatic plants increased in a time-dependent manner ( Figure 2—figure supplement 4E ) . Infection with Fny-CMV was confirmed by measurements showing that the amount of CMV MP RNA increases in a time-dependent manner ( Figure 2—figure supplement 4F ) . We next challenged Arabidopsis plants with Fny-CMV and found that the symptoms of infection were more severe in atfh2 plants than in WT plants ( Figure 2F , G ) . This was quantified by the time-dependent increase in the relative proportion of symptomatic plants ( Figure 2H ) and by the relative accumulation of CMV MP RNA ( Figure 2I ) in atfh2 mutants . In addition , the growth of Arabidopsis leaves was inhibited more severely in atfh2 compared to WT plants ( Figure 2J , K ) . These data together suggest that the SEL of PD increases in atfh2 plants . The cell-to-cell trafficking phenotype is quite subtle in atfh2 . We wondered whether this is due to the functional redundancy of AtFH2 with other Class I formins ( Figure 3A ) . To determine whether there are additional PD-localized Class I formins in Arabidopsis , we examined the localization of the N-terminus of all the other Arabidopsis Class I formins except AtFH3 ( At4g15200 ) , which is pollen-specific ( Ye et al . , 2009 ) , and AtFH7 ( At1g59910 ) , which lacks a TM domain ( Cvrcková et al . , 2004 ) ( Figure 3A ) . We found that AtFH1 ( At3g25500 ) , AtFH4 ( At1g24150 ) , AtFH8 ( At1g70140 ) and AtFH9 ( At5g48360 ) form obvious dot-like structures along the cell membrane whereas AtFH5 ( At5g54650 ) , AtFH6 ( At5g67470 ) , AtFH10 ( At3g07540 ) and AtFH11 ( At3g05470 ) are distributed uniformly on the cell membrane ( Figure 3B–I ) . To determine whether the AtFH1- , AtFH4- , AtFH8- and AtFH9-decorated dot-like structures are PD , we performed simultaneous callose staining or colocalization with AtFH2 ( Figure 3J–O ) and confirmed that AtFH1 and AtFH9 localize to PD ( Figure 3J , K , N and O ) . To determine whether the localization of Class I formins to PD is conserved in plants , we examined the intracellular localization of Class I formins in monocotyledon rice . Among 11 Class I rice formins ( Figure 3—figure supplement 1A; Cvrcková et al . , 2004 ) , we found that OsFH8 ( Os03g0204100 ) , OsFH11 ( Os07g0545500 ) , OsFH15 ( Os09g0517600 ) and OsFH16 ( Os02g0739100 ) are able to localize to PD ( Figure 3—figure supplement 1B–E ) . Given that AtFH1 and AtFH2 belong to the same subclass ( Figure 3A ) , we generated atfh1 atfh2 double mutants ( Figure 4—figure supplement 1 ) to examine whether they redundantly regulate cell-to-cell trafficking via PD . We did not detect overt developmental phenotypes in atfh1 atfh2 mutants compared to WT , atfh1 and atfh2 plants ( Figure 4—figure supplement 1 ) . However , we found that atfh1 atfh2 mutants have stronger cell-to-cell trafficking phenotypes than atfh1 and atfh2 single mutants ( Figure 4; Figure 2—figure supplement 1C ) . This suggests that AtFH1 and AtFH2 indeed act redundantly in regulating cell-to-cell trafficking via PD . Given that the accumulation of callose was shown to be involved in the regulation of PD permeability ( Cui and Lee , 2016; Han et al . , 2014; Simpson et al . , 2009 ) , we wondered whether loss of function of AtFH1 and/or AtFH2 might alter the actin dynamics at PD and consequently affect the accumulation of callose at PD . We therefore performed callose staining with aniline blue and found that the amount of callose accumulated at PD in atfh1 , atfh2 and atfh1 atfh2 mutants was not overtly different to that in WT plants ( Figure 4—figure supplement 2 ) . This suggests that the alteration in cell-to-cell trafficking in atfh1 , atfh2 and atfh1 atfh2 mutants is not due to an alteration in the accumulation of callose at PD . We next generated several recombinant AtFH2 proteins , with increasingly large N-terminal deletions , to determine their effect on actin dynamics in vitro ( Figure 5A , B ) . We initially tested whether they are able to nucleate actin assembly , which is a characteristic feature of formins ( Chesarone et al . , 2010; Goode and Eck , 2007 ) . The bona fide actin nucleator AtFH1 ( Michelot et al . , 2005 ) , which efficiently enhances actin assembly , was used as a positive control ( Figure 5C ) . Unexpectedly , AtFH2 recombinant proteins failed to nucleate actin assembly; instead , they slightly inhibited actin assembly ( Figure 5C ) . This was confirmed by directly visualizing the effect of AtFH2 recombinant proteins on actin assembly by total internal reflection fluorescence microscopy ( TIRFM ) ( Figure 5D , E ) . In addition , we found that recombinant AtFH2 proteins failed to nucleate actin assembly from profilin-actin complexes ( Figure 5F ) . Thus , our results showed that AtFH2 fails to nucleate actin assembly from actin or actin bound to profilin in vitro . To determine whether AtFH2 has the characteristic barbed end capping activity of formins , we performed the seeded actin elongation assay . We found that recombinant AtFH2-ΔN , AtFH2-FH1FH2 and AtFH2-FH2 prevented the addition of profilin-actin complexes in a dose-dependent manner ( Figure 6A , B ) , suggesting that AtFH2 does indeed have barbed end capping activity . From three independent experiments , the average dissociation constant ( Kd ) ( mean ± SE , n = 3 ) was determined to be 80 . 3 ± 12 . 9 nM , 163 . 3 ± 22 . 2 nM and 354 . 6 ± 49 . 0 nM for AtFH2-ΔN , AtFH2-FH1FH2 and AtFH2-FH2 , respectively . The barbed end capping activity of AtFH2 was also confirmed by showing that AtFH2 blocks the annealing of actin filaments ( Figure 6C , D ) and inhibits their elongation ( Figure 6E , F ) . Similar to the features reported for other barbed end capping proteins , we found that AtFH2 protects actin filaments from dilution-mediated actin depolymerization in a dose-dependent manner ( Figure 6G ) . Thus , our results demonstrated that AtFH2 caps the barbed end of actin filaments and stabilizes them in vitro . To directly examine whether the involvement of AtFH2 in the regulation of cell-to-cell trafficking is via its interaction with the actin cytoskeleton , we genetically manipulated AtFH2 in order to alter its interaction with actin filaments . It was reported that substitution of Ile1431 with alanine ( A ) or of Lys1601 with aspartic acid ( D ) within the FH2 domain of the yeast formin Bni1p completely abolished the actin binding activity of its FH2 domain since it fails to form dimers ( Xu et al . , 2004 ) . We therefore mutated the corresponding residues in AtFH2 in order to inhibit its dimer formation and disrupt its interaction with actin filaments . We found that two residues , Ile519 and Lys672 , in AtFH2 correspond to the conserved Ile1431 and Lys1601 in Bni1p ( Figure 7A ) . We initially replaced Ile519 with A and Lys672 with D in AtFH2-FH2 and generated the recombinant AtFH2-FH2I519A , K672D protein ( Figure 7B; named as AtFH2-FH2M hereafter ) . We found that , unlike AtFH2-FH2 , AtFH2-FH2M failed to inhibit actin nucleation ( Figure 7C ) and actin filament elongation ( Figure 7D ) . In contrast to AtFH2-FH2 , which is mainly in the dimeric form , AtFH2-FH2M is mainly in the monomeric form ( Figure 7E ) . These data together suggest that AtFH2-FH2M fails to dimerize and interact with actin filaments . To determine how the mutation affects the role of AtFH2 in regulating PD-mediated cell-to-cell trafficking , we initially found that AtFH2M-eGFP still localizes to PD ( Figure 7F ) . In contrast to AtFH2-eGFP , we found that AtFH2M failed to rescue the cell-to-cell trafficking phenotype in atfh2 mutants assayed with eGFP diffusion ( Figure 7G ) and virus infection ( Figure 7H ) experiments . Thus , these data together suggest that the regulatory function of AtFH2 on PD permeability depends on its interaction with the actin cytoskeleton .
Here , we demonstrate that the Class I formin AtFH2 localizes to PD through its TM domain and is involved in the regulation of cell-to-cell trafficking via PD . Given that AtFH2 is a simple actin filament barbed end capper that stabilizes actin filaments in vitro ( Figures 5 and 6 ) , we propose that AtFH2 acts as an anchor that tethers actin filaments to the membrane at PD and stabilizes the filaments through its barbed end capping activity . In the absence of AtFH2 , the amount of actin filaments is assumed to be reduced at PD . In this regard , our findings that loss of function of AtFH2 increases the SEL of PD ( Figures 2 and 4 ) are consistent with previous results , obtained from actin-based pharmacological treatments , showing that destabilization of actin filaments increases the SEL of PD ( Ding et al . , 1996; Su et al . , 2010 ) . These data together allow us to conclude that the stability and/or the amount of actin filaments is crucial for their role in regulating the permeability of PD . Our study thus significantly enhances our understanding of actin-mediated regulation of intercellular trafficking via PD in plants . Decades of studies have confirmed the presence of the actin cytoskeletal system at PD and demonstrated its involvement in the regulation of PD-mediated intercellular trafficking ( Ding et al . , 1996; Fernandez-Calvino et al . , 2011; Su et al . , 2010 ) , but the underlying molecular and cellular mechanisms remain largely unknown . Studies in this area have progressed slowly for at least two reasons . First , we still lack approaches to specifically visualize actin at PD since they are small and buried deeply in the cell wall . Consequently , the existence of filamentous forms of actin at PD remains controversial . Second , we still lack approaches to specifically perturb actin dynamics at PD . Identification of the native components of the actin cytoskeletal system that specifically localize to PD may provide targets for the genetic manipulation of the actin cytoskeleton in this region , which will enhance our understanding of how actin at PD regulates cell-to-cell trafficking . Identification of the localization of AtFH2 to PD and demonstration of its involvement in PD permeability open a window for us to dissect the status , organization and functions of actin at PD . Several lines of evidence show that loss of function of AtFH2 increases the SEL of PD ( Figures 2 and 4; Figure 2—figure supplement 1B , C ) . Although we cannot directly visualize the organization of actin at PD in WT or atfh2 plants , we speculate that the amount of F-actin is reduced at PD in atfh2 plants since AtFH2 caps and stabilizes actin filaments even though it lacks the characteristic actin nucleation activity ( Figures 5 and 6 ) . Our data allow us to propose that PD-localized AtFH2 recruits actin filaments to PD by capping the barbed end of actin filaments and stabilizing them , which consequently regulates cell-to-cell trafficking . Given that AtFH2 is a simple actin barbed end capper that stabilizes actin filaments in vitro ( Figures 5 and 6 ) and loss of function of AtFH2 increases the permeability of PD ( Figures 2 and 4 ) , it is likely that there are fewer actin filaments at PD in atfh2 mutants , so that intercellular trafficking is upregulated . Our results , along with previous data from actin-based pharmacological treatments showing that destabilization of actin filaments increases SEL of PD ( Ding et al . , 1996; Su et al . , 2010 ) , suggest that actin filaments at PD may act as the physical barrier to regulate its permeability ( Chen et al . , 2010 ) . To gain further insights into the function of actin filaments in the regulation of the permeability of PD , it will be necessary in the future to determine where AtFH2-tethered actin filaments localize along the transport path of PD . The reconstructed three-dimensional ultrastructure of post-cytokinesis plasmodesmata using electron tomography showed that the gap between the plasma membrane ( PM ) and the ER ( called the cytoplasmic sleeve ) within plasmodesmata pores is less than 10 nm ( Nicolas et al . , 2017 ) . This suggests that there is unlikely to be enough space to fit more actin filaments within the cytoplasmic sleeve of PD . The authors also discovered a second PD morphotype ( type I ) that lacks a visible cytoplasmic sleeve but is capable of non-targeted movement of macromolecules ( Nicolas et al . , 2017 ) , which urges us to rethink the relationship between the size of the cytoplasmic sleeve and the permeability . Nonetheless , these findings suggest that the contact between ER and PM is very dynamic . In this regard , AtFH2-tethered actin filaments might be involved in the regulation of the dynamic connection between ER and PM , and may consequently modulate the permeability of PD , although the underlying mechanism remains largely unknown at the moment . Certainly , it might also be possible that AtFH2-tethered actin filaments localize at the neck region of PD to regulate cell-to-cell transportation . Again , it remains to be established how AtFH2-tethered actin filaments at the neck of PD act as a barrier in cell-to-cell trafficking . It is intriguing that loss of function of both AtFH1 and AtFH2 does not have an overt effect on Arabidopsis growth and development ( Figure 4—figure supplement 1H , I ) , even though it significantly increases the permeability of PD ( Figure 4A–D ) . The biological significance of formin-mediated regulation of PD permeability is supported by the finding that loss of function of AtFH1 and/or AtFH2 increases the sensitivity of Arabidopsis plants to infection by Fny-CMV ( Figure 4E–I ) . This suggests that the processes involved in normal plant development may be less sensitive to changes in formin-mediated regulation of PD permeability than the processes involved in coping with biotic stress . Given that AtFH2 is a simple actin barbed end capper ( Figures 5 and 6 ) , it is still an outstanding question how the barbed end capping activity of AtFH2 is tightly regulated in response to developmental and environmental cues in order to fine-tune the permeability of PD . Surprisingly , and unlike previously characterized formins ( Blanchoin and Staiger , 2010; Chesarone et al . , 2010; Goode and Eck , 2007; Kovar , 2006 ) , AtFH2 lacks actin nucleation activity in the presence or absence of profilin but retains the actin filament barbed end capping activity in vitro ( Figures 5 and 6 ) . The crystal structure of the FH2 domain derived from yeast Bni1p showed that it forms a stable yet flexible dimer ( Xu et al . , 2004 ) . The later co-crystal of Bni1p FH2 in complex with actin demonstrates the flexibility of the FH2 dimer , which permits the addition of actin monomers onto the barbed end of actin filaments ( Otomo et al . , 2005 ) . Based on the structural and biochemical data , it was proposed that the FH2 dimer exists in a dynamic equilibrium between a ‘closed’ and an ‘open’ state ( Kovar , 2006 ) . It is possible that FH2 dimers in AtFH2 might be less flexible , and are able to cap the barbed end of actin filaments but cannot facilitate the further addition of actin or actin-profilin complexes . In the future , careful examination of the molecular details underlying the unique activity of AtFH2 may provide insights into the mechanism of action of formin in regulating actin nucleation and elongation . Nonetheless , based on the unique biochemical properties of AtFH2 , we propose that it serves well as an anchor for tethering actin filaments at the membrane of PD by capping the barbed end of actin filaments . Since AtFH2 presumably only interacts with filamentous actin through its barbed end capping activity ( Figures 5 and 6 ) , and the interaction of AtFH2 with actin filaments was demonstrated to be crucial for its in vivo functions ( Figure 7G , H ) , this study to some extent provides another piece of evidence supporting the presence of filamentous actin at PD . Interestingly , we found that AtFH1 , the closest homolog of AtFH2 , is also able to target to PD ( Figure 3 ) and functions redundantly with AtFH2 in the regulation of intercellular trafficking ( Figure 4 ) . Biochemically , their functional redundancy is likely due to the fact that both have capping activity ( Figure 6; [Michelot et al . , 2005] ) . Considering that AtFH1 is a non-processive actin nucleation factor ( Michelot et al . , 2006 ) , AtFH1 and AtFH2 might coordinately regulate actin dynamics at PD , with AtFH1 promoting actin polymerization to generate more actin filaments around PD , and AtFH2 subsequently taking over the binding to the barbed end of actin filaments after the dissociation of AtFH1 . We cannot completely rule out the possibility that AtFH2 might have actin nucleation activity when bound to other proteins and/or after post-translational modification in vivo . More work needs to be undertaken to examine these possibilities in the future . Nonetheless , our findings provide insights into the functional specification of members of the formin protein family and the functional adaptation of the actin cytoskeletal system in plants in general . Our results showed that the localization of AtFH2 is determined by its TM domain ( Figure 1 ) , which suggests that the TM domain of AtFH2 has evolved to allow it to adapt to regulate actin dynamics at PD . Given that the membrane of PD has a unique phospholipid composition ( Grison et al . , 2015 ) , the TM domains of AtFH1 and AtFH2 may bind preferentially to certain phospholipid ( s ) when compared to other Class I formins in Arabidopsis . Strikingly , we showed that several rice Class I formins localize to PD ( Figure 3—figure supplement 1B–E ) , suggesting that the involvement of Class I formins in regulating actin dynamics at PD might be evolutionarily conserved in plants . In summary , we found that AtFH2 localizes to PD and functions redundantly with AtFH1 to regulate cell-to-cell trafficking in Arabidopsis . Based on the in vitro biochemical data that AtFH2 caps and stabilizes actin filaments , we propose that AtFH2 regulates cell-to-cell trafficking by tethering actin filaments to the membrane at PD and stabilizing them through its barbed end capping activity . Our study suggests that the stability and/or the amount of actin filaments at PD is crucial for the permeability of PD .
T-DNA insertion lines GK_066D02 , GK_396H03 , Salk_032981 and Salk_143939 were designated as atfh2-1 , atfh2-2 , atfh1-1 and atfh1-3 , respectively . Information about the mutant atfh1-1 has been presented previously ( Rosero et al . , 2013 ) . Double mutant atfh1 atfh2 was generated by crossing atfh2-1 with atfh1-3 , and atfh1-3 atfh2-1 is named as atfh1 atfh2 throughout the manuscript . Arabidopsis Columbia-0 ( Col-0 ) ecotype was used as wild-type ( WT ) and Arabidopsis plants were grown in soil or media under a 16-h-light/8-h-dark photoperiod at 22°C . Nicotiana benthamiana plants were grown in pots placed in growth-rooms at 25°C under a 16-h-light/8-h-dark cycle . The genotyping of atfh2-1 and atfh2-2 was performed with primer pairs atfh2-1 LP/atfh2-1 RP and atfh2-2 LP/atfh2-2 RP ( Supplementary File 1 ) in combination with left border primer GK_LB ( Supplementary File 1 ) , and the genotyping of atfh1-1 and atfh1-3 was performed with primer pairs atfh1-1 LP/atfh1-1 RP and atfh1-3 LP/atfh1-3 RP ( Supplementary File 1 ) in combination with left border primer Salk_LB ( Supplementary File 1 ) , respectively . RT-PCR was performed to determine the transcript levels of AtFH1 and AtFH2 in their loss-of-function mutants . Total RNA was isolated from leaves of 4-week-old plants using Trizol reagent ( Invitrogen ) and reverse transcribed using MMLV reverse transcriptase ( Promega ) according to the manufacturer’s instructions . To perform semi-quantitative RT-PCR analysis , two primer pairs AtFH2CDSFOR/AtFH2CDSREV and AtFH1CDSFOR/AtFH1CDSREV ( Supplementary File 1 ) were used to amplify full-length AtFH2 and AtFH1 from WT and their loss-of-function mutants . To perform quantitative RT-PCR analysis , two primer pairs Q1/Q2 and Q3/Q4 ( Supplementary File 1 ) were used to determine the transcript levels of AtFH2 and AtFH1 , respectively . eIF4A was used as an internal control and was amplified with primers eIF4AFOR/eIF4AREV ( Supplementary File 1 ) . The 2−ΔΔCt method was used to quantify the qRT-PCR results . 2 × RealStar Power SYBR Mixture ( GeneStar ) was used for the amplification . AtFH2-eGFP fusion constructs , driven either by the AtFH2 promoter or the 35S promoter were generated to determine the subcellular localization of AtFH2 . The promoter of AtFH2 was amplified with AtFH2PROFOR and AtFH2PROREV primers ( Supplementary File 1 ) using Arabidopsis genomic DNA as the template , and the coding sequence ( CDS ) of AtFH2 was amplified from WT Arabidopsis cDNA using AtFH2CDSFOR and AtFH2CDSREV primers ( Supplementary File 1 ) . The promoter and CDS were cloned into pEASY-Blunt vector ( TransGen ) . The CDS of AtFH2 was subsequently moved into pCAMBIA1301-35S:eGFP-NOS or pCAMBIA1301-35S:mCherry-NOS to generate pCAMBIA1301-35S:AtFH2-eGFP-NOS or pCAMBIA1301-35S:AtFH2-mCherry-NOS plasmids , or moved into pCAMBIA1301-eGFP-NOS or pCAMBIA1301-mCherry-NOS along with the promoter of AtFH2 to generate pCAMBIA1301-AtFH2p:AtFH2-eGFP-NOS or pCAMBIA1301-AtFH2p:AtFH2-mCherry-NOS plasmids . Substitution of isoleucine-519 with alanine ( I519A ) and lysine-672 with aspartic acid ( K672D ) was achieved by changing AT to GC and AAA to GAT by PCR with primers containing the corresponding nucleotide changes ( M1/M2 and M3/M4; see Supplementary File 1 ) using pEASY-Blunt-AtFH2 plasmid as the template . The resulting pEASY-Blunt-AtFH2M plasmid was digested with SalI/BamHI and the resulting AtFH2M fragment was moved into pCAMBIA1301-35S:eGFP-NOS or pCAMBIA1301-NOS along with the promoter of AtFH2 to generate pCAMBIA1301-35S:AtFH2M-eGFP-NOS and pCAMBIA1301-AtFH2p:AtFH2M-NOS . Those plasmids were introduced into Agrobacterium tumefaciens strain GV3101 and transformed into either WT or atfh2 plants via the floral dip method ( Clough and Bent , 1998 ) . The T3 homozygous plants were used for subsequent analyses . To determine whether PD localization of AtFH2 is determined by its N-terminus , three N-terminal fragments , designated as AtFH2N437 , AtFH2N282 and AtFH2N175 , were amplified with primer pairs , AtFH2CDSFOR/AtFH2N437REV , AtFH2CDSFOR/AtFH2N282REV and AtFH2CDSFOR/AtFH2N175REV ( see Supplementary File 1 ) , respectively , using AtFH2 cDNA as the template . They were subsequently moved into pBI101-35S:eGFP-NOS to generate pBI101-35S:AtFH2N437-eGFP-NOS , pBI101-35S:AtFH2N282-eGFP-NOS and pBI101-35S:AtFH2N175-eGFP-NOS plasmids . To determine whether the TM domain of AtFH2 is necessary for its PD localization , the N282 fragment without the TM domain ( named as AtFH2N282-ΔTMD ) was amplified with primer pairs AtFH2N282-ΔTMDFOR/AtFH2N282-ΔTMDREV and AtFH2CDSFOR/AtFH2N282REV using pBI101-35S:AtFH2N282-eGFP-NOS as the template and the amplified fragment was subsequently moved into pBI101-35S:eGFP-NOS to generate pBI101-35S:AtFH2N282-ΔTMD-eGFP-NOS . To determine whether other Class I Arabidopsis formins localize to PD , the N-terminus of AtFH1 , AtFH4 , AtFH6 , AtFH8 , AtFH9 , AtFH10 and AtFH11 was amplified from WT Arabidopsis cDNA using primer pairs AtFH1 ( N ) FOR/AtFH1 ( N ) REV , AtFH4 ( N ) FOR/AtFH4 ( N ) REV , AtFH6 ( N ) FOR/AtFH6 ( N ) REV , AtFH8 ( N ) FOR/AtFH8 ( N ) REV , AtFH9 ( N ) FOR/AtFH9 ( N ) REV , AtFH10 ( N ) FOR/AtFH10 ( N ) REV and AtFH11 ( N ) FOR/AtFH11 ( N ) REV ( see Supplementary File 1 ) , respectively . Error-free PCR fragments digested with SalI/BamHI were moved into pBI101-35S:eGFP-NOS restricted with SalI/BamHI to generate the corresponding plasmids . To determine whether some rice Class I formins localize to PD , the N-terminus of Class I OsFH8 , OsFH11 , OsFH15 and OsFH16 was amplified from Oryza sativa Japonica cDNA using primer pairs OsFH8 ( N ) FOR/OsFH8 ( N ) REV , OsFH11 ( N ) FOR/OsFH11 ( N ) REV , OsFH15 ( N ) FOR/OsFH15 ( N ) REV and OsFH16 ( N ) FOR/OsFH16 ( N ) REV ( Supplementary File 1 ) , respectively , and the error-free PCR fragments were subsequently moved into pBI101-35S:eGFP-NOS restricted with XbaI/BamHI . The resulting eGFP fusion constructs were transiently expressed in Nicotiana benthamiana leaves via Agrobacterium-mediated transformation ( see below ) . To indicate the primary cell targeted by particle bombardment , the ER marker HDEL ( Batoko et al . , 2000 ) was amplified with primers HDEL-mCherryFOR and HDEL-mCherryREV ( Supplementary File 1 ) using the plasmid 1301-Lat52:HDEL-mCherry-NOS as the template and subsequently moved into a pdGN vector to replace eGFP ( Lee et al . , 2005 ) . To generate recombinant AtFH2 proteins , three fragments , AtFH2-ΔN , AtFH2-FH1FH2 and AtFH2-FH2 , were amplified with primer pairs AtFH2ΔNFOR/AtFH2ΔNREV , AtFH2-FH1FH2FOR/AtFH2-FH1FH2REV and AtFH2-FH2FOR/AtFH2-FH2REV ( Supplementary File 1 ) , respectively , using AtFH2 cDNA as the template . The amplified fragments were moved into pET28a or pET23a to generate prokaryotic expression plasmids , pET28a-AtFH2-ΔN , pET28a-AtFH2-FH1FH2 and pET23a-AtFH2-FH2 , respectively . For substitution of isoleucine-519 with alanine ( I519A ) and lysine-672 with aspartic acid ( K672D ) in AtFH2-FH2 , AtFH2-FH2 was amplified with primer pairs M5/M6 and M7/M8 ( Supplementary File 1 ) using pEASY-Blunt-AtFH2M as the template . The amplified fragment , named as AtFH2-FH2M , was moved into pGEX-KG or pET23b to generate pGEX-KG-AtFH2-FH2M or pET23b-AtFH2-FH2M plasmid . They were transformed into E . coli Tuner ( DE3 ) pLysS strain . To determine the colocalization of CMV MP with AtFH2 , AtFH2M and rice formins , Agrobacterium-mediated transient expression was performed in N . benthamiana , which was essentially according to the previously published method ( Voinnet et al . , 2003 ) . Agrobacterium tumefaciens stain GV3101 transformed with pGD-CMV MP-mCherry ( Su et al . , 2010 ) , pCambia1301-35S:AtFH2-eGFP-NOS , pCambia1301-35S:AtFH2M-eGFP-NOS , pBI101-35S:OsFH8 ( N ) -eGFP-NOS , pBI101-35S:OsFH11 ( N ) -eGFP-NOS , pBI101-35S:OsFH15 ( N ) -eGFP-NOS , pBI101-35S:OsFH16 ( N ) -eGFP-NOS and GV3101 strain carrying p19 were grown at 28°C in Luria Broth culture medium containing 50 μg/mL kanamycin , 50 μg/mL rifampicin , 10 mM MES and 40 μM acetosyringone for 16 hr . Cells were collected by centrifugation and resuspended in 10 mM MgCl2 and 200 μM acetosyringone and the cell suspension was adjusted to OD600 = 1 . 0 . Cells were left at room temperature for 3 hr . Cultures were mixed with GV3101 transformed with pGD-CMV MP-mCherry or p19 and subsequently infiltrated into 3-week-old N . benthamiana leaves . N . benthamiana leaf epidermal cells were visualized and the images were captured by laser scanning confocal microscopy after 48–60 hr . The localization of fluorescently tagged proteins was assessed using a laser scanning confocal microscope . eGFP or YFP was excited by a 488 nm argon laser and emission was captured in the range of 505–545 nm . mCherry or propidium iodide ( PI ) was excited by a 543 nm HeNe laser and emission was captured in the range of 590–625 nm . Arabidopsis seedlings were cultured on 1/2 Murashige and Skoog culture medium for 3 to 5 days before being observed under the microscope . For PI staining , Arabidopsis seedlings were incubated with 50 μg/mL PI ( Sigma-Aldrich , P4864 ) diluted with the solution containing 9% glucose and 5% glycerol for 5 min at room temperature as described previously ( Wang and Huang , 2014 ) . To demonstrate whether AtFH2 localizes to PD , PD was revealed by staining with aniline blue ( see below ) or colocalization with CMV MP-mCherry , PDLP1-GFP ( Thomas et al . , 2008 ) or YFP-PDCB1 ( Simpson et al . , 2009 ) . To simultaneously visualize AtFH2 and actin filaments , GFP-ABD2 marker ( Sheahan et al . , 2004 ) was introduced into Arabidopsis plants expressing pCAMBIA1301-35S:AtFH2-mCherry by genetic crossing . To determine whether the organization of actin filaments is required for the localization of AtFH2 , 7-day-old Arabidopsis seedlings were treated with 1 μM latrunculin B for 1 hr before imaging . Plasmodesmata was revealed by staining with aniline blue . To observe colocalization of AtFH2 and callose , 7-day-old Arabidopsis seedlings were used . To quantify the accumulation of callose at PD , the fifth leaves of 4-week-old Arabidopsis were used . Briefly , Arabidopsis seedlings or detached leaves were initially treated with a solution containing 300 μM MBS for 30–60 min , which was followed by incubation with aniline blue solution ( 0 . 1% aniline blue in double-distilled water and 1 M glycerol , pH 9 . 5 , at a volume ratio of 2:3 ) for 30 min at room temperature as described previously ( Levy et al . , 2007; Simpson et al . , 2009 ) . To quantify the accumulation of callose at PD in WT , atfh1 , atfh2 and atfh1 atfh2 plants , confocal images were subjected to intensity analysis using ImageJ software ( http://rsbweb . nih . gov/ij/; version 1 . 51 ) as described previously ( Simpson et al . , 2009 ) . The average fluorescence intensity and the size of the aniline blue-stained callose foci were measured and compared between WT and formin loss-of-function mutants . For each genotype , more than 100 data sets were collected from at least 20 separate images . As the data did not show a normal distribution , statistical comparison between different genotype was performed with a Mann-Whitney U test using IBM SPSS Statistics version 25 software . To capture the organization of actin filaments in leaf epidermal cells , 7-day-old Arabidopsis seedlings expressing 35S:GFP-ABD2 were observed under a laser scanning confocal microscope and the Z series images were collected with the step size set at 0 . 5 μm . The density of actin filaments in the leaf epidermal cells was determined by measuring the value of occupancy as described previously ( Higaki et al . , 2010 ) . Briefly , the Z-stack images were converted to binary images by thresholding and were subsequently skeletonized and processed with maximum intensity projection . The occupancy is determined as the proportion of the pixels representing filaments out of the total pixels in the selected region . More than 50 data sets were collected from at least 10 plants for each genotype . All measurements and image processing were performed by ImageJ software ( version 1 . 51 s , http://imagej . nih . gov/ij ) . Cytoplasmic streaming was quantified according to the method described previously ( Okamoto et al . , 2016 ) . The hypocotyl cells were observed in a bright field under an IX71 microscope ( Olympus ) equipped with a × 40 objective . Digital images were collected at 1 s intervals for 10–15 min with a Retiga Exi Fast 1394 CCD camera ( QImaging ) using Image-Pro Express 6 . 3 software . For estimation of the velocity , more than 50 data sets were collected from at least five plants for each genotype . The moving plastids were randomly selected and measured using ImageJ software . Cell-to-cell trafficking was examined by determining the diffusion of eGFP across cells or the cell-to-cell movement of CMV MP-eGFP , which is described in more detail at Bio-protocol ( Diao et al . , 2019 ) . In brief , the transient expression of eGFP or CMV MP-eGFP was achieved by a biolistic DNA delivery system in Arabidopsis leaves . Bombardment of leaves with plasmid DNA was performed basically as described by Liarzi and Epel ( 2005 ) with slight modifications . Briefly , approximately 2 μg of plasmid DNA were mixed with 0 . 5 mg gold particles ( diameter , 1 μm , Bio-Rad ) in the presence of 20 μL of fresh N-[3-aminopropyl]−1 , 4-butanediamine ( spermidine ) ( 0 . 1 M ) and 50 μL of CaCl2 ( 2 . 5 M ) before vortexing uninterruptedly for 3 min . The mixture was centrifuged for 10 s before precipitating for 5 min . The pellet was subsequently washed with 70% ethanol and absolute ethanol , and finally resuspended in 15 μL absolute ethanol . Rosette leaves from 3-week-old Arabidopsis plants which had not yet bolted were placed in a petri dish containing 20 mL of MS medium with 0 . 8% agar before bombardment . The target leaf was placed at a distance of 12 cm from the gene gun . The leaves were incubated for 24 hr or 48 hr at 23°C in the dark . To quantify the number of layers of cells into which eGFP diffused , images of eGFP in the lower epidermis cells were captured by a Zeiss LSM 510 META at 24 hr and 48 hr . Statistical analysis was performed as described by Levy et al . ( 2007 ) . The cell that was transformed by DNA bombardment was defined as layer 0 , and the cells that share a common cell wall with layer 0 were defined as layer 1 . Cells that share a common cell wall with layer 1 cells , but not with layer 0 cells , were defined as layer 2 cells . Cells that expressed GFP but showed no diffusion were not counted to avoid the situation in which damage was caused by the bombardment . To indicate the bombarded cell , pdGN-35S:HDEL-mCherry plasmid was used for bombardment along with the pdGN plasmid expressing eGFP . Both pdGN and pdGN-35S:HDEL-mCherry plasmids at 2 μg were mixed with 1 μm gold particles ( Bio-Rad ) for the bombardment . Results were analyzed with the nonparametric Mann-Whitney U test using IBM SPSS Statistics version 25 software . Fny-CMV infected N . benthamiana leaves were used for virus purification . After 7 days , the extract of virions at a concentration of 100 μg/ml was rub inoculated onto Arabidopsis seedlings at the 5–6 true-leaf stage ( Du et al . , 2014 ) . The extent of infection was assessed by determining the amount of CMV MP RNA and quantifying the morphology of leaves . To detect the amount of CMV MP RNA in plants , a quantitative RT-PCR was performed . Primers Oligo ( dT ) 18 and CMV MPREV were used to reverse transcribe plant total RNA and RNA of CMV MP respectively . Primer pairs CMV MP-RTFOR/CMV MP-RTREV ( Supplementary File 1 ) were used to determine the transcript levels of CMV MP . The relative amount of CMV MP transcripts was quantified using the 2-ΔΔCt method with eIF4A as the internal control . AtFH2-FH2 , AtFH2-FH1FH2 , AtFH2ΔN and AtFH2-FH2M either fused with GST or 6 × His were expressed in the E . coli Tuner ( DE3 ) pLysS strain by induction with 0 . 4 mM isopropyl β-D-thiogalactopyranoside overnight at 16°C . E . coli cells were collected by centrifugation at 6000 rpm ( Beckman JA-10 ) , and resuspended either in 1 × PBS ( 140 mM NaCl , 2 . 7 mM KCl , 10 mM Na2HPO4 , 1 . 8 mM KH2PO4 ) or in 1 × binding buffer ( 25 mM Tris-HCl , pH 8 . 0 , 250 mM KCl , 5 mM imidazole , 2 mM DTT ) with proteinase inhibitor cocktail for the purification of GST fusion proteins or His-tag fusion proteins . After sonication and centrifugation at 27 , 216 g for 30 min , the supernatant was incubated with glutathione-sepharose ( Amersham ) or Ni-NTA resin ( Novagen ) , using purification procedures according to the manufacturers’ instructions . All purified proteins were dialyzed against 5 mM Tris-HCl , pH 8 . 0 , 5% glycerol , flash frozen in liquid nitrogen , and stored at −80°C . They were preclarified at 200 , 000 g for 30 min at 4°C before the subsequent analyses . Actin was isolated from acetone powder of rabbit skeletal muscle ( Spudich and Watt , 1971 ) . Monomeric Ca-ATP-actin was subsequently purified by chromatography on Sephacryl S-300 at 4°C in Buffer G ( 5 mM Tris-HCl , pH 8 . 0 , 0 . 2 mM ATP , 0 . 1 mM CaCl2 , 0 . 5 mM DTT , 0 . 1 mM NaN3 ) ( Pollard , 1984 ) . Actin was labeled on Cys-374 with pyrene iodoacetamide to monitor the kinetic process of actin polymerization and depolymerization ( Pollard , 1984 ) or labeled with Oregon-green as described previously ( Amann and Pollard , 2001 ) . Human profilin I was purified as described previously ( Fedorov et al . , 1994 ) , and AtFH1-FH1FH2 was purified according to Michelot et al . ( 2005 ) . The actin nucleation assay procedure was performed as described previously ( Higgs et al . , 1999 ) . Mg-ATP-actin or actin-profilin complexes ( 3 μM , 10% pyrene labeled ) were incubated with various concentrations of AtFH2 for 3 min at room temperature before the addition of one-tenth volume of 10 × KMEI ( 500 mM KCl , 10 mM MgCl2 , 10 mM EGTA , 100 mM imidazole-HCl , pH 7 . 0 ) . Polymerization of actin was traced by monitoring pyrene fluorescence by a QuantaMaster Luminescence QM 3 PH fluorometer ( Photo Technology International , Inc . ) with the excitation and emission wavelengths set at 365 nm and 407 nm , respectively . A seeded actin elongation assay , to determine the affinity of AtFH2 for the barbed end of actin filaments , was performed roughly according to the method described previously ( Huang et al . , 2003 ) . Actin filaments at a final concentration of 0 . 8 μM were incubated for 3 min at room temperature with various concentrations of AtFH2 . Actin elongation was initiated by the addition of actin-profilin complexes ( 1 μM , 10% pyrene labeled ) to the actin filament mixture . The equilibrium binding constant ( Kd ) of AtFH2 proteins for the barbed end of actin filaments was determined by plotting the initial elongation rate as a function of the concentration of AtFH2 proteins using the following equation:Vi = Vif + Vib-VifKd+ ends+ AtFH2-Kd + ends + AtFH22-4endsAtFH2 2endswhere Vi is the observed rate of elongation , Vif is the rate of elongation when all the barbed ends are free , Vib is the rate of elongation when all the barbed ends are capped , [ends] is the concentration of barbed ends , and [AtFH2] is the concentration of AtFH2 . The data were modeled with KaleidaGraph software ( version 4 . 03 , http://kaleidagraph . software . informer . com/ ) . The dilution-mediated actin depolymerization assay was performed as described previously ( Huang et al . , 2003 ) . 5 μM F-actin ( 50% pyrene-labeled ) was incubated with various concentrations of AtFH2 for 3 min at room temperature before being diluted 25-fold into Buffer G . Actin depolymerization was traced by monitoring the changes in pyrene fluorescence by a QuantaMaster Luminescence QM 3 PH fluorometer ( Photo Technology International , Inc . ) with the excitation and emission wavelengths set at 365 nm and 407 nm , respectively . The annealing of actin filaments was visualized and quantified by fluorescence light microscopy as described previously ( Huang et al . , 2003 ) . Briefly , actin filaments were labeled with equimolar rhodamine-phalloidin ( Sigma-Aldrich ) and then they were broken by sonication in the absence or presence of different concentrations of AtFH2 . After incubation for a certain period of time , actin filaments were visualized under an Olympus IX71 microscope equipped with a × 60 , 1 . 42–numerical aperture oil objective and the images were acquired by a Retiga EXi Fast 1394 CCD camera ( QImaging ) with Image-Pro Express 6 . 3 software . The effect of AtFH2 on actin filament annealing was quantified by measuring the length of actin filaments . The effect of AtFH2 on the dynamics of single actin filaments was determined by TIRFM as described previously ( Zheng et al . , 2012 ) . The effect of AtFH2 on actin nucleation was analyzed by counting the number of actin filaments within microscope fields , and the effect of AtFH2 on the growth of actin filaments was analyzed by performing kymograph analysis by tracing the elongating end of actin filaments using ImageJ software ( http://rsbweb . nih . gov/ij/; version 1 . 51 ) . Size exclusion chromatography was performed to determine the size of AtFH2-FH2 and AtFH2-FH2M . Protein samples in 1 mL of running buffer ( 100 mM Tris-HCl , pH 8 . 5 , 150 mM NaCl , 1 mM DTT and 5% glycerol ) were loaded onto a Superdex 200 Increase 10/300 gel filtration column ( GE ) pre-equilibrated with running buffer , and the column was run at the speed of 0 . 4 mL/min with running buffer . Protein fractions of 0 . 5 ml were collected and separated by SDS-PAGE . Molecular weights were determined according to the manufacturer’s instructions . The statistical analysis of the datasets was performed using IBM SPSS Statistics version 25 software . The normality of the datasets was initially assessed by Shapiro-Wilk tests . If the data were normally distributed , the Student’s t-test was applied for the subsequent statistical analyses . However , if the data were not normally distributed , the Mann-Whitney U test was applied for the subsequent statistical analyses . Differences were considered significant when p<0 . 05 and differences were considered extremely significant when p<0 . 01 . The statistical tests and number of replicates are provided in the figure legend . | Plant cells communicate with each other via narrow channels embedded across adjacent cell walls . These channels , called plasmodesmata , allow molecules to pass between cells , thereby enabling plants to grow normally and develop tissues and organs . But plasmodesmata also serve as passageways that viruses can exploit to infect more and more cells . Given these pros and cons , plants must regulate how permeable their plasmodesmata are so they can transport necessary materials cell-to-cell while still defending against the spread of infection . Each cell within plants , animals , and fungi , contains a protein skeleton that helps to stabilize it . A threadlike fiber called actin filament , one of the key components that makes up the cell’s skeleton , presumably extends out to the plasmodesmata , which lie across the cell’s external wall . Previous research has shown that actin helps regulate cell-to-cell traffic through the plasmodesmata and that drug treatments involving actin disturb normal traffic . But techniques to visualize actin at the plasmodesmata are lacking , and both how plants control their plasmodesmata and actin’s involvement remain unclear . Diao et al . used a confocal microscope , fluorescent tags , and staining procedures in experiments that analyzed how plasmodesmata and actin interact within a small flowering plant called thale cress . These experiments showed that a protein known to regulate actin , called Formin 2 , positions itself at the plasmodesmata where it caps off actin threads and anchors them to the channels . Diao et al . also generated thale cress that cannot produce Formin 2 . These mutant plants had more permeable plasmodesmata and were more susceptible to a virus . By stably tethering actin to the plasmodesmata , Formin 2 plays a key part in regulating the permeability of these cell-to-cell channels , with unstable actin threads resulting in more penetrable plasmodesmata . These findings provide further evidence that plants rely on actin to regulate plasmodesmata , and they establish that Formin 2 is involved . Further research will clarify how actin and Formin 2 work together to adjust the structure of plasmodesmata channels . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2018 | Arabidopsis formin 2 regulates cell-to-cell trafficking by capping and stabilizing actin filaments at plasmodesmata |
The architecture of normal and diseased tissues strongly influences the development and progression of disease as well as responsiveness and resistance to therapy . We describe a tissue-based cyclic immunofluorescence ( t-CyCIF ) method for highly multiplexed immuno-fluorescence imaging of formalin-fixed , paraffin-embedded ( FFPE ) specimens mounted on glass slides , the most widely used specimens for histopathological diagnosis of cancer and other diseases . t-CyCIF generates up to 60-plex images using an iterative process ( a cycle ) in which conventional low-plex fluorescence images are repeatedly collected from the same sample and then assembled into a high-dimensional representation . t-CyCIF requires no specialized instruments or reagents and is compatible with super-resolution imaging; we demonstrate its application to quantifying signal transduction cascades , tumor antigens and immune markers in diverse tissues and tumors . The simplicity and adaptability of t-CyCIF makes it an effective method for pre-clinical and clinical research and a natural complement to single-cell genomics .
Histopathology is among the most important and widely used methods for diagnosing human disease and studying the development of multicellular organisms . As commonly performed , imaging of formalin-fixed , paraffin-embedded ( FFPE ) tissue has relatively low dimensionality , primarily comprising Hematoxylin and Eosin ( H&E ) staining supplemented by immunohistochemistry ( IHC ) . The potential of IHC to aid in diagnosis and prioritization of therapy is well established ( Bodenmiller , 2016 ) , but IHC is primarily a single-channel method: imaging multiple antigens usually involves the analysis of sequential tissue slices or harsh stripping protocols ( although limited multiplexing is possible using IHC and bright-field imaging [Stack et al . , 2014; Tsujikawa et al . , 2017] ) . Antibody detection via formation of a brown diamino-benzidine ( DAB ) or similar precipitates are also less quantitative than fluorescence ( Rimm , 2006 ) . The limitations of IHC are particularly acute when it is necessary to quantify complex cellular states and multiple cell types , such as tumor infiltrating regulatory and cytotoxic T cells ( Postow et al . , 2015 ) in parallel with tissue and pharmaco-dynamic markers . Advances in DNA and RNA profiling have dramatically improved our understanding of oncogenesis and propelled the development of targeted anticancer drugs ( Garraway and Lander , 2013 ) . Sequence data are particularly useful when an oncogenic driver is both a drug target and a biomarker of drug response , such as BRAFV600E in melanoma ( Chapman et al . , 2011 ) or BCR-ABL in chronic myelogenous leukemia ( Druker and Lydon , 2000 ) . However , in the case of drugs that act through cell non-autonomous mechanisms , such as immune checkpoint inhibitors , tumor-drug interaction must be studied in the context of multicellular environments that include both cancer and non-malignant stromal and infiltrating immune cells . Multiple studies have established that these components of the tumor microenvironment strongly influence the initiation , progression and metastasis of cancer ( Hanahan and Weinberg , 2011 ) and the magnitude of responsiveness or resistance to immunotherapies ( Tumeh et al . , 2014 ) . Single-cell transcriptome profiling provides a means to dissect tumor ecosystems at a molecular level and quantify cell types and states ( Tirosh et al . , 2016 ) . However , single-cell sequencing usually requires disaggregation of tissues , resulting in loss of spatial context ( Tirosh et al . , 2016; Patel et al . , 2014 ) . As a consequence , a variety of multiplexed approaches to analyzing tissues have recently been developed with the goal of simultaneously assaying cell identity , state , and morphology ( Giesen et al . , 2014; Gerdes et al . , 2013; Micheva and Smith , 2007; Remark et al . , 2016; Gerner et al . , 2012 ) . For example , FISSEQ ( Lee et al . , 2014 ) enables genome-scale RNA profiling of tissues at single-cell resolution , and multiplexed ion beam imaging ( MIBI ) and imaging mass cytometry achieve a high degree of multiplexing using antibodies as reagents , metals as labels and mass spectrometry as a detection modality ( Giesen et al . , 2014; Angelo et al . , 2014 ) . Despite the potential of these new methods , they require specialized instrumentation and consumables , which is one reason that the great majority of basic and clinical studies still rely on H&E and single-channel IHC staining . Moreover , methods that involve laser ablation of samples such as MIBI inherently have a lower resolution than optical imaging . Thus , there remains a need for highly multiplexed tissue analysis methods that ( i ) minimize the requirement for specialized instruments and costly , proprietary reagents , ( ii ) work with conventionally prepared FFPE tissue specimens collected in clinical practice and research settings , ( iii ) enable imaging of ca . 50 antigens at subcellular resolution across a wide range of cell and tumor types , ( iv ) collect data with sufficient throughput that large specimens ( several square centimeters ) can be imaged and analyzed , ( v ) generate high-resolution data typical of optical microscopy , and ( vi ) allow investigators to customize the antibody mix to specific questions or tissue types . Among these requirements the last is particularly critical: at the current early stage of development of high dimensional histology , it is essential that individual research groups be able to test the widest possible range of antibodies and antigens in search of those with the greatest scientific and diagnostic value . This paper describes a method for highly multiplexed fluorescence imaging of tissues , tissue-based cyclic immunofluorescence ( t-CyCIF ) , inspired by a cyclic method first described by Gerdes et al . ( 2013 ) . t-CyCIF also extends a method we previously described for imaging cells grown in culture ( Lin et al . , 2015 ) . In its current implementation , t-CyCIF assembles up to 60-plex images of FFPE tissue sections via successive rounds of four-channel imaging . t-CyCIF uses widely available reagents , conventional slide scanners and microscopes , manual or automated slide processing and simple protocols . It can , therefore , be implemented in most research or clinical laboratories on existing equipment . Our data suggest that high-dimensional imaging methods using cyclic immunofluorescence have the potential to become a robust and widely-used complement to single-cell genomics , enabling routine analysis of tissue and cancer morphology and phenotypes at single-cell resolution .
Cyclic immunofluorescence ( Gerdes et al . , 2013 ) creates highly multiplexed images using an iterative process ( a cycle ) in which conventional low-plex fluorescence images are repeatedly collected from the same sample and then assembled into a high-dimensional representation . In the implementation described here , samples ~5 µm thick are cut from FFPE blocks , the standard in most histopathology services , followed be dewaxing and antigen retrieval either manually or on automated slide strainers in the usual manner ( Shi et al . , 2011 ) . To reduce auto-fluorescence and non-specific antibody binding , a cycle of ‘pre-staining’ is performed; this involves incubating the sample with secondary antibodies followed by fluorophore oxidation in a high pH hydrogen peroxide solution in the presence of light ( ‘fluorophore bleaching’ ) . Subsequent t-CyCIF cycles each involve four steps ( Figure 1A ) : ( i ) immuno-staining with antibodies against protein antigens ( three antigens per cycle in the implementation described here ) ( ii ) staining with a DNA dye ( commonly Hoechst 33342 ) to mark nuclei and facilitate image registration across cycles ( iii ) four-channel imaging at low- and high-magnification ( iv ) fluorophore bleaching followed by a wash step and then another round of immuno-staining . In t-CyCIF , the signal-to-noise ratio often increases with cycle number due to progressive reductions in background intensity over the course of multiple rounds of fluorophore bleaching . This effect is visible in Figure 1B as the gradual disappearance of an auto-fluorescent feature ( denoted by a dotted white oval and quantified in Figure 1—figure supplement 1; see detailed analysis below ) . When no more t-CyCIF cycles are to be performed , the specimen is stained with H&E to enable conventional histopathology review . Individual image panels are stitched together and registered across cycles followed by image processing and segmentation to identify cells and other structures . t-CyCIF allows for one cycle of indirect immunofluorescence using secondary antibodies . In all other cycles antibodies are directly conjugated to fluorophores , typically Alexa 488 , 555 or 647 ( for a description of different modes of CyCIF see Lin et al . , 2015 ) . As an alternative to chemical coupling we have tested the Zenon antibody labeling method ( Tang et al . , 2010 ) from ThermoFisher in which isotype-specific Fab fragments pre-labeled with fluorophores are bound to primary antibodies to create immune complexes; the immune complexes are then incubated with tissue samples ( Figure 1—figure supplement 2 ) . This method is effective with 30–40% of the primary antibodies that we have tested and potentially represents a simple way to label a wide range of primary antibodies with different fluorophores . Imaging of t-CyCIF samples can be performed on a variety of fluorescent microscopes each of which represent a different tradeoff between data acquisition time , image resolution and sensitivity ( Table 1 ) . Greater resolution ( a higher numerical aperture objective lens ) typically corresponds to a smaller field of view and thus , longer acquisition time for large specimens . Imaging of specimens several square centimeters in area at a resolution of ~1 µm is routinely performed on microscopes specialized for scanning slides ( slide scanners ) ; we use a CyteFinder system from RareCyte ( Seattle WA ) configured with 10 × 0 . 3 NA and 40 × 0 . 6 NA objectives but have tested scanners from Leica , Nikon and other manufacturers . Figure 2A–B show an H&E image of a ~10 × 11 mm metastatic melanoma specimen and a t-CyCIF image assembled from 165 individual image tiles . The assembly process involves stitching sequential image tiles from a single t-CyCIF cycle into one large image panel , flat-fielding to correct for uneven illumination and registration of images from successive t-CyCIF cycles to each other; these procedures were performed using ImageJ , ASHLAR , and BaSiC software as described in materials and methods ( Peng et al . , 2017 ) . In the t-CyCIF image ( Figure 2B ) tumor cells staining positive for S100 ( a melanoma marker in green [Henze et al . , 1997] ) are surrounded by CD45-positive immune cells ( CD45RO+ cells in white ) and by stromal cells expressing the alpha isoform of smooth muscle actin ( α-SMA in red ) . By zooming in on one tile , single cells can be identified and characterized ( Figure 2C ) ; in this image , CD4+ and CD8+ T-lymphocytes and proliferating pRB+ positive cells are visible . At 60X resolution on a confocal GE INCell Analyzer 6000 , kinetochores stain positive for the phosphorylated form of the Aurora A/B/C kinase and can be counted in a mitotic cell ( yellow arrowhead in Figure 2D ) . Nominally super-resolution imaging on a GE OMX Blaze Structured Illumination Microscope ( Carlton et al . , 2010 ) ( using a 60 × 1 . 42 Plan Apo objective ) reveals very fine structural details including differential expression of Lamin isotypes ( in a melanoma , Figure 2E and Figure 2—figure supplement 2 ) and mitotic spindle fibers ( in cells of a xenograft tumor; Figure 2F and Figure 2—figure supplement 2 ) . These data show that t-CyCIF images have readily interpretable features at the scale of an entire tumor , individual tumor cells and subcellular structures . Little subcellular ( or super-resolution ) imaging of clinical FFPE specimens has been reported to date ( but see Chen et al . , 2015 ) , but fine subcellular morphology has the potential to provide dramatically greater information than simple integration of antibody intensities across whole cells . To date , we have tested commercial antibodies against ~200 different proteins for their compatibility with t-CyCIF; these include lineage makers , cytoskeletal proteins , cell cycle regulators , the phosphorylated forms of signaling proteins and kinases , transcription factors , markers of cell state including quiescence , senescence , apoptosis , stress , etc . as well as a variety of non-antibody-based fluorescent stains ( Table 2 ) . Multiplexing antibodies and stains makes it possible to discriminate among proliferating , quiescent and dying cells , identify tumor and stroma , and collect immuno-phenotypes ( Angelo et al . , 2014; Giesen et al . , 2014; Goltsev , 2017 ) . Use of phospho-specific antibodies and antibodies against proteins that re-localize upon activation ( e . g . transcription factors ) makes it possible to assay the states of signal transduction networks . For example , in a 10-cycle t-CyCIF analysis of human tonsil ( Figure 3A ) subcellular features such as membrane staining , Ki-67 puncta ( Cycle 1 ) , ring-like staining of the nuclear lamina ( Cycle 6 ) and nuclear exclusion of NF-ĸB ( Cycle 6 ) can easily be demonstrated ( Figure 3B ) . The five-cycle t-CyCIF data on normal skin in Figure 3C shows tight localization of auto-fluorescence ( likely melanin ) to the epidermis prior to pre-bleaching and images of three non-antibody stains used in the last t-CyCIF cycle: HCS CellMask Red Stain for cytoplasm and nuclei , Actin Red , a Phalloidin-based stain for actin and Mito-tracker Green for mitochondria . In the current work , we rely exclusively on commercial antibodies that have previously been validated using IHC or conventional immunofluorescence; when feasible we confirm that staining by t-CyCIF resembles what has previously been reported for IHC staining . This does not constitute a sufficient level of testing or validation for discovery science or clinical studies and the patterns of staining described in this paper should therefore be considered illustrative of the t-CyCIF approach rather than definitive descriptions; we are currently developing a database of matched t-CyCIF and IHC images across multiple tissues and knockdown cell lines to address this issue and share validation test data with the wider research community . The efficiency of fluorophore inactivation by hydrogen peroxide , light and high pH varies with fluorophore but only minimally with the antibody to which the fluorophore is coupled ( Alexa Fluor 488 is inactivated more slowly than Alexa Fluor 570 or 647; Figure 4B and Figure 4—figure supplement 1 ) . We typically incubate specimens in bleaching conditions for 60 min , which is sufficient to reduce fluorescence intensity by 102 to 103-fold ( Figure 4C ) . When testing new antibodies or analyzing new tissues , imaging is performed after each bleaching step and prior to initiation of another t-CyCIF cycle to ensure that fluorophore inactivation is complete . In preliminary studies , we have tested a range of other fluorophores for their compatibility with t-CyCIF including FITC , TRITC , phycoerythrin , Allophycocyanin , eFluor 570 and eFluor 660 ( eBioscience ) . We conclude that it will be feasible to increase the number of t-CyCIF channels per cycle from four to at least six ( 3 to 5 antibodies plus a DNA stain ) . However , all the images in this paper are collected using a four-channel method . The primary limitation on the number of t-CyCIF cycles that can be performed is the integrity of the tissue: some tissues samples are physically more robust and can withstand more staining and washing procedures than others ( Figure 4D ) . To study the effect of cycle number on tissue integrity , we performed a 10-cycle t-CyCIF experiment on a tissue microarray ( TMA ) comprising a total of 40 cores from 16 different tissues and tumor types . After each t-CyCIF cycle , the number of nuclei remaining was quantified for each core relative to the initial number . For example , Figure 4D shows breast , bladder , lung and prostate cores in which cell number was reduced after 10 cycles by ~2% and an unusually high 46% ( apparent increases in cell number in these data are caused by fluctuation in the performance of cell segmentation routines and are not statistically significant ) . Cells that were lost appear red in these images . The data show that cell loss is often uneven across samples , preferentially affecting regions of tissue with low cellularity . Overall , we found that the extent of cell loss varied with tissue type and , within a single tissue type , from core to core ( six breast cores are shown; Figure 4E ) . For many tissues , we have not yet attempted to optimize cycle number and the experiments performed to date do not fully control for pre-analytical variables ( Vassilakopoulou et al . , 2015 ) such as fixation time and the age of tissue blocks . As a rule , we find that normal tonsil , skin , glioblastoma , ovarian cancer , pancreatic cancer and melanoma can be subjected to >15 cycles with less than 25% cell loss . Figure 4F shows a melanoma specimen subjected to 20 t-CyCIF cycles with good preservation of cell and tissue morphology ( Figure 4G ) . We conclude that t-CyCIF is compatible with multiple normal tissues and tumor types but that some tissues and/or specimens can be subjected to more cycles than others . One requirement for high cycle number appears to be cellularity: samples in which cells are very sparse tend to be more fragile . We expect improvements in cycle number with additional experimentation and the use of fluidic devices that deliver staining and wash liquids more gently . One potential concern about cyclic immunofluorescence is that the process is relatively slow; each cycle takes 6–8 hr and we typically perform one cycle per day . However , a single operator can easily process 30 slides in parallel , and in the case of TMAs , 30 slides can comprise over 2000 different samples . Under these conditions , the most time-consuming step in t-CyCIF is collecting the 200–400 fields of view needed to image each slide . Time could be saved by imaging fewer cells per sample , but the results described below ( demonstrating substantial cellular heterogeneity in a single piece of a tumor resection ) strongly argue in favor of analyzing as large a fraction of each tissue specimen as possible . As a practical matter , data analysis and data interpretation remain more time-consuming than data collection . We also note that the throughput of t-CyCIF compares favorably with other tissue-imaging platforms or single-cell transcriptome profiling . Because t-CyCIF assembles multiplex images sequentially , it is sensitive to factors that alter immunogenicity as cycle number increases . To investigate such effects , we performed a 16-cycle t-CyCIF experiment in which the order of antibody addition was varied between two immediately adjacent tissue slices cut from the same tissue block ( Figure 5A; Slides A and B ) ; the study was repeated three times , once with tonsil and twice with melanoma specimens with similar results ( ~1 . 8 × 105 cells were used for the analysis and overall cell loss was <15% ) . This experiment made it possible to judge: ( i ) the repeatability of staining a single specimen using the same set of antibodies ( Figure 5A , denoted by yellow highlight ) ( ii ) the similarity of staining between slides A and B ( blue highlight ) and ( iii ) the effect of swapping the order of antibody addition ( cycle number ) between slides A and B ( blue lines ) . Comparisons within a single slide were made on a cell-by-cell basis but because slides A and B contain different cells , comparisons between slides were made at the level of intensity distributions ( computed on a per-cell basis following segmentation ) . The repeatability of staining ( as measured in cycles 3 , 7 , 12 and 16 ) was performed using anti-PCNA-Alexa 488 , anti-Vimentin-Alexa 555 and anti-Tubulin- Alexa 647 which bind abundant proteins with distrinct cellular distributions ( Figure 5B ) . Repeated staining of the same antigen is expected to saturate epitopes , but we reasoned that this effect would be less pronounced the more abundant the antigen . For PCNA , the correlation in staining intensities across four cycles was high ( ρ = 0 . 95 to 0 . 99 ) and somewhat lower in the case of Vimentin and Tubulin ( ρ = 0 . 80 to 0 . 95; Figure 6A; a more extensive comparison is shown in Figure 6—figure supplement 1 ) . When we examined the corresponding images , it was readily apparent that Tubulin , and to a lesser extent Vimentin , stained more intensely in later than in earlier t-CyCIF cycles ( see intensity distributions in Figure 6A and images in Figure 6B ) . When images were scaled to equalize the intensity range ( by histogram equalization ) , staining patterns were indistinguishable across all cycles and loss of cells or specific subcellular structures was not obviously a factor ( Figure 6B , left vs right panels and Figure 6C ) . Thus , for at least a subset of antibodies , staining intensity increases rather than decreases with cycle number whereas background fluorescence falls . As a consequence , dynamic range , defined here as the ratio of the least to the most intense 5% of pixels , frequently increases with cycle number ( Figure 6A and Figure 6—figure supplement 1 ) . These effects were reproducible across slides A and B in all three experiments performed . When we compared staining between slides A and B for the same antibodies and cycle number , the overlap in intensity distributions was high ( >0 . 85 ) , demonstrating good sample to sample reproducibility ( Zhou and Liu , 2012 ) . The overlap remained high for the majority of antibodies even when they were used in different cycles on slides A and B , but for some antibodies , signal intensity clearly increased or decreased with cycle number ( Figure 6D; blue and red outlines ) . In the case of eight antibodies for which the effect of cycle number was greatest ( including tubulin , as discussed above ) , the overlap in intensity distributions was <0 . 6 as a consequence of both increases and decreases in staining intensity ( Figure 6E ) . Overall , we found that the repeatability of staining between two biological samples was highest when the antibodies were used in the same cycle on both samples , lower when the antibodies were used in different cycles on the sample , and lowest when both the order and sample were different ( Figure 6F ) . The reasons for changes in staining intensity with cycle number are not known , but the fact that the same changes were observed across multiple experiments ( for any single antibody ) suggests that they arise not from irreproducibility of the t-CyCIF procedure but rather from changes in epitope accessibility . Even in these cases , it appears that it is absolute intensity rather than morphology that is variable . Thus , while changes in staining intensity with cycle number are a concern for a subset of t-CyCIF antibodies , it should be possible to minimize the problem by staining all samples in the same order . Other approaches will also be important; for example , using calibration standards and identifying antibodies exhibiting the least variation with cycle number . One way to reduce artefacts generated by differences in the order of antibody addition is to create a single high-plex antibody mixture and then stain all antigens in parallel . This approach is not compatible with t-CyCIF but is feasible using methods such as MIBI or CODEX ( Angelo et al . , 2014; Goltsev , 2017 ) . However , there is substantial literature showing that the formulation of highly multiplex immuno-assays is complicated by interaction among antibodies ( Ellington et al . , 2010 ) that has a physicochemical explanation in some cases in weak self-association and viscosity ( Wang et al . , 2018 ) . Consistent with these data , we have observed that when eight or more unlabeled antibodies are added to a t-CyCIF experiment , the intensity of staining can fall , although the effect is smaller than observed with antibodies most sensitive to order of addition . We conclude that the construction of sequentially applied t-CyCIF antibody panels and of single high-plex mixtures will both require optimization of specific panels and their method of use . Review of large histopathology specimens by pathologists involves rapid and seamless switching between low-power fields to scan across large regions of tissue and high-power fields to study cellular morphology . To mimic this integration of information at both tissue and cellular scales , we performed eight-cycle t-CyCIF on a large 2 × 1 . 5 cm resection specimen that includes pancreatic ductal adenocarcinoma ( PDAC ) and adjacent normal pancreatic tissue and small intestine ( Figure 7A–C ) . Nuclei were located in the DAPI channel and cell segmentation performed using a watershed algorithm ( Figure 7—figure supplement 1: see Materials and methods section for a discussion of the method and its caveats ) yielding ~2 × 105 single cells each associated with a vector comprising 25 whole-cell fluorescence intensities . Differences in subcellular distribution were evident for many proteins , but for simplicity , we only analyzed fluorescence intensity on a per-antigen basis integrated over each whole cell . Results were visualized by plotting intensity value onto the segmentation data ( Figure 7D ) , by computing correlations on a cell-by-cell basis ( Figure 7E ) , or by using t-distributed stochastic neighbor embedding ( t-SNE ) ( Maaten and Hinton , 2008 ) , which clusters cells in 2D based on their proximity in the 25-dimensional space of image intensity data ( Figure 8A ) . The analysis in Figure 7E shows that E-cadherin , keratin and β-catenin levels are highly correlated with each other , whereas vimentin and VEGFR2 receptor levels are anti-correlated , recapitulating the known dichotomy between epithelial and mesenchymal cell states in normal and diseased tissues . Many other physiologically relevant correlations are also observed , for example between the levels of pERKT202/Y204 ( the phosphorylated , active form of the kinase ) and activating phosphorylation of the downstream kinase pS6S235/S236 ( r = 0 . 81 ) . When t-SNE was applied to all cells in the specimen , we found that those identified during histopathology review as being from non-neoplastic pancreas ( red ) were distinct from PDAC ( green ) and also from the neighboring non-neoplastic small intestine ( blue ) ( Figure 8B–D ) . Vimentin and E-Cadherin had very different levels of expression in PDAC and normal pancreas as a consequence of epithelial-to-mesenchymal transitions ( EMT ) in malignant tissues as well as the presence of a dense tumor stroma , a desmoplastic reaction that is a hallmark of the PDAC microenvironment ( Mahadevan and Von Hoff , 2007 ) . The microenvironment of PDAC was more heavily infiltrated with CD45+ immune cells than the normal pancreas , and the intestinal mucosa of the small intestine was also replete with immune cells , consistent with the known architecture and organization of this tissue . The capacity to image samples that are several square centimeters in area with t-CyCIF can facilitate the detection of signaling biomarker heterogeneity . The WNT pathway is frequently activated in PDAC and is important for oncogenic transformation of gastrointestinal tumours ( Jones et al . , 2008 ) . Approximately 90% of sporadic PDACs also harbor driver mutations in KRAS , activating the MAPK pathway and promoting tumourigenesis ( Vogelstein et al . , 2013 ) . Studies comparing these pathways have come to different conclusions with respect to their relationship: some studies show concordant activation of MAPK and WNT signaling and others argue for exclusive activation of one pathway or the other ( Jeong et al . , 2012 ) . In t-SNE plots derived from images of PDAC , multiple sub-populations of cells representing negative , positive or no correlation between pERK and β-catenin levels can be seen ( marked with labels ‘a’ , ‘b’ or ‘c’ , respectively in Figure 8A ) . The same three relationships can be found in non-neoplastic pancreas and small intestine ( Figures 8A and 7C ) . In PDAC , malignant cells can be distinguished from stromal cells , to a first approximation , by high proliferative index , which can be measured by staining for Ki-67 and PCNA ( Bologna-Molina et al . , 2013 ) . When we gated for cells that were both Ki67high and PCNAhigh , and thus likely to be malignant , the co-occurrence of different relationship between pERK and β-catenin levels on a cellular level was again evident . While we cannot exclude the possibility of phospho-epitope loss during sample preparation , it appears that the full range of possible relationships between the MAPK and WNT signaling pathways described in the literature can be found within a specimen from a single patient , illustrating the impact of tissue context on the activities of key signal transduction pathways . Immuno-oncology drugs , including immune checkpoint inhibitors targeting CTLA-4 and the PD-1/PD-L1 axis are rapidly changing the therapeutic possibilities for traditionally difficult-to-treat cancers including melanoma , renal and lung cancers , but responses are variable across and within cancer types . The hope is that tumor immuno-profiling will yield biomarkers predictive of therapeutic response in individual patients . For example , expression of PD-L1 correlates with responsiveness to the ICIs pembrolizumab and nivolumab ( Mahoney and Atkins , 2014 ) but the negative predictive value of PD-L1 expression alone is insufficient to stratify patient populations ( Sharma and Allison , 2015 ) . In contrast , by measuring PD-1 , PD-L1 , CD4 and CD8 by IHC on sequential tumor slices , it has been possible to identify some immune checkpoint inhibitor-responsive melanom patients ( Tumeh et al . , 2014 ) . To test t-CyCIF in this application , eight-cycle imaging was performed on a 1 × 2 cm specimen of clear-cell renal cell carcinoma using 10 antibodies against multiple immune markers and 12 against other proteins expressed in tumor and stromal cells ( Figure 9A–B; Supplementary file 4 ) . A region of the specimen corresponding to tumor was readily distinguishable from non-malignant stroma based on α-SMA expression ( α-SMAhigh regions denote stroma and α-SMAlow regions high density of malignant cells ) . In the α-SMAlow domain , CD3+ or CD8+ lymphocytes were fourfold enriched ( Figure 9C ) and PD-1 and PD-L1-positive cells were 13 to 20-fold more prevalent as compared to the surrounding tumor stroma ( α-SMAhigh domain ) ; CD3+ CD8+ double positive T-cells were found almost exclusively in the tumor . Suppression of immune cells is mediated by binding of PD-L1 ligand , which is commonly expressed by tumor cells , to the PD1 receptor expressed on immune cells ( Tumeh et al . , 2014 ) . To begin to estimate the likelihood of ligand-receptor interactions , we quantified the degree of co-localization of cells expressing the two molecules . The centroids of PD-1+ or PD-L1+ cells were determined from images ( PD-1 , red; PD-L1 , green , Figure 9E ) and co-localization ( highlighted in yellow , Figure 9F ) computed by k-nearest neighbor analysis . We found that co-localization of PD-1/PD-L1 was ~2 . 7-fold more likely ( Figure 9—figure supplement 1 ) in tumor and stroma and was concentrated on the tumor-stroma border consistent with previous reports on melanoma ( Tumeh et al . , 2014 ) . These data demonstrate the potential of spatially resolved immuno-phenotyping to quantify state and location of tumor infiltrating lymphocytes; such data may ultimately yield biomarkers predictive of sensitivity to immune checkpoint inhibitor ( Tumeh et al . , 2014 ) . To explore the general utility of t-CyCIF in a range of healthy and cancer tissues we applied eight cycle t-CyCIF to TMAs containing 39 different biopsies from 13 healthy tissues and 26 biopsies corresponding to low- and high-grade cancers from the same tissue types ( Figure 10A and Figure 10—figure supplement 1 , Supplementary file 3 for antibodies used , Supplementary file 5 for TMA details and naming conventions ) and then performed t-SNE and clustering on single-cell intensity data ( Figure 10B ) . The great majority of TMA samples mapped to one or a few discrete locations in the t-SNE projection ( compare normal kidney tissue - KI1 , low-grade tumors - KI2 , and high-grade tumors – KI3; Figure 10C ) , although ovarian cancers were scattered across the t-SNE projection ( Figure 10D ) ; overall , there was no separation between normal tissue and tumors regardless of grade ( Figure 10E ) . In a number of cases , high-grade cancers from multiple different tissues of origin co-clustered , implying that transformed morphologies and cell states were closely related . For example , while healthy and low-grade pancreatic and stomach cancer occupied distinct t-SNE domains , high-grade pancreatic and stomach cancers were intermingled and could not be readily distinguished ( Figure 10F ) , recapitulating the known difficulty in distinguishing high-grade gastrointestinal tumors of diverse origin by histophathology ( Varadhachary and Raber , 2014 ) . Nonetheless , t-CyCIF might represent a means to identify discriminating biomarkers by efficiently sorting through large numbers of alternative antigens and antigen localizations . Data from single-cell genomics reveals extensive heterogeneity in many types of cancer ( Turner and Reis-Filho , 2012 ) but our understanding of this phenomenon requires spatially resolved data ( Giesen et al . , 2014 ) . We performed eight-cycle imaging on a 2 . 5 cm x 1 . 8 mm resected glioblastoma ( GBM ) specimen imaging markers of neural development , cell cycle state and signal transduction ( Figure 11A–B , Supplementary file 6 ) . GBM is a highly aggressive and genetically heterogeneous ( Brennan et al . , 2013 ) brain cancer commonly classified into four histologic subtypes ( Olar and Aldape , 2014 ) . Following image segmentation , phenotypic heterogeneity was assessed at three spatial scales corresponding to: ( i ) 1 . 6 × 1 . 4 mm fields of view ( 252 total ) each of which comprised 103 to 104 cells ( ii ) seven macroscopic regions of ~104 to 105 cells each , corresponding roughly to tumor lobes and ( iii ) the whole tumor comprising ~106 cells . To quantify local heterogeneity , we computed the informational entropy on a-per-channel basis for 103 randomly selected cells in each field ( Figure 11C; see online Materials and methods for details ) . In this setting , informational entropy is a measure of cell-to-cell heterogeneity on a mesoscale corresponding to 10–30 cell diameters . For a marker such as EGFR , which can function as a driving oncogene in GBM , informational entropy was high in some areas ( Figure 11C; red dots ) and low in others ( blue dots ) . Areas with high entropy in EGFR abundance did not co-correlate with areas that were most variable with respect to a downstream signaling protein such as pERK . Thus , the extent of local heterogeneity varied with the region of the tumor and the marker being assayed . Semi-supervised clustering using expectation–maximization Gaussian mixture ( EMGM ) modeling of all cells in the tumor yielded eight distinct clusters , four of which encompassed 85% of all cells ( Figure 12A and Figure 12—figure supplement 1 ) . Among these , cluster one had high EGFR levels , cluster two had high NGFR and Ki67 levels and cluster six had high levels of vimentin; cluster five was characterized by high keratin and pERK levels . The presence of four highly populated t-CyCIF clusters is consistent with data from single-cell RNA-sequencing of ~400 cells from five GBMs ( Patel et al . , 2014 ) . Three of the t-CyCIF clusters have properties reminiscent of established histological subtypes including: classical , cluster 1; pro-neural , cluster 3; and mesenchymal , cluster 6 , but additional work will be required to confirm such assignments . To study the relationship between phenotypic diversity and tumor architecture , we mapped each cell to an EMGM cluster ( denoted by color ) . Extensive intermixing was observed at all spatial scales ( Figure 12B ) . For example , field of view 147 was highly enriched for cells corresponding to cluster 5 ( yellow ) , but a higher magnification view revealed extensive intermixing of four other cluster types on a scale of ~3–5 cell diameters ( Figure 12C ) . At the level of larger , macroscopic tumor regions , the fraction of cells from each cluster also varied dramatically ( Figure 12D ) . None of these findings was substantially different when the number of clusters was set to 12 ( Figure 12—figure supplement 2 ) . These results have several implications . First , they suggest that GBM is phenotypically heterogeneous on a spatial scale of 5–1000 cell diameters and that cells corresponding to distinct t-CyCIF clusters are often found in the vicinity of each other . Second , sampling a small region of a large tumor has the potential to misrepresent the proportion and distribution of tumor subtypes , with implications for prognosis and therapy . Similar concepts likely apply to other tumor types with high genetic heterogeneity , such as metastatic melanoma ( Tirosh et al . , 2016 ) , and are therefore relevant to diagnostic and therapeutic challenges arising from tumor heterogeneity .
PKS is a member of the Scientific Advisory Board of RareCyte Inc . , which manufactures the CyteFinder slide scanner used in this study; research with RareCyte is funded by NIH grant R41 CA224503 ( PI E . Kaldjian ) . PKS is also co-founder of Glencoe Software , which contributes to and supports the open-source OME/OMERO image informatics software used in this paper . Other authors have no competing financial interests to disclose .
Formalin fixed and paraffin embedded ( FFPE ) tissues from were retrieved from the archives of the Brigham and Women’s Hospital as part of discarded/excess tissue protocols or obtained from commercial vendors . The Institutional Review Board ( IRB ) of the Harvard Faculty of Medicine last reviewed the research described in this paper on 2/16/2018 ( under IRB17-1688 ) and judged it to ‘involve no more than minimal risk to the subjects’ and thus eligible for a waiver of the requirement to obtain consent as set out in 45CFR46 . 116 ( d ) . Tumor tissue and FFPE specimens were collected from patients under IRB-approved protocols ( DFCI 11–104 ) at Dana-Farber Cancer Institute/Brigham and Women’s Hospital , Boston , Massachusetts . Tonsil samples used in Figure 1 were purchased from American MasterTech ( CST0224P ) . Tissue microarrays for analyses in Figure 4D and E were obtained from Biomax ( Cat . MTU481 ) ; detailed information can be found online at https://www . biomax . us/tissue-arrays/Multiple_Organ/MTU481 . Tissue microarrays ( TMA ) for diverse healthy tissues and tumor analyses were obtained from Protein Biotechnologies ( Cat . TMA-1207 ) . All conjugated and unconjugated primary antibodies used in this study are listed in Table 2 . Indirect immunofluorescence was performed using secondary antibodies conjugated with Alexa-647 anti-Mouse ( Invitrogen , Cat . A-21236 ) , Alexa-555 anti-Rat ( Invitrogen , Cat . A-21434 ) and Alexa-488 anti-Rabbit ( Invitrogen , Cat . A-11034 ) . 10 mg/ml Hoechst 33342 stock solution was purchased from Life Technologies ( Cat . H3570 ) . 20xPBS was purchased from Santa Cruz Biotechnology ( Cat . SC-362299 ) . 30% hydrogen peroxide solution was purchased from Sigma-Aldrich ( Cat . 216763 ) . PBS-based Odyssey blocking buffer was purchased from LI-COR ( Cat . 927–40150 ) . All reagents for the Leica BOND RX were purchased from Leica Microsystems . HCS CellMask Red Stain and Mito-tracker Green stains were purchased from ThermoFischer ( catalog numbers H32712 , R37112 and M751 , respectively ) . All primary antibodies ( fluorophore-conjugated and unconjugated ) were diluted in Odyssey blocking buffer . Slides carrying tissues that had been subjected to pre-staining , or to a previous t-CyCIF stain and bleach cycle , were incubated at 4°C for ~12 hr with diluted primary or fluorophore-conjugated antibody ( 250–500 μl per slide ) in a moisture chamber . Long incubation times were a matter of convenience and many antibodies only require short incubation with sample . Slides were then washed four times in 1x PBS by dipping in a series of vertical staining jars . For indirect immunofluorescence , slides were incubated in diluted secondary antibodies in a moisture chamber for 1 hr at room temperature followed by four washes with 1xPBS . Slides were incubated in Hoechst 33342 at 2 μg/ml in Odyssey blocking buffer for 15 min at room temperature , followed by four washes in 1xPBS . Stained slides were mounted prior to image acquisition ( see the Mounting section below ) . Stained slides from each round of CyCIF were imaged with a CyteFinder slide scanning fluorescence microscope ( RareCyte Inc . Seattle WA ) using either a 10X ( NA = 0 . 3 ) or 40X long-working distance objective ( NA = 0 . 6 ) . Imager5 software ( RareCyte Inc . ) was used to sequentially scan the region of interest in four fluorescence channels . These channels are referred to by the manufacturer as a: ( i ) ‘DAPI channel’ with an excitation filter having a peak of 390 nm and half-width of 18 nm and an emission filter with a peak of 435 nm and half-width of 48 nm; ( ii ) ‘FITC channel’ having a 475/28 nm excitation filter and 525/48 nm emission filter ( iii ) ; ‘Cy3 channel’ having a 542/27 nm excitation filter and 597/45 nm emission filter and ( iv ) ; ‘Cy5 channel’ having a 632/22 nm excitation filter and 679/34 nm emission filter . Imaging was performed with 2 × 2 binning to increase sensitivity , shorten exposure time and reduce photo bleaching . We have tested slide scanners from several other manufacturers ( e . g . a Leica Aperio Digital Pathology Slide Scanner , GE IN-Cell Analyzer 6000 and GE Cytell Cell Imaging System ) and found that they too can be used to acquire images from samples processed by t-CyCIF . Slides can also be analyzed on conventional microscopes , but the field of view is typically smaller , and an automated stage is required for accurate stitching of individual fields of view into a complete image of a tissue . We acquired 3D-SIM images on a Deltavision OMX V4 Blaze ( GE Healthcare ) with a 60x/1 . 42N . A . Plan Apo oil immersion objective lens ( Olympus ) and three Edge 5 . 5 sCMOS cameras ( PCO ) . Two to three micron z-stacks were collected with a z-step of 125 nm or 250 nm and with 15 raw images per plane . To minimize spherical aberration , immersion oil matching was used for each sample as described by Hiraoka et al . ( 1990 ) . except that we measured point spread functions of point-like structures within the sample as opposed to beads on a separate slide . DAPI fluorescence was excited with a 405 nm laser and collected with a 477/35 emission filter , Alexafluor 488 with a 488 nm laser and a 528/48 emission filter , Alexa fluor 555 with a 568 nm laser and a 609/37 emission filter , and Alexa fluor 647with a 642 nm laser and a 683/40 emission filter . All stage positions were saved in softWorX to be revisited later . Super-resolution images were computationally reconstructed from the raw data sets with a channel-specific , measured optical transfer function and a Wiener filter constant of 0 . 001 using CUDA-accelerated 3D-SIM reconstruction code based on Gustafsson et al . ( 2008 ) . A comparison of properties of different imaging platforms used in this study are shown in Table 1 . Quantitative analysis of tissue images is challenging , in large part because cells are close together and embedded in a complex extracellular environment . Background can be uneven across large images and signal-to-noise ratios relatively low , particularly in the case of tissues with high auto-fluorescence and low signal antibodies ( e . g . phospho-protein antibodies ) . We have only started to tackle these issues in the case of high-dimensional t-CyCIF data and users are encouraged to check for updates on www . cycif . org and implement their own approaches . We purchased a TMA ( MTU481 , Biomax Inc , https://www . biomax . us/tissue-arrays/Multiple_Organ/MTU481 ) to test the impact of cycle number on tissue integrity . Images were captured and processed as described above . The registered image stacks were then segmented and nuclei counts for each core and each cycle were recorded . All values were normalized to the number of nuclei from the first cycle of a particular core biopsy and the fractional normalized nuclei count shown at each staining cycle . To compare staining patterns between different cycles within the same specimen , we calculated overlap integrals . First , we determined the distribution of intensity data averaged over each single cell and for each t-CyCIF cycles . The area under the curve of these distributions was calculated by trapezoidal numerical integration using ‘trapz’ function in Matlab ( Gustafsson et al . , 2008 ) . The ratio of the area under the curve ( AUC ) for different cycles , samples or antibodies was calculated and the overlap scores then computed as:Overlapscore=overlapAUC/totalAUC The dynamic range ( DR ) of fluorescence intensities for a given antibody was calculated as a rough estimate of the signal-to-noise ratio; SNR . The calculation was performed as follows: first , pixel-by-pixel intensity data was extracted from a t-CyCIF image; the DR was then calculated as the ratio of the intensities of the 95th and 5th percentile values and represented on a log scale . High DR values indicate a favorable SNR . Intensities below the 5th percentile were considered to be background noise . Raw intensity data generated from registered and segmented images were imported into Matlab and converted to comma separated value ( csv ) files . The viSNE implementation of t-SNE and EMGM algorithms from the CYT single-cell analysis package were obtained from the Pe’er laboratory at Columbia University ( Amir et al . , 2013 ) . Intensity-based measurements ( such as flow cytometry or imaging cytometry ) of protein expression have approximately log-normal distribution ( Bagwell , 2005 ) , hence , t-CyCIF raw intensity values were first transformed in log or in inverse hyperbolic sine ( asinh ) using the default Matlab function or the CYT package ( Amir et al . , 2013 ) , respectively . Between-sample variation was normalized on a per-channel basis by using the CYT package to align intensity measurements that encompass values between 1st and the 99th percentile . Data files were aggregated and used to generate viSNE plots . All viSNE/t-SNE analyses used the following settings: perplexity −30 , epsilon = 500 , lie factor = 4 for initial 100 iterations and lie factor −1 for remaining iterations . To determine whether PD-1 and PD-L1 expressing cells are sufficiently close for the receptor and ligand to interact , the spatial densities for PD1+ and PDL1+ cells were estimated using a k nearest neighbors ( kNN ) model with k = 4 , corresponding to a ~10 µm smoothing window . Since the density in space of the PD1+ or PDL1+ cells at any point in that space is proportional to the probability of that cell having a centroid there , the co-occurrence probability at a point was therefore proportional to the product of the spatial densities for both cell types at a point . To normalize for the difference in total PDL1+ or PD1+ cells between regions of the tissue corresponding to tumor and stroma , we calculated spatial probabilities for the different regions in the specimen separately . Figure 9—figure supplement 1 shows the distribution of co-occurrence densities for stroma and tumor relevant to a clear-cell carcinoma shown in Figure 9 . Images were divided into regular grids and 1000 cells from each region used to calculate the non-parametric Shannon entropy as follows:ShanonEntropy ( s ) = −∑isi2log ( si2 ) where si is the per-pixel intensity of signal s at a given point . Normalized Shannon entropy as calculated as Enormalized = Eregion/Esample . To determine an appropriate number of clusters ( k ) for analysis of the GBM tumor shown in Figures 11 and 12 and in Figure 12—figure supplement 2 we determined negative log-likelihood-ratios for various values of k . For each choice of cluster number n , the likelihood-ratio was calculated for a Gaussian mixture model with n = k-1 and with n = k and the ratio then plotted relative to k . The EMGM algorithm was initialized 30 times for each value of k and it converged in all instances . The inflection at k = 8 ( red arrow ) suggests that inclusion of additional clusters ( k > 8 ) explains a smaller , distinct source of variation in the data ( Figure 12—figure supplement 1 ) . As an alternative , k = 12 was also explored in Figure 12—figure supplement 2 . Intensity values from all antibody channels ( plus area and Hoechst intensity ) were used for clustering . All data generated or analyzed during this study are included in the manuscript and supporting files . Intensity data used to generate figures is available in supplementary materials and can be downloaded from the HMS LINCS Center Publication Page ( http://lincs . hms . harvard . edu/lin-elife-2018/ ) ( RRID:SCR_016370 ) . Code and scripts used in this study are listed in Key resources table and also on-line at the HMS LINCS Center publication page ( http://lincs . hms . harvard . edu/lin-elife-2018/ ) . ImageJ is available at https://imagej . nih . gov/ij/ BaSic is available at https://www . helmholtz-muenchen . de/icb/research/groups/quantitative-single-cell-dynamics/software/basic/index . html . Matlab scripts used in this paper and the ASHLAR registration/stitching algorithm is available at our GitHub repositories ( https://github . com/sorgerlab/cycif and https://github . com/sorgerlab/ashlar ( Muhlich , 2018; Lin , 2018 ) . A Jupyter notebook for futher exploration of data in Figures 5 and 6 is available at https://github . com/sorgerlab/lin_elife_2018_tCyCIF_plots ( Muhlich and Wang , 2018; copy archived at https://github . com/elifesciences-publications/lin_elife_2018_tCyCIF_plots ) . All images can be obtained from an OMERO image database via links found at the HMS LINCS Center Publication Page http://lincs . hms . harvard . edu/lin-elife-2018/ ( RRID: SCR_016370 ) . Stitched and registered image composites can be obtained at www . cycif . org . ( RRID:SCR_016267 ) and via links found there . | To diagnose a disease such as cancer , doctors sometimes take small tissue samples called biopsies from the affected area . These biopsies are then thinly sliced and treated with dyes to identify healthy and cancerous cells . However , clinicians and scientists often need to look into what happens inside individual cells in the tissues so they can understand how cancers arise and progress . This helps them to identify different types of tumor cells and to tailor the best treatment for the patient . To do so , a number of proteins ( the molecules involved in nearly all life’s processes ) need to be tracked in healthy and diseased cells and tissues . This can be done thanks to a range of methods known as immunofluorescence microscopy , but following different proteins on the same slice of a sample is difficult . However , a new type of immunofluorescence known as t-CyCIF may be a solution . With this technique , a fluorescent compound is applied that will bind to a specific protein of interest . A microscope can pick up the light from the compound when the sample is imaged , which reveals the protein’s location in the cell or tissue . Then , a substance is used that deactivates the fluorescence signal . After this , another compound that binds to a new type of protein is used , and imaged . This cycle is repeated several times to locate different proteins . Lastly , the individual images are processed and stitched together to reveal the cells and their internal structures . Here , Lin , Izar et al . showed that t-CyCIF could be used to study biopsies and to obtain images that covered a large area of healthy human tissues and tumors . The technique helped to track over 60 different proteins in normal and tumor tissue samples from human patients . Several sets of experiments showed that t-CyCIF could uncover the molecular mechanisms that are disrupted during cancer , but also reveal the complexity of a single tumor . In fact , as shown with biopsies of brain cancer , cancerous cells in a tumor can be strikingly different , even when they are close to each other . Finally , the method helped to pinpoint which types of immune cells are involved in fighting a kidney tumor . Overall , such information cannot be obtained with conventional methods , yet is crucial for diagnosis and treatment . Most laboratories can readily use t-CyCIF since the technique is open source and requires equipment that is easily accessible . In fact , the technique should soon be used to assess how well certain drugs help the immune system combat cancer . Ultimately , better use of biopsies is key to customizing cancer care . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"tools",
"and",
"resources",
"cancer",
"biology"
] | 2018 | Highly multiplexed immunofluorescence imaging of human tissues and tumors using t-CyCIF and conventional optical microscopes |
RNA viruses rapidly diversify into quasispecies of related genotypes . This genetic diversity has long been known to facilitate adaptation , but recent studies have suggested that cooperation between variants might also increase population fitness . Here , we demonstrate strong cooperation between two H3N2 influenza variants that differ by a single mutation at residue 151 in neuraminidase , which normally mediates viral exit from host cells . Residue 151 is often annotated as an ambiguous amino acid in sequenced isolates , indicating mixed viral populations . We show that mixed populations grow better than either variant alone in cell culture . Pure populations of either variant generate the other through mutation and then stably maintain a mix of the two genotypes . We suggest that cooperation arises because mixed populations combine one variant’s proficiency at cell entry with the other’s proficiency at cell exit . Our work demonstrates a specific cooperative interaction between defined variants in a viral quasispecies .
The evolution of RNA viruses is characterized by high mutation rates and large population sizes , which together create genetically diverse populations known as quasispecies ( Eigen , 1971; Holland et al . , 1982; Lauring and Andino , 2010; Andino and Domingo , 2015 ) . High levels of standing genetic diversity can provide a substrate for selection and rapid adaptation , an advantage for viruses that experience strong and varied selective pressures to escape immune recognition , develop drug resistance , and adapt to new hosts ( Najera et al . , 1995; Pfeiffer and Kirkegaard , 2005; Dutta et al . , 2008 ) . Recently , several studies have suggested that cooperative interactions between variants in a quasispecies can also increase population-level fitness ( Vignuzzi et al . , 2006; Ciota et al . , 2012; Shirogane et al . , 2012; Ke et al . , 2013; Bordería et al . , 2015 ) . Vignuzzi et al . , 2006 found that genetically diverse poliovirus populations were required for wild-type neurotropism and pathogenesis , leading the authors to suggest that unknown cooperative interactions among minor variants promoted the overall fitness of the population . More recent studies have suggested cooperation between distinct variants of measles ( Shirogane et al . , 2012 ) , hepatitis B virus ( Cao et al . , 2014 ) , and Coxsackie virus ( Bordería et al . , 2015 ) . However , specific examples of robust cooperative interactions between defined variants in viral quasispecies remain rare ( Holmes , 2010 ) . Here , we demonstrate that cooperation between two distinct variants of human H3N2 influenza promotes viral growth in cell culture . The two variants differ by a single amino-acid mutation in the neuraminidase ( NA ) protein , which normally mediates viral exit from the host cell . Both variants are reported numerous times in human H3N2 NA sequences deposited in the GISAID EpiFlu database , and both have been observed in mixed populations when clinical specimens are passaged in cell culture . We show that the two variants grow better together than apart , and that serial passage repeatably selects for mixed populations . We suggest that the cooperation arises because one variant is proficient at cell entry while the other is proficient at cell exit . Overall , our work represents a clear example of selection to generate and maintain two cooperating genotypes within a viral quasispecies .
Over the last decade , several groups have reported that mutations arise rapidly and repeatedly at residue NA 151 when human H3N2 influenza is passaged in cell culture ( Table 1 ) ( McKimm-Breschkin et al . , 2003; Lin et al . , 2010; Tamura et al . , 2013; Lee et al . , 2013; Chambers et al . , 2014; Mishin et al . , 2014; Mohr et al . , 2015 ) . Residue 151 is in the NA active site and is highly conserved; until recently , it had an amino-acid identity of D in virtually all N2 NAs . Ordinarily , NA mediates viral exit from the host cell by cleaving sialic-acid receptors to release newly produced virions . The D151G mutation ablates the catalytic activity of NA and instead causes it to bind the receptors that it typically cleaves ( Zhu et al . , 2012 ) . Mutations at this site seem to be more common in viruses that have been passaged in cell culture compared to the original clinical isolates ( Deyde et al . , 2009; Lin et al . , 2010; Okomo-Adhiambo et al . , 2010; Tamura et al . , 2013; Lee et al . , 2013; Chambers et al . , 2014; Mishin et al . , 2014 ) . As a result , mutations at site 151 have been categorized as lab adaptations ( Okomo-Adhiambo et al . , 2010; Tamura et al . , 2013; Lee et al . , 2013; Mishin et al . , 2014 ) . 10 . 7554/eLife . 13974 . 003Table 1 . Prior reports of variation at neuraminidase site 151 when H3N2 clinical specimens are passaged in cell culture . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 003ReferenceSummary ( McKimm-Breschkin et al . , 2003 ) Sanger sequencing of 38 oseltamivir- and zanamivir-resistant MDCK-passaged clinical isolates found that 7 had G , N , E , or V at site 151 . ( Lin et al . , 2010 ) Sanger sequencing of 18 isolates after passage in MDCK or MDCK-SIAT1 cells found 4 isolates as D+G , 3 as D+N , and 2 as D+A . Pyrosequencing detected low frequencies of G151 and N151 in some clinical samples . ( Tamura et al . , 2013 ) Pyrosequencing of 150 isolates after 1-4 passages in MDCK cells found that 85% developed mixed populations at site 151; 29% did so after a single passage . Mixed populations consisted of D+N , D+G , D+G+N , and D+G+A genotypes . T148I/K/P mutations were also observed in 23% of isolates . ( Lee et al . , 2013 ) 77 clinical specimens were Sanger-sequenced before and after a single passage in MDCK cells . 18 acquired a mutation at NA site 151: 10 were D+N , 7 were D+G and one fixed D151N at the limit of detection . No mutations were detected in the unpassaged specimens . ( Chambers et al . , 2014 ) 9 A/Victoria/361/11-like clinical specimens were passaged twice in MDCK cells and Sanger-sequenced before and after expansion . 4 isolates developed NA-dependent cell binding; 3 had D151G , the other D151N . ( Mishin et al . , 2014 ) Pyrosequencing of 150 MDCK-grown isolates found that 42 were D+G , 34 were D+N , and 57 were D+G+N . Pyrosequencing of 50 matched clinical specimens detected no variation at site 151 . ( Mohr et al . , 2015 ) 16 pairs of isolates cultured in parallel in MDCK cells and in eggs were sequenced using Ion Torrent . 5 MDCK isolates were D+N , 4 were D+G , and 2 were D+N+G . No egg-passaged isolates had mutations at site 151 . T148I/K mutations were observed in 7 MDCK isolates . We examined whether mutations at site 151 exhibited patterns consistent with simple lab adaptation by determining the frequencies of amino acids at this position in human H3N2 NA sequences in the GISAID EpiFlu database for each year from 2000 to 2014 ( Figure 1 ) . Most isolates in this database are first passaged in eggs or cell culture , and then the consensus sequence of the viral population is determined by Sanger sequencing . Beginning in 2007 , the frequency of mutations at NA site 151 rose dramatically , with mutant genotypes representing about a quarter of the sequences . G151 and N151 are each reported in about 1% of sequences , but ambiguous nucleotide calls at the codon make up the majority of non-wild-type sequences . Because these sequences usually represent consensus calls from Sanger sequencing , the ambiguous nucleotides likely indicate the presence of mixed D151+G151 and D151+N151 populations . The relative abundance of mixed populations in strains deposited in the GISAID EpiFlu database is consistent with the fact that , in tissue culture , mutations at site 151 arise frequently but fix rarely ( Table 1 ) ( Lin et al . , 2010; Okomo-Adhiambo et al . , 2010; Tamura et al . , 2013; Lee et al . , 2013; Mishin et al . , 2014; Mohr et al . , 2015 ) . 10 . 7554/eLife . 13974 . 004Figure 1 . Ambiguous identities are common at NA site 151 after 2007 . ( A ) Shown are the number of human H3N2 influenza NA sequences in the GISAID EpiFlu database with the given identity at site 151 for each year from 2000 to 2014 . Since 2007 , ambiguous amino-acid identities have been present at residue 151 in about 20% of sequences . Sequences from ( B ) 2000 to 2006 and ( C ) 2007 to 2014 were classified into groups based on their passage history . Ambiguous amino-acid identities were present almost exclusively in isolates that had been passaged in cell culture . Sequences were classified as 'undetermined' if the passage history was difficult to interpret and as 'not listed' if the passage history was absent altogether . Mixed genotypes were inferred on the basis of IUPAC nucleotide ambiguity codes; for instance , the triplet GRT could refer to GAT or GGT , corresponding to amino acids D and G , respectively . Genotypes are indicated if they exceeded a frequency of 0 . 5% among all analyzed sequences; otherwise , they are categorized as 'other . ' The computer code used for analysis is available in Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 00410 . 7554/eLife . 13974 . 005Figure 1—source data 1 . This 7-zip archive contains the source code used for Figure 1 ( the analysis of mutation frequencies at site 151 in naturally occurring sequences ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 005 When viruses are passaged , they experience culture-specific selective pressures . To understand how variation at site 151 might depend on the passaging procedures , we classified sequences according to their passage histories as annotated in the GISAID EpiFlu database . Prior to 2007 , site 151 is almost always reported to be D , regardless of passage history ( Figure 1B ) . From 2007 onward , variants at site 151 are reported almost exclusively in isolates that have been passaged in cell culture ( Figure 1C ) . Whereas nearly a third of isolates that have been passaged in cell culture are reported to have an amino acid other than D at site 151 , only two of nearly 1800 unpassaged isolates and five of nearly 400 egg-passaged isolates are reported to have non-D amino acids at site 151 . These observations accord with reports in the literature that mutations at site 151 are observed primarily in cell-culture-passaged isolates ( Tamura et al . , 2013; Lee et al . , 2013; Mohr et al . , 2015 ) . However , it is important to remember that the methods used to determine most sequences in the GISAID EpiFlu database lack sensitivity to detect minority variants in a viral population . For instance , the experimental results that we describe below suggest that it is exceedingly unlikely that any of the more than 100 reported G151 sequences actually reflect the complete fixation of this mutation . Overall , the results in Figure 1 indicate that the G151 mutation tends to occur in mixed populations . We therefore sought to experimentally characterize the growth of pure D151 and G151 viral variants to determine whether mixed populations represent incomplete fixation of a lab-adaptation mutation or whether they are the product of active selection . We compared the growth of the D151 and G151 variants both alone and in mixed populations by generating viruses of defined genotypes using reverse genetics . The A/Hanoi/Q118/2007 strain , henceforth referred to as Hanoi/2007 , is a human H3N2 strain with a G151 genotype . Its NA protein sequence is identical to that of several other sequenced isolates , except that the other strains have D at site 151 . We created reverse-genetics plasmids encoding the protein sequences of both the D151 and G151 variants of the Hanoi/2007 NA , as well as the HA from this strain . The internal genes were derived from the lab-adapted A/WSN/33 influenza strain with GFP packaged in the PB1 segment ( Bloom et al . , 2010 ) . The two viral variants were therefore isogenic except for the variation at site 151 . We generated virus using reverse genetics by co-transfecting cells with plasmids encoding the D151 variant , the G151 variant , or an equal mix of the two , together with isogenic plasmids for the other viral genes , and we quantified the resulting titers ( Figure 2A ) . Surprisingly , given its widespread designation as a lab adaptation , the G151 variant grows extremely poorly . However , a mixed population of D151 and G151 variants grows to substantially higher titers than the corresponding pure population of D151 viruses . The growth advantage of the mixed population suggests that cooperation between the two variants improves viral growth . 10 . 7554/eLife . 13974 . 006Figure 2 . Mixed populations grow to higher titers than either pure population alone . ( A ) Pure and mixed populations were generated by reverse genetics . Cells were transfected with a Hanoi/2007 NA plasmid encoding D151 , G151 , or an equal mix of the two , along with isogenic plasmids for the other genes . The total amount of NA plasmid was the same in all cases; that is , the pure populations were transfected with 250 ng of the indicated variant , and the mixed populations were transfected with 125 ng of each variant . The HA was also derived from Hanoi/2007 , and the other genes were derived from the lab-adapted A/WSN/33 strain with GFP packaged in the PB1 segment . The titer was determined after 72 hr using the GFP reporter . Black lines indicate the mean and standard error of the titers for three biological replicates , with titers for each replicate plotted as points . Figure 2—figure supplement 1 shows a comparable effect when the virus does not package GFP . Figure 2—figure supplement 2 shows that growth of the G151 variant is improved by adding oseltamivir . ( B ) Cells were infected at an MOI of 0 . 2 with pure D151 virus , pure G151 virus , or an equal mix of the two . The total MOI of infecting virus was the same in all cases . The main plots show titers averaged across three biological replicates , with each replicate plotted individually in the small insets . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 00610 . 7554/eLife . 13974 . 007Figure 2—figure supplement 1 . A mixed population outgrows either pure population when viruses are generated by reverse genetics with an unmodified PB1 gene . The data here differ from Figure 2A in that the virus populations were generated by reverse genetics using the unmodified A/WSN/33 PB1 gene rather than the PB1 segment modified to package GFP . Titers were determined at 74 hr post-transfection by staining for nucleoprotein in infected cells . Black lines indicate the mean and standard error of the titers for three biological replicates , with titers for each replicate plotted as points . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 00710 . 7554/eLife . 13974 . 008Figure 2—figure supplement 2 . Growth of the G151 variant is improved by adding oseltamivir during the generation of viral populations by reverse genetics . Presumably , this improvement occurs because oseltamivir blocks the binding of G151 NA to receptor , allowing newly formed virions to be released more efficiently . Oseltamivir also slightly increases the growth of the D151 variant . We speculate this is because oseltamivir blocks receptor cleavage by the D151 NA , leaving more receptors that can be bound by HA during secondary viral replication . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 008 We next sought to determine whether there was also a cooperative effect when the D151 and G151 variants were mixed in direct infections , since generation of influenza virus by reverse genetics is a complex process that involves co-transfecting cells with plasmids encoding each of the eight viral genes . We generated pure populations of D151 and G151 viruses by reverse genetics , growing both populations in the presence of 50 nM oseltamivir , a small molecule that competes with sialic acid for binding to the NA active site . The addition of oseltamivir increases the titers of the G151 variant ( Figure 2—figure supplement 2 ) and presumably prevents selection for de novo NA mutations by suppressing both the cleavage and binding activity of this protein . We then infected cells with pure D151 viruses , pure G151 viruses , or an equal mix of both variants at a total multiplicity of infection ( MOI ) of 0 . 2 . One hour post-infection , we washed the cells to remove residual oseltamivir and then monitored viral replication . These experiments were performed in full biological triplicate , beginning with triplicate independent creations of each pure population by reverse genetics . Once again , the mixed populations consistently grew more rapidly and reached higher maximal titers than either pure population ( Figure 2B ) . The trends in the direct co-infections were similar to those observed when generating the viruses by reverse genetics . The pure G151 populations grew very poorly , again showing that this variant has very low fitness on its own . The pure D151 populations grew reasonably well on their own , but the mixed populations grew even better . These results show that cooperation between the D151 and G151 variants improves growth of the overall population . Interestingly , viral titers increased sharply late in the passage in some G151 populations . One possibility is that de novo mutations to the D151 variant create a mixed population with higher fitness . To explore the possibility of de novo emergence of cooperation , we serially passaged pure and mixed populations as described below . If the D151 and G151 variants cooperate , then we expect mixed populations to emerge by de novo mutation and to be stably maintained when they already exist . To test this prediction , we serially passaged pure and mixed viral populations and performed targeted deep sequencing of the NA gene at the end of each passage to assess changes in allele frequency at site 151 . We again used reverse genetics to generate triplicate pure populations of D151 and G151 viral variants in the presence of 50nM oseltamivir , then infected cells with D151 viruses , G151 viruses , or an equal mix of the two at a total MOI of 0 . 2 , washing the cells one hour post-infection to remove residual oseltamivir . We verified that the D151 and G151 populations used to inoculate the first passage were pure within our limit of detection of approximately 1% , which we determined by deep-sequencing pure plasmid . We performed a total of five serial passages for each replicate , in each case seeding the new passage with the supernatant from the previous one at a total MOI of 0 . 2 . The mixed D151+G151 populations maintained an approximately equal mix of the two variants through all five passages ( Figure 3 ) . In the pure populations , the opposite variant arose by de novo mutation , then rose in frequency as the population converged towards a roughly equal mix of the two variants . The D151 variant emerged rapidly during passage of the G151 populations , exceeding a frequency of 20% by the end of the second passage in all three replicates . The G151 variant was slower to arise in the D151 populations but had reached a substantial frequency by the end of passage 4 in all three replicates . The changes in allele frequency during serial passage demonstrate that selection acts to balance the proportion of these two genotypes in the population . 10 . 7554/eLife . 13974 . 009Figure 3 . Serial passage selects for a stable mix of the two variants . Shown are the allele frequencies at NA site 151 over five tissue-culture passages of initially pure D151 viruses , pure G151 viruses , or an equal mix of the two . Each passage was seeded at a total MOI of 0 . 2 . Passage 0 refers to the ratio of variants in the viral inoculum for passage 1 . Allele frequencies were determined by targeted Illumina deep-sequencing of the NA gene . Based on sequencing of pure plasmid , the error rate was less than 1% . The raw data and computer code are available in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 00910 . 7554/eLife . 13974 . 010Figure 3—source data 1 . This 7-zip archive contains the data and source code used for Figure 3 ( the analysis of mutation frequencies at site 151 after serial passage in the lab ) . The code and all the FASTQ files are also available at http://dx . doi . org/10 . 5061/dryad . s3rs0 . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 01010 . 7554/eLife . 13974 . 011Figure 3—figure supplement 1 . The N151 variant also cooperates with D151 . Shown are the titers after reverse genetics with the indicated variant of the Hanoi/2007 NA . The experiments here parallel those in Figure 2A . Black lines indicate the mean and standard error of the titers for three biological replicates , with titers for each replicate plotted as points . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 011 In one of the D151 populations , N151 also emerged spontaneously , and by the end of passage 5 , the population consisted of a mix of D151 , N151 , and G151 . Like G151 , N151 commonly occurs in mixed populations with D151 in sequences in the GISAID EpiFlu database ( Figure 1 ) and is mentioned in reports of mutations at site 151 in cell culture ( Table 1 ) ( McKimm-Breschkin et al . , 2003; Lin et al . , 2012; Okomo-Adhiambo et al . , 2010; Tamura et al . , 2013; Lee et al . , 2013; Chambers et al . , 2014; Mishin et al . , 2014; Mohr et al . , 2015 ) . We verified that N151 cooperates with D151 by creating the N151 variant of the Hanoi/2007 NA and generating pure and mixed populations by reverse genetics ( Figure 3—figure supplement 1 ) . N151 viruses behave similarly to G151 viruses: they grow very poorly on their own , but cooperate with D151 to outgrow either pure population . These results show that serial passage selects for mixed populations of D151 and G151 variants , even when the starting population is isogenic . Furthermore , mixed populations are stably maintained; the G151 variant does not sweep to fixation , as would be expected for a simple lab adaptation . Cooperation between the D151 and G151 variants evidently selects for the generation and maintenance of a genetically diverse quasispecies . Since each influenza virion typically packages only a single copy of the NA gene , co-infection of a cell by multiple viruses is likely to increase opportunities for interactions among viral variants . Figure 2 shows that at an MOI of 0 . 2 , the mixed populations of D151 and G151 variants have an advantage over pure populations . At a lower MOI , co-infection is less likely . We therefore sought to test whether cooperation also promotes growth at higher and lower MOIs . We infected cells with pure and mixed viral populations in biological triplicate at an MOI of 0 . 02 and an MOI of 0 . 5 , and then monitored viral titers over the next 40 hr as in Figure 2 . At an MOI of 0 . 02 , the mixed populations grew similarly to or slightly worse than the D151 populations for the first 24 hr post-infection ( Figure 4A ) . Later in the infection , however , the mixed populations grew to substantially higher titers than the D151 populations . In contrast , at MOIs of 0 . 2 ( Figure 2B ) and 0 . 5 ( Figure 4B ) , the mixed populations grew better than the pure populations throughout the entire infection . 10 . 7554/eLife . 13974 . 012Figure 4 . Cooperative dynamics depend on multiplicity of infection . Cells were infected at an MOI of ( A ) 0 . 02 or ( B ) 0 . 5 with pure D151 virus , pure G151 virus , or an equal mix of the two . The total MOI of infecting virus was the same across the mixed and pure populations for infection at each MOI . The main plots show titers averaged across three biological replicates , with each replicate plotted individually in the small insets . The experiments here parallel those in Figure 2B . Black lines indicate the mean and standard error of the titers for three biological replicates , with titers for each replicate plotted as points . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 012 We note that the effective MOI of an infection increases as the infection progresses as newly produced viruses accumulate in the supernatant ( Wilke et al . , 2004 ) . For the infections inoculated at an MOI of 0 . 02 , the sharp increase in titers for the mixed population late in the infection are likely a result of this higher effective MOI . We therefore conclude that the dynamics of cooperation depend on the multiplicity of infection , with the cooperative effect decreased at lower MOIs . Mutations at NA site 151 become common in the EpiFlu database only starting in 2007 ( Figure 1A ) , suggesting that other mutations to the influenza genome around that date might have affected the potential for cooperation among NA variants at site 151 . A candidate gene for these potentiating mutations is HA . Good viral growth requires a balance between the receptor binding of HA and the receptor cleaving of NA ( Wagner et al . , 2002; Gulati et al . , 2005; Neverov et al . , 2015 ) . For reasons that remain unclear , the HAs of recent human H3N2 influenza have lost their affinity for many types of sialic acid ( Lin et al . , 2012; Gulati et al . , 2013 ) . We therefore hypothesized that recent mutations in HA might have potentiated cooperation by making it advantageous for viral populations to acquire the NA-mediated receptor-binding of the G151 variant ( Zhu et al . , 2012 ) to compensate for reduced HA binding . To test this hypothesis , we examined the effects of the D151 and G151 NA variants in viruses that had the HA of an earlier H3N2 strain , A/Wisconsin/67/2005 , henceforth referred to as Wisconsin/2005 . We cloned the HA gene from the Wisconsin/2005 strain into a reverse-genetics plasmid and generated pure and mixed populations of D151 and G151 variants in the genetic background of either the Hanoi/2007 HA or the Wisconsin/2005 HA . Cooperation between the D151 and G151 variants was eliminated in the Wisconsin/2005 HA background ( Figure 5 ) . Therefore , some of the changes to HA that distinguish the Wisconsin/2005 and Hanoi/2007 homologs are important for potentiating cooperation between the NA variants . 10 . 7554/eLife . 13974 . 013Figure 5 . Changes in HA between 2005 and 2007 potentiated cooperation . Cooperation occurs between the D151 and G151 NA variants in viruses with HA from the Hanoi/2007 strain , but not in viruses with HA from the Wisconsin/2005 strain . Shown are the titers after reverse genetics with the indicated HA and NA . The experiments here parallel those in Figure 2A . Black lines indicate the mean and standard error of the titers for three biological replicates , with titers for each replicate plotted as points . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 013 If decreased HA receptor-binding potentiates cooperation between the receptor-cleaving D151 and receptor-binding G151 NA variants , then viral growth should depend entirely on this cooperation if NA is the only protein able to bind the receptor . To test this hypothesis , we used an HA that has been heavily engineered to eliminate its receptor-binding activity ( Hooper and Bloom , 2013 ) . We used reverse genetics to generate pure and mixed populations of D151 and G151 NA variants paired with this binding-deficient HA , and we measured viral titers ( Figure 6 ) . In the absence of HA receptor binding , neither the D151 nor the G151 variant alone reached appreciable titers . However , the mixed population was still able to grow with the binding-deficient HA . These results show that cooperation becomes obligate in the absence of HA receptor binding , presumably because NA must serve as the sole source of both binding and cleaving . 10 . 7554/eLife . 13974 . 014Figure 6 . Cooperation is obligate when HA lacks receptor-binding activity . Shown are the titers after reverse genetics with the indicated Hanoi/2007 NAs in combination with an engineered binding-deficient H3 HA with multiple mutations to the receptor-binding pocket ( Hooper and Bloom , 2013 ) . The experiments here parallel those in Figure 2A . Black lines indicate the mean and standard error of the titers for three biological replicates , with titers for each replicate plotted as points . DOI: http://dx . doi . org/10 . 7554/eLife . 13974 . 014
We have shown that cooperation between two distinct variants of human H3N2 influenza promotes viral growth in cell culture . These variants differ by a single amino-acid mutation in NA , and each variant is present in many human H3N2 isolates that have been analyzed by Sanger sequencing after passage in the lab . Prior work has assumed that the less common G151 variant is a lab-adaptation mutant that emerges as the more common D151 variant is passaged in cell culture . Our work shows , however , that evolution in cell culture selects for a balanced mix of both variants . The G151 variant can barely replicate alone , but it cooperates with the D151 variant to increase population fitness . After multiple serial passages , both pure and mixed populations converge to an equilibrium in which both variants are present at approximately equal frequencies . Our work therefore represents a clear example of cooperation between distinct variants in a viral quasispecies . We propose that cooperation arises because one variant is proficient at cell entry , while the other is proficient at cell exit . Viruses with wild-type D151 NAs always exit cells efficiently , since their NAs cleave sialic-acid receptors to facilitate viral release . However , the HAs of recent human H3N2 strains have reduced affinity for many sialic-acid receptors ( Lin et al . , 2012; Gulati et al . , 2013 ) , reducing the efficiency with which those viruses can attach to many cells via HA . G151 viruses are proficient at cell entry , since their NA binds strongly to sialic acid , but they cannot detach effectively from host cells due to a lack of catalytic activity ( Zhu et al . , 2012 ) . But in combination , D151 and G151 enable both efficient cell exit and entry . Indeed , our experiments with a binding-deficient HA indicate that in a mixed D151 and G151 population , NA can act as the exclusive source of both receptor binding and receptor cleaving ( Figure 6 ) . Our results evoke prior work showing that fitness in a Coxsackie virus population is enhanced by the combination of multiple receptor-binding variants ( Bordería et al . , 2015 ) . How do the D151 and G151 variants collaborate to enable both viral entry and exit at the level of individual virions ? Co-infection of the same cell with both D151 and G151 variants would produce progeny that have both NA variants on their surface , even though each new virion would package only a single copy of the NA gene . We suspect that much of the observed cooperation may result from co-infections that produce such mixed-NA virions , which would then carry proteins that make them proficient at both cell entry and cell exit . We found that MOI affects cooperative dynamics , supporting this interpretation ( Figure 4 ) . However , other mechanisms could also contribute . In a mixed population , D151 viruses may cleave G151 viruses from the cell surface without both protein variants being present on the same infectious particle , since sialidase activity can promote viral growth in trans ( Liu and Air , 1993 ) . More detailed molecular characterization of virions in mixed populations will be necessary to establish the exact mechanism of cooperation . It remains unclear whether the D151 and G151 variants cooperate in clinical infections . When we analyzed the passage histories of sequenced isolates , we found that mixed populations were reported almost exclusively in isolates that had been passaged in cell culture ( Figure 1C ) . Several groups have reported that mutations emerge at NA site 151 when clinical isolates are expanded in cell culture – but with one exception ( Lin et al . , 2010 ) , Sanger sequencing or pyrosequencing of matched clinical and passaged isolates has so far failed to detect variation in site 151 in unpassaged isolates ( Lee et al . , 2013; Chambers et al . , 2014; Mishin et al . , 2014 ) ( see also Table 1 ) . However , the sequencing methods used by these studies are relatively insensitive to low-frequency variation . Given how quickly and frequently D151 mutations sometimes arise—one group found that nearly a quarter of isolates showed variation at appreciable frequencies after a single passage in MDCK cells ( Lee et al . , 2013 ) —pre-existing variation at site 151 in the original clinical isolates could contribute to the observed evolution . More sensitive deep sequencing of unpassaged clinical isolates will be necessary to resolve these questions . Our work demonstrates that cooperation between distinct viral variants can enhance the population’s overall fitness . This cooperation is not a rare event; the cooperating variants that we describe emerge rapidly and repeatedly , both in our own experiments and apparently in hundreds of clinical isolates passaged by numerous labs . Our work emphasizes that genetic diversity in viral populations can be more than a transient state that facilitates adaptation: it can itself be a beneficial trait that is generated and maintained by selection . As the deep sequencing of viruses becomes increasingly common in microbiology and epidemiology , it will be important to better understand the broader role that cooperation plays in the evolution and maintenance of population-level diversity .
We downloaded the set of 15 , 079 sequences in the Global Initiative on Sharing All Influenza Data ( GISAID ) EpiFlu database ( Bogner et al . , 2006 ) corresponding to all full-length NA coding regions from human H3N2 influenza A isolates collected from January 1 , 2000 to December 31 , 2014 . We pairwise aligned each sequence to the A/Hanoi/Q118/2007 ( H3N2 ) coding sequence ( Genbank accession CY104446 ) using the program needle from EMBOSS version 6 . 6 . 0 ( Rice et al . , 2000 ) , which implements a Needleman-Wunsch alignment . We identified the genotype of each sequence at site 151 and parsed the sequence metadata to determine the year in which it was collected and the sequence’s passage history . Mixed genotypes were assigned on the basis of IUPAC nucleotide ambiguity codes; for instance , the triplet GRT could refer to GAT or GGT , corresponding to amino acids aspartic acid ( D ) and glycine ( G ) , respectively . We occasionally observed the triplet RRT , which could correspond to a mix of aspartic acid ( D; GAT ) , glycine ( G; GGT ) , asparagine ( N; AAT ) , and serine ( S; AGT ) . We chose to annotate triplet RRT as a mix of D , G , and N , given that this mixed population has been previously observed by multiple groups ( Tamura et al . , 2013; Mishin et al . , 2014; Mohr et al . , 2015 ) , whereas serine is two mutations away from the D consensus identity and is not present in the H3N2 GISAID sequences that we analyzed . Passage histories are not recorded in a standardized fashion and are frequently missing altogether . In parsing the passage histories of isolates in the EpiFlu database , therefore , we sought only to sort sequences into broad categories of which we could be reasonably certain: egg-passaged , cell-culture-passaged , and unpassaged isolates . For instance , sequences with passage annotations containing 'MDCK , ' 'SIAT , ' 'RHMK , ' 'MEK , ' and various other cell-culture signifiers were combined into the broad category of cell-culture-passaged isolates . Our exact parsing procedures and the computer code used for analysis are available in Figure 1—source data 1 . HA and NA sequences from the A/Brisbane/10/2007 ( H3N2 ) strain were cloned into the bidirectional pHW2000 backbone to generate virus by reverse genetics ( Hoffmann et al . , 2000 ) . We performed site-directed mutagenesis on the HA and NA to match the amino-acid ( but not nucleotide ) sequence from A/Hanoi/Q118/2007 ( Genbank accessions AEX34134 and AEX34137 for the HA and NA , respectively ) , which has a G at NA site 151 ( Bao et al . , 2008 ) . The HA and NA protein sequences are identical to those from A/California/UR06-0565/2007 ( Genbank accessions ABW40191 and ABW40194 for HA and NA , respectively ) aside from a single site in HA , as well as the genotype at NA site 151 . We performed further rounds of site-directed mutagenesis to generate the D151 and N151 variants of the NA . The HA sequence from the A/Wisconsin/67/2005 ( H3N2 ) influenza strain ( Genbank accession CY163744 ) was similarly cloned into the bidirectional pHW2000 backbone for reverse genetics . The binding-deficient HA is derived from the A/Hong Kong/2/1968 ( H3N2 ) HA and contains extensive mutations and deletions that eliminate receptor-binding activity; this is the variant referred to as the 'PassMut HA' in ( Hooper and Bloom , 2013 ) . Coding sequences for the HA and NA genes used in this study are available in Supplementary file 1 . The remaining six viral genes were expressed from bidirectional reverse-genetics plasmids derived from the A/WSN/33 strain ( pHW181-PB2 , pHW182-PB1 , pHW183-PA , pHW185-NP , pHW187-M , and pHW188-NS ) and were kind gifts from Robert Webster of St . Jude Children’s Research Hospital . For all experiments not otherwise indicated , we used a plasmid ( PB1flank-eGFP ) that carried GFP flanked by PB1 packaging signals in place of pHW182-PB1 plasmid , and propagated the viruses in 293T and MDCK-SIAT1 cells expressing the WSN PB1 under the control of a CMV promoter as described in ( Bloom et al . , 2010 ) . To generate GFP-expressing virus using reverse genetics , we transfected co-cultures of 293T-CMV-SIAT-PB1 and MDCK-SIAT1-CMV-PB1 cells with plasmids encoding the eight viral genes , with PB1flank-GFP rather than PB1 as described in ( Bloom et al . , 2010 ) . We plated 2 x 105 293T-CMV-PB1 cells and 0 . 2 x 105 MDCK-CMV-PB1 cells per well in six-well dishes in D10 ( Dulbecco modified Eagle medium supplemented with 10% heat-inactivated fetal bovine serum [FBS] , 2 mM L-glutamine , 100 U/mL penicillin , and 100 µg/mL streptomycin ) and transfected each well with 2 µg plasmid DNA , corresponding to 250 ng of each of the eight plasmids , using the BioT transfection reagent ( Bioland Scientific , Paramount , California ) . At 12 to 18 hr post-transfection , the cells were washed once with phosphate-buffered saline ( PBS ) , and the media was changed to low-serum influenza growth media ( IGM; Opti-MEM supplemented with 0 . 01% heat-inactivated FBS , 0 . 3% bovine serum albumin , 100 U/mL penicillin , 100 µg/mL streptomycin , and 100 µg/mL calcium chloride ) . TPCK ( toylsulfonyl phenylalanyl chloromethyl ketone ) -trypsin was added to IGM at 3 µg/mL immediately before use . For reverse genetics carried out in the presence of oseltamivir , we added the indicated concentration of oseltamivir carboxylate ( kindly provided by Roche ) to the IGM at this point as well . We collected viral supernatant at 72 hr post-transfection , clarified by centrifugation at 285xg for 4 min , aliquoted , and froze at -80 degrees C before thawing aliquots for titering . To generate viral populations that expressed the A/WSN/33 PB1 gene rather than the PB1 segment packaging GFP , we substituted 293T and MDCK-SIAT1 cells for 293T-CMV-PB1 and MDCK-SIAT1-CMV-PB1 cells in the protocol above . For viruses grown with the PB1flank-eGFP gene , titers were determined using flow cytometry . We plated 105 MDCK-SIAT1-CMV-PB1 cells per well in 12-well plates in IGM and infected them 4–6 hr later with 0 . 1 , 1 , 10 , or 100 µL of viral supernatant . At 16 hr post-infection , we collected the cells into PBS with 1% paraformaldehyde from wells in which approximately 1–10% of cells were GFP-positive and used flow cytometry to determine the exact proportion of GFP-positive cells . We used the Poisson equation to calculate the number of infectious particles in the original inoculum as: [titer , in infectious particle per µL] = -log ( 1 – [fraction of GFP-positive cells] ) * [number of cells plated , in this case 105] / [inoculum volume , in µL] For viruses grown with the WSN PB1 gene , titers were determined by staining for intracellular NP . Similar to the GFP titering described above , MDCK-SIAT1 cells were infected with serial dilutions of viral supernatant . At 12 hr post-infection , the cells were collected , fixed and permeabilized with the BD Cytofix/Cytoperm kit ( product number 554722 , BD Biosciences , Franklin Lakes , New Jersey ) following the manufacturer’s protocol but omitting the GolgiPlug , stained with a 1:20 dilution of mouse anti-NP FITC-conjugated antibody ( clone A1 from MAB8257F , EMD MilliPore , Darmstadt , Germany ) , washed twice , and analyzed by flow cytometry to count NP-positive cells . The viral titer was computed from the fraction of positive cells using the Poisson equation above . All titers are plotted with the lower bound of the y-axis set at the limit of detection of this assay , approximately 10–1 infectious particles/µL . Note that all of these titering methods quantify the number of virions that enter cells and express a functional polymerase complex that produces large amounts of the mRNA encoding the protein product being detected ( GFP or NP ) . Unlike in a TCID50 or plaque assay , not all detected virions are necessarily able to undergo multi-cycle infections . For each passage , we plated 105 MDCK-SIAT1-CMV-PB1 cells per well in six-well plates in IGM and infected them 4–6 hr later with D151 viruses , G151 viruses , or an equal mix of the two at a total MOI of 0 . 2 . An hour after the viruses were added , we washed the cells with PBS and added fresh IGM supplemented with 3 µg/mL TPCK-trypsin to dilute the effect of any oseltamivir remaining from reverse genetics for the first passage . We collected viral supernatant at 40 hr post-infection , clarified by centrifugation at 285xg for 4 min , aliquoted , and froze it at -80 degrees C before thawing aliquots for titering . We collected cells remaining at the end of each passage in 1mL Trizol reagent and froze them at -20 degrees C . From passage 4 onwards , mutation accumulation in the PB1flank-eGFP gene caused widespread loss of GFP in many viral populations; by the end of this passage , cytopathic effect was clearly visible even though GFP fluorescence was not . To inoculate passage 5 , we infected cells with 5uL viral supernatant from passage 4 , a volume corresponding approximately to an MOI of 0 . 2 based on the titers for earlier passages . We extracted RNA from cells remaining at the end of each passage using Trizol reagent and performed reverse-transcription using the primers CAGGAGTGAAAATGAATCCAAATCAAAAGATAATAACGATTG and TTGCGAAAGCTTATATAGGCATGAGATTGATG , which target the full-length NA gene . We then used primers CTTTCCCTACACGACGCTCTTCCGATCTxxxCAACACTAAACAACGTGCATTCAAATGAC and GGAGTTCAGACGTGTGCTCTTCCGATCTCCTAACTCATTCATCAATAGGGTCCGATAAGG to amplify a targeted region of the NA gene surrounding site 151 and add the first half of the Illumina sequencing adaptor and a three-mer in-read barcode , represented here as xxx . We performed 25 cycles of amplification at an annealing temperature of 55 degrees C and an extension time of 40 s . We purified the PCR product using 1 . 5X Ampure beads and used this product as template for a second round of PCR using primers AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCC and CAAGCAGAAGACGGCATACGAGATxxxxxxGTGACTGGAGTTCAGACGTGTGCTCTTCC , which add the second half of the Illumina sequencing adaptors and a six-mer barcode , represented here as xxxxxx . To sequence the viral stocks used to inoculate the first passage ( these inoculating stocks are referred to as 'Passage 0' in Figure 3 ) , we plated 105 MDCK-SIAT1-CMV-PB1 cells per well in six-well plates in IGM and infected them 4–6 hr later with viral stocks at an MOI of 0 . 02 . An hour after the viruses were added , we washed the cells with PBS and added fresh IGM to dilute the effect of any oseltamivir remaining from the generation of virus by reverse genetics . We collected the cells in 1mL Trizol reagent at 16 hr post-infection and froze them at -20 degrees C . The purpose of this inoculation was to ensure that sequencing of viral stocks detected only infectious particles . We then prepared PCR amplicons from these samples for deep sequencing as described above . Reads were first screened to verify sequencing quality and correct identity . Reads were discarded if any position had a Q-score below 25 or if the read had more than 4 mismatches relative to the plasmid reference sequence . We translated the reads in the NA reading frame and tallied the amino-acid identities at each position , recording “X” for positions with a discrepancy between the forward and reverse reads . The FASTQ files and the computer code used to analyze them are available in Figure 3—source data 1 . | Viruses like influenza mutate fast . When you get the flu , the virus hijacks your cells , and your body becomes home to millions of viruses , many of which are genetically different from each other . Previous research had suggested that variants of rapidly evolving viruses sometimes cooperate with one another to survive , but few studies have pinpointed specific cooperative interactions . In the past decade , influenza surveillance groups noticed that one particular mutant virus appears again and again when influenza viruses are grown in the laboratory . Xue et al . thought that this mutant virus shouldn’t be able to grow on its own because the mutation disrupted the protein that influenza viruses use to detach from host cells . So , they asked if the mutant was cooperating with the non-mutated virus to survive . Xue et al . revealed that the two influenza viruses , which differ by just one mutation , cooperate with each other when grown in cells in the laboratory . The mutant virus often appeared following a random mutation in a population of non-mutated virus , and vice versa . Instead of competing with each other until one virus went extinct , the two viruses actually grew better together than they did apart . Xue et al . suggest that this is because one of the viruses is good at entering new cells , while the other is better at exiting cells to spread the infection . A mixed population combines these two strengths . Following on from this work , it remains unclear whether influenza viruses cooperate in other settings – for example , during infections in people . Further studies are also needed to determine exactly how the two viruses help each other at the molecular level . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Cooperation between distinct viral variants promotes growth of H3N2 influenza in cell culture |
Lineage tracing approaches have provided new insights into the cellular mechanisms that support tissue homeostasis in mice . However , the relevance of these discoveries to human epithelial homeostasis and its alterations in disease is unknown . By developing a novel quantitative approach for the analysis of somatic mitochondrial mutations that are accumulated over time , we demonstrate that the human upper airway epithelium is maintained by an equipotent basal progenitor cell population , in which the chance loss of cells due to lineage commitment is perfectly compensated by the duplication of neighbours , leading to “neutral drift” of the clone population . Further , we show that this process is accelerated in the airways of smokers , leading to intensified clonal consolidation and providing a background for tumorigenesis . This study provides a benchmark to show how somatic mutations provide quantitative information on homeostatic growth in human tissues , and a platform to explore factors leading to dysregulation and disease .
In adults , stem cells reside at the apex of proliferative hierarchies and , either directly , or through a sequence of terminal divisions , give rise to the specialist differentiated cells that provide the functional properties of tissue . To conserve their number , following division , one cell must on average remain in the stem cell compartment while the other must commit to a differentiation pathway . This asymmetry may be enforced at the level of individual cells , or it may be achieved on a population basis , so that stem cell proliferation is perfectly compensated by the differentiation of others ( Clayton et al . , 2007; Klein et al . , 2010; Gomes et al . , 2011 ) . In recent years there has been significant progress in defining common strategies of stem cell self-renewal using lineage tracing assays in mice ( Simons and Clevers , 2011 ) . By tracing the clonal evolution of marked cells following genetic pulse-labelling of transgenic animals , statistical methods have been used to discern the pattern of adult stem cell fate in several actively cycling mammalian tissues . In skin ( Clayton et al . , 2007; Doupe et al . , 2010 ) , oesophagus ( Doupe et al . , 2012 ) , gut ( Lopez-Garcia et al . , 2010; Snippert et al . , 2010 ) and testis ( Klein et al . , 2010 ) , such methods have demonstrated that tissue maintenance involves the continuous stochastic loss and replacement of stem cells . However , although such approaches provide key insights into the mechanisms of stem cell maintenance in transgenic animal models , their relevance and application to human tissues is unknown . To date , our understanding of human airway homeostasis is limited and inferred indirectly through in vitro studies of human cell culture and murine airway models ( Hong et al . , 2004a; Schoch et al . , 2004; Hajj et al . , 2007; Rawlins et al . , 2007; Giangreco et al . , 2009; Rawlins et al . , 2009; Rock et al . , 2009 ) . In both humans and mice , airways are composed of basal , ciliated , secretory , and chemosensory cell populations . Collectively , these cells produce antimicrobial proteins , eliminate bacteria and toxic compounds via the mucociliary escalator , warm inspired air , and act as physical barriers to exogenous particulate matter ( Knight and Holgate , 2003 ) . Previous studies of human tissue have demonstrated that basal cells can exhibit multipotent growth and differentiation properties when grown in vitro under specific conditions , while in vivo murine studies demonstrate that basal cells of the major airways function as common multipotent progenitors ( Rock et al . , 2009 , 2010 ) . Using lineage tracing methods , involving chimeric tissue analysis ( Giangreco et al . , 2009 ) and genetic cell labelling ( Rawlins et al . , 2009 ) , murine airways appear to be maintained in homeostatic state by abundant progenitor cells located throughout airways . However , the range and identity of the stem cell compartment , and the frequency of stem cell loss and replacement , has not been quantified . Moreover , the relevance of these studies for the maintenance of human airways remains unexplored . In this study , we show that the continuous accumulation of somatic mutations that occur in the mitochondrial genome provides a clonal record from which the self-renewal properties of the human airway stem cell population can be inferred . As well as providing new insight into human airway maintenance , this study exemplifies a general methodological scheme that can serve as a template to study adult stem cell fate in other human tissues . Moreover , we show that the characterisation of normal tissue maintenance provides a quantitative platform from which we can study factors promoting the dysregulation of cells leading to the development of disease . In order to directly assess the in vivo identity and potency of individual human airway epithelial progenitor cells , we make use of the predisposition of mtDNA to develop spontaneous mutations that affect expression of the cytochrome c oxidase ( CCO ) gene . CCO gene mutations occur spontaneously in all cells in a stochastic manner , do not significantly affect cellular function , and are unrelated to cellular toxicant exposure ( Elson et al . , 2001; Taylor et al . , 2001; Carew and Huang , 2002; Taylor et al . , 2003; Taylor and Turnbull , 2005; Greaves et al . , 2006; McDonald et al . , 2008; Fellous et al . , 2009; Gutierrez-Gonzalez et al . , 2009; Lin et al . , 2010; Gaisa et al . , 2011a; Nicholson et al . , 2011 ) . Thus , the division and accumulation of CCO-deficient cells leads to the formation of clonal patches of CCO-deficient cells within tissues , including the normal airway , and their examination provides a unique , histologically traceable record of airway progenitor cell fate . We use genetic sequencing to confirm the clonal origin of individual CCO patches and immunofluorescence to assess the cellular composition of these clones . Then , using statistical modelling of the frequency and size distribution of CCO-deficient clones visualised using whole mount labelling , we establish the cellular hierarchy and the in vivo pattern of airway homeostasis , making an explicit comparison between non-smokers and smokers . From a detailed and quantitative analysis of the size and composition of CCO-deficient clones , we provide evidence that the maintenance of upper human airways relies upon multipotent progenitor cells that reside within the basal cell population . Further , we show that these cells maintain homeostasis through a process of population asymmetry in which their chance loss following commitment to differentiation is perfectly balanced by the duplication of others . This stochasticity leads to a natural process of age-associated airway clonal consolidation , which is notably accelerated in smokers , most likely due to increased rates of cellular turnover . As well as its intrinsic interest to human airway stem and progenitor cell biology , this study provides the benchmark to show how quantitative insights can be obtained from in vivo lineage tracing studies in human tissues , with obvious implications for studies of clonal progression in neoplasia .
To detect CCO-deficient cell patches of airway epithelial cells , we combined immunofluorescence labelling for CCO ( Figure 1A , B , green ) , with the pan-mitochondrial protein porin ( Figure 1A , B , red ) . Cells deficient in CCO , but marked by porin , indicate cell patches with CCO mitochondrial DNA mutation ( Nicholson et al . , 2011 ) . Using lung whole-mount imaging of seven patients of varying age ( Table 1 ) , we identified and quantified CCO-deficient patches of cells within the third generation bronchi of human upper airways ( Figure 1A , B , C ) . These patches were rare and randomly distributed within the airways ( Figure 1A ) . Consistent with previous observations , no CCO-deficient individual cells , or patches of cells , were found in the 25 year old patient , despite examination of over one million cells , placing a constraint on the time taken for the chance clonal selection of a single mitochondrial mutation within an individual cell ( Greaves et al . , 2006 ) . Within cells , there are thousands of mitochondria , each containing multiple copies of mtDNA . mtDNA mutations are random and increase with age ( Brierley et al . , 1998; Michikawa et al . , 1999 ) . Through chance expansion these mutations can be present in all copies of the mitochondrial genome ( homoplasmy ) or a proportion thereof ( heteroplasmy ) . For the mutated mitochondrial CCO genotype to result in a loss of CCO expression , homoplasmy or high levels of heteroplasmy must be present ( Sciacco et al . , 1994 ) . 10 . 7554/eLife . 00966 . 003Figure 1 . CCO-deficient human epithelial patches in the upper airway demonstrate multipotent differentiation . ( A and B ) Third generation bronchi stained for CCO ( green ) and counterstained with mitochondrial marker porin ( red ) show CCO-deficient patches . The black spaces are Goblet cells with vesicles containing mucus . ( C ) Schematic showing the location of the third generation human bronchi . ( D ) Schematic showing lung epithelial cell types of the upper airways . ( E ) CCO-deficient clonal patches contain keratin 5 positive basal cells ( blue; arrowhead ) , ( F ) acetylated tubulin positive ciliated cells ( blue; arrowhead ) and ( G ) Mucin 5AC positive goblet cells ( blue: arrowhead ) . ( H ) There was no difference in lineage-specific differentiation between the CCO-deficient patches and the CCO-active lung epithelial cells . Cell percentages were calculated after counting all cells from 11 CCO-deficient patches and 11 CCO-active patches ( basal cells–34 . 7 ± 2 . 1 vs 36 . 7 ± 6 . 5; ciliated cells–45 . 7 ± 1 . 7 vs 44 ± 7 . 2 , Goblet cells–19 . 6 ± 2 . 1 vs 19 . 3 ± 13 . 5 [mean ± SD] ) . Scale bars—50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 00310 . 7554/eLife . 00966 . 004Figure 1—figure supplement 1 . CCO-deficient human epithelial patches in the lower airway demonstrate multipotent differentiation . ( A ) Schematic showing lung epithelial cell types of the lower airways . ( B ) CCO-deficient clonal patches contain keratin 5 positive basal cells ( blue; arrowhead ) , ( C ) Clara cell secretory protein positive Clara cells ( blue; arrowhead ) , ( D ) acetylated tubulin positive ciliated cells ( blue; arrowhead ) and ( E ) Mucin 5AC positive goblet cells ( blue; arrowhead ) . ( F ) There was no difference in lineage-specific differentiation between the CCO-deficient patches and the CCO-active lung epithelial cells . Cell percentages were calculated after counting all cells from 11 CCO-deficient patches and 11 CCO-active patches ( basal cells–31 . 9 ± 3 . 6 vs 27 . 8 ± 2 . 3; ciliated cells–38 . 8 ± 2 . 6 vs 51 . 8 ± 6 . 5 , Clara cells–23 . 7 ± 3 . 5 vs 16 . 3 ± 5 . 6 , Goblet cells–5 . 6 ± 0 . 4 vs 4 . 1 ± 1 . 3 [mean ± SD] ) . Scale bars—50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 00410 . 7554/eLife . 00966 . 005Figure 1—figure supplement 2 . c-kit cells stained ( red ) did not stain with epithelial markers ( green ) and were not present in most patches . ( A and B ) c-kit positive cells ( red ) were positive for CD45 , a leucocyte common antigen , ( green ) making cells appear yellow and ( C ) were not positive for K5 basal cell marker ( green ) and they were not present in most patches . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 00510 . 7554/eLife . 00966 . 006Table 1 . Patient characteristicsDOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 006PatientAge ( Surgery ) SexSmoking ( pack years ) 139M30255F40379M40425FNon smoker547MNon smoker657FNon smoker765MNon smoker Normal lung airway contains two regions . The upper airway epithelium contains basal , ciliated , and goblet cells ( Figure 1D ) , while the lower conducting airways include Clara cell secretory protein positive cells ( Clara cells ) . To confirm a normal epithelial make up of the epithelial CCO-deficient patches we performed immunostaining of the two regions . These studies revealed that large CCO-deficient patches in upper human airways contain the expected bronchial epithelial cell types including keratin 5 positive basal cells , acetylated tubulin positive ciliated cells , and Mucin 5AC positive goblet cells ( Figure 1E–G ) . Based on the percentage of constituent cell types , we found that there was no significant difference in cell type composition between CCO-deficient clonal patches and the neighbouring CCO-active lung epithelial cells ( Figure 1H ) . On examination of the lower conducting airways , the frequency of Clara cells in CCO-active lung epithelial cells matched with that of the neighbouring tissue ( Figure 1—figure supplement 1 ) . Of note , we observed rare c-kit positive cells , recently proposed as a putative marker for pulmonary stem cells in the airway . However , these c-kit positive cells were stained uniformly with CD45 ( leucocyte common antigen ) , but not epithelial markers , and they were absent in most patches ( Figure 1—figure supplement 2 ) . Cells within individual CCO-deficient patches were demonstrated as clonally derived by single cell laser microdissection and subsequent mitochondrial genome sequencing ( Figure 2A–F ) . Mitochondrial sequencing of single cells requires frozen tissue , and patches were identified on frozen tissue sections using dual colour enzyme histochemistry , simultaneously detecting enzyme activity of the mtDNA-encoded CCO and nuclear DNA-encoded succinate dehydrogenase ( SDH ) . CCO-deficient cells , or patches of cells ( stained blue ) , were seen to be surrounded by CCO-expressing adjacent lung epithelial cells ( stained brown ) , and individual cells were laser dissected ( Figure 2A , D ) . 10 . 7554/eLife . 00966 . 007Figure 2 . CCO-deficient patches are clonal with distinct mutations from neighbouring patches . ( B and D ) Histochemistry demonstrates CCO-deficient patch ( dark blue ) surrounded by CCO-active adjacent lung epithelial cells ( brown ) with individual cells laser micro-dissected . ( B and E ) Homoplasmic mutation shared by all CCO-deficient ( blue ) lung epithelial cells both within the main patch and blue cells scattered around the edges of the patch . ( C and F ) CCO-active cells show wild-type genotype . Scale bars—100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 007 Within the patch shown in Figure 2A , a 9850 T>C homoplasmic mutation was found in all CCO-deficient lung epithelial cells ( Figure 2B ) while all CCO-normal cells maintained their wild-type genotype ( Figure 2C ) . This 9850 T>C mtDNA mutation is present in the mitochondrial cytochrome c oxidase III gene , and results in a Leu215Pro amino acid substitution in cytochrome c oxidase protein predicting the identified CCO deficiency . The second patch , shown in Figure 2D–F , also demonstrated that CCO-deficient lung epithelial cells shared the same mtDNA mutation not present in adjacent CCO-normal cells . In total , seven CCO-deficient patches of varying size from three patients were genetically analysed and confirmed that each patch was clonally distinct , as evidenced by patch-unique mtDNA mutations ( Table 2 ) . This demonstrates clonal expansion of proximate epithelial cells within morphologically normal airways . 10 . 7554/eLife . 00966 . 008Table 2 . Analysis of mtDNA mutations of seven CCO-deficient patches , from three patientsDOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 008PatientPatchCCO-deficient cells mutationGene119850 T>CMT-CO326708 G>AMT-CO136838 T>CMT-CO146690 G>AMT-CO156692 A deletionMT-CO1269478 T deletionMT-CO3376087 G>AMT-CO1Multiple single cells laser captured from each patch confirmed the same sporadic mutation and hence patch clonality . There was no significant difference in the proliferation index ( as measured by Ki67 ) between CCO-positive epithelium and CCO-negative patches ( Figure 3A , B ) . Significantly , on examining the cell types within CCO-deficient patches , only ciliated and basal cells were always present in CCO-deficient patches with more than five cells , consistent with murine data identifying basal cells as the most likely host of multipotent progenitors of the upper airway ( Figure 3C , D ) . While the majority of clones with five cells or less contain one or more basal cells ( Figure 3E ) , a few small clones lack basal cells altogether suggesting basal progenitors are able to undergo terminal division leading to clone loss ( Figure 3—figure supplement 1A–B ) . Indeed , this behaviour is corroborated by measurements of the clone density . In total we identified 844 CCO-deficient patches among the 12 . 7 million cells studied in the seven patients . However , despite the ongoing marking of cells through sporadic mutation , the number of surviving clones does not increase substantially between 39 to 79 year old patients ( Figure 4A ) . As well as providing further evidence of clone loss through differentiation , this result suggests that the frequency of loss and replacement may be high . 10 . 7554/eLife . 00966 . 009Figure 3 . CCO-deficient upper airway patches are representative of the upper airway . ( A ) Ki67 immunofluorescence showing a positive CCO-deficient cell ( arrowhead ) and a positive CCO-active cell ( arrow ) . ( B ) There was no difference in the proliferation index between CCO-deficient clonal patches and CCO-active lung epithelia . CCO-active patches ( n = 11 ) vs CCO-deficient patches ( n = 11 ) ( mean ± SD , 1 . 8 ± 0 . 2 vs 1 . 5 ± 0 . 1 ) . ( C ) Cell type examination of patches greater than five cells showed that only ciliated and basal cells were present in all CCO-deficient patches . ( D ) Shows a clone greater than five cells with basal cells—arrowhead points to violet basal cells due to an expression of Porin ( red ) and Krt5 ( blue ) within a CCO-deficient patch . ( E ) Shows a small clone with a basal cell—arrowhead points to violet basal cells due to an expression of Porin ( red ) and Krt5 ( blue ) within a CCO-deficient patch . ( CCO , green; Porin , red; Keratin 5 , blue ) . The black spaces in D and E are Goblet cells with vesicles containing mucus . Scale bars—50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 00910 . 7554/eLife . 00966 . 010Figure 3—figure supplement 1 . Shows rare small clones with no basal cells . CCO-deficient cells ( red ) after porin ( red ) + CCO ( green ) staining show occasional small clones with no basal cells ( blue ) . Scale bars—50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 01010 . 7554/eLife . 00966 . 011Figure 4 . Density of CCO-deficient clones and their size distribution . ( A ) Measured clone density ( number per unit area ) of CCO-deficient clones measured in three smokers and three non-smokers of different ages . Note that , with increasing age of patients , the clone density changes little , while the overall density in the non-smoker is significantly smaller than smokers of a similar age . ( B and C ) Cumulative clone size distribution of CCO-deficient clones , Cn ( t ) , for the three smokers ( B ) and three non-smokers ( C ) showing the probability that a clone has a size larger than n cells in patients of age t = 39 years , 55 years , and 79 years for the smokers and 47 years , 57 years and 65 years for the non-smokers . ( Errors denote SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 01110 . 7554/eLife . 00966 . 012Figure 4—figure supplement 1 . Raw clone size data . ( A ) Smokers and ( B ) non-smokers . Each data point depicts an individual clone . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 012 Although these findings are consistent with the existence of a multipotential self-renewing stem cell within the basal cell population , they leave open the question of stem cell identity , potency , and pattern of fate . To classify the behaviour of the airway progenitor cell population , we turned to a more detailed quantitative analysis of the size distribution of CCO-deficient clones in the upper airways by analysing large areas of whole-mount airways . To develop a more quantitative analysis of the clonal fate data , we recorded in detail the total sizes of CCO-deficient clones for all patients including smokers and non-smokers . With clonal densities of around 100/cm2 ( Figure 4A ) , we estimate a typical separation between clones to be around 100 cell diameters . With typical clone sizes of around 100 cells or less , this suggests that errors due to clone merger events are likely to play only a minor role . The results , shown in Figure 4B , C , are depicted through the cumulative clone size distribution , Cn ( t ) , representing the frequency of clones larger than size n cells . For example , referring to the 39 year old smoker , some 25% of clones have a total size of more than 10 cells , etc . From these studies , we find a broad distribution of clone sizes with clones as large as 30–40 cells coexisting with clones with only one or two cells . While the clones show a small drift to larger sizes with increasing age , the distributions remain broad . Although the characterisation of CCO-deficient clones by size provides indirect access to the fate behaviour of marked cells and their progeny , the interpretation of these data sets is not straightforward . First , as a result of the sporadic nature of cell labelling , the clonal history of ‘young’ clones is not readily disentangled from that of clones induced in the distant past . For example , small clones may derive from recently induced cells or they may be associated with chance expansion and contraction through differentiation and cell loss in older early marked cells . Furthermore , the interpretation of the clonal fate data , and the association of clone size with a measure of the number of constituent progenitor cells in the clone , may be further complicated by the existence of a transit-amplifying cell hierarchy . Therefore , to critically assess the clonal data for signatures of equipotency and progenitor cell fate , it is useful to develop a simple and robust biophysical modelling scheme , the validity of which can be checked self-consistently through characteristics in the clone fate data . To develop our model , let us suppose that maintenance of lung epithelium involves the steady turnover of a single , multipotent , functionally equivalent ( i . e . , equipotent ) , tissue-maintaining population , a necessary condition of long-term tissue homeostasis . Following division of these proliferative cells , the daughters can adopt one of three possible fate outcomes—two progenitor cells , two cells that have commited to a differentiation pathway ( either directly or through a series of terminal divisions ) , or one progenitor and one differentiating cell ( Figure 5A ) . At this stage , we do not discriminate between different differentiating cell types and , in doing so , presume that the balance between proliferation and differentiation is not correlated with particular fate choice . To ensure long-term homeostasis , symmetric divisions leading to progenitor cell duplication must be perfectly balanced by those leading to differentiation and loss . In this paradigm , clones are predicted to follow a process of ‘neutral drift’ in which the chance expansion of some clones through proliferation is compensated by the contraction and extinction of others through differentiation and subsequent loss ( Figure 5B ) . 10 . 7554/eLife . 00966 . 013Figure 5 . Maintenance of lung epithelium involves neutral competition . ( A ) Schematic showing the model hypothesis used to interpret and analyse the clonal fate data . According to this model , maintenance of the human airway epithelium involves the balanced stochastic fate of tissue-maintaining cells in which cell division results in all three fate outcomes: symmetric duplication , asymmetric division , or symmetric differentiation . λ denotes the corresponding cell division rate , and r controls the fraction of divisions that lead to symmetric fate outcome . Note that differentiation can lead to any one of the differentiated cell types , with probabilities commensurate with the tissue composition . Here , for simplicity , we associate tissue-maintaining cells with the basal progenitor population . However , if tissue-maintaining cells constitute only a subpopulation of these basal progenitors , we would obtain the same long-term clone fate behaviour , Equation 1 , while the overall fraction of tissue-maintaining cells , ρ , would have to be adjusted accordingly . ( B ) Schematic depicting the pattern of clonal evolution following pulse-labelling of tissue . As tissue turns over , chance clonal loss is perfectly compensated by the expansion of other clones so that the overall number of labelled progenitor cells remains approximately constant—a process reminiscent of ‘neutral drift’ . ( C ) Through the spontaneous acquisition of somatic mtDNA mutation single cells become clonally marked throughout adulthood . As result , the clones distribution at any given time point represents the amalgamation of clones of different ages from those induced at the earliest times when CCO-deficient cells first become visible , to those marked within the recent past . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 013 Following this pattern of stochastic fate choice , previous studies have shown that the size distribution of clones derived from a single progenitor cell converge onto a scaling form in which the chance of finding a surviving clone with n>0 progenitor cells after a time t post-labelling is given by , Pnsurv . ( t ) = ( 1/N ( t ) ) exp[−n/N ( t ) ] , with the average size , N ( t ) = ( 1+rλt ) /ρ , growing linearly with time ( Clayton et al . , 2007 ) . Here λ denotes the progenitor cell division rate , r specifies the balance between symmetric and asymmetric fate outcome ( Figure 5A ) , and ρ defines the total number of progenitor cells as a fraction of the total cell population . This growth in the average size of surviving clones perfectly compensates for clones that are lost through commitment to differentiation so that the average number of labelled progenitor cells remains fixed at one per induced clone . Note that , since the proliferative capacity of any transit-amplifying cell progeny is strictly limited , its effect can influence the value of ρ , but does not effect the long term scaling form of the size distribution . In the present case , since mitochondrial DNA mutations led to a steady accumulation of marked cells over time , the effective size distribution of surviving clones involves the sum over all of these histories ( Figure 5C ) , and is predicted to take the form ( Klein et al . , 2010 ) , ( 1 ) Pnsurv . ( t ) =1ln[ρN ( t ) ]exp[−n/N ( t ) ]n . Here N ( t ) denotes the average size reached by surviving clones induced in the first round of visible mutations ( i . e . , after a time , t ) . Although the predicted size dependence , Equation 1 , can be immediately compared with the observed clonal fate data ( Figure 4B , C ) , it is important to consider which parts of the dataset are valuable ( Klein et al . , 2010 ) . For n≪N ( t ) ( approximately Figure 4B , 55y ) , the form of the clone size distribution will be dominated by clones created in the ‘recent past’ . For these clones , the sum over possible histories ( i . e . , clone birthdates ) masks the divergence of individual clone sizes due to stochastic fate choice , and leads to the featureless 1/n dependence that dominates the distribution , Equation 1 , for small n . However , for n≥N ( t ) ( approximately Figure 4B , 55y ) , the clone size distribution becomes sensitive to the ‘front’ of clones that were marked at the earliest time point . In the present context , this translates to the age at which the CCO-deficient clones first become visible . In this limit , the clone size distribution , Equation 1 , is dominated by the exponential dependence . From a fit of the model prediction , Equation 1 , to the raw experimental data ( Figure 4B , C , Figure 4—figure supplement 1 ) , we find excellent agreement ( Figure 6A , B , Figure 6—figure supplement 1 ) for all six patients ( smokers and non-smokers ) with CCO-deficient clones , with the inferred values of average clone size N ( t ) , shown in Figure 6C . ( Here , to amplify the tail of the clone size distribution , we have studied the behaviour of a ‘derivative’ of the cumulative distribution known as the first incomplete moment . For further details , see ‘mathematical analysis’ ) From the fit of the experimental data with the predicted form of the distribution , we conclude that the multipotent progenitor cells that line the lung epithelium of the upper airways form a single , equipotent population that maintain tissue through a process of population asymmetry . In the course of turnover , the chance loss of progenitors through differentiation is perfectly compensated by duplication of neighbouring progenitors leading to a neutral drift in the size of surviving clones and a continual depletion in clonal diversity of tissue ( Figure 5A ) . 10 . 7554/eLife . 00966 . 014Figure 6 . Quantitative analysis of the clonal fate data . Comparison and fit of the first incomplete moment , a derivative of the cumulative clone size distribution ( Figure 4B , C ) , to the model prediction , Equation 1 ( for details , see ‘Mathematical analysis’ ) for ( A ) smokers and ( B ) non-smokers . Points show data and lines show the model prediction . From these comparisons , we obtain the one fitting parameter , N ( t ) , defining the average size of clones created at the earliest time point , as described in the main text . The resulting values of N ( t ) from these fits are shown in ( C ) for the three smokers and three non-smokers . Error bars depict the approximate range of N ( t ) values that provide satisfactory fits to the measured clonal fate data . Note that these inferred values are consistent with the predicted linear dependence of N ( t ) on age t providing further corroboration of the model dynamics . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 01410 . 7554/eLife . 00966 . 015Figure 6—figure supplement 1 . First incomplete moment distribution , derived from the cumulative clone size distribution . ( A ) Smokers and ( B ) non-smokers ( for details of the definition of the distribution , see ‘Mathematical analysis’ ) . Points show data and lines show a fit to an exponential size dependence , as predicted by the model dynamics . DOI: http://dx . doi . org/10 . 7554/eLife . 00966 . 015 As well as providing a signature of population asymmetry , from the analysis of the average clone size , N ( t ) , and clone density , we can infer the frequency of progenitor cell loss and replacement . In particular , for both smokers and non-smokers , the values of N ( t ) , inferred from the quantitative analysis of the clone fate data follow approximately the predicted linear time-dependence ( Figure 6C ) , with a rate constant of rλ/ρ = 2 . 7 ± 0 . 5/year for smokers and rλ/ρ = 1 . 5 ± 0 . 8 for non-smokers . Moreover , from the extrapolation of N ( t ) to zero , we can deduce that CCO-deficient mutations become visible in patients at approximately 20 ± 10 years of age , consistent with the absence of CCO-deficient patches in the 25 year old patient . If we suppose that the basal cell population comprises the entire population of tissue-maintaining progenitor cells , from their proportion of the total lung epithelial cell population ( Figure 1H ) we can estimate ρ = 1/3 , from which it follows that the progenitor cell loss/replacement rate is given by 2rλ = 1 . 8 ± 0 . 4/year . Intriguingly , this compares to the inferred value for non-smokers of 2rλ = 1 ± 0 . 5/year , a factor of around two smaller . Further support for this conclusion is found from the observation of single cell clones , which are more prevalent in non-smokers than smokers ( Figure 4—figure supplement 1 ) , consistent with a higher rate of turnover in the later . Although this methodology provides the means to estimate the progenitor cell loss/replacement rate , the total cell division rate is not accessible . Cell divisions leading to asymmetric fate have no impact on clone size . As a result , we cannot say whether the acceleration of loss/replacement in smokers is simply a reflection of enhanced proliferation , or represents a tilt in the balance between symmetric and asymmetric cell division . Alongside the clone size distribution , the neutral drift model also predicts the accumulation rate of surviving clones . In homeostasis , for a constant mutation rate , R , the surviving clone density ( clones per unit area of tissue ) is predicted to rise only slowly ( logarithmically ) over time varying according to the relation , σ= ( ρR/rλ ) ln[N ( t ) ] ( ‘Mathematical analysis’ ) . Such a slow growth is a manifestation of the neutrality of clonal evolution leading to the chance loss of the majority of induced progenitors . This dependence of clone density on the parameters of cell fate can be used to further challenge the model . From the small increase in clone density in smokers from 39 to 79 years ( Figure 4A ) , as expected from the predicted logarithmic growth dependence , using the inferred values of rλ/ρ and N ( t ) , we find a mutation rate of R = 100 ± 10/year/cm2 . With a cell density of around 106/cm2 , this translates to the acquisition of CCO deficiency at a rate of around 10−4/cell/year . By contrast , applied to the data for non-smokers from age 47 to 65 years , we find a mutation rate with a lower figure of R = 60 ± 10/year/cm2 .
In summary , these results show that maintenance of human lung epithelium involves the ongoing stochastic loss and replacement of a single , functionally equivalent , multipotent , progenitor cell population . Although the clonal fate data does not unambiguously disclose the identity of this population , the findings are consistent with tissue-maintaining cells belonging to the pool of basal progenitors . Moreover , this study provides strong evidence that the rate of progenitor cell loss and replacement is increased in smokers , providing a background on which the process of field cancerisation can operate . In this study we exploited the natural occurrence of mitochondrial DNA mutations as markers of clonal expansion . Since only stem cells have long enough life spans to accumulate mtDNA mutations to detectable levels of homoplasmy , a clonal population marked by a particular mtDNA mutation likely represents the progeny of a single mutated stem cell ( Elson et al . , 2001; Greaves et al . , 2006; Fellous et al . , 2009 ) . Using this lineage tracing methodology , stem cell niches have been located in many human tissues ( Elson et al . , 2001; McDonald et al . , 2008; Fellous et al . , 2009; Gutierrez-Gonzalez et al . , 2009; Lin et al . , 2010; Gaisa et al . , 2011b ) . Our use , for the first time , of whole-mount human lung imaging allowed us to use quantitative analysis of these clones to demonstrate both that basal cells are the likely multipotent progenitor cells in the human upper airways and that airway homeostasis is maintained in a stochastic manner . The demonstration that a few small clones lack basal cells is consistent with terminal differentiation of a stem cell within a said clone , which would eventually lead to its loss . Curiously , using elegant lineage tracing studies following injury , the presence of both unipotent and multipotent basal cells airway progenitors have been proposed in mice ( Hong et al . , 2004b ) . By contrast , the absence of large clones ( >5 cells ) with only basal cells suggests that , at least under normal homeostatic conditions , the basal population of the upper airways does not include a long-lived unipotent progenitor . However , further studies of the murine system , using 3D whole mount imaging of homeostatic tissue is required to establish whether these differences are genuine . Lastly , we have systematically examined the upper large airways and do not discard a role for other cell types such as Clara cells in the maintenance of homeostasis of smaller human airways . In our whole mount imaging of large bronchi Clara cells are virtually absent , as found by others ( Boers et al . , 1998 ) , but in distal conducting airway epithelium of normal human lungs Clara cells make up between 11 and 22% of the epithelial population and 9% of the overall proliferative compartment ( Boers et al . , 1999 ) . In analysing the data , we have placed emphasis on a model in which progenitor cell fate choice follows from cell-autonomous ( intrinsic ) regulation , that is it is insensitive to environmental factors . However , previous studies have shown that , in the two-dimensional geometry pertinent to the lung epithelium , a process in which progenitor cell loss was compensated by duplication of neighbouring progenitors through a coordinated process would lead to the same long-term exponential clone size distribution considered here , while the average clone size would acquire a small additional logarithmic dependence on time , which would not be resolved by the current data ( Klein et al . , 2010 ) . As such , the question of the regulatory control underlying stochastic fate specification and balance is left open . In addition , these findings do not rule out the potential for a stem/progenitor cell hierarchy , with progenitors supported by a second quiescent stem cell population , as recently resolved in normal interfollicular epidermis ( Mascre et al . , 2012 ) . Finally , as well as its implications for the maintenance of the human airway epithelium , this study provides a benchmark to show how clones derived from the acquisition of somatic mutations can be used quantitatively to explore the pattern of homeostatic growth in other human tissues , providing a platform to investigate factors leading to dysregulation and disease .
Human lung lobes were obtained from patients undergoing lung resection . Airways were taken at least 5 cm away from any tumour . The airways were histologically analysed and found normal . Specimens were either fixed in 4% paraformaldehyde and embedded in paraffin or snap-frozen in liquid nitrogen cooled isopentane . Ethical approval was sought and obtained from University College London Hospital Research Ethics Committee ( REC reference 06/Q0505/12 ) . This study was carried out in accordance with the Declaration of Helsinki ( 2000 ) of the World Medical Association . Frozen lung samples were mounted in OCT compound for sectioning . Same section sequential histochemical staining for cytochrome c oxidase ( CCO ) , a component of complex IV of the respiratory chain enzyme , and succinate dehydrogenase histochemistry , a component of complex II of the respiratory chain ( the presence of which was used to highlight the absence of CCO activity ) was carried out on 10 µm transverse sections as follows . Sections were first incubated in CCO incubation medium ( 100 µM cytochrome c/4 mM diaminobenzidine tetrahydrochloride/20 µg/ml catalase in 0 . 2 M phosphate buffer , pH 7 . 0 ) for 50 min at 37°C . They were then washed in PBS , pH 7 . 4 , for 3 × 5 min and incubated in SDH incubation medium ( 130 mM sodium succinate/200 µM phenazine methosulphate/1 mM sodium azide/1 . 5 mM nitroblue tetrazolium in 0 . 2 M phosphate buffer , pH 7 . 0 ) for 45 min at 37°C . Sections were washed in PBS for 3 × 5 min , dehydrated in a graded ethanol series ( 70% , 95% , and 100% ) and left to air dry for 1 hr . Single cells from CCO-deficient and CCO-positive airways were cut into sterile 0 . 5-ml PCR tubes using the Leica ( Deerfield , IL , USA ) Laser Microdissection ( AS-LMD ) System . Cell digestion and DNA extraction were performed by overnight incubation in a DNA extraction kit ( PicoPure , Arcturus; Molecular Devices , Sunnyvale , CA , USA ) at 65°C and then 95°C for 10 min to denature the proteinase K . The extracted DNA was used to sequence the entire mitochondrial genome from microdissected areas . A two-round amplification method was followed , whereby the first round consisted of amplifying nine fragments spanning the entire genome , and the second round consisted of 36 M13-tailed primer pairs to amplify overlapping segments of the first-round products . Sequencing was performed using the BigDye terminator cycle sequencing method on an ABI Prism Genetic Analyzer ( Applied Biosystems , Foster City , CA ) and compared with the revised Cambridge reference sequence using sequence alignment software of the European Molecular Biology Open Software Suite ( EMBOSS , http://www . ebi . ac . uk/emboss/align/ ) . 4 µm formalin fixed paraffin lung sections were cut . Human OxPhos Complex IV Subunit Monoclonal antibody ( CCO ) ( Invitrogen , Carlsbad , CA ) , Porin ( Abcam , Cambridge , UK ) , Clara Cell Secretory Protein ( gift from Barry Strip’s lab—Goat anti—CCSP , Goat number 899 ) , Keratin 5 ( Abcam ) , acetylated tubulin ( Sigma-Aldrich , St . Louis , MO ) , Ki67 ( Dako , Glostrup , Denmark ) , Mucin 5AC ( Sigma ) , c-kit ( Dako ) and cd45 ( Dako ) antibodies were used for immunostaining following EDTA ( pH 9 . 0 ) antigen retrieval . Secondary antibodies included appropriate species-specific Alexa488 , Alexa555 , and Alexa633 dyes ( Invitrogen ) . Isotype-matched and without primary controls revealed no nonspecific staining . Images were obtained using a Leica TCS Tandem confocal at 10x , 20x and 40x objective magnification ( Leica Microsystems , Milton Keynes , UK ) . Small patches were assessed for cell type by contiguous sectioning . The samples mathematically analysed using the whole mount technique were from the third generation bronchi . Lobectomy specimens were used that had two clear features . First the surgical cut was across the second generation bronchus ( meaning the upper on the right and left; or lower lobe on the left; or lower or middle lobe on the right ) . The resection margin was retained by pathology , and the next/third generation bronchi dissected out by the research team . Second any resected tumour had to be in the distal lung parenchyma to avoiding close proximity to the whole mount area . Importantly the distal end of the samples were blocked and shown to be normal to regular histology assessment . Human lung lobes were dissected open to expose the airways to antibodies staining . Whole-mount airways were fixed with 4% vol/vol paraformaldehyde , washed in 0 . 1 Triton X-100 in PBS , subjected to antigen retrieval by boiling in 10 mM citrate buffer for 40 min , and incubated in blocking buffer ( 3 hr; 10% FBS , 0 . 05% fish skin solution in PBS ) . Airways were then incubated overnight with antibodies against OxPhos Complex IV Subunit Monoclonal antibody ( CCO ) and Porin and were then washed in PBS ( 0 . 05% Tween 20 ) and incubated with Alexafluor-488 or -555—conjugated anti-mouse or anti-rabbit secondary reagents ( Invitrogen ) for 6 hr and then washed . Nuclei were stained with DAPI . Images were acquired using an Olympus epi-fluorescence microscope . A total number of 12 . 7 million cells included in seven different airways were analysed using the 100X objective of the Olympus epi-fluorescence microscope . Among them 20 , 002 CCO-deficient cells included in 844 CCO-deficient patches were identified . To analyse the clonal fate data , we have made use of a simple biophysical modelling scheme . In keeping the focus of the main text on the principle results , important aspects of the mathematical analysis have been sacrificed . Here , in this section , we develop the modelling scheme more fully , stating clearly the assumptions on which it relies , and the experimental consistency checks that we can make to validate the approach . The method is one that is inspired by an earlier study on the time-evolution of p53 mutant clones in epidermis following exposure to UV-B radiation ( Klein et al . , 2010 ) . This section is divided into three sub-sections . In the first , we consider the potential pattern of clonal growth in a homeostatic tissue following ‘pulse-labelling’ of cells . In the following section , we discuss how these results must be revised to accommodate on-going induction due to the continuous accumulation of sporadic mtDNA mutations . The application of these results is discussed in the main text and figures . To address the lineage tracing data , we need a platform to interpret the clonal evolution . In the following , we will assume that , in adult , the lining of the human airway epithelium undergoes a slow and constant rate of turnover . Later , we will present experimental evidence in support of this assumption . Further , let us assume that the tissue-maintaining cell population is functionally equivalent ( i . e . , equipotent ) , a necessary condition of long-term homeostatic turnover ( Klein and Simons , 2011 ) , and able choose any of the three possible fate outcomes following division -symmetric duplication , asymmetrical cell division in which one of the progeny commits to a differentiation pathway , or terminal division in which both cells undergo commitment ( Figure 5A ) . Formally , we can denote this behavior by the process , ( 2 ) S→λ{S+SPr . rS+DPr . 1−2rD+DPr . r , D→Γ∅ , where λ denotes the division rate of proliferative cells , S , r controls the balance between symmetric and asymmetric cell fates , and Γ represents the rate at which differentiated cells , D , are lost ( denoted as Ø ) . Here we do not discriminate the different differentiating cell types , D Rather , we assume that , once a cell has committed to a differentiation pathway , its proliferative potential and the lifetime of the differentiated cell progeny are both strictly limited , and characterised by the loss rate Γ . Although this model admits all three possible fate outcomes , if we choose r = 0 , we recover a process of invariant asymmetrical cell division in which each and every progenitor cell persists long-term . Models of this kind have been studied extensively in the literature , and notably in relation to the problem of interfollicular epidermal homeostasis , where lineage tracing studies in mice show that tissue is maintained through this pattern of stochastic cell fate with r≈0 . 1 ( Clayton et al . , 2007 ) . In this incarnation of the model , the fate choice is considered as a cell-autonomous or intrinsic process . Of course , in the present case , we can not rule out the possibility that the balance between proliferation and differentiation is regulated through extrinsic cues as would occur , for example , through neutral competition for limited niche space . Crucially , in the two-dimensional setting relevant to an epithelial tissue , long-term clonal evolution does discriminate between these possibilities ( Klein and Simons , 2011 ) . We will therefore follow this scheme , noting that the clonal fate data cannot shed light on this important issue . Although the clonal size dependence of the model cannot be determined straightforwardly by analytic computation , we can gain intuition and recover a good approximation for the clonal dynamics by studying clonal evolution of the proliferative cell population alone . For these cells , asymmetrical cell division does not change progenitor cell number . We are therefore led to the following ‘birth-death’ process , ( 3 ) S→2rλ{S+SPr . 1/2∅Pr . 1/2 , For such a critical process , it is straightforward to show that the chance of finding a surviving clone with nS>0 progenitor cells at a time t after the induction of single cells is given by , ( 4 ) PnSsurv . ( t ) =1rλt ( 1+1rλt ) =nS , with the average size of these surviving clones growing linearly with time as nSsurv . ( t ) ≡〈nS〉surv . =1+rλt . At the same time , the survival probability falls Psurv . ( t ) =1/ ( 1+rλt ) leading to the expected conservation law nssurv . ( t ) Psurv . ( t ) =1 , that is over time , the number of surviving clones forever diminishes while their size increases so that the average number of labelled cells remains constant . Although these results are formally exact , they can be further simplified in the long-time limit . In particular , for rλt≫1 , ( 5 ) PnSsurv . ( t ) ≈1nSsurv . ( t ) exp[−nSnSsurv . ( t ) ] , the clone size distribution acquires an exponential from . In this ‘scaling limit’ , we can also determine a good approximation for the dynamics of the total cell population . Since each progenitor cell will be associated with a ‘clonal unit’ of differentiating progeny of size 1/ρ , where ρ is set by the balance between cell production and loss , ρλ = ( 1−ρ ) Γ , we have the same exponential size dependence for the total clone size , ( 6 ) Pnsurv . ( t ) ≈1N ( t ) exp[−nSN ( t ) ] , where now the average size of the surviving clones grows as ( 7 ) N ( t ) =1+rλtρ . With this platform we now turn to consider the process of on-going clone induction following the sporadic accumulation of somatic mutations . If we assume that these mutations accumulate at a constant rate , starting at some initial time ( which we define as ‘time zero’ ) , the chance that a clone will have a total of n>0 cells is given by ( Klein et al . , 2010 ) ( 8 ) Pnsurv . ( t ) =1ln[ρN ( t ) ]exp[−n/N ( t ) ]n . while the clone survival probability is given by , ( 9 ) Psurv . ( t ) =ρrλtln[N ( t ) ] . From this last result , it follows that the clone density grows slowly ( logarithmically ) as ( 10 ) σ=RtPsurv . ( t ) =ρRrλln[N ( t ) ] . where R denotes the mutation rate ( per unit area ) . In particular , the leading dependence is on the ( inverse ) progenitor cell loss/replacement , 2rλ . For a higher rate of loss/replacement , we expect a smaller number of clones to persist . Although Equation 8 provides a concrete prediction for the clone size distribution , Pnsurv . ( t ) , its direct application to the experimental data is compromised by fluctuations due to small number statistics . Therefore , to address the experimental data , following ( Klein et al . , 2010 ) , we will use this result to construct a related distribution function from which a reliable comparison can be made . In particular , we may note that the cumulative sum , ( 11 ) μ1 ( n , t ) =∫0ndm mPmsurv . ( t ) ≈μ1 ( ∞ , t ) ( 1−exp[−n/N ( t ) ] ) , known as the first incomplete moment , is predicted to follow an exponential form . Therefore , if we plot the quantity 1−μ1 ( n , t ) /μ1 ( ∞ , t ) , we expect a pure exponential decay with a decay constant ρ/N ( t ) . With this result , we now turn to consider the experimental data . As discussed in the main text , we have acquired clonal fate data for six patients , three smokers and three non-smokers . First , to sensibly interpret the clonal fate data , it is necessary to make sure that the data is not compromised by clonal merger due to chance induction events at near-neighbouring sites in the tissue . For the smokers , we find a clone density of around 100 patches/cm2 . This translates to a typical clone separation of around a∼1000 microns , around two orders of magnitude larger than the typical cell diameter . Since , on this background , the chance of finding a clone with a separation r is given by ( 12 ) Y ( r ) =2a ( ra ) exp[− ( ra ) 2] , we can deduce that the chance a clone of radius r overlaps with another clone is given approximately by ∫0rdr P ( r ) =1−exp[− ( r/a ) 2] . For clones of size n cells , this equates to some exp[−n/κa2] , where κ is the cell density ( number per unit area ) . For clones of approximately 50 cells or more , this equates to around 1 in 20 clones . For these sizes , merger can play a role . However , the effect of merger is mitigated by the following effect: since small clones are more abundant , their contribution to larger clones will not significantly effect the size . But their absence from the cohort of smaller clones will influence the distribution . With these preliminaries , we now turn to the raw clone fate data ( Figure 4—figure supplement 1A , B ) . Although the first incomplete moment shows a characteristic exponential dependence over a wide range of clone sizes ( Figure 6—figure supplement 1A , B ) , for the largest clone sizes , the departure of the data from exponential behavior is significant . How then , can we understand this departure , and how can we use the data to extract N ( t ) and with it the growth parameters ? The departure at large clone sizes may derive from at least three independent sources . The first involves rare events: Although theory may correctly predict the existence of very large clones , their frequency may be very small . As a result , one would have to make a vast number of measurements to resolve the tail of the distribution . In a typical experiment , where the number of clones is limited , such clones may not appear at all . Such rare event phenomena will give rise to strong statistical fluctuations that will plague the data for large enough clone sizes . Second , the epithelial lining of the human airways is not flat but involves a ductal structure . While clones are small , the geometry will be flat and the clone sizes will be expected to conform to the basic paradigm . However , when clone sizes become sufficiently large , the warping of the clone can also lead to constraints that confine the growth of the clone . Again , this will lead to a reduction in clone fraction at larger sizes . Thirdly , as we have seen , the infrequent chance clone merger can influence the statistics leading to a mis-representation of clone size that , by the nature of the first incomplete moment , would most adversely effect large clones . Finally , the departure of theory and data for very large clone sizes could signal a breakdown of the basic modelling scheme . For example , if tissue plays host to a rare slow-cycling stem cell population , the growth dynamics could vary . In practice , since such clones are expected to persist long term , it is likely that such a population would compromise the smaller clone sizes where exponential distribution is most robust . Therefore , in the following , we will assume that the departure of the measured distribution from exponential is a pathology associated with rare event phenomena and small number statistics . Fortunately , to extract N ( t ) for the majority smaller clones , we can circumvent these difficulties by focusing on a statistic closely related to the first incomplete moment . In particular , if we consider the function , ( 13 ) μ1 ( n , t ) μ1 ( M , t ) ≈1−exp[−n/N ( t ) ]1−exp[−M/N ( t ) ] , where M denotes a large clone size cut-off beyond which we can expect rare fluctuations to dominate , we expect ( and indeed observe ) μ1 ( n , t ) /μ1 ( M , t ) to depend only weakly on M for a range of M values . From this fit , we can obtain N ( t ) . The results are shown in Figure 6A , B . | As air flows into our lungs , the lining of the nasal cavity , the throat and the rest of the respiratory tract prevents microbes , bacteria , dust and other small particles from entering the lungs . The lining of these airways is made up of many different types of cells , which must be continuously replaced as they become damaged . Experiments in mice have shown that cells called basal cells act as progenitor cells to keep the lining supplied with new cells . Progenitor cells are similar to stem cells: they divide to make , on average , one copy of themselves and one mature cell of another type ( such as a secretory cell ) . This ensures that healthy supply of progenitor cells is maintained for the future . However , it is not clear whether this process takes place at the level of individual progenitor cells or as an average for a population of cells . Teixeira et al . have now performed a study which shows that basal cells achieve this balance as a result of averaging . The study took advantage of the fact that cellular organelles called mitochondria have their own DNA , which gradually accumulates mutations over time . This makes it possible to identify groups of cells that are descended from a single progenitor cell because they will all contain the same mitochondrial mutation . By studying lung tissue from seven individuals , Teixeira et al . were able to identify clusters of related cells and found that , as expected , the size of the clusters increased with age . And by applying a mathematical model across all the cells in the study , it was discovered that whenever one basal progenitor cell committed to a particular fate , another progenitor cell duplicated itself: however , this balancing process happened in a random manner across a large number of cells , and not at the level of individual progenitor cells . Interestingly , it was found random cell division happened among smokers too , but was accelerated . This leads to clusters of identical cells forming more quickly in smokers than in non-smokers . In addition to providing further insights into the origins of lung cancer , the statistical methods developed by Teixeira et al . could be used to analyse the behaviour of many other types of stem or progenitor cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2013 | Stochastic homeostasis in human airway epithelium is achieved by neutral competition of basal cell progenitors |
The embryonic mouse lung is a widely used substitute for human lung development . For example , attempts to differentiate human pluripotent stem cells to lung epithelium rely on passing through progenitor states that have only been described in mouse . The tip epithelium of the branching mouse lung is a multipotent progenitor pool that self-renews and produces differentiating descendants . We hypothesized that the human distal tip epithelium is an analogous progenitor population and tested this by examining morphology , gene expression and in vitro self-renewal and differentiation capacity of human tips . These experiments confirm that human and mouse tips are analogous and identify signalling pathways that are sufficient for long-term self-renewal of human tips as differentiation-competent organoids . Moreover , we identify mouse-human differences , including markers that define progenitor states and signalling requirements for long-term self-renewal . Our organoid system provides a genetically-tractable tool that will allow these human-specific features of lung development to be investigated .
During mouse lung development the distal tip epithelial cells are SOX9+ , ID2+ and function as multipotent progenitors producing first bronchiolar , and then alveolar , descendants ( Alanis et al . , 2014; Rawlins et al . , 2009 ) . Between ~E10–15 cells that exit the distal tip turn off SOX9 , upregulate SOX2 and differentiate along bronchiolar lineages . Whereas , ~E16–18 cells exiting the tip turn off SOX9 and co-express markers of alveolar type 1 ( AT1 ) and alveolar type 2 ( AT2 ) fate . As morphogenesis proceeds these bipotent cells line developing alveolar sacs and differentiate as mature AT1 or AT2 cells ( Desai et al . , 2014; Treutlein et al . , 2014 ) . Many factors controlling self-renewal and differentiation in the developing mouse lung epithelium have been identified . By contrast , relatively little is known about human lung development . This is largely due to practical considerations about tissue availability and culture system limitations ( Benlhabib et al . , 2015; Haitchi et al . , 2009; Rajatapiti et al . , 2010 ) . A small number of human studies show the detailed expression of specific genes ( Al Alam et al . , 2015; Gonzalez et al . , 1996; Khoor et al . , 1993 , Khoor et al . , 1994; Laresgoiti et al . , 2016; Stahlman et al . , 2007; Zhang et al . , 2012 ) . Whereas , transcriptomics has provided a genome-wide view of human lung developmental transitions , but currently lacks cellular resolution ( Feng et al . , 2014; Kho et al . , 2010 , 2016 ) . Improved in vitro models of human disease are needed to complement available mouse models . One recent approach to disease modelling is to use self-renewing human organoids which recapitulate aspects of organ morphogenesis/physiology ( Dekkers et al . , 2013; Ettayebi et al . , 2016; Huch et al . , 2015 ) . Human organoids are typically derived from adult stem cells limiting their use for studying paediatric disease and disease progression . An alternative is to derive the organ of interest from pluripotent stem cells by directed differentiation ( Dye et al . , 2016; Merkle and Eggan , 2013 ) . The ability to in vitro self-renew and differentiate bona fide human lung tip progenitors could provide a genetic system for fundamental developmental biology and paediatric disease modelling . Moreover , an improved understanding of human lung progenitor states and human fate specification would facilitate strategies for directed differentiation of pluripotent cells . We have extensively characterized human epithelial tip progenitors , and the early stages of fate specification , revealing mouse-human differences in the expression of key marker genes including SOX2 and pro-SFTPC . We developed methods to grow human tip epithelium as long-term self-renewing , branching , organoids and therefore investigate tip signalling requirements and differentiation capacity . Our human tip organoids can be directed to differentiate towards alveolar or bronchiolar fate in vitro and can engraft into the adult or developing mouse lung . These experiments confirm that human and mouse tips are analogous in function and validate our organoid conditions . These conditions are sufficient to convert differentiating human embryonic lung epithelial stalks to tip fate , illustrating the plasticity of the developing lung . However , they do not support the long-term self-renewal of mouse tips , highlighting species-specific regulatory differences . Our organoid system thus provides a genetically-tractable in vitro model to accelerate studies of human lung development .
We characterized the evolution of marker gene expression in human embryonic lungs 6–21 pcw ( post-conception weeks; Figure 1—figure supplement 1 ) . This period covers the pseudoglandular stage ( ~5–16 pcw ) when the bronchiolar tree is established and the early canalicular stage ( ~16–21 pcw ) when alveolar sac formation begins ( Burri , 1984; Rackley and Stripp , 2012 ) . Throughout the time-course , we observed SOX9 ( SRY-box 9 ) localized to the distal epithelial tips ( Figure 1A , B ) . In the adult human lung , pro-SFTPC ( Surfactant Protein C , or SPC ) and a monoclonal antibody , HTII-280 , are markers of AT2 cells ( Figure 1C ) ( Barkauskas et al . , 2013; Gonzalez et al . , 2010 ) . We first detected low levels of HTII-280 in SOX9- epithelium at 11 pcw , adjacent to the distal tip ( Figure 1A ) . Similarly at 14 and 17 pcw ( Figure 1—figure supplement 2A ) . At 20 pcw , HTII-280 was more heterogeneous in the alveolar sacs with cells having either high , or undetectable , levels ( Figure 1B ) . We confirmed that HTII-280 was not in the larger airways ( Figure 1B , arrowhead ) , but it was ubiquitous in the columnar epithelium of the terminal bronchioles ( Figure 1B , arrow ) . By contrast , pro-SFTPC was first detected at very low levels at 17 pcw , particularly in tip epithelium ( Figure 1—figure supplement 3D ) , and more robustly in distal squamous cells by 20 pcw ( Figure 1D–F ) . At 20 pcw , pro-SFTPC was mostly co-expressed with HTII-280 ( Figure 1F’ ) , although we also observed a small number of pro-SFTPC+ , HTII-280- cells and many more pro-SFTPC- , HTII-280+ cells . This human embryonic pro-SFTPC staining differs from that observed in mouse development in which pro-SFTPC is expressed throughout the epithelium from ~E10 , increases in canalicular stage tips and is further up-regulated in differentiating AT2 cells ( Laresgoiti et al . , 2016; Wuenschell et al . , 1996 ) . However , our human data are consistent with previous reports of pro-SFTPC staining in human embryos ( Khoor et al . , 1994 ) . We previously used LPCAT1 ( Lysophosphatidylcholine acyltransferase 1 ) as a specific marker of alveolar/AT2 fate in mouse development ( Laresgoiti et al . , 2016 ) . However , it was expressed widely throughout the epithelium of the developing human lung from six pcw , although specific to AT2 cells in the adult ( Figure 1—figure supplement 2E , F ) . ABCA3 ( ATP binding cassette subfamily A member three ) is present on the surface of lamellar bodies in mature AT2 cells , but has not been detected in human embryonic lungs prior to 28 weeks gestation ( Stahlman et al . , 2007 ) . Consistent with this , we could not detect ABCA3 in any of the embryonic lungs we stained ( 6–21 pcw ) , but reproducibly saw expression in adult AT2 cells ( Figure 1—figure supplement 2F ) . This suggests that mature lamellar body-containing AT2 cells are not present in human embryonic lungs by 21 pcw consistent with previous analysis ( Oulton et al . , 1980 ) . 10 . 7554/eLife . 26575 . 003Figure 1 . Evolution of alveolar and bronchiolar marker gene expression during human embryonic lung development . Sections of human embryonic and adult lungs . ( A , B ) 11 and 20 pcw . Green: HTII-280; red: SOX9 ( tips ) ; white: ECAD ( epithelial cells ) . Arrow = HTII-280 positive terminal airway . ( C ) Adult . Green: HTII-280 ( AT2 cells ) ; red: pro-SFTPC ( AT2 cells ) . Arrow = selected AT2 cells . ( D–F ) 11 , 17 and 20 pcw . Green: HTII-280; red: pro-SFTPC . ( G–H ) 11 and 20 pcw . Green: HTII-280; red: HOPX; white: SOX2 ( bronchiolar cells ) . Dotted bracket = terminal airway cells co-expressing SOX2 and HTII-280 . Arrowheads = differentiating AT1 cells . ( I , J ) 11 and 20 pcw . Green: HTI-56; red: HOPX . K . Adult . Red: HOPX ( AT1 cells ) . Arrowheads = HOPX+ nuclei . ( L ) 7 pcw . Green: TP63; red: SOX2 . Arrowheads = TP63/SOX2 dual-positive cells in the more proximal airway region . Boxed region is shown as an inset in L’ with channels separated . ( M ) Adult . Red: TP63 ( basal cells ) . Arrowheads = TP63+ cells in an intra-lobar bronchiole . ( N ) Adult . Green: FOXJ1 ( ciliated cells ) ; red: SCGB1A1 ( secretory cells ) . ( O , P ) 11 pcw proximal and distal airway from the same lung . Green: TP63; red: SOX2 . Arrowheads = TP63/SOX2 dual-positive cells in the more distal airway . ( Q–S ) 16 pcw proximal and distal airway from the same individual . Green: KRT5; red: TP63 . Arrows = developing sub-mucosal glands . Arrowheads = KRT5+ , TP63- cells . Dashed lines = patches of KRT5+ , TP63+ cells . ( T ) 16 pcw proximal airway . Green: MUC5AC . Arrows = developing sub-mucosal gland . Arrowheads = mucous cells . Blue: DAPI ( nuclei ) . Bars = 100 μm ( A , B , D , E , F , G , H , I , J , Q . R ) ; 50 μm ( B’ , C , F’ , H’ , H’’ , J’ , K , L , L’ , M , N , O , O’ , P , P’ , S , T ) ; 25 μm ( insets in C , F’ , I , L’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 00310 . 7554/eLife . 26575 . 004Figure 1—figure supplement 1 . Representative morphology of human embryonic lung samples . Wholemount images of human embryonic lungs . ( A ) CS15 ( ~5 pcw ) . ( B ) 6 pcw . ( C ) 8 pcw . ( D ) 10 pcw . ( E ) 16 pcw . Bars = 2 . 5 mm ( A , B ) ; 5 mm ( C , D ) ; 3 cm ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 00410 . 7554/eLife . 26575 . 005Figure 1—figure supplement 2 . The evolution of alveolar marker gene expression in the human embryonic lung . Sections of human embryonic and adult lungs . ( A ) 14 and 17 pcw . Green: HTII; red: SOX9 ( tips ) ; white: ECAD ( epithelial cells ) . ( B ) 14 and 17 pcw . Green: HTII-280; red: HOPX; white: SOX2 ( bronchiolar cells ) . ( C ) 20 pcw Green: HTII-280; red: HOPX; white: SOX2 ( bronchiolar cells ) . Dotted line indicates SOX2+ cells that do not express the alveolar markers . Arrow heads = triple-positive ( SOX2+ , HTII-280+ , HOPX+ ) cells . Green/white and red/white channels also shown separately for clarity . ( D ) 11 , 14 and 17 pcw . Green: HTI-56; red: SOX2 . ( E ) 16 pcw . Green: LPCAT1; red: SOX2 . ( F ) Adult . Green: LPCAT1 ( AT2 cells ) ; red: ECAD ( epithelial cells ) . ( F’ ) Adult . Green: ABCA3 ( AT2 cells ) ; red: ECAD ( epithelial cells ) . Blue: DAPI ( nuclei ) . Bars = 50 μm all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 00510 . 7554/eLife . 26575 . 006Figure 1—figure supplement 3 . Epithelial PDPN expression is not specific to AT1 cells during human lung development . Sections of human and adult embryonic lungs . ( A , A’ ) nine pcw . Green: NKX2-1; red: PDPN . Arrowheads = low levels of PDPN in the developing airways . Arrows = PDPN in lymphatic endothelial cells . ( B ) 14 pcw . Green: ECAD ( epithelial cells ) ; red: PDPN . Arrowheads = PDPN is stronger in the more distal epithelium than the developing airways ( arrow ) . ( C ) 17 pcw . Green: SOX9 ( epithelial tips ) ; red: PDPN; white: ECAD ( epithelial cells ) . Low levels of PDPN can be seen in the tips , likely due to the angle of the section . ( D ) 17 pcw . Green: pro-SFTPC; red: PDPN; white: HTII-280 . Arrowheads = PDPN and HTII-280 are largely co-expressed in cells that have already exited the tips . ( E ) 20 pcw . Green: SOX9 ( epithelial tips ) ; red: PDPN . Arrowheads = PDPN is observed in cells that have exited the epithelial tips . ( F ) 17 pcw . Green: pro-SFTPC; red: PDPN; white: HTII-280 . The majority of cells lining the developing alveolar sacs are PDPN+ , HTII-280+ dual-positive , although levels of each protein vary ( seen clearly in F’ and F’’ ) . ( G ) 20 pcw . Green: NKX2-1; red: PDPN . Epithelial cells lining the alveolar sacs are mostly PDPN+ , low levels of PDPN are still observed in the developing bronchioles ( arrow ) . ( H , H’ ) 20 pcw . Green: AQP5; red: PDPN . Line = rows of continuous epithelial cells lining parts of the developing alveolar sacs are AQP5+ , PDPN+ dual-positive . Arrows = distinct , individual , cells are AQP5+ , PDPN+ dual-positive and express a higher level of AQP5 ( likely developing AT1 cells ) . Arrowhead = individual cells that are AQP5- , PDPN- dual-negative ( likely developing AT2 cells ) can also be distinguished . ( I ) Adult lung . Green: pro-SFTPC; red: PDPN . Arrowheads = adult AT2 cells are pro-SFTPC+ and PDPN- . All panels: blue = DAPI ( nuclei ) . Bars = 200 μm A; 50 μm B , C , D , E , F , F’ , F’’ , G , H , H’’ , I; 25 μm inset in D . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 00610 . 7554/eLife . 26575 . 007Figure 1—figure supplement 4 . Epithelial AQP5 expression is not specific to AT1 cells during human lung development . Sections of human and adult embryonic lungs . ( A–D ) 11 , 14 , 15 and 20 pcw . Green: AQP5; red: ECAD ( epithelial cells ) . Patches of diffuse , apical AQP5 staining can be observed from 11 pcw . Arrowheads = elongated cells with more intense AQP5 staining observed from 20 pcw . However , AQP5 expression is widespread and also observed in cells with a more columnar epithelial appearance ( arrows ) . ( E ) 20 pcw . Green: SOX9 ( epithelial tips ) ; red: AQP5; white: ECAD ( epithelium ) . AQP5 extends to the epithelial tips . ( F ) Adult alveolar region . Red: AQP5; white ECAD ( epithelial cells ) . Arrowheads = AT1 staining . ( G ) Adult terminal bronchiole . Green: TP63 ( basal cells ) ; red: AQP5; white: ACT ( ciliated cells ) . Arrowheads = AQP5+ secretory cells . All panels: blue = DAPI ( nuclei ) . Bars = 50 μm all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 00710 . 7554/eLife . 26575 . 008Figure 1—figure supplement 5 . NKX2-1 and FOXA2 are expressed in all human lung epithelial cells up to 20 pcw . Sections of human embryonic lungs . ( A–D ) . 15 , 17 , 20 and 21 pcw . Green: NKX2-1; red: FOXA2; white: HTII-280 . Arrowheads in D’ = NKX2-1- cells apparently lining the developing alveolar sacs . E-G . 20 pcw . ( E ) Green: NKX2-1; red: PDPN . Arrowheads = apparently PDPN+ , NKX2-1- cells lining the developing alveolar sacs . ( F ) Green: NKX2-1; red: PDPN; white: ECAD ( epithelial cells ) . Arrowheads = NKX2-1- cells are also ECAD- and therefore not epithelial . ( G ) Green: NKX2-1; red: VECAD ( vascular endothelium ) . Mesenchymal cells can appear to line the alveolar sacs in thin sections . Arrowheads = NKX2-1- , VECAD+ endothelial cells apparently lining the developing alveolar sacs . All panels: blue = DAPI ( nuclei ) . Bars = 100 μm A; 50 μm B , C , D’ , E , F , G; 200 μm D . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 008 In mouse development the AT1 marker HOPX ( Hop Homeobox ) is first detected ubiquitously in cells that have exited the canalicular stage distal tip , subsequently becoming heterogeneous and marking future AT1 cells ( Jain et al . , 2015; Laresgoiti et al . , 2016 ) . Similarly , we detected HOPX in cells that had exited the distal tip at 11 pcw , although there was also co-expression of HOPX and SOX2 ( SRY-Box 2 ) in the smaller airways ( Figure 1G ) . HOPX and HTII-280 were frequently co-expressed between 11 and 17 pcw ( Figure 1—figure supplement 2B ) . By 20 pcw , a mixture of single and co-expressing distal epithelial cells were observed lining the developing alveolar sacs ( Figure 1H ) . Interestingly , at 20 pcw cells in the terminal bronchioles co-expressed SOX2 , HTII-280 and HOPX ( Figure 1H’ and Figure 1—figure supplement 2C ) . A monoclonal antibody , HTI-56 , has been reported as human AT1 cell specific ( Dobbs et al . , 1999; Gonzales et al . , 2015 ) . Consistent with this , we observed the highest levels of expression in HOPX+ cells adjacent to the distal tip from 11 pcw ( Figure 1I ) . By 20 pcw HTI-56 was becoming heterogeneous , whilst HOPX remained ubiquitous ( Figure 1J , J’ ) . However , we also observed a low level of HTI-56 expression throughout the SOX2+ airways , beyond the terminal bronchioles , making it less useful as an AT1 marker in vitro ( Figure 1—figure supplement 2C ) . In adult human alveoli a relatively small number of HOPX+ nuclei can be observed whilst the entire alveolar surface is covered by HOPX+ cell membranes extending from the AT1 cells ( Figure 1K ) . PDPN ( Podoplanin , also known as T1α ) protein is detected in mouse lung epithelium from the time of alveolar specification at ~E16 . 5 and becomes restricted to differentiating AT1 cells by E18 . 5 ( Laresgoiti et al . , 2016 ) . It is also a marker of airway basal cells and lymphatic endothelium ( Breiteneder-Geleff et al . , 1999; Farr et al . , 1992 ) . We detected low levels of apical PDPN expression in columnar ( not basal ) cells of the developing human bronchioles from 9 pcw ( Figure 1—figure supplement 3A ) . At later stages expression is stronger in the more distal regions where PDPN and HTII-280 are co-expressed ubiquitously on the developing alveolar surface at 17 pcw , but PDPN is absent from the distal tips ( Figure 1—figure supplement 3B–G ) . By 20 pcw PDPN can be observed to be co-expressed with AQP5 ( Aquaporin 5 ) in distinct , individual cells , rather than running continuously through the alveolar duct ( Figure 1—figure supplement 3H ) . It is strongly expressed in adult AT1 cells as expected ( Figure 1—figure supplement 3H ) . AQP5 has been described in mouse as specifically detected in differentiating AT1 cells and not expressed in the bipotent alveolar progenitors which co-express AT1 and 2 markers ( Desai et al . , 2014 ) . It is therefore a putative marker of differentiating AT1 cells . We first detect AQP5 in distal regions of the lung from 11 pcw ( Figure 1—figure supplement 4A–C ) . By 20 pcw it can be observed to be expressed ubiquitously in the low columnar epithelium that lines the developing alveolar ducts and extends to the distal tips . It is expressed at higher levels in elongating cells with very short membrane extensions ( Figure 1—figure supplement 4D , E; these are similar in appearance to the HOPX+ cells marked with arrowheads in Figure 1H’’ and are likely to be differentiating AT1 cells that are beginning to extend their membranes ) . In the adult alveoli AQP5 is highly expressed in AT1 cells , but it can also be observed on the surface of non-ciliated cells in the lower airways ( Figure 1—figure supplement 4F , G ) . NKX2-1 is ubiquitously expressed in lung epithelial cells from the time of lung specification in the foregut ( Lazzaro et al . , 1991 ) . NKX2-1 levels have been reported to be heterogeneous in mature alveolar cells in the adult mouse and rat ( Desai et al . , 2014; Liebler et al . , 2016 ) . We observed that NKX2-1 was co-expressed with FOXA2 in all human embryonic lung epithelial cells from 9 to 21 pcw , albeit at higher levels in alveolar-fated than bronchiolar-fated cells ( Figure 1—figure supplement 5A–C ) . NKX2-1- , FOXA2- cells which apparently lined the more mature alveolar ducts were identified , but closer inspection revealed that these were ECAD- non-epithelial cells their presence at the alveolar surface likely being an artefact due to the angle of sectioning ( Figure 1—figure supplement 5D–G ) . Our results show that the broad picture of mouse alveolar development ( low-level co-expression of AT1 and 2 markers in cells that exit the canalicular stage distal tip , followed by lineage-specific expression as morphogenesis proceeds ) is conserved in human . However , even in this limited analysis there are striking mouse-human differences . The timing of alveolar marker gene onset in human is ~11 pcw , preceding the canalicular stage by approximately five weeks . In addition , the relative timing of expression of specific markers can differ ( pro-SFTPC , LPCAT1 , AQP5 ) . Moreover , the developing human terminal bronchioles are SOX2+ , but co-express alveolar markers which to our knowledge has never been observed in mouse ( Laresgoiti et al . , 2016 ) . The first signs of cellular heterogeneity/differentiation in the bronchioles were observed from 7 pcw where we saw that a sub-set of SOX2+ cells co-expressed low levels of TP63 ( Tumour Protein P63 ) in the larger airways ( Figure 1L , arrowheads ) . These are likely to be differentiating basal cells which are TP63+ in the adult ( Figure 1M , arrowheads ) ( Rock et al . , 2009 ) . FOXJ1 and SCGB1A1 were also readily detected in adult airways ( Figure 1N ) . By 11 pcw , TP63 staining was much stronger and localised to a sub-set of basally-located airway nuclei ( Figure 1O ) . TP63 was also stronger in more proximal , versus distal , bronchioles within the same lungs ( Figure 1O , P ) . We first detected KRT5 ( Keratin 5 ) at 16 pcw in the larger airways , particularly in invaginating submucosal gland buds ( Figure 1Q–S ) . KRT5 was also seen in patches of TP63+ airway cells , likely differentiating basal cells ( Figure 1—figure supplement 1S , dotted lines ) . KRT5+ , TP63- cells were also observed , frequently in a non-basal position ( Figure 1—figure supplement 1S , arrowheads ) . The first signs of columnar cell differentiation were detected at 16 pcw when rare MUC5AC+ ( Mucin 5AC ) cells were identified in the proximal airways ( Figure 1T ) . This is consistent with a proximal-distal pattern of airway differentiation as in other species ( Plopper et al . , 1992; Toskala et al . , 2005 ) . In the pseudoglandular stage mouse lung , there is a clear demarcation between SOX9+ tip and SOX2+ stalk ( differentiating bronchiole ) cells and multiple signalling mechanisms regulate the boundary between the two populations ( Hrycaj et al . , 2015; Mahoney et al . , 2014; Wang et al . , 2013 ) . We observed a tip-stalk boundary in the pseudoglandular stage human lungs with SOX9 restricted to the tip and SOX2 expressed highly in the stalk . Differentiating αSMA+ ( α-Smooth Muscle Actin ) smooth muscle was observed around the SOX2+ future airways and SOX9 was present throughout the mesenchyme at low levels ( Figure 2A ) . We noted that human tip epithelium was more proliferative than stalk ( Figure 2B ) , consistent with mouse results ( Okubo et al . , 2005 ) . However , in contrast to the mouse , low levels of SOX2 were co-expressed with SOX9 in the distal tip epithelium throughout the pseudoglandular stage ( Figure 2B; Figure 2—figure supplement 1A–C ) . SOX2 , SOX9 co-expression at the tip was confirmed by qRT-PCR in microdissected tip and stalk cells ( Figure 2—figure supplement 1D ) . Further examination of our time-course revealed that SOX2 gradually decreased over time and disappeared from the tip epithelium at the transition to the canalicular stage of development . This happened heterogeneously throughout the lung . For example , at 17 pcw we observed a mixture of SOX2+ and SOX2- distal tips within individual lungs ( Figure 2C , D; Figure 2—figure Supplement 2 ) . However , by 20 pcw all distal tips were SOX2- ( Figure 2E; Figure 2—figure Supplement 2 ) . Moreover , there was a SOX2- , SOX9- zone adjacent to the 20 pcw distal tips which corresponds to the developing saccules where markers of alveolar differentiation are expressed ( compare Figure 2E with Figure 1H ) . 10 . 7554/eLife . 26575 . 009Figure 2 . The tip and stalk epithelial cell populations are clearly demarcated in branching human , pseudoglandular stage , lungs . ( A–E ) Sections of human embryonic lungs . ( A ) 11 pcw . Green: SOX9 ( tip ) ; red: SOX2 ( stalk ) ; white: α-SMA ( smooth muscle ) . ( B ) 8 pcw . Green: SOX9 ( tip ) ; red: SOX2 ( stalk ) ; white: KI67 ( proliferating cells ) . ( C , D ) 17 pcw . ( E ) 20 pcw . Green: SOX9 ( tip ) ; red: SOX2 ( stalk ) . Arrowheads = SOX9+ , SOX2- co-expressing tips . Arrows = SOX9+ , SOX2- tips . Blue: DAPI ( nuclei ) . ( F ) Experimental schematic for tip versus stalk RNAseq . ( G ) Venn diagram showing common and differentially-expressed transcripts based on a fold-change of at least 2 . ( H ) Unsupervised hierarchical clustering of tip , stalk and published foetal lungs based on the differentially-expressed genes . ( I ) Chart to show the percentage of the gene ontology classes represented in the differential expression data . ( J ) List of transcription factors enriched at least two-fold in the tips . Number in brackets indicates fold-change over the stalk . * indicates reported mouse tip expression , see Supplementary file 2 . ( K ) List of differentiation markers enriched in the stalk . ( L ) List of transcription factors enriched in stalks that were previously reported as expressed in the mesenchyme . Scale bars = 50 μm ( A , C , D ) ; 100 μm ( B , E ) ; 2 mm ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 00910 . 7554/eLife . 26575 . 010Figure 2—figure supplement 1 . Pseudoglandular stage human lung tips co-express SOX9 and SOX2 . ( A ) E13 . 5 mouse lung staining illustrating absence of SOX2 in SOX9+ tip cells . Green: SOX9; red: SOX2; white: ECAD . ( B ) Whole-mount staining of a 5 pcw human embryonic lung showing the primary branches and SOX2/SOX9 co-expression in the tips ( boxed area in B is magnified in B’ ) . Green: SOX9; red: SOX2; white: ECAD . ( C ) Confocal image of a 8 pcw human embryonic lung illustrating SOX2 expression in SOX9+ tip cells . Green: SOX9; red: SOX2 . ( D ) qRT-PCR of microdissected tip and stalk compared to whole lung: SOX9 and SOX2 data confirm immunostainings . TBX4 is a mesenchymal marker . ID2 is another well-characterised mouse tip gene . Scale bars = 50 μm ( A , C ) ; 200 μm ( B ) ; 100 μm ( B’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01010 . 7554/eLife . 26575 . 011Figure 2—figure supplement 2 . Human lung tips down-regulate SOX2 during the canalicular stage . ( A–G ) Sections of human embryonic lungs at 11 , 14 , 17 and 20 pcw stained for green: SOX9; red: SOX2 . Slides were stained simultaneously and imaged using the same microscope settings such that protein levels can be compared between lungs . Each row shows representative images from a different individual lung . SOX2 is also shown as a separate channel . SOX2 is significantly down-regulated in the epithelial tips by 17 pcw , although still present in some tips . It is absent in the tips by 20pcw . Arrows indicate SOX9+ tips . ( A ) 11 pcw . ( B , C ) 14 pcw . ( D–F ) 17 pcw . G . 20 pcw . Blue = DAPI ( nuclei ) . Bar = 50 μm in all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01110 . 7554/eLife . 26575 . 012Figure 2—figure supplement 3 . Transcriptional differences and similarities between human pseudoglandular stage tip and stalk populations . ( A ) Chart to show the percentage of the expressed genes in the major gene ontology classes represented in the data . ( Note: tip and stalk data are reproduced from Figure 2I ) . ( B–E ) Human embryonic lung sections . ( B ) 11 pcw . Green: NKX2 . 1 ( lung epithelium ) ; red: FOXA2 ( lung epithelium ) ; white: ECAD ( epithelial cells ) . ( C ) 11 pcw . Green: β-III-Tubulin ( differentiating neurons , including neuroendocrine cells ) ; red: SOX2; white: PECAM ( endothelial cells ) . ( D ) 12 pcw . Green: β-III-Tubulin ( differentiating neurons , including neuroendocrine cells ) ; red: SOX2 . Arrowheads = β-III-Tubulin+ , SOX2+ neuroendocrine cells . ( E ) 10 pcw . Green: SPRY2; red: SOX2; white: ECAD ( epithelial cells ) . ( F , G ) 8 pcw . Green: FGFR2; red: SOX2; white: ECAD ( epithelial cells ) . Scale bars = 100 μm ( A , B , C , E , F , G ) ; 50 μm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01210 . 7554/eLife . 26575 . 013Figure 2—figure supplement 4 . Specific transcription factors are enriched at the protein level in human distal epithelial lung tips . ( A–D ) Sections of 8 pcw human embryonic lung . ( A ) Green: ETV5; red: SOX2; white: ECAD ( epithelial cells ) . ( B ) Green: HMGA2; red: SOX2; white: ECAD ( epithelial cells ) . ( C ) Green: HNF1B; red: SOX2; white: ECAD ( epithelial cells ) . ( D ) Green: ID2; red: HMGA1; white: ECAD ( epithelial cells ) . Arrows = distal tips . Blue = DAPI ( nuclei ) . Bars = 50 μm all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01310 . 7554/eLife . 26575 . 014Figure 2—figure supplement 5 . Tip gene expression is highly conserved between mouse and human . ( A ) Pie chart showing over-lap of orthologous human-mouse genes called as present in 6–7 pcw human tips and E11 . 5 mouse tips . ( B ) Scatter plot to estimate the relative levels of expression of orthologous genes between mouse and human tips . We reasoned that the microarray signal saturates and therefore generated a scatter plot of mean microarray signal , versus mean log-transformed RPKM for each orthologous gene identified . r2 = 0 . 423; p<10−16 . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 014 To compare tip-stalk gene expression globally we microdissected tip and stalk epithelium from four individual 6–7 pcw lungs and performed RNAseq ( Figure 2F ) . We detected a total of 15 , 599 transcripts , with the majority ( 86% ) expressed in both tip and stalk ( Figure 2G ) . This high level of similarity was expected since the stalk cells are the immediate descendants of the tip population . Using a two-fold difference in expression level cut-off , 2208 genes that were enriched in stalk or tip were also identified . Non-hierarchical clustering of the samples based on these 2208 transcripts revealed a clear separation into distinct tip and stalk populations ( Figure 2H ) . Moreover , these genes were expressed at similar levels in published whole human foetal lung RNAseq ( Figure 2H ) ( Bernstein et al . , 2010 ) . Gene Ontology ( GO ) analysis of the differentially expressed transcripts included categories related to cell signalling , proliferation , adhesion , motility , transcription and developmental processes expected for embryonic progenitors ( Figure 2I ) . These GO categories were also a major feature of genes that were co-expressed in both tip and stalk ( Figure 2—figure supplement 3A ) . NKX2 . 1 and FOXA2 , which we detect as ubiquitous throughout the developing human lung epithelium ( Figure 2—figure supplement 3B ) , were not enriched in tip or stalk . The human tip-enriched data set contained 37 genes annotated as transcription factors of which 54% ( 20/37 ) had previously been characterized as expressed in mouse tips ( Figure 2J; Supplementary files 1 and 2 ) . These included GATA6 , HMGA2 , MYCN and SOX9 which have documented tip-specific functions in mouse ( Chang et al . , 2013; Rockich et al . , 2013; Singh et al . , 2014; Zhang et al . , 2008 ) . Moreover , we were able to confirm human tip-enrichment of ETV5 , HMGA1 , HMGA2 , HNF1B and ID2 at the protein level ( Figure 2—figure supplement 4 ) . Among the tip transcription factors , we identified only one gene , MEIS2 , which is likely to be mesenchymally expressed ( Diez-Roux et al . , 2011; Herriges et al . , 2012 ) , suggesting a very low level of mesenchymal contamination . A comparison of our human tip RNAseq with a previously published mouse tip microarray ( Laresgoiti et al . , 2016 ) found that 96% of orthologous genes that were expressed in human tips were also present in mouse ( Figure 2—figure supplement 5A ) . There was also a good correlation between levels of mouse-human orthologous gene expression ( Figure 2—figure supplement 5B ) . We detect the first signs of cellular heterogeneity in the stalk by 7 pcw ( Figure 1L ) . Consequently , we were able to identify stalk-specific transcripts that are characterized as expressed in differentiating airway cells ( Figure 2K ) . These included , basal , neuroendocrine and club cell markers ( Guha et al . , 2012; Ito et al . , 2000; Jia et al . , 2015; Lan and Breslin , 2009; Rock et al . , 2009 ) . Consistent with this , we observed neuroendocrine cells in pseudoglandular stage airways ( Figure 2—figure supplement 3C , D ) . When we examined the stalk transcripts annotated as transcription factors in more detail , we noticed that 52% ( 67/128 ) had published mouse airway expression patterns . However , 55% of these ( 37/67 ) were expressed in mouse mesenchyme , rather than epithelium ( Figure 2L ) , suggesting a high level of mesenchymal contamination in the stalk samples . Within the RNAseq data , we identified components of the EGF , FGF , Hedgehog , IGF , Notch , Retinoic Acid , TGF-β super-family and WNT signalling pathways ( Supplementary file 1 ) . We noted that while core downstream signalling components were transcribed in both tip and stalk; ligands , receptors and inhibitors were more likely to be tip or stalk specific . FGF Receptor signalling is central to mouse lung branching morphogenesis ( Volckaert and De Langhe , 2015 ) and protein expression of several FGF pathway components was confirmed by antibody staining ( Figure 2—figure supplement 3E–G ) . There were various subtle differences in tip signalling pathways between mouse and human . For example , BMP2 and BMP7 were highly enriched in human tips compared with Bmp4 in the mouse ( Bellusci et al . , 1996 ) and IHH in human where mouse has Shh ( Bellusci et al . , 1997 ) . This analysis suggests that the human tip epithelium is analogous to the mouse population with a highly conserved transcriptome and similar signalling pathway activity . However , we also observe differences that are likely to be functionally significant . The mouse distal tip population is a long-lived progenitor that self-renews extensively throughout normal lung development . Moreover , our recent heterochronic grafting experiments demonstrated that its behaviour is largely controlled by extrinsic signals ( Laresgoiti et al . , 2016 ) . We therefore reasoned that we should be able to capture tip self-renewing behaviour in vitro by supplying the correct combination of factors . This would be analogous to the long-term self-renewal of blastocyst inner cell mass as ES ( Embryonic Stem ) cells . We microdissected human epithelial tips from 5 to 9 pcw lungs ( as in Figure 2F ) and plated them in Matrigel in the presence of 7 factors: EGF , FGF7 , FGF10 , NOG ( Noggin ) , RSPO1 ( R-spondin 1 ) , a GSK3β inhibitor CHIR99021 and a TGFβ inhibitor SB431542 . Factor choice was based on the conditions used to grow adult foregut derivatives as organoids , the extensive literature on mouse lung development and our RNAseq analysis ( Huch et al . , 2013a , 2013b; Sato et al . , 2009; Swarr and Morrisey , 2015; Yin et al . , 2014 ) . In these conditions human lung epithelial tips formed organoids with 100% colony forming efficiency ( n = 303 tips from 13 individuals ) . Tips formed spheres within 12 hr , expanded spherically for 6–8 days and then branched; by culture day 14 the organoids resembled a mass of tips ( Figure 3A; Video 1 ) . We passaged the organoids every 2 weeks by mechanically breaking into smaller pieces and re-plating . Growth continued in a similar fashion in later passages , although the morphological appearance of the cultures became more heterogeneous depending on the extent of breakage during passaging ( Figure 3A ) . Tip organoids retained SOX2 and SOX9 expression over multiple passages ( Figure 3B ) . Moreover , they expressed the lung-specific transcription factor NKX2-1 and the tip-specific marker proteins that we have validated ( Figure 3—figure supplement 1 ) . Organoids possessed a lumen and were composed of a single epithelial layer ( Video 2 ) similar to the in vivo morphology of the tip epithelium ( Figure 2—figure supplement 1B , C ) . Every tip organoid line we isolated continued to grow in the same way for at least 9 passages over 4 months ( n = 11 organoid lines were maintained for >9 passages without apparent change in morphology , or SOX2 and SOX9 expression; some cultures have been maintained for up to 9 months ) . Organoid karyotypes were also normal ( Figure 3—figure supplement 2 ) . We concluded that activation of EGF , FGF and WNT signalling , and inhibition of BMP and TGFβ , are sufficient to grow human epithelial tips as long-term , self-renewing organoids with an initial colony forming efficiency of 100% . 10 . 7554/eLife . 26575 . 015Figure 3 . Long-term , self-renewing organoid culture of human lung epithelial tip cells with a initial colony forming efficiency of 100% . ( A ) Frames from Video 1 showing bright field images of a single microdissected tip taken every 24 hr for 12 days . Representative bright field images of tip organoid cultures from P0 , P6 and P15 . A typical organoid after matrigel removal is shown and after further microdissection of branched structures ( inset ) . ( B ) Confocal images of tip organoids at P0 , P6 and P15 . Green: SOX9; red: SOX2; white: ECAD . ( C ) Bright field images of stalk organoids cultured in self-renewing medium at P0 , P1 , P5 . ( D ) Confocal images of stalk organoids at P3 . Green: SOX9; red: SOX2; white: ECAD . ( E ) Multidimensional scaling plot showing the distribution of fresh tip and stalk transcriptomes and cultured organoids . ( F ) Heat map of selected tip , stalk and differentiation markers . Bars = 1 mm ( A , C ) ; 50 μm ( B , D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01510 . 7554/eLife . 26575 . 016Figure 3—figure supplement 1 . Organoids passaged in self-renewing medium retain tip-specific transcription factor proteins . ( A–F ) Cryosections of self-renewing tip organoids between passage 5 and 7 . ( A ) Green: SOX9; red: SOX2; white: ECAD ( epithelial cells ) . ( B ) Green: HMGA2; red: SOX2; white: ECAD ( epithelial cells ) . ( C ) Green: NKX2-1; red: SOX2; white: ECAD ( epithelial cells ) . ( D ) Green: ETV5; red: SOX2; white: ECAD ( epithelial cells ) . ( E ) Green: HNF1B; red: SOX2; white: ECAD ( epithelial cells ) . ( F ) Green: ID2; red: HMGA1; white: ECAD ( epithelial cells ) . Bar = 50 μm all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01610 . 7554/eLife . 26575 . 017Figure 3—figure supplement 2 . Organoids passaged in self-renewing medium retain a normal karyotype . ( A–G ) Karypotype of 7 organoid lines between passage 4 and 8 . 7/7 organoid lines tested retained a normal karypotype . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01710 . 7554/eLife . 26575 . 018Figure 3—figure supplement 3 . SOX2- , SOX9+ canalicular stage human embryonic tips can be grown as SOX2+ , SOX9+self-renewing organoids . ( A–D ) Bright field images of 19 pcw tips growing as self-renewing organoids . ( E ) 19 pcw tip organoids co-express SOX2 and SOX9 . Green: SOX9; red: SOX2; white: ECAD . Blue: DAPI . Scale bars = 2 mm ( A–D ) ; 100 μm ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01810 . 7554/eLife . 26575 . 019Figure 3—figure supplement 4 . Box plots of selected tip and stalk specific genes showing transcript levels in fresh tissue and cultured organoids . Normalised RNAseq data was used to generate box plots of selected genes . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 01910 . 7554/eLife . 26575 . 020Figure 3—figure supplement 5 . E12 . 5 mouse tips do not long-term self-renew in the growth medium developed for human tips . ( A , C ) Bright-field images of mouse E12 . 5 tip epithelium growing as organoids in self-renewal medium +/- SB431542 . ( Factors adjusted to be mouse-specific ) . ( B , D ) Confocal images of mouse organoids . These were initially SOX9+ , SOX2- similar to the in vivo tissue , but they decreased SOX9 expression and turned on SOX2 over time . Green: SOX9; red: SOX2; white: ECAD . Blue: DAPI . Scale bars = 2 mm ( A , C ) ; 100 μm ( B , D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02010 . 7554/eLife . 26575 . 021Video 1 . Human lung epithelial tip growing into an organoid over 11 days . Imaged every 12 hr in bright-field on a Nikon Biostation . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02110 . 7554/eLife . 26575 . 022Video 2 . Organoid structure is a single layer epithelium with a hollow lumen . Confocal z-stack of P6 tip organoid . Nuclei ( DAPI , blue ) and epithelial cells ( ECAD , white ) illustrating typical organoid morphology . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 022 We have recently shown that differentiating mouse bronchiolar cells can produce alveolar descendants ( Laresgoiti et al . , 2016 ) . We therefore tested if human stalk cells , which are undergoing the first signs of bronchiolar differentiation ( Figure 1L; 2K ) , could be induced to grow as tip-like organoids using our tip culture conditions . The human stalks formed organoids with a 100% colony forming efficiency ( n = 17 stalks from five individuals ) . These grew spherically , branched with a similar morphology to tip organoids and could be passaged over many months ( Figure 3C ) . Consistent with their branching behaviour , by the end of P0 the SOX2+ stalks had become SOX2+ , SOX9+ organoids and they retained expression of both proteins over multiple passages ( Figure 3D ) . Similarly , we asked if the SOX9+ , SOX2- tip cells from 19 pcw lungs ( Figure 2E ) could be cultured as self-renewing organoids in the same conditions . The 19 pcw tips also grew as branching organoids for multiple passages with a 100% colony forming efficiency ( organoids from three individuals maintained for at least six passages ) . The 19 pcw tips also became SOX2+ , SOX9+ organoids in vitro ( Figure 3—figure supplement 3 ) . Thus the tip organoid self-renewal conditions are sufficient to convert SOX2+ , SOX9- stalks and SOX2- , SOX9+ tips into long-term self-renewing SOX2+ , SOX9+ organoids . As a first test of our in vitro self-renewing organoid system , we compared the transcriptome of organoids to that of freshly-isolated tips and stalks . When the whole transcriptome is compared freshly isolated tips and stalks can be separated into two distinct clusters on a multi-dimensional scaling plot ( Figure 3E ) , although it should be remembered that they are very similar cell types ( Figure 2G ) . The organoid transcriptomes lie between the freshly isolated tips and stalks , showing that they retain many transcriptional features of the cells from which they are derived ( Figure 3E ) . Analysis of markers that we have characterized as tip or stalk enriched supports this conclusion ( Figure 3F , Figure 3—figure supplement 4 ) . Thus the organoids are transcriptionally very similar to the starting progenitor population , with a tendency to also express stalk markers . We could detect no signs of organoid differentiation into mature lung cell lineages , or towards non-lung cell types . We next asked if our human tip organoid growth conditions were sufficient to support the growth of mouse tips as long-term self-renewing organoids . We plated E12 . 5 mouse tips in identical growth conditions , and also without SB431542 . In both media conditions mouse tips formed branching organoids with 100% efficiency ( Figure 3—figure supplement 5 ) . Mouse organoids grew at a faster rate than the human organoids and were initially SOX9+ , SOX2- , reflecting the in vivo mouse situation . However , by passage six the mouse tip organoids had decreased SOX9 and started to express SOX2 ( Figure 3—figure supplement 5B , D ) . We concluded that the human growth conditions are not sufficient to support the long-term self-renewal of undifferentiated mouse tip cells , consistent with our data that there are differences in signalling gene expression between mouse and human tip epithelium . During organoid establishment diffuse mesenchymal cells were always visible in the cultures and more prevalent in the stalk-derived organoids . We estimated the proportion of mesenchymal cells in freshly microdissected tips and stalks ( Figure 4A , B ) . There was 0 . 04 ± 0 . 07% mesenchyme in the tip dissections and 3 . 8 ± 0 . 47% in the stalk ( n = 4 samples; mean ± standard deviation ) . This estimate is in agreement with the observation that mesenchymal genes were present in the RNAseq of microdissected stalk ( Figure 2J , L ) . We therefore cannot exclude that the presence of the mesenchyme plays a role in the establishment of organoid cultures . However , when we looked for mesenchyme in our passaged human tip organoids , we were able to observe ECAD- cells at P0 , P1 and , rarely , at P2 , but never in higher passage number organoids ( Figure 4C ) . Thus , mesenchymal cells are not required for organoid maintenance . 10 . 7554/eLife . 26575 . 023Figure 4 . All factors added to the medium are required for culture establishment . ( A ) Whole-mount staining for DAPI and ECAD was performed to estimate the fraction of mesenchyme in a microdissected tip ( dotted area ) and stalk ( dashed box ) . ( B ) Quantitation of percentage of mesenchyme in four microdissected tips and stalks . ( C ) Presence of mesenchyme was assessed in six organoid lines over multiple passages by staining for ECAD ( arrowhead = mesenchymal cells ) . ( D ) Bright field and confocal images of P0 Day 13 organoids cultured in self-renewing medium , or without the indicated factors . Green: SOX9; red: SOX2; white: ECAD . ( E ) . Bright field and confocal images of organoids at P1 Day 6 cultured in self-renewing medium , or without FGF10 , or Noggin . Green: SOX9; red: SOX2; white: ECAD . Boxed area is magnified in inset . ( F ) Quantitation of organoid forming efficiency and size with , or without , TGFβ inhibition . Bars = SEM . Three biological replicates were analysed; 61 tip cultures without TGFβ inhibition , 38 tip cultures with TGFβ inhibition . ( G ) Established organoid lines were grown for 3 days in self-renewing , or indicated test medium , and levels of SOX2 and SOX9 assessed by qRT-PCR ( values normalized to one for self-renewing controls ) . Three independent tip organoid lines at P9 , 15 and 21 were used . Bars = SEM . * = p-value<0 . 05 . Bars = 50 μm ( A , C; D , E confocal images ) ; 1 mm ( D , E bright field ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02310 . 7554/eLife . 26575 . 024Figure 4—source data 1 . Individual data points for Figure 4B . Percentage of cells scored as mesenchyme in freshly dissected tips and stalks . Expressed as percentage of non-ECAD+ cells from total DAPI+ cells . Quantitation done using the mesenchyme-macro . ijm for FIJI which is available as a supplemental file . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02410 . 7554/eLife . 26575 . 025Figure 4—source data 2 . Individual data points for Figure 4F . Percentage organoid forming efficiency and size increase from D1-11 of culture when fresh tips are plated in self-renewing medium +/- SB43125 . Three independent experiments were performed on different lung samples . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02510 . 7554/eLife . 26575 . 026Figure 4—source data 3 . Raw qRT-PCR data for Figure 4G . Fold-change was normalised to one for self-renewing medium . Data for three individual organoid lines are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 026 We focussed on human tip epithelium to determine whether all seven factors ( EGF , FGF7 , FGF10 , NOG , RSPO1 , CHIR99021 and SB431542 ) were required for organoid establishment . We plated tips from the same lung in our self-renewing conditions , or in eight other media combinations in which specific factors were removed . To our surprise , SOX2+ , SOX9+ organoids grew in all conditions tested ( Figure 4D; n = 4 experiments with different individual lungs ) . However , compared to our self-renewing condition , all other media produced smaller organoids . These organoids could mostly be passaged ( except in the absence of CHIR99021 where organoids disintegrated at passaging ) , but they remained extremely small at P1 and subsequent passaging was not attempted ( Figure 4E ) . Interestingly , in the absence of NOG the mesenchymal cells present in the cultures expanded much more than in self-renewing medium . These mesenchymal cells were initially SOX9+ , as in vivo , but lost SOX9 expression by P1 . Whereas , in the absence of FGF10 we noted that by P1 the organoids had distinct SOX2+ , SOX9+ and SOX2- , SOX9+ domains ( Figure 4E ) , suggesting that regionalization was occurring spontaneously . We repeated organoid derivation in self-renewal medium , versus self-renewal medium without SB431542 , and quantified the effects on organoid-forming efficiency and size . In the presence of SB431542 organoid-forming efficiency was 100% and organoids were larger ( Figure 4F ) . It has recently been noted that growth of human organoids usually requires TGFβ inhibition , whereas mouse organoids usually do not ( Huch and Koo , 2015 ) . We grew established organoid lines for three days in media in which one or more of the factors was altered to test the effects on SOX2 and SOX9 levels by qRT-PCR ( Figure 4G; n = 3 experiments on three different tip organoid lines ) . None of the conditions resulted in any effect on organoid morphology over the 3 days . Growth in basal medium caused a significant reduction in SOX9 and a less-reproducible increase in SOX2; consistent with a major function of the factor combination to promote tip , and inhibit bronchiolar , fate . Removal of SB431542 , or FGF7 and FGF10 , or CHIR99021 caused a significant reduction in SOX9 levels . Moreover , removal of FGF7 and FGF10 , or CHIR99021 , increased SOX2 . Removal of FGF7 , or 10 , alone had no significant effects . These data suggest that FGF and WNT signalling are both required to promote tip self-renewal at the expense of differentiation , consistent with known functions in mouse ( Volckaert et al . , 2013 ) . To functionally test if the tip organoids retained their lung identity after long-term culture , we tested if they were capable of integrating into adult mouse lungs in vivo . We injured immune-compromised NOD-scid-IL2rg-/- ( NSG ) mice with a low dose of bleomycin and intra-tracheally transplanted 6 × 105 single cells isolated from self-renewing organoids ( Figure 5—figure supplement 1A , B ) . At day two post-transplant small groups of cells were visible in 100% ( 4/4 ) of transplanted mice and these had grown into much larger patches by day 8 ( Figure 5—figure supplement 1C–E ) . Grafts were usually observed in the region spanning the bronchioles and alveoli . Human cells retained SOX2 and SOX9 co-expression ( Figure 5—figure supplement 1G ) . However , they frequently turned off NKX2-1 but retained FOXA2 ( Figure 5—figure supplement 1F ) . They showed the first signs of airway differentiation with sub-sets of graft cells expressing KRT5 , TRP63 and MUC5AC ( Figure 5—figure supplement 1H , I ) . These data strongly suggest that lung identity is retained within the organoids . However , it is not known if embryonic human cells engrafted into adult mouse lungs can receive appropriate differentiation cues . We therefore performed a similar set of experiments using the kidney capsule environment for chimeric human-mouse embryonic grafts . E13 . 5 mouse lungs were dissociated , mixed with dissociated human tip organoids , formed into a cell pellet and transplanted under the kidney capsule of NSG mice ( Figure 5A ) . All kidneys harvested contained mouse/human chimeric lung grafts ( 9/9 ) in which the mouse and human cells tended to segregate ( Figure 5B ) . The mouse parts of every graft contained either squamous epithelium with differentiated alveolar cells ( Figure 5B; Supplement 2A , B ) , or columnar epithelium with differentiated bronchiolar cells ( Figure 5—figure supplement 2C–E ) . By contrast , human cells were always assembled into columnar epithelial airway-like structures surrounded by mouse mesenchyme . Rare human pro-SFTPC+ cells were found in all samples ( Figure 5C; Figure 5—figure supplement 2F ) . Human cells were clearly SOX9+ , SOX2+ at 3 weeks , although patches of cells down-regulating SOX9 were visible ( Figure 5D ) . By 7 and 12 weeks SOX9 was expressed at very low levels in only some human cells . Rare patches of differentiated human airway were identified from 2/3 organoid lines in kidneys harvested at 12 weeks . These were lined with basal , goblet and ciliated cells similar to the in vivo human airways ( Figure 5E , F ) . The remaining human airway-like structures contained goblet cells only and were found in every graft . 10 . 7554/eLife . 26575 . 027Figure 5 . Disscoiated self-renewing organoids are competent to differentiate under the kidney capsule in the presence of embryonic mouse lung cells . ( A ) Experimental schematic . Dissociated organoids were mixed with dissociated E13 . 5 mouse lungs ( either MF1 outbred strain for grafts harvested at 3 and 7 weeks , or Rosa26R-mT-mG strain for grafts harvested at 12 weeks . ) Animals culled at 12 weeks received 3x dexamethasone injections 1 week before culling . The 7 and 12 week grafts looked identical and thus no effects on the grafts were observed from the dexamethasone injections . Three independent organoid lines were used and three mice ( one per organoid line ) culled at each time point . Grafts were clearly visible growing beneath the kidney capsule in all nine kidneys harvested . ( B ) Chimeric mouse/human lung structures were found in all kidneys . Green: ECAD ( epithelium ) ; red: HuNu ( human nuclei ) . ( B’ ) It can clearly be seen that mouse and human cells tend to segregate within the grafts , possibly due to their differing size or surface properties . Wide-spread regions of mouse pro-SFTPC+ cells were always visible . Green: ECAD; red: HuNu; white: pro-SFTPC . ( C ) Rare human pro-SFTPC+ cells ( arrow ) were identified in all samples . Green: ECAD; red: HuNu; white: pro-SFTPC . ( D ) Human cells arranged into airway-like structures were strongly SOX9+ , SOX2+ at 3 weeks , although patches of cells which were down-regulating SOX9 were visible ( bracket ) . By 7 and 12 weeks SOX9 was expressed at very low levels in some human cells . Green: SOX9; red: HuNu; white: SOX2 . ( E , F ) Rare patches of differentiated human airway cells were identified in 2/3 organoid lines in kidneys harvested at 12 weeks . These were lined with basal , goblet and ciliated cells similar to the in vivo human airways . The remaining human airway-like structures contained goblet cells only and were found in every graft harvested ( e . g . arrows in E and F ) . ( E ) Green: KRT5 ( basal cells ) ; red: Td-Tomato ( mouse cells ) ; white: MUC5AC ( goblet cells ) . ( F ) Green: KRT5 ( basal cells ) ; red: Td-Tomato ( mouse cells ) ; white: ACT ( cilia ) . Bars = 50 μm ( C , D , E’ , F’ ) ; 2 mm ( A ) ; 0 . 5 mm ( B ) ; 200 μm ( B’ ) ; 100 μm ( E , F ) ; 20 μm ( E’’ , F’’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02710 . 7554/eLife . 26575 . 028Figure 5—figure supplement 1 . Cells isolated from self-renewing organoids are competent to engraft into adult mouse lungs . ( A ) Oropharyngeal administration of 1 mg/ml bleomycin at a dose of 1 μl per gram body weight is sufficient to injure adult male NSG mouse lungs . Representative haematoxylin and eosin stained lung sections . B . Experimental schematic: 6 × 105 cells were administered 2 days post-bleomycin . C . Grafts were only detected in animals that received both injury and cells . Bars = SEM . D . Grafted human cells ( arrow or bracket ) identified at day 2 and day 8 post-cell administration . Green: ECAD ( epithelial cells ) ; red: HuNu ( Human nuclear marker ) . E . Graph showing number of cells per graft counted at day 2 and day 8 . F-I . Sections of day eight grafts . F . Green: NKX2-1; red: HuNu ( Human nuclear marker ) ; white: FOXA2 . Arrow = NKX2-1+ grafted cell; although the majority of the graft is NKX2-1- , FOXA2+ . G . Green: SOX9; red: HuNu ( Human nuclear marker ) ; white: SOX2 . H . Green: MUC5AC ( goblet cells ) ; red: TRP63 ( basal cells ) ; white: ECAD ( epithelial cells ) . I . Green: KRT5 ( basal cells ) ; red: HuNu ( Human nuclear marker ) ; white: ECAD ( epithelial cells ) . Arrows = KRT5+ graft cells . Bars = 1 mm ( A ) ; 50 μm ( D , F–I ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 02810 . 7554/eLife . 26575 . 029Figure 5—figure supplement 2 . Mouse regions of chimeric human-mouse kidney capsule grafts differentiate efficiently . The extent of differentiation of the mouse cells was indistinguishable at 3 , 7 and 12 weeks post-grafting and a selection of representative images from the various times are shown . Regions of mouse alveoli are found throughout the grafts illustrated by pro-SFTPC staining ( A ) and HOPX staining ( B ) . Mouse airway epithelial cells can be KRT5+ , MUC5AC+ ( C ) , SCGB1A1+ ( D ) and ACT+ ( E ) . Human airway epithelial cells very occasionally have a low level of SCGB1A1 staining ( arrow in D ) . F . Very rare human cells express pro-SFTPC . These are always at the periphery of a tubular structure ( arrow ) . Green: ECAD ( epithelium ) ; red: HuNu ( Human Nuclei ) ; white: pro-SFTPC . HOPX+ human cells were not identified . Bars = 100 μm ( A , B , D ) ; 50 μm ( C , E , F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 029 These grafting experiments show that the organoids retain the ability to assemble into lung structures and robustly generate differentiated airway cells , but alveolar differentiation is either premature or ineffective . However , one limitation of these experiments is that the human cells may be unable to respond efficiently to the mouse differentiation signals; possibly because they require different signalling inputs . We therefore asked if it is possible to differentiate the human tip organoid lines to bronchiolar and alveolar fate in vitro . To test the ability of our tip organoids to differentiate into bronchiolar structures we grew them for 2–4 weeks in an established human airway differentiation medium , PneumaCult . In high passage number organoids this mostly resulted in differentiation into SOX2+ , MUC5AC+ goblet cells , although a smaller number of organoids also contained KRT5+ basal cells ( Figure 6—figure supplement 1; n = 3 organoid lines ) . In low passage number organoids , in which a small amount of mesenchyme was still present , growth in PneumaCultTM resulted in expansion of the mesenchyme and differentiation of TP63+ basal cells , MUC5AC+ goblet cells and rare ACT+ ciliated cells in the epithelium ( Figure 6A–C; n = 3 , P2 organoid lines ) . SOX2 , SOX9 co-expression was observed in the latter experiments , suggesting that airway differentiation was not complete . 10 . 7554/eLife . 26575 . 030Figure 6 . In vitro differentiation of self-renewing organoids towards bronchiolar and alveolar lineages . ( A ) Experimental schematic for 3 weeks organoid differentiation in Pneumacult-ALI medium . ( B ) In low passage number organoids , mesenchyme expanded ( arrows ) , basally-located TP63+ basal cells differentiated and rare ACT+ ciliated cells were seen . Dashed arrow = ciliated cell . Arrow heads = basal cells . Arrow = mesenchyme . Green: ACT ( cilia ) ; red: TP63 ( basal cells ) ; white: SOX2 . ( C ) Differentiated organoids were predominantly composed of MUC5AC+ goblet cells , although low levels of SCGB1A1 were observed in some cells . Cells retained SOX9 suggesting that differentiation was not complete . Green: MUC5AC ( mucous ) ; red: SOX9; white: SCGB1A1 . ( D ) Experimental schematic for 3 week alveolar differentiation experiment . ( E ) All organoid cells retained a columnar appearance and expressed relatively low levels of apical pro-SFTPC . Red: pro-SFTPC; white: ECAD . ( F ) Experimental schematic for 3 week alveolar differentiation in the presence of freshly-isolated 19 pcw human mesenchyme . ( G ) Mesenchymal cells were observed in the cultures ( arrowheads ) , moreover cells expressing higher levels of pro-SFTPC with a more squamous appearance were also obtained . Red: pro-SFTPC; white: ECAD . ( H ) Experimental schematic for 3 week alveolar differentiation of 19 pcw organoid in the presence of expanded human 20 pcw mesenchymal cells . DCI = dexamethasone , cAMP , IBMX . ( I ) Organoid epithelium took on a more squamous appearance and was surrounded by mesenchymal cells ( arrowheads ) . White: ECAD . ( J ) AT2 markers were expressed robustly . Green: HTII-280; red: pro-SFTPC . ( K ) AT1 and AT2 markers were co-expressed . Green: HOPX; red: pro-SFTPC . Blue: DAPI . Bars = 1 mm ( A , H ) ; 100 μm ( B , C ) ; 20 μm ( B’ , E’ ) ; 10 μm ( C’ , G’ ) ; 50 μm ( E , G , I , J , K ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 03010 . 7554/eLife . 26575 . 031Figure 6—figure supplement 1 . Exposure of self-renewing organoids to PneumacultTM medium leads to efficient goblet cell differentiation with rare patches of KRT5+ basal cells . ( A ) Experimental schematic . ( B ) Every organoid examined had many MUC5AC+ goblet cells and little or no SOX9 . Green: MUC5AC ( mucous ) ; red: SOX9 . ( C ) More rarely , some organoids contained patches of KRT5+ basal cells . Green: MUC5AC ( mucous ) ; red: KRT5 ( basal cells ) . Blue: DAPI . Bar = 50 μm ( B , C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 03110 . 7554/eLife . 26575 . 032Figure 6—figure supplement 2 . Testing media conditions for ability to promote human organoid alveolar differentiation . ( A ) Experimental schematic . ( B ) Alveolar differentiation media tested . DCI = dexamethasone , cAMP , IBMX . ( C-D ) Self-renewing controls co-express SOX2 and SOX9 , whereas SOX2 is retained and SOX9 is lost following 2 weeks in most media tested . Green: SOX9; red: SOX2; white: ECAD . ( F ) Green: HTII-280; red: pro-SFTPC cannot be detected in self-renewing control conditions . ( G–M ) The majority of media tested resulted in patchy loss of SOX2 and upregulation of pro-SFTPC . Green: SOX2; red: pro-SFTPC; white: ECAD . ( N ) HTII-280 expression was very rarely observed in any differentiation condition . Green: HTII-280; red: pro-SFTPC . Scale bars = 1 mm ( A ) ; 50 μm ( C–N ) ; 10 μm ( insets ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 03210 . 7554/eLife . 26575 . 033Figure 6—figure supplement 3 . A combination of canalicular stage lung embryonic mesenchyme and alveolar differentiation medium together promote the most efficient organoid alveolar differentiation . ( A ) Experimental schematic for 3 week alveolar differentiation of psedoglandular stage-derived organoids in the presence of freshly-isolated 19 pcw human mesenchyme . DCI = dexamethasone , cAMP , IBMX . ( B , C ) Regions of organoid that have turned off SOX2 and turned on pro-SFTPC can readily be distinguished . Green: SOX2; red: pro-SFTPC; white: ECAD . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 03310 . 7554/eLife . 26575 . 034Figure 6—figure supplement 4 . Expanded canalicular stage mesenchyme and alveolar medium can promote organoid alveolar differentiation . ( A ) Experimental schematic for 3 week alveolar differentiation of psedoglandular and canalicular stage-derived organoids in the presence of expanded 19 or 20 pcw human mesenchyme . ( B–D ) Pseudoglandular stage organoids took on a squamous appearance and expressed pro-SFTPC and , more rarely , HTII-280 . The AT1 marker , HOPX was also expressed . ( E–I ) Canalicular stage organoids also took on a squamous appearance and expressed higher levels of AT2 markers , in addition to HOPX and PDPN . Cells retained co-expression of AT1 and AT2 markers . Results shown are from two independent 19 pcw organoid experiments are shown . B , C , E , F . Green: HTII-280; red: pro-SFTPC . ( D , G ) Green: HOPX; red: HTII-280 . ( H , I ) Green: NKX2-1; red: PDPN; white: HTII-280 . Blue = DAPI . Bars = 50 μm all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 03410 . 7554/eLife . 26575 . 035Figure 6—figure supplement 5 . Expanded fibroblasts used for organoid co-cultures are a heterogeneous population expressing various lung embryonic fibroblast markers . ( A–D ) 20 pcw lung sections stained for mesenchymal markers . ( A ) Green: PDGFRA; red: PDPN ( alveolar epithelium and lymphatic endothelium ) ; white: SMA . ( B ) Green: PDGFRB; red: PDPN ( alveolar epithelium and lymphatic endothelium ) . ( C ) Green: CD90; red: PDPN ( alveolar epithelium and lymphatic endothelium ) . ( D ) Green: VECAD ( vascular endothelium ) ; red: PDPN ( alveolar epithelium and lymphatic endothelium ) . ( E–F ) Passage three fibroblasts expanded from 19 pcw human embryonic lung . ( E ) Green: PDGFRA; red: SMA . ( F ) . Green: PDGFRB; red: CD90 . Expanded fibroblasts are all PDGFRB+ , but co-express a variety of other markers . All panels: blue = DAPI ( nuclei ) . Bars = 50 μm all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 035 There is no established culture medium for growing human alveolar epithelium . We therefore tested media conditions for their ability of turn off SOX2 and SOX9 and activate pro-SFTPC and/or HTII-280 , compared to our self-renewing condition , over two weeks ( Figure 6—figure supplement 2 ) . SOX9 was lost in all conditions tested , possibly due to prolonged exposure to Dexamethasone which turns off Sox9 in mouse lungs ( Figure 6—figure supplement 2C–E ) ( Alanis et al . , 2014 ) . In general , we obtained somewhat patchy differentiation with regions where SOX2 was retained and regions where pro-SFTPC ( or more rarely HTII-280 ) was expressed ( Figure 6—figure supplement 2F–N ) . We next cultured organoids for 3 weeks in the most efficient alveolar medium ( CHIR99021 , FGF7 , FGF10 , Dexamethosone , cAMP , IBMX , T3 , DAPT ) . This resulted in uniformly SOX2- , SOX9- , pro-SFTPC+ organoids ( Figure 6E , F; n = 2 organoid lines ) . These expressed low levels of pro-SFTPC and retained a columnar epithelial appearance , suggesting alveolar-fate rather than differentiation . HTII-280 expression was not detected . We reasoned that addition of human canalicular stage mesenchyme would provide additional cues and promote improved alveolar differentiation . We isolated mesenchyme from 19 pcw human lungs by dispase digestion and micro-dissection , mixed it directly with human tip organoids and cultured in alveolar differentiation medium for three weeks . This resulted in heterogeneous organoids with patches of SOX2+ cells and patches of pro-SFTPC+ cells ( Figure 6G , H; Figure 6—figure supplement 3A–C; n = 1 line for this preliminary experiment ) . Importantly , pro-SFTPC expression was stronger than without mesenchyme and the pro-SFTPC+ cells had a more squamous appearance , similar to endogenous differentiating AT2 cells . However , there were many regions of the cultures where mesenchymal cells were not observed . To increase the proportion of mesenchyme in the co-cultures , we expanded fibroblasts from 19 or 20 pcw lungs and cultured with tip organoids . This resulted in more highly branched organoids with squamous epithelium surrounded by mesenchymal cells , reminiscent of the canalicular stage lung ( Figure 6I , J ) . These organoids were SOX2- , SOX9- and pro-SFTPC+ , HOPX+; rare HTII-280+ cells were also observed ( Figure 6—figure supplement 4B–D; n = 3 organoid lines ) , consistent with alveolar fate; most likely bipotential alveolar progenitors . When tip organoids derived from 19 pcw lungs were used there were extensive regions of pro-SFTPC , HTII-280 , HOPX and PDPN co-expression and NKX2-1 was retained . This again indicates that we have successfully differentiated to the bipotent progenitor stage and also suggests that the organoids derived from the canalicular stage lungs are intrinsically easier to differentiate towards alveolar fate ( Figure 6H–K; Figure 6—figure supplemental 4 E–I E-I; n = 2 , 19 pcw organoid lines ) . Antibody staining showed that the expanded fibroblasts used in these experiments were PDGFRB+ , but an otherwise heterogeneous mixture of cells expressing markers consistent with the mesenchyme observed in 20 pcw lung sections ( Figure 6—figure supplement 5 ) .
It is routinely assumed that the pseudoglandular stage human embryonic lung distal tip epithelium is a multipotent progenitor population . We now provide several lines of evidence that this hypothesis is correct . Firstly , we show that typical differentiation markers are detectable at a protein level only after cells exit the tip . Secondly , that 96% of human tip genes with mouse orthologues are also expressed in mouse tips . Thirdly , we demonstrate that the human tips have the ability to long-term self-renew and differentiate into bronchiolar and alveolar epithelium in vitro . Moreover , we have established conditions to grow human embryonic lung epithelial tips as long-term self-renewing organoids . These organoids retain many transcriptional similarities to tip cells and show no signs of spontaneous differentiation . They can also integrate into adult and embryonic lungs and be induced to differentiate towards bronchiolar or alveolar lineages in vitro . The organoid culture conditions that we have established provide a new tool for in vitro genetic studies of human lung development . Although we find an extremely high level of transcriptome conservation between human and mouse tips we also report multiple differences , including in genes used as definitive cell-type specific markers . The most striking difference is in SOX2 . Mouse tip progenitors are SOX2- , SOX9+ throughout development . By contrast , human pseudoglandular tips , which are producing bronchiolar descendants , are SOX2+ , SOX9+ and tips become SOX2- , SOX9+ during the canalicular stage . The functional significance of tip SOX2 expression is currently unknown , although it may reflect lineage-priming . However , this observation is highly relevant to attempts to establish human pluripotent stem cell ( PSC ) differentiation protocols for lung epithelium . Based on mouse data , such human studies typically focus on the derivation of multipotent lung tip progenitors as NKX2-1+ , SOX2- , SOX9+ , ID2+ cells . These are likely to be human canalicular tip progenitors . Differentiation marker expression in human PSC differentiation experiments is also based on mouse data . More robust , efficient protocols for human PSC differentiation are likely to be developed if human lung development is taken into account . We have observed that mouse tip progenitors are unable to long-term self-renew in human culture conditions , even though mouse-specific versions of the factors were provided . Hence , the tip transcriptome is highly conserved between mouse and human , but the differences are of functional significance . There is therefore a strong imperative to study human lung development alongside more traditional mouse studies . We have used organoid culture to show that differentiating human lung stalks can be reprogrammed to tip fate analogous to classical rodent studies ( Alescio and Cassini , 1962 ) ; FGF and WNT are required to maintain human tip SOX2 and SOX9; and activation of EGF , FGF and WNT , with inhibition of BMP and TGFβ , signalling is sufficient to maintain human tip self-renewal . Moreover , our long-term self-renewing organoids are competent to differentiate , although in vitro differentiation requires further optimisation . We have also introduced plasmids to the self-renewing organoids by electroporation ( Fujii et al . , 2015 ) and been able to freeze-thaw organoids for longer-term storage . Therefore , a genetic system for the study of human lung development is now available . We propose that this will also be useful for disease modelling and informing the differentiation of human PSCs .
Human embryonic and foetal lungs were obtained from terminations of pregnancy from Cambridge University Hospitals NHS Foundation Trust under permission from NHS Research Ethical Committee ( 96/085 ) and the Joint MRC/Wellcome Trust Human Developmental Biology Resource ( London and Newcastle , grant 099175/Z/12/Z , www . hdbr . org ) . Their age ranged from 5 to 20 weeks developmental age , also known as post-conception weeks , pcw ( this corresponds to 7–22 weeks gestational age ) . Samples were staged according to their external physical appearance and measurements , and not to the estimated last menstrual period . Detailed guidelines for embryonic samples ( <8 pcw ) : http://hdbr . org/downloads/embryo_staging_guidelines . doc; and for foetal samples: http://hdbr . org/downloads/fetal_staging_guidlines . doc . Samples used had no known genetic abnormalities . Fresh healthy adult lung tissue ( background tissue from lobectomies for lung cancer ) was obtained from Papworth Hospital NHS Foundation Trust ( Research Tissue Bank Generic REC approval , Tissue Bank Project number T01939 ) and processed for both cryo- and paraffin sectioning . All experiments were approved by local ethical review committees and conducted according to Home Office project licenses PPL 70/8012 ( Emma Rawlins , University of Cambridge ) and 70/7607 ( Adam Giangreco , UCL ) . Mouse strains ( Rosa26R-mT/mG , formally known as Gt ( ROSA ) 26Sortm4 ( ACTB-tdTomato , -EGFP ) Luo/J; RRID:IMSR_JAX:026862 ) ( Muzumdar et al . , 2007 ) and NOD-scid-IL2rg-/- ( NSG; RRID:IMSR_JAX:005557 ) ( Ishikawa et al . , 2005; Shultz et al . , 2005 ) have been described . Wild-type mice were outbred MF1 strain . Lung lobes were incubated for 2 min in Dispase ( ThermoFisher Scientific , Gibco , UK , 8 U/ml ) at room temperature . Mesenchyme was dissected away using tungsten needles and tips and stalks were isolated by cutting the very end of a branching tip , or alternatively a stalk area more proximally . Five tips , or 2 pieces of stalk tissue , were transferred into 30 μl Matrigel ( Corning , UK , 356231 ) . A dissecting microscope was used to guide aspiration of 25 μl Matrigel containing the tissue pieces , which was transferred into a well of a 48 well low-attachment plate ( Greiner , UK ) . The plate was incubated for 5 min at 37°C to solidify the Matrigel , following which at least 250 μl of self-renewing was added ( Table 1 , or Table 2 ) . Plates were incubated under standard tissue culture conditions ( 37°C , 5% CO2 ) . 10 . 7554/eLife . 26575 . 036Table 1 . Self-renewal ( Human ) DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 036ReagentCompanyCat noFinal concentrationAdvanced DMEMThermoFisher Scientific , Invitrogen12634–010Base mediumPenicillin/StreptomycinThermoFisher Scientific , Invitrogen15140–122100 U/ml ( Pen ) 100 μg/ml ( Strep ) HepesThermoFisher Scientific , Invitrogen15630–05610 mMGlutamaxThermoFisher Scientific , Invitrogen35050–0382 mMN2ThermoFisher Scientific , Invitrogen17502–0481:100B27 ( -Vit A ) ThermoFisher Scientific , Invitrogen12587–0101:50N-acetylcysteineSigma-AldrichA91651 . 25 mMMatrigel ( growth factor reduced; specific lots of matrigel with at least 8 mg/ml protein concentration were used ) Corning356231undilutedR-spondin1 conditioned mediumStem Cell Intitute , University of CambridgeFrom 293T-HA-Rspo1-Fc cell line made by Calvin Kuo , Stanford5% v/vEGFPeprotech , UKAF-100–1550 ng/mlNogginR and D Systems6057 NG-100100 ng/mlFGF10R and D Systems345-FG-025100 ng/mlFGF7Peprotech100–19100 ng/mlCHIR 99021Stem Cell Institute , University of Cambridgen/a3 μMSB 431542Tocris161410 μM48 well plates ( Greiner Cellstar ) Sigma-AldrichM9437n/a10 . 7554/eLife . 26575 . 037Table 2 . Self-renewal ( Mouse ) DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 037ReagentCompanyCat noFinal concentrationAdvanced DMEMThermoFisher Scientific , Invitrogen12634–010n/aPenicillin/StreptomycinThermoFisher Scientific , Invitrogen15140–122100 U/ml ( Pen ) 100 μg/ml ( Strep ) HepesThermoFisher Scientific , Invitrogen15630–05610 mMGlutamaxThermoFisher Scientific , Invitrogen35050–0382 mMN2ThermoFisher Scientific , Invitrogen17502–0481:100B27 ( -Vit A ) ThermoFisher Scientific , Invitrogen12587–0101:50N-acetylcysteineSigma-AldrichA91651 . 25 mMMatrigelCorning356231UndilutedR-spondin conditioned mediumStem Cell Institute , University of CambridgeFrom 293T-HA-Rspo1-Fc cell line made by Calvin Kuo , Stanford5% v/vmEGFR and D Systems2028-EG-20050 ng/mlNogginR and D Systems6057 NG-100100 ng/mlFGF10R and D Systems345-FG-025100 ng/mlmFGF7R and D Systems5028 KG_025100 ng/mlCHIR 99021Stem Cell Institute , University of Cambridgen/a3 μMSB 431542Tocris161410 μM 92 mm plates were coated with type I collagen ( Sigma , UK , C3867-1VL ) mixed with 0 . 02 N acetic acid ( 1:72 . 5 ) using a total volume of 6 ml , then left to evaporate in a tissue culture hood for about 4 hr . Fresh human foetal lung was cut into small pieces and incubated at 37°C , 30 min in 24 U/ml Dispase ( Gibco ) , 10 μg/ml DNase in PBS . DMEM/F12 with 10% ( v/v ) FBS ( ThermoFisher Scientific , Life Technologies ) was added and lung pieces spun 200 g , 5 min . The supernatant was aspirated and the pellet was resuspended in DMEM/F12 with 10% ( v/v ) FBS and 1:100 Penicillin/Streptomycin ( ThermoFisher Scientific , Life Technologies ) . The lung pieces and culture medium were transferred evenly onto the collagen coated plate . The plate was incubated for 5 days without medium change . On day 5 , the lung pieces were removed and fresh medium added . Medium change was twice a week until confluence . Cells were split using 0 . 1% ( w/v ) trypsin for 2 min at 37°C , inactivated with DMEM/F12 with 10% ( v/v ) FBS , centrifuged and then plated on 92 mm plates . For antibody staining fibroblasts were passaged onto collagen-coated coverslips . ( Consumable details , Table 3 ) 10 . 7554/eLife . 26575 . 038Table 3 . Human foetal lung mesenchymeDOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 038ReagentCompanyCat noFinal concentrationDish Nunc T/C 92 mmThermoFisher Scientific10508921n/aType I collagenSigma-AldrichC3867-1VL55 μg/mlAcetic acidThermoFisher Scientific103049800 . 02 NDMEM/F12 with L-glutamineThermoFisher Scientific , Invitrogen11320–074n/aDispaseThermoFisher Scientific , Invitrogen1710504124 U/mlFBSSigma-AldrichF966510%Penicillin/StreptomycinThermoFisher Scientific , Invitrogen15140–122100 U/ml ( Pen ) 100 μg/ml ( Strep ) DNase IQiagen , UK7925410 μg/mlTrypsin ( from porcine pancreas ) Sigma-AldrichT47990 . 1% Organoids were cultured in Matrigel ( Corning , 356231 ) in 48-well plates with self-renewing medium ( Advanced DMEM/F12 supplemented with 1x GlutaMax , 1 mM Hepes and Penicillin/ Streptomycin ( P/S ) , 1:50 B27 supplement ( without Vitamin A ) , 1:100 N2 supplement , 1 . 25 mM n-Acetylcysteine , 5% ( v/v ) R-spondin1 conditioned medium , 50 ng/ml recombinant human EGF , 100 ng/ml recombinant human Noggin , 100 ng/ml recombinant human FGF10 , 100 ng/ml recombinant human FGF7 , 3 μM CHIR99021 and 10 μM SB431542 ( Table 1 ) . For mouse cultures , mouse specific EGF and FGF7 were used ( Table 2 ) . Medium was changed twice a week , and organoids were passaged every 10–14 days depending on cell confluence and Matrigel stability . Plates were incubated under standard tissue culture conditions ( 37°C , 5% CO2 ) . Unless otherwise stated organoid lines were used between passage 4 and passage 16 for experiments . Organoids were usually split 1:4 to 1:6 after 10–14 days of culture . The medium was aspirated and fresh cold base medium: Advanced DMEM with Glutamax , P/S and Hepes ( AdvDMEM+++ ) added to each well . The Matrigel in each well was sucked into a P1000 pipette tip and transferred into a 15 ml tube . Cold AdvDMEM+++ was added up to 10 ml and then the sample was centrifuged at 100 g at 4°C for 5 min . 8 . 5 ml of medium was then aspirated , and the remaining organoids triturated using a flame polished glass pipette . Cold AdvDMEM+++ was added up to 10 ml and the sample was again centrifuged at 220 g at 4°C . The pellet was resuspended in undiluted Matrigel ( in a volume depending on the splitting ratio ) and 25 μl of Matrigel containing the split organoids was plated onto a well of a 48 well low attachment plate . The plate was incubated for 5 min at 37°C to allow the Matrigel to solidify , upon which at least 250 μl culture medium was added per well ( Table 1 or Table 2 ) . For bronchiolar or alveolar differentiation different media were used as outlined in the figures ( Table 4 , Table 5 ) . Organoids , and other primary cells , were tested regularly for mycoplasma . 10 . 7554/eLife . 26575 . 039Table 4 . Human Bronchiolar differentiationDOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 039ReagentCompanyCat noFinal concentrationPneumaCultTM-ALI mediumStem Cell Technologies05001n/aMatrigelCorning/SLS356231undiluted10 . 7554/eLife . 26575 . 040Table 5 . Human Alveolar differentiationDOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 040ReagentCompanyCat noFinal concentrationAdvanced DMEMThermoFisher Scientific , Invitrogen12634–010n/aPenicillin/StreptomycinThermoFisher Scientific , Invitrogen15140–122100 U/ml ( Pen ) 100 μg/ml ( Strep ) HepesThermoFisher Scientific , Invitrogen15630–05610 mMGlutamaxThermoFisher Scientific , Invitrogen35050–0382 mMN2ThermoFisher Scientific , Invitrogen17502–0481:100B27 ( -Vit A ) ThermoFisher Scientific , Invitrogen12587–0101:50N-acetylcysteineSigma-AldrichA91651 . 25 mMMatrigelCorning/SLS356231UndilutedCHIR 99021Stem Cell Institute , University of Cambridgen/a1 μM or 3 μMFGF10R and D345-FG-025100 ng/mlFGF7Peprotech100–19100 ng/mlDexamethasoneSigma-AldrichD4902-25MG50 nMcAMPSigma-AldrichB5386-5MG0 . 1 mMIBMXSigma-AldrichI5879-100MG0 . 1 mMDAPTSigma-AldrichD5942-25MG50 μMTri-iodothyronine ( T3 ) Sigma-AldrichT6397-100MG6 . 7 ng/mlHuman IGF-1R and D Systems291-G1-2001 μg/mlInterleukin-6R and D Systems206-IL-0101 μg/ml Organoid lines could be frozen and thawed without apparent change in behaviour . For freezing , organoids were removed from matrigel and triturated with a flame polished glass pipette as for passaging . They were then pelleted and resuspended in cold freezing medium ( Invitrogen , 12648010 ) at 500 μl per well in a cryovial which was transferred into a pre-cooled Mr . Frosty Freezing Container ( ThermoFisher Scientific , Invitrogen , 5100–0001 ) at −80°C overnight followed by longer-term storage in liquid nitrogen . Cryovials were thawed for 2 min in a 37°C waterbath and organoids plated in matrigel in self-renewing medium supplemented with 10 μl Rho kinase inhibitor ( Y27632 , Sigma-Aldrich , Y0503-1MG ) . When fresh , or expanded , human mesenchyme was added to the cultures , 250 , 000 mesenchymal cells per well were mixed with the organoids immediately prior to the final spin and resuspension in Matrigel . All differentiation experiments were performed in at least three technical replicates . Prior to fixation , or lysis , organoids were removed from Matrigel using Corning Matrigel Cell Recovery Solution ( Corning , 354253 ) . First , organoids were harvested into a 15 ml tube using a wide Pasteur pipette and washed with 10 ml of cold washing medium ( Advanced DMEM/F12 , 1X GlutaMax , 1 mM Hepes and Penicillin/Streptomycin ) . The 15 ml tube was inverted every 2 min for 10 min , followed by 5 min incubation on ice before organoids were spun 200 g at 4°C . This was repeated once and then Corning Cell Recovery Solution ( Corning , 354253 ) was used to further remove the Matrigel ( incubation on ice for 30 min with inversion once after 15 min ) . Organoids were washed with cold PBS , spun down at 200 g 4°C . For 5–9 pcw lungs fixation was overnight hour at 4°C in 4% PFA . Organoids were recovered from the Matrigel using Corning Cell Recovery Solution ( Corning , 354253 ) as above and fixed 4% ( w/v ) paraformaldehyde ( PFA ) for 30 min at 4°C . After washing in PBS organoids were transferred to a round-bottom 96 well plate using wide Pasteur pipettes . Permeabilisation in 0 . 5% ( v/v ) Triton-X in PBS for 30 min was followed by washing in 0 . 5% ( w/v ) Bovine Serum Albumin ( BSA ) , 0 . 2% Triton-X in PBS ( washing solution ) . Blocking was for at least 1 hr at room temperature in 1% BSA , 5% NDS ( normal donkey serum ) , 0 . 2% Triton-X in PBS . Primary antibodies ( Table 6 ) in blocking solution used at 4°C overnight . The following day washes were performed at 4°C and secondary antibodies ( 1:2000 dilution; Table 7 ) in 5% NDS , 0 . 2% Triton-X in PBS incubated overnight at 4°C . The following day washes were performed at 4°C and DAPI ( Sigma ) added to the washing solution for 30 min at 4°C . Samples were processed to 2’−2’-thio-diethanol ( TDE , Sigma , 166782 ) for clearing/mounting: 10% ( v/v ) TDE in 1x PBS; 25%; 50% 1 hr , 97% TDE overnight at 4°C on a rocker . The following day , organoids were transferred onto a slide with an imaging spacer ( diameter 20 mm; thickness 0 . 12 mm; Sigma GBL654006 ) containing 65 μl 97% ( v/v ) TDE and coverslipped . 10 . 7554/eLife . 26575 . 041Table 6 . Primary antibodiesDOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 041AntibodyCompanyCat noHost speciesFinal dilutionAntigen retrieval ( Cryo ) Antigen retrieval ( Paraffin ) Research Resource Identifier ( RRID ) ABCA3Seven Hills Bioreagents , Cincinnati , OHWRAB-ABCA3Rabbit1:500NoYesRRID:AB_577286Acetylated tubulin ( ACT ) SigmaT7451 , clone 6-11B-1Mouse1:3000NoNoRRID:AB_609894AQP5Santa Cruz Biotechnology , Dallas , TexasSc9890 , clone G19Goat1:100YesYesRRID:AB_2059877β3-TUBULINBiolegend ( Covance ) , UKPRB-435PRabbit1:1000NoN/ARRID:AB_291637CD90Novus Biologicals , UKNBP2-37330 ( clone 7E1B11 ) Mouse1:200NoN/ARRID:AB_2665376E-CADHERINThermoFisher Scientific Invitrogen13–1900Rat1:3000NoN/ARRID:AB_2533005E-CADHERINBD Biosciences , UK610182Mouse1:500NoYes ( citrate ) RRID:AB_397581ETV5Santa Cruz BiotechnologySc-22807Rabbit1:200YesN/ARRID:AB_2101008FGFR2Santa Cruz BiotechnologySC-122Rabbit1:200NoN/ARRID:AB_631509FOXA2Santa Cruz BiotechnologySC-6554Goat1:200NoN/ARRID:AB_2262810FOXJ1Thermo Fisher Scientific14-9965-82Mouse1:200YesYes ( citrate ) ; needs streptavidin-biotin amplificationRRID:AB_1548835HMGA1BR and D Systems , UKAF5956Sheep1:50YesN/ARRID:AB_1964602HMGA2Proteintech , UK20795–1-APRabbit1:100NoN/ARRID:AB_2665377HNF1BProteintech12533–1-APRabbit1:100YesN/ARRID:AB_2116758HOPXSanta Cruz BiotechnologySC-30216Rabbit1:50NoYes ( citrate ) RRID:AB_2120833HTI-56Gift from Leland Dobbsn/aMouse1:100NoN/ARRID:AB_2665380HTII-280Gift from Leland Dobbsn/aMouse IgM1:100NoNoRRID:AB_2665381Human Nuclei ( HuNu ) Merck , UKMAB1281Mouse1:3000No ( needs streptavidin-biotin amplification ) N/ARRID:AB_11212527ID2Abcam , UKAb52093Rabbit1:200YesN/ARRID:AB_880731KRT5CovancePRB-160P-100Rabbit1:500NoYes ( citrate ) RRID:AB_291581KI67BD Transduction Laboratories , UK550609 , clone B56Mouse1:100NoYes ( citrate ) RRID:AB_393778LPCAT1Proteintech16112–1-APRabbit1:500NoYes ( citrate ) RRID:AB_2135554MUC5ACThermoFisher ScientificMS-145PMouse1:500NoYes ( citrate ) RRID:AB_62731NKX2-1AbcamAb76013Rabbit1:500YesYes ( citrate ) RRID: AB_1310784PDGFRACell Signalling3174 ( clone D1E1E ) Rabbit1:1000NoN/ARRID:AB_2162345PDGFRBCell Signalling3169 ( clone 28E1 ) Rabbit1:100NoN/ARRID:AB_2162497PDPNProteintech11629–1-APRabbit1:200NoYesRRID:AB_2162067PDPNR and D SystemsAF3670Sheep1:200NoYesRRID:AB_2162070PECAM ( CD31 ) AbcamAb9498Mouse1:200NoN/ARRID:AB_307284SCGB1ASanta Cruz BiotechnologySC-25555Rabbit1:200NoYes ( citrate or trypsin ) RRID:AB_2269914pro-SFTPCMillipore , UKAb3786Rabbit1:500NoYes ( citrate ) RRID:AB_91588SMASigmaA5228 , clone 1A4Mouse1:500NoN/ARRID:AB_262054SOX2Santa Cruz BiotechnologySC-17320Goat1:250No/YesYes ( citrate ) RRID:AB_2286684SOX9Santa Cruz BiotechnologySC-20095Rabbit1:200NoYes ( citrate ) RRID:AB_661282SOX9Abcamab196450Rabbit1:200NoN/ARRID:AB_2665383SPRY2Abcamab50317Rabbit1:200YesN/ARRID:AB_882688TP63Cell Signaling13109Rabbit1:200YesYes ( citrate ) ; needs streptavidin-biotin amplificationRRID:AB_2637091VECADR and D SystemsAF938Goat1:400NoN/ARRID:AB_35572610 . 7554/eLife . 26575 . 042Table 7 . Secondary antibodiesDOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 042AntibodyCompanyCat noFinal dilutionResearch Resource Identifier ( RRID ) Donkey α-mouse 488Thermo Fisher ScientificA212021:2000RRID:AB_141607Donkey α-rabbit 488Thermo Fisher ScientificA212061:2000RRID:AB_2535792Donkey α-rat 488Thermo Fisher ScientificA212081:2000RRID:AB_2535794Donkey α-mouse 546Thermo Fisher ScientificA100361:2000RRID:AB_2534012Donkey α-rabbit 546Thermo Fisher ScientificA100401:2000RRID:AB_2534016Donkey α-goat 555Thermo Fisher ScientificA214321:2000RRID:AB_2535853Donkey α-rat 594Thermo Fisher ScientificA212091:2000RRID:AB_2535795Donkey α-sheep 594Jackson Immunoresearch , West Grove , PA713-585-1471:2000RRID:AB_2340748Donkey α-mouse 647Thermo Fisher ScientificA315711:2000RRID:AB_162542Donkey α-rabbit 647Thermo Fisher ScientificA315731:2000RRID:AB_2536183Goat α-rat 647Thermo Fisher ScientificA212471:2000RRID:AB_141778Streptavidin 594Jackson Immunoresearch016-580-0841 . 8 μg/mlRRID:AB_2337250Biotin-SP-conjugated Donkey α-mouseJackson Immunoresearch715-065-1501:500RRID:AB_2307438 Mouse embryonic lungs were fixed at 4°C in 4% PFA for 1 hr . Fixation was at 4°C in 4% PFA overnight for mouse kidneys and human embryonic , foetal and adult lungs; after nine pcw the lungs were divided into pieces prior to fixation , preferably intact lung lobes , in order to fit in 15x15 × 5 mm moulds . Post-fixation PBS washes and sucrose protection ( 15% , 20% , 30% w/v sucrose in PBS 1 hr each ) were at room temperature . Samples were incubated 1:1 in 30% sucrose: optimal cutting temperature compound ( OCT ) overnight at 4°C , 1 hr room temperature 100% OCT wash for small tissue fragments only , then embedded in OCT . 7 μm sections were cut and stored at −80°C . Tissue was permeabilised using 0 . 3% Triton-X in PBS . Antigen retrieval was by heating slides in 10 mM Na Citrate buffer at pH6 in a full power microwave for 5 min . Blocking at least 1 hr , room temperature in 5% NDS , 1% BSA , 0 . 1% Triton-X in PBS . Primary antibodies ( Table 6 ) were diluted in block and incubated overnight 4°C . After PBS washes , secondary antibodies ( 1:2000; Table 7 ) were added in 5% NDS , 0 . 1% Triton-X in PBS and incubated 2–3 hr at room temperature . When biotin-coupled secondaries were used , sections were incubated in 1 . 8 μg/ml Streptavidin-594 in 1% BSA for 30 min at room temperature . DAPI ( Sigma ) was added for 20 min , followed by PBS washes and mounting in Fluoromount ( Sigma ) . Images were collected on a Zeiss Axiophot microscope , or Leica SP8 confocal where stated . For sections of human lungs , for each antibody at least 3 different 5–8 pcw lungs; 2 different 11 pcw lungs; 2 different 14–15 pcw lungs; 2 different 16–17 pcw lungs; one 19 pcw lung and 2 different 20 pcw lungs were stained . At least two technical replicates were performed for each immunostaining . Human embryonic and adult lungs were cut to fit 15x15 × 5 mm moulds and fixed at 4°C in 4% ( w/v ) PFA overnight . Organoids were removed from the Matrigel following the Corning Cell Recovery Solution protocol ( above ) and fixed for 30 min at 4°C . Following PBS washes , small tissue ( e . g . organoids ) was embedded in 3% ( w/v ) Low Melting Point Agarose ( Sigma , A2790 ) . All samples were dehydrated to 100% ethanol and then processed to paraffin wax for embedding . Paraffin blocks were sectioned at 5 μm and slides dried at 50°C for at least 30 min . Deparaffinisation was performed by 2x xylene washes , followed by rehydration to distilled water then PBS rinse . Antigen retrieval was heating in 10 mM Na Citrate buffer at pH6 , or 0 . 05% ( w/v ) trypsin in PBS as appropriate . Blocking at least 1 hr , room temperature in 5% NDS , 1% BSA , 0 . 1% Triton-X in PBS . Primary antibodies ( Table 6 ) were diluted in block and incubated overnight 4°C . After PBS washes , secondary antibodies ( 1:2000; Table 7 ) were added in 5% NDS , 0 . 1% Triton-X in PBS and incubated 2–3 hr at room temperature . When biotin-coupled secondaries were used , sections were incubated in 1 . 8 μg/ml Streptavidin-594 in 1% BSA for 30 min at room temperature . DAPI ( Sigma ) was added for 20 min , followed by PBS washes and mounting in Fluoromount ( Sigma ) . Images were collected on a Zeiss Axiophot microscope , or Leica SP8 confocal where stated . Haematoxylin and Eosin staining followed standard protocols . Organoids were removed from the Matrigel following the Corning Cell Recovery Solution protocol ( above ) and lysed using 500 μl RLT buffer . Freshly isolated tips and stalks were lysed using 350 μl RLT buffer . RNA extraction was performed according to the RNeasy Mini Kit protocol ( Qiagen , UK ) . RNA concentrations were measured using Nanodrop ( ThermoFisher Scientific ) . First Strand cDNA synthesis was performed using 1 μg RNA and the Superscript III RT system ( ThermoFisher Scientific ) . cDNA was diluted 1:10 and 2–4 μl was used for one qPCR reaction with Taqman assays ( ThermoFisher Scientific; Table 8 ) . Relative gene expression was calculated using the ΔΔCT method relative to GAPDH control . P-values were obtained using an unpaired two-tailed student’s t-test with unequal variance . 10 . 7554/eLife . 26575 . 043Table 8 . RT-PCR primersDOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 043TaqMan primerCompanyCat noId noGAPDHThermo Fisher Scientific# 4326317En/aSOX2Thermo Fisher Scientific# 4331182Hs01053049_s1SOX9Thermo Fisher Scientific# 4331182Hs01001343_g1TBX4Thermo Fisher Scientific# 4331182Hs00218515_m1 Four age-matched biological replicates ( 6–7 pcw ) were selected based on size and hand/foot morphology . Fresh tips and stalks were microdissected and cleaned of mesenchyme using tungsten needles following 2 min in Dispase ( Gibco , 16 U/ml final concentration ) at room temperature . Microdissected tips ( ~8 ) and stalks ( ~8 ) were transferred by mouth pipette into 50 μl extraction buffer using the PicoPure RNA Isolation Kit ( ThermoFisher Scientific ) in DNA LoBind tubes ( Eppendorf ) . Organoids were removed from the Matrigel following the Corning Cell Recovery Solution protocol ( above ) and transferred into 50 μl extraction buffer . RNA extraction was performed according to the PicoPure RNA Isolation Kit protocol . Total RNA concentration and quality using RIN score was assessed using RNA 6000 Pico Kit ( Agilent ) . Only biological replicates with RIN score >8 were used . Reverse transcription and cDNA amplification was performed according to Ovation RNA-Seq Systems V2 protocol ( NuGEN ) . For each sample a minimum total input RNA amount of 500 pg was used . Quality check was performed using the Agilent DNA 1000 kit and RNA-Seq library preparation was performed according to the NuGEN Ovation Rapid DR Multiplex System 1–8 protocol . Sequencing was performed at the Gurdon Institute on a HiSeq 1500 in rapid run mode ( Illumina , San Diego , CA; single read 50 nucleotides ) . All RNAseq data deposited in GEO: https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? token=chyfeyegvxurngn&acc=GSE95860 . For comparing freshly isolated tips to freshly isolated stalks ( Figure 2 ) , Fastq files were filtered for low quality reads ( <Q20 ) and low quality bases were trimmed from the ends of the reads ( <Q20 ) using Sickle . The resulting reads were mapped to the human reference genome ( UCSC GRCh37/hg19 ) using TopHat 2 . 0 . 6 ( Kim et al . , 2013 ) guided by RefSeq gene models ( UCSC ) . Raw counts per transcripts were obtained using featureCounts and differentially expressed genes ( >2 fold difference ) identified using edgeR 2 . 6 . 12 ( Robinson et al . , 2010 ) . Hierarchical unsupervised clustering was performed using published foetal lung RNAseq data as a comparison ( Table 9 ) . Gene Ontology and Panther Pathway analysis was performed in DAVID ( Huang et al . , 2009a , 2009b ) . 10 . 7554/eLife . 26575 . 044Table 9 . Published whole foetal lung RNAseq ( Bernstein et al . , 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26575 . 044Sample labelDescriptionSourceDonor IDAccession #Lung_A_6Foetal day 105GEO DatasetsH-24005GSM1101693Lung_F_2Foetal day 105GEO DatasetsH-24111GSM1101708Lung_F_3Foetal day 108GEO DatasetsH-23887GSM1101684Lung_F_4Foetal day 91GEO DatasetsH-23914GSM1101685Lung_F_5Foetal day 96GEO DatasetsH-24089GSM1101699Lung_F_6Foetal day 98GEO DatasetsH-23964GSM1101687 To compare the human and mouse embryonic tip transcriptome , we compared our human tip RNA seq with previously published mouse E11 . 5 tip microarrays ( GEO accession numbers: GSM1968996 , GSM1968997 , GSM1968998 , GSM1968999 , GSM1969000 ) . We first assessed whether transcripts of orthologous mouse/human genes ( defined by the HomoloGene database ) were present in each data-set ( Figure 2—figure supplement 3A ) . Genes were excluded which had RPKM values < 1 ( RNAseq ) and expression values of <5 ( microarray ) . To estimate the relative levels of these transcripts between the mouse microarray and human RNAseq data , we reasoned that the microarray signal saturates and therefore generated a scatter plot of mean microarray signal , versus mean log-transformed RPKM for each orthologous gene identified in mouse and human ( Figure 2—figure supplement 3B ) . RNAseq for the cultured organoids was performed on an independent sequencing run . To compare freshly isolated human tip and stalk samples with the cultured organoids , the RUVSeq R package ( ruvg using housekeeping genes ) was used to control for the batch effect in the data ( Risso et al . , 2014 ) using hidden factor k = 1 . Multi-dimensional scaling plot , heat map and box plots ( Figure 3 and Figure 3—figure supplement 3 ) were produced in R using the batch-corrected data . Three days after splitting , colcemid was added to each well in self-renewing ( SN ) medium at 0 . 1 μg/ml for 48 hr . Organoids were then incubated in TrypLETM Express Enzyme ( ThermoFisher Scientific , 12604013 ) 37°C for 2 min and Advanced DMEM with 5% ( v/v ) foetal bovine serum ( FBS ) added . After centrifugation at 4°C , 300 g for 5 min , the cell pellet was resuspended in Advanced DMEM with 5% FBS and cells were triturated with a flame polished glass pipette . After further centrifugation at 1200 rpm for 7 min , 5 ml of 0 . 055 M KCl hypotonic solution was added to each tube which was gently inverted twice to mix . Tubes were centrifuged at 1200 rpm for 7 min . 500 μl of 3:1 100% methanol:glacial acetic acid fixative was added to each cell pellet dropwise down the side of the tube , then 1 . 5 ml was added in one go . Tubes were centrifuged at 1200 rpm for 7 min . The pellet was resuspended in fixative and stored at −20°C . All fixed cells were delivered to the Cytogenetics Laboratory at Cambridge University Hospitals NHS Foundation Trust for karyotyping . E13 . 5 whole mouse lungs were microdissected from Rosa26R-mT/mG heterozygous embryos ( 10 lungs ) and from MF1 embryos ( 24 lungs ) and cut into small pieces . Lungs were incubated in TrypLETM Express Enzyme ( ThermoFisher Scientific , 12604013 ) at 37°C for 2 min then 20 ml of Advanced DMEM with 5% FBS added . After centrifugation at 4°C , 300 g for 5 min and aspiration , the cell pellet was resuspended in Advanced DMEM supplemented with 0 . 5% ( v/v ) BSA and cells were triturated with a flame polished glass pipette ( no cell strainer was used ) . Cells were counted manually . Three biological replicates of human embryonic lung organoids were selected ( ~120 wells in total ) and processed as the whole mouse lungs . Human and mouse cells were then combined at a ratio of 1:4 ( human to mouse ) with 2 million cells in total per kidney capsule graft . For each of the three biological replicates , three human cell samples were mixed with MF1 lungs and one with Rosa26R-mT/mG/+ lungs . In order to facilitate kidney capsule transplantation , the human/mouse cell mixture was prepared into a cell aggregate ( Sheridan et al . , 2009 ) . Cells were spun at 300 g for 5 min and 100 μl of the cell suspension was aspirated with a sterile non-filtered P200 tip , the end part of which was shortened slightly . The end of the tip was then pushed into folded parafilm and secured by combining all folds upwards ( to prevent the cell suspension from leaking ) and then the whole tip was centrifuged in a 15 ml tube at 300 g for 5 min . A cell pellet became visible on the end part of the pipette nearest to the parafilm seal . The pipette tip was held horizontally to remove the parafilm and then the cell pellet was transferred directly onto a polycarbonate filter ( Millipore ) floating in a well of a 24 well plate . Overnight culture in self-renewing medium supplemented with 5% FBS was performed . The next morning , cell aggregates were transferred to Advanced DMEM . NSG male mice were anaesthetised and each cell transplanted into the left kidney capsule . Kidneys were harvested at 3 , 7 and 12 weeks post-transplant and fixed overnight 4% PFA at 4°C . Chimeric human-mouse MF1 grafts were harvested at 3 and 7 weeks and the human cells distinguished using HuNu antibody staining . Human–mouse Rosa26R-mT/mG grafts were harvested at 12 weeks and the human cells distinguished by the absence of red membranes . The 12 week hosts received three daily intraperitoneal injections of 0 . 5 mg/Kg body weight dexamethasone 1 week before culling . NSG male mice were used . Mice were anaesthetised using 2% ( v/v ) isoflurane and then exposed to bleomycin oropharyngeally through controlled aspiration on day −2 . Bleomycin ( clinical grade purchased from UCL pharmacy ) was prepared as a 1 mg/ml stock using sterile 0 . 9% ( w/v ) normal saline and administered as 1 μl/g body weight to each mouse . Human embryonic lung organoids were expanded as described above . On the day of transplantation organoids were harvested in 15 ml tubes using cold washing medium , centrifuged at 4°C , 300 g for 5 min and incubated in TrypLETM Express Enzyme ( ThermoFisher Scientific , 12604013 ) at 37°C for 2 min then 20 ml of Advanced DMEM with 5% ( v/v ) FBS added . After centrifugation at 4°C , 300 g for 5 min and aspiration , the cell pellet was resuspended in DMEM/F12 supplemented with 0 . 5% ( v/v ) BSA and cells were triturated with a flame polished glass pipette . Viable cells were counted manually and single cell morphology confirmed . 600 , 000 cells ( in 25 μl of DMEM/F12 supplemented with 0 . 5% ( v/v ) BSA ) were administered intratracheally to each mouse under isoflurane anaesthesia on day 0 . Lungs were harvested by culling the animals through intraperitoneal injection of sodium thiopental . Lungs were insufflated with 4% ( w/v ) PFA intratracheally , upon which the most proximal part of the trachea was tied with dental floss . The whole lung was then immersed in 4% ( w/v ) PFA and incubated overnight at 4°C . Tissue was then processed for cryo-sectioning . The following microscopes were used: Compound microscope: Zeiss Axiophot . Confocal microscope: 1 ) Leica SP8 , 2 ) Olympus FV1000 Inverted . Confocal z stacks were acquired at an optical resolution of 1024 × 1024 with an optical z slice every 1 μm for 40x images and every 2 . 3 μm for 20x images . Movies of growing organoids were captured by culturing in a Nikon Biostation and capturing bright-field images every 12 hr for up to 11 days . For estimating the proportion of mesenchyme after microdissection of the epithelium for RNAseq and organoid culture , a macro for Fiji was written by Richard Butler , Gurdon Institute Imaging Facility ( Mesenchyme_Macro . ijm is available as a supplemental file ) . The macro estimates the number of mesenchymal cells inside the 3D projection of a selected 2D region of interest by subtracting an E-Cadherin signal mask from a DAPI signal mask and dividing the remaining volume by a user-defined predicted nucleus volume . Quantitation of number of human grafts seen in bleomycin-injured mouse lungs was performed manually by counting the number of grafts , and number of cells per graft , seen per 20 consecutive 20x fields in 1 section of 1 random slide for each of the four mouse lungs in each of the four experimental groups . Quantitation of the number of human cells per graft was also manual . Quantitation of organoid forming efficiency with , or without , TGFβ inhibition was done based on the definition of an organoid as structures which had at least doubled in size compared to a fresh tip and had also branched . Quantitation of organoid size ( diameter at the widest point ) with , or without , TGFβ inhibition was expressed as a percentage increase comparing Day 11 to Day 1 . | Degenerative lung disease occurs when the structure of the lungs breaks down , which makes it harder to get enough oxygen into the bloodstream . Most , but not all , cases occur in smokers and ex-smokers or people who have been exposed to a lot of air pollution . Currently , there is no way to reverse the damage , and even slowing the progress of the disease is extremely difficult . Some researchers are looking for ways to treat patients with degenerative lung diseases by regenerating the surface of their lungs . However , it is still not clear what the most effective route towards this long-term goal will be . One approach to lung regeneration is to use findings from developmental biology to understand how embryos normally build the gas exchange surfaces in the lungs . This knowledge may allow scientists to trigger a similar process in an adult lung to renew or replace any diseased tissue . Alternatively , cells could be collected from patients , reprogrammed and then coaxed into becoming a gas exchange surface in the laboratory . Such a “lung-in-a-dish” could be used to understand how degenerative diseases develop , to discover and test new drugs , or even to treat the patient directly via a transplant . To date , the embryonic development of lungs has mostly been studied using mouse lungs as a model system . However , it was not clear if human lungs actually develop in similar ways to mouse lungs , and whether using mice is a valid research strategy . Nikolić et al . compared embryonic lungs from humans and mice and showed that they are indeed very similar in terms of the cell types that they contain and how they mature . However , some key differences were identified that can only be explored in human cells and tissue . Nikolić et al . went on to identify conditions that allowed them to grow cells from human embryonic lungs indefinitely in a dish . These cells can now be used to investigate the aspects of lung development that are specific to humans . Together these findings provide a useful guide to allow scientists to coax human cells growing in a laboratory to become lung cells . Further improvements to this process will make the lungs-in-a-dish more true to the real organs , meaning that they could be used to better understand lung disease and identify new medicines . In the longer term , Nikolić et al . hope to gain enough insight from the human lung-in-a-dish model to eventually be able to regenerate the lungs of patients with degenerative lung disease . However , this possibility is still many years away . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2017 | Human embryonic lung epithelial tips are multipotent progenitors that can be expanded in vitro as long-term self-renewing organoids |
Metazoan transcriptional repressors regulate chromatin through diverse histone modifications . Contributions of individual factors to the chromatin landscape in development is difficult to establish , as global surveys reflect multiple changes in regulators . Therefore , we studied the conserved Hairy/Enhancer of Split family repressor Hairy , analyzing histone marks and gene expression in Drosophila embryos . This long-range repressor mediates histone acetylation and methylation in large blocks , with highly context-specific effects on target genes . Most strikingly , Hairy exhibits biochemical activity on many loci that are uncoupled to changes in gene expression . Rather than representing inert binding sites , as suggested for many eukaryotic factors , many regions are targeted errantly by Hairy to modify the chromatin landscape . Our findings emphasize that identification of active cis-regulatory elements must extend beyond the survey of prototypical chromatin marks . We speculate that this errant activity may provide a path for creation of new regulatory elements , facilitating the evolution of novel transcriptional circuits .
Metazoan transcriptional circuitry features activation and repression signals that constitute robust regulatory networks important for the unfolding of developmental programs . In the Drosophila embryo , localized transcriptional repressors provide essential patterning information that establishes the primary anterior-posterior and dorsal–ventral axes of the organism . The action of transcriptional repressors is heterogeneous and can exhibit context effects; one of the most striking aspects involves the different classes of repressors that mediate distinct chromatin changes on target genes . Short-range acting proteins Snail and Knirps interfere with transcription only when their cognate binding sites are located within close range of the activator binding sites ( Gray and Levine , 1996 ) . These proteins interact with evolutionarily conserved corepressors that possess chromatin modifying activities ( Nibu et al . , 1998; Payankaulam and Arnosti , 2009 ) . Paradoxically , these same cofactors are also recruited by another class of repressors , the long-range transcriptional repressors , exemplified by the Hairy factor ( Paroush et al . , 1994; Barolo and Levine , 1997; Poortinga et al . , 1998 ) . This protein is a founding member of the Hairy/Enhancer of Split ( HES ) transcription factors , which play essential roles in animal development , including segmental gene patterning in the early embryo and specification of neuronal differentiation in response to Notch signaling ( Kageyama et al . , 2007 ) . Thus , elucidation of molecular mechanisms of Hairy activity will shed light on a number of important gene circuits that are prominently represented in key developmental pathways . The biochemical function of Hairy is associated with long-range chromatin modifications , which endow this factor with the ability to interfere with multiple cis-regulatory regions , including activators bound over 1 kb distal to the Hairy binding sites . The long-range effect has been proposed to be due to the recruitment of the corepressor Groucho ( Gro ) , that can oligomerize to spread over large areas of the genome , and colocalization of HDAC to the target genes resulting in deacetylation of specific lysine residues in histones H3 and H4 ( Courey and Jia , 2001; Martinez and Arnosti , 2008 ) . In our previous studies , we showed that Hairy induced extensive tracts of deacetylation on ftz , a segmental patterning gene expressed early in embryogenesis ( Li and Arnosti , 2011 ) . While potent in repression potential , Hairy and other long-range repressors are apparently restricted in their ability to exercise transcriptional effects by the local cis-regulatory context in which binding sites are located . Hairy was demonstrated to lack long-range effects on a distal RACE enhancer in the embryonic dorsal ectoderm , when Hairy binding motifs were situated in an element with activators that are restricted to mesoderm/neurectoderm regions . Furthermore , the Dorsal protein , when itself acting as a long-range repressor , is dependent on neighboring Cut and Dri transcription factor motifs to function , indicating that long-range repression complexes may require specific cis-regulatory grammar ( Cai et al . , 1996; Nibu et al . , 2001 ) . The action of eukaryotic transcriptional repressors involves a number of biochemical activities , including direct antagonism of transcriptional activators and assembly of chromatin-associated factors that are correlated with gene silencing ( Perissi et al . , 2010 ) . Specific types of covalent histone modifications , such as H3 and H4 deacetylation , H3K9 trimethylation and H3K27 trimethylation are correlated with repressed genes , but there is still no general understanding of how important in a quantitative sense such modifications are for inhibition of transcription at specific genes . Context effects for a particular transcriptional repressor can influence what sort and how much of a response will be generated . At a genome-wide level , specific chromatin features correlate with transcriptionally repressed genes ( e . g . , H3K9 and 27 methylation , reduced levels of H3 and H4 acetylation , binding of HP1 ) , however these marks are also found within highly active loci ( modENCODE Consortium et al . , 2010 ) . The epigenetic signature of transcriptional repression is thus context-dependent , consistent with a revised picture of the simple ‘histone code’ hypothesis . In the context of specific transcriptional repressors , we know little about how the context of distinct factors present at cis-regulatory elements shapes their action . Genome-wide information obtained from chromatin immunoprecipitation experiments should provide information about molecular targets and action of transcription factors , however , in addition to bona fide regulatory targets , metazoan transcription factors typically associate with a large number of in vivo binding sites of unknown significance . Recent studies have suggested that these interactions represent off-target genomic interactions , driven by low binding specificity of transcription factors and a general affinity for open chromatin of active enhancers ( MacArthur et al . , 2009 ) . A survey of possible ‘off target’ binding elements suggested that these tend to be of lower affinity and are transcriptionally inert ( Fisher et al . , 2012 ) . As noted above , previous studies of Hairy suggested that the protein is unable to mediate transcriptional repression in the absence of other factors co-occupying regulatory elements ( Nibu et al . , 2001 ) . Identification of functional properties of Hairy transcends the simple biochemical elucidation of repression; this protein is representative of the regulatory factors comprising conserved gene regulatory networks ( GRN ) that constitute the basis of animal development . Molecular studies have demonstrated that the acquisition or loss of binding sites or entire regulatory modules appears to drive significant changes in gene expression that initiate critical evolutionary transitions , such as elaboration of novel limb structures ( Khila et al . , 2009; Pavlopoulos et al . , 2009; Tanaka et al . , 2011 ) . Significantly , although relatively subtle changes have been linked to such important evolutionary innovations , it appears that functional conservation of gene expression is also compatible with major changes in the structure of transcription control regions ( Hare et al . , 2008 ) . The constraints for reorganization of existing cis-regulatory elements , or appearance of such elements de novo , are poorly understood; in some cases , the exact placement of multiple transcription factor motifs is essential for transcriptional function , while the composition of other genetic switches appears to be very loosely organized ( Arnosti and Kulkarni , 2005 ) . The existence of a large fraction of ‘off-target’ binding sites both complicates the analysis of important functional links , and the interpretation of potential evolutionary changes . Thus , elucidation of the functional targets and chromatin effects of Hairy can provide important insights on the basic substance of evolutionary variation . In this study , we use genetic tools to mediate induction of Hairy on a short time scale , permitting us to identify direct regulatory targets and chromatin effects on a genome-wide level . In addition to identifying common features of Hairy repression mechanisms across many targets , we also show that this protein exerts pervasive biochemical activity to change chromatin states at many loci unlinked to gene expression , revealing a possible pathway to evolution of novel gene regulatory connections .
To study transcriptional repression at the genome-wide level at this important developmental stage , we profiled changes in transcriptome , epigenome and RNA polymerase II ( Pol II ) binding regulated by Hairy in the blastoderm embryo using an inducible system as described previously to capture direct effects with high temporal resolution ( Li and Arnosti , 2011 ) ( Figure 1A ) . Hairy is first expressed in the Drosophila blastoderm embryo in a seven stripe pattern , which is important in controlling downstream pair rule genes that direct segmentation ( Ish-Horowicz and Pinchin , 1987 ) . Here , we express Hairy with a brief heatshock , throughout the embryo , which is sufficient to completely repress target genes such as ftz ( Figure 1A , B ) . We treated the control embryos identically to embryos carrying the inducible Hairy transgene to test for possible nonspecific effects of heat shock on gene expression and chromatin marks . In this system , heat shock alone has no effect on the expression patterns of the pair rule and other genes analyzed , and the chromatin marks in heat shocked control embryos were indistinguishable from chromatin patterns previously reported for untreated embryos ( Li and Arnosti , 2011 and K Kok , data not shown ) . In total , we identified 241 down-regulated and 146 up-regulated transcripts in response to induction of Hairy ( Figure 1C ) . Our microarray analysis captured previously identified targets of Hairy , showing downregulation of en , edl , Impl2 , and prd , as well as ftz , all of which were previously found to be derepressed in h embryos ( Ish-Horowicz and Pinchin , 1987; Bianchi-Frias et al . , 2004 ) . 10 . 7554/eLife . 06394 . 003Figure 1 . Global analysis of Hairy regulation . ( A ) Schematic expression of Drosophila embryo system used for Hairy repression , with outline of the genome-wide analysis of transcription , chromatin , and RNA polymerase II ( Pol II ) . ( B ) Repression of ftz , odd , comm and esg revealed by in situ hybridization in wild-type ( wt ) and Hairy transgenic embryos ( hs-hairy ) after 20 min induction . Similar repression of 18w , HLHm7 and erm was also observed ( not shown ) . ( C ) Transcriptionally regulated ( red , down; blue , up ) and Hairy bound genes identified by microarray and ChIP–chip ( MacArthur et al . , 2009 ) . A larger fraction of down-regulated genes were physical targets of Hairy than for up-regulated genes ( significance: p = 3 . 8e-95 and p = 3 . 5e-08 respectively , hyper-geometric test ) . Differentially expressed genes are selected based on p < 0 . 05 and fold change >2 . ( D ) Validation of microarray data by RT-qPCR , showing concordance between these methods . Genes are ranked by the fold change from the microarray measurements . Significance was tested by Student's t-test . y-axis values were normalized as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 00310 . 7554/eLife . 06394 . 004Figure 1—figure supplement 1 . Similarity between binding of endogenous Hairy and overexpressed Hairy protein . ( A ) Similarity of promoter proximities . Histogram shows the global distribution of Hairy peaks around TSS identified from ChIP-seq of induced Flag tagged Hairy protein ( right panel ) and by previous ChIP–chip detection of endogenous Hairy binding ( left panel ) ( MacArthur et al . , 2009 ) . ( B ) Genomic annotation of peaks shows similar binding distributions on genic and intergenic regions . ( C ) Area-proportional Venn diagram showing significant overlap between endogenous and induced Hairy binding ( p = 2 . 15e-159 ) . ( D ) De novo motif analysis reveals similar motifs enriched under peaks of both data , including canonical Hairy binding site ( CACGCG ) . We used the ChIP–chip data from MacArthur et al . ( 2009 ) for our analysis because the Flag epitope gave low signals overall , although high-confidence functional targets such as ftz , Impl2 , odd , h , 18w , wg , tup , pros , nht , and en were found . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 004 Differentially regulated genes were compared to those physically bound by Hairy ( MacArthur et al . , 2009 ) ; 70% of down-regulated genes are bound by Hairy , suggesting that most of these are likely to be direct targets ( Figure 1C ) . In contrast , only 30% of up-regulated genes are bound by Hairy , indicating that majority of these genes may be indirect targets . In situ hybridization and RT-qPCR confirmed the repression of a number of target genes we identified ( Figure 1B , D ) . Many of these genes , including odd , comm , comm2 , edl , en , Impl2 , prd , and 18w , have striped expression patterns complementary to that of Hairy , supporting direct regulation by the repressor . Furthermore , consistent with known biological functions of Hairy , gene ontology analysis showed that categories for down-regulated genes are significantly enriched in transcriptional regulation , cell fate commitment and neurogenesis ( p < 3 . 7e-18 ) . GO categories for the set of upregulated genes were of lower statistical significance , and included reproductive processes ( p < 0 . 03 ) ( Supplementary files 1 , 2 ) . Expression of the majority of genes bound by Hairy did not change ( Figure 1C ) , consistent with previous observations that metazoan transcription factors have apparently many ‘nonfunctional’ interaction sites in the genome ( Cusanovich et al . , 2014 ) . Identification of functional and physical Hairy targets allowed us to study gene-specific chromatin changes associated with repression . We performed epigenomic profiling via chromatin immunoprecipitation-high throughput sequencing ( ChIP-seq ) of chromatin marks that are often correlated with specific features of cis-regulation; H4Ac , H3K27Ac , and H3K4me1 at promoters and enhancers; H3K4me3 at transcription start sites ( TSS ) ; H3K36me3 at gene body regions; and H3K9me3 at repressed regions of chromatin ( Zhou et al . , 2011 ) . The measured signals for specific marks were highly reproducible in separate biological replicates , and Hairy-induced changes in histone marks were consistently observed at specific loci , such as the widespread loss of the H4Ac signal on the ftz locus , with little change to the overall global chromatin landscape ( Figure 2—figure supplement 1A ) . As was apparent from comparison of control chromatin profiles , the induction of Hairy did not cause a global impact on histone marks . In the presence or absence of induced Hairy , the genome features for multiple chromatin marks are virtually identical , except in very discrete regions where there are significant changes ( Figure 2—figure supplement 1A–C ) . 10 . 7554/eLife . 06394 . 005Figure 2 . Examples of coupled , large-scale chromatin changes mediated by Hairy . Chromatin immunoprecipitation-high throughput sequencing ( ChIP-seq ) tracks for H4Ac , H3K27Ac , H3K4me1 and H3K4me3 are shown at repressed genes before ( − ) and after Hairy ( + ) induction , with gene models below . ( A–D ) Coupled reduction of active histone marks was observed in a wide-spread fashion on ftz , h , 18w and odd genes ( scale at top left ) . ( E–H ) Relatively smaller blocks of chromatin changes were detected on HLHm7 , gogo , pros and tup genes . Significantly changed regions ( shaded boxes ) were identified by the diffReps program . Hairy binding ( top track ) from MacArthur et al . ( 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 00510 . 7554/eLife . 06394 . 006Figure 2—figure supplement 1 . ChIP-seq reproducibility of biological replicates and variation between wild-type ( wt ) and transgenic embryos ( H ) . ( A ) Specific reduction of H4Ac signal at ftz locus ( red box ) in three biological replicates after induction of Hairy ( H ) . ( B ) H4Ac peaks were not altered globally in genome by Hairy expression . Heatmaps show 5 kb window centered on called H4Ac peaks , ranked by peak height . ( C ) Measured global chromatin features were similar in wt and H samples , indicating that Hairy does not affect the majority of chromatin features throughout the genome . Scatter plots indicate the correlation ( r = Pearson's correlation coefficient ) between wt and H embryos for H4Ac , H3K27Ac , H3K4me1 , H3K4me3 , H3K36me3 and H3K9me3 marks . Each dot represents a peak . ChIP-seq read counts on the axis are transformed to log2 base . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 006 Using single gene techniques , we previously found that Hairy induces a widespread histone H4 deacetylation throughout the entire ftz locus ( Li and Arnosti , 2011 ) . To determine if these are general properties of Hairy , we compared all affected loci genome-wide . We observed that on a number of transcriptionally repressed target genes , H4 deacetylation is coupled with loss of the active marks H3K27Ac and H3K4me1 . Widespread reduction of these active marks affecting >1 kb blocks was observed on many genes repressed by Hairy , including ftz and other segmentally expressed genes such as h and 18w ( Figure 2 ) . Notably , Hairy regulates its own transcription by chromatin alteration , consistent with autoregulatory mechanism of related mammalian HES proteins ( Kageyama et al . , 2007 ) . In addition to removal of enhancer marks , repression on h and 18w resulted in demethylation of the promoter mark H3K4me3 . Furthermore , action of Hairy on another pair rule gene , odd , was limited to removal of acetyl marks on H4 and H3K27; methylation marks on H3K4 are untouched ( Figure 2D ) . These results suggest that Hairy mediates coordinated sets of chromatin transitions . The chromatin changes did however exhibit heterogeneous characteristics; the sizes of altered chromatin domains varied on different repressed genes . For example , changes in levels of H4Ac involved blocks with a range of sizes; generally larger than 1 kb , with the average ∼2 . 5 kb . Somewhat smaller chromatin blocks were associated with repression of the HLHm7 , gogo , pros and tup genes , which showed just as robust regulation of transcription as those genes with large tracts of chromatin modification ( Figure 2E–H ) . The largest ranges of size in chromatin domains were observed for H4Ac , but similar , although smaller ranges were also seen for H3K27Ac and H3K4me1 marks ( Figure 3 , Figure 3—figure supplement 1A , B and Supplementary file 3 ) . We found strong correlations between the sizes of the domains of chromatin modification and the direct action of Hairy . Hairy-bound blocks of deacetylation were significantly larger than those not bound by Hairy , and smaller correlations were noted for other modifications , indicating that deacetylation is especially likely to show ‘spreading’ characteristics ( Figure 3 , Figure 3—figure supplement 1A , B and Supplementary file 3 ) . 10 . 7554/eLife . 06394 . 007Figure 3 . Direct Hairy target genes exhibit broad domains of chromatin effects . Distribution of genome-averaged ChIP-seq signals before ( straight line ) and after ( dashed line ) Hairy induction , showing 4 kb window around affected regions . ( A ) Distributions of histone H4Ac and H3K27Ac marks of direct Hairy targets were significantly broader than for regions ( B ) not bound by Hairy ( p = 2 . 55e-92 and p = 5 . 63e-70 respectively; KM test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 00710 . 7554/eLife . 06394 . 008Figure 3—figure supplement 1 . Distinct chromatin profiles associated with direct and indirect Hairy targeted loci . Histograms show the distribution of averaged ChIP-seq signals in a window of 4 kb centered on differentially changed regions associated with Hairy bound ( A ) and unbound ( B ) genes in the wild-type ( wt , solid lines ) and Hairy induced ( H , dashed lines ) embryos for H4Ac , H3K27Ac , H3K4me1 , H3K4me3 , H3K36me3 and H3K9me3 . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 00810 . 7554/eLife . 06394 . 009Figure 3—figure supplement 2 . Little correlation between height of Hairy peaks or width of Hairy-bound region and extent of H4 deacetylation blocks and width ( A ) or height ( B ) of Hairy peaks . Other marks also exhibited little correlation between Hairy peak width and height and range of chromatin alterations ( not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 009 These results suggest that widespread effects found at H4Ac , H3K27Ac and H3K4me1 marks are dependent on presence of Hairy and are consistent with a long-range ‘spreading’ repression mechanism . We saw no correlation between the height or extent of Hairy binding sites and the range of chromatin alteration , suggesting that the effectiveness of this protein is not merely a function of number of binding sites ( Figure 3—figure supplement 2A , B ) . Other local factors may dictate how extensively modifications are propagated on individual genes . Therefore , Hairy induces diverse chromatin transitions associated with gene silencing , indicating that there are gene-specific features dictating how repression is mediated at individual genes . These observations suggest there are context-specific aspects to chromatin modifications directed by Hairy . To determine the nature of changing chromatin states at different genomic loci , we compared the complete set of significant alterations in all measured chromatin marks observed after Hairy induction , regardless of transcriptional effects on the neighboring genes . We observed both loss and gain of these marks on hundreds of regions . Most frequently observed were changes in H4Ac , H3K27Ac , H3K4me1 and H3K36me3; changes in some chromatin marks were much more frequent than in others , indicating that there is some heterogeneity in the impact of Hairy on different regions ( Figure 4A ) . The changes in levels of these marks is not simply due to increased or decreased histone density , as histone H3 levels generally were unchanged ( Figure 4A ) . The roughly equal abundance of regions showing loss or gain of acetylation and methylation would indicate that either secondary effects are common , or that Hairy may exert distinct biochemical activities on different loci . The correlation of Hairy-bound regions with repressed transcripts , as well as the association of Hairy binding with longer-range deacetylations , but not with increased acetylation , supports the idea that indirect effects are common . Indeed , focusing specifically on genes targeted by Hairy , we found that H4 histone deactylation was strongly enriched compared to acetylation gains , suggesting that deacetylations are direct effects ( Figure 4B and Figure 4—figure supplement 1A ) . Further support comes from consideration of the actual Hairy occupancy of the chromatin blocks in question; there was significant correlation between Hairy binding and chromatin blocks exhibiting decreased , but not increased acetylation ( Figure 4—figure supplement 2 ) . 10 . 7554/eLife . 06394 . 010Figure 4 . Pervasive genome-wide chromatin effects of Hairy . ( A ) All reduced ( top ) and increased ( bottom ) chromatin marks in the genome for H4Ac , H3K27Ac , H3K4me1 , H3K4me3 , H3K36me3 , H3K9me3 and H3 shown as heatmaps for 5 kb windows from the center of significantly affected regions before ( − ) and after ( + ) Hairy induction . The number of affected regions indicated below each mark . ( B ) Affected chromatin regions associated with Hairy-bound genes show preferential enrichments for H4Ac , H3K27Ac , and H3K4me1 . All affected regions were assigned to closest genes , and those in the vicinity of Hairy-bound genes are shown . ( C ) Subset of modified regions from ( B ) that were linked to genes transcriptionally regulated by Hairy . Significance of enrichment for chromatin modifications shown in Figure 4—figure supplement 1A , B . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 01010 . 7554/eLife . 06394 . 011Figure 4—figure supplement 1 . Significance of individual histone modifications associated with Hairy bound genes and transcriptionally regulated genes . ( A ) Strongest link between loss of H4Ac , gain of H3K4me1 , and presence of Hairy on genes . ( B ) Transcriptionally repressed genes associated with loss of H4Ac , H3K27Ac , and gain or loss of H3K4me1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 01110 . 7554/eLife . 06394 . 012Figure 4—figure supplement 2 . Strong correlation between the presence of Hairy binding and chromatin alterations on specific chromatin blocks . Reduced H4Ac , H3K27Ac , and H3K4me1 significantly associated with Hairy binding . Hairy bound regions overlapped with chromatin blocks by at least 1 bp . y-axis indicates p-value ( logln ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 012 With respect to another chromatin mark , changes in histone methylation revealed an unexpected and interesting trend . Both decreases and increases in H3K4me1 signals were significantly associated with Hairy-bound genes; decreases were especially found in those regions directly bound by Hairy ( Figure 4—figure supplement 1A , Figure 4—figure supplement 2 and Figure 4B ) . At the same time , about one-quarter of the genes that were transcriptionally silenced by Hairy showed increases in H3K4me1 , although these regions of increase did not overlap with Hairy binding . The increase in this mark may represent a reaction of proximal promoter chromatin to distal enhancer silenced by Hairy . H3K36me3 modification is often associated with active transcription . We found a small fraction of transcriptionally regulated genes that exhibited changes in the mark upon transcriptional repression ( Figure 4C and Figure 4—figure supplement 1B ) . These findings indicate that Hairy repression does not require H3K36me3 changes . Indeed , the direct effect of H3K36me3 on transcription is complex , as has been found for many other histone marks . For example , upregulation of KDM4A histone demethylase target genes in Drosophila occurs without increases in H3K36me3 ( Crona et al . , 2013 ) . Similar studies with elongation factor Spt6 in Drosophila further indicate that Hsp70 gene expression is not correlated to H3K36me3 levels ( Ardehali et al . , 2009 ) . In fact , H3K36me3 may in some contexts contribute to gene silencing , due to its presence in heterochromatic domains ( Chantalat et al . , 2011 ) and in other cases , removal of H3K36me3 is required to promote transcriptional elongation ( Kim and Buratowski , 2007 ) . A smaller number of H3K9me3 regions were observed to change globally , or on genes that were associated with Hairy ( Figure 4A , B ) . Very few repressed genes showed any alteration in this mark , thus it appears that repression mediated by Hairy does not require changes in such repressive histone modifications ( Figure 4C ) , consistent with our previous report that repression on ftz did not change H3K27me3 levels ( Li and Arnosti , 2011 ) . Indeed , other studies have found that these marks are not always simply coupled to repression . For example , only a modest correlation between H3K9me3 and H3K27me3 levels and gene silencing was observed in human cells ( Barski et al . , 2007; Zhang et al . , 2012 ) . In the differentiation of T and B cells , only a small fraction of repressed genes ever acquire H3K27me3 ( McManus et al . , 2011; Zhang et al . , 2012 ) . Interestingly , H3K9me3 was found to be enriched in many active promoters and associated with transcriptional elongation in vertebrates ( Vakoc et al . , 2005; Squazzo et al . , 2006 ) . Consequently , of the assessed modifications , it appears that Hairy predominantly works to modify acetyl and methyl marks of H4 , H3K27 and H3K4 and represses gene expression primarily by eliminating active marks . Of all chromatin regions impacted by Hairy , only a small number are associated with genes demonstrating measurable transcriptional changes ( Figure 4C ) . Thus , it is striking that the majority of chromatin changes are decoupled from any detectable effect on gene expression ( Figure 4—figure supplement 1B ) . For the many cases where chromatin effect was unlinked to changed gene expression , we observed extensive chromatin alterations associated with both silent and active genes . For example , chromatin transitions occur on transcribed genes not functionally repressed by Hairy , as seen on the pyr gene ( Figure 5A ) . In this case , the gene may remain active because the necessary cis-regulatory elements are located distally and are still able to interact with the promoter and activate it . In other cases , chromatin changes flank silent loci; nht undergoes widespread deacetylation and demethylation even though it is silent during this developmental stage of embryos ( Figure 5B ) . In some cases , binding and changing chromatin near inactive genes by Hairy in the blastoderm embryo may involve the interaction of Hairy with DNA elements that will become active at a later developmental stage , however , this seems unlikely in the case of nht , a testes-specific gene . Here , the physical binding by Hairy and subsequent impact on chromatin may represent ‘errant targeting’ . Overall , chromatin changes were observed to correlate with over half of the regions bound by Hairy , suggesting that in most cases , this protein is biochemically active on chromatin , whether or not the changes lead directly to gene repression ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 06394 . 013Figure 5 . Examples of chromatin-modified loci unlinked to changes in gene expression . ( A ) pyr is actively transcribed , and not significantly repressed by Hairy , ( B ) while nht is not expressed at this stage . ChIP-seq tracks for H4Ac , H3K27Ac , H3K4me1 and H3K4m3 are shown before ( − ) and after Hairy ( + ) induction . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 01310 . 7554/eLife . 06394 . 014Figure 5—figure supplement 1 . Global association of Hairy binding with histone mark alterations . Changes in histone marks , predominantly reductions , were detected for more than half of the Hairy bound genes . Genes were divided into two groups; no detectable histone mark changes vs at least one change , and then ranked ( right to left ) by height of Hairy peaks and total number of changes in histone marks . Differential changed regions of histone marks and Hairy peaks were assigned to genes with the closest TSS . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 014 The individual cases described in Figure 2 suggest that Hairy organizes a coordinated set of chromatin changes involving both deacetylation and demethylation of multiple histone residues . To determine if such alterations are a general property of the repressor , we assessed the extent of coordination of modifications on all individual blocks of affected chromatin . Changes in H4Ac , H3K27Ac and H3K4me1 marks were significantly correlated at many loci ( Figure 6A ) . Deacetylation events were also strongly correlated with loss of both H3K4me1 and H3K4me3 , indicating that Hairy may form complexes containing both deacetylase and demethylase activities . Indeed , the CtBP cofactor is known to bind both of these classes of enzymes . However , Hairy is not mediating only one average type of transformation; removal of methyl groups from H3K4me1 and H3K4me3 is catalyzed by distinct classes of enzymes; Hairy is likely to interact with both , allowing for removal of H3K4me1 marks on distal sites and H3K4me3 at TSS ( Figure 6B ) . A very similar pattern of correlations between acetylation marks , and between acetylation and methylation marks was observed for regions with increased acetylation and methylation . These elements may represent to a large extent indirect targets of Hairy , as no significant overlap between Hairy binding and these modified regions was found ( Figure 4—figure supplement 2 ) . 10 . 7554/eLife . 06394 . 015Figure 6 . Coordination in changes of specific chromatin modifications by Hairy . ( A ) Very strong overlap between decreases in regions of H4Ac , H3K27Ac , H3K4me1 ( heat map , upper left quadrant ) . Similar coordination between increases of H4Ac , H3K27Ac , H3K4me1 was noted ( lower right quadrant ) . Combined increases and decreases of different marks were rarely observed . ( B ) Distribution of modified blocks by genomic regions show preferential action of Hairy at a distance from transcription start site ( TSS ) . Affected regions were mapped to intergenic regions , promoter , exon etc . Overall distribution of genomic peaks for measured marks shown at right; the distributions for affected H4Ac and H3K27Ac regions deviated from the genomic averages ( left , decreased , and center , increased levels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 015 Where does Hairy most commonly mediate significant chromatin modifications ? We compared the location of individual histone marks genome-wide to those altered by Hairy expression . Although a third of Hairy binding sites are promoter-proximal , where the majority of H4 and H3K27 actetylation occurs , the large majority of affected chromatin sites were found on intergenic and intronic regions , suggesting that successful alterations are targeted to distal sites that may represent transcriptional enhancers ( Figure 6B ) . By contrast , changes in the methylation marks H3K4me1 , H3K4me3 , H3K36me3 , and H3K9me3 are found in the genomic regions where they are naturally enriched ( Figure 6B ) . For instance , H3K4me3 marks are enriched at TSS , as are the bulk of the altered chromatin sites . Hairy may thus have privileged sites on which it is more likely to induce chromatin changes; promoter regions may be in general more resistant to acetylation changes if strong activators are replenishing acetylation marks at these loci . In addition , transcriptional targets of Hairy are enriched in developmentally regulated genes , which typically possess larger cis-regulatory regions with multiple distal enhancers ( Supplementary file 1 ) ( Nelson et al . , 2004 ) . To directly assess the influence of Hairy on transcriptional machinery , we compared the genome-wide occupancy of RNA Pol II before and after Hairy induction . 75 of 241 repressed genes exhibited changes in Pol II occupancy ( Figure 7A ) . Only three of those are not directly bound by Hairy , indicating a direct regulation by Hairy in loss of Pol II signal . A marked decrease of Pol II occupancy was observed at the ftz promoter , gene body and distal downstream region ( Figure 7B ) . Loss of binding at the promoter , or the body of the gene , or both was detected on other loci ( Figure 7C–I ) . Thus , the loss of Pol II on the promoter and gene body of ftz is not universally associated with transcriptional repression; on other genes , silencing of a distal enhancer may interfere with promoter release without blocking polymerase recruitment to the promoter , consistent with recent studies implicating transcriptional signaling in promoter escape , rather than promoter recruiting ( Lagha et al . , 2013 ) . As expected , genes with associated chromatin changes without any impact on transcription did not show any change on Pol II occupancy ( Figure 7J , K ) . 10 . 7554/eLife . 06394 . 016Figure 7 . Diverse impact on RNA Pol II occupancy by Hairy . ( A ) A minority of genes show significant changes in Pol II occupancy after Hairy repression , although a larger proportion of the directly targeted genes have measureable decreases in Pol II . ‘Repressed genes’ shows entire set of transcriptionally downregulated genes , with reduced Pol II occupancy shown in dark gray . Subsets of genes directly bound or not bound by Hairy shown in center and at right . ( B–I ) Pol II occupancy on transcriptionally regulated genes before ( − ) and after ( + ) Hairy induction . Pol II occupancy decreases in the promoter and gene body of ftz and odd , only on the gene body of h , 18w and pros , and only at the promoter of HLHm7 and gogo . Pol II signal was not changed significantly on tup . ( J , K ) Consistent with lack of transcriptional effects on other genes with associated chromatin modifications , Pol II occupancy on pyr is not changed , and absent on nht . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 016 95 repressed genes bound by Hairy did not show any change in Pol II occupancy ( Figure 7A ) . It is possible that Hairy induces a slower transit rate of Pol II without any detectable change in Pol II binding . It has been suggested that repression through elongation control may cause no change in Pol II binding on slp1 and Hsp70 ( Adelman et al . , 2006; Wang et al . , 2007; Ardehali et al . , 2009 ) . Our previous analysis of eve repression by short-range repressor Knirps showed similar effects ( Li and Arnosti , 2011 ) . Therefore , Hairy may interfere with gene expression at different steps of the transcription cycle , as also suggested for repression by the glucocorticoid receptor , indicating gene specific repression mechanisms ( Gupte et al . , 2013 ) . An additional consideration is that genes featuring poised polymerase at the promoter in many or most nuclei , but are only expressed in a few nuclei , will have weak signals at the body of the gene . Therefore , the lack of change in Pol II levels on the gene body would reflect the inherently low signal , rather than a distinct biochemical mechanism . This explanation may account for a considerable number of affected genes where no changes in Pol II levels are observed after repression . The complexity of chromatin transitions observed genome-wide in the wake of Hairy expression prompted us to ask which features best predict successful repression of a target gene , vs those genes with no chromatin responses or exhibiting errant targeting by Hairy . Here , we alter the expression of only one regulatory factor , rather than the many changes in regulatory factors observed over a developmental time course , therefore our data sets are enriched for direct action of Hairy , potentially simplifying the search space . We sought out correlations between dynamic histone marks , Pol II , Hairy , CtBP and Gro and the repression of targeted genes . Direct inspection reveals that occupancy by Hairy , Gro , and decreases in Pol II are strong predictors of repression , as are several histone marks , compared to genes unaffected or those activated ( Figure 8A ) . However , there are numerous loci that do not fit these simple generalizations . To more systematically assess the connections between these different observed states and transcriptional repression , we applied machine learning to analyze features that may be implicated in the activity of Hairy . We tested 41 features , including the number of observed peaks for Hairy , CtBP , and Gro; the number , width , and magnitude of altered chromatin blocks , and distance to TSS for 583 genes ( 241 repressed , 146 activated and 196 unaffected genes; activated and unaffected genes were grouped as nonrepressed genes ) . To identify the most informative features , four different feature selection algorithms were used to rank the information content of the 41 measured properties associated with the genes; the top twenty of these features were then used for predictions ( Supplementary file 4 ) . We then tested four classifiers , using 90% of the data for training and 10% for predictions , with 10-fold cross-validation . Overall , each of the classifiers performed better than background , with Random Forests showing superior performance of ∼75% accuracy for repressed and nonrepressed genes ( Figure 8B ) . Three of the feature selection algorithms used with this classifier employed very similar features to achieve this high level of accuracy ( Supplementary file 4 ) , indicating that certain features are most informative . The presence and properties of Hairy and Gro peaks are good indicators , although not sufficient information by themselves . RNA Pol II properties , transcript levels , and chromatin modifications , especially H3K4me1 and H4Ac , whether causal or not , are also a close reporter of gene activity . The overall performance differences in these methods are frequently observed in machine learning studies , and likely reflect the underlying data structure and types of features available for analysis . Genes that were correctly predicted as repression targets generally had the most differential features , including binding by Hairy and Gro , and changes in histone modifications . The genes that were least successfully called had one or no differential features , and may represent genes that are expressed in fewer cells and at lower levels where measurement of chromatin changes in a global population is difficult ( Figure 8C ) . The nonrepressed gene pyr was consistently called as ‘repressed’ by the machine learning algorithms , as it exhibited chromatin signatures similar to those found on genes that were actually repressed ( Figure 8C ) . In this case , we propose that the relevant enhancers lie outside of the chromatin regions affected by Hairy . Such genes may represent loci that are poised for capture in the Hairy regulatory network through stepwise acquisition of activator binding sites . Overall , this analysis indicates that from the perspective of Hairy biochemistry , there are intuitive and some non-intuitive combinations of chromatin dynamics that typify this protein's action in the context of transcriptional repression , rather than a ‘practice’ site , but other factors predominate in many instances . The missing information likely relates to the activity of bona fide cis- regulatory elements that are acting on genes in the vicinity of Hairy , which is partially but incompletely known from genome-wide studies ( Kvon et al . , 2014 ) . 10 . 7554/eLife . 06394 . 017Figure 8 . Machine learning reveals complex chromatin code for repression of Hairy target genes . ( A ) Changes in histone marks , Pol II occupancy and Hairy , Gro and CtBP binding on repressed ( red ) , activated ( green ) and unaffected ( black ) genes upon Hairy induction . Genes were grouped by change in expression , then subgrouped into Hairy bound or unbound , and finally ranked by fold change in gene expression . Activated and unaffected genes were grouped as nonrepressed genes . ( B ) Relative success rate at calling repressed and nonrepressed genes for four different machine learning models . Background prediction for this entire set is expected to be 58%; Random Forests , Naive Bayes , KNN classifiers had an average success of 75% overall , while the SVM classifier was not better than background . Classifiers were used in conjunction with Information Gain , Symmetrical Uncertainty , Chi Square and Relief feature selection algorithms . The average prediction accuracies of each method are shown in the first column . Expected random success ( 42% ) for repressed genes ( middle column ) shown on heat map scale bar . ( C ) Model predictions for subset of repressed genes including those identified in Figure 1; top 19 were successfully predicted by almost all methods . fra , Optix , dib , and onecut were genes with disparate predictions that had few measureable chromatin features . At bottom , uniform false ‘repressed’ calls for pyr , which was not transcriptionally repressed . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 017
By testing direct effects of the Hairy repressor in the embryo , we conclude that this protein coordinates a stereotypical set of chromatin modifications , modulated by local context , that underlie its function as a long-range repressor . Most remarkably , these changes on chromatin impact large segments of the genome that are not directly relevant to gene expression in this developmental context . We speculate that these off-target activities may provide an easy entry point for evolution of novel regulatory switches ( Figure 9 ) . Our mechanistic analysis of Hairy provides insights into likely mechanisms of related HES factors , as well as other transcriptional repressors that serve as scaffolds for chromatin modifying complexes . Hairy interacts with the widely utilized cofactors Gro , CtBP , and the Sir2 HDAC , and here we provide for the first time a genome-wide picture of the biochemical activities of this archetypal repressor . 10 . 7554/eLife . 06394 . 018Figure 9 . Pervasive biochemical activities on ‘off-target’ loci may represent molecular exaptations that generate novel edges between nodes of a standing gene regulatory networks ( GRN ) . Functional and nonfunctional interactions of Hairy with chromatin are depicted . ( A ) Hairy repression of target genes results in loss of active histone marks such as H4Ac , H3K27Ac , and H3K4me1 ( dark gray peaks; gene x ) . Hairy interacts with many other nonfunctional targets where it carries out biochemical activities similar to those seen on transcriptionally controlled loci ( gene y ) . The latter chromatin changes are inconsequential and unlikely to be evolutionarily selected . ( B ) Gain of activator sites in a region of Hairy-modified chromatin may generate an on/off switch and result in functional targeting . ( C ) Schematic representation of cooption of Hairy physical interaction into modified GRN . DOI: http://dx . doi . org/10 . 7554/eLife . 06394 . 018 How is transcription actually controlled by Hairy ? The associated chromatin modifications may be effects , rather than direct causes of gene silencing . Our previous studies indicated that Hairy modulated transcription independent of activator occupancy or SAGA co-activator occupancy ( Martinez and Arnosti , 2008 ) . These previous observations raised the possibility that Hairy acts through entirely independent pathways from that employed by activators to block transcription . Our work here indicates that Hairy does indeed directly reverse chromatin marks associated with activators , and may therefore work through a dynamic competition with these activators , undoing their positive influence on the chromatin environment that would be necessary for RNA polymerase initiation and/or elongation ( Figure 9A ) . Indeed , Hairy repression is readily reversible , with genes showing reversion to an active state minutes after depletion of the overexpressed repressor ( K Kok , unpublished results ) . The genome-wide analysis of repression by Hairy revealed an unexpected facet of chromatin activity and highlights the need to consider the activity of ‘off target’ sites in generating novel elements , particularly because for Hairy at least ( and likely other factors that employ the same cellular machinery ) they are ‘shovel ready’ and not constrained by complex cis-regulatory grammar . Metazoan transcription factors typically interact with thousands of discrete sites in the genome , but only a small subset of these interactions correlate with observable effects on gene expression . In this study , we combined analysis of gene expression and chromatin dynamics in a way that allowed us to attribute effects directly to the induction of Hairy , inferences that would be difficult with a loss-of-function assay due to kinetics of depletion and secondary effects . In contrast , many other genome-wide data sets provide a static snapshot of the extant chromatin landscape or track complex changes through development , which represents the combined contributions of many activators and repressors . Previous studies have noted the presence of detectable but lowly-occupied sites , which have been suggested to reflect non-specific , non-functional interactions that are unavoidable by-products of proteins binding to large genomes ( Fisher et al . , 2012 ) . Other studies have emphasized that transcription factors may have general nonspecific interaction with HOT sites that represent preferences for open chromatin ( Gerstein et al . , 2010; modENCODE Consortium et al . , 2010 ) . In general , the overall view is that whether or not these interactions are conserved , they may be of little functional consequence , and are not important for activity of GRNs ( Cusanovich et al . , 2014 ) . Importantly , considering our finding that ‘off-target’ Hairy sites still appear to regulate chromatin structure , we should fundamentally reconsider how we interpret genome-wide data sets . Frequently , an increase in H3K27 acetylation is taken as an indication that the element is an active enhancer , without further functional tests ( e . g . , Villar et al . , 2015 ) . Of course , correlated gene expression measurements indicate that such elements are likely to be enhancers in many cases , but genomic consideration of chromatin marking must not automatically equate changes in certain active marks with enhancers . Our study provides a new perspective on these previous observations , in that essentially trivial biological interactions may have consequences in evolutionary time . We show that Hairy is engaged apparently in errant targeting of chromatin on many loci during the period when it is expressed , and demonstrate that in many cases , little distinguishes the types of chromatin effects observed on functionally repressed targets compared to ‘non-functional’ interactions on other loci ( Figure 9A , B ) . Thus , unlike an earlier model for Hairy action , in which the protein is active only when embedded in a previously active enhancer ( Nibu et al . , 2001 ) , our work demonstrates that Hairy is able to mediate biochemical activities in most bound regions , indicating that there is little context necessary for the protein to function . Therefore , Hairy may be relatively nonselective about where it can attract chromatin-modifying agents across the genome . Much molecular biology research has emphasized the high degree of cooperativity necessary for metazoan transcription factors to work well . Enhanceosomes , patterning elements and other enhancers give aberrant readouts if correct stoichiometries and spacings are not respected . These findings suggest that random individual sites are less likely to generate a suitable transcriptional readout . At least for repressors such as Hairy , the demands for generating biochemical activity are lower than anticipated , indicating that enhancers may have a lower threshold for formation that we might have expected . Although some of the targeted genes that are not transcriptionally affected may represent ectopic binding events of the induced Hairy protein , most sites are found in ChIP analysis of endogenous Hairy . The unresponsive genes may in some cases represent later targets of Hairy , may be already repressed by endogenous Hairy , or may have responses too small to measure in this system , however it is likely that there are hundreds of changed chromatin regions that not formally part of the functional Hairy GRN . Thus , a large fraction of the genomic interactions are likely to be with regions that are not strongly selected on an evolutionary timescale . As long as the induced chromatin changes are inconsequential , these effects will not be selected against during genomic evolution . This biochemical activity , however , may provide a unique molecular exaptation to generate novel edges between nodes of a standing GRN ( Figure 9B ) . Most enhancers involve the combined action of transcriptional activators and repressors , thus errant targeting may facilitate formation of new modules with gain of a few activator binding sites ( Gould and Vrba , 1982 ) ( Figure 9C ) .
The heat-inducible hairy gene was created by introducing a multiple cloning site containing Kozak sequence , initiator ATG and HindIII/BglII sites into the 5′ portion of the hairy ORF in the pCaSpeR-hsh using EcoRI/BstEII sites as described previously ( Li and Arnosti , 2011 ) . 400 bp of upstream promoter , 5′ UTR , Kozak sequence , initiator ATG , HindIII/BglII sites , coding sequence and entire hsp70 3′ UTR from the modified pCaSpeR-hsh were amplified using 5′ and 3′ primers with AgeI/KpnI sites and subcloned to the modified pattB vector ( Sayal et al . , 2011 ) . Oligonucleotides with sequence encoding the double Flag epitope , as described in Zhang and Arnosti ( 2011 ) , was inserted 5′ of the coding sequence after the ATG using HindII/BglII sites , so that Hairy protein was expressed with the double Flag tag at the N terminus . For chromatin analysis 2–3 . 5 hr embryos were collected and 20 min heat-shock treated for induction of transgenes as described previously ( Li and Arnosti , 2011 ) . We treated the wild-type embryos similar to embryos carrying inducible transgene to control for possible nonspecific effects of heat shock . Heat shock alone has no effect on the expression or chromatin patterns ( data not shown ) . For analysis of gene expression by in situ hybridization , embryos were fixed and stained using anti-digoxigenin-UTP-labeled RNA probe for ftz as described previously ( Struffi , 2004 ) . Total RNA from embryos was purified using RNeasy columns ( Qiagen ) , and reversed transcribed using a High Capacity cDNA Reverse Transcription Kit from Invitrogen/Applied Biosystems . The cDNA was then analyzed by real-time PCR using the primer pairs located at transcription units . Data was normalized to act5c . Values for wild-type embryos were set to 1; results represent the average of 2–8 biological replicates . Statistical significance was tested using Student's t-test and p < 0 . 05 . Amplicons were designed using Primer Express and Primer-BLAST . Total RNA from 2–3 hr embryos was purified using RNeasy columns ( Qiagen , Valencia , CA ) . Samples were amplified and labeled using the Quick AMP Labeling kit ( Agilent , Santa Clara , CA ) and hybridized to 8 × 15K Customized Drosophila Genome Oligo Microarrays ( Agilent ) according to the manufacturer's instructions . Slide image data was quantified using Agilent's Feature Extraction software . Four biological replicates were performed for each sample . Differential gene expression analysis was performed with the GeneSpring program ( Agilent ) . Functional annotation of down- and up-regulated genes was done using the Database for Annotation , Visualization and Integrated Discovery ( Dennis et al . , 2003 ) . Differentially regulated gene symbols and their fold changes are listed in Supplementary file 5 . Heat shocks and ChIPs were performed as described previously ( Li and Arnosti , 2011 ) , with the exceptions that embryos were sonicated for a total of 20 times using a Branson sonicator in 1 ml of sonication buffer . After precipitation of chromatin-antibody complexes , protein A beads were washed twice with low-salt buffer , once with high-salt buffer , once with LiCl buffer and twice with Tris-EDTA . We used the following antibodies: rabbit IgG ( 5 μl , Santa Cruz Biotechnology ) , rabbit anti-H3 ( 1 μl , Abcam , Cambridge , MA ) , rabbit anti-acetyl H4 ( 1 μl , Upstate , EMD Millipore , Billarica , MA ) , rabbit anti-acetyl H3K27 ( 1 μl , Abcam ) , rabbit anti-monomethyl H3K4 ( 1 μl , Abcam ) , rabbit anti-trimethyl H3K4 ( 1 μl , Abcam ) , rabbit anti-trimethyl H3K36 ( 2 μl , Abcam ) , rabbit anti-trimethyl H3K9 ( 3 μl , Abcam ) , rabbit anti-Flag ( 5 μl , Sigma-Aldrich , St . Louis , MO ) , rabbit anti-Rpb3 ( 5 μl , gift from Carla Margulies , LMU University of Munich ) . We used the differential changes of H4Ac , H3K27Ac , H3K4me1 , H3K4me3 , H3K36me3 , H3K9me3 and Pol II in response to Hairy as features for our analysis here . The genomic blocks detected as significantly altered by diffReps are annotated to closest TSS . We considered four features for each of the ChIP-seq data; number of blocks linked to the same gene , range of blocks , fold change of ChIP-seq signal at blocks , and distance of blocks to closest TSS . Four features from ChIP–chip data sets of Hairy ( MacArthur et al . , 2009 ) , CtBP , and Gro ( Nègre et al . , 2011 ) were used; number of peaks linked to the same gene , width of peaks , peak signal , and distance of peaks to closest TSS . In addition , expression of transcripts in wild-type embryos was included as a feature . In total , these 41 features were collected for 583 genes ( 241 repressed , 146 activated and 196 unaffected genes; activated and unaffected genes were grouped as nonrepressed genes ) in this study . Differentially regulated genes and their fold changes are listed in Supplementary file 5 and randomly selected unaffected genes are listed in Supplementary file 7 . Important features were first identified with four feature selection algorithms ( Information Gain , Symmetrical Uncertainty , Chi Square and Relief ) . Then , to predict genes in the repressed and nonrepressed categories , four classifiers ( Support Vector Machine ( SVM ) , k-Nearest Neighbors ( KNN ) , Naive Bayes and Random Forests ) were employed . To perform this analysis , we wrote Python and Java codes to partition our dataset into 10 parts to perform feature selection and 10-fold cross validation classification utilizing the Weka machine learning software ( http://www . cs . waikato . ac . nz/ml/weka/ ) . To increase the robustness of our results we performed 50 iterations of the above procedure and combined the predicted classes for each gene to create a new aggregate predicted class for that gene . Here we took the class that has been predicted more than 50% of the 50 iterations as the predicted class of the gene . We have applied every combination of the four feature selection algorithms and four classification algorithms to the data to obtain the optimal classification methodology for our dataset . The results of our analysis are summarized in the main text . | The genes encoded in DNA contain the instructions to make proteins and other molecules important for cell behavior . Only a fraction of genes are ‘expressed’ at any particular time; proteins called transcriptional repressors keep many in a silent state . One such repressor in the fruit fly is called Hairy , and its activity is essential for embryos to develop correctly . Similar Hairy-related proteins are crucial regulators of development in mammals . A central mechanism of controlling gene expression involves the wrapping of DNA around histone proteins to form a structure called chromatin . Attaching chemical tags to histones changes how accessible the genes are within the chromatin—the more accessible the genes are , the more likely they are to be active . Some tags promote gene activation , while other tags block expression . Previous research showed that Hairy reduces gene expression by influencing which tags are added to , or removed from , the chromatin . Kok et al . have now tracked the effects of the Hairy protein on the entire genome of Drosophila fruit fly embryos . This revealed the genes that Hairy directly targets and the corresponding effects this targeting has on chromatin structure . Hairy altered chromatin chemical tags over large blocks of DNA on silenced genes , with some of the changes being specific to particular genes . However , many areas of chromatin activity were not associated with changes in gene expression . Instead , many genes ignore Hairy-mediated changes in their vicinity , while in other cases chromatin changes occurred on genes that were already silent . Previous studies have suggested that regulatory factors like Hairy frequently bind to many sites on the genome and have no function . Kok et al . now suggest that these sites—previously regarded as representing ‘inert’ sites—are biochemically very active . Genomic studies that label regulatory sites solely by changes to their chromatin modifications may be fooled by the apparent activity of such ‘errantly targeted’ sites , assuming that they are critical for gene regulation . At the same time , these sites may represent regions that are particularly likely to evolve regulatory properties . Kok et al . therefore propose that errant targeting by Hairy may help new regulatory elements to evolve that could eventually influence how genes are expressed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"genetics",
"and",
"genomics"
] | 2015 | Genome-wide errant targeting by Hairy |
An organised spindle is crucial to the fidelity of chromosome segregation , but the relationship between spindle structure and function is not well understood in any cell type . The anaphase B spindle in fission yeast has a slender morphology and must elongate against compressive forces . This ‘pushing’ mode of chromosome transport renders the spindle susceptible to breakage , as observed in cells with a variety of defects . Here we perform electron tomographic analyses of the spindle , which suggest that it organises a limited supply of structural components to increase its compressive strength . Structural integrity is maintained throughout the spindle's fourfold elongation by organising microtubules into a rigid transverse array , preserving correct microtubule number and dynamically rescaling microtubule length .
It has long been proposed that the morphology of organisms could reflect both evolutionary design principles and underlying physical laws ( Thompson , 1942 ) . The organelle responsible for faithful segregation of the genome in eukaryotic cells , known as the mitotic spindle , exhibits tremendous morphological diversity between different cell types ( Walczak and Heald , 2008; Glotzer , 2009; Wühr et al . , 2009; Goshima and Scholey , 2010 ) . Although the mitotic spindle has been studied intensively for more than a century ( Mitchison and Salmon , 2001 ) , our understanding of the mechanisms that give rise to the remarkable precision of eukaryotic chromosome segregation remains incomplete . A challenge facing mechanistic studies of the mitotic spindle is its inherent complexity , with a large number of essential protein components ( Walczak and Heald , 2008; Glotzer , 2009 ) and an intricate , extended structure with many details that cannot be visualised by light microscopy ( McDonald et al . , 1992; Mastronarde et al . , 1993 ) . This complexity is compounded by the substantial morphological and temporal variability that often exists between different cells of the same type . Another complication is that the spindle is usually involved in performing multiple interrelated functions in the cell at any one time . During prometaphase , for example , the spindle must simultaneously capture mitotic chromosomes and form a bipolar structure ( Walczak and Heald , 2008; Glotzer , 2009 ) . We chose the anaphase B spindle in fission yeast as a model system to address the relationship between spindle form and function as it circumvents many of these conceptual difficulties . The first advantage of studying mitosis in Schizosaccharomyces pombe is that the rate of elongation and the timing of each mitotic stage are highly consistent in different cells ( Mallavarapu et al . , 1999; Fu et al . , 2009 ) . This enables the dynamics of mitosis to be related directly to the comprehensive , nano-scale reconstructions of the spindle architecture that can be acquired using electron tomography . A second advantage is that anaphase B spindle elongation represents a clearly defined morphogenetic transition that is driven by outward sliding of overlapping microtubules by plus-end directed molecular motors located at its centre ( Tolić-Nørrelykke , Sacconi , et al . , 2004b; Khodjakov et al . , 2004; Fu et al . , 2009; Glotzer , 2009 ) . Microtubule minus-ends are static and remain anchored at the spindle pole body , whilst growth of inter-polar microtubules at their plus-ends is coordinated with outward sliding to maintain an overlap at the spindle centre ( Mallavarapu et al . , 1999 ) . A final key advantage of studying the anaphase B spindle is that the forces to which it is subjected have a well-defined directionality . Spindle severance experiments have revealed that spindle elongation is powered by internal forces , and resisted by external compressive loads ( Tolić-Nørrelykke , Sacconi , et al . , 2004b; Khodjakov et al . , 2004 ) . These external loads can compromise the fidelity of chromosome segregation , as it has been observed that buckling of the spindle followed by its breakage is a common failure mode for cells that have defects in chromosome condensation , the nuclear envelope or in the organisation of spindle microtubules ( Courtheoux et al . , 2009; Khmelinskii et al . , 2009; Yam et al . , 2011; Petrova et al . , 2013 ) . In this study , we reconstruct the architecture of wild-type fission yeast spindles using electron tomography ( ET ) ( Höög et al . , 2007; Roque et al . , 2010 ) . This technique has several advantages ( Soto et al . , 1994 ) over the serial-section electron microscopy method that was used previously to determine the structures of spindles in cdc25 . 22 fission yeast and budding yeast cells ( Ding et al . , 1993; Winey et al . , 1995 ) . We use the analyses of the EM spindle reconstructions to build computational models of the spindle . These models imply that the fission yeast spindle architectures organise a limited supply of structural components to increase their resistance to compressive forces , thus demonstrating a direct link between the morphology of a mitotic spindle and its function . We also investigate the effects of external mechanical reinforcement ( Pickett-Heaps et al . , 1997; Mitchison et al . , 2005; Brangwynne et al . , 2006 ) on the forces that the spindle can bear during its elongation .
We began by using the very uniform mitotic progression of cells containing GFP-labelled tubulin ( Figure 1A; Video 1 , 2 ) to assign ET reconstructions to a specific time during mitosis . Our ET reconstructions confirmed that the spindle is composed of two opposing arrays of pole-nucleated microtubules that interdigitate at the spindle midzone ( Figure 1B; Videos 3–5 ) . A similar gross spindle organisation was observed in previous serial-section reconstructions of cdc25 . 22 fission yeast cells ( Ding et al . , 1993 ) and the related budding yeast spindle ( Winey et al . , 1995 ) . 10 . 7554/eLife . 03398 . 003Figure 1 . The Architecture and Dynamics of the Fission Yeast Spindle . ( A ) Fission yeast cell expressing GFP-labelled tubulin . The dashed red line shows cell outline at mitotic entry . Interval between frames = 1 min , scale bar = 0 . 5 μm . ( B ) ET reconstructions of three anaphase B spindles . Microtubules are coloured according to the pole from which they originate . Dashes represent the approximate locations of the cross-sections shown in C . ( C ) Longitudinal and transverse spindle architecture . Diagrams on the left show the contour length and number of microtubules within each spindle . Black arrowheads mark the three anaphase B spindles depicted in B . Circle-and-stick diagrams on the right represent cross-sections near to the poles and midzone of each spindle . Microtubule radii are drawn to scale . Black lines are drawn between neighbouring anti-parallel microtubules . ( D ) Raw electron tomographic images of second-longest anaphase B spindle shown in C . Microtubules are marked by yellow arrows . Scale bar = 25 nm . ( E ) Histograms of nearest-neighbour microtubule distances and packing angles for the three spindles shown in panel B . Blue lines represent histograms for the polar regions of the spindle and red lines the midzone . Coloured arrows mark the median of each distribution . The midzone is defined as the region of microtubule overlap that contains at least two microtubules from each pole . Kolmogorov–Smirnov tests were used to determine the probability that distributions for the polar and midzone regions were drawn from the same parent . All of the comparisons , shown here , have a p-value <10−4 . ( F ) Schematic representation of microtubule angle and distance distributions at the poles and midzone of the spindle . ( G ) Number of spindle microtubules with respect to pole-to-pole spindle length Ls . The blue error bars summarise results for spindles divided into metaphase ( Ls < 2 . 5 μm ) , early anaphase B ( 2 . 5 < Ls < 4 μm ) , and late anaphase B ( Ls > 4 μm ) stages . ( H ) Estimates of the ratio of fluorescent tubulin incorporated into the spindle for twenty mitotic cells ( grey curves ) with mean and standard deviations shown in blue . The red crosses indicate the total length of microtubules polymerised within each ET spindle . ( I ) Elongation profile for twenty wild-type spindles obtained from automatic tracking of SPBs . The traces are aligned temporally by defining time-zero as the frame at which spindles first reach a length of 3 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 00310 . 7554/eLife . 03398 . 004Figure 1—figure supplement 1 . Live-cell imaging confirms that the mass of tubulin polymerised in the spindle is conserved throughout anaphase B . Linear calibration curve for relating polymer mass to spindle intensity . Coefficient of determination R2 = 0 . 40 . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 00410 . 7554/eLife . 03398 . 005Video 1 . Spindle formation and elongation in fission yeast . Frames are shown at intervals of 1 min . The green channel shows maximum intensity projections of cells expressing GFP-tubulin ( SV40:GFP-Atb2 ) , with the magenta channel showing SPBs labelled with Cut12-tdTomato . The cut12-tdTomato images were processed using a deconvolution algorithm . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 00510 . 7554/eLife . 03398 . 006Video 2 . Summed intensity projections of GFP-tubulin in mitotic fission yeast cell augmented with tracking and segmentation results . The cell outline is shown in magenta . The magenta circles represent tracking of the SPBs from the cut12-tdTomato channel ( not shown ) . These are used to define the green box , which is used to compute spindle intensity . Tracking of the spindle intensity is ceased after the spindle elongates beyond 9 μm . At later stages of elongation more pronounced buckling of the spindle is observed , and microtubules begin to be nucleated in the vicinity of the cytokinetic ring . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 00610 . 7554/eLife . 03398 . 007Video 3 . Electron Tomogram ( ET ) reconstruction of short anaphase B spindle . from Figure 1B , top . Scale bar = 0 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 00710 . 7554/eLife . 03398 . 008Video 4 . Electron Tomogram ( ET ) reconstruction of intermediate length anaphase B spindle . from Figure 1B , middle . Scale bar = 0 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 00810 . 7554/eLife . 03398 . 009Video 5 . Electron Tomogram ( ET ) reconstruction of long anaphase B spindle . from Figure 1B , bottom . Scale bar = 0 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 009 In order to obtain estimates of the spindle's cross-sectional ( or transverse ) organisation that are independent of any spindle distortions ( Figure 1B; see subsequent discussion ) , we developed a new algorithm , termed Isotropic Fibre Tracking Analysis ( IFTA ) . This procedure enabled us to determine the spindles' contour length and to investigate the transverse organisation of microtubules ( ‘Materials and methods’ ) . A visualisation of the IFTA results ( Figure 1C ) reveals that microtubules at the spindle midzone form arrays with a degree of square-packed order from late metaphase onwards . We frequently observe square arrays containing a 3 × 3 ‘checkerboard’ arrangement of microtubules at the midzone of spindles from the end of metaphase and early in anaphase B , whilst all of the later spindles contain at least four microtubules arranged as the vertices of a square . The polar regions of the spindle are loosely packed in metaphase but become more tightly packed in anaphase B , where we observe three microtubules arranged in an equilateral triangle motif in all of the reconstructed spindles . An inspection of the raw tomographic images ( Figure 1D ) reveals good agreement with the results of IFTA , and also shows that thin filaments of electron density that bridge the spaces between microtubules are present at both the midzone and the flanking regions . An analysis of the lengths of microtubules with respect to the spindle's contour length ( Figure 1C ) shows that the microtubule overlap spans across the spindle in metaphase . In anaphase B , the midzone occupies the central portion of the spindle with a symmetric width of around 2 μm . Statistical analyses of IFTA results from anaphase B spindles ( Figure 1E ) confirm the square-packed character ( θ ∼ 90° ) with a large spanning distance ( d ∼ 40 nm ) at the spindle midzone . Conversely , denser packing ( d ∼ 30 nm ) with hexagonal character ( θ ∼ 60° ) is present in the regions that flank the midzone ( Figure 1F ) . The square packing suggests that the microtubules are bundled in an anti-parallel orientation at the spindle midzone ( Ding et al . , 1993; Janson et al . , 2007 ) ; whilst hexagonal packing in the regions closer to the spindle poles is indicative of microtubules being bundled in parallel ( McDonald et al . , 1979 ) . A prominent feature of the ET reconstructions is the relatively small number of constituent microtubules . The ET shows that the spindle contains around thirty microtubules during metaphase , and that this number is reduced dramatically by kinetochore-fibre depolymerisation in anaphase A ( Figure 1G ) . Upon entry into anaphase B , the spindle is constructed from only ten microtubules , which are then lost progressively from the spindle until only six remain prior to its disassembly . Concomitant with the changes in microtubule number , the total microtubule length polymerised within the spindle peaks at around 30 μm at the end of metaphase , but then remains roughly constant in anaphase B as the spindle length increases by a factor of four ( Figure 1H ) . This result was confirmed by live-cell imaging , which indicated that the intensity of the GFP-labelled tubulin incorporated into the spindle also plateaus shortly after the metaphase-to-anaphase transition ( Figure 1H; Figure 1—figure supplement 1; Videos 1 and 2 ) . This analysis also affirmed the highly stereotypic spindle elongation profile in fission yeast cells ( Figure 1I ) . The close agreement between the live-cell imaging and static ET enabled us to estimate the critical and total concentration of tubulin subunits in fission yeast cells by calibrating the fluorescent intensity measurements to the total tubulin polymerised in ET reconstructions of the spindle ( Table 1 ) . This technique yields an estimate for the tubulin concentration ( 4 . 3 ± 0 . 8 μM ) that is a factor of five lower than in metazoan cells ( Gard and Kirschner , 1987 ) , and is in good agreement with mass spectrometry estimates of the abundance of tubulin isoforms ( Marguerat et al . , 2012 ) . Since all microtubule dynamics occur at the plus-tips of spindle microtubules during anaphase B ( Mallavarapu et al . , 1999 ) , the conservation of polymer mass suggests that microtubule depolymerisation and recycling of tubulin subunits into the free pool ( Walker et al . , 1988 ) may be required for growth of the surviving interpolar microtubules , as suggested by previous EM reconstructions of fission yeast cells containing the cdc25 . 22 mutation ( Ding et al . , 1993 ) . Interestingly , cells from the cdc25 . 22 genetic background are enlarged upon entry into mitosis ( West et al . , 2001 ) , and can be compared with the wild-type spindle to provide insight into how the spindle's architecture is altered by increases in cytoplasmic volume . 10 . 7554/eLife . 03398 . 010Table 1 . Measurements for calculation of critical and absolute tubulin concentration in fission yeastDOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 010Cell dimensionsRadiusRc = 1 . 6 ± 0 . 1 μmMaximal cell diameter/2 ( Foethke et al . , 2009 ) LengthLc = 14 . 3 ± 0 . 9 μmDistance between cell tips at mitotic entry ( Martin and Berthelot-Grosjean , 2009; Moseley et al . , 2009 ) VolumeVc = 2πRc3 . ( Lc/2Rc−1/3 ) Vc = 106 . 4 ± 13 . 4 μm3Assumes cell is sphero-cylindricalSpindle propertiesSteady-state polymer27 . 1 ± 4 . 2 μmSum of all the microtubule's length . This studyα/β-tubulin heterodimers ( 4 . 4 ± 0 . 6 ) × 104 dimersEach MT has 13-protofilaments with a length of 8 nm ( Howard , 2001 ) Relative intensity0 . 16 ± 0 . 02Mean fluorescent intensity ratio of spindles longer than 4 μmTubulin dimer concentrationAbundance ( 20 . 0 ± 4 . 0 ) × 104Number of dimers in the cellFree pool3 . 61 ± 0 . 68 μMCorresponds to the critical concentration of MT assemblyPolymerized pool0 . 68 ± 0 . 13 μMTotal concentration4 . 30 ± 0 . 81 μM The earlier spindle reconstructions of spindles in cdc25 . 22 cells were performed using serial-section electron microscopy ( Ding et al . , 1993 ) , which involves sectioning fixed cells in a particular orientation , imaging each section using transmission electron microscopy and then recording the transverse microtubule positions ( Soto et al . , 1994 ) . These two-dimensional slices are then registered and used to approximate the full three-dimensional object . The main disadvantage of this technique is that movements of the microtubules perpendicular to the imaging plane cannot be fully accounted for , which can cause straightening of microtubules during the registration step . The technique provides accurate estimates of microtubule length and transverse organisation in linear microtubule bundles , but it may be impossible to track strongly curved microtubules , which , by definition , have an orientation that varies along their length . These difficulties are largely overcome by electron tomography where the sample is imaged at numerous orientations ( i . e . a tilt series ) with respect to the imaging plane ( Soto et al . , 1994 ) . An interesting feature of the tomographic spindle reconstructions of wild-type cells is the curvature that is present in all of the anaphase B spindles ( Figure 1B ) . These deflections are more pronounced in the early anaphase B spindles , where they appear to be inconsistent with the linear spindle morphology that can be observed in cells via light microscopy ( Mallavarapu et al . , 1999; Tolić-Nørrelykke , Sacconi , et al . , 2004b; Khodjakov et al . , 2004 ) . This suggests that the deflections are caused partially or entirely by the standard preparation methods for electron tomography ( Giddings et al . , 2001; Höög and Antony , 2007; Buser , 2010 ) , and that native spindles have a straighter morphology . In the absence of further confirmation of the degree of microtubule curvature in live cells; we developed the IFTA algorithm to trace the path of the spindle in three dimensions . This enabled us to computationally straighten the spindle , and to infer the microtubule arrangements in the plane perpendicular to the spindle's main axis , independently of any spindle curvature . Together with the microtubule lengths , this corrected information was used in the subsequent analyses . The organisation of the spindle in wild-type cells resembles the serial-section reconstructions of cdc25 . 22 fission yeast spindles in many respects ( Ding et al . , 1993 ) , but there are also several notable differences . In both genetic backgrounds , the spindle is highly symmetric from late metaphase onwards with an almost identical number of microtubules nucleated from each spindle pole ( Figure 1C ) . The total length of tubulin polymerised in spindle microtubules also appears to remain constant throughout anaphase B , in both conditions , which is accommodated by a gradual reduction in microtubule numbers . The most striking difference is that cdc25 . 22 spindles contain around twice the quantity of polymerised tubulin . The likely cause for this discrepancy is the cdc25 . 22 mutation , which causes the allele of the mitotic phosphatase cdc25p to be inactivated by high temperature ( Russell and Nurse , 1986 ) . The cdc25 . 22 allele is hypomorphic under the permissive temperatures used by Ding et al . , and consequently cells divide at lengths that are around twice that of wild-type cells ( Hagan et al . , 1990 ) . This suggests that the increased quantity of polymerised tubulin in cdc25 . 22 mutant cells arises from the increased cytoplasmic volume and the corresponding twofold increase in the abundance of tubulin and other spindle assembly factors . Other differences between the wild-type and cdc25 . 22 spindles are associated with the transverse organisation of microtubules . In particular , we have observed that , in wild-type cells , the square-packed organisation with a regular 40 nm centre-to-centre microtubule separation is established during metaphase whereas spindles from the same stage in the cdc25 . 22 mutants appear to have a less ordered midzone architecture ( Ding et al . , 1993 ) . Square packing is present at the midzone of anaphase B spindles in both genetic backgrounds , but the tight hexagonal packing nearer the poles of anaphase B spindles was not observed in cdc25 . 22 mutant cells , where a more variable transverse architecture was present . These two differences could be explained by the spindle morphology being perturbed by the increased microtubule polymer in cdc25 . 22 fission yeast cells or the technical differences in the electron microscopy . The ordered wild-type spindle architecture , the well-defined forces to which it is subjected ( Tolić-Nørrelykke , Sacconi , et al . , 2004b; Khodjakov et al . , 2004 ) and the limited quantity of tubulin subunits that appear to be available for its assembly prompted us to investigate the mechanical properties of the complex assembly of microtubules that form the spindle . The mechanical response of slender elastic beams under compression can be described using the Euler-Bernoulli beam theory ( Landau and Lifshitz , 1986 ) , which states that a beam will remain straight if subjected to forces below a certain threshold but will buckle if the critical force is exceeded . The critical force depends on the length of the beam , the elastic properties of the constituent material ( specifically , its Young's modulus , E ) and the beam's cross-section ( Figure 2 ) . The Young's modulus is a measure of the stiffness of a particular material . Intuitively , it can be thought of as the constant that relates the degree of contraction ( or strain ) in a block of material to the pressure ( or stress ) exerted on its ends . 10 . 7554/eLife . 03398 . 011Figure 2 . Effects of Transverse Organisation on the Critical Force of Prismatic Beams . ( A ) The Ι-beam is a structural element that is commonly used in civil engineering . If the beam is clamped in a horizontal position at the end furthest from view ( grey rectangle ) , and a transverse force is applied at the other end ( in this case upwards ) , then the beam's resistance is defined by the scalar transverse stiffness , EIxx , which is the product of the Young's modulus , E , and the area moment of inertia , Ixx . The area moment of inertia can be computed by dividing the cross-section into area elements , dA , and multiplying each area by the square of their distance , y , from the neutral axis ( dotted line ) . A sum or integration of this quantity can then be used to compute the area moment of inertia , Ixx . The neutral axis is perpendicular to the applied force and passes through the centre-of-mass of the beam's cross-section . Intuitively , an increased stiffness can be achieved by placing the beam's material as far from the neutral axis as possible . This property is the rationale behind the design of the I-beam , which typically bears loads in the direction given by Ixx . ( B ) A similar calculation to A can be used to calculate the beam's stiffness in an orthogonal direction . It is usually the case that Ixx > Iyy for Ι-beam designs . ( C ) The generalised response of a beam to forces in an arbitrary direction , denoted by the unit vector n , can be represented by the area moment of inertia tensor , J , with entries Ixx , Iyy , Ixy . The tensor , J , is a matrix quantity that relates the direction of the applied force to the beam's deflection ( Landau et al . , 1986 ) . For a general tensor matrix , there exist two orthogonal vectors ( known as eigenvectors ) that represent the directions of the beam's maximal and minimal stiffness . ( D ) The eigenvectors or principal axes of the Ι-beam point along the x and y-axes . ( E ) Under purely compressive forces a prismatic beam with length , L , will buckle in the direction of the most compliant principal axis , which defines the critical force , Fc , for beams of this type . ( F ) Beams with a mechanically isotropic transverse organisation , such as microtubules , have a scalar stiffness tensor ( Imin = Imax ) and degenerate principal axes . The ratio Imax/Imin ≥ 1 can be used to quantify the beam's degree of anisotropy . The ratio is one for mechanically isotropic structures such as a single microtubule or a bundle of 4 microtubules arranged in a 2 × 2 square motif . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 011 The influence of the cross-section on the beam's response to simple loads can be summarised by two orthogonal vectors associated with the scalar values Imin and Imax ( Figure 2A–D ) . These principal axes define the direction of the beam's maximal and minimal resistance to bending forces . In civil engineering , anisotropic structures such as the I-beam are typically used as horizontal supports with the stiffer axis oriented in the same direction as the major load to which the beam is subjected ( Gere and Goodno , 2012 ) . However , under purely compressive forces beams with a constant cross-section will typically buckle in the direction of the most compliant principal axis ( Figure 2E ) , and thus the minimal transverse stiffness , Imin , determines the critical force for beams of this type . It is for these reasons that the columns used as vertical supports in buildings , and perhaps also microtubules themselves , have rotational symmetry and an isotropic bending resistance ( Howard , 2001; Gere and Goodno , 2012 ) . We first considered how the transverse stiffness increases with microtubule number for idealised hexagonal and square arrays ( Figure 2F and Figure 3A , B ) . In these models , we assume that microtubules are hollow , elastic cylinders , and that cross-linkers form fixed attachments to the microtubule surface ( Figure 3A ) . We initially assume that the cross-linkers contribute a negligible density to the area moment of inertia tensor ( w = 0 in Figure 3A ) ( Claessens et al . , 2006 ) . All stiffness estimates are normalised to the flexural rigidity of a single , 13-protofilament microtubule . 10 . 7554/eLife . 03398 . 012Figure 3 . The Cross-Sectional Microtubule Organisation Enhances the Spindle's Transverse Stiffness . ( A ) Schematic representation of a pair of bundled microtubules . Microtubules are modelled as hollow cylinders that are linked by rectangular support elements with identical material properties . In all moment of inertia calculations we used r1 = 7 . 5 nm and r2 = 12 . 5 nm . ( B ) Transverse organisation of idealised hexagonal and square-packed arrays of microtubules as additional fibres are added . ( C , D ) Eigenvalues of the stiffness tensor for idealised hexagonal and square-packed microtubule arrays . Results are shown for centre-to-centre separations ( equal to h+2 r2 ) that match those obtained from ET reconstructions . These are 30 nm and 40 nm for hexagonal ( polar ) and square ( midzone ) arrays , respectively . The blue and red curves show the small and large eigenvalues of the stiffness tensor . Black arrows represent microtubule organisations with degenerate eigenvalues and isotropic stiffness . The solid black curves show the mean of the large and small eigenvalues , and the dashed black curves a quadratic fit to the mean stiffness . The y-axis ( labelled as MT units ) shows the stiffness of the bundle divided by the stiffness of one microtubule . ( E ) Minimal transverse stiffness , EImin , for three anaphase B spindles with identical units to C , D . ( F ) Stiffness anisotropy ratio , Imax/Imin , color-coded over the ET spindle reconstructions . Arrows mark the positions of the regions shown in G and H . ( G ) An apparent loss of cross-linker integrity at the pole of the spindle is a cause of stiffness anisotropy . Microtubules are drawn to scale , and are numbered consistently between slices . Arrows show the approximate extent of the bundle in two orthogonal directions . The two sections are separated by a distance of 240 nm along the spindle axis . ( H ) Transverse sections showing stiffness anisotropy at the transition from the square-packed crystalline phase to hexagonal phase . Each slice is separated by an axial distance of 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 01210 . 7554/eLife . 03398 . 013Figure 3—figure supplement 1 . Transverse Stiffness of Cross-linked Microtubule Bundles . ( A ) Minimal transverse stiffness of a cross-linked hexagonal array of microtubules . Curves reflect the small eigenvalue of the stiffness tensor for cross-linkers of increasing width . Other constants are identical to those described in panel D from the main text . Arrows mark the position of hexagonal structural motifs . ( B ) Minimal transverse stiffness of square-packed bundles of microtubules with variable cross-link width . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 013 This analysis shows that the beam's minimal transverse stiffness contains peaks corresponding to specific microtubule numbers and organisations ( Figure 3C , D ) . These structural motifs have far higher minimal transverse stiffness than arrays containing one fewer microtubule , and almost identical minimal stiffness to arrays that contain an additional microtubule . For example , removing a single microtubule from the 2 × 2 square motif in Figure 3D reduces the minimal transverse stiffness by ∼1100% while the addition of a further microtubule only gives rise to an increase of ∼3% . These large increases in the minimal stiffness correspond to mechanically isotropic organisations , where Imin = Imax . Although the cellular abundances of midzone proteins , such as ase1p and klp9p , are ∼30-fold lower than the tubulin isoforms ( Marguerat et al . , 2012 ) , these proteins have a high molecular weight and a very high local concentration at the spindle midzone in anaphase B ( Fu et al . , 2009 ) . We modelled the effects of these proteins by assuming they have identical material properties to microtubules and form dense bridges in the regions of overlap . The uncertainty in the cross-linker density is then modelled by varying the width , w , of the connections between microtubules ( Figure 3A ) . This analysis confirms that peaks in the minimal transverse stiffness are present irrespective of the cross-linker density , and are therefore associated with the geometry of square and hexagonal arrays ( Figure 3—figure supplement 1 ) . The structural motifs , such as the triangular motif , the 2 × 2 square motif and the 3 × 3 square motif that are observed in late metaphase and early anaphase B , thus appear to represent features that increase the spindle's stiffness and the force that it can tolerate before buckling . A notable feature of the stiffness of the idealised arrays , shown in Figure 3C , D , is that the mean stiffness increases more rapidly for the square-packed than for the hexagonal bundles . This leads , for example , to four hexagonally-packed microtubules having a mean stiffness of ∼21EIMT compared with the 2 × 2 motif's stiffness of ∼34EIMT . This effect has two causes with the first being the intrinsically higher hexagonal packing density , which is 2/√3 times greater than equivalent square packing . The second cause is the wider bridging distance between cross-linked microtubules at the midzone compared with the polar regions of the spindle ( Figure 1E , F ) . Both of these effects tend to increase the separation between each microtubule centre and the bundle's neutral axis and thus increase the transverse stiffness of the midzone . Since , the bridging distance between microtubules is set by the span of the cross-linking molecules , selection for increased bundle stiffness could be one of the explanations for the relatively large size of the major classes of microtubule motors within the spindle ( Carter et al . , 2011; Scholey et al . , 2014 ) . We next applied the transverse stiffness tensor calculations ( with w = 0 ) to the cross-sections of ET-reconstructed fission yeast spindles . This calculation can be used to estimate the minimal transverse stiffness ( Figure 3E ) , and the anisotropy ratio , Imax/Imin ( Figure 3F ) of the three exemplary spindles marked with arrows in Figure 1C . These maps illustrate the decrease in the transverse stiffness that is associated with elongation of the anaphase B spindle . This effect is caused by the conservation of polymer mass , which is maintained by completely depolymerising a sub-set of the interpolar microtubules from the elongating spindle . This reduction in microtubule numbers , in turn , leads to the decrease in the spindle's transverse stiffness ( Figure 3C , D ) . We also observed that the square-packed architecture , larger number of microtubules and increased bridging distance between cross-linked microtubules at the spindle midzone result in a greater transverse stiffness than in the polar regions for all phases of spindle elongation ( Figure 3E ) . The estimates of the spindle's transverse anisotropy ratio reveal that all of the anaphase B spindles have low mechanical anisotropy throughout their extent apart from localised stretches ( Figure 3F ) . These regions may be associated with local loss of cross-linker integrity in early anaphase B spindles , where recruitment of anaphase-specific microtubule bundling factors is likely to be incomplete , and at the transitions from square to hexagonal microtubule packing arrangements in later spindles ( Figure 3G , H ) . The narrow width of these transitional regions ( also known as phase boundaries ) thus appears to enable fission yeast spindles to establish mechanically isotropic microtubule organisations , throughout most of their extent . Remarkably , inspection of anaphase B spindles from cdc25 . 22 mutant fission yeast cells ( Ding et al . , 1993 ) and budding yeast ( Winey et al . , 1995 ) indicate that they also contain similar structural microtubule motifs . The enlarged cytoplasmic volumes and greater polymer mass in cdc25 . 22 cells ( Ding et al . , 1993 ) leads to spindles that retain the nine microtubules at the spindle midzone in a 3 × 3 configuration late into anaphase B . Conversely , late anaphase B spindles in budding yeast contain around 20 μm of polymer , and are often arranged with two long microtubules emanating from each pole ( Winey et al . , 1995 ) , and a 2 × 2 square-packed array in the central spindle . This suggests that the mechanisms of spindle assembly that give rise to mechanically isotropic arrangements are conserved in other yeast species , and are adaptable to the increased abundance of tubulin in enlarged fission yeast cells . As the spindle's cross-sections have striking geometric properties that appear to enhance its mechanical properties , an open question concerns whether the length and number of spindle microtubules also increase the critical force that the structure can tolerate before buckling . Since the Euler-Bernoulli beam theory only applies to prismatic beams with a constant transverse organisation , it is also unclear whether the local design rules for microtubule packing endow the non-prismatic spindles with high critical forces ( Landau et al . , 1986; Gere and Goodno , 2012 ) . To address these questions , we constructed a simplified computational model of each anaphase B spindle using the Cytosim simulation software ( Nedelec and Foethke , 2007 ) . The model spindles were then subjected to increasing compressive loads to induce buckling , which then enabled us to calculate the effective stiffness EIeff of the structure , using the relation EIeff = Fc ( Ls/π ) 2 , where Ls is the spindle length . In each simulation , the computational model incorporates the microtubule lengths determined by electron microscopy and many of S . pombe spindle's known biophysical properties ( Table 2 ) . 10 . 7554/eLife . 03398 . 014Table 2 . Physical and numerical constants for simulations of spindle stiffnessDOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 014DescriptionValueNotesGlobalkBT0 . 0042 pN . μmThermal energy at T = 27°CViscosity1 pN pN . s . μm−2Viscosity of fission yeast cytoplasm ( Tolić-Nørrelykke , Munteanu , et al . , 2004a ) Time step0 . 001 sSimulations with smaller time-steps produce similar resultsMicrotubulesSegmentation0 . 1 μmFlexural rigidity20 pN . μm2This is EIMT ( Gittes et al . , 1993 ) Steric radius30 nmMicrotubule outer radius + Debye lengthSteric stiffness200 pN . μm−1per microtubule segmentMidzone width2 . 5 μm ( Loïodice et al . , 2005; Yamashita et al . , 2005 ) Spindle Pole BodiesRadius60 nmObserved from ETDepth100 nmObserved from ETStiffness 11000 pN . μm−1Appropriate for ( Khodjakov et al . , 2004 ) ( Tolić-Nørrelykke , Sacconi , et al . , 2004b ) ( Toya et al . , 2007 ) Stiffness 220 pN . μm−1Appropriate for ( Kalinina et al . , 2013 ) Cross-linkersNumber300Less than abundance of ase1p ( ∼900 dimers/cell ) and klp9p ( ∼1300 dimers/cell ) ( Marguerat et al . , 2012 ) . Larger numbers do not alter simulation resultsBridging length50 nmApproximate centre-to-centre distance of microtubules bundled by Map65 proteins ( Subramanian et al . , 2010 ) . Link stiffness1000 pN . μm−1Force is Hookean with a non-zero resting length ( Howard , 2001 ) In most eukaryotes , the anaphase midzone is defined by the localisation of the anti-parallel bundling protein , ase1p ( Glotzer , 2009 ) , which binds strongly to the spindle at anaphase onset and occupies a region at the spindle's centre . In fission yeast , the midzone has a width that is approximately constant throughout anaphase B ( Loïodice et al . , 2005; Yamashita et al . , 2005; Janson et al . , 2007 ) , which was reproduced in the simulations of each spindle by confining cross-linkers with specificity for anti-parallel microtubules to a cylindrical region at the spindle centre with the same constant width , Lm = 2 . 5 μm . For simplicity , the model does not include binding of cross-linkers to regions flanking the midzone . After initialising a bipolar spindle with the correct gross organisation ( Figure 4A , B; Video 6 ) , the rate with which cross-linkers detach from the microtubule lattice was set to zero to approximate the slow turnover of midzone proteins ( Schuyler et al . , 2003; Loïodice et al . , 2005; Fu et al . , 2009 ) . The poles were then subjected to a linearly increasing force to probe the spindle's elastic response . Simulation results for the longest fission yeast spindle , after subjecting it to forces that increase at different rates , are shown in Figure 4C . These numerical experiments indicate that the simulated spindle withstands the increasing force until a threshold is reached and buckling occurs . All of the stochastic simulations were initialised in the same state and behaved identically until a force is applied . However , there are substantial differences in the peak force that can be sustained before the spindle buckles . This transient effect is a known signature of the buckling instability , and becomes more pronounced as spindles are compressed more rapidly . It is for this reason that the estimates of the critical force of the spindle architectures were determined by averaging the force in the final 50 s of the simulation when the system is in equilibrium . The responses from computational models of wild-type spindles , compressed at the same rate ( Figure 4D ) , shows that the critical force decreases with spindle length . 10 . 7554/eLife . 03398 . 015Figure 4 . Computational Reconstructions of the Spindle can be used to Estimate its Effective Stiffness . ( A ) Computational reconstruction of an early anaphase B spindle before and after the critical force is exceeded . The spindles are subjected to compression by attaching a spring between the spindle poles . The elastic constant is then increased to probe the spindle's response to force . ( B ) Stochastic initialisation conditions can reproduce the 3 × 3 organisation at the midzone of a spindle in early anaphase B . Microtubules are depicted with a diameter of 25 nm . ( C ) Model response of the longest fission yeast spindle to forces that increase at different rates . The SPBs are held at the spindles' contour separation for the first 50 s of the simulation , when cross-linkers are allowed to form attachments between the two halves of the bipolar spindle . At t = 50 s , the elastic constant of the spring connecting the two SPBs is then set to zero and its resting length reduced to Δ = 1 μm less than the SPB's resting separation . The elastic constant is then increased linearly over the subsequent 150 s of the simulation to exert progressively larger force on the spindle . Each colour , represented in the legend , denotes a different force increase in units of pNs−1 . The spindles bear the increasing compressive loads until a threshold is reached and the force decays before plateauing at the equilibrium ( critical ) force , Fc . In the final 50 s of the simulation ( marked quantification ) , the spring constant of the elastic element connecting the SPBs is maintained at its maximal value , and the critical force quantified by averaging the force-response profile . ( D ) Response of fission yeast spindle models to compressive forces . The contour length of each spindle is noted in the right-hand column . Curves show the median critical force from N = 100 stochastic simulations . ( E ) Dependence of critical force on spindle length for models of w . t . S . pombe , cdc25 . 22 S . pombe and Saccharomyces cerevisiae anaphase B spindles . Each point indicates the median critical force calculated from N = 100 simulations . Curves show Fc = ALs−4 fits to the simulation results , where the only fit parameter is the pre-factor , A . Inset shows normalised force Fc/A = Ls−4 for all spindle types . This rescaling highlights the universality of the relationship between spindle length and critical force . ( F ) Comparison of the critical force of fission yeast spindles ( grey histograms , N = 100 ) with a null statistical model ( blue histograms , N = 103 ) . p-values refer to the probability that a random spindle has a critical force greater than the median wild-type spindle . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 01510 . 7554/eLife . 03398 . 016Figure 4—figure supplement 1 . Compressive Strength and Optimality of Yeast Spindles . ( A , B ) Response of budding yeast and cdc25 . 22 spindle models to compressive forces . The pole-to-pole length of each spindle is noted in column to the right of each plot . ( C–E ) Comparisons of the critical force of anaphase B spindles in wild-type fission yeast , budding yeast and cdc25 . 22 fission yeast cells with a null model of each spindle architecture . The spindle length is shown in the top-left corner of each plot . The blue histograms show the critical force of 1000 realisations of the null model , in which the lengths and number of microtubules are sampled randomly . The grey histograms show the critical force estimated from N = 200 simulations of the wild-type spindle architecture with the medians represented by black arrows . p-values refer to the probability that a random spindle has a critical force greater than that of the median wild-type spindle . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 01610 . 7554/eLife . 03398 . 017Video 6 . Simulation of Spindle Subjected to Compressive Forces . In the first 50 s of the simulation , the force exerted on the poles of the spindle is zero , and cross-linkers are allowed to bind and unbind from the microtubule lattice to form attachments between the two halves of the bipolar spindle . After 50 s the force on the spindle increases linearly until buckling is induced ( Figure 4C , D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 017 A comparison of the critical forces for all of the simulations of ET-reconstructed spindles shows that it scales as Fc ∝ Ls−4 with spindle length ( Figure 4E ) . This fourth-order dependence has a contribution from the beam theory , Fc ∝ EIeffLs−2 , and another contribution from the conserved polymer mass ( Figure 1H ) . The polymer mass conservation gives rise to a quadratic decay in the spindles' effective transverse stiffness ( EIeff ∝ Ls−2 , see ‘Materials and Methods’ ) . This adverse scaling law suggests that the combination of a pushing mode of force generation with recycling of tubulin subunits and telescopic extension acts to severely limit the forces that can be generated by a spindle architecture of this type . We next considered whether the number and length of microtubules within each spindle maximise the spindles' critical force . This was addressed by constructing a model to sample alternative spindle architectures with the same polymer mass as wild-type spindles . In this model , the number of microtubules nucleated at each pole is sampled from a stochastic process and the total polymer is distributed randomly between the nucleated microtubules . This comparison reveals that the wild-type fission yeast spindles are substantially stronger than the random model ( Figure 4F ) ; a property that also applies to the cdc25 . 22 and the majority of budding yeast spindles ( Figure 4—figure supplement 1 ) . This analysis also indicates that alternative spindle architectures with a larger critical force occur infrequently under this null model , and suggests that the lengths of wild-type microtubules are regulated to increase the spindle's effective stiffness . To gain insight into the origin of the high critical force of wild-type spindles , we investigated alternative architectures that are expected to have the highest resistance to compressive forces . In spindles where anti-parallel cross-linkers are confined to a region at the spindle centre , microtubules that are too short to reach the midzone cannot form associations with a partner from the opposing SPB ( Figure 5A ) . These microtubules make only a small contribution to the stiffness of the structure , and represent an inefficient use of the fixed length of polymerised tubulin , LT , that is available for building the spindle . The span of a microtubule that projects beyond the midzone region also makes an inefficient contribution to the spindle's stiffness . Maximally efficient overlap at the spindle midzone can thus be achieved by ensuring that all microtubules terminate at the furthest edge of the midzone , and have a uniform length , LMT = ( Ls + Lm ) /2 , where Ls is the spindle length , and Lm the length of the overlap ( Figure 5B ) . The conservation of polymer mass then gives the following equation for the number of microtubules , NMMO , that are present in these maximal midzone-overlap ( MMO ) spindle architecturesNMMO=2LTLs+Lm10 . 7554/eLife . 03398 . 018Figure 5 . Fission Yeast Spindles Scale Microtubule Lengths and Number for Maximal Overlap at the Spindle Midzone . ( A ) Schematic representation of a fission yeast spindle of length Ls . Green lines represent microtubules and magenta lines anti-parallel cross-linkers , which bind to the midzone region of width Lm . Blue shaded ellipses represent the spindle pole bodies ( SPBs ) . ( B ) Maximal anti-parallel bundling of microtubules at the spindle midzone is achieved if microtubules have a monodisperse length LMT = ( Ls + Lm ) /2 . ( C ) Number of interpolar microtubules in wild-type and cdc25 . 22 mutant fission yeast cells with respect to spindle length ( blue and red diamonds , respectively ) . Trend lines show theoretical calculation of microtubule number in MMO spindle architecture based on the spindle length , Ls , and average total polymer , LT . For wild-type cells LT = 27 . 1 S . D . 4 . 2 μm ( blue curve with standard deviation represented by grey shaded area ) whilst for cdc25 . 22 cells LT = 61 . 2 S . D . 7 . 8 μm ( red curve ) . ( D ) Comparison of interpolar microtubule number with predictions of the MMO model . The black line shows trend for exact agreement with grey bars representing 1 microtubule deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 01810 . 7554/eLife . 03398 . 019Figure 5—figure supplement 1 . Critical force of Wild-type and Cdc25 . 22 Mutant Fission Yeast Cells . ( A ) Critical force for simulations of wild-type spindles compared with MMO spindles with the same quantity of polymer mass . Points reflect the median from 200 stochastic simulations of each condition . ( B ) Ratio of median critical force of wild-type spindle architectures compared with the distribution of critical forces for MMO spindles . Error bars represent a single standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 019 Simulations probing the force-response of MMO spindles reveal critical forces that are , on average , only 8% higher than those measured for reconstructed wild-type and cdc25 . 22 spindles ( Figure 5—figure supplement 1 ) . This suggests that fission yeast cells exert precise global control on microtubule number and length . A comparison of the observed number of interpolar microtubules in wild-type spindles , Nobs , with the theoretical prediction from equation 1 ( Figure 5C , blue curve ) confirms that microtubule numbers are under precise control in fission yeast . This curve also indicates that the midzone width and total length of polymerised tubulin in wild-type cells enable the spindle to conform to the first 3 × 3 midzone organisation at the start of anaphase B where spindle-derived forces must remove residual catenation between the chromosome arms ( Renshaw et al . , 2010; Petrova et al . , 2013 ) and remodel the nuclear envelope ( Yam et al . , 2011 ) . The 2 × 2 motif can also be maintained throughout spindle elongation up to lengths of around 10 μm ( Figure 5C ) , which coincides with the length at which GFP-tubulin spindles are observed to begin disassembly ( Figure 1A; Videos 1 , 2 ) . Remarkably , the theoretical curve for microtubule number in cdc25 . 22 spindles containing around twice the mass of polymerised tubulin ( Figure 5C , red curve ) , is also in very close agreement with observed number of interpolar microtubules . A direct comparison between Nobs and NMMO for wild-type and cdc25 . 22 spindles ( Figure 5D ) reveals that all but four of the fourteen spindles deviate from the theoretical model by less than a single microtubule , thus indicating precise scaling of microtubule number to changes in spindle length and the availability of polymerised tubulin . Whilst microtubule number and length appear to be regulated precisely in the fission yeast spindle , the critical force of the longer spindles , at around 10 pN ( Figure 4E , F ) , is relatively small compared with other intracellular forces . For example , whilst this force exceeds the viscous drag on daughter nuclei in wild-type cells ( see ‘Materials and methods’ ) it is comparable to the force generated by two kinesin motors ( Gittes et al . , 1996; Visscher et al . , 1999 ) . The late anaphase B spindle would therefore appear to have an inefficient architecture for generating forces to segregate lagging or concatenated chromosomes ( Pidoux et al . , 2000; Courtheoux et al . , 2009; Petrova et al . , 2013 ) . We therefore investigated whether external reinforcement ( Brangwynne et al . , 2006 ) of the spindle might lead to an increase in the forces that the interpolar microtubules can support under these conditions . In simulations of the spindle , exceeding the critical forces leads to the spindle being buckled into a shape with a single maximum , which we refer to as primary-mode buckling ( Figure 4A ) . Whilst this behaviour is predicted by the Euler-Bernoulli beam theory ( Landau et al . , 1986 ) , the theory also predicts that the beam can buckle with shorter wavelengths . However , since these higher-order modes ( i . e . with two or more maxima ) occur at higher critical forces ( and energy ) than buckling in the primary mode , they are expected to relax into the primary mode over longer time scales . It has , however , been shown , using experiments on cells and macroscopic analogues ( Brangwynne et al . , 2006 ) , that laterally reinforcing microtubules with a dense elastic meshwork can lead to stable buckling over much shorter wavelengths . For microtubules subjected to an elastic confinement , the critical force that can be supported is ( Brangwynne et al . , 2006 ) fc=8π2EIλ2which is similar in form to the classical Euler-Bernoulli formula but , in this case , the critical force depends on the buckling wavelength , λ , rather than the length of the beam . The buckling wavelength is determined by the transverse stiffness of the fibre and a parameter , α , which measures the strength of the elastic confinementλ=2π ( EIα ) 1/4 We investigated this effect by imposing an elastic confinement on the spindles from early , intermediate and late stages of elongation ( Figure 6A–C; Videos 7–9 ) . As before , the spindles were subjected to increasing compressive loads until they underwent buckling . A visualisation of the shape adopted by the shortest anaphase B ( Figure 6A ) reveals that its profile transitions from buckling in the primary mode when unreinforced ( α = 0 ) to second-order buckling for an elastic confinement α = 40 Pa , with intermediate configurations containing both components ( α = 20 Pa ) . A similar transition occurs for longer spindles ( Figure 6B , C ) but with lower elastic confinement . 10 . 7554/eLife . 03398 . 020Figure 6 . Elastic Reinforcement Enhances the Compressive Loads that can be Borne by the Spindle . ( A–C ) Equilibrium shapes adopted by exemplary anaphase B spindles after the critical force is exceeded . The degree of lateral reinforcement , α , is shown alongside each simulation . Scale bars = 1 μm . ( D ) Force-response of reconstructed anaphase B spindles ( median critical force calculated from the results of 50 stochastic simulations ) . ( E ) Relationship between the critical force of laterally reinforced spindles and their length . Lines reflect power law fits Fc = ALp , with exponent , p , shown to the right of each curve . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 02010 . 7554/eLife . 03398 . 021Video 7 . Simulation of Short , Intermediate and Long Example Spindles Subjected to Compressive Forces with Differing Degrees of Lateral Reinforcement . Scale is set by the size of the enclosing cell , which measures 11 μm in length between the opposite cell tips . From the image at the top downwards , the spindles are subjected to increasing confinement with α = 0 , 20 , 40 , 200 Pa , respectively for the short and intermediate length spindles . For the longest spindle , α = 0 , 5 , 40 , 200 Pa . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 02110 . 7554/eLife . 03398 . 022Video 8 . Simulation of Short , Intermediate and Long Example Spindles Subjected to Compressive Forces with Differing Degrees of Lateral Reinforcement . Scale is set by the size of the enclosing cell , which measures 11 μm in length between the opposite cell tips . From the image at the top downwards , the spindles are subjected to increasing confinement with α = 0 , 20 , 40 , 200 Pa , respectively for the short and intermediate length spindles . For the longest spindle , α = 0 , 5 , 40 , 200 Pa . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 02210 . 7554/eLife . 03398 . 023Video 9 . Simulation of Short , Intermediate and Long Example Spindles Subjected to Compressive Forces with Differing Degrees of Lateral Reinforcement . Scale is set by the size of the enclosing cell , which measures 11 μm in length between the opposite cell tips . From the image at the top downwards , the spindles are subjected to increasing confinement with α = 0 , 20 , 40 , 200 Pa , respectively for the short and intermediate length spindles . For the longest spindle , α = 0 , 5 , 40 , 200 Pa . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 023 A comparison of the force-response curves for all of the simulated wild-type spindle architectures ( Figure 6D ) shows that the critical force that can be withstood increases with greater elastic confinement . This effect is most dramatic for the longest anaphase B spindles where large elastic confinements can increase the critical forces by around two orders of magnitude . This effect is caused by the critical force of laterally reinforced spindles decreasing much more slowly with increasing spindle length ( Figure 6E ) . These results are in good agreement with the theory for reinforced elastic rods ( Brangwynne et al . , 2006 ) , which predicts a critical force , Fc∝EI . In the case of the spindle , this leads to a Fc ∝ Ls−1 dependence , which is caused by the reduction in the effective transverse stiffness ( EIeff ∝ Ls−2 ) that occurs as a consequence of spindle elongation . These results suggest that a small degree of lateral reinforcement , compared with the ∼1 kPa found in interphase mammalian cells ( Brangwynne et al . , 2006 ) , could greatly increase the forces that the anaphase B spindle can sustain . The scaling law relating the critical force to the length of a reinforced spindle could also facilitate its elongation in enlarged fission yeast cells .
In the case of the yeast spindle , the relatively high abundance of midzone proteins ( Fu et al . , 2009; Marguerat et al . , 2012 ) and the force-resistant , non-slip binding of kinesin molecules to the microtubule lattice ( Gittes et al . , 1996; Visscher et al . , 1999 ) suggest that the spindle is densely cross-linked at the midzone . Furthermore , the dramatic remodelling of the spindle's transverse architecture at the boundary between the midzone and flanking regions of the spindle ( Figure 3G , H ) further suggests that the cross-linkers that bind to the polar regions also form rigid bridges between pairs of microtubules . Finally , the slow but uniform rate of spindle elongation ( Figure 1I ) suggests that the midzone-bound motors operate in a regime that is far from their stall force . In simulations of the anaphase B spindle , introducing a sufficient number of crosslinkers leads to reduced movement of the filaments at the overlap zone , and causes this part of the spindle to behave elastically ( Claessens et al . , 2006 ) . An important distinction between the spindle and macroscopic machines is that it must self-organise from local interactions between protein components that are many orders of magnitude smaller in size . This necessitates the use of chemical and physical properties of these proteins to generate a particular microtubule organisation . The key feature that endows the yeast spindle with rigidity is the crystalline microtubule architecture that is present throughout its extent , and which takes the form of hexagonal packing in the polar regions of the spindle and square-packing at the spindle midzone . The advantage of crystals over more disordered microtubule packing arrangements is that the number of interactions between a microtubule and its neighbours is increased ( Figure 7A ) . This increased connectivity enhances the spindle's stiffness by ensuring that transverse rearrangements of microtubules that lead to bundle warping are prevented , and that the structural integrity of the bundle is thus maintained . Arranging the crystalline units in each cross-section with rotational symmetry enhances the minimal transverse stiffness still further . This combination produces stiff microtubule arrays with isotropic responses to force; similarly to the designs for load-bearing columns used in civil engineering since antiquity ( Thompson , 1942; Gere and Goodno , 2012 ) . 10 . 7554/eLife . 03398 . 024Figure 7 . Spindle Resistance To Compressive Forces . ( A ) Applying compressive force to a sparsely connected microtubule bundle leads to warping . This increases bundle anisotropy and leads to a reduction in the critical force . In a crystalline microtubule array , microtubules can associate with a larger number of neighbouring microtubules . This constrains microtubule movement and increases the transverse stiffness . Arranging the crystal unit cells with rotational symmetry leads to architectures with an isotropic resistance to bending forces and an increased minimal transverse stiffness . ( B ) In late anaphase B , the forces driving spindle elongation are balanced by the viscoelastic response of the cytoplasm and an elastic force . After ablation of the spindle midzone , this elastic force is opposed by the viscous drag . DOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 024 Similar bundle design principles have also been observed in horseshoe sperm acrosomes , which are long , finger-like actin projections that enable the sperm cell to penetrate the 30 μm thick jelly coat that surrounds the egg ( Schmid et al . , 2004; Shin et al . , 2004 ) . These dense , highly cross-linked , crystalline actin bundles are likely to have been selected for resistance to compressive forces ( Schmid et al . , 2004 ) , and the composite structure has a Young's modulus that is similar to that of the constituent actin filaments ( Shin et al . , 2004 ) . The convergent evolution of a crystalline architecture in cytoskeletal bundles that resist compressive forces may be one factor that has led to the distinct transverse arrangements that are observed in different populations of spindle microtubules ( McDonald et al . , 1979 , 1992; Mastronarde et al . , 1993 ) and may reflect the relative contributions of tensile and compressive forces in each system ( Wühr et al . , 2009; Goshima and Scholey , 2010 ) . Two additional features of yeast spindles that also make contributions to their high critical forces are the control of microtubule length and number . We have provided evidence that the regulation of these properties enables the fission yeast spindle to use the available microtubule polymer to form specific square-packed midzone motifs at critical points during mitosis ( Figure 1C; Figure 5C ) . This property also applies , albeit less precisely , to the budding yeast spindle , where a lower midzone width ( Lm ≈ 2 μm , ( Schuyler et al . , 2003 ) ) enables the reduced quantity of polymerised tubulin ( LT = 17 . 9 ± 6 . 5 , ( Winey et al . , 1995 ) ) to be used to form 3 × 3 square-packed motifs at the onset of anaphase , and to maintain 2 × 2 motifs up to a length of around 8 μm . It remains to be seen whether a similar pattern , with coevolution between microtubule polymer mass , midzone width and final spindle extension , is present in other ascomycete yeast species with a similar spindle organisation ( Horio and Oakley , 2005; Roca et al . , 2010 ) . It will also be interesting to investigate how motors ( Fu et al . , 2009 ) , cross-linking molecules ( Janson et al . , 2007 ) and regulators of microtubule stability ( Bratman and Chang , 2007 ) collaborate to form cytoskeletal bundles with this remarkable degree of precision . A surprising feature of the spindle is that the regulation of its architecture is maintained in enlarged cdc25 . 22 cells . This suggests that the mechanisms of spindle assembly can adapt its morphology to changes in cell size . This adaptive control may be required to enable the fully elongated spindle to reach the cell poles in fission yeast cells of variable size ( Hagan et al . , 1990 ) . This is a property that fission yeast shares with metazoan embryos , which contain spindles that are observed to scale in size with cytoplasmic volume during early development ( Good et al . , 2013; Hazel et al . , 2013 ) . However , the evolutionary pressure that has led to this property being selected in unicellular yeast cells is less obvious . One possible explanation is that the spindle has evolved to be robust to natural variations in the size of the cell at mitotic entry ( Martin and Berthelot-Grosjean , 2009; Moseley et al . , 2009 ) or noise in the abundance of different spindle assembly factors ( Kaern et al . , 2005 ) . Another possibility is that spindle scaling has been selected to respond to differentiation of yeast cells into the alternate forms that occur during mating or hyphal growth ( Niki , 2014 ) . A vital consideration in the design of engineered systems is tolerance , which specifies the range of performance that is required from a component or device . In studies of spermatocyte spindles and related systems ( Nicklas , 1988 ) , it was found that the maximal force that can be exerted on chromosomes is typically several orders of magnitude larger than that required to overcome viscous drag from the surrounding cytoplasm . The forces that are applied to chromosomes are likely to vary according to the phase of mitosis and the state of the attachments between kinetochores and the spindle . However , direct measurements of these forces have only been possible in a small number of cell types ( Nicklas , 1983 ) . In vitro studies of purified budding yeast kinetochores ( Akiyoshi et al . , 2010 ) found that the tensile force exerted on a single microtubule influences the life-time of its association with the kinetochore . The most long-lived kinetochore-microtubule attachments take place under loads of around 5 pN , with weaker or stronger forces leading to increased rates of detachment ( Akiyoshi et al . , 2010 ) . This behaviour is consistent with the in vivo regulation of kinetochore attachments , where tension is used to detect chromosomes that are properly bi-oriented on the metaphase spindle ( Bouck et al . , 2008; Tanaka , 2010 ) . As the positions of the 16 kinetochore pairs in budding yeast are highly consistent during metaphase ( Joglekar et al . , 2009 ) , when microtubule attachments are at their most long-lived ( Tanaka , 2010 ) , this suggests that the total tensile force on the budding yeast spindle could be around 80 pN at the end of metaphase . Fission yeast spindles have complex centromeres , which more closely resemble those of higher eukaryotes ( Malik and Henikoff , 2009 ) , with around three microtubules occupying each kinetochore ( Ding et al . , 1993 ) compared with the single microtubule in budding yeast ( Tanaka , 2010 ) . The coordination between microtubules within composite kinetochore-fibres is currently unknown but extrapolating the in vitro estimates from budding yeast to the three chromosomes in fission yeast would suggest a force in the range of ∼50 pN . This value is well below the critical force of unsupported early anaphase spindles , and suggests that shorter , stiffer metaphase spindles ( Figure 1C ) can easily generate the required forces without external mechanical reinforcement . The forces that are exerted on the anaphase B spindle have a well-defined directionality ( Tolić-Nørrelykke , Sacconi , et al . , 2004b; Khodjakov et al . , 2004 ) , but their magnitude at the critical early stages of elongation is unknown due to the complex organisation of chromatin around the spindle ( Stephens et al . , 2013 ) and the need for re-modelling of the nuclear envelope ( Yam et al . , 2011 ) . However , in later anaphase B , after the nuclear envelope splits into a ‘dumb-bell’ morphology , the major external loads opposing spindle elongation are likely to arise from cytoplasmic resistance to the movement of daughter nuclei ( Lim et al . , 2007 ) . A rough estimate for the compressive forces acting on the spindle at this stage of mitosis can thus be estimated by treating the cytoplasm as a viscoelastic fluid . The viscous drag exerted on the poles of the spindle can be calculated using various properties of the fission yeast cell ( Foethke et al . , 2009 ) ( Table 3 ) , whilst the relaxation of nuclei in ablated spindles ( Khodjakov et al . , 2004 ) can be used to determine the relative magnitude of the viscous and elastic forces ( Figure 7B ) . The final estimate for the total resistive force acting on the poles of the elongating spindle is subject to a large degree of uncertainty , but is of the same order of magnitude ( ∼1–10 pN ) as the critical force of longer anaphase B spindles . This suggests that the wild-type spindle can support the drag forces that resist its elongation under normal conditions . However , the spindle may be subjected to larger forces if chromosomes are concatenated or are not properly bi-oriented in metaphase ( Pidoux et al . , 2000; Courtheoux et al . , 2009; Petrova et al . , 2013 ) . 10 . 7554/eLife . 03398 . 025Table 3 . Estimates of forces resisting spindle elongationDOI: http://dx . doi . org/10 . 7554/eLife . 03398 . 025DescriptionVariableValueNotesCell radiusrcell1 . 6 μm ( Foethke et al . , 2009 ) Daughter nuclei radiusrnucleus0 . 8–1 . 0 μmThis study ( ET estimate ) −21/3 × radius of interphase nucleus ( Foethke et al . , 2009 ) Cytoplasmic viscosityηcell1 pN . s . μm−2 ( Tolić-Nørrelykke , Munteanu , et al . , 2004a ) Spindle pole separation speedvs0 . 9 μm . min−1This studyDrag coefficientγ25-110 pN . μm−1 . sSee ‘Materials and methods’Relaxation timeτ = γ/k9 . 7 sExponential fit to collapse of long spindle ( Khodjakov et al . , 2004 ) , Figure 7BEquilibrium positionx01 . 4 μmSecond fit parameterDrag forceFv = γ . vs/20 . 2–0 . 8 pNElastic forceFe = x0 . γ/ τ3–16 pNThis force is dramatically decreased after the nucleus adopts a dumbbell configurationTotal forceFe + Fv4–17 pN Under increased loads , the klp9p motors that bind to the spindle midzone ( Visscher et al . , 1999; Fu et al . , 2009; Marguerat et al . , 2012 ) could theoretically generate up to ∼103 pN , while each depolymerising microtubule , of which there are around nine in fission yeast cells during anaphase A ( Ding et al . , 1993 ) , can produce a maximal force of ∼40 pN ( Grishchuk et al . , 2005 ) . A possible mechanism to ensure tolerance of the anaphase B spindle against chromosome segregation errors would be to externally reinforce the spindle with an elastic material ( Mitchison et al . , 2005; Brangwynne et al . , 2006 ) . It remains to be determined whether the anaphase B spindle is reinforced , but candidates for providing this mechanical support have been identified and include actin , which is required for accurate chromosome segregation in fission yeast ( Gachet et al . , 2001; Meadows and Millar , 2008 ) , the nuclear envelope ( Lim et al . , 2007; Yam et al . , 2011 ) and mitotic chromosomes themselves , which surround the interpolar microtubule axis of yeast spindles ( Stephens et al . , 2013 ) . Whilst this study has provided evidence that the architecture of the beam-like fission yeast spindle is sculpted for the generation of pushing forces , a more general question concerns why this linear morphology was selected in yeast cells compared with the more conventional ellipsoid spindles observed in metazoans and many other eukaryotes . One possible functional explanation for the change in spindle morphology is that fungal cells have evolved a stiff polysaccharide cell wall that encloses the plasma membrane ( Bowman and Free , 2006 ) . This innovation could then have reduced the cell's dependence on the cytoskeleton for providing motility and structural support , and enabled reduced expression of the tubulin and actin isoforms ( Pollard et al . , 2000; Marguerat et al . , 2012 ) . In yeast cells with a small cytoplasmic volume , the beam-like spindle morphology may have been selected to segregate chromosomes with reduced use of structural components and their associated metabolic cost to the cell . In this respect , the yeast spindle is indeed remarkably efficient , as chromosomes are transported over similar distances to those in many somatic mammalian cell types ( Goshima and Scholey , 2010 ) , but with a structure that has a 1000-fold smaller volume . The mitotic spindle displays a diversity of sizes and shapes across different cell types . We have identified structural features of the fission yeast spindle that enable it to use a small quantity of raw material to exert large forces that drive the segregation of chromosomes . This represents a first step towards an understanding of the relationship between the mechanisms of force generation and the architecture of the cell division machinery in eukaryotic cells .
Standard S pombe genetic and molecular biology techniques , including media , were used as described in the ‘Nurse Lab Manual’ ( Roque et al . , 2010 ) . Cells were grown and imaged in minimal medium EMM2 supplemented with amino acids where necessary ( final concentration of 250 mg/l ) . For image acquisition , cells were collected and placed in a 35 mm glass-bottom culture dish ( MatTek Corp . , Ashland , MA , USA ) coated with lectin ( from Bandeiraea simplicifolia , Sigma L2380 ) . Cells were further immobilised with a 2% agarose/EMM2-coated coverslip , which was attached to the glass-bottom dish using VALAP . The partially sealed coverslip was submerged in liquid EMM2 media to prevent drying and deoxygenation of the growth media . Confocal images were obtained with a Carl Zeiss Axiovert 200 M microscope equipped with a PerkinElmer RS Dual spinning disc system . The Argon Krypton line laser was used at wavelength of 488 nm for GFP signal detection . Images were collected using a 63× oil immersion objective ( Plan-Apochromat , NA 1 . 4 ) coupled to a Hamamatsu C9100-50 EMCCD camera ( Hamamatsu , Japan ) with a pixel size of 8 μm . The microscope was controlled with the Ultraview acquisition software ( Perkin Elmer , Foster City , CA US ) . All experiments were performed inside a climate control box ( EMBL , Heidelberg , Germany ) at a constant temperature of 28°C . Live-cell imaging of fission yeast spindles was carried out using spinning disk confocal microscopy from 15 focal planes separated by a distance of 0 . 5 μm , a set-up similar to that used to quantify the cellular abundance of proteins localised to the cytokinetic ring in fission yeast ( Wu and Pollard , 2005 ) . A maximum likelihood image deconvolution algorithm was applied to the images to improve the signal-to-noise ratio of both the tubulin fluorescent marker ( Bratman and Chang , 2007 ) and the spindle pole body ( SPB ) marked with cut12-tdTomato ( Samejima et al . , 2010 ) . After deconvolution , maximum-intensity projections of the SPB marker and summed-projections of the GFP-tubulin marker were used to track the poles of the spindle and to segment the cytoplasmic volume , respectively , but were not used to quantify the intensity of either the spindle or cytoplasm . The segmentation of the cells was performed using a three-dimensional region-growing algorithm ( Gonzalez , 2010 ) . A seed-point in the cellular background was selected and propagated to the entire image stack , using a threshold of 30% of the mean region intensity as the voxel inclusion criterion . Erosion and dilation operations were then used to remove isolated bright pixels and smooth the cellular outlines . Apart from deconvolution , all analyses of images were carried out using Matlab ( The MathWorks inc . ) . Scripts for performing image analysis are provided in Source Code 1 . The tracking of SPB positions was performed using the μ-Track image analysis package ( Jaqaman et al . , 2008 ) . The splitting of a single SPB into two distinct and persistent tracks was used to identify mitotic events in a field of asynchronously dividing cells . Each mitotic event was then mapped to a single segmented cell that was used in subsequent analyses . The background signal was estimated by fitting a histogram to the pixel intensities in each field of view . The background signal was subtracted from the pixel intensity measurements to obtain an absolute estimate of fluorescent intensity . For each cell in mitosis , the background-corrected pixel intensity was integrated over the cytoplasmic volume to obtain a measurement of the total fluorescent intensity within the cell . Large reductions in the integrated cellular intensity were used to diagnose cells that moved significantly during the acquisition period . All acquired cells were also checked by manual inspection of the images augmented with the segmentation and tracking results . The intensity of GFP-labelled spindle was determined by considering a rectangular window centred on the spindle with a length defined by the positions of the two SPBs . The width of the window was 1 . 02 microns ( 8 pixels ) , which is sufficient to collect the light diffracted by the objective lens ( λ ≈ 200 nm ) . Increasing the width of the window to as much as 1 . 53 microns did not lead to a qualitative change in the results . Pixel intensity values within the window were determined by linear interpolation to account for rotation and translation of the pixel positions with respect to the sensor array . These intensity values were corrected for the local cytoplasmic fluorescent intensity , and summed to obtain an estimate of the fraction of the cellular fluorescence that is associated with spindle microtubules . In order to calibrate the spindle intensities to the tubulin polymerised in microtubules , the polymer within each ET spindle was compared to the intensities of the mitotic cells where pole-to-pole length matches that of the ET spindle most closely ( Figure 1—figure supplement 1 ) . This relationship was fit with a linear function a . x + b . The small value of b , and the relatively high coefficient of determination support a linear scaling relationship between fluorescent intensity and polymer mass . Estimates of polymer mass obtained using electron tomography and live-cell imaging indicate that the tubulin within the spindle plateaus in early anaphase . This implies that the transverse density of microtubules within the anaphase B spindle decreases linearly with the length . Since , the transverse spindle width is below the resolving power of conventional epifluorescence microscopy , a useful proxy for tubulin density is the spindle's average fluorescent intensity , which indeed declines linearly with spindle length during anaphase B ( Figure 1A ) . The correspondence between fluorescent tubulin intensity and the mass of tubulin polymerised in the spindle can also be used to calculate the abundance and critical concentration of tubulin subunits within the fission yeast cell ( Table 1 ) . This calculation uses the well-defined structure of microtubules ( Howard , 2001 ) to estimate the total number of tubulin subunits within the spindle . The intensity ratio between GFP-labelled tubulin in the spindle and cytoplasm can then be used to extrapolate this quantity to the abundance of tubulin subunits in the cytoplasm . Finally , the regular spherocylindrical shape of fission yeast cells , with a uniform radius and a well-defined length upon entry into mitosis , enables accurate estimates of the intracellular volume and thus concentration to be obtained . The estimates of tubulin abundance are in good agreement with mass-spectrometry studies ( Marguerat et al . , 2012 ) , and yield an estimate for the critical tubulin concentration in fission yeast that is around 25% lower than the 4 . 9 ± 1 . 6 μM measured from in vitro tubulin polymerisation experiments ( Walker et al . , 1988; Janson and Dogterom , 2004 ) . Yeast samples were prepared as in ( Roque et al . , 2010 ) . Briefly , log yeast cultures were high pressure frozen using an EMPATC 2 ( Leica Microsystems , Wetzlar , Germany ) and fixed by freeze substitution with 0 . 1% dehydrated glutaraldehyde , 0 . 25% uranyl acetate and 0 . 01% osmium tretraoxide in dry-acetone . Freeze substitution was performed at −90°C for 56 hr after which the temperature was raised to −45°C in steps of 5°C/hr . Several washes of 15 m in dry-acetone were followed by lowicryl resin infiltration at −45°C in graded steps of 3:1 , 1:1 , 1:3 acetone:HM20 resin ( EMS , cat . 14 , 340 ) for 1 hr each followed by 3 steps of 2 hr in 100% HM20 resin . After 100% HM20 over-night , fresh resin was added and polymerisation by UV light initiated while the samples were at −45°C and carried on while the samples were raised to 20°C in steps of 5°C/hr . The samples were then kept for 24 hr more at 20°C under UV light . Tilt-series of relevant areas were acquired from ±60° at 1° intervals in a Tecnai F30/F20/T12 ( FEI , Oregon USA ) at the magnification of 15 , 500×/14 , 500×/11 , 000× , respectively using the SerialEM software . Tilt-series were reconstructed , joined in the cases where spindles spanned several adjacent subsections , and tracked manually using the software package IMOD ( Kremer et al . , 1996 ) . In the electron tomograms , microtubules appear as cylindrical tubes with a diameter of around 18 nm compared with the 25 nm expected for 13-protofilament microtubules . The microtubule shrinkage is caused by the freeze-substitution methods used to fix the cells , but other aspects of the microtubule structure appeared unperturbed . All measurements were thus scaled uniformly by γ = 25/18 to obtain more accurate estimates of the spindles' physical properties . The granularity of the isotropic fibre tracking analysis ( IFTA ) solution is controlled by a single parameter , Rs , which specifies the separation of neighbouring coordinates that define the axis of the spindle in three-dimensions . This parameter is set to 200 nm for tracking anaphase B spindles in fission yeast and at 100 nm for computing the histograms of nearest-neighbour distances and microtubule packing angles . The IFTA algorithm proceeds by first detecting the pole with largest number of microtubule minus-ends that lie within the threshold distance , Rs . The centre-of-mass of these ends is used as the first point defining the spindle . A sphere with a radius , Rs , is then centred at the pole , and used to detect the positions where microtubules intersect the ball's surface . A constrained optimisation calculation is then used to determine the point lying on the sphere's surface that minimises the mean-squared distance from the microtubule crossings . This point is selected as the second position in the three-dimensional interpolation of the spindle axis whilst microtubule coordinates that lie within the sphere are masked . This procedure is repeated until the spindle coordinates reach the opposite spindle pole . The transverse organisation of the spindle is then determined by detecting the positions where microtubules intersect a plane halfway between and perpendicular to the points that define the spindle axis . The calculation of the transverse stiffness is made by assuming that microtubules are uniform , hollow cylinders and that the cross-linkers are rectangular support elements with identical material properties . The parameters are taken from the known dimensions of microtubules and the microtubule numbers and organisation determined in this study . For a slender , elastic beam with length , L , constructed from a material with Young's modulus E , the compressive force that can be sustained before buckling is given by F = π2EI/L2 , provided that the two ends of the beam are allowed to pivot freely ( Landau et al . , 1986 ) , as is likely to be the case for the fission yeast spindle ( Tolić-Nørrelykke , Sacconi , et al . , 2004b; Kalinina et al . , 2013 ) . Alternative boundary conditions lead the critical force to be altered by a multiplicative constant . The third property of the beam that determines the critical force is the area moment of inertia ( Figure 2 ) . The contribution that a single microtubule with its centre a distance , y , from the neutral axis makes to the bundle's area moment of inertia in the y-direction , Ixx , is given by the following equation ( 2 ) Ixx=∫ area ( y+rsinθ ) 2dA=∫r1r2[y2+r2sin2θ+2yrsinθ]rdrdθ=πy2 ( r22−r12 ) +π4 ( r24−r14 ) =y2AMT+IMTwhere r is a radial coordinate representing distance from the microtubule axis , and r2 and r1 are the outer and inner radii of the cylindrical microtubule . The term AMT is the microtubule cross-sectional area and IMT is the ( isotropic ) moment of inertia of a single microtubule about its axis . A symmetric expression exists for the moment of inertia in the x-direction , and the product moment of inertia Ixy , is given by ( 3 ) Ixy=−∫ area ( y+rsinθ ) ( x+rcosθ ) dA=−πxy ( r22−r12 ) =−xyAMT The components of the moment of inertia tensor can be obtained by summing the contributions of individual microtubules . The position of the neutral axis about which the bundle of microtubules will undergo bending is given by the centre of mass of the microtubule centres , which gives the following expression for a bundle containing N microtubules with centres at ( xi , yi ) with a mean positions ( x¯ , y¯ ) in the plane perpendicular to the bundle ( 4 ) Ixx=NIMT+AMT∑i=1N ( yi−y¯ ) 2Ixy=−AMT∑i=1N ( yi−y¯ ) ( xi−x¯ ) The formula for the area moment of inertia tensor , J , can thus be written compactly in matrix notation as ( 5 ) J= ( IxxIxyIxyIyy ) =N[IMTI2+AMT ( tr ( Σ ) I2−Σ ) ]where Ι2 is the 2 × 2 identity matrix , tr ( . ) refers to the matrix trace operator and Σ is the covariance matrix of the microtubule centres in the xy-plane . A similar expression can be obtained for a composite beam containing microtubules and cross-linkers , if it is assumed that the cross-linkers are rectangular support elements ( with a width , w , and a height , h ) and the same material properties as microtubules . However , since the rectangular cross-linkers , themselves , have an anisotropic area moment of inertia , the first term in equation must also take into account the orientation , θ , of each cross-linker , which we define with respect to the y-axis as ( 6 ) Jlinker ( θ ) = ( wh12 ) ( h2cos2θ+w2sin2θ ( w2−h2 ) sinθcosθ ( w2−h2 ) sinθcosθh2sin2θ+w2cos2θ ) The effect that polymer mass conservation has on the spindle's stiffness can be investigated by treating the spindle as a solid , homogeneous cylinder . If the volume of the cylinder , V , is held constant , to model the conservation of polymer mass , then the cylindrical beam becomes thinner as it elongates . The relationship between cylinder radius , rc , and its length , Ls , ( rc=V/πLs ) and the fourth-order scaling of transverse stiffness with beam radius , EI ∝ rc4 ( Landau et al . , 1986 ) then lead to the EI ∝ Ls−2 scaling of transverse stiffness with length . This scaling relation is consistent with the MMO model ( Figure 5 ) , which posits that the number of microtubules in a cross-section should be approximately LT/ ( Ls + Lm ) , because the transverse stiffness , EI , depends quadratically on the number of microtubules in a cross-section ( Figure 3C , D ) . The response of the spindle to compressive forces was investigated using the cytoskeletal modelling software Cytosim , which solves the over-damped Langevin equations of cytoskeletal filaments using an implicit numerical integration scheme ( Nedelec and Foethke , 2007 ) . The code was compiled and run on the EMBL High Performance Computing cluster , with jobs submitted to the Platform LSF scheduler using custom Python scripts . Simulation results were analysed using Matlab ( The MathWorks inc . ) . The simulations were designed to reproduce the morphology and biophysical characteristics of each spindle sufficiently closely to estimate the spindle's critical force . The SPBs were modelled as cylindrical elastic solids , associated with a scalar drag coefficient . Each microtubule was connected to the SPB by a pair of Hookean springs . The first of these was coupled to the minus-end of the microtubule , and was given a large elastic constant to model the high resistance of wild-type SPBs to pushing forces ( Toya et al . , 2007 ) . The steric exclusion between microtubules was implemented using a one-sided quadratic potential with a minimum at the steric radius of 30 nm ( Loughlin et al . , 2010 ) . The number and length of microtubules in the models of spindles from wild-type fission yeast cells were determined directly from ET reconstructions , whilst the SPB separation ( or spindle length ) was set to the IFTA-derived contour length between the poles of the ET spindle . The lengths of microtubules in budding yeast and cdc25 . 22 fission yeast cells were measured from the line representations of each spindle in the respective publications ( Ding et al . , 1993; Winey et al . , 1995 ) . The budding yeast spindles contain short microtubules ( with lengths less than 0 . 5 μm ) that are likely to represent kinetochore fibres that were not fully depolymerised during anaphase A . These fibres are unlikely to contribute to the structural integrity of the spindle , and account for a small proportion ( 7 . 8 ± 4 . 7% , ( S . D . ) ) of the total polymerised tubulin . These microtubules were therefore neglected in the spindle models . Two of the budding yeast spindles also contain pole-to-pole microtubules that cannot be unambiguously assigned to a specific spindle pole . In the first spindle ( numbered 12 in Winey et al . ) the two pole-to-pole microtubules are assigned symmetrically to each SPB . The second spindle ( numbered 14 in Winey et al . ) contains a single pole-to-pole microtubule that is assigned to the SPB with the lowest number of long nucleated microtubules . Having determined the spindle length , the number of microtubules and their lengths , the positions of the microtubule minus-ends at each SPB are set by sampling a random position on the circular face of each SPB using a Monte Carlo method . Rejection sampling was first used to sample the area on the disk's surface with uniform probability . The Euclidean distance between a candidate point and the other microtubules was determined , and the point was only accepted if its separation was greater than the twice the steric radius of each microtubule . This process was repeated until the SPB was populated with the correct number of microtubules . This procedure set the position of the microtubule minus-ends with respect to the SPB , with the position of plus-ends and the transverse organisation of microtubules at the midzone determined by simulating cross-linker attachment and detachment from the microtubule lattice . The SPBs were confined to the x-axis throughout the simulation to aid visualisation . The simulation of microtubule organisation at the spindle midzone was performed using cross-linkers that only bind to pairs of anti-parallel microtubules ( Janson et al . , 2007 ) . The cross-linkers were also confined to cylindrical region with a total length of 2 . 5 μm at the centre of the spindle , to approximate the width of the midzone in yeast cells ( Loïodice et al . , 2005; Yamashita et al . , 2005 ) . When bound to a pair of microtubules , cross-linkers behave as elastic bridging elements that set the centre-to-centre between pairs of microtubules at 50 nm . The stiffness of the cross-linkers is consistent with the known Young's modulus of the alpha-helical class of proteins to which the dimeric kinesins and Map65 proteins belong . The microtubules were also subjected to weak ( 20 Pa ) centring forces to prevent rotational diffusion of the microtubules away from the spindle axis and thus ensure that cross-links were formed between the two halves of the bipolar spindle . Throughout the initialisation of spindle architecture , the SPBs were connected to each other by a stiff spring with the same resting length as the spindle to prevent the SPB separation being altered by diffusion . After simulating the spindle for fifty seconds , almost all of the cross-linkers are bound to the midzone and the two halves of the spindle are strongly connected . Under these conditions , the cross-linkers are capable of forming the idealised square-packed arrays observed in yeast spindles , albeit with lower efficiency than we observe in electron tomogram reconstructions of wild-type spindles ( data not shown ) . Upon completion of the initialisation step , the rate with which cross-linkers detach from the microtubule lattice was set to zero in order to probe the spindles' elastic response to increasing forces . This was effected by decreasing instantaneously the resting length of the spring connecting the two SPBs by 1 μm , and then linearly increasing the spring constant from a starting value of zero to a maximum of 240 pNμm−1 over the remaining 150 s of the simulation . The stiffness of the elastic element was maintained at the maximal value for a further 50 s , during which time the critical force on the spindle was determined . This was carried out by averaging the forces borne by the elastic element connecting the pair of SPBs . In simulations of spindles with elastic reinforcement , each microtubule model point was confined by an elastic potential with a given degree of stiffness ( Figure 6A–D ) . The stiffness of the spring compressing the spindle poles was also increased to a maximum value of 1500 pNμm−1 to ensure that the critical force was exceeded , but all other simulation parameters were identical . Simulation parameters are provided in Table 2 . The simulations of the null models of spindle architecture were identical to those used to determine the critical force of wild-type spindles , except that microtubule length and number were sampled from probability distributions . The objective of this procedure was to sample a large number of alternative spindle morphologies to investigate the degree to which wild-type spindles are mechanically optimal . In constructing the null statistical models , the number of microtubules emanating from each SPB was sampled from a Poisson distribution with a mean equal to that observed in the wild-type spindle with the same length . The lengths of the microtubules in each random model were determined by randomly partitioning the total polymer present in the wild-type spindle between the NMT microtubules . In cases where the length of a sampled microtubule exceeded the spindle length , the microtubule was truncated to the length of the spindle with the remaining polymer used to set the length of one or more additional microtubules that were assigned to one of the two SPBs at random . This procedure increased slightly the average number of microtubules in the random spindles but ensured that the overall polymer mass was conserved . At SPBs that contained in excess of six microtubules , the radius of the circular face of the SPB was increased so that its area increased linearly with microtubule number , and that the density of microtubules on the surface of the SPB was constant . The viscous drag on the fission yeast nucleus is substantially larger than is predicted by Stokes' law due to the narrow separation between the nuclear envelope and the enclosing cell wall ( Foethke et al . , 2009 ) . The equation for translational motility of the nucleus in the cell geometry can be approximated asγ=9π22ηcellrnucleus4ϵ5/2where ε= ( rcell−rnucleus ) /rnucleus is the cell clearance . A description of the other variables is shown in Table 3 . After ablation of the spindle midzone , the forces acting on the daughter nuclei areFviscous=−Felastic−γdxdt=kxwhich has a solution , x ( t ) = x0 . e ( −t/τ ) , where τ = γ/k represents the characteristic time for the spindle to reach equilibrium , and x0 is the initial displacement . An exponential fit to the relaxation data provides values for these variables , which can then also be combined to give a rough estimate of the total force resisting spindle elongation ( Table 3 ) . | Before a cell divides to form two new cells , it duplicates its entire set of chromosomes . These chromosomes need to be equally distributed between the new cells—if cells receive too many or too few chromosomes , it can cause developmental defects or cancer . In cells that have a nucleus , a structure called the mitotic spindle ensures that chromosomes are partitioned equally between the dividing cells . The spindle consists of long , thin protein fibers called microtubules , which grow from small structures known as centrosomes that are present on either side of a cell . While some of the microtubules from each centrosome overlap in the middle of the spindle in a region called the spindle midzone , another set of microtubules attaches to the chromosomes , allowing the spindle to pull each of the chromosomes in a pair in opposite directions . The size and shape of the mitotic spindle varies widely between different species , and how the structure of the spindle helps it to do its job was unclear . However , it is known that the spindle has to be strong and fairly rigid in order to separate the chromosomes . Ward , Roque et al . studied the chromosome separation process in a species of yeast that has unusually consistent growth and cell division rates in different cells . In a technique called electron tomography , an electron microscope took images of the spindle from many different angles , and these images were combined computationally to produce a three-dimensional structure of the entire spindle . Ward , Roque et al . observe that the number and length of microtubule fibers in the spindle is the same in each yeast cell . The spindle also has a striking geometric pattern . In the spindle midzone , microtubules are ordered into a highly regular square-packed array , while the rest of the spindle contains microtubules arranged hexagonally . This hexagonal arrangement maximizes the interactions between a microtubule and its neighbors , which makes the spindle stronger and prevents it from buckling under the physical forces that act on it . Engineers have incorporated this type of design in man-made structures for decades . A future challenge is to explain how the properties of the spindle components have been tuned to be able to always assemble into a structure with such reliable properties . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"physics",
"of",
"living",
"systems"
] | 2014 | Mechanical design principles of a mitotic spindle |
Heterochromatin is a conserved feature of eukaryotic chromosomes with central roles in regulation of gene expression and maintenance of genome stability . Heterochromatin formation involves spreading of chromatin-modifying factors away from initiation points over large DNA domains by poorly understood mechanisms . In Saccharomyces cerevisiae , heterochromatin formation requires the SIR complex , which contains subunits with histone-modifying , histone-binding , and self-association activities . Here , we analyze binding of the Sir proteins to reconstituted mono- , di- , tri- , and tetra-nucleosomal chromatin templates and show that key Sir-Sir interactions bridge only sites on different nucleosomes but not sites on the same nucleosome , and are therefore 'interrupted' with respect to sites on the same nucleosome . We observe maximal binding affinity and cooperativity to unmodified di-nucleosomes and propose that nucleosome pairs bearing unmodified histone H4-lysine16 and H3-lysine79 form the fundamental units of Sir chromatin binding and that cooperative binding requiring two appropriately modified nucleosomes mediates selective Sir recruitment and spreading .
The packaging of eukaryotic nuclear DNA with histones and other proteins into chromatin is critical for the regulation of transcription , recombination , replication and DNA damage repair . The basic unit of chromatin folding is the nucleosome , in which 147 base pairs of DNA are wrapped around an octamer of two copies each of histones H2A , H2B , H3 , and H4 ( Kornberg , 1977; Luger et al . , 1997 ) . Site-specific DNA-binding proteins or RNA-based mechanisms recruit histone-modifying proteins that mediate histone posttranslational modifications such as acetylation and methylation ( Holoch and Moazed , 2015; Jenuwein and Allis , 2001; Kouzarides , 2007; Li et al . , 2007a; Schreiber and Bernstein , 2002; Strahl and Allis , 2000 ) . These modifications provide binding sites for numerous effector proteins that activate or silence transcription and are often associated with large domains of DNA , which can range in size from 2–3 to several hundred kilobases . The assembly of large domains of DNA with activating or repressive histone modifications allows regional and coordinated regulation of gene expression and maintenance of landmark chromosome structures , but the mechanisms that mediate the spreading of histone modification over large DNA domains are poorly understood . Heterochromatic DNA domains are a conserved feature of eukaryotic chromosomes and provide the most striking examples of regional control ( Moazed , 2001; Richards and Elgin , 2002 ) . Heterochromatin forms at repetitive DNA regions in order to prevent recombination and maintain genome integrity as well as at developmentally regulated genes ( Richards and Elgin , 2002 ) . Heterochromatin tends to spread from defined initiation sites , leading to the inactivation of genes in a sequence-independent manner ( Talbert and Henikoff , 2006; Wang et al . , 2014 ) . The mechanism of spreading of heterochromatin involves the recruitment of chromatin-modifying complexes , which have coupled histone-binding and histone-modifying activities , to specific nucleation sites ( Canzio et al . , 2011; Grunstein , 1997; Hoppe et al . , 2002; Luo et al . , 2002; Moazed , 2001; Rusche et al . , 2002 , 2003 ) . This is then followed by iterative cycles of histone modification and histone binding , which are thought to be coupled with the formation of homo and heterotypic interactions between silencing factors . In these models , silencing factors are proposed to form bridges that span binding sites on the same nucleosome and 'sticky ends' that extend away from nucleation points and mediate interactions across neighboring nucleosomes ( Canzio et al . , 2011; Moazed , 2001 ) . The continuous nature of interactions between silencing proteins , both across single nucleosomes and across neighboring ones , amounts to the formation of proteinaceous chromatin-bound oligomers . However , this model has not been tested with chromatin templates that allow the extent to which self-association of silencing factors contributes to specific binding and spreading to be determined . The nature of silencing factor self-associations , how they bridge nucleosomes , and their relationship to the mechanism of spreading therefore remain ambiguous . Silent chromatin in the budding yeast S . cerevisiae serves as a major model system for studies of heterochromatin establishment and inheritance . Silencing at the silent mating type loci ( HML and HMR ) and sub-telomeric regions requires silent information regulator ( Sir ) proteins , Sir2 , Sir3 , and Sir4 , which together form the SIR complex ( Aparicio et al . , 1991; Klar et al . , 1979; Liou et al . , 2005; Moazed et al . , 1997; Rine and Herskowitz , 1987; Rudner et al . , 2005; Rusche et al . , 2003 ) . During the initiation step , the Sir2 and Sir4 proteins , which together form a stable Sir2/4 heterodimer ( Moazed et al . , 1997 ) , and Sir3 , are recruited to the silencer through interactions with silencer-specificity factors , ORC , Abf1 , and Rap1 ( Hoppe et al . , 2002; Luo et al . , 2002; Moretti et al . , 1994; Moretti and Shore , 2001; Rusche et al . , 2002; Triolo and Sternglanz , 1996 ) . The Sir2 subunit , which is an NAD-dependent deacetylase ( Imai et al . , 2000; Landry et al . , 2000; Smith et al . , 2000 ) , then deacetylates silencer-proximal nucleosomes , particularly the H4K16 residue , creating a binding site for Sir3 ( Armache et al . , 2011; Carmen et al . , 2002; Liou et al . , 2005; Wang et al . , 2013 ) . Subsequent iterative cycles of deacetylation and Sir-Sir interactions lead to spreading of SIR complexes ( Hoppe et al . , 2002; Luo et al . , 2002; Rusche et al . , 2002 ) along multiple kilobases of chromatin away from the silencer . Many of the key activities of the SIR complex have been mapped to specific domains in its subunits and provide important guides for further studies . Modification-sensitive nucleosome binding occurs via a conserved domain at the N terminus of Sir3 , called the bromo-adjacent homology ( BAH ) domain ( Figure 1A ) ( Buchberger et al . , 2008; Onishi et al . , 2007 ) . The AAA ATPase-like ( AAAL ) domain of Sir3 also interacts with histones and nucleosomes ( Hecht et al . , 1995 ) . However , this interaction is at least an order of magnitude weaker than the BAH-mediated chromatin interactions ( Martino et al . , 2009; Wang et al . , 2013 ) . Sir4 forms stable dimers via a coiled-coil domain at its C terminus ( Sir4CC ) , which also forms a binding surface for two Sir3 molecules , linking the Sir2 histone deacetylase to the nucleosome binding subunit of the complex ( Chang et al . , 2003; Moazed et al . , 1997; Rudner et al . , 2005 ) . Finally , Sir3 forms dimers via a winged helix ( wH ) domain at its C terminus ( Oppikofer et al . , 2013 ) . Although all of the above interaction domains are critical for silencing , how they promote the spreading of silencing remains unknown . 10 . 7554/eLife . 17556 . 003Figure 1 . Cooperative association of Sir3 with DiN . ( A ) Schematic diagram of the Sir3 primary sequence showing the location of the BAH , AAAL , and wH domains . ( B ) Models for the association of Sir3 dimers with chromatin . ( C , D ) Representative EMSA showing Sir3 binding to unmodified MonoN ( C ) and DiN ( D ) . Purified Sir3 proteins were titrated onto a constant amount of MonoN or DiN at 3 nM . Samples were separated on native gels , nucleosomes were stained with SYBR Gold , and the amount of unbound nucleosomes was quantified by the staining intensity of the unshifted nucleosome band . * , higher mobility shifted band that may result from either bridging of MonoN by Sir3 or other minor high molecular weight Sir3-MonoN complexes . Band 1 ( B1 ) likely reflects Sir3-DiN in bridged conformation , whereas Band 2 ( B2 ) shows additional binding to single nucleosome surfaces . ( E ) Quantification and analysis of Sir3 binding to MonoN . Binding curves from three experiments performed as in ( C ) and ( D ) were fitted with the Hill Equation . The apparent KD values for Sir3 binding to MonoN ( blue dotted line ) and DiN ( red dotted line ) are indicated . ( F ) Bio-layer interferometry ( BLI ) assay detects changes in binding of proteins to surface-immobilized nucleosome templates by measuring the wavelength shift of the light reflected from the surface . ( G , H ) Binding of Sir3BAH ( a . a . 1–382 ) ( G ) and full-length Sir3 ( H ) to MonoN and DiN after background correction and normalization to min-max of binding signal was fit with the Hill equation ( see Materials and methods and Table 1 ) . Data from 2–3 replicate experiments ( >30 data points ) ( G ) and from 3 or more replicate experiments ( >30 data points ) ( H ) were pooled for model fitting . The apparent KD for Sir3BAH and Sir3 binding to MonoN ( blue ) and DiN ( red ) is indicated . See Table 1 for parameter values . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 00310 . 7554/eLife . 17556 . 004Figure 1—figure supplement 1 . In vitro reconstitution of MonoN , DiN , TriN , and TetraN . ( A ) Purification and analysis of the yeast histone octamer ( HO ) using gel filtration . HO refolded from individual histones was purified on a gel filtration column . Fractions from the gel filtration column ( left ) were run on denaturing SDS polyacrylamide gel ( right ) . ( B–E ) Reconstitution of MonoN ( B ) , DiN ( C ) , TriN ( D ) , and TetraN ( E ) using a titration of different HO:DNA ratios . The reconstitutions that produced a single sharp band ( red dotted box ) were used for biochemical assays . ( F ) Left: Illustration of the DiN DNA construct , and the predicted restriction enzyme digestion pattern for naked DNA and fully reconstituted DiN . Right: Restriction enzyme protection assay of reconstituted DiN . U: uncut , S: digested with ScaI , A: digested with AluI . Note that ScaI but not AluI digestion converts slower migrating DiN to faster migrating MonoN , whereas naked DNA is cleaved with both enzymes , as expected . ( G ) Purified Sir3-3XFLAG used in binding experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 00410 . 7554/eLife . 17556 . 005Figure 1—figure supplement 2 . Association of the Sir3 with DiN and Sir3BAH domain with MonoN and DiN . ( A ) EMSA experiments showing binding of the Sir3 domain to DiN with and without ScaI digestion . ( B ) EMSA experiments showing binding of the Sir3BAH domain to MonoN and DiN . The continuous up-shift of bands may be due to BAH-BAH interactions ( Armache et al . , 2011 ) or nonspecific BAH-MonoN interactions in gel buffer at high BAH concentrations . ( C ) Binding curves from three experiments were fitted with the Hill Equation . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 00510 . 7554/eLife . 17556 . 006Figure 1—figure supplement 3 . Measurements of Sir3 binding to MonoN with BLI . ( A ) MonoN and DiN reconstitution using histone octamers containing histone H2A-K120C covalently linked to biotin-PEG2 maleimide . ( B ) Restriction enzyme protection assay of biotinylated DiN . ( C ) BLI assay of Sir3 binding to MonoN with differently biotinylated MonoN substrates yields identical results . The titration shows normalized equilibrium binding signal , integrating all modes of Sir3 binding to MonoN . H2A-biotin-PEG2 MonoN: Maleimide-PEG2-Biotin moiety covalently attached to histone H2A K120C SH group , 167bp-DNA-biotin: biotin moiety is covalently attached to the 5’-end at the beginning of DNA template ( 20 bp extension + 147 bp 601 positioning sequence ) . ( D ) Equilibrium binding signal to unmodified ( dark blue ) or acetylated ( light blue ) nucleosomes bearing H2A-biotin-PEG2 were measured by BLI at [Sir3] = 1 . 0–1 . 5 µM . Nucleosomes were loaded on streptavidin-coated BLI sensors to the same level before being used for Sir3 titration experiments . Histograms show the average and standard deviation of the equilibrium binding signal of the three BLI sensograms shown on the left . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 00610 . 7554/eLife . 17556 . 007Figure 1—figure supplement 4 . Analysis of BLI binding profiles . ( A ) Representative Sir3-MonoN complex decay curves for full-length ( Sir3 ) , Sir3∆wH , and BAH ( Sir3 a . a . 1–382 ) proteins indicate that deletion of AAA-ATPase like domain , unlike the wH domain has a strong effect on nucleosome residence time of Sir3 . For Sir31–382 , best fit to dissociation curve is shown instead of raw data . Dissociation curves are normalized to their ranges for easier visual comparison . ( B , C ) Real-time association ( B ) and dissociation ( C ) profiles of Sir3 binding to unmodified DiN at indicated concentrations of Sir3 . Higher concentrations are not shown to highlight the range of concentrations in which Sir3 binds cooperatively to di-nucleosome . The black lines indicate the best fit mono-exponential models ( see Materials and methods ) with associated fit residuals shown in the bottom plot . ( D , E ) Real-time association ( D ) and dissociation ( E ) profiles of Sir3 binding to unmodified MonoN at indicated concentration of Sir3 . Fits with mono- and bi-exponential models ( black lines , see Materials and methods ) at representative Sir3 concentrations are shown on the right . The systematic deviation of the mono-exponential model from data results in a periodic ( non-random ) change of sign in residuals ( bottom plots ) between positive and negative values , indicating model inadequacy . Fits with bi-exponential model show a substantial decrease and random distribution around zero of fit residuals . ( F , G ) Real-time association ( F ) and dissociation ( G ) profiles of Sir3BAH binding to unmodified MonoN at indicated concentrations of Sir3BAH showing mono-phasic association and dissociation . In protein concentrations above 1 µM a slow association/dissociation phase with small amplitude appeared , which likely reflected nonspecific interactions with nucleosomes . Measurements at different protein concentrations in B–G are performed by separate BLI sensors , pre-loaded with biotinylated nucleosomes to identical levels in each titration ( ± 5% variation ) . Dissociation profiles ( C , E , G ) are aligned at time zero to an arbitrary value . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 007 In this study , we use equilibrium and kinetic binding experiments to compare the association of Sir3 and its subfragments with in vitro reconstituted mono- , di- , tri- , and tetra-nucleosomal chromatin templates ( MonoN , DiN , TriN , and TetraN , respectively ) , and determine how this association is affected by the Sir4 coiled-coil ( Sir4CC ) domain . Our analysis shows that , at physiological concentrations , Sir3 binds to DiN with maximal cooperativity by a mechanism that requires the Sir3wH dimerization domain and is enhanced by the Sir4CC domain . In contrast , although each nucleosome contains two Sir3 binding sites , the association of Sir3 with MonoN , in the presence or absence of Sir4CC , occurs by a non-cooperative mechanism , suggesting that these interactions mediate lateral Sir3-Sir3 bridging across two nucleosomes , rather than on the same nucleosome . Moreover , we show that H4K16 acetylation and H3K79 trimethylation , two well-established anti-silencing modifications , work together to dramatically reduce the affinity of Sir3 for nucleosomes . Together , our findings suggest that spreading of the Sir proteins on chromatin involves the cooperative recruitment of new SIR complexes to pairs of nucleosomes lacking H4K16 acetylation and H3K79 methylation independently of interactions with already bound SIR complexes . This inter-nucleosomal cooperative mode of binding suggests that interrupted Sir3 bridges across neighboring nucleosomes , stabilized by Sir4 , are the primary driving force for heterochromatin spreading .
Sir3 molecules self-associate with high affinity ( KD ~2 nM , Liou et al . , 2005 ) to form mostly homo-dimers and to a lesser extent oligomers ( Figure 1A ) ( Liou et al . , 2005; McBryant et al . , 2006 ) . Due to the multivalent nature of Sir3 dimers , potential interactions between Sir3 proteins bound to the same ( intra-nucleosomal bridging ) or adjacent nucleosomes ( inter-nucleosomal bridging ) may cooperatively stabilize Sir3 interactions with properly modified chromatin ( Figure 1B ) . These different modes of binding predict different Sir3 affinities for mono- and di-nucleosomes ( MonoN and DiN , respectively ) that will depend on Sir3 dimerization . In order to investigate the mechanism of Sir3 binding to chromatin and to distinguish intra- and inter-nucleosomal contributions to Sir3-chromatin association , we studied binding of Sir3 with MonoN and DiN . We reconstituted defined nucleosome arrays by the salt-gradient dialysis method , as described previously ( Huynh et al . , 2005; Luger et al . , 1999 ) , using the 601 nucleosome positioning sequence and S . cerevisiae histones purified from E . coli ( Figure 1—figure supplement 1A–E ) . The quality of reconstituted nucleosomes was assessed by electrophoretic mobility shift assays ( EMSA ) and restriction enzyme protection ( Figure 1—figure supplement 1F ) . Purified overexpressed Sir3 from S . cerevisiae ( Liou et al . , 2005 ) was used for all assays ( Figure 1—figure supplement 1G ) to maintain the Nɑ-acetylation of Sir3 , which is required for its efficient binding to nucleosomes ( Arnaudo et al . , 2013; Onishi et al . , 2007; Yang et al . , 2013 ) and silencing in vivo ( Wang et al . , 2004 ) . Various purification strategies resulted in Sir3 proteins with identical purity and nucleosome binding behavior ( see Materials and methods ) . Binding experiments were performed in salt concentrations that had previously been determined to render Sir3 binding strongly sensitive to H4K16Q mutation , resembling the in vivo behavior of Sir3 ( Johnson et al . , 2009; Swygert et al . , 2014 ) . To determine the affinity of Sir3 for MonoN and DiN , we performed EMSA ( Buchberger et al . , 2008; Johnson et al . , 2009; Liou et al . , 2005 ) . We found that Sir3 bound to MonoN with a KD around 1 . 7 μM ( Figure 1C , E ) and to the DiN with a KD around 0 . 17 µM ( Figure 1D , E ) , indicating a ~10 fold increase in affinity for DiN relative to MonoN , much higher than what would be expected by the increase in the number of independent binding sites . ScaI digestion of the linker DNA in the DiN resulted in a reduction in binding affinity to that observed for the MonoN ( Figure 1—figure supplement 2A ) , indicating that specific binding to DiN , not extra DNA content , is responsible for higher affinity binding ( Other assays ruling out DNA binding are described in Materials and methods ) . In contrast , Sir3BAH domain bound to DiN with only around 2-fold higher affinity than to MonoN ( Figure 1—figure supplement 2B , C; Table 1 ) , as expected from the higher number of binding sites on the DiN . 10 . 7554/eLife . 17556 . 008Table 1 . Thermodynamic parameters describing the binding of Sir3 protein to nucleosomes . Data from more than 2 replicate titration experiments were pooled and the BLI data were fit with Hill equation ( see Materials and methods ) . Uncertainties show 68% confidence intervals around fit parameters ( ±1 SD ) reported by fitting algorithm . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 008Binding experiments BLIEMSA*Apparent KD ( µM ) Hill coefficient Apparent KD ( µM ) MonoNSir31 . 4 ± 0 . 06 1 . 3 ± 0 . 1 1 . 7 ± 0 . 20 Sir3+Sir4CCN/AN/A1 . 4 ± 0 . 10 Sir3∆wH1 . 2 ± 0 . 10 0 . 93 ± 0 . 07 1 . 0 ± 0 . 10 Sir3∆wH+Sir4CCN/AN/A0 . 9 ± 0 . 05 Sir3BAH1 . 4 ± 0 . 10 1 . 5 ± 0 . 2 2 . 1 ± 0 . 20 DiNSir30 . 12 ± 0 . 01 1 . 9 ± 0 . 2 0 . 17 ± 0 . 10 Sir3+Sir4CCN/AN/A0 . 08 ± 0 . 01 Sir3∆wH1 . 1 ± 0 . 05 1 . 2 ± 0 . 1 0 . 62 ± 0 . 10 Sir3∆wH+Sir4CCN/AN/A0 . 12 ± 0 . 01 Sir3BAH1 . 6 ± 0 . 10 1 . 4 ± 0 . 1 1 . 40 ± 0 . 20 Sir3acMonoNN/AN/A4 . 0 ± 0 . 20 meMonoNN/AN/A5 . 2 ± 0 . 20 acDiNN/AN/A0 . 7 ± 0 . 05 meDiNN/AN/A0 . 8 ± 0 . 05 ac/meMonoNN/AN/A>11†ac/meDiNN/AN/A>3†*Hill coefficients obtained from EMSA appeared unreliable due to assay artifacts , such as non-specific binding to DNA , and are not reported . † Nonspecific binding could not be measured accurately due to low affinity . Although EMSA allows for direct observation of complex formation and may even elucidate assembly intermediates , it is a quasi-equilibrium method , which might be affected by fast dynamics of the complex in the gel matrix or spurious Sir3-nucleosome interactions due to the low ionic strength and temperature of EMSA buffer . To validate the EMSA observations using an equilibrium assay , we examined Sir3-nucleosome interactions at physiological ionic strength and temperature using the BioLayer Interferometry ( BLI ) assay ( Abdiche et al . , 2008 ) . To perform BLI measurements , we immobilized biotinylated MonoN or DiN ( Figure 1—figure supplement 3A , B ) on the surface of streptavidin-coated biosensors and studied changes in the number of Sir3 molecules bound to nucleosomes by monitoring in real time the wavelength shifts of the reflected light from the biosensor surface ( Figure 1F , see Materials and methods ) . The binding signal at equilibrium reflects the number of Sir3 molecules bound to nucleosomes at any given Sir3 concentration . We reconstructed binding isotherms by plotting normalized binding signals at equilibrium vs . Sir3 concentration ( see Materials and methods for further details ) . As controls , binding of Sir3 to immobilized nucleosomes was insensitive to whether the biotin moiety was attached to histone H2A via a flexible 30 Å linker or to the end of nucleosomal DNA with a 20 bp extension ( Figure 1—figure supplement 3C ) . In contrast , it was strongly sensitive to the acetylation of nucleosomes ( Figure 1—figure supplement 3D ) , indicating that the BLI signal reflected specific Sir3-nucleosome interactions . We observed that Sir3BAH bound to MonoN and DiN with the nearly identical apparent affinity of ~1 . 4 µM ( Figure 1G , Table 1 ) , consistent with EMSA measurements ( Table 1 ) . Similarly , analysis of the binding profiles of full-length Sir3 to MonoN and DiN resulted in calculated apparent KD values of around 1 . 4 and 0 . 11 µM , respectively ( Figure 1H , Table 1 ) , which are similar to those obtained by EMSA ( Table 1 ) . The Hill coefficient of Sir3 and Sir3BAH binding to MonoN was nearly identical and close to 1 ( Figure 1G , Table 1 ) . In fact , binding of Sir3 or Sir3BAH to MonoN could be satisfactorily described by a simple model based on saturation of two identical and independent ( i . e . non-cooperative ) binding sites with a macroscopic dissociation constant of 1 . 2–1 . 4 µM ( see Materials and methods ) , confirming the absence of cooperativity in binding of Sir3 to MonoN . In contrast , we observed cooperative binding of Sir3 to DiN , reflected in the Hill coefficient of ~2 ( Figure 1H; Table 1 ) . Therefore , the complex of Sir3 with DiN , but not with MonoN , is stabilized by a cooperative mechanism , supporting the inter-nucleosomal mode of bridging depicted in Figure 1B . Our findings above show that Sir3-Sir3 contacts across different nucleosomes ( Figure 1B , right ) play an important role in stabilizing Sir3-chromatin complexes . In contrast , even though there are two binding surfaces on each nucleosome for Sir3 , intra-nucleosomal Sir3 interactions are prohibited ( Figure 1B , left ) . Therefore , one may conclude that in the context of chromatin arrays , unmodified DiN units act as independent high affinity binding sites for Sir3 dimers . To directly test this hypothesis , we compared the binding of Sir3 to DiN versus larger nucleosomal arrays using the BLI assay . Positive interactions between binding sites , beyond those present in the DiN template ( equivalent to Sir3 oligomerization on chromatin ) , would result in higher apparent binding affinity and cooperativity for longer nucleosome arrays compared to DiN . In stark contrast to the difference in binding affinity between MonoN and DiN templates ( Figure 1E , H ) , Sir3 binding to tri- and tetra-nucleosome templates ( TriN and TetraN , respectively ) displayed very similar binding affinity and cooperativity as binding to DiN ( Figure 2A ) . This result suggests that even in the context of nucleosome arrays Sir3-chromatin interactions that contribute to binding stability are limited to sites on only two different nucleosomes ( Figure 1B , right ) . We therefore conclude that binding sites on two different nucleosomes form the fundamental unit of chromatin binding for Sir3 dimers . 10 . 7554/eLife . 17556 . 009Figure 2 . Sir3 displays maximal cooperative binding to DiN that is mediated by its winged helix ( wH ) . ( A ) Binding of full-length Sir3 to DiN , TriN , and TetraN templates after background correction and normalization to min-max of binding signal was fit with the Hill equation ( see Materials and methods and Table 1 ) . Data from 3 or more replicate experiments ( >30 data points ) were pooled for model fitting . Vertical dotted line indicates the apparent KD for Sir3 binding to DiN , TriN and TetraN . ( B ) EMSA showing the binding of Sir3ΔwH to MonoN ( blue ) and DiN ( red ) . Binding curves from three experiments performed ( see Figure 2—figure supplement 1 ) were fitted with the Hill Equation . ( C ) Sir3ΔwH MonoN and DiN binding data from BLI assays normalized to the range of binding signal and fit with Hill equation ( see Materials and methods ) . Data from two independent replicates were pooled before model fitting . ( D ) Kinetics of Sir3-nucleosome complex dissociation measured by the BLI assay reveals that the Sir3wH domain is required for the cooperative stabilization of Sir3-DiN and Sir3-triN complexes . Representative dissociation profiles obtained at 0 . 2 µM Sir3 were self-normalized for easier visual comparison . See Table 1 for parameter values . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 00910 . 7554/eLife . 17556 . 010Figure 2—figure supplement 1 . Purification of Sir3∆wH and EMSA assays with Sir3∆wH . ( A ) Coomassie-stained gels showing purified Sir3ΔwH-CBP ( left ) and Sir3ΔwH-3XFLAG ( right ) proteins used in EMSA and BLI experiments . ( B ) EMSA experiments comparing the binding of Sir3ΔwH-3XFLAG to unmodified MonoN or DiN . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 010 Sir3 forms dimers and oligomers in vitro ( King et al . , 2006; Liou et al . , 2005; McBryant et al . , 2006; Moretti et al . , 1994 ) , and its C-terminal winged helix ( wH ) domain is necessary and sufficient for dimerization ( Oppikofer et al . , 2013 ) ( see Figure 1A for Sir3 domains ) . Furthermore , deletion of the wH domain abolishes Sir3 association with silent chromatin regions and disrupts silencing at both the mating-type loci and telomeres ( Oppikofer et al . , 2013 ) . We therefore investigated whether and how the wH domain may contribute to the cooperative mechanism of Sir3 binding to DiN by studying the binding of affinity purified Sir3 lacking the wH domain ( Sir3ΔwH ) ( Figure 2—figure supplement 1A ) to both MonoN and DiN . Both EMSA and BLI assays showed that Sir3ΔwH bound to MonoN with a KD around 1 . 1 μM ( Figure 2A , B and Figure 2—figure supplement 1B , Table 1 ) , similar to the KD value ( 1 . 4 μM ) we observed for the association of full-length Sir3 with MonoN . Deletion of the wH domain therefore did not affect Sir3 affinity for MonoN . Moreover , in contrast with ~10 fold increase in the apparent affinity of full-length Sir3 for DiN compared to MonoN ( Figure 1G , H , Table 1 ) , we did not observe a significant difference between binding of Sir3ΔwH to DiN and MonoN ( Figure 2B , C , Table 1 ) . Further analysis of the BLI binding data with the Hill equation revealed that in contrast to full-length Sir3 ( Figure 1G , H ) , Sir3∆wH , like Sir3BAH , bound to MonoN and DiN without cooperativity ( Table 1 ) . We therefore conclude that the wH dimerization domain is required for Sir3 cooperative and high affinity binding to DiN , without affecting Sir3 binding to MonoN . Furthermore , measurement of complex dissociation rates by BLI revealed that loss of wH domain reduced the half-life of Sir3-DiN and Sir3-TriN complexes by ~40% , while having little or no effect on Sir3-MonoN complex half-life ( Figure 2D , Table 2 ) . These findings support our conclusion regarding Sir3 dimerization in forming inter-nucleosomal bridges and the absence of intra-nucleosomal Sir3 contacts , and suggest that inter-nucleosomal bridges may contribute to forming temporally stable heterochromatin domains in vivo . 10 . 7554/eLife . 17556 . 011Table 2 . Kinetic parameters describing the binding of Sir3 protein to nucleosomes . Rates and amplitudes , representing the average of 3 or more measurements at different Sir3 concentrations in the ranges specified below , were obtained by fitting data with appropriate rate equations ( Suppl . Materials and methods ) . Values in parentheses indicate standard deviations . In the low range of concentrations , binding rates to mono- and di-nucleosomes were within the expected range of diffusion-limited rate of protein interactions ( Schreiber et al . , 2009 ) . At higher concentrations , however , binding proceeded with rates slower than diffusion limit , probably due to the competition with other modes of binding to nucleosome surfaces . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 011Association Aobs , 1 kobs , 1 ( s−1 ) Aobs , 2 kobs , 2 ( s−1 ) [Sir3] < 0 . 3 µM MonoNSir351% ( 11% ) 0 . 25 ( 0 . 07 ) 49% ( 11% ) 0 . 04 ( 0 . 01 ) Sir3∆wH73% ( 11% ) 0 . 14 ( 0 . 06 ) 27% ( 11% ) 0 . 03 ( 0 . 01 ) Sir3BAH100% ( 0% ) 0 . 65 ( 0 . 34 ) 0% ( 0% ) N/AN/ADiNSir3100% ( 0% ) 0 . 06 ( 0 . 01 ) 0% ( 0% ) N/AN/ASir3∆wH11% ( 22% ) 0 . 16N/A89% ( 22% ) 0 . 08 ( 0 . 01 ) Sir3BAH100% ( 0% ) 0 . 40 ( 0 . 18 ) 0% ( 0% ) N/AN/A[Sir3] 1 . 5–4 µM MonoNSir356% ( 1% ) 0 . 62 ( 0 . 02 ) 44% ( 1% ) 0 . 08 ( 0 . 01 ) Sir3∆wH74% ( 4% ) 0 . 70 ( 0 . 07 ) 26% ( 4% ) 0 . 10 ( 0 . 02 ) Sir3BAH*100% ( 0% ) 0 . 53 ( 0 . 06 ) 0% ( 0% ) N/AN/ADiNSir355% ( 2% ) 0 . 51 ( 0 . 10 ) 45% ( 2% ) 0 . 08 ( 0 . 00 ) Sir3∆wH56% ( 3% ) 0 . 47 ( 0 . 05 ) 44% ( 3% ) 0 . 09 ( 0 . 02 ) Sir3BAH*100% ( 0% ) 0 . 46 ( 0 . 03 ) 0% ( 0% ) N/AN/Ai †Aoff , 1 τoff , 1 ( S ) Aoff , 2 τoff , 2 ( S ) kon , 1 ( M−1s−1 ) ‡kon , 2 ( M-1s−1 ) ‡ [Sir3] < 0 . 3 µM MonoNSir3238% ( 3% ) 7 . 4 ( 3 . 5 ) 62% ( 3% ) 78 . 5 ( 23 . 7 ) 2 . 4E + 05 ( 6 . 5E + 4 ) 6 . 9E + 04 ( 2 . 0E + 4 ) Sir3∆wH256% ( 2% ) 8 . 4 ( 0 . 9 ) 44% ( 2% ) 75 . 3 ( 3 . 6 ) 1 . 1E + 0 ( 4 . 3E + 4 ) 3 . 6E + 04 ( 1 . 4E + 4 ) Sir3BAH2100% ( 0% ) 5 . 6 ( 1 . 9 ) 0% ( 0% ) N/A N/A1 . 4E + 06 ( 9 . 2E + 5 ) N/AN/ADiNSir3192% ( 14% ) 51 . 6 ( 3 . 8 ) 8% ( 14% ) N/A N/A3 . 8E + 05 ( 6 . 8E + 4 ) N/AN/ASir3∆wH452% ( 2% ) 10 . 8 ( 1 . 2 ) 48% ( 2% ) 83 . 8 ( 6 . 5 ) 9 . 7E + 04 ( 3 . 7E + 4 ) 4 . 4E + 04 N/A Sir3BAH4100% ( 0% ) 5 . 0 ( 1 . 4 ) 0% ( 0% ) N/A N/AN/AN/AN/AN/A[Sir3] 1 . 5–4 µM MonoNSir3252% ( 3% ) 7 . 7 ( 0 . 5 ) 48% ( 3% ) 82 . 2 ( 3 . 7 ) 9 . 1E + 04 ( 2 . 4E + 4 ) 1 . 3E + 04 ( 4 . 0E + 3 ) Sir3∆wH269% ( 2% ) 3 . 4 ( 0 . 5 ) 31% ( 2% ) 44 . 0 ( 12 . 4 ) 8 . 5E + 04 ( 3 . 2E + 4 ) 1 . 6E + 04 ( 7 . 3E + 3 ) Sir3BAH2100% ( 0% ) 5 . 2 ( 0 . 5 ) 0% ( 0% ) N/A N/A5 . 5E + 04 ( 1 . 3E + 4 ) N/AN/ADiNSir3442% ( 3% ) 5 . 8 ( 0 . 1 ) 58% ( 3% ) 48 . 6 ( 2 . 8 ) 3 . 7E + 04 ( 8 . 1E + 3 ) 7 . 2E + 03 ( 3 . 5E + 3 ) Sir3∆wH463% ( 1% ) 5 . 2 ( 0 . 6 ) 37% ( 1% ) 48 . 5 ( 3 . 8 ) 3 . 2E + 04 ( 7 . 5E + 3 ) 8 . 7E + 03 ( 2 . 9E + 3 ) Sir3BAH4100% ( 0% ) 5 . 2 ( 0 . 5 ) 0% ( 0% ) N/A N/A2 . 1E + 04 ( 7 . 1E + 3 ) N/AN/A* Slow dissociation phase was not quantified due to small amplitude . † Presumed number of binding sites used in the calculation of kon . ‡ Rates are calculated from ( kobs , 1 , koff , 1 ) and ( kobs , 2 , koff , 2 ) pairs . Furthermore , Sir3 proteins lacking both the AAAL and wH domains ( Sir3BAH ) , but not wH domain alone ( Sir3∆wH ) , displayed drastically reduced Sir3-MonoN complex half-life ( Figure 1—figure supplement 4A , see also Materials and methods ) . Therefore , although AAAL domain on its own interacts weakly with chromatin ( Wang et al . , 2013 ) , it plays an important role in stabilizing the BAH-mediated Sir3-nucleosome complex . This effect is distinct from the wH domain-mediated stabilization of Sir3 inter-nucleosomal bridges discussed above . Sir4 forms dimers , via its C-terminal coiled-coil ( CC ) domain , and this dimerization activity is required for silencing at both telomeres and the mating-type locus ( Figure 3A ) ( Chang et al . , 2003; Chien et al . , 1991; Murphy et al . , 2003 ) . Sir4CC also interacts with Sir3 , through the Sir3 AAA ATPase-like ( AAAL ) domain , and this interaction is required for Sir3 recruitment and silencing in vivo ( Figure 3A ) ( Chang et al . , 2003; Ehrentraut et al . , 2011; King et al . , 2006; Rudner et al . , 2005 ) . We therefore speculated that Sir4CC might further stabilize Sir3-nucleosome interactions . There are at least two possible ways that Sir4 may interact with Sir3-nucleosome complexes . In the first mode , the Sir4CC mediates interactions between Sir3 molecules bound to each side of the same nucleosome forming an intra-nucleosomal bridge ( Figure 3B , left ) . In the second model , the Sir4CC interacts with Sir3 molecules bound to two neighboring nucleosomes , adding a second layer of inter-nucleosomal interactions ( Figure 3B , right ) . Similar to the different Sir3 binding modes described above , these two models predict different Sir4CC effects on the binding affinity of Sir3 for the MonoN and DiN , and therefore can be distinguished by binding experiments . Intra-nucleosomal bridging predicts that Sir3/Sir4CC has a higher binding affinity towards MonoN than Sir3 alone . In contrast , inter-nucleosomal bridging predicts ( 1 ) higher binding affinity of Sir3/Sir4CC towards DiN compared with Sir3 alone , because of the dimerizing Sir3wH domain and the interaction of Sir4CC with Sir3 AAAL domains , and ( 2 ) no change in the binding affinity of Sir3 towards MonoN upon the addition of Sir4CC . 10 . 7554/eLife . 17556 . 012Figure 3 . Sir4CC does not affect Sir3 binding to MonoN , but increases its affinity towards DiN . ( A ) Schematic diagram of Sir4 primary sequence showing the location of the coiled-coil ( CC ) domain and the Sir2 interaction domain ( aa747–893 ) . ( B ) Models for the association of Sir4 with Sir3-bound nucleosomes . ( C , D ) EMSA experiments comparing Sir3 binding to MonoN ( C ) and DiN ( D ) in the presence or absence of Sir4CC . ( E , F ) Binding curves from three experiments performed as in ( C ) and ( D ) , respectively , were fitted with the Hill Equation . ( G , H ) Comparison of Sir3ΔwH binding to MonoN ( G ) and DiN ( H ) in the presence or absence of Sir4CC . Binding curves from three experiments performed as in Figure 3—figure supplement 1A and B were fitted with the Hill equation . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 01210 . 7554/eLife . 17556 . 013Figure 3—figure supplement 1 . EMSA assays with Sir3∆wH and Sir4CC . ( A ) EMSA experiments comparing the binding of Sir3ΔwH to unmodified MonoN in the presence or absence of 2X molar excess of Sir4CC . Quantitative analysis of the results is presented in Figure 3G . ( B ) Comparison of Sir3ΔwH binding to DiN in the presence or absence of 2X molar excess of Sir4CC . Quantitative analysis of the results is presented in Figure 3H . ( C ) Comparison of Sir3ΔwH binding to MonoN and DiN in the presence of 2-fold molar excess of Sir4CC . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 013 EMSA experiments showed that although the addition of Sir4CC caused a slight upshift of Sir3-MonoN band , it did not affect Sir3-MonoN binding affinity , as the KD values were similar with or without Sir4CC ( Figure 3C , E , Table 1 ) . In contrast , Sir4CC decreased the apparent KD value of Sir3 binding to DiN about 2 fold , from 0 . 17 to 0 . 08 μM ( Figure 3D , F , Table 1 ) , suggesting that Sir4 binds to Sir3 proteins that bridge neighboring nucleosomes . Consistent with the above binding results ( Figure 4C–F ) , Sir4CC did not affect the binding affinity of Sir3ΔwH towards MonoN ( Figure 3G , Figure 3—figure supplement 1A , C , Table 1 ) , but increased its binding affinity towards DiN about 4-fold ( Figure 3H , Figure 3—figure supplement 1B , C , Table 1 ) . We conclude that both Sir3wH and Sir4CC dimerization domains contribute to inter-nucleosomal bridging , but they may perform partially redundant functions in this respect . 10 . 7554/eLife . 17556 . 014Figure 4 . Sir3 crosslinks free mono-nucleosomes in solution . ( A ) Illustration of the crosslinking assay . ( B ) A representative native polyacrylamide gel showing the fluorescent DNA pulled down from different reaction mixtures . ( C ) A representative western blot showing Sir3 and Sir3ΔwH in all reactions bound to nucleosomes efficiently . I , input; B , bound , U , unbound . ( D ) Percentage of input Alexa 647- nucleosomes that was pulled down in different reaction mixtures where 2 µM Sir3 or Sir3ΔwH protein concentration was used . Quantification for 6 experiments is presented . ( E ) Same as D but using 0 . 2 µM Sir3 or Sir3ΔwH protein . Quantification for 4 experiments is presented . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 014 Since Sir3 binds to DiN cooperatively , we tested whether it could act as a bridge linking free MonoN in solution . To achieve this , we devised a crosslinking assay in which the ability of a nucleosome immobilized on a solid resin to bind to a free nucleosome could be tested ( Figure 4A ) . We assembled MonoN with 5’ biotinylated 601 DNA containing a 20 bp linker 5’ to the 601 sequence to allow sufficient space and flexibility of the nucleosome away from the solid support . The reconstituted biotinylated nucleosome was conjugated to streptavidin magnetic beads and incubated with Sir3 or Sir3ΔwH , either alone or in combination with Sir4CC ( Figure 4A ) . This was followed by incubation with Alexa-647-labeled MonoN . The beads and their associated proteins and nucleosomes were then recovered by magnetic concentration and washed prior to elution of nucleosomal DNA from the beads with 2 M NaCl . The resulting supernatant was analyzed by gel electrophoresis , and the amount of pulled down Alexa-647 nucleosomal DNA was quantified by the intensity of its fluorescent band . The addition of Sir3 to immobilized nucleosomes promoted the recovery of free labeled nucleosomes and this recovery was not affected by the addition of Sir4CC ( Figure 4B , lanes 4 , 5 , and 6; Figure 4D ) , even at lower Sir3 concentrations that still mediated substantial bridging ( Figure 4E ) . In contrast , Sir3ΔwH , alone or in the presence of Sir4CC did not promote the recovery of free nucleosomes ( Figure 4B , lanes 7–9; Figure 4D ) . Western Blot analysis showed that similar amounts of proteins were loaded onto the resin ( Figure 4C ) , ruling out the possibility that differences in bridging activity of different proteins were caused by unequal loading of proteins onto nucleosomes . Therefore , this result indicated that Sir3 , through its wH domain , acts as a bridge linking mono-nucleosomes in solution . We note that the inability of Sir4CC to stimulate bridging , despite its effect in increasing the affinity of Sir3 and Sir3∆wH binding to DiN ( Figure 3F , H ) , may suggest that Sir3-Sir4CC interactions are too weak to endure the stringent test of the pull-down assay . In this assay , the concentration of nucleosomes relative to each other is much lower ( 100 nM ) than the effective local nucleosome concentration in the DiN EMSA . Furthermore , weak or highly dynamic complexes may partially disassemble during the wash steps of the assay . We suspect that while Sir4CC can readily link Sir3 bound to adjacent DNA-linked nucleosomes , the contact is not strong enough by itself to stably bring together separate Sir3-bound nucleosomes . A weak Sir4CC-Sir3 interaction in solution is consistent with most Sir3 not being associated with Sir4 in yeast extracts ( Moazed et al . , 1997; Rudner et al . , 2005 ) . In contrast , the wH-wH interactions can mediate relatively stable dimers in solution ( Oppikofer et al . , 2013 ) . Both H4K16 and H3K79 play important roles in silencing in S . cerevisiae ( Braunstein et al . , 1993; Johnson et al . , 1990; Ng et al . , 2002; van Leeuwen et al . , 2002 ) . Previous work showed that H4K16 acetylation ( H4K16ac ) and H3K79 methylation ( H3K79me ) each inhibits Sir3 binding to histone peptides and nucleosomes , but the difference in binding constants between unmodified and singly modified nucleosomes is rather modest ( Johnson et al . , 2009; Martino et al . , 2009; Swygert et al . , 2014; Wang et al . , 2013 ) . We quantified the effect of H4K16ac and H3K79me and , more importantly , the effect of co-existence of both modifications in the same nucleosome on Sir3 binding . As both H4K16ac and H3K79me are markers for euchromatin , and are deposited globally in S . cerevisiae ( Kimura et al . , 2002; Ng et al . , 2002; Suka et al . , 2002 ) , it is highly likely that nucleosomes within euchromatic regions harbor both histone modifications at the same time . We used the piccolo histone acetyltransferase ( HAT ) complex to acetylate H4K16 and the methyl-lysine analog ( MLA ) method to generate KC79me3 H3 histones ( Simon et al . , 2007 ) , which were then reconstituted into MonoN and DiN ( Figure 5—figure supplement 1 ) . We chose H3K79me3 for our binding studies because it has been shown that the trimethylated state of H3K79 is the predominant in vivo state ( Frederiks et al . , 2008; Ng et al . , 2002 ) . Consistent with previous results ( Johnson et al . , 2009; Martino et al . , 2009 ) , H4K16ac and H3KC79me3 each decreased the affinity of Sir3 binding for MonoN by 4–5 fold , with KD values of about 4 . 5 μM and 5 . 0 μM , respectively ( Figure 5A–D , Table 1 ) . Each modification also reduced the affinity of Sir3 for DiN with KD values of 0 . 7 µM and 0 . 8 µM , respectively ( Figure 5E–F and Figure 5—figure supplement 2A , B , Table 1 ) , which represented about a 5-fold decrease in affinity relative to unmodified DiNs ( Figure 1E , H ) . However , the combination of the two modifications inhibited Sir3 binding to MonoN and DiN , so that we could not obtain specifically shifted bands at the highest Sir3 concentration ( 11 μM ) used in the assay ( Figure 5—figure supplement 2C , D , Table 1 ) . This observation suggests that the two modifications act together to strongly inhibit Sir3 binding . We noted slightly up-shifted bands in ac/me-modified DiN at Sir3 concentrations above 3 . 6 µM . As the up-shift continues to increase with higher Sir3 concentrations , but never reaches the specifically shifted band observed for unmodified Sir3-DiN complex , we surmise that the observed binding is likely nonspecific . 10 . 7554/eLife . 17556 . 015Figure 5 . H4K16 acetylation and H3K79 methylation act together to strongly inhibit the binding of Sir3 to nucleosomes . ( A , B ) Comparison of Sir3 binding to unmodified and H4K16ac MonoN ( A ) , or unmodified and H3KC79me3 MonoN ( B ) by EMSA . We note that Sir3 bound H4K16ac and H3KC79me3 MonoN and DiN shift to a lower position than Sir3 bound unmodified nucleosomes , suggesting that Sir3 binds to modified nucleosomes in a different conformation than to unmodified nucleosomes . ( C ) Binding curves from three experiments performed as in ( A ) were fitted with the Hill Equation . ( D ) Binding curves from three experiments performed as in ( B ) were fitted with the Hill Equation . ( E ) Comparison of Sir3 binding to unmodified and H4K16ac DiN . Binding curves from three experiments performed as in Figure 5—figure supplement 2A were fitted with the Hill Equation . ( F ) Comparison of Sir3 binding to unmodified and H3KC79me3 DiN . Binding curves from three experiments performed as in Figure 5—figure supplement 2B were fitted with the Hill Equation . In C–F , curves in lighter colors show model fits to binding data of unmodified nucleosomes for visual comparison . Data and fits are shown in Figure 1E . Blue and Red dotted lines indicate the apparent KD for Sir3 binding to MonoN and DiN , respectively . See Table 1 for parameter values . Note that the titrations of modified nucleosomes are not fully saturated at the highest concentrations tested and the calculated apparent affinities may be overestimated . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 01510 . 7554/eLife . 17556 . 016Figure 5—figure supplement 1 . Reconstitution of H4K16ac and H3KC79me3 nucleosomes . ( A ) Coomassie stained gel showing purified Piccolo HAT complex . ( B ) Quantification , by quantitative Western Blot , of Piccolo HAT mediated acetylation on unmodified nucleosomes . Increased amounts of Piccolo HAT complex were titrated against a constant amount of unmodified nucleosomes . The resulting H4K16ac level was quantified using anti-H4K16ac antibody . There was no increase in acetylation signal with increasing amount of Piccolo HAT , indicating saturating levels of acetylation . ( C ) Quantification of acetylation reactions as in ( B ) but on H3KC79me3 nucleosomes . ( D ) Nucleosome acetylation causes a slight shift of nucleosome , consistent with previous findings . Modified nucleosomes are completely shifted , indicating complete acetylation . ( E ) Mass Spectrometry verified the complete alkylation of H3KC79me3 . ( F ) Reconstitution of H3KC79me3 MonoN . ( G ) Reconstitution of H3KC79me3 DiN . Nucleosome reconstitutions that produced a single band were used for biochemical assays . ( H ) Assembly of unmodified , singly , or doubly modified mono- and DiN . H4K16ac DiN gel was taken from lanes 5 and 7 in ( D ) ; H3KC79me3 MonoN gel was taken from lanes 1 and 6 in ( F ) ; H3KC79me3 DiN gel was taken from lanes 1 and 5 in ( G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 01610 . 7554/eLife . 17556 . 017Figure 5—figure supplement 2 . Sir3 binding to singly and doubly modified H4K16 acetylated and H3K79 methylated MonoN and DiN . ( A ) EMSA experiments comparing Sir3 binding to the unmodified and H4K16ac DiN . ( B ) EMSA experiments comparing Sir3 binding to unmodified and H3KC79me3 DiN . We note that Sir3 bound H4K16ac and H3KC79me3 MonoN and DiN ( Figure 5A–B ) shift to a lower position than Sir3 bound unmodified nucleosomes , suggesting that Sir3 binds to modified nucleosomes in a different conformation than to unmodified nucleosomes . ( C ) EMSA experiments showing Sir3 binding to doubly modified H4K16ac/H3KC79me3 MonoN . There was about 10% MonoN shift at 10 . 7 μM Sir3 . ( D ) EMSA experiments showing Sir3 binding to unmodified and doubly modified H4K16ac/H3KC79me3 DiN . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 017 We next investigated the relative contribution of each the Sir3wH and the Sir4CC domains to Sir3 association with chromatin in vivo . To this end , we performed chromatin immunoprecipitation followed by high throughput sequencing ( ChIP-seq ) using an antibody that recognizes Sir3 . The results confirmed previous studies , which have shown that each domain is required for Sir3 association with silent loci . As shown in Figure 6A ( tracks 1–5 ) for the left telomere of chromosome 1 ( Chr1L ) , the deletion of either the wH domain ( sir3ΔwH ) , or a mutation of the Sir4CC that abolishes its interaction with Sir3 ( sir4-I1311N ) ( Chang et al . , 2003; Rudner et al . , 2005 ) , resulted in a strong loss of the Sir3 ChIP signal . Furthermore , the association of Sir3 with Chr1L in the double mutant ( sir3ΔwH , sir4-I1311N ) was reduced to the same level observed in sir3Δ cells . The results of the double mutant strain hinted at a greater loss in binding than in cells carrying each of the single mutations . However , our ChIP-seq assays were not sensitive enough to reliably detect possible differences between single and double mutant strains when Sir3 was expressed at wildtype levels ( see Figure 6—figure supplement 1A for the result of all telomeres ) . 10 . 7554/eLife . 17556 . 018Figure 6 . Sir3wH and Sir4CC are both required for SIR complex spreading in vivo but play partially redundant roles when Sir3 is overexpressed . ( A ) Normalized ChIP-seq read densities for Sir3 at the left telomere of chromosome 1 ( Chr1L ) from cells with the indicated genotypes . The core X element ( black box ) and the ACS within ( C-ACS filled red box ) are indicated below the tracks . Sir3 was overexpressed from a 2 micron plasmid ( SIR3 2 µ ) . Colorshaded areas indicate regions assayed by ChIP-qPCR , shown in panel C . Full scale plots of WT and WT/Sir3 2µ strains are shown in Figure 6—figure supplement 1B . ( B ) Ensemble plot of Sir3 ChIP-seq signals with Sir3 overexpression , from 30 subtelomeres , excluding TEL01R and TEL13R , aligned at C-ACS . ChIP signal at all 30 subtelomeric regions was summed up , normalized to the sir3Δ sample , and plotted as a function of distance from C-ACS . Negative values are towards the chromosome end , and positive values are towards the centromere . Similar plot with endogenous Sir3 expression level is shown in Figure 6—figure supplement 1A . ( C ) Sir3 ChIP-qPCR results from cells with the indicated genotypes normalized to sir3∆ cells . Error bars indicate the standard deviation of three biological replicates . PCR primer sets ( Table 4 ) were chosen to assay regions shown with corresponding colors in panel A . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 01810 . 7554/eLife . 17556 . 019Figure 6—figure supplement 1 . The requirement of Sir3wH and Sir4CC for the association of Sir3 with heterochromatin in vivo . ( A ) Ensemble plot of Sir3 ChIP-seq signals with endogenous Sir3 expression level , from 30 subtelomeres , excluding TEL01R and TEL13R , aligned at C-ACS . Negative values are towards chromosome end , and positive values are towards centromere . ( B ) Normalized ChIP-seq read densities for Sir3 and overexpressed Sir3 ( Sir3 2 µ ) at the left telomere of chromosome 1 ( Chr 1 L ) . Full-scale plot for comparison with Figure 6A . ( C , E ) Ensemble plots of Sir3 ChIP-seq signals , without ( C ) and with ( E ) Sir3 overexpression , from 30 subtelomeres , excluding TEL01R and TEL13R , aligned at chromosome ends . ChIP signal at all 30 subtelomeric regions was summed up , normalized to the sir3Δ sample , and plotted as a function of distance from chromosome end . ( D , F ) Ensemble plots of Sir3 ChIP-seq signals , without ( D ) and with ( F ) Sir3 overexpression , from 30 subtelomeres , excluding TEL01R and TEL13R , aligned at C-ACS . Negative values are towards the chromosome end , and positive values are towards the centromere . ChIP signal at all 30 subtelomeric regions was summed up , normalized to the signal from sir3Δ cells , and plotted as a function of distance from C-ACS . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 019 Therefore , we next performed these experiments in cells with Sir3 overexpressed from a high copy 2 micron plasmid ( SIR3 2 µ ) . Sir3 overexpression has been suggested to partially bypass the requirement for Sir4 in spreading of silencing and may provide a way to assess possible Sir4-independent contribution of the wH domain to spreading ( Renauld et al . , 1993; Strahl-Bolsinger et al . , 1997 ) . Consistent with previous studies ( Strahl-Bolsinger et al . , 1997 ) , we observed domains of Sir3 association at telomeres that were expanded upon Sir3 overexpression ( Figure 6 , compare tracks 1 and 6 , Figure 6—figure supplement 1B ) . Under conditions of Sir3 overexpression , Sir3 remained detectably associated with Chr1L in both sir3ΔwH and sir4-I1311N cells , albeit at reduced levels ( Figure 6A , tracks 7 and 8 ) . The degree of association was higher in sir4-I1311N than in sir3ΔwH cells , suggesting that wH-mediated Sir3 dimerization played a more important role in spreading than the association of Sir3 with Sir4 . In double mutant cells , however , Sir3 binding was reduced to background levels ( Figure 6A , track 9 ) . Therefore , Sir3wH and Sir4CC domains each make important contributions to Sir3 spreading , which become partially redundant when Sir3 is overexpressed . The data for Chr1L was representative of Sir3 association with most telomeric regions . Ensemble plots of the ChIP-seq data , aligned either to chromosome ends ( Figure 6—figure supplement 1E ) or to the ACS sequence in the subtelomeric X elements ( Figure 6B and Figure 6—figure supplement 1F ) , previously suggested to be the initiation site of SIR complex spreading in subtelomeric regions ( Ellahi et al . , 2015; Pryde and Louis , 1999; Radman-Livaja et al . , 2011; Tham and Zakian , 2002; Thurtle and Rine , 2014; Zill et al . , 2010 ) , supported our conclusions above . Finally , Sir3 ChIP-qPCR analysis of strains shown in Figure 6B at four loci across chromosome 1 confirmed greater Sir3 binding deficiency in the double mutant and the more pronounced effect of Sir3∆wH compared to Sir4I1311N mutation ( Figure 6C ) . We conclude that the effects of Sir3wH and Sir4CC domains on Sir3 spreading in vivo correlate with their respective contributions to the stability of the internucleosomal Sir3 bridge in vitro .
The interrupted spreading mechanism described here is distinct from previously proposed oligomerization-based models . For example , Swi6/HP1 in S . pombe has been proposed to associate with chromatin via chromo shadow domain-mediated dimerization across adjacent nucleosomes as well as chromo domain-chromo domain interactions on the same nucleosome ( Canzio et al . , 2011 ) . This 'sticky ends' mode of binding would result in the formation of continuous Swi6-Swi6 interactions across silent chromatin domains . However , oligomerization beyond stable dimers proved to be weak , even in very high ( 20 µM or higher ) Swi6/HP1 concentrations ( Canzio et al . , 2011 ) , and absent in mouse HP1β , even at 30 µM concentrations ( Muller-Ott et al . , 2014 ) . Therefore , oligomerization may not contribute to the in vivo mechanism of action of HP1 proteins . In fact , subsequent re-analysis of the Swi6/HP1 binding isotherms ( Canzio et al . , 2011 ) revealed that a simpler model , lacking direct interactions among neighboring Swi6/HP1 dimers that were suggested in the original study , could explain the observations ( Teif et al . , 2015 ) . Furthermore , recent studies of the mammalian HP1α and HP1β proteins suggest that they associate with chromatin primarily as nucleosome bridges ( Hiragami-Hamada et al . , 2016; Kilic et al . , 2015 ) . Similarly , Sir3 can form oligomers in vitro and this oligomerization was previously suggested to mediate SIR complex spreading along chromatin ( Liou et al . , 2005; McBryant et al . , 2006 ) . In contrast , the discontinuous mode of binding revealed here suggests that recruitment of new SIR complexes does not rely on contacts between newly recruited and already bound complexes , but instead requires association of Sir3 with a pair of nucleosomes unmodified at H4K16 and H3K79 . We note that Sir3 protein , at sub-micromolar concentrations and in buffers containing physiological salt concentrations , is found largely as a monomers and dimers ( McBryant et al . , 2006; Swygert et al . , 2014 ) . Therefore , the ability of Sir3 to form higher order oligomers in low salt and at high concentrations may not play a role in its cellular function . However , it remains possible that Sir3 oligomerization is regulated by unknown factors in vivo that may be absent in our experiments , or may result from increased effective concentration of nucleosome-bound Sir3 . Active and silent chromatin regions are associated with stereotypical patterns of histone post-translational modifications with each type of region containing several different modifications ( Ernst et al . , 2011; Filion et al . , 2010; Kharchenko et al . , 2011; Taverna et al . , 2007 ) . The combined action of multiple histone modifications potentially provides better binding specificity ( Du and Patel , 2014; Ruthenburg et al . , 2007 ) . In a few cases , combinations of modification states are recognized by the same ( Eustermann et al . , 2011; Iwase et al . , 2011; Moriniere et al . , 2009; Ramón-Maiques et al . , 2007 ) , or different domains in a single protein or different subunits of a complex ( Dhalluin et al . , 1999; Li et al . , 2006; 2007b; Rothbart et al . , 2013; Ruthenburg et al . , 2011 ) . In budding yeast , previous studies indicate that H4K16 acetylation and H3K79 methylation each reduces Sir3 binding to nucleosomes , and therefore negatively regulate heterochromatin formation ( Braunstein et al . , 1996; Johnson et al . , 2009; Liou et al . , 2005; Martino et al . , 2009; Onishi et al . , 2007; Swygert et al . , 2014; van Leeuwen et al . , 2002 ) . Moreover , recent structural analysis of Sir3BAH bound to MonoN indicates that both the H4K16 and H3K79 regions interact directly with the BAH domain ( Armache et al . , 2011; Arnaudo et al . , 2013; Wang et al . , 2013 ) . Consistent with these findings , H4K16 acetylation and H3KC79 trimethylation each reduces the affinity of Sir3 binding to nucleosomes around 4-fold ( this study ) , and substitution of H4K16 with Q reduces the affinity of Sir3 for oligo-nucleosomes to a similar degree ( Swygert et al . , 2014 ) . More strikingly , when both modifications were present in the same nucleosome , Sir3 binding affinity is reduced to a level that could not be measured in our experiments ( >11 µM ) . Since both H3K79 methylation and H4K16 acetylation are present at high levels in euchromatic genes and absent in silent chromatin regions ( Kurdistani et al . , 2004; van Leeuwen et al . , 2002 ) , their combined action , together with the strong cooperative binding to surfaces on two unmodified nucleosomes , is likely to be sufficient for the specific localization of Sir3 to silencer-proximal nucleosomes lacking H3K79 and H4K16 modifications . Cooperative modes of association , relying on multiple weak interactions rather than one strong interaction interface , are widespread in biology and contribute to modularity of regulatory networks , their robustness against noise , and their ability to display bistability ( Ptashne , 2009; Williamson , 2008 ) . The cooperative mechanism of Sir3 binding to unmodified DiN units strongly biases Sir3 binding away from association with randomly occurring deacetylated nucleosomes that may arise throughout the genome . Independent measurements have reported a wide range of Sir3 and Sir4 molecules per cell ( Chong et al . , 2015; Gerber et al . , 2003; Ghaemmaghami et al . , 2003; Kulak et al . , 2014 ) . Since Sir3 is primarily concentrated inside the nucleus ( volume ~2 . 8 µm3 ) ( Jorgensen et al . , 2007 ) , Sir3 intranuclear concentration may be roughly estimated at 0 . 4–0 . 8 µM . In this range of concentrations , unmodified nucleosome pairs are ~50–80 fold more likely to be bound by Sir3 than isolated single nucleosomes ( see Materials and methods ) . This selectivity is in large part due to binding cooperativity: even if Sir3 bound to di-nucleosomes with 10-fold higher affinity than to mono-nucleosomes , in the absence of cooperativity , di-nucleosomes would be favored over isolated nucleosomes by only ~12 fold . Therefore , cooperative association of reader proteins with nucleosome pairs may endow heterochromatin domains with robustness against random noise .
S . cerevisae strains and plasmids used in this study are listed in Table 3 . All deletions and mutations were confirmed by PCR and sequencing . Epitope-tagged strains were constructed by a PCR-based gene targeting method ( Longtine et al . , 1998; Rudner et al . , 2005 ) . 10 . 7554/eLife . 17556 . 021Table 3 . List of yeast strains and plasmids used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 021Name Yeast strain genotype Source W303-1a ( SF1 ) MATa ade2-1 can1-100 his3-11 leu2-3 , 112 trp1 ura3-1 GAL J . RineSF4sir3△::TRP1 in SF1J . RineDMY4350sir3△wH::TRP1 in SF1This StudyADR2973sir4-I1311N in SF1 ( Longtine et al . , 1998; Rudner et al . , 2005 ) DMY4351sir4-I1311N sir3△wH::TRP1 in SF1This StudyDMY3315W303-1a sir3Δ::KanR hmrΔE::TRP1 TELVII-L::URA3 ( Buchberger et al . , 2008 ) Name Plasmid genotype Source pRS315CEN/ARS LEU2 ( Sikorski and Hieter , 1989 ) pJR104 ( pDM602 ) SIR3 under endogenous promoter in YEp24 ( Kimmerly and Rine , 1987 ) pDM1798sir3△wH under endogenous promoter in YEp24This StudypDM832Sir3-3XFLAG under endogenous promoter in pRS315 ( Buchberger et al . , 2008 ) pDM1799Sir3△wH-3XFLAG under endogenous promoter in pRS315This StudyAll deletions and mutations were confirmed by PCR and sequencing . Epitope-tagged strains were constructed by a PCR-based gene targeting method ( Longtine et al . , 1998; Rudner et al . , 2005 ) . Sir3-3XFLAG and BAH-3XFLAG were purified from S . cerevisiae as described previously ( Buchberger et al . , 2008; Huang et al . , 2006; Liou et al . , 2005 ) . Sir3ΔwH-3XFLAG was constructed by deleting the winged helix ( wH ) region on pDM1009 ( GAL-Sir3-3XFLAG 2μ plasmid ) , and purified by the same FLAG purification protocol . Sir3-CBP and Sir3∆wH-CBP were prepared by TEV cleavage of affinity purified Sir3-TAP and Sir3∆wH-TAP proteins , followed by anion-exchange and size-exclusion chromatography as described previously ( Buchberger et al . , 2008; Huang et al . , 2006; Liou et al . , 2005 ) . We tested a number of purification strategies to optimize protein yield and purity , including the addition of 500 mM KCl and 250 mM guanidine hydrochloride in size exclusion chromatography buffers . These conditions strongly decrease Sir3 oligomerization without affecting its tertiary structure ( McBryant et al . , 2006 ) . While high salt and small amounts of chaotropic agent increased purification yield , none of our purification schemes affected the final purity or binding behavior of Sir3 to nucleosome templates . Sir4CC ( 1198–1358 ) was cloned into the pET47b ( + ) plasmid , and the protein was purified from E . coli by Ni+-affinity purification , followed by PreScission protease cleavage and gel filtration , to remove the His tag . A minor degradation product co-purified with Sir4CC ( 1198–1358 ) . The mass Spectrometry analysis identified this fragment as Sir4 ( 1242–1358 ) , which covers the entire Sir4CC core domain , and should therefore have the same Sir3 binding activity as the larger Sir4 ( 1198–1358 ) . Sir4CC ( 1198–1358 ) was also cloned into pGEX6P-1 , and the resulting GST-Sir4CC was affinity purified from E . coli . S . cerevisiae histones were overexpressed and purified from E . coli as previously described ( Johnson et al . , 2009 ) . H3KC79me3 histone was prepared as previously described ( Simon et al . , 2007 ) . Histone H2A K120C was prepared by standard PCR-based mutagenesis and reacted with EZ-Link Maleimide-PEG2-Biotin ( Thermo Fisher Scientific ) following the manufacturer’s protocol . The catalytic Piccolo subcomplex of the NuA4 histone acetyltransferase ( HAT ) complex was purified from E . coli as previously described ( Barrios et al . , 2007; Johnson et al . , 2009; Selleck et al . , 2005 ) . MonoN and nucleosome arrays were reconstituted using gradient salt dialysis as described previously ( Luger et al . , 1999 ) with modifications for arrays encompassing more than 2 nucleosomes , according to Huynh et al . ( Huynh et al . , 2005 ) . The MonoN DNA template contains the 147 bp 601 positioning sequence ( Lowary and Widom , 1998 ) . The array DNA templates contain defined number of direct repeats of the 601 sequence , separated by a 20 bp linkers . The 601 tetramer template also contains 20 bp DNA before and after the array . The biotinylated nucleosomal DNA template contains the 601 sequence , with an extra 20 bp linker added upstream by PCR reactions using a 5’ biotinylated primer ( Integrated DNA Technologies ) . The Alexa-647 labeled MonoN DNA template was also made by PCR reactions using 5’ Alexa-647 labeled primer ( Integrated DNA Technologies ) . Internucleosomal linker DNA in the S . cerevisiae silent chromatin regions has a heterogeneous length distribution ( Brogaard et al . , 2012; Ravindra et al . , 1999; Weiss and Simpson , 1998 ) . We chose the linker DNA to be 20 bp , which reflects the average linker DNA length in S . cerevisiae ( Arya et al . , 2010 ) . Nucleosome acetylation was carried out as described previously ( Johnson et al . , 2009 ) . Briefly , nucleosomes were incubated with 1/10th molar ratio of the Piccolo HAT complex and 100X molar excess of acetyl-CoA in the HAT buffer ( 20 mM Tris . HCl , pH 8 . 0 , 50 mM KCl , 5% glycerol , 5 mM DTT , 1 mM PMSF , and 0 . 5 mg/ml BSA ) at 30˚C for 1 hr . The completion of acetylation was assessed by the complete shift of the nucleosome band , and by quantitative Western blot using antibody against H4K16ac , where saturated signal was achieved . DiN were incubated with 10U of either ScaI or AluI restriction enzyme ( New England Biolabs ) in 1XNEB CutSmart Buffer , at 37°C for 1 hr . The resulting digestion products were separated on native polyacrylamide gels , and visualized by staining with ethidium bromide . Different amounts of Sir3 protein were incubated with 3 nM MonoN or DiN in binding buffer ( 25 mM Tris . HCl ( pH 7 . 5 ) , 50 mM NaCl , and 5 mM DTT ) at 4°C for 1 hr . Samples were then run on native polyacrylamide gels , stained with SYBR Gold ( Invitrogen ) , visualized on a Typhoon FLA7000 imager ( GE Healthcare ) , and quantified using ImageQuant software . Sir3 binding to nucleosomes was quantified by the decrease in the intensity of the unbound nucleosome band . The apparent KD ( protein concentration at transition midpoint ) and Hill coefficient for each binding reaction was calculated by fitting the binding curves with the Hill Equation ( see Analysis of binding cooperativity section below ) using MATLAB ( Mathematica ) . BLI sensors were pre-incubated in loading buffer ( 20 mM Tris . HCl pH 7 . 5 , 1 mM EDTA , 200 mM NaCl , 1 mM DTT , 0 . 5 mg/ml BSA , 0 . 02% Tween-20 ) before incubation with 10 nM biotinylated MonoN or DiN in the same buffer for 5–10 min . To eliminate artifacts due to surface crowding and ligand walking , nucleosome binding to sensors were first optimized to determine conditions where sensors were loaded at <25% capacity and binding behavior was insensitive to nucleosome density on sensor . All sensors for each titration experiment were loaded with the same number of nucleosomes ( ± 5% , as monitored by BLI loading signal ) . Binding experiments were performed using Octet RED384 system ( Pall Life Sciences ) at 30˚C in the same buffer used for loading nucleosomes , except that the NaCl concentration was reduced to 50 mM . To determine the effect of nonspecific protein-sensor interactions , we measured binding signal of empty sensors in various concentrations of Sir3 as well as nucleosome-loaded sensors in buffer solution without Sir3 . Nonspecific Sir3-sensor interactions gave rise to linearly rising baselines , which were subsequently subtracted from the signal using a linear extrapolation procedure . Measurements were repeated with Sir3 protein from at least two independent purifications ( using FLAG or TAP tags ) and two separate reconstitutions of nucleosomes templates . For larger nucleosome arrays , two biotinylation densities ( all or 33% of histone octamers biotinylated ) were tested in the reconstitution of nucleosome arrays to ensure that sensor immobilization does not interfere with Sir3 binding . Addition of 0 . 5 µM competitor DNA ( salmon sperm genomic DNA physically sheared to average 150–200 bp length ) did not affect Sir3 binding to nucleosomes . Therefore , average binding profiles shown in Figure 1G–H include experiments performed with and without competitor DNA . Furthermore , the presence of 20 bp extra linker DNA on MonoN did not affect binding of Sir3 ( Figure 1—figure supplement 3C ) . Consistent with the above observations , loading of at least 10-fold higher free DNA on biosensors was necessary to obtain measurable signal changes as a result of Sir3 binding to free DNA . We therefore concluded that , the weak affinity of Sir3 for nonspecific binding to free DNA does not contribute to our nucleosome binding assays . Association and dissociation rates and amplitudes were calculated by nonlinear least-square fitting of data with mono- or bi-exponential saturation models ( see Measurement of Sir3 binding kinetics with BLI section below ) . The amplitudes were normalized and plotted versus protein concentration to reconstruct titration curves which were fit with the Hill equation , or when possible , with a model describing binding to identical independent sites ( see Analysis of binding cooperativity section below ) . All model fitting procedures were performed by the nonlinear least squares method implemented in MATLAB ( Mathematica ) . The following equations were used to quantify binding experiments , as indicated in the main text . Simple association and dissociation of Sir3 with nucleosomes is described by mono-phasic exponential functions: Association: Signal=Aon . ( 1− e−kobst ) Dissociation: Signal=Aoff . e−kofft+B , where Aonand kobs are the amplitude and rate of saturation and Aoff and koff are the amplitude and rate of dissociation . B represents the baseline signal . Protein-nucleosome association rates ( kon ) were calculated from kobs and koff values kon= kobs−koffi . [S] , where i is the presumed number of binding sites for the protein on nucleosome , and [S] is protein concentration . While this model was sufficient to quantify binding of Sir3 to di-nucleosomes and larger templates ( Figure 1—figure supplement 4B , C ) , we observed that it fails to capture Sir3 association and dissociation with mono-nucleosomes . Instead , a biphasic binding model , indicating two binding processes with different rates , was minimally required to fit the data ( Figure 1—figure supplement 4D , E ) :Signal=Aon , 1 . ( 1− e−kobs , 1t ) +Aon , 2 . ( 1− e−kobs , 2t ) Signal=Aoff , 1 . e−koff , 1t+Aoff , 2 . e−koff , 2t+B Shortening of Sir3 incubation time with MonoN from 60 s to 20 s or biotinylation of nucleosomes on DNA instead of histone H2A did not affect the biphasic dissociation of Sir3 , ruling out heterogeneities in Sir3 or nucleosome preparations as the cause of biphasic binding behavior . Since Sir3 can engage multiple sites on nucleosomes through its BAH or AAAL domain ( Hecht et al . , 1995; Liou et al . , 2005; Martino et al . , 2009; Onishi et al . , 2007 ) , the two binding phases may represent different modes of Sir3-nucleosome interactions . Consistent with this hypothesis , the Sir3BAH domain alone bound to and dissociated from mono-nucleosomes with a single rate that was comparable to the fast rate of Sir3 binding and dissociation ( Figure 1—figure supplement 4F , G , Table 2 ) . More importantly , the apparent affinity of Sir3BAH for MonoN ( apparent KD = 1 . 4 ± 0 . 1 µM , Figure 1G ) closely resembled that of the fast forming fraction of the Sir3-MonoN complex ( KD = 1 . 4 ± 0 . 1 µM , Figure 1H ) , while the slow-forming fraction showed a considerably lower affinity ( apparent KD > 4 µM ) . Deletion of both AAAL and wH domains , but not the wH domain alone , caused a strong ( >3 fold ) decrease in Sir3-MonoN complex half-life ( Figure 1—figure supplement 4A ) , confirming either a direct or synergistic contribution of AAAL domain to the interaction of Sir3 with mono-nucleosomes . Therefore , thermodynamic , kinetic , and domain deletion experiments reveal that in addition to the BAH-mediated binding , Sir3 can engage mono-nucleosomes in other modes , most likely through the AAAL domain ( Ehrentraut et al . , 2011; Hecht et al . , 1995 ) , which are precluded in binding to larger nucleosome arrays ( [Sir3] <0 . 5 µM ) . Therefore , we compared binding of Sir3 to larger nucleosome templates with its fast phase of binding to MonoN ( Figure 1H ) . Nucleosomes assembled with biotinylated DNA were conjugated to Dynabeads M-280 streptavidin ( Invitrogen ) at RT for 1 hr with rotation , using 36 μl of beads slurry per μg of nucleosomes in the binding buffer ( 20 mM Tris . HCl ( pH 7 . 5 ) , 0 . 3 mM EDTA , 50 mM NaCl , 10% glycerol , 5 mM DTT , 1 mg/ml BSA , and 0 . 02% NP–40 ) . Bead-conjugated nucleosomes were washed , and resuspended in equal volume of binding buffer as the initial volume of beads taken . Equal amount of conjugated nucleosomes , in a final concentration of 100 nM , was added to tubes containing Sir3 , Sir3/Sir4CC , Sir3ΔwH , Sir3ΔwH/Sir4CC , Sir4CC , or buffer alone , and incubated with rotation at 4°C for 1 hr . The concentration of Sir3 proteins was 2 μM in the case of high protein concentration binding assay , and 200 nM in the case of low protein concentration binding assay . Sir4CC was in 2X molar excess of Sir3 proteins . Subsequently , Alexa-647 labeled nucleosomes were added into each reaction at a final concentration of 100 nM , and reactions were incubated for another 1 hr at 4°C . Finally , the beads were washed twice in the binding buffer before magnetic concentration . Alexa-647 nucleosomal DNA from the crosslinked nucleosomes was stripped from the beads with 2M NaCl , separated on native polyacrylamide gels , and quantified by the fluorescent intensity of the band . Cells were cultured overnight in YEPD medium , or selective media for cells harboring overexpression plasmids ( YEp24 2 µ plasmid with Sir3 or Sir3△wH expressed from Sir3 endogenous promoter ) , diluted into fresh media to OD600 = 0 . 4 , and harvested at late log phase ( OD600 = 1 . 5 ) . Cells were fixed with 1% formaldehyde for 15 min at room temperature ( RT ) , then quenched with 130 mM glycine for 5 min at RT , harvested by centrifugation , washed twice with TBS ( 50 mM Tris . HCl pH 7 . 6 , 150 mM NaCl ) , and flash frozen . Cell pellets were resuspended in 600 µl lysis buffer ( 50 mM HEPES-KOH pH 7 . 5 , 150 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 0 . 1% Na-Deoxycholate , 0 . 1% SDS , 1 mM PMSF , protease inhibitor tablet ( Roche ) ) , and disrupted by bead beating ( MagNA Lyser , Roche ) for 6 × 30 s at 4500 rpm with 0 . 5 mm glass beads . Tubes were punctured and the flow-through was collected in a new tube by centrifugation . After sonication for 3 × 20 s at 40% amplitude ( Branson Digital Sonifier ) , the extract was centrifuged ( Eppendorf 5415R ) for 15 min at 13 , 000 rpm . The soluble chromatin was then transferred to a fresh tube . Sir3 antibody ( Rudner et al . , 2005 ) was preincubated with washed Dynabeads Protein A , and for each immunoprecipitation , 2 μg antibody coupled to 100 μl beads was added to soluble chromatin . Samples were incubated for 2 hr at 4°C with rotation , after which the beads were collected on magnetic stands , and washed 3 times with 1 ml lysis buffer and once with 1 ml TE , and eluted with 100 μl preheated buffer ( 50 mM Tris . HCl pH 8 . 0 , 10 mM EDTA , 1% SDS ) at 65°C for 15 min . The eluate was collected , and 150 μl 1XTE/0 . 67% SDS was added . Immunoprecipitated samples were incubated overnight at 65°C to reverse crosslink , and treated with 50 μg RNase A at 37°C for 1 hr . 5 μl proteinase K ( Roche ) was added and incubation was continued at 55°C for 1 hr . Samples were purified using a PCR purification kit ( Qiagen ) . Libraries for Illumina sequencing were constructed as described previously ( Wong et al . , 2013 ) , starting with ~5 ng of immunoprecipitated DNA fragments . Each library was generated with custom-made adapters carrying unique barcode sequences at the ligating end . Barcoded libraries were mixed and sequenced with Illumina HiSeq 2500 . Raw reads were separated according to their barcodes and mapped to the S . cerevisiae S288C genome using Bowtie . Mapped reads were normalized to reads per million and visualized in IGV . Ensemble plots aligned at chromosome ends were generated by aligning all 30 telomeres , excluding TEL01R and TEL13R , at chromosome ends , and calculating the total ChIP-seq signal across 14 kb regions towards the centromere . The cumulative ChIP-seq reads was then normalized , on a per-base basis , to that of the Δsir3 sample . Ensemble plots aligned at the ACS within the core X element ( C-ACS ) were generated in a similar manner , except that each telomere was aligned at C-ACS , and total ChIP-seq signal was computed for 14 kb in each direction . ChIP was performed essentially as described in the ChIP-seq above , except for the modification described below . Extracts were sonicated for 3 × 20 s at 50% amplitude using a sonicator ( Branson Digital Sonifier ) . After centrifugation for 15 min at 13 , 000 rpm , the soluble chromatin was transferred to a fresh tube and normalized for protein concentration by the Bradford assay . For each immunoprecipitation , 2 µg Sir3 antibody coupled to 30 µl Dynabeads Protein A was used . Immunoprecipitation , washes , elution and reverse crosslinking were performed as described in the ChIP-seq section . 60 µg glycogen , 44 µl of 5M LiCl and 250 µl TE were added and the samples were extracted with phenol/chloroform , and ethanol precipitated . DNA was resuspended in 100 µl of 10 mM Tris pH 7 . 5 , 50 mM NaCl . 2 . 5 µl of immunoprecipitated DNA was used for qPCR . Primers used are listed in Table 4 . qPCR was performed in the presence of SYBR Green using an Applied Biosystems 7900HT light cycler . Fold enrichments were calculated using the ∆CT method and average values of three biological replicates were normalized to the cup1+ gene . Enrichment relative to the control ( sir3∆ ) was calculated after normalization . 10 . 7554/eLife . 17556 . 022Table 4 . List of qChIP PCR primers used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 17556 . 022Primer Sequence RB139 ( C-ACS distal 1 , F ) TTC TGC CCA TAC GAT ACC TRB140 ( C-ACS distal 1 , R ) AGT TAC GCG TGC TAC ATT ACRB141 ( C-ACS distal 2 , F ) GTT CTA CTG ACA GGA TGG AAT AGRB142 ( C-ACS distal 2 , R ) GTG AAG GAG GGC ATG AAA TRB143 ( C-ACS proximal 1 , F ) CGT ACT TAC ACA GGC CAT ACRB144 ( C-ACS proximal 1 , R ) GTT TGA GCC ACT ACC GTA TTARB145 ( C-ACS proximal 2 , F ) CTT GTG GTA GCA ACA CTA TCARB146 ( C-ACS proximal 2 , R ) GGC CTG TGT AAG TAC GAA AT The Boltzmann distribution describes the probability distribution of a system consisting of multiple free energy states . For a system consisting of n states , the probability pi of a given state i is calculated aspi=e−ΔGikT∑i=1ne−ΔGikT , where ΔGi is the Gibbs free energy of the state i with respect to a common reference state , k is the Boltzmann constant , and T is absolute temperature . In our analysis , the system consists of three types of Sir3-nucleosome complexes: unmodified di-nucleosome units , unmodified isolated MonoN , and acetylated nucleosomes . For each complex state ( i ) , the free energy ( relative to the unbound state ) at a given Sir3 concentration [S] isΔGi=RTlnKD=RTln ( funbound1−funbound ) =−niRTln ( [S]KD , i ) , where KD , i and ni are experimentally determined apparent dissociation constants and Hill coefficients , respectively ( Table 1 ) . | Inside plant , fungi and animal cells , DNA wraps around disc-shaped histone proteins to form structures called nucleosomes . Chains of nucleosomes , each with a small stretch of DNA , help to package meters of genetic material into a compact form called chromatin in the cell’s nucleus . Changes to how chromatin is organized can affect how genes switch on and off . Critically , this allows cells to respond to changes in their environment and to develop into the many cell types required to build animals ranging from worms to humans . For example , specialized groups of proteins that bind to nucleosomes , spread along specific sites of chromatin and can change its structure into an inaccessible form called heterochromatin thereby switching off genes . Proteins that bind to specific nucleosomes control the spreading , gene activity , and even memory properties of heterochromatin . However , it is not clear how these proteins spread from their original binding point on the chromatin to other nucleosomes . Now , Behrouzi , Lu et al . show how heterochromatin spreads to form large , stable structures in budding yeast . Their experiments reveal that heterochromatin proteins attach to sites on neighbouring nucleosomes , forming bridges between them . These findings conflict a long-held view as they show that pairs of nucleosomes , rather than individual nucleosomes , are the natural binding partners for heterochromatin proteins . Also , because these proteins cannot bridge from one side of a nucleosome to the other , they are unlikely to form a continuous chain across multiple nucleosomes on the chromatin . Instead , Behrouzi , Lu et al . observed that a series of short bridges between nucleosomes helps heterochromatin to spread . To fully understand why bridging only happens between separate nucleosomes , the atomic structure of heterochromatin proteins bound to pairs of nucleosomes needs to be determined . In addition , it will be essential to develop more experimental methods to study the spreading of heterochromatin inside cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"biochemistry",
"and",
"chemical",
"biology"
] | 2016 | Heterochromatin assembly by interrupted Sir3 bridges across neighboring nucleosomes |
Hepatocytes are regarded as the only effective cell source for cell transplantation to treat liver diseases; however , their availability is limited due to a donor shortage . Thus , a novel cell source must be developed . We recently reported that mature rodent hepatocytes can be reprogrammed into progenitor-like cells with a repopulative capacity using small molecule inhibitors . Here , we demonstrate that hepatic progenitor cells can be obtained from human infant hepatocytes using the same strategy . These cells , named human chemically induced liver progenitors ( hCLiPs ) , had a significant repopulative capacity in injured mouse livers following transplantation . hCLiPs redifferentiated into mature hepatocytes in vitro upon treatment with hepatic maturation-inducing factors . These redifferentiated cells exhibited cytochrome P450 ( CYP ) enzymatic activities in response to CYP-inducing molecules and these activities were comparable with those in primary human hepatocytes . These findings will facilitate liver cell transplantation therapy and drug discovery studies .
Expansion of functional human hepatocytes is a prerequisite for liver regenerative medicine . Human hepatocytes are currently regarded as the only competent cell source for transplantation therapy ( Fisher and Strom , 2006 ) ; however , their availability is limited due to a shortage of donors . Moreover , the therapeutic application of hepatocytes is hampered by their inability to proliferate in vitro . To overcome this , researchers have sought to generate expandable cell sources as alternatives to primary hepatocytes . Such cell sources include embryonic stem cell- and induced pluripotent stem-cell-derived hepatic cells ( Carpentier et al . , 2014; Liu et al . , 2011; Takebe et al . , 2013; Woo et al . , 2012; Zhu et al . , 2014 ) , lineage-converted hepatic cells ( induced hepatic cells; ( Du et al . , 2014; Huang et al . , 2014 ) , and facultative liver stem/progenitor cells ( LPCs ) residing in adult liver tissue ( Huch et al . , 2015 ) . However , while primary hepatocytes efficiently repopulate injured mouse livers ( repopulation indexes ( RIs ) > 50% ) , the repopulation efficiency of these laboratory-generated hepatocytes is limited , with reported RIs generally less than 5% ( reviewed in Rezvani et al . , 2016 ) . Researchers have also attempted to expand primary human hepatocytes ( PHHs ) in vitro . Several studies reported the expansion of these cells ( Hino et al . , 1999; Shan et al . , 2013; Utoh et al . , 2008; Walldorf et al . , 2004; Yamasaki et al . , 2006 ) , suggesting that they are potentially applicable for transplantation therapy . However , the growth rate and proliferative lifespan of PHHs are limited . For example , Yoshizato’s group reported that PHHs can be cultured for several passages , but their growth rate is slow ( population doubling time of 20–300 days ) ( Yamasaki et al . , 2006 ) . This finding indicates that culture of PHHs must be improved for the clinical application of these cells . We recently reported that a cocktail of small molecule signaling inhibitors reprograms rodent adult hepatocytes into culturable LPCs , named chemically induced liver progenitors ( CLiPs ) ( Katsuda et al . , 2017 ) . Notably , rat CLiPs extensively repopulate chronically injured mouse livers without causing any tumorigenic features . Here , using the same strategy , we demonstrate that human infant hepatocytes can be also converted into proliferative LPC-like cells , which are named human CLiPs .
In a pilot study , we tested whether the combination of Y-27632 ( Y ) , A-83–01 ( A ) , and CHIR99021 ( C ) , the chemical cocktail used to reprogram rodent hepatocytes , also induced proliferation of commercially available cryopreserved adult PHHs ( APHHs ) ( donor information is summarized in ( Table 1 ) . In contrast with the basal culture medium ( small hepatocyte medium ( SHM ) ) , culture in YAC-containing SHM ( SHM+YAC ) induced the proliferation of cells that morphologically resembled epithelial cells ( Figure 1A ) . These cells were small and had a higher nucleus-to-cytoplasm ratio than hepatocytes , which is a typical morphological feature of LPCs . When colonies became densely packed , rat and mouse CLiPs exhibited a compact polygonal cell shape delimited by sharply defined refractile borders with bright nuclei in phase contrast images ( Figure 1B and C ) . However , unlike rat and mouse CLiPs , the morphology of human cells did not clearly change after colonies became densely packed ( Figure 1A ) . Although we did not perform further characterization , these proliferating cells likely arose from non-hepatic cells , such as biliary epithelial cells ( BECs ) or so-called liver epithelial cells , the origins of which are not well-defined ( Mitaka et al . , 1999 ) . Thus , we speculated that human hepatocytes require additional proliferative stimuli . Therefore , we tested the ability of fetal bovine serum ( FBS ) to support the proliferation of these cells . One of three lots of APHHs formed proliferative and densely packed colonies , and exhibited a hepatocytic morphology upon culture in medium supplemented with YAC and 10% FBS ( FYAC ) ( Figure 1D ) . By contrast , all three lots of APHHs formed proliferative colonies with hepatic morphologies upon culture in medium supplemented with AC and 10% FBS ( FAC ) ( Figure 1E ) . However , the proliferative capacity of these hepatic colony-forming cells was limited , and the number of these cells markedly decreased after the first passage , while non-parenchymal cells ( NPCs ) with non-hepatic morphologies became the dominant population ( data not shown ) . Next , considering the previous finding that PHHs derived from young donors are optimal for in vitro expansion ( Walldorf et al . , 2004; Yamasaki et al . , 2006 ) , we tested whether infant PHHs ( IPHHs ) expanded more efficiently in the presence of small molecules and FBS . Using IPHHs derived from a 10-month-old donor ( lot FCL ) , we performed a mini-screen using all possible combinations of Y , A and C in 10% FBS-supplemented SHM . The water-soluble tetrazolium salt-based ( WST ) assay demonstrated that these cells proliferated in the presence of A , YA , AC and YAC ( Figure 2A ) . Consistent with the observations made in APHHs ( Figure 1E ) , these cells proliferated most efficiently in FAC and thus we used this medium in all subsequent experiments . Robust proliferation of hepatocytes was not supported by culture in the presence of AC or FBS alone , but was synergistically supported by culture in the presence of both AC and FBS ( Figure 2B ) . Although proliferating cells cultured in FAC did not morphologically resemble hepatocytes when the cell density was low , they spontaneously acquired a hepatocyte-like morphology as colonies became densely packed ( Figure 2C ) . This observation strongly suggests that human proliferative cells cultured in FAC more closely resembled rodent CLiPs than those cultured in the presence of YAC . Unlike APHHs , IPHHs proliferated efficiently and became the predominant population over 2 weeks of culture . Two other lots of IPHHs ( lot DUX from an 8-month-old donor and lot JFC from a 1-year-old donor ) ( Table 1 ) also proliferated in this culture condition , although the proliferative capacity varied among the lots: FCL , DUX and JFC proliferated 49 . 2 ± 9 . 34 ( at day 14 ) , 46 . 2 ± 2 . 12 ( at day 14 ) and 3 . 66 ± 0 . 321 ( at day 12 ) folds , respectively ( mean ± SEM , determined by two repeated experiments for each lot ) . We also confirmed by microscopy that FAC enabled two more donors ( 11 months and 2-year-old ) -derived IPHHs and one juvenile donor ( 7-year-old ) -derived hepatocytes to proliferate and spontaneously change their morphologies to hepatocyte-like ones in the densely packed region of the proliferating colonies ( Figure 2—figure supplement 1 ) . These proliferating cells expressed multiple surface markers of LPCs , including EPCAM , CD44 , PROM1 ( also known as CD133 ) , CD24 , and ITGA6 ( Figure 2D , Figure 2—figure supplement 2A ) . It should be noted that PHHs before plating minimally expressed these LPC markers ( Figure 2—figure supplement 2B ) . Although we used PHHs which underwent a freeze-thaw cycle , we confirmed that expression of DPP4/CD26 , a general hepatocyte marker , was still preserved , supporting the validity of this flow cytometry analysis . We observed that PHHs were slightly positive for ITGA6 , but this is likely to be a nature of primary hepatocytes , because lineage-traced mouse primary hepatocytes also exhibited a slight signal of Itga6 , while none of the other analyzed LPC markers , Epcam , Prom1 and Cd24 were detected ( Figure 2—figure supplement 2C ) . Next , we performed microarray-based transcriptome analysis of previously identified BEC/LPC marker genes to further characterize these cells . Expression of many of these genes was induced during the 2 weeks of culture ( Figure 2E ) . Some of these genes , such as PROM1 and SPP1 , were expressed at comparable levels regardless of whether cells were cultured in the presence of AC , suggesting that their expression was spontaneously induced by the basal culture conditions ( Figure 2E ) . However , expression of multiple BEC/LPC marker genes , including EPCAM , SOX9 , KRT19 , TACSTD2 , AXIN2 and PROX1 , was increased in cells cultured in FAC ( Figure 2E and F ) . Of these , expression of EPCAM , SOX9 , and KRT19 was affected not only by the presence of AC but also by the culture duration , suggesting that AC-induced expression of these genes during in vitro culture . By contrast , expression of AXIN2 and PROX1 was maintained , but not increased , upon culture in the presence of AC . Gene signature enrichment analysis ( GSEA ) comparing cells cultured in the presence of FBS and those cultured in FAC demonstrated that the majority of gene sets enriched in the latter cells were related to hepatic function ( Figure 2G , Supplementary file 1 ) , suggesting that AC also helped to maintain the hepatocytic characteristics of cultured hepatocytes . Although cell-cycle-related gene sets were also identified by GSEA , their enrichment scores were relatively low ( Figure 2—figure supplement 3A , Supplementary file 1 ) . This is likely because cell proliferation was also increased in part by culture in FBS alone . Indeed , proliferation-related gene sets were enriched both in cells cultured in FBS only and in FAC compared with D1 hepatocytes ( Figure 2—figure supplement 3B and C , Supplementary file 2 , 3 ) . In summary , two small molecules , AC , together with FBS , support the proliferation of hepatic epithelial cells with characteristics of both hepatocytes and LPCs/BECs . To investigate the difference regarding the responsiveness to FAC of IPHHs and APHHs , we compared their transpcriptome by microarray analysis . Hierarchical clustering of the whole transcriptome demonstrated that IPHHs cultured in FAC for 7 or 14 days formed a cluster distinct from those cultured in FBS ( Figure 3A ) . In contrast , APHHs cultured in FAC for 7 or 14 days were not clearly separated from those cultured in FBS . These results suggest that APHHs are less sensitive to AC than IPHHs . GSEA indicated that many of the signaling pathways enriched for IPHHs cultured in FAC for 7 days compared with APHHs were cell-cycle-related pathways ( Figure 3—figure supplements 1 and 2 ) ( we avoided comparing cells at D14 , because lot 187271 APHHs were severely contaminated with NPCs at D14 , as shown in Figure 3—figure supplement 3 , ( iv ) ) . In contrast , pathways enriched for APHHs included hepatic function-associated ones ( Figure 3—figure supplements 1 and 2 ) . These results suggest that APHHs were not susceptible to the pro-proliferative effect of AC . Intriguingly , we observed a relatively similar expression profile of LPC marker genes between IPHHs and APHHs ( Figure 3B ) , except for EPCAM and ANPEP . We then asked whether APHHs indeed responded to A83-01 and CHIR99021 . GSEA indicated enrichment of Wnt signaling in IPHHs compared with APHHs ( nominal p-value=0 . 019 ) , suggesting that APHHs less efficiently responded to CHIR99021 ( Figure 3—figure supplements 1 and 4 ) . In contrast , to our surprise , TGFβ signaling was enriched in IPHHs compared with APHHs ( nominal p-value<0 . 001 ) ( Figure 3—figure supplements 1 and 4 ) . We further investigated the expression of individual genes which are known to be in the downstream of Wnt signaling ( Russell and Monga , 2018 ) and TGFβ signaling ( Cicchini et al . , 2015; Fabregat and Caballero-Díaz , 2018 ) ( Figure 3C and D ) . Both IPHHs and APHHs upregulated typical Wnt target genes in hepatocytes , such as GLUL and CYP1A2 in the presence of AC ( Figure 3C ) . On the other hand , we found that LPC-related Wnt-target genes , AXIN2 and LGR5 , were expressed at higher levels in IPHHs than APHHs ( Figure 3C ) . We also confirmed that TGFβ downstream genes were downregulated in both IPHHs and APHHs treated with FAC compared with their FBS counterparts ( Figure 3D ) . In addition , we confirmed that expression levels of some of these genes , for example VIM , SNAI1 and ZEB1 , were higher in IPHHs than APHH , which explains the reason for the enrichment of TGFβ signaling in IPHHs by GSEA ( Figure 3—figure supplements 1 and 4 ) . Another signaling pathway enriched in IPHHs was mTORC1 signaling ( Figure 3—figure supplement 4 ) . mTORC1 is activated specifically in pericentral hepatocytes in a Wnt signaling-dependent manner , and suggested to regulate their growth in normal liver ( Adebayo Michael et al . , 2019 ) . mTORC1 is also reported to be essential for BEC expansion during ductular reaction in regenerating liver as well as BEC organoid formation in vitro ( Planas-Paz et al . , 2019 ) . In summary , although a more detailed analysis is needed , the low proliferative capacity of APHHs might be partly explained by their lower responsiveness to Wnt signaling . A hepatic differentiation capacity is an important feature of LPCs , particularly for their potential use as a candidate cell source for transplantation therapy . To investigate the hepatic differentiation capacity of these proliferative cells , we passaged and cultured them in the presence of oncostatin M ( OSM ) , dexamethasone and Matrigel , which induce maturation of LPCs into hepatocytes ( Kamiya et al . , 2002 ) . As noted in Figure 2C , the proliferative cells spontaneously acquired hepatic morphologies when they reached 100% confluency , even in the absence of hepatic maturation inducers ( Figure 4A , Figure 4—figure supplement 1A , middle panels for each lot ) . However , this morphological change was more evident in the presence of hepatic maturation inducers ( Figure 4A , Figure 4—figure supplement 1A , right panels for each lot ) . In particular , cells acquired a polygonal and cytoplasm-rich morphology , which is similar to that of PHHs ( Figure 4B ) . Accordingly , microarray analysis confirmed that expression of representative hepatic marker genes , including ALB , TDO2 and SERPINA1 was increased after hepatic maturation induction ( Figure 4C ) . However , the expression levels of these genes were not markedly changed in cells from lot JFC . This is presumably because expression of hepatic maturation genes was already high in these cells even before hepatic induction . In contrast with the hepatic marker genes , expression of the BEC/LPC marker genes including SOX9 , KRT19 , and KRT7 was decreased , suggesting that the proliferative cells lost their BEC/LPC phenotype and acquired a mature hepatic phenotype ( Figure 4—figure supplement 1B ) . Hierarchical cluster analysis of genes that were differentially expressed between cells cultured in the presence of hepatic maturation inducers ( Hep-i ( + ) ) and cells cultured for the same duration in the absence of hepatic maturation inducers ( Hep-i ( - ) ) indicated that the characteristics of Hep-i ( + ) cells were relatively similar to those of PHHs ( Figure 4D ) . Overrepresented pathways in Hep-i ( + ) cells in comparison with Hep-i ( - ) cells were associated with the immune response and metabolic processes ( Figure 4E ) , both of which are important functions of the liver . These findings were further validated by GSEA ( Figure 4F , Supplementary file 4 ) . By contrast , overrepresented pathways in Hep-i ( - ) cells in comparison with Hep-i ( + ) cells were associated with developmental processes and morphogenesis , implying that Hep-i ( - ) cells were functionally immature compared with Hep-i ( + ) cells ( Figure 4—figure supplement 1C ) . In addition , cell cycle-related genes were overrepresented in Hep-i ( - ) cells ( Figure 4—figure supplement 1D , Supplementary file 5 ) , which is consistent with the general notion that progenitor cells have a greater proliferative capacity than cells with a more mature phenotype . Taken together , proliferative cells derived from human hepatocytes via culture in FAC lost their immature phenotype and acquired a mature hepatocyte-like phenotype in response to hepatic maturation inducers . Thus , we hereafter designate these proliferative cells as human CLiPs ( hCLiPs ) . Cytochrome P-450 ( CYP ) enzymes play a central role in the metabolic functions of the liver . Thus , we investigated the metabolic functions of hCLiP-derived hepatocytes . As noted in the previous section , overrepresented pathways in Hep-i ( + ) cells were associated with metabolism ( Figure 4E and F , Supplementary file 5 ) . In addition , pathways involving CYPs were enriched in Hep-i ( + ) cells , as characterized by GSEA using both the KEGG and Reactome databases , although the p-values for these gene sets were not lower than 0 . 05 ( Figure 4—figure supplement 1E ) . A heatmap revealed that expression of several CYP genes was higher in Hep-i ( + ) cells than in Hep-i ( - ) cells ( Figure 5A ) . These genes included CYP2B6 , CYP2D6 , CYP2E1 , CYP2C9 and CYP3A4 , which play crucial roles in metabolic functionality of the human liver ( Martignoni et al . , 2006 ) . The enzymatic activities of multiple CYPs were investigated by liquid chromatography tandem mass spectrometry ( LC-MS/MS ) using a cocktail of substrates ( Figure 5B ) ( Ohtsuki et al . , 2012 ) . This revealed that the enzymatic activities of CYP1A2 , CYP2C19 , CYP2C9 , CYP2D6 and CYP3A were comparable , if not the same , in Hep-i ( + ) cells derived from lots FCL and JFC as in PHHs , but were lower in Hep-i ( + ) cells derived from lot DUX ( Figure 5B ) . Expression of CYP1A2 , CYP2B6 and CYP3A4 is induced in hepatocytes via transcriptional activation in response to certain chemicals . Thus , we investigated whether the expression and activities of these CYPs were increased in hCLiP-derived hepatocytes treated with prototypical inducers of each CYP isoform , namely , omeprazole ( aryl hydrocarbon receptor ligand ) for CYP1A2 , phenobarbital ( indirect activator of constitutive active androstane receptor ) for CYP2B6 and CYP3A4 , and rifampicin ( pregnane X receptor ligand ) for CYP3A4 . These CYP genes were markedly upregulated in cells derived from the three lots in response to the corresponding inducer ( Figure 5—figure supplement 1A , B ) . Although enzymatic activities of these CYPs were increased in both Hep-i ( - ) and Hep-i ( + ) cells upon treatment with the corresponding inducer , these increases were relatively larger in the latter cells ( Figure 5C , Figure 5—figure supplement 1C ) , consistent with the changes in gene expression ( Figure 5—figure supplement 1B ) . We also directly quantified CYP protein expression by mass spectrometry . Protein expression of CYP1A2 and CYP3A4 in hCLiP-derived hepatocytes was increased in response to the corresponding inducer ( Figure 5D ) . In addition , activities of the phase II enzymes sulfotransferase ( SULT ) and UDP-glucuronosyltransferase ( UGT ) were comparable in hCLiP-derived hepatocytes and PHHs ( Figure 5E ) . These results demonstrate that hCLiPs differentiate into cells that are metabolically mature after induction of hepatic maturation and thus are potentially applicable for drug metabolism studies . Long-term culture of hepatocytes or LPCs with a sustained proliferative capacity is of great interest for liver regenerative medicine and drug discovery studies . Thus , we investigated the feasibility of long-term culture of hCLiPs . Cells derived from lots FCL and DUX could be serially passaged until at least passage 10 ( P10 ) without growth arrest ( Figure 6A ) or obvious morphological changes ( Figure 6—figure supplement 1A ) . The population doubling times of FCL and DUX hCLiPs were 1 . 27 ± 0 . 0066 and 1 . 43 ± 0 . 0086 d , respectively ( mean ± SEM , determined by three repeated experiments for each lot ) . However , non-hepatic cells with a fibroblast-like morphology were also observed ( Figure 6—figure supplement 1A , arrows ) , and the percentage of these cells varied among repeated experiments for each lot , as assessed by flow cytometric analysis of the epithelial-cell surface marker proteins EPCAM and CD24 ( Figure 6—figure supplement 1B ) . Cultures of cells from lot JFC contained more fibroblast-like cells than cultures of cells from lots FCL and DUX ( Figure 6—figure supplement 1A ) . Upon culture of cells from lot JFC , the percentage of fibroblastic cells increased with the passage number and fibroblastic cells overwhelmed hCLiPs by P5 , as assessed by microscopic observation ( n = 3 repeated experiments ) ( Figure 6—figure supplement 1A ) and flow cytometric analysis of LPC markers ( n = 1 experiment ) ( Figure 6—figure supplement 1B ) . However , when EPCAM+ cells were sorted from primary hCLiPs at the first passage , proliferative epithelial cells were observed for at least the next three passages ( total of four passages ) with their population doubling time 1 . 24 d ( n = 1 experiment ) between P1 and P4 ( Figure 6A , Figure 6—figure supplement 1A ) , confirming the proliferative capacity of hCLiPs obtained from lot JFC . Although expression of surface markers varied among experimental batches at later passages ( Figure 6—figure supplement 1B ) , it was relatively stable up to P5 in cells derived from lots FCL and DUX ( Figure 6—figure supplement 1B ) . We also investigated the karyotype of cells derived from lots FCL and DUX at P7 ( Figure 6B ) . hCLiPs derived from lot JFC were contaminated by an increased percentage of fibroblast-like cells; therefore , we karyotyped FACS-sorted EPCAM+ cells ( at the first passage ) which were then passaged four times after sorting ( Figure 6B ) . None of the analyzed cells exhibited any chromosomal abnormality ( 20 cells analyzed per lot ) and all the analyzed cells were diploid ( 50 cells analyzed per lot ) ( Figure 6B ) . This implies that hCLiPs were derived from diploid hepatocytes , which is consistent with our previous observations in rat CLiPs ( Katsuda et al . , 2017 ) . We further investigated transcriptomic changes in hCLiPs derived from lots FCL and DUX between P0 and P10 using cells from the experimental batches that maintained higher levels of EPCAM and CD24 expression ( Figure 6—figure supplement 1B ) ( experimental batch #3 and #2 for lots DUX and FCL , respectively ) . A heatmap of genes that were differentially expressed between P0 and P10 showed that the phenotype of hCLiPs gradually changed ( Figure 6—figure supplement 1C ) . As indicated on the right in Figure 6—figure supplement 1C , genes whose expression decreased included those related to hepatic functions , indicating that hCLiPs lost their hepatic phenotypes during repeated passage . Nonetheless , the heatmap suggested that hCLiPs retained at least some of their original characteristics until approximately P5 ( Figure 6—figure supplement 1C ) . Thus , we investigated the hepatic phenotype of hCLiPs at P3 and P5 . qRT-PCR analysis of hCLiPs derived from each lot indicated that absolute expression levels of hepatic genes consistently decreased as the passage number increased ( Figure 6C ) . Nevertheless , hCLiPs derived from each lot , particularly lots FCL and DUX , could undergo hepatic differentiation ( Figure 6C ) . Immunocytochemistry revealed that Hep-i ( + ) cells derived from lot FCL expressed hepatic marker proteins at P3 ( Figure 6—figure supplement 1D ) . We also investigated CYP enzymatic activities in these cells . Although the CYP enzymatic activities clearly decreased upon repeated passage , the basal activities of these enzymes , with the exception of CYP2C19 , were maintained at P3 and P5 ( Figure 6D ) . Induction of CYP3A enzymatic activity in response to rifampicin and phenobarbital was relatively stable even at P3 and P5 , especially in Hep-i ( + ) cells ( Figure 6E ) . In summary , functional decline of hCLiP-derived hepatocytes during continuous culture is unavoidable; however , CYP3A , the most important CYP in human drug metabolism , is still induced in these cells . We then asked whether the loss of the original phenotype of hCLiPs , especially the hepatic phenotype , during serial passages would be caused by their own phenotypic change or by expansion of contaminated NPCs . Using antibodies against three LPC makers ( EPCAM , PROM1 and CD24 ) and a NPC marker THY1/CD90 , which particularly characterizes fibroblastic cells , we sorted each LPC-marker+THY1- population and LPC-marker-THY1+ population from FCL-hCLiPs at P0 ( Figure 7A , Figure 7—figure supplement 1 ) . qRT-PCR clearly demonstrated that each LPC marker enabled enrichment of cells with hepatic phenotype ( ALB , TTR , GJB1 ) , whereas THY1-enriched cells consistently exhibited mesenchymal phenotype as characterized by the expression of ACTA2 and VIM in addition to THY1 ( Figure 7B ) . After subsequent three passages ( 2–3 weeks ) , LPC marker-enriched cells relatively retained their LPC/hepatic phenotypes as assessed by qRT-PCR ( Figure 7C ) and flow cytometry ( Figure 7—figure supplement 1 ) compared to THY1-enriched cells . However , we noted that compared to the cells at P0 , such LPC/hepatic phenotypes were largely reduced after three passages ( Figure 7—figure supplement 2 ) . These results demonstrated that , even after enrichment of LPC marker+ cells , phenotypic deterioration of hCLiPs is unavoidable . Since hCLiPs did not exhibit remarkable morphological changes until P10 of culture ( Figure 6—figure supplement 1A ) , the results obtained here call attention to the need for a careful quality control of hCLiPs by quantitative analyses , including flow cytometry and qRT-PCR . The capacity to repopulate injured livers is the most important and stringent criterion of a candidate cell source for liver regenerative medicine . Depending on the disease , 1–15% of hepatocytes must be replaced to achieve and sustain a therapeutic effect ( Jorns et al . , 2012; Rezvani et al . , 2016 ) . Laboratory-generated hepatocytes typically have RIs of less than 5% ( Rezvani et al . , 2016 ) , but a few studies reported maximum RIs of 20% or 30% in individual animals ( Carpentier et al . , 2014; Du et al . , 2014 ) . Moreover , in a recent study , Zhang et al . ( 2018 ) achieved much higher RI ( >60% ) by transplanting expandable hepatic cells named ProliHH , which were generated from PHHs as with hCLiPs . Thus , it is important to evaluate the repopulative capacity of hCLiPs from a comparative point of view . We assessed the repopulative capacity of hCLiPs in immunodeficient mice with chronically injured livers . Our previous study revealed that rat CLiPs repopulate the liver of cDNA-uPA/SCID mice ( Katsuda et al . , 2017 ) ; therefore , we first transplanted hCLiPs derived from lots FCL , DUX and JFC at P0–P2 into this model . After intrasplenic transplantation of primary hCLiPs that had been expanded in vitro for approximately 2 weeks ( 11–13 days ) ( hereafter designated P0-hCLiPs ) , the human ALB ( hALB ) level was exponentially increased in the blood of some , but not all , mice ( Figure 8A , red lines ) . The maximum hALB level in blood was >10 mg/ml , which is comparable with that observed following transplantation of PHHs in this animal model ( Tateno et al . , 2015 ) . Immunohistochemistry ( IHC ) of human-specific CYP2Cs ( including CYP2C9 and other CYP2Cs according to the manufacturer’s datasheet ) demonstrated extensive repopulation in mouse livers extracted at 10–11 weeks after transplantation ( Figure 8B ) . Although the RI varied among mice ( 32 . 2 ± 13 . 5% for lot FCL , n = 11; 39 . 3 ± 13 . 5% for lot JFC , n = 11; 17 . 8 ± 16 . 4% for lot DUX , n = 4 , mean ± SEM ) , it reached >90% in some animals ( Figure 8C ) . This maximum RI is comparable with that achieved after transplantation of PHHs ( Rezvani et al . , 2016 ) . The repopulative capacity declined as the culture period increased ( Figure 8A and C ) . Nonetheless , one mouse transplanted with FCL-P1-hCLiPs ( hCLiPs derived from lot FCL that were passaged once before transplantation ) ( 67 . 4% ) and two mice transplanted with JFC-P2-hCLiPs ( hCLiPs derived from lot JFC that were passaged twice before transplantation ) ( 83 . 1% and 91 . 1% ) exhibited high RIs . It should be noted that FCL-P1-hCLiPs and JFC-P2-hCLiPs underwent approximately 1000- and 400-fold expansion from the initial PHHs , respectively . These fold expansion is comparable to that achieved by ProliHH developed by Zhang et al . ( 2018 ) . They reported high repopulative capacity of ProliHH at P4-P6 . The fold expansion based on the initial number of PHHs , ProliHH at P4-P6 in their culture system underwent approximately 400–1000-fold expansion . Thus , hCLiPs have repopulative capacity at the comparable levels with ProliHH . We further confirmed the repopulative capacity of FCL-P0-hCLiPs using another model , namely , TK-NOG mice ( Hasegawa et al . , 2011 ) . In this model , the serum hALB level was dramatically elevated to at most 8 . 1 mg/ml ( Figure 8D ) . The maximum RI was lower in TK-NOG mice ( 57 . 5% ) than in cDNA-uPA/SCID mice ( 96 . 0% ) ( Figure 8E and F ) . However , engraftment was more efficient in TK-NOG mice than in cDNA-uPA/SCID mice; significant repopulation ( >15% RI ) with FCL-P0-hCLiPs was observed in 83% ( 5/6 mice ) of TK-NOG mice ( Figure 8F ) , but only in 50% ( 3/6 mice ) of cDNA-uPA/SCID mice ( Figure 8C ) . Examination of the area repopulated by hCLiPs by staining with an antibody against human mitochondria showed that repopulating human cells expressed MDR1 and TTR , which are associated with hepatic function ( Figure 8G and H ) . MDR1 was detected on the apical side of adjacent mouse and human hepatocytes , suggesting that hCLiP-derived cells successfully reconstructed the normal liver architecture ( Figure 8G , arrows ) . Accordingly , hepatic zonation was correctly established in the repopulated regions , as assessed by expression of glutamate-ammonia ligase ( GLUL , also known as glutathione synthetase ) ( Figure 8H ) , CYP1A2 and CYP3A4 ( Figure 8I ) . We next tested whether , after several passages , hCLiPs would still retain repopulative capacity . First , we transplanted FCL-P4-hCLiPs ( hCLiPs derived from lot FCL that were passaged four times before transplantation ) to cDNA-uPA/SCID mice . Based on the growth curves ( Figure 6A ) , we estimated that these cells underwent 2 . 2 ± 0 . 94×106 fold ( n = 3 , mean ± SEM ) expansion from the initial PHHs . We observed increase of blood hALB levels in these mice , but the hALB increasing rates were much slower than in the animals transplanted with FCL-P0-hCLiP or FCL-P1-hCLiPs ( Figure 8—figure supplement 1A ) . The hALB levels 8 weeks after transplantation was 10 μg/ml at most ( Figure 8—figure supplement 1A ) , which was approximately 1/1000 of the mice with high RI ( >60% RI ) . As expected , hCYP2C staining indicated that all the mice transplanted with FCL-P4-hCLiPs exhibited very low RI ( <1% ) ( Figure 8—figure supplement 1B ) . We also transplanted FCL-P3-hCLiPs to TK-NOG mice . In this experiment , we prepared EPCAM-expressing cells at the first passage by magnetic activated cell sorting ( MACS ) , and cultured them for another two passages ( three passages in total ) . During the culture at P3 , we separated these cells to two groups , one with hepatic induction ( P3 Hep-i ( + ) ) and the other without hepatic induction ( P3 Hep-i ( - ) ) . After transplantation to TK-NOG mice , we observed serum hALB increase in these mice ( Figure 8—figure supplement 1C ) . Importantly , we confirmed that Hep-i ( + ) cell-transplanted group showed consistently higher hALB levels than the Hep-i ( - ) cell-transplanted group . However , the serum hALB levels of FCL-P3-hCLiP-Hep-i ( + ) -transplanted mice at 8 weeks were still only 9 . 1 ± 1 . 8 μg/ml ( n = 4 , mean ± SEM ) , which were again much lower than the mice transplanted with FCL-P0-hCLiPs . These results collectively show that hCLiPs unavoidably decrease their repopulative capacity following extended in vitro culture . Finally , we isolated human cells from chimeric mouse livers and investigated their functionality because it has been argued that some types of laboratory-generated hepatocytes are not fully functional after repopulation ( Rezvani et al . , 2016 ) . We first performed microarray-based transcriptomic analysis . After isolating hepatocytes from chimeric livers of cDNA-uPA/SCID mice by a two-step collagenase perfusion method , we eliminated mouse cells using a magnetic bead separation system . Microscopic observation revealed that 32 . 7% , 16 . 8% and 33 . 1% of hepatocytes isolated from chimeric livers of mice transplanted with hCLiPs derived from lots FCL , JFC and DUX bound to magnetic beads conjugated with a specific anti-mouse antibody prior to magnetic separation , respectively , while these percentages were reduced to 2 . 9% , 0 . 0% , and 1 . 6% after magnetic separation , respectively . Thus , we assumed that the results of experiments performed with these cells should be mostly ascribed to human cells . Magnetically separated human cells exhibited typical morphologies of mature hepatocytes ( Figure 9A ) . However , unexpectedly , hierarchical clustering and principle component analysis ( PCA ) of the entire transcriptome showed that chimeric liver-derived human cells were distinct from PHHs ( Figure 9B and C ) . A control sample of human hepatocytes isolated from chimeric livers following transplantation of IPHHs ( lot JFC ) yielded similar results as human hepatocytes isolated from chimeric livers following transplantation of hCLiPs ( Figure 9B and C ) , indicating that the transcriptomic difference between human hepatocytes in chimeric livers and PHHs is due to environmental differences between human and mouse livers . Surprisingly , GSEA demonstrated that multiple hepatic function-related gene sets were overrepresented in human hepatocytes isolated from chimeric livers in comparison with PHHs ( Supplementary file 6 ) . The majority of these gene sets were associated with metabolic pathways . Other hepatic functions were also enriched , such as pathways associated with coagulation and complement production ( Figure 9D , Supplementary file 6 ) . BEC/LPC marker genes were underrepresented in hCLiP-derived hepatocytes isolated from chimeric livers and PHHs in comparison with hCLiPs ( Figure 9—figure supplement 1A ) , demonstrating that hCLiPs underwent hepatic maturation after repopulating mouse livers . We also investigated whether hCLiP-derived hepatocytes isolated from chimeric livers exhibited CYP activities . As expected based on the transcriptomic analysis , hCLiP-derived cells isolated from chimeric livers exhibited basal enzymatic activities of major CYPs at the levels comparable with those in PHHs ( Figure 9E ) . Enzymatic activities of CYP1A2 , CYP2B6 and CYP3A were markedly induced in hCLiP-derived hepatocytes isolated from chimeric livers upon treatment with rifampicin , phenobarbital and omeprazole ( Figure 9F ) . Consistently , qRT-PCR analysis demonstrated that expression of CYP1A2 , CYP2B6 and CYP3A4 was dramatically upregulated upon treatment with CYP inducers ( Figure 9—figure supplement 1B ) . Finally , activities of the phase II enzymes SULT and UGT in hCLiP-derived hepatocytes isolated from chimeric livers were comparable with those in PHHs ( Figure 9G ) . These results indicate that although their transcriptomic profiles are not identical to those of PHHs , including IPHHs and APHHs , hCLiPs functionally mature in the mouse liver .
In this study , we demonstrated that hCLiPs can repopulate chronically injured livers of immunodeficient mice . An efficient repopulative capacity is one of the most important requirements of a candidate cell source for transplantation therapy; however , it is very challenging to develop such a cultured cell source . Laboratory-generated hepatic cells , such as pluripotent cell-derived hepatic cells and those transdifferentiated from cells of different lineage origins , have a poor repopulative capacity ( Rezvani et al . , 2016 ) . The RI of laboratory-generated hepatocytes is typically less than 5% ( Rezvani et al . , 2016 ) . After our report of rodent CLiPs ( Katsuda et al . , 2017 ) , four groups recently reported methods for in vitro generation of proliferative liver ( progenitor ) cells from human hepatocytes ( Fu et al . , 2018; Hu et al . , 2018; ; Kim et al . , 2019; Zhang et al . , 2018 ) . In three of these studies ( Fu et al . , 2018; Hu et al . , 2018; ; Kim et al . , 2019 ) , the generated cells exhibited relatively low repopulative efficiency with approximately 13% of RI at maximum . In contrast , Zhang el al . reported strikingly high repopulation efficiency with as high as 64% of RI ( Zhang et al . , 2018 ) . Importantly , although the proliferative efficiency is limited compared with IPHHs , they succeeded in induction of proliferative hepatic cells even from APHHs . Moreover , contrary to the dichotomous repopulation of hCLiPs in our study ( RI >80% or nearly 0% ) , Zhang et al demonstrated highly stable repopulation among transplanted animals . Our study is , thus , not the first one to report substantial repopulation using an in vitro-generated human hepatic cell source . Nonetheless , to solidify a novel concept , more evidence must be provided independently from multiple laboratories . As such , we still believe that our work also plays an important role in pioneering this new field . Another important finding in this study is that hCLiPs may be a novel cell source for drug discovery studies . The major criterion for the application of cultured hepatic cells in drug discovery studies , particularly to evaluate the functions of drug-metabolizing enzymes , is the inducibility of CYP enzymatic activities . CYP enzymes play central roles in the metabolism of clinically used drugs and xenobiotics . In general , CYP induction accelerates the clearance of xenobiotics , leading to beneficial or harmful outcomes depending on the context . Thus , recapitulation of CYP induction in cultured hepatocytes or their equivalents is important to precisely predict the effects of a tested drug on hepatocytes . However , PHHs lose their hepatic functions , including CYP inducibility , upon in vitro culture . Laboratory-generated hepatocytes reportedly exhibit basal CYP activities after maturation ( Baxter et al . , 2015; Kanninen et al . , 2016; Liu et al . , 2011; Takayama et al . , 2018; Takayama et al . , 2014 ) . Although a few groups described CYP inducibility in terms of enzymatic activity ( Inamura et al . , 2011; Pettinato et al . , 2016; Takayama et al . , 2012 ) , such reports are very limited , to the best of our knowledge . We propose that hCLiPs are a novel platform for drug discovery studies . An issue yet to be addressed is clarification of the mechanism underlying the small molecule-mediated conversion of PHHs to hCLiPs . Mini-screen of three small molecules Y , A and C demonstrated that A and C individually accelerated proliferation of PHHs , while Y alone exhibited no beneficial effect on proliferation , and even negatively affected proliferation when combined with AC ( in comparison of YAC with AC ) . This is in line with our previous observation in rodent hepatocyte culture , in which Y minimally affected the proliferation of these three small molecules ( Katsuda et al . , 2017 ) . AC substantially induced the proliferation of rodent hepatocytes at the comparable , if not at the same , level with YAC ( Katsuda et al . , 2017 ) . Thus , the synergistic effect of A83-01 and CHIR99021 is the key to hepatocyte proliferation in rodent and human hepatocytes . Importantly , comparative analysis of APHHs and IPHHs suggested that activity of Wnt signaling in response to CHIR99021 may partly explain the proliferative ability of IPHHs . On the other hand , APHHs responded to A83-01 equally or even more efficiently than IPHHs , leaving a question how A83-01 affected the proliferation of IPHHs , but not APHHs . Since A83-01 is essential to IPHH proliferation as assessed in the mini-screen ( Figure 1A ) , this small molecule might affect IPHHs in a TGFβ-independent manner . Further investigation is needed to fully understand the difference between the proliferative ability endowed by FAC between IPHHs and APHHs . Another important issue to be considered is the requirement for FBS in hCLiP induction , which is not the case for rodent CLiP induction . FBS-derived factor ( s ) essential for hCLiP induction should be identified in a future study . Comparison of our study with the recently reported four studies provides hints to mechanistic understanding of in vitro PHH expansion ( Fu et al . , 2018; Hu et al . , 2018; Kim et al . , 2019; Zhang et al . , 2018 ) . Notably , hepatocyte growth factor ( HGF ) , which is not included in our culture condition , is used in all these four studies , suggesting its critical role . Indeed , Kim et al . ( 2019 ) particularly emphasizes its essential role in the presence of AC . On the other hand , Zhang et al . ( 2018 ) ascribe the proliferative capacity of PHHs particularly to Wnt signaling ( Zhang et al . , 2018 ) . Interestingly , these authors reported that Wnt3a plays an essential role , while neither CHIR99021 nor Wnt signaling amplifier Rspo1 substituted for the pro-proliferative effect of Wnt3a . Moreover , these authors proposed a unique idea that hypoxic culture condition supports the stable proliferation of PHHs by suppressing PHH senescence . In line with this observation , Fu et al . also demonstrated that a sirtuin suppressor nicotinamide decreases proliferation of PHH . This finding highlights the difference between human and rodent PHHs: nicotinamide is known to induce proliferation of rat hepatocytes ( Mitaka et al . , 1991 ) and thus is frequently added to hepatocyte culture medium ( including ours ) , but this may not be the case for induction of PHH proliferation . Collectively , these findings , including ours , provide important insight to optimization of the methodology of PHH expansion .
Infant primary human hepatocytes ( IPHHs ) ( lots FCL , DUX , JFC and MRW ) were purchased from Veritas Corporation ( Tokyo , Japan ) . Adult primary human hepatocytes ( APHHs ) ( lots HC1-14 , HC3-14 , HC5-25 , and HC7-4 ) were purchased from Sekisui XenoTech ( KS ) . IPHH lot 187273 and APHH lot 187271 were purchased from Biopredic ( Saint-Gregoire , France ) . Donor information is summarized in Table 1 . The basal medium for culture of PHHs was SHM ( DMEM/F12 ( Life Technologies , MA ) containing 2 . 4 g/l NaHCO3 and L-glutamine ) ( Chen et al . , 2007; Katsuda et al . , 2018 ) supplemented with 5 mM HEPES ( Sigma , MO ) , 30 mg/l L-proline ( Sigma ) , 0 . 05% bovine serum albumin ( Sigma ) , 10 ng/ml epidermal growth factor ( Sigma ) , insulin-transferrin-serine-X ( Life Technologies ) , 10−7 M dexamethasone ( Sigma ) , 10 mM nicotinamide ( Sigma ) , 1 mM ascorbic acid-2 phosphate ( Wako , Osaka , Japan ) , and antibiotic/antimycotic solution ( Life Technologies ) . Depending on the experiment , this basal medium was supplemented with 10% FBS ( Life Technologies ) , as well as small molecules , namely , 10 μM Y-27632 ( Wako ) , 0 . 5 μM A-83–01 ( Wako ) , and 3 μM CHIR99021 ( Axon Medchem , Reston , VA ) . After a mini-screen of these three small molecules , PHHs were routinely cultured in SHM supplemented with 10% FBS , 0 . 5 μM A-83–01 , and 3 μM CHIR99021 . IPHHs were thawed in a water bath set to 37°C and suspended in 10 ml Leibovitz’s L-15 Medium ( Life Technologies ) supplemented with Glutamax ( Life Technologies ) and antibiotic/antimycotic solution . After centrifugation at 50 × g for 5 min , the cells were resuspended in William’s E medium supplemented with 10% FBS , GlutaMAX , antibiotic/antimycotic solution , and 10−7 M insulin ( Sigma ) . The number of viable cells was determined using trypan blue ( Life Technologies ) . IPHHs from lot JFC were seeded in collagen I-coated plates ( IWAKI , Shizuoka , Japan ) at a density of approximately 5 × 103 viable cells/cm2 . IPHHs from lots FCL and DUX barely attached to the plates , and many of the small number that did attach subsequently detached prior to D3 , as monitored by time-lapse imaging using a BZ-X700 microscope ( Keyence , Osaka Japan ) ( data not shown ) . Therefore , IPPHs from lots FCL and DUX were seeded at a density of approximately 2 × 104 viable cells/cm2 , which was approximately 4-fold higher than the seeding density of IPHHs from lot JFC . To determine the fold change in cell number during in vitro culture , the number of adherent cells on D3 was counted based on micrographs acquired at 10 × magnification ( 5–10 fields per experiment ) . Cells were harvested using TrypLE Express ( Life Technologies , MA ) when they reached 70–100% confluency and then re-plated into a 10 cm collagen-coated plate at a density of 1–2 × 105 cells/dish . Numbers of viable cells were estimated based on the WST-8 assay using Cell Counting Kit 8 ( Dojindo , Kumamoto , Japan ) , according to the manufacturer’s instructions . Flow cytometry and cell sorting were performed using a S3e Cell Sorter ( BioRad , Hercules , CA ) . Cells were labeled with APC-conjugated mouse anti-human CD44 ( 1:20; G44-26; BD , Franklin Lakes , NJ ) , APC-conjugated mouse anti-human EPCAM ( 1:20; EBA-1; BD ) , PE/Cy7-conjugated anti-human/mouse CD49f ( ITGA6 ) ( 1:20; GoH3; Biolegend ) , PE/Cy7-conjugated anti-human CD24 ( 1:20; ML5; Biolegend ) , APC-conjugated human anti-PROM1/CD133 ( 1:11; AC133; Miltenyi Biotech ) , APC-conjugated mouse anti-human CD26/DPP4 ( 1:11 , FR10-11G9; Miltenyi Biotech ) , and FITC-conjugated mouse anti-human CD90/THY1 antibodies . An APC-conjugated mouse IgG1 , κ isotype control antibody ( Biolegend , MOPC-21 ) and a PE-Cy7-conjugated mouse IgG2b , κ isotype control ( BD , 27–35 ) were used as controls . An adult RosaYFP/YFP mouse received retro-orbital injection of AAV-TBG-cre at the dose of 2 . 5 × 1011 viral particles , and hepatocytes were harvested 3 weeks later using a standard two-step collagenase perfusion method . Isolated RosaYFP/YFP hepatocytes were stained with APC anti-mouse CD326/EPCAM ) ( 1:100 , G8 . 8 , Biolegend ) or APC anti-human/mouse CD49f/Itga6 antibody ( 1/100 , GoH3 , Biolegend ) . For staining with Prom1/Cd133 and Cd24 , cells were incubated with purified rat anti-mouse CD133/Prom1 antibody ( 1:100 , 315–2 C11 , Biolgend ) and rat anti-mouse CD24 antibody ( 1:100 , Biolegend , M1/69 ) followed by staining with Alexa647-conjugated donkey anti-rat antibody ( 1:300 , Jackson ImmunoResearch ) . APC-conjugated rat IgG2a or IgG2b , κ isotype control antibody was used as control . DAPI was added to stain dead cells . Attune NxT Flow Cytometer ( Lifetechnologies ) was used for data collection . B6 wild-type adult hepatocytes , which were stained with only DAPI , was used for making the threshold of YFP signal . One-color microarray-based gene expression analysis was performed using a SurePrint G3 Human Gene Expression v3 8 × 60K Microarray Kit ( Agilent , Santa Clara , CA ) following the manufacturer’s instructions . The 75th percentile shift normalization was performed using GeneSpring software ( Agilent ) . hCLiPs were harvested using TrypLE Express ( Life Technologies ) and reseeded into a collagen I-coated 24-well plate at a density of 5 × 104 cells/well ( 2 . 5 × 104 cells/cm2 ) . When cells reached approximately 50–80% confluency , culture medium was replaced by SHM supplemented with 2% FBS , 0 . 5 mM A-83–01 , and 3 mM CHIR99021 in the absence ( Hep-i ( - ) ) or presence ( Hep-i ( + ) ) of 5 ng/ml human OSM ( R and D ) and 10−6 M dexamethasone . Cells were cultured for a further 6 days and fresh medium was provided every 2 days . On D6 , cells were overlaid with a mixture of Matrigel ( Corning , Corning , NY ) and the aforementioned hepatic induction medium at a ratio of 1:7 and cultured for another 2 days . Thereafter , Matrigel was removed via aspiration , samples were washed with Hank’s Balanced Salt Solution supplemented with Ca2+ and Mg2+ ( Life Technologies ) , and cells were used for RNA extraction or CYP induction experiments . SHM containing 2% FBS , but not A-83–01 or CHIR99021 , was used as basal medium . CYP3A and CYP2B6 were induced via treatment with 10 μM rifampicin and 1 mM phenobarbital . An equal volume of methanol ( 1/100 dilution ) and H2O ( 1/1000 dilution ) was used as the vehicle control for rifampicin and phenobarbital , respectively . CYP1A2 was induced via treatment with 50 μM omeprazole , and methanol ( 1/100 dilution ) was used as the vehicle control . Each CYP induction medium was replaced by freshly prepared medium every day . After 3 days , CYP activity was measured by LC-MS/MS . Cells were cultured in phenol red-free William’s E medium supplemented with a cocktail of substrates ( 1/100 dilution ) at 37°C for 1 hr . This cocktail contained 40 μM phenacetin as a CYP1A2 substrate , 50 μM bupropion as a CYP2B6 substrate , 0 . 1 μM amodiaquin as a CYP2C8 substrate , 5 μM diclofenac as a CYP2C9 substrate , 100 μM S-mephenytoin as a CYP2C19 substrate , 5 μM bufuralol as a CYP2D6 substrate , 5 μM midazolam as a CYP3A substrate , and 100 μM 7-hydroxycoumarin as a UGT and SULT substrate . Thereafter , the culture supernatant was harvested and metabolites were quantified by LC-MS/MS as described previously ( Kozakai et al . , 2012 ) with minor modifications . CYP protein levels were measured as described previously ( Kawakami et al . , 2011 ) with minor modifications . After trypsin digestion of cells , the target peptide of each CYP isoform was absolutely quantified by LC-MS/MS . The expression levels of each CYP were quantified using previously described peptide standards ( Kawakami et al . , 2011 ) . The cellular DNA content was measured to estimate the number of cells for CYP induction experiments . Following removal of Matrigel via aspiration , cells were washed once with phosphate-buffered saline ( PBS ) and any remaining Matrigel was removed by treating cells with Cell Recovery Solution ( Corning ) at 4°C for approximately 30 min . Thereafter , cells were washed once with PBS , and the cellular DNA content was determined using a DNA Quantity Kit ( Cosmobio , Tokyo , Japan ) . To estimate the cell number from the DNA content , the correlation between these two parameters was determined using a dilution series of hCLiPs derived from each lot . Total RNA was isolated using an miRNeasy Mini Kit ( QIAGEN , Venlo , The Netherlands ) . Reverse transcription was performed using a High-Capacity cDNA Reverse Transcription Kit ( Life Technologies ) according to the manufacturer’s guidelines . cDNA was used for PCR with Platinum SYBR Green qPCR SuperMix UDG ( Lifetechnologies ) . Expression levels of target genes were normalized against that of ACTB as an endogenous control . The primers used for qRT-PCR are listed in the following table . GeneForwardReverseACTBACTCTTCCAGCCTTCCTTCCAGCACTGTGTTGGCGTACAGALBGCAAGGCTGACGATAAGGAGCCTAAGGCAGCTTGACTTGCTATATCTCTGTTATGGGGCGTTGACTAACCGCTCCGTGAACTCTTRATCTCCCCATTCCATGAGCCATTCCTTGGGATTGGTGACTDO2GGTGGTTCCTCAGGCTATCATGTCGGGGAATCAGGTATGTG6PCCCTTGCTGCTCATTTTCCTCTGTGGATGTGGCTGAAAGTTCYP1A2CCCCAAGAAATGCTGTGTCTAGGGCTTGTTAATGGCAGTGCYP2B6GGGGCACTGAAAAAGACTGAAGTTCTGGAGGATGGTGGTGCYP3A4ATTGGCATGAGGTTTGCTCTCGGGTTTTTCTGGTTGAAGAEPCAMTGGACATAGCTGATGTGGCTTACCAGGATCCAGATCCAGTTGPROM1AGTCGGAAACTGGCAGATAGCGGTAGTGTTGTACTGGGCCAATCD24AGGCGCGGACTTTTCTTTGATGCTGGGTGCTTGGAGGJB1CTGCTCTACCCTGGCTATGCGTAGACGTCGCACTTGACCATHY1ACCTACACGTGTGCACTCCAGCCCTCACACTTGACCAGTTACTA2CTGTTCCAGCCATCCTTCATGGCAATGCCAGGGTACATAGVIMTCTGGATTCACTCCCTCTGGGGTCATCGTGATGCTGAGAA The antibodies used for ICC are listed in the table below . Cells were fixed in chilled methanol ( −30°C ) on ice for 5 min . In some experiments , cells were fixed in 4% paraformaldehyde ( PFA ) ( Wako , Osaka , Japan ) at room temperature for 15 min and permeabilized by treatment with PBS containing 0 . 05% Triton X-100 for 15 min . Thereafter , cells were washed three times with PBS , incubated in Blocking One solution ( Nacalai Tesque , Kyoto , Japan ) at 4°C for 30 min , and labeled with primary antibodies at room temperature for 1 hr or at 4°C overnight . The primary antibodies were detected using Alexa Fluor 488- or Alexa Fluor 594-conjugated secondary antibodies ( Life Technologies ) . Nuclei were counterstained with Hoechst 33342 ( Dojindo ) . AntibodyHost animalCatalog #DilutionManufacturerFixationCYP3A4RabbitAb35721:200AbcamMethanolMRP2MouseAb33731:200AbcamMethanolHNF4ARabbitsc-89871:200Santa Cruz4% PFAMDR1Rabbitsc-532411:200Santa CruzMethanolCYP2CMousesc-532451:200Santa CruzMethanolCYP1A2Mousesc-532411:200Santa CruzMethanolTTRRabbitAb758151:500AbcamMethanol The antibodies used for IHC are listed in the table below . Formalin-fixed paraffin-embedded ( FFPE ) tissue samples were prepared . Following dewaxing and rehydration , heat-induced epitope retrieval was performed by boiling specimens in ImmunoSaver ( Nissin EM , Tokyo , Japan ) diluted 1/200 at 98°C for 45 min . Endogenous peroxidase was inactivated by treating specimens with methanol containing 0 . 3% H2O2 at room temperature for 30 min . Thereafter , specimens were permeabilized with 0 . 1% Triton X-100 , treated with Blocking One solution at 4°C for 30 min , and incubated with primary antibodies at room temperature for 1 hr or at 4°C overnight . Sections were stained using ImmPRESS IgG-peroxidase kits ( Vector Labs , Burlingame , CA ) and a metal-enhanced DAB substrate kit ( Life Technologies ) , according to the manufacturers’ instructions . Finally , specimens were counterstained with hematoxylin , dehydrated , and mounted . FFPE tissue samples were used for fluorescence IHC unless otherwise stated . Following dewaxing and rehydration , heat-induced epitope retrieval was performed by boiling specimens in ImmunoSaver ( Nissin EM ) diluted 1/200 at 98°C for 45 min and then the following staining steps were performed . Fresh frozen tissue blocks prepared using Tissue-Tek O . C . T . Compound ( Sakura Finetek , Tokyo , Japan ) were used for CYP1A2 and CYP3A4 staining . Fresh frozen liver sections prepared using a cryostat ( Leica ) were fixed in chilled ( −30°C ) acetone ( Wako ) for 5 min , washed three times with PBS , permeabilized with 0 . 1% Triton X-100 , and treated with Blocking One solution at 4°C for 30 min . Thereafter , specimens were incubated with primary antibodies at room temperature for 1 hr or at 4°C overnight and then stained with a mixture of an Alexa Fluor 488-conjugated antibody ( Invitrogen ) ( 1:500 ) and an Alexa Fluor 594-conjugated antibody ( Invitrogen ) ( 1:500 ) at room temperature for 1 hr . Stained sections were mounted using Vectashield mounting medium containing DAPI ( Vector Laboratories ) . AntibodyHost animalCatalog #DilutionManufacturerTissue typeCYP2CMousesc-532451:200Santa CruzFFPE/frozenMDR1Rabbitsc-532411:200Santa CruzFFPEHuman MitochondriaMouseab928241:1000AbcamFFPEHuman TTRRabbitab758151:500AbcamFFPEGLULRabbitab735931:1000AbcamFFPEHuman CYP1A2RabbitBML-CR3130-01001:200EnzoFrozenHuman CYP3A4RabbitBML-CR3340-01001:200EnzoFrozen Liver repopulation assay using cDNA-uPA/SCID mice hCLiPs derived from three lots of cells were used . For lots FCL and JFC , primary cultured cells at D11–14 ( P0-hCLiPs ) , cells passaged once ( P1-hCLiPs ) , and cells passaged twice ( P2-hCLiPs ) were used . For lot FCL , P4-hCLiP transplantation was also performed . For lot DUX , P0-hCLiPs were used . After harvesting cells using TrypLE Express , 0 . 2–1 × 106 cells/mouse were intrasplenically transplanted into 2–4 week-old cDNA-uPA/SCID mice ( PhoenixBio Co . , Ltd , Higashihiroshima , Japan ) under isoflurane anesthesia . From 2 weeks after transplantation , 10 μl blood was retro-orbitally collected each week and the hALB concentration was measured using a Human Albumin ELISA Quantitation Kit ( Bethyl , TX ) or a Latex agglutination turbidimetric immunoassay with a BioMajesty analyzer ( JCA-BM6050; JEOL , Tokyo , Japan ) . Livers were extracted at 8–11 weeks after transplantation and histologically analyzed . The transplantation experiments were approved by animal care committee . FCL-P0-hCLiPs were used . Seven-week-old TK-NOG mice were obtained from the Central Institute of Experimental Animals ( Kawasaki , Japan ) . One day after arrival at the National Cancer Center , mice were intraperitoneally injected with 10 mg/ml ganciclovir ( Mitsubishi Tanabe Pharma Corporation , Osaka , Japan ) at a dose of 10 μl/g body weight to induce thymidine kinase-mediated injury in host mouse hepatocytes . One day after injection , approximately 30 μl blood was obtained from the tail . Serum was separated and diluted 1/5 with PBS , and the serum ALT level was measured using a DRI-CHEM 3500 analyzer ( Fujifilm , Tokyo , Japan ) . Mice with serum ALT levels of 500–1600 U/l were chosen as host animals for transplantation . At 1–3 days after ALT measurement , 0 . 4–1 × 106 cells were intrasplenically transplanted into these mice under isoflurane anesthesia . From 2 weeks after transplantation , approximately 20 μl blood was collected each week from the tail and the hALB concentration was measured using a Human Albumin ELISA Quantitation Kit ( Bethyl , Montgomery , TX ) . Livers were extracted at 8–10 weeks after transplantation and histologically analyzed . The transplantation experiments were approved by animal care committee . For transplantation of FCL-P3-hCLiPs , EPCAM+ FCL-hCLiPs were magnetically sorted at the first passage ( using P0 hCLiPs ) by MACS cell sorting system using MidiMACS Separator ( Miltenyi ) using CD326 ( EpCAM ) MicroBeads ( Miltenyi ) . These EPCAM+ cells were subjected to another two passages ( P3 in total ) . During the culture at P3 , we separated these cells to two groups , one with hepatic induction ( P3 Hep-i ( + ) ) and the other without hepatic induction ( P3 Hep-i ( - ) ) . 1 × 106 cells/mouse were transplanted into GCV-treated TK-NOG mice as described before . Throughout this experiment , hCLiPs were cultured in a slightly different condition from other experiments ( FBS concentration was 5% instead of 10% ) , but we do not think the obtained results were severely changed by this minor modification . Hepatic induction was conducted in SHM supplemented with 5%FBS and AC with 10 ng/ml hOSM ( matrigel was not used ) . The transplantation experiments were approved by animal care committee . Unless otherwise stated , RIs were estimated based on CYP2C positivity using image analysis software and a Keyence BZX-710 microscope . RIs in chimeric mice that were sacrificed to isolate primary hepatocytes were estimated based on magnetic bead separation , as described in the following section . Hepatocytes were isolated from chimeric livers of cDNA-uPA/SCID mice at 10 weeks after transplantation of FCL-P1-hCLiPs , DUX-P0-hCLiPs , and JFC-P0-hCLiPs using a two-step collagenase perfusion method . To remove contaminating mouse hepatocytes , isolated cells were incubated with the 66Z antibody , which recognizes the surface of mouse hepatocytes , but not of human hepatocytes ( Yamasaki et al . , 2010 ) . Cells were washed with DMEM containing 10% FBS and then incubated with Dynabeads M450-conjugated sheep anti-rat IgG ( Dynal Biotech , Milwaukee , WI ) for 30 min on ice . The tube was placed in a Dynal MPC-1 holder ( Dynal Biotech ) for 1–2 min to remove 66Z+ mouse hepatocytes . Human hepatocytes were collected as 66Z- cells . 66Z+ and 66Z- hepatocytes were counted using a hemocytometer before and after magnetic separation to estimate the repopulation efficiency and purity of human hepatocytes after separation , respectively . Magnetically purified human hepatocytes were resuspended in SHM containing 2% FBS and seeded into a 24-well collagen I-coated plate . One day later , RNA was prepared from cells in some wells for microarray-based transcriptomic analysis . As a control , RNA was also prepared from hepatocytes isolated from the chimeric liver of a mouse transplanted with IPHHs ( lot JFC ) immediately after thawing the original cell suspension ( kindly prepared by PhoenixBio Co . , Ltd ) . Other hCLiP-derived hepatocytes were used for the CYP activity assay , as described above . Data represent the mean ± SEM of independently repeated experiments or the mean ± SD of technical replicates in separate culture wells . Two groups were statistically compared using the Student’s t-test , unless otherwise stated . Time-dependent alteration of gene expression was analyzed by the linear mixed models using IBM SPSS Statistics 23 ( SPSS Inc , Chicago , IL ) . Group allocation ( FBS or FAC ) , time ( culture period [day] ) , and the interaction of group and time were included in the model as fixed effects . A p-value less than 0 . 05 was considered statistically significant . Microarray transcriptome data are available with accession numbers GSE133776 ( Reprogramming of primary human hepatocytes ( PHHs ) into hCLiPs ) ; GSE133777 ( Hepatic induction of hCLiPs ) ; GSE133778 ( Characterization of long term-cultured of hCLiPs ) ; GSE133779 ( Transcriptomic analysis of PHHs isolated from hCLiP-transplanted mouse chimeric liver ) . GSE133776-GSE133779 are included in Superseries GSE133797 . Comparative analysis of IPHH and APHH transcriptome is available with an accession number GSE134672 . | One of the most successful treatments for liver disease is transplanting a donor liver into a patient . But demands for donor livers far outstrips supply . A promising alternative could be , rather than replacing the whole organ , to transplant patients with individual liver cells called hepatocytes . These cells can then move into the liver , replace damaged cells , and help support the organ . However , hepatocytes are also in short supply , as despite the liver’s amazing regenerative abilities , these cells struggle to divide outside of the body . Improving how these cells multiply , could therefore help more people receive hepatocyte transplants . In 2017 , researchers found a way to convert mouse and rat hepatocytes into cells that could divide more rapidly using a cocktail of three small molecules . These 'chemically induced liver progenitors' , or CLiPs for short , were able to mature into working hepatocytes and support injured mouse livers . But , discoveries made in rats and mice are not always applicable to humans . Now , Katsuda et al . – including some of the researchers involved in the 2017 work – have set out to investigate whether CLiPs can also be made from human cells , and if so , whether these cells can be used for hepatocyte transplantations . Using a similar cocktail of molecules , Katsuda et al . managed to convert infant human hepatocytes into CLiPs . As with the rodent cells , these human CLiPs were able to turn back into mature , working liver cells . When transplanted into mice with genetic liver diseases , the human CLiPs moved into the liver and became part of the organ . These transplanted cells were able to reconstruct the liver tissue of diseased mice , and in some cases , replaced more than 90% of the liver’s damaged cells . Developing human CLiP technology could provide a new way to support people on the waiting list for liver transplantation . But there are some obstacles still to overcome . At present the technique only works with hepatocytes from infant donors . The next step is to improve the method so that it works with liver cells donated by adults . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"cell",
"biology"
] | 2019 | Generation of human hepatic progenitor cells with regenerative and metabolic capacities from primary hepatocytes |
In simple organisms such as Caenorhabditis elegans , whole brain imaging has been performed . Here , we use such recordings to model the nervous system . Our model uses neuronal activity to predict expected time of future motor commands up to 30 s prior to the event . These motor commands control locomotion . Predictions are valid for individuals not used in model construction . The model predicts dwell time statistics , sequences of motor commands and individual neuron activation . To develop this model , we extracted loops spanned by neuronal activity in phase space using novel methodology . The model uses only two variables: the identity of the loop and the phase along it . Current values of these macroscopic variables predict future neuronal activity . Remarkably , our model based on macroscopic variables succeeds despite consistent inter-individual differences in neuronal activation . Thus , our analytical framework reconciles consistent individual differences in neuronal activation with macroscopic dynamics that operate universally across individuals .
Advances in neuronal imaging ( Kato et al . , 2015; Ahrens et al . , 2013; Berényi et al . , 2014; Jorgenson et al . , 2015; Venkatachalam et al . , 2016; Nguyen et al . , 2016; Schrödel et al . , 2013 ) are now making it possible to simultaneously record activity in a large number of neurons simultaneously during execution of behaviors . Most analytic techniques used to simplify such complex datasets involve dimensionality reduction ( Kato et al . , 2015 ) , clustering ( Venkatachalam et al . , 2016 ) , correlations between activity of neuronal populations and behavior ( Georgopoulos et al . , 1986 ) or features of the sensory stimuli ( Luo et al . , 2014 ) , and connectivity among neurons ( Varshney et al . , 2011 ) . Although having sufficiently detailed experimental observations is absolutely essential , even when analyzed using these sophisticated statistical techniques , detailed information about activation of individual neurons does not always automatically lead to greater understanding of the laws that give rise to the temporal evolution of neuronal activity or the relationship between neuronal activity and the ‘computations’ performed by the brain ( Frégnac , 2017; Jonas and Kording , 2017 ) . Most modeling approaches aimed at understanding how the observed neuronal activity unfolds in time proceed in a bottom-up fashion . In simple nervous systems , such as the stomatogastric nervous system ( Hartline , 1979 ) , feeding central pattern generator in Aplysia ( Susswein et al . , 2002 ) , and locomotor circuitry in nematode Caenorhabditis elegans ( Kunert et al . , 2014 ) realistic models built on biophysics of individual neurons and properties of their connections can be constructed . Attempts have been made to model more complex neural networks such as a cortical column at the level of biophysical properties of individual neurons ( Markram , 2006; Markram et al . , 2015 ) . Although these modeling approaches can prove successful in some settings , the bottom-up approach is limited in several fundamental ways . Even in the simplest nervous systems biophysically realistic models can rarely be sufficiently constrained by the available experimental measurements ( Selverston , 1980 ) . Biophysical properties of individual neurons and their connections change dynamically as a function of neuromodulation and neuronal activity ( Bargmann and Marder , 2013; Marder , 2012 ) . Because of many nonlinear interactions among the components of even simple neuronal networks , detailed models are not necessarily conceptually revealing ( Selverston , 1980 ) and are computationally costly ( Izhikevich , 2003; Markram , 2006 ) . Finally , bottom-up approaches typically assume that the microscopic parameters measured in a typical experiment such as neuronal connectivity or biophysics of individual neurons and synapses must be tuned to specific values in order to assure proper functioning of the brain . Variations around these values are typically seen as noise . Thus , microscopic parameters are routinely averaged across iterations of the same experiment and across individuals . Yet , biophysically realistic simulations of even simple neuronal networks in crustaceans ( Prinz et al . , 2004 ) show that the relationship between the microscopic parameters and global behavior of the network is highly degenerate . Many disparate microscopic configurations lead to almost indistinguishable macroscopic behavior . Because of non-linearities , however , averaging microscopic parameters disrupts the global behavior of the system ( Golowasch et al . , 2002 ) . Therefore , in order to adequately constrain a realistic model of a neuronal network , many microscopic parameters need to be simultaneously measured in the same animal ( Golowasch et al . , 2002 ) . Yet , such a detailed model is not guaranteed to be generalizable across individuals . Thus , while on the one hand there is a desire to create sufficiently realistic models , it is likely that ultimately these bottom-up approaches need to be combined with more abstract phenomenological models of neuronal dynamics . Here , we describe a general methodology capable of extracting neuronal dynamics from neuronal imaging in nematode C . elegans ( Brennan and Proekt , 2017 ) . To demonstrate the power of this approach we show that our model is capable of predicting future motor commands on a cycle-by-cycle basis and is valid across multiple individual C . elegans despite consistent inter-individual differences in neuronal activation . Locomotion of C . elegans is one of the very few biological systems where experimental measurements of brain activity and behavior can be performed with sufficient granularity for developing and testing a quantitative model of brain dynamics at a behaviorally relevant scale . All 302 neurons ( White et al . , 1986 ) in C . elegans and all their connections are known ( Izquierdo and Beer , 2013; Bargmann and Marder , 2013; Varshney et al . , 2011 ) . Simultaneous recordings of the majority of the neurons in the brain ( head ganglia ) of C . elegans have been performed in vivo using calcium imaging ( Kato et al . , 2015; Nguyen et al . , 2016; Prevedel et al . , 2014; Tian et al . , 2009 ) ( Figure 1A , Materials and methods ) . The graded activity of most C . elegans neurons ( see Liu et al . , 2018 , however ) make them better suited for calcium imaging compared to vertebrate nervous systems in which the utility of calcium imaging is limited by the slow speed of calcium indicators relative to the temporal precision of spike trains ( Rad et al . , 2017 ) . Biomechanics of locomotion of C . elegans are well-described by just a few movement modes ( Stephens et al . , 2008 ) suggesting that the dynamics of the nervous system that controls locomotion are likely to be simple enough to be inferred from relatively short recordings of neuronal activity . Locomotor behaviors fall into well-characterized individual distinct stereotyped behavioral subtypes ( Kato et al . , 2015; Li et al . , 2014; Luo et al . , 2014; Larsch et al . , 2013 ) ( Figure 1B ) . The final fundamental advantage of C . elegans as a model organism is that neurons can be individually identified in different genetically identical animals ( Kato et al . , 2015 ) . Thus , C . elegans is an ideal model system for the proof of principle that a model of neuronal dynamics can be constructed on the basis of imaging of neuronal activity .
One plausible explanation of variability in neuronal activity is that a particular neuron is irrelevant for a specific behavior and therefore its activity is not adequately constrained . An example of this type of variability is ALA – a neuron involved in quiescence regulation and mechanosensation ( Van Buskirk and Sternberg , 2007; Sanders et al . , 2013; Hill et al . , 2014; Nelson et al . , 2014 ) . Since experiments analyzed herein were performed in immobilized worms and no quiescence was observed , as expected , ALA activation is quite variable from one cycle of reversal to the other in each individual animal . Note , however , that there are no statistically significant differences between ALA activity during reversals across different individuals ( p-value ≈ 0 . 9 , Materials and methods ) ( Figure 1C ) . As a result , neuronal activity averaged across animals at each phase of behavior is representative of neuronal activity observed in each animal individually . In contrast variability of activation of RIML – a command neuron known to activate AVA which , in turn , elicits backwards locomotion ( Guo et al . , 2009 ) – is paradigmatically distinct . During backwards locomotion , RIML activation differs significantly between animals ( p-value < 0 . 001 , Materials and methods ) . These differences are not simply random noise superimposed onto a common activation template . As a result , averaging RIML activity across animals during backward locomotion yields a pattern of activity that does not resemble that observed in any one of the individual C . elegans . Yet , during a different behavior – dorsal turn – RIML activation is consistent across individuals ( p-value ≈ 0 . 1 ) , Materials and methods . This makes it unlikely that the observed differences in RIML activation during backwards locomotion are an artifact of neuron misidentification . Consistent differences in activity of individual identified neurons between genetically identical animals performing the same behavior are not unique to RIML . To show this , we quantify inter-individual differences in activity of each neuron during each locomotor behavior ( Figure 1D ) . The p-values in Figure 1D reflect the probabilities that activation of a particular neuron is consistent among individuals . For most neurons involved in locomotion activity differs from animal to animal during execution of at least one type of locomotor behavior . Many neurons can be consistently activated in one locomotor behavior but be highly inconsistent among individuals in another type of locomotion . Only three neurons were consistent in all behaviors . One of these neurons ( ALA ) is not known to play a direct role in locomotion beyond quiescence ( Van Buskirk and Sternberg , 2007; Sanders et al . , 2013; Hill et al . , 2014; Nelson et al . , 2014 ) not observed in this dataset . Consistent with this observation , ALA did not exhibit any appreciable activation during any locomotor behavior . AVB and RID were the only locomotion-associated ( White et al . , 1986; Lim et al . , 2016 ) neurons whose activity failed to exhibit statistically significant differences among individual animals in any locomotor behavior . These inter-individual differences in neuronal activation is the primary reason why principal component analysis performed on neuronal activity in each individual successfully reveals cycles in neuronal dynamics ( Kato et al . , 2015 ) but attempts at projecting data from all individuals onto a common set of principal components fails to reveal any meaningful structure ( Figure 1—video 1 ) . To further illustrate the consistent differences in neuronal activation among individuals , we attempted to decode the behavioral state on the basis of neuronal activity . Half of all instances of backing behavior were used to compute the average activity of each neuron at the onset of backing behavior . Mutual information between this snapshot of neuronal activity and behavioral state ( Materials and methods ) was then used as the basis for decoding the other half of backing behaviors either within each animal or across animals . Using this strategy , we reliably decoded the onset of backwards locomotion based on ∼100 neurons recorded in each animal individually ( p-value < 0 . 001 relative to shuffle control , Materials and methods ) . The ability to decode did not degrade appreciably when just 15 neurons identified in each animal were used ( p-value ≈ 0 . 5 within animal - all neurons vs . within animal - 15 neurons , Materials and methods ) . This limited subset of neurons ( ∼1/20th of the entire nervous system ) , therefore , still contains most of the essential information about initiation of backwards locomotion and confirms that neuronal activation is consistent in each animal . This is not surprising as the 15-neuron subset contains most of the known command neurons that control the direction of locomotion . Yet , activity from one animal cannot be used to reliably decode the onset of backing behavior in another animal . When activity from four animals was used to decode the 5th ( leave one out ) the correct decoding rate was indistinguishable from chance ( p-value ≈ 0 . 3 , Materials and methods ) ( Figure 1E ) . Thus , mutual information between neuronal activity and behavioral state is degraded when neuronal activity is averaged among genetically identical individuals during locomotion in a simple environment . This inter-individual variability is the fundamental reason why simple averaging of activation of individual neurons fails to yield a meaningful model of neuronal dynamics . Although there are potentially many different classifiers that could be built to decode the behavioral state on the basis of neuronal activity , a classifier based on mutual information is a parsimonious strategy that succeeds in decoding behavior in each individual . Thus , it is unlikely that our ability to decode the behavioral state on the basis of neuronal activity will be dramatically improved by a different classification strategy . Consistent differences in activation of individual neurons do not necessarily imply that global dynamics of the brain are distinct in different individual C . elegans . It is possible that distinct activity combinations observed in different individuals give rise to an equivalent behavioral strategy implemented at the level of global brain dynamics . An example of this state dependence of neuronal activity is known in the olfactory system of C . elegans ( Gordus et al . , 2015 ) . This degeneracy of neuronal activation complicates analysis of individual microscopic components taken in isolation or averaged across individuals . There is a fundamental distinction between neuronal activity and neuronal dynamics ( Churchland et al . , 2012; Salinas and Sejnowski , 2001 ) . Neuronal dynamics are the laws of motion that govern the temporal evolution ( flux ) of neuronal activity in the space spanned by the relevant variables ( phase space ) . Thus , rather than focusing on individual neurons , the dynamical systems description is focused on identifying the salient variables that make up the phase space and on the laws of motion that act to move the state of the system along a trajectory in phase space . The observed neuronal activity is governed by the biophysics of individual neurons and synapses ( Seung , 1996; Beer , 1995; Miller and Selverston , 1982 ) as well as activity of other neurons not reliably identified in all experiments . These biophysical processes influence neuronal activity and are in turn influenced by it . Yet , these processes cannot be directly inferred from the observed activation of neurons . In the appendix , we illustrate a novel method – Asymmetric Diffusion Map Modeling – that allows for the extraction of neuronal dynamics from high-dimensional , noisy and non-linear neuronal activity time series recordings . The final output of this method is a two dimensional approximation of the neuronal dynamics which describes the time evolution of the system as a flux along distinct loops in phase space . One fundamental advantage of having an approximation of neuronal dynamics is that neuronal activity in C . elegans can be efficiently simulated ( Figure 2A ) . The validity of the simulated dynamics can then be explicitly tested by comparing these newly simulated traces of neuronal activity to those experimentally observed in C . elegans . This simulation is first performed in the phase space . As the system evolves in phase space it traces out neuronal activation ( Figure 2A ) ( Materials and methods and Appendix ) . Note , that the simulated neuronal activity does not merely recapitulate experimental observations but rather yields new neuronal activity traces . These simulated activity traces are in good qualitative agreement with experimental observations . Both the observed and the simulated traces exhibit abrupt coordinated transitions between levels of activity of multiple neurons . Further note that the correlations in activation across neurons are preserved . Finally , note that the activity of the simulated AVA neuron ( Figure 2B ) exhibits bouts of activations interspersed with prolonged periods of inactivity corresponding to backward and forward locomotion respectively . These bouts are in good qualitative agreement with the experimental observations . The first instance of backward locomotion in a bout is distinct from subsequent instances . It is associated with stronger activation of the AVA neuron ( Figure 1B ) . Remarkably , transient activation is also a salient feature of the simulated AVA during the first instance of backing behavior in a bout ( Figure 2B ) . Because there is an element of stochasticity in the neuronal dynamics , the total number of instances and durations of locomotor behaviors are variable both in the experimentally observed and simulated neurons . To quantitatively compare the simulated and observed neuronal activation , we computed the spectra of each of the 15 neurons identified across all individuals to the spectra of simulated neurons ( Materials and methods ) . With the exception of the very low frequencies ( <0 . 05 Hz ) most strongly affected by the finite dataset effects , the spectra of all simulated neurons are statistically indistinguishable from experimentally observed neuronal activity ( Figure 2—figure supplement 1 ) . To determine whether the model of neuronal dynamics reproduces behavioral statistics , we assigned each time point in a simulation a behavioral state . This was accomplished by sampling the empirically derived distribution of behaviors at each point in phase space . The experimentally observed and de novo simulated distributions of dwell times in different behavioral states are in excellent agreement ( Figure 2B ) . Note that the simulations reproduce not just the time scale of individual behaviors ( forward and backward locomotion ) but also sequences of behaviors that we refer to as backing bouts . This is remarkable because the model of the dynamics was constructed by estimating probability of transition between two states on the time scale of one time step dictated by data acquisition and GCAMP kinetics ( ∼ 1/3 of a second ) . Yet , the simulation reproduces the dynamics on the time scale longer than 100 s . Note that PCA previously applied to neuronal activity ( Kato et al . , 2015 ) does not directly yield a quantitative model that can be used to simulate new neuronal activity . Thus , inter-individual variability aside , PCA in and of itself does not yield any quantitative predictions concerning neuronal dynamics . Based on the observations of abrupt stereotyped transitions in activity of many neurons ( e . g . Figure 1A ) and dwell times of locomotor behaviors , it has been argued that switching between different modes of locomotion in C . elegans is stochastic ( Roberts et al . , 2016; Srivastava et al . , 2009 ) . If so , then timing of behavioral transitions on each individual cycle of behavior should be unpredictable and the entirety of information concerning behavioral switching should be contained in the dwell time distributions . Thus , the most compelling test of the neuronal dynamics model is the ability to predict future abrupt changes in neuronal activation that signal switches in locomotor behavior solely on the basis of initial position in phase space . To test this prediction , we make use of a new dataset of calcium imaging in C . elegans from Nichols et al . ( 2017 ) ( Materials and methods ) . We restricted our analysis to the prelethargus N2 animals ( n = 11 ) that were subjected to similar experimental conditions and imaging to those from Kato et al . ( 2015 ) dataset . Critically , no data from the Nichols et al . dataset was used for the construction of the model . Animals in the dataset ( Nichols et al . , 2017 ) shared between 8 and 13 neurons with the neurons recorded by Kato et al . ( 2015 ) on the basis of which the neuronal dynamics model was constructed ( Materials and methods ) . Simulations started from several initial positions ( phase bins ) associated with backwards locomotion were used to estimate the expected distribution of times to the start of forward locomotion ( Figure 3A , orange ) for each phase bin . To compare these predictions to the experimental observations , we identify all points in the validation dataset from Nichols et al . that pass through the same phase bins and note the experimentally observed time until the start of forward locomotion signaled by abrupt change in AVA activity ( Figure 3A , blue ) . For most phase bins , the expected time of simulated behavioral switch was indistinguishable from experimentally observed switch in motor command . In contrast , the predictions made by the null model based solely on behavioral dwell time distributions deviate significantly from the timing of observed transitions . To quantify the success of the predictions , we compute the correlation between simulated time to initiation of forward locomotion and that observed by Nichols et al . for each phase bin ( Figure 3B ) . Consistent with observations in ( Figure 3A ) simulation-based predictions ( filled circles ) were strongly correlated with observed timing of behavioral transitions ( correlation coefficient 0 . 74 ) ( Figure 3B ) . In contrast , predictions based solely on the dwell time distributions were less well correlated ( p < 0 . 0001 ) with experimental observations . Further , note that the dispersion around the best fit line is smaller for the simulation-based than for dwell-time based predictions . Thus , dynamics-based predictions are more precise and accurate than those based on behavioral statistics alone . Because definition of behavioral states relies heavily on observed activity of the AVA neuron , we sought to determine whether including AVA critically affects the results . We removed AVA from the Kato et al . ( 2015 ) dataset used for model construction and the Nichols et al . ( 2017 ) dataset used for model validation . Even in the absence of the AVA , manifold predictions correlated strongly with the observed time of behavioral transitions ( Slope 0 . 9; R2 0 . 8 ) and outperformed predictions based solely on dwell time distribution Figure 3—figure supplement 1 . Therefore , our modeling approach reveals a strong contribution of deterministic dynamics to abrupt changes in locomotor direction in C . elegans . These predictions do not depend strongly on activity of AVA–the command neuron for backward locomotion . It should be noted , however , that by construction the Asymmetrical Diffusion Map Method is a stochastic model . Thus , in addition to the deterministic cyclic fluxes , stochastic forces also contribute to the observed neuronal activity . Remarkably , the method reveals that the transition probability between neuronal activity patterns is a function of the macroscopic variables such as phase of the cyclic flux . Knowing the initial conditions is sufficient to predict the expected time of transitions between different modes of locomotion 30 s before they are experimentally observed ( Figure 3 ) . Remarkably , these predictions are valid across individuals observed years apart . Therefore , neuronal dynamics model can be applied universally across individuals despite significant inter-individual differences in neuronal activation and undersampling of neuronal activity . Although it is likely that the simulation-based predictions could be improved with addition of more neurons , the fact that the animals in the validation dataset shared as few as eight neurons with the original data suggests that using our methodology one can uncover macroscopic dynamics even when only a small subset of the nervous system can be recorded and unequivocally identified . In principle , our methodology ( Materials and methods and Appendix ) could be used to uncover system dynamics from activity of any single component of a tightly coupled system ( Harnack et al . , 2017 ) . Thus , we attempted to reconstruct dynamics of C . elegans nervous system on the basis of activity of a single neuron . We used a single neuron from the Kato et al . ( 2015 ) for model construction . The quality of predictions was assessed using dwell time statistics ( Materials and methods ) Figure 2—figure supplement 4 . The quality of predictions varied substantially between neurons . Models built on some neurons involved in backwards locomotion ( e . g . AVAL , AVAR , AVER , and RIML ) yielded predictions comparable to those obtained for a set of 15 neurons . In contrast , neurons that play limited role in locomotion such as the ALA were not predictive . Interestingly , although RIML is known to play a role in backward locomotion , its activity varied significantly among individual animals ( Figure 1C ) during backwards locomotion . Nevertheless , models based solely on RIML were ∼75% as informative as models built upon the entire 15 neuron set . Thus , at least in the simple nervous system of C . elegans a predictive model can be constructed on the basis of a single experimentally observed neuron as long as activation of this neuron is tightly coupled to the network that mediates the observed behaviors . The ability to simulate neuronal activity , behavioral dwell-time statistics , and even predict timing of individual behavioral transitions implies that trajectories traced by the state of the brain as it evolves in phase space are remarkably conserved among individuals . If the dynamics that give rise to neuronal activity were purely deterministic , then such trajectories would never cross ( Sugihara et al . , 2012; Strogatz , 2014 ) . However , any experimental system is bound to have noise due to both measurement error and stochastic processes that affect the trajectories traversed in phase space . Noise inevitably causes trajectories to tangle . Nevertheless , in the limit of low noise ( Materials and methods ) , trajectories will form bundles in phase space . A collection of such trajectory bundles is referred to as the manifold . To determine whether the manifolds are conserved among individuals , we applied the manifold reconstruction method ( Materials and methods and Appendix ) to neuronal activity of C . elegans . The manifold in Figure 4A was constructed on the basis of all 107 neurons recorded in one animal . This illustrates that our methodology is able to reconstruct the global dynamics in the limit of relatively large fraction ( ∼ 1/3 ) of all neurons ( Figure 4—figure supplement 1 ) and can be applied to time series consisting of at least 100 neurons . In the C . elegans nervous system , the phase space ( Materials and methods ) is too high dimensional to be shown graphically in its entirety . Nevertheless , trajectories spanned by a broad class of noisy dynamical systems ( Wang et al . , 2008 ) will form loops – a low-dimensional object in the high-dimensional phase space . Thus , a position of the system can be approximated just by two parameters: the identity of the loop α and the phase along it θ . Identifying these variables from neuronal activity ( Materials and methods ) allows us to project neuronal activity averaged with respect to θ and α onto the first three principal components . This coordinate system which we refer to as DPCA plays no role in simulating neuronal dynamics and is used purely for visualization purposes ( Figure 4—video 1 ) . The width of the manifold represents the density of points or , equivalently , decreases in phase velocity dθ/dt . The direction of phase velocity is shown by arrows . For instance , in the region associated with forward locomotion ( blue ) phase velocity is relatively small . Thus , transit through this region of phase space is dominated by stochastic processes . In contrast , reversal behaviors ( red and purple ) are associated with high dθ/dt . Therefore , duration of reversals has a characteristic time scale dominated by phase velocity . The sequence of behaviors is dictated by the arrangement of different locomotor behaviors along the phase of the manifold . The distribution of locomotor behaviors as a function of position in the manifold is shown by color . The final color of the manifold is a blend of the colors for each behavior according to their prevalence . Note that although behavioral assignments were not used in the construction of the manifold ( Materials and methods ) , most regions of the manifold are associated with just one type of locomotor command . In other words , different locomotor commands are localized to different regions in the phase space . While the two trajectory loops are well separated , the system is quite deterministic . When the two loops pass near each other , conversely , the future state of the system is dominated by stochastic processes . Several lines of evidence converge on the fact that , unlike activity of individual neurons , the phase space ( θ , α ) is universal across animals . The manifold in Figure 4B was constructed on the basis of activity from all five animals using only 15 neurons identified in each animal . In contrast to averaging neuronal activity by applying PCA ( Figure 1—figure supplement 1 ) reconstruction of neuronal dynamics is possible even when only 15 neurons ( ∼5% ) are consistently identified in each individual . This is especially remarkable given the inter-individual differences in activation of the common neuronal subset . The structure of the manifold constructed on the basis of 15 neurons across individuals is nearly identical to the manifold constructed on the basis of 107 neurons in a single animal ( Figure 4A ) . Position in phase space ( θ , α ) preserves behavioral information across animals . As a result , the assigned behavioral state can be correctly decoded 83% of the time solely on the basis of position along the manifold in Figure 4B . This is the median successful decoding probability computed across across all locations in the manifold binned into 426 bins ( total of 15405 predictions in all five animals , χ2 19 . 8 , p-value 5 . 5×10-4 ) . Note that because of limited temporal precision with which behavioral states can be experimentally assigned , some uncertainty about the behavioral state is expected especially around the times of behavioral transitions . To further strengthen the argument for universality of the global dynamics , we constructed a manifold based on the data from four out of the five animals in the Kato et al . ( 2015 ) dataset . We then used the manifold to project the neuronal activity from the fifth animal not used for manifold construction onto the manifold space ( Figure 4C ) . Behavioral states of the left out animal align well with the distribution of behavioral states along the manifold . Correct behavioral state assignment in the excluded animal can be decoded 81% of time ( median correct decoding probability across all 426 manifold bins and all five animals left out in turn ) . The difference in the median correct prediction probability based on the all worm manifold and the leave one out manifold ( Figure 4C ) is not statistically significant ( χ2 3 . 8 , p-value 0 . 44 ) . The probability of obtaining this quality of decoding by chance is p=0 . 0014 ( χ2 17 . 7 ) . Thus , averaging neuronal activity with respect to its position in phase space , rather than across individual neurons , preserves most of the behavioral information and can be universally applied across individuals even when only ∼5% of neurons are uniquely identified . This conserved shape of the manifold in the phase space is what allows the predictions of timing of switching of motor commands across different animals . Yet , the salient variables that span the phase space are not directly apparent from recordings of individual neurons even when most locomotor control circuitry is recorded in a simple environment .
Here , we developed a method for extracting salient dynamical features from complex , multivariate , nonlinear , and noisy time series . We apply this method to neuronal imaging in C . elegans to demonstrate its success in simulating activity of the nervous system and predicting switches between different motor commands . The manifold in C . elegans nervous system is composed of two loops . While the system is in either one of the loops , its fate is largely predictable . Yet , in the neighborhood where the loops merge , the behavior cannot be clearly predicted and stochastic forces play a stronger role . This leads us to hypothesize that the region where the two loops merge is a decision point where the nervous system is most susceptible to noise and/or sensory inputs ( Gordus et al . , 2015 ) . The manifold shape is conserved among individuals and initial position in the manifold space is sufficient to predict future switches in motor commands . This suggests that the macroscopic variables such as loop identity and phase along it express behaviorally relevant information . Intriguingly , we find that even in genetically identical organisms consistent differences in neuronal activity associated with motor commands are the norm . This striking observation is not without precedent . Hodgkin-Huxley models of conductances measured in individual AB neurons in crustacean stomatogastric ganglion exhibit bursting akin to the biological neuron . However , averaging conductance measurements across AB neurons in different individuals yields models that fail to burst ( Golowasch et al . , 2002 ) . Virtually indistinguishable network activity patterns can arise from distinct biophysical mechanisms ( Prinz et al . , 2004; Chiel et al . , 1999; Beer et al . , 1999 ) . This suggests that differences between individual AB neurons ( Goldman et al . , 2001; Prinz et al . , 2004 ) or individual C . elegans are not simply random deviations from a common template that can be averaged away at the microscopic level . This nontrivial inter-subject variability is the fundamental difficulty impeding the construction of biophysically-realistic models of even simple nervous systems . In order to sufficiently constrain such models many parameters have to be simultaneously measured in each individual . This is not currently possible even in the simplest neuronal networks . Even more troubling is the observation that such detailed models may not be generalizable between highly similar individuals . Therefore , a more abstract phenomenological approach to modeling neuronal dynamics will be helpful for understanding circuit-level function . We hypothesize that the nontrivial degeneracy between microscopic biophysical processes and circuit-level dynamics arises because evolutionary selection operates at the macroscopic level of organismal behavior ( Lässig and Valleriani , 2008 ) embodied by the global dynamics of the brain . Thus , there is no explicit selective pressure for each individual to produce identical neuronal activation during behavior . Nor is there an explicit pressure for an AB neuron to express a particular number of each of the ion channels on its surface . All that is required is that the overall system gives rise to an adaptive behavioral strategy ( Beer , 2000 ) . Although undoubtedly there are important constraints imposed by the biomechanics of the animal , the connectome , and other variables , any microscopic solution that gives rise to the appropriate macroscopic dynamics yields the same behavioral strategy . This is equivalent to David Marr’s ( Marr , 1982; Frégnac , 2017 ) proposal that the biophysical details of neuronal circuits are constrained by the computation implemented by the circuit as a whole , rather than the traditional bottom-up approach ( Markram , 2006; Markram et al . , 2015 ) which assumes the opposite . Thus , one should not necessarily expect a detailed model of the nervous system to be equally valid for different , seemingly identical , individuals . Our methodology can be used to construct a model of macroscopic dynamics despite consistent differences in neuronal activation in different individuals . To appreciate the full computational significance of macroscopic dynamics , future work can apply similar methodology to determine how these dynamics are altered by interaction with the environment ( Clark , 1998; Beer , 2000; Linderman et al . , 2019 ) . The model in this work was constructed on the basis of immobilized animals . Although Kato et al . ( 2015 ) established some essential similarities between activation of neurons in the immobilized and freely moving C . elegans , there are also important differences ( Nguyen et al . , 2016; Venkatachalam et al . , 2016; Scholz et al . , 2018 ) . One important difference is that repeated bouts of backing behavior are not observed in the freely moving animal . Yet , neurons associated with backing behavior ( e . g . RIM , AVA , AVE , AIB ) were consistently activated during backing in freely moving animals and during fictive locomotion in the immobilized worms . The manifold of C . elegans dynamics consists of two loops dominated by forward and backward locomotion . The decrease frequency of backward behavior in the freely moving C . elegans , therefore , may correspond to the decreased probability of entering the backward locomotion loop rather than a fundamental differences in the shape of the manifold . Decoupling the motor commands from the behavioral output can prolong the duration of backing behaviors as evidenced by prolonged depolarization of RIM in the immobilized state . This could correspond to the decrease in the phase velocity along the corresponding loop of the manifold . Kato et al . ( 2015 ) show that silencing the AVA – a command neuron for backward locomotion – eliminates backing behaviors in the freely moving animal . Silencing of the AVA slightly attenuated the activation of RIM and AVE but did not affect the phase relationship between activation of RIM and AVE and other neurons . Thus , although it is possible to uncouple the dynamics of the motor command circuitry from the actual execution of behavior , the macroscopic dynamics remain qualitatively similar . Yet , in general , it is very likely that the manifold shape and properties will depend strongly on the interactions with the environment . Thus , behavioral significance of neuronal dynamics could only be clearly established by reconstructing the neuronal dynamics in animals engaged in their natural behaviors . Nevertheless , our methodology for extracting neuronal dynamics should still apply . The principal innovation of our methodology is to find loops in non-linear , multivariate and noisy neuronal activity . Oscillations in neuronal activity are well known in nervous systems from leach swimming ( Kristan and Calabrese , 1976 ) , to stomatogastric ganglia of crustaceans ( Selverston and Moulins , 1985 ) , to locomotion in primates ( Churchland et al . , 2012 ) and others . Although oscillations in neuronal activity are expected during rhythmic behaviors , behaviors that are not themselves obviously rhythmic – such as preparation for movement ( Churchland et al . , 2010 ) or reaching ( Churchland et al . , 2012 ) – are also associated with rotations in phase space . Thus , we expect that our methodology will be broadly useful for characterizing dynamics in diverse nervous systems . Several issues need to be considered before applying this methodology to other organisms . The graded potentials of C . elegans neurons can be thought of as similar to fluctuations in the firing rate of vertebrate neurons . Yet , it is not always clear whether timing of individual action potentials conveys meaningful information ( Theunissen and Miller , 1995 ) . In principle , the methodology could be adapted to utilize spike train distances ( Victor and Purpura , 1997 ) . However , as the number of dimensions of neuronal activity grows , the notions of local neighborhoods become complicated ( Aggarwal et al . , 2001 ) and may require modifications to the distance measures . Furthermore , the choice of distance measure and the size of the local neighborhood can effect the coarseness with which neuronal trajectories are combined into the same bundle or split between different bundles of the manifold . Our ability to build a single model that captures the dynamics in different individuals relies on the ability to identify the same neuron in different C . elegans . Neuron identification is challenging even in simple systems such as C elegans and is generally impossible for complex nervous systems of vertebrates . The fact that the model can be built on a small subset of neurons suggests a possibility that models constructed for different individuals can nevertheless be combined in the manifold space rather than in the space spanned by neuronal activity . In order to accomplish this , future work will need to develop a methodology to robustly compare diffusion maps constructed on the basis of neuronal activity without relying on neuronal identification . In C . elegans , we are able to successfully build a manifold on the basis of ∼100 neurons . The effective dimensionality of the data , however , is much smaller . Indeed , we are able to construct a manifold on only 15 neurons and still faithfully simulate the dynamics . Furthermore , the animals in the validation dataset shared as few as eight neurons with the manifold . Nevertheless , the predictions based on the manifold were highly accurate . Because nonlinear dynamical systems are best thought of as wholes rather than a collection of individual components ( Harnack et al . , 2017 ) , the phase space of the nervous system can theoretically be extracted from any individual neuron ( Takens , 1981 ) . Consistent with this notion , we showed that recording of a single neuron can be used to construct a meaningful model . The reconstruction is only possible , however , when the components of the system are tightly coupled . Only some neurons yielded meaningful predictions in C elegans . Thus , recordings from more complex nervous systems may have to first be separated into weakly coupled component parts before the dynamics can be adequately modeled . There is clearly still much work to be done before dynamics of arbitrarily complex and noisy neuronal circuits can be reliably modeled . Nonetheless , our success in modeling the global dynamics of C . elegans in a simple environment illustrates the potential power of our method and promises a fruitful new approach to analysis of complex nervous systems .
Here , we developed a novel method for the extraction of the global dynamics which give rise to observed neuronal activity . We call this method Asymmetric Diffusion Map Modeling . This section will strive to give an overview of the method and a basic intuition as to why it works . A full treatment of the mathematics of the method can be found below . First we will define several distinct representations of the data which the method utilizes . Then , we will discuss how and why the data is transformed from one representation to the next . Activity space contains experimental observations of neuronal activity . A vector in this space represents the instantaneous activation of all individual neurons at a single time point . Each component of this vector represents the instantaneous activity of a single identified neuron . The ultimate goal of the method is to efficiently model the temporal sequences of neuronal activation . To do this , we first need to extract relevant variables sufficient to fully describe the dynamics which give rise to neuronal activity . This collection of variables is known as the phase space . In phase space each dimension represents a unique relevant variable . In contrast to neuronal activity , these variables may not necessarily be directly observed . We will approximate the time evolution of the system in phase space by constructing a transition probability matrix . Each element ( i , j ) of this matrix corresponds to the probability that a system observed at location i in phase space will transition to location j after one time step ( see below , Figure 4—figure supplement 2 ) . This n×n representation , where n is the number of observations , gives an approximation of the velocity of the system at each observed point in phase space . Finally , we will simplify this table of velocities to extract manifold space – allowing for a minimal representation of the dynamics . Temporal evolution of the system in the manifold space can then be readily simulated to yield quantitative predictions about future neuronal activity . The global dynamics of a nervous system depends on biophysical processes beyond neuronal firing . It is experimentally intractable to record all such processes including time and voltage dependent currents , neurotransmitter and neuromodulator release , hormonal signaling , plasticity , etc . However , the key variables that make up phase space can be extracted from the observations using methods known as delay embedding ( Takens , 1981; Packard et al . , 1980 ) . The main idea behind delay embedding is that one can use the experimental observations ( neuronal activation and its time-derivative ) to extract independent measurements that together form the phase space . To extract independent measurements from a single time series ( e . g . neuronal activity ) , the delay time τ is chosen such that correlation between two points in the activity space separated by τ is negligible . These delayed versions of the time series correspond to different dimensions of the reconstructed phase space . According to Takens’ theorem ( Takens , 1981 ) , this reconstructed space preserves essential features of the dynamics which are required for model construction . When phase space is well approximated , points that are close to each other have similar velocities . Consequentially , if two trajectories in the time series data are close in phase space they will continue to evolve in time along similar trajectories – giving rise to recurrent coherent trajectories in the dynamics . The process of delay embedding dramatically inflates the dimensionality of the data making it unusable for complex time series such as activation of many neurons . Thus , the final critical step of the method will reduce the dimensionality of the system . The goal of this step will be to enumerate the phase space dynamics into a discrete transition probability matrix 𝐌 . The i𝑡ℎ row of this matrix tabulates the probability that a system starting out in state i will transition in any other state j after one time step . In this case , the state of the system is described by delay embeddings of observed neuronal activity . To assign transition probabilities , we use diffusion mapping ( Nadler et al . , 2006; Coifman and Lafon , 2006; Lian et al . , 2015 ) – a non-linear dimensionality reduction technique . Similar to local linear embedding ( Roweis and Saul , 2000 ) or isomap ( Tenenbaum et al . , 2000 ) , diffusion maps seek to preserve local relationships between nearby points . Points that are close together in phase space will be assigned high transition probabilities . However , points that are far away ( Equation 13 ) in phase space are not directly connected ( i . e . transition probability is zero ) . After appropriate normalization which ensures that the sum of all probabilities in a row adds up to 1 , this diffusion map can be used to simulate the time evolution of the system . To simulate evolution after N time steps 𝐌 is exponentiated N times . In standard diffusion maps the transition probabilities between points are assumed to be symmetric ( i . e . transition probability Pi→j=Pj→i ) . Yet , this approach does not take into account the fact that neuronal activity is ordered in time . We therefore modify the transition probability calculation to include temporal information . To take temporal information into account , we compute transition probability between the state of the system 𝐃t at time t and points in the neighborhood of the next experimentally observed state 𝐃t+1 . These transition probabilities are computed as a Gaussian centered at 𝐃t+1 , ( 1 ) kFP ( Dt , Dj ) =exp ( −‖Dt+1−Dj‖222σ2 ) , where 𝐃j is a point in the local neighborhood of 𝐃t+1 , and ∥⋅∥22 is the Euclidean distance . σ2 is a normalization term that sets the size of the local neighborhood ( see below for details ) . The result is that time evolution of neuronal activity given by asymmetrical 𝐌 preserves the temporal order of neuronal activity . Although 𝐌 can be used to simulate neuronal activity , it is not in itself a particularly useful model . 𝐌 does not directly inform dominant features of neuronal dynamics and simulations of 𝐌 can only generate reordered versions of the experimentally observed time series . This limitation is due to the fact that 𝐌 is only defined in terms of the observed states of the system . However , spectral analysis of 𝐌 can be used to extract salient features of neuronal dynamics ( fluxes ) . Because 𝐌 is not symmetrical , it can give rise to rotational dynamics . To identify the most salient rotational fluxes , we perform spectral analysis of 𝐌 ( see below ) . As a result , each point in 𝐌 is assigned a phase along the rotational flux . To identify the most dynamically salient fluxes , we find the complex eigenvalues of 𝐌 with the largest modulus . A pair of complex conjugate eigenvectors associated with this eigenvalue relate states of the nervous system 𝐃t to the phase of the rotational flux . This allows us to bin points with similar phase . Because in C . elegans there are multiple rotational fluxes , it is not a priori clear which rotational flux is associated with a given phase . This can be resolved using clustering analysis of 𝐌 ( see below ) . As a result of eigendecomposition and clustering , each point in 𝐌 is assigned to a single bin defined by the identity of the flux and the phase along it . We refer to the transition probability matrix simplified in this fashion as the manifold . Simulations of the manifold are sufficient to predict behavioral statistics , sequences of behaviors , timing of individual behavioral transitions , and neuronal activation . Furthermore , simulations in manifold space yield novel neuronal activity patterns not directly observed in the experiment . In this section , we will present a theoretical argument ( see also ) ( Wang et al . , 2008 ) which suggests that cyclic fluxes are likely to be a common feature of neuronal dynamics . This argument motivates the manifold reconstruction method ( see below ) . Neuronal systems are inherently noisy . Thus , the most sensible approach is to model the dynamics of the nervous system using both deterministic dynamics and stochastic processes ( Yan et al . , 2013 ) , ( 2 ) d𝐗dt=𝐅 ( 𝐗 ) +ϵ , where 𝐅 ( 𝐗 ) is the driving force which quantifies the deterministic aspect of neuronal dynamics , 𝐗 is the position in state space , and ϵ is noise . Because of noise , it is not possible to precisely model the trajectory of any single point starting out at some location in 𝐗 . It is possible , however , to model the temporal evolution of a cloud of points – or more precisely a probability distribution of points – P ( 𝐗 ) ( Pathria , 1996 ) . We begin with the law of probability conservation , ( 3 ) dP ( 𝐗 ) dt=-∇𝐉 ( 𝐗 , t ) , which states that the change in probability P is due to the local flux , 𝐉 ( 𝐗 , t ) , in that region . In systems with homogeneous ( constant in space ) noise , the flux is defined by: ( 4 ) 𝐉 ( 𝐗 , t ) =𝐅 ( 𝐗 ) P-∇P , where 𝐅 ( 𝐗 ) is the driving force . We now assume that the system is at steady state during the time course of the experiment . Mathematically , this corresponds to the assumption that probability distribution is constant . Thus , ( 5 ) dP ( X ) dt=0=∇J ( X , t ) . From a neuroscience standpoint , this statement corresponds to the assumption that the nervous system is not changing ( e . g . learning ) during the experiment . This is a reasonable assumption for the datasets in this manuscript which last ∼15 min per recording . Over long-term recordings , this assumption can be invoked in a piecewise fashion over shorter time intervals . One well-known solution to Equation 5 is a purely stochastic case where the deterministic flux of the system vanishes at all 𝐗 , 𝐉 ( 𝐗 , t ) =0 . In this case , the only meaningful measure of neuronal activity is the probability of different activity patterns . This assumption is invoked in stochastic models of neuronal activity such as maximum entropy models ( Tkačik et al . , 2013; Tang et al . , 2008 ) , Hopfield networks ( Hopfield , 1982 ) , and others . Yet , another class of solutions exist when the flux does not vanish at steady state ( Yan et al . , 2013 ) . The key insight is that in order to keep the distribution of states P ( 𝐗 ) constant , the flux must be purely cyclic , ( 6 ) 𝐉 ( 𝐗 , t ) =∇×A , where A is an arbitrary vector field . Such fluxes are divergence free , and will always form complete loops . Intuitively , this means that a system that evolves around a cyclical orbit will at once have a deterministic flux J ( 𝐗 , t ) ≠0 and satisfy the steady state assumption . For such systems , the driving force is ( 7 ) 𝐅 ( 𝐗 ) =𝐃∇P ( 𝐗 ) P ( 𝐗 ) +𝐉 ( 𝐗 ) P ( 𝐗 ) , where 𝐉 ( 𝐗 ) is the flux at steady state . Note that Equation 7 is a form of the Fokker-Planck equation . The driving force is made of two distinct terms . The first term corresponds to diffusion , while the second corresponds to a deterministic cyclic flux . The purpose of the manifold reconstruction method is to discover this deterministic cyclic flux in neuronal recordings . The ultimate goal of the manifold extraction method is to express neuronal dynamics as a linear stochastic dynamical system . This requires the construction of a transition probability matrix 𝐌 based on empirical observations of neuronal activity where each element ( i , j ) is given by , ( 8 ) 𝐌ij=∥si→sj∥∥si∥ , where ∥si→sj∥ is the number of times the system transitions from state i to state j and ∥si∥ is the total number of times the system is found in state i can be used to simulate the time evolution of the system for a single time step by ( 9 ) 𝐗t+1=𝐌𝐗t , or for some arbitrary time t ( 10 ) 𝐗t=𝐌t𝐗o , where 𝐗o is the initial state of the system . Alternatively this equation can be rewritten in terms of the eigenmodes of 𝐌 , ( 11 ) 𝐗t=∑iciλitϕi , where λi are the eigenvalues , ϕi are the eigenvectors and ci is the projection of the initial state of the system onto the i-th eigenvector . Under a broad range of conditions , the largest eigenvalue of 𝐌 is λ=1 . This corresponds to an assertion that such systems come to a single steady state . The associated eigenvector corresponds to the steady state distribution of the system . If 𝐌 is symmetrical , that is 𝐌i , j=𝐌j , i , then all eigenvalues of 𝐌 are real , and the resulting system is purely stochastic . Asymmetry can give rise to complex eigenmodes . Then Equation 11 becomes an equation of a decaying wave in the plane spanned by a pair of complex conjugate eigenvectors . These decaying spirals correspond to the cyclic fluxes of Equation 7 . In the long time limit , all eigenmodes with complex eigenvalues whose modulus is much less than one damp out . Complex modes with eigenvalues near one heavily shape the dynamics of the system even in the long time limit . These eigenmodes are used to identify the cyclic fluxes of neuronal activity . In order to construct 𝐌 , two steps are required: definition of the state of the system and definition of distances between two points in the state space . The distances between points in state space are used to define transition probabilities . We extract state space from the data using delay embedding ( see below ) , and then use diffusion mapping to define distances between points in the delay embedded coordinates . There are several algorithms for finding a good delay embedding parameters and number of delays ( Packard et al . , 1980; Sauer et al . , 1991; Buzug and Pfister , 1992 ) . The key point is that maximally independent measurements are chosen . Here , we used autocorrelation as a measure of interdependence to estimate delay τ such that autocorrelation becomes ∼0 . For C . elegans manifolds , we used τ=10 frames ( ∼4 s . ) . We explored a range of number of delays . The number of delays used to generate the figures is five but the results are fairly robust to changes in this parameter . Kato et al . ( 2015 ) notice that derivatives of neuronal activity in C . elegans are useful for analysis of neuronal dynamics . Building upon their result here , we used the adjoint space formed by the raw neuronal activity and its derivative ( akin to position and velocity ) . At every time t , the position of the system in the raw neuronal activity space 𝐀t can be mapped to the delay embedded space 𝐃t using the following formula: ( 12 ) ⟨At…At−5τ , At′…At−5τ′⟩→Dt , where 𝐀𝐭 is a snapshot of neuronal activity , 𝐀𝐭′ is a snapshot of the derivative of neuronal activity , ⟨…⟩ denotes concatenation of vectors , and 𝐃𝐭 is the position of the system in the delay embedded coordinates at time t . As discussed in the manuscript , delay embedded neuronal activity of even simple nervous system of C . elegans is too high dimensional to be useful for characterizing system dynamics . For instance for the common 15 neuron dataset 𝐃t is a 180-dimensional vector . Yet , as has been shown by Coifman and Lafon ( 2006 ) , there is a fundamental connection between the eigenvectors of the Markov chain ( Equation 11 ) and dimensionality reduction . This connection is the motivation for a class of methods known as diffusion mapping . The basic idea behind diffusion map is to cast distances between two nearby points in state space as transition probabilities ( Nadler et al . , 2006; Coifman and Lafon , 2006 ) . Diffusion maps have two fundamental advantages: they are nonlinear and preserve local structures . The former is critical here because neuronal dynamics can be safely assumed to be nonlinear . The latter is important because large distances in complex high-dimensional and nonlinear datasets are meaningless ( Aggarwal et al . , 2001 ) . This local geometry assumption is common to a number of nonlinear dimensionality reduction techniques such as isomap , locally linear embedding , and kernel PCA . Traditional applications of diffusion maps have been in dimensionality reduction . For these purposes , the diffusion map is assumed to be symmetric . Here , we modify the formalism slightly to account for the possibility of cyclic fluxes and therefore allow for the possibility of asymmetry in the transition probabilities i→j and j→i . This asymmetry arises naturally if the diffusion map is constructed such that experimentally observed order of neuronal activation is preserved . We accomplish this simply by centering the kernel of a diffusion map kFP on the next empirically observed data point as follows: ( 13 ) kFP ( Dt , Dj ) =exp ( −‖Dt+1−Dj‖222σ2 ) , where 𝐃t is the position of the system in the delay embedded coordinates at time t , 𝐃t+1 is the next empirically observed state of the system , 𝐃j is a point in the local neighborhood of 𝐃t+1 , and ∥⋅∥22 is the Euclidean distance . σ2 is a normalization term that sets the size of the local neighborhood . The key mathematical insight is that after appropriate normalization , diffusion maps converge to the Fokker-Planck ( Nadler et al . , 2006 ) operator . Under these conditions , Equation 9 is an approximation of Equation 7 and thus diffusion maps constitutes a natural way to cast distances between points along a trajectory generated by a stochastic dynamical system . To see this , note that if the local neighborhood is decreased such that it only contains a single point 𝐃t+1 , Equation 13 will exactly reproduce the observed neuronal activity in the correct temporal order . In other words the matrix 𝐌 constructed by applying Equation 13 to all pairs of states will have 1’s for all 𝐌i , i+1 and zeros elsewhere . This matrix , however , is not particularly useful for simulating neuronal dynamics because it will only exactly recapitulate experimental observations . To overcome this limitation , normalization term , σ2 , sets the amount of noise around the experimentally observed neuronal trajectories and allows the simulation to deviate from the actual experimental measurements . Although it is likely that several choices of σ2 will work , here we chose ( 14 ) σ2=σl ( Dt+1 ) σl ( Dt ) ⟨kFP⟩XY , where σl ( ⋅ ) is the standard deviation of the data in a 12 time-step temporal window centered at time t , and ⟨kFP⟩𝐗𝐘 is the mean value of the kernel ( Equation 13 ) over all data points in the neighborhood of 𝐃t+1 . For C . elegans , we compute kFP for the 12 nearest neighbors to each point 𝐃t+1 . The method is robust to the exact number of nearest neighbors used . Equation 13 was then evaluated for all observed states of C . elegans neuronal activity . This results in an n×n ( where n is the number of delay embedded snapshots of neuronal activity ) matrix . This matrix is normalized such that the sum along each row is equal to 1 . This normalization converts the distance matrix to a right stochastic ( Markov ) matrix 𝐌 . The complex eigenvalue with the largest modulus of 𝐌 defines the dominant cyclic flux . The projection of the associated pair of complex conjugate eigenvectors onto elements of 𝐌 define the phase along the cyclic flux θ associated with each delay embedded neuronal activity state . If there are multiple cyclic fluxes as in C . elegans CNS , then in addition to the phase one needs to also know the identify of the flux . To identify fluxes we preform clustering on the data . Any standard clustering algorithm will suffice , and this section will only detail one of many possible choices ( Rubinov and Sporns , 2010 ) that can be used . We did not explore the effects of the choice of clustering and suspect that , as is the case with many clustering applications , the best choice will depend on the specifics of the dataset . We use a maximum modularity algorithm ( Newman , 2006 ) on the transition probability matrix constructed according to Equation 13 . By construction , the transition probability matrix is sparse ( only transitions in local neighborhoods are allowed ) . Therefore , in its raw form the system given by this matrix will not explore the manifold sufficiently as it will be trapped in each individual isolated neighborhood . To overcome this problem , the matrix is exponentiated N times until a minimum fraction of elements of each row are non-zero ( 25% in the C . elegans data ) . Conceptually , this corresponds to finding the evolution of the system after N time steps and is closely related to the ‘diffusion distance’ ( Coifman and Lafon , 2006 ) . Specific choice of N does not have a strong influence on the results , so long as the resultant matrix is not too sparse . Note that the exponentiation of 𝐌 does not change its eigenvalues . Two major features are found in the transition probability matrix ( Figure 4—figure supplement 2 ) : patches and diagonals . Square patches identify locations where the system exhibits Brownian motion near a point attractor . In these patches , the matrix is approximately symmetric and therefore stochastic processes dominate . Diagonal traces identify coherent trajectories where deterministic fluxes are dominant . The square patches are already suitable for clustering . If two elements of the matrix belong to the same point attractor , they will be found in the same square patch . The situation is slightly more complex for coherent trajectories identified by diagonal bands . To determine whether two elements of state space belong to the same coherent trajectory , we compute the maximum correlation of each row ( distances from each element of state space ) and time lagged copies all the other rows max ( corr ( rowi , shift ( rowj , t ) ) ) . Where rowi is the ith row , shift moves all elements in the row t steps to the right and the maximum is taken over all t . This newly formed matrix has the same dimensions as the original transition probability matrix . We apply standard maximum modularity clustering using the community_louvain function from the Brain Connectivity Toolbox to this matrix ( Rubinov and Sporns , 2010 ) . Clustering assigns flux ID α to each point 𝐃t in the delay embedded neuronal activity space . Together with the phase θ , assigned by eigenmode decomposition , ( θ , α ) span the phase space of neuronal dynamics . The phase space spanned by θ and α , rather than raw neuronal activity provide a proper basis with respect to which neuronal activity can be averaged . These averages are shown as manifolds in Figure 4 . For each α , we sort the delay embedded neuronal activity according to its phase θ . We then convolve this activity with a sliding Gaussian window over θ . The width of the Gaussian smooths neuronal activity but does not play any appreciable role in setting the dynamics over a broad range of values . To visualize these θ-averaged trajectories in Figure 4 , we project them onto the first three principal components . Because phase identity α is discrete , θ-averaged trajectories form disjoint bundles . For the purposes of visualization ( Figure 4 ) these bundles are joined together by interpolating a spline ( over both position and direction ) from the end of one bundle to the beginning of the next bundle . This interpolation is performed solely for visualization and plays no role in quantitative analyses – which are all done in the manifold space ( θ , α ) . We make use of a network of two neurons ( Appendix 1—figure 1 ) whose simplified biophysics are modeled by Ermentrout ( 1998 ) ; Beer ( 1995 ) ( 15 ) dAdt=-A+σ ( 8A-6B-0 . 34 ) +ϵ , ( 16 ) 6dBdt=-B+σ ( 16A-2B-2 . 5 ) +ϵ , ( 17 ) σ ( x ) ≡1/ ( 1+exp ( −x ) ) , where the noise term , ϵ , is drawn independently from a Gaussian distribution ϵ∼𝒩 ( 0 , 0 . 1 ) at each time step . A schematic of the system , along with an illustration of the asymptotic behavior of the dynamics are given in Appendix 1—figure 1—figure supplement 1 . Appendix 1—figure 1—figure supplement 1D shows an example trace used in the construction of the manifold in Appendix 1—figure 1 . Here , we analyze Ca2+ imaging data published by Kato et al . ( 2015 ) and Nichols et al . ( 2017 ) . The deviation of fluorescence from baseline ( ΔF/F ) is considered as a proxy for neuronal activity . The manifold was constructed on the data from Kato et al . ( 2015 ) . The validation of the predictions concerning timing of behavioral switching was performed using the dataset from Nichols et al . ( 2017 ) . The dataset were obtained as MATLAB files and were preprocessed by the Zimmer Lab to account for the effects of bleaching . C . elegans were immobilized in a microfluidic device ( Schrödel et al . , 2013 ) under environmentally constant conditions . The 107 to 131 neurons detected in each worm in the Kato et al . ( 2015 ) span all head ganglia , all head motor neurons and most of the sensory neurons and interneurons along with most of the anterior ventral cord motor neurons ( White et al . , 1986 ) . Of the identified neurons for each worm there is a subset of 15 neurons ( AIBL , AIBR , ALA , AVAL , AVAR , AVBL , AVER , RID , RIML , RIMR , RMED , RMEL , RMER , VB01 , VB02 ) which were unambiguously identified in each worm . This set of neurons is used to build the manifold . We adopt the same behavior states defined by Kato et al . ( 2015 ) . The four primary behavioral states are forward locomotion , turns ( FALL ) , reversals ( RISE ) and backwards locomotion ( Figure 1 ) . FALL and RISE were further split into two distinct motor command states by performing k-means clustering on the RISE and FALL phase timing vectors separately . More details of the experiment can be found in Kato et al . ( 2015 ) . All analyses were implemented in MATLAB . In addition to the processing steps by Kato et al . ( 2015 ) which account for bleaching of the GCaMP proteins , we smooth the ( ΔF/F ) time series for each neuron with a Gaussian filter ( σ=1 ) and convert the filtered time series to z-scores . Note that the amount of smoothing applied is orders of magnitude less than the autocorrelations found in the data ( Figure 2—figure supplement 2 ) . Because the experimental data are dominated by forward and backward locomotion , we focus our predictions on just these two behaviors . We do not attempt to predict dorsal or ventral turns or the two types of reversals ( 1 and 2 ) defined by Kato et al . because these behaviors occupy a small fraction of the observed time series . To compare neuronal activity in different instances of the same type of locomotor behavior , we convert from raw time to ‘behavioral phase’ ϕb as follows ϕb= ( ti-tstart ) / ( tend-tstart ) where ti is the raw time , tstart and tend are the beginning and end times of the behavior respectively . This time warping normalizes ϕb such that it ranges from 0 to 1 ( beginning and end ) of each individual instance of behavior . In order to average neuronal activity across different instances of the same behavior , we sample ϕb in equally spaced 100 intervals . Prior to averaging neuronal activity , constant shift in the ΔF/F signal was subtracted ( i . e . the mean of neuronal activity across ϕb for each individual instance of behavior is zero ) . Thus , differences in neuronal activity between two different individuals reflect differences in the temporal pattern of activation rather than shifts in the overall level of activity . Neuronal activity normalized in this fashion and averaged across instances of a particular locomotor behavior in each animal is plotted as a function of ϕb in Figure 1 . Lack of overlap between 95% confidence intervals around the mean neuronal activity observed in different animals in Figure 1 signifies statistically significant differences between neuronal activity in different individual C . elegans . To quantify these differences for each neuron and each type of locomotion , we constructed an n×m matrix T , where n is the number of instances of behavior ( observed in all 5 C . elegans ) and m is the number of ϕb bins . Because neuronal activity is smooth , activation in nearby phases is highly correlated . To remove these correlations , T was subjected to principal component analysis ( PCA ) and projected onto the first principal component ( PC1 ) . This results in n scalars ( one for each neuronal activity trajectory ) . This quantity reflects the similarity between projections onto the first principal component ( mean neuronal activity trajectory across all animals shown by dashed line in Figure 1 ) and each individual cycle of behavior . We subjected this PC1 projection to a one-way ANOVA ( with animal ID as the categorical variable ) . p-values for ANOVA obtained for each combination of locomotor behavior and neuron ID are shown in Figure 1 . For statements concerning statistical significance in the text we used ( α=0 . 05 ) after a Bonferroni correction for multiple comparisons . A subset of neuronal activity was chosen as the training set while the remaining neuronal activity were used as a validation set . A template was constructed by averaging activity of each neuron at the onset of each backing behavior in the training set . This template was convolved with neuronal activity in the validation set to yield similarity score between the template and neuronal activity at each time point in the validation dataset . For the decoding in Figure 1 , we chose a threshold of this score such that the overlap between distribution of scores associated with true events ( initiation of backing behavior ) and distribution of scores of false events ( all other behaviors ) is minimized ( see below ) . To minimize the effect of noise and compensate for the low probability of true positives , we only considered local maxima of the score . To compensate for the inherent imprecision of assigning behavioral states we considered all peaks found within 10 frames ( ≈3s . ) of the initiation of backing behavior as true events . The probability of correctly identifying a behavioral event given a specific threshold Xthres is ( 18 ) p ( θ=1∣X≥Xthres ) =∥Xθ=1≥Xthres∥∥Xθ=1∥+∥Xθ=0≥Xthres∥ , where θ=1 are true events , Xθ=1 are the scores of true events , Xθ=0 are scores of false events , and ∥⋅∥ denotes the number of elements in the set . Optimal threshold X~thresh is found as argmax of Equation 18 with respect to Xthresh in each training dataset individually and used to compute correct decoding probability in Figure 1 . For single animal predictions , 1/2 of the backing behaviors in each animal was used as the training dataset while the remaining 1/2 of backing behaviors in the same animal was used as validation . For the cross animal predictions , we used 1/2 of the backing behaviors in four animals to construct a training set and used the 1/2 of the events in the left out animal as the validation dataset . For the shuffled control , we used random time points as true events in the training dataset . To obtain errors around decoding probability in Figure 1 , we bootstrapped this procedure for multiple partitions of the data into training and validation datasets . Box plot in Figure 1 shows the distribution of the decoding probability across all bootstraps . Manifold space was divided into Gaussian bins each centered at a particular phase θ where Δθ≈0 . 05 . Total of 426 bins were used for the entire data set . The likelihood that a given point 𝐃t belongs to each θ bin was computed and 𝐃t was assigned to the most likely bin . Each point 𝐃t was assigned a behavioral state by Kato et al . ( 2015 ) . Thus , for each θ bin , we attain a distribution of assigned behaviors . This distribution is encoded in color of the manifold ( Figure 4 ) . If θ did not reflect behaviorally relevant information , then the distribution of behavioral states in θ will be the same as in the dataset as a whole . This constitutes the null hypothesis against which manifold-based decoding of behavioral state were tested . Similar approach was taken for the ‘leave one out’ manifold prediction . Manifold was constructed as above on the basis of data from four animals from Kato et al . Distribution of behavioral states for each θ-bin was estimated on the basis of only these four animals . Then data from the fifth animal left out of manifold construction was delay embedded as above . Each snapshot of delay embedded activity of the fifth worm was assigned to the nearest θ-bin . In an attempt to decode the behavioral state of the fifth worm , in each θ-bin we compare the behavioral state of the left out animal to the most likely behavioral state in the θ-bin comprised of data from the remaining four animals . The null hypothesis relative to which quality of decoding was compared is that the prevalence of each behavioral state in a given θ-bin is the same as the prevalence of the behavioral state in the whole dataset . This procedure was repeated by leaving out the data from each one of the five worms in the Kato et al . dataset in turn . The distributions of behavioral states from left out animals and the other four animals used to construct the manifold was compared using χ2 . χ2 averaged over all θ-bins and all five left out animals is reported in the manuscript . As a result of the manifold construction method ( see above ) , each point in the observed neuronal time series 𝐃t is assigned to a single bin in the manifold space ( θi , αi ) . Thus , rather than describing the time series in terms of activation of neurons , we have a 2-D description of the state of the system at each point in time . This allows us to directly estimate transition probability between two states ( θi , αi ) → ( θj , αj ) by Equation 8 . The time evolution of the system can now be readily simulated using Equation 9 . This simulation gives rise to a new time series . To map from manifold space back to neuronal activity or behavior , each point in ( θ , α ) is assigned to a ( delay embedded ) neuronal activity by reversing the relationship in Equation 12 . Recall that each point in manifold space ( θ , α ) corresponds to a cloud of points in the delay embedded space . Here , for the purposes of simulation of neuronal activity we parsimoniously assigned each point in ( θ , α ) the mean of the delay embedded neuronal activity that was assigned to this bin . Alternatively , a random sample from this distribution of points can be chosen . Behavioral state that is associated with this newly simulated neuronal activity snapshot was assigned by sampling the distribution of behavioral states in each phase bin . Simulated dwell time statistics are calculated by assigning to each time point in the simulation a behavior based on the most prevalent behavior in that time point’s corresponding phase bin . This behavioral time sequence is then smoothed by a median filter with a size of 11 time steps ( ∼3 . 5 s . ) . Turns and reverses are transients and constitute a small fraction of the dataset and are thus highly under sampled . Thus , we restrict our analysis to only forward and backwards locomotion . Backing bouts are periods in which the animal sustains backing locomotion with minimal forward locomotion . These events are defined as periods in which the forward locomotion state fails to last for more than 30 frames ( ∼10 s ) . Dwell time distributions are shown in Figure 2 . Experimentally observed and simulated dwell time histograms are smoothed using the ksdensity function in MATLAB . r2 values are calculated from these smoothed histograms . Each data point 𝐃𝐭 is characterized by two independent quantities: time since the onset of the behavior tstart and position in manifold space ( θi , αi ) . The null hypothesis is the expected time to behavioral transition is based solely the dwell time distribution . This corresponds to finding the survival function given by the right tail of the dwell time distribution from tstart to infinity , ( 19 ) Pnull ( t ) =P ( t+tstart ) , where P ( t ) is the probability of the transition occurring at time t is calculated by averaging the time since the onset of the behavior over all points in a given phase bin ( θi , αi ) . To find the corresponding manifold-based prediction the distribution of times until behavioral switch is explicitly found in the simulated neuronal activity data . We identify all points belonging to a particular phase bin and determine the distribution of times until the behavior is terminated . In order to apply the same analysis to the data presented in Nichols et al . ( 2017 ) , we restricted our analysis to the prelethargus N2 animals ( n = 11 ) . These are most genetically similar to the five animals used in the construction of our manifold . First , we selected the subset of neurons that were uniquely identified in each animal from the Nichols et al . ( 2017 ) dataset and the 15 neurons on the basis of which the manifold was constructed using the Kato et al . data . The number of common neurons varied between 8 and 13 . Neuronal activity from Nichols et al . was delay embedded as above yielding a set of 𝐃ts . Each 𝐃t from the validation dataset was assigned to the closest phase bin as in the ‘leave one out’ validation . The only exception here is that the distance to the closest phase bin was computed by omitting the neurons that were missing from the animal in the validation dataset . The distribution of times to behavioral transition in the validation dataset was empirically estimated by observing the switching times of all points in the validation dataset assigned to a given phase bin ( θi , αi ) . The Nichols et al . ( 2017 ) animals use a different behavioral assignment paradigm than those in Kato et al . ( 2015 ) and so we normalized the behavioral assignments by assigning forward locomotion to any point in which the z-score of AVAL was below a given threshold and backward locomotion to any point in which the z-score of AVAL was above that threshold . This method does not preserve finer details such as the timing of turns and reversals , and so our predictions do not attempt to address those behaviors . Starting from the motor command dwell time distributions as shown in Figure 2C we calculate the Kullback–Leibler divergence between the experimentally observed distribution and the simulated distribution . Because the exact binning heavily effects information theoretic quantities such as KL divergence , we scan over a range of bin counts between 40 and 200 and choose the minimum KL divergence in this range . Finally , to normalize these quantities for easy comparison we calculate the ratio between the original total information and the modified model ( different numbers of neurons or different parameters ) by: ( 20 ) Itot=∑b∈ℬDKL ( Pobs||Psim ) , ( 21 ) Irel=Itotoriginal/Ttotmodified , where DKL is the KL divergence , ℬ is the set of three motor command distributions in Figure 2C ( forward locomotion , backwards locomotion and backwards bouts ) and Pobs and Psim are the observed and simulated dwell time distributions respectively . For the robustness tests in Figure 2—figure supplement 4 and Figure 3—figure supplement 1 , the models are built with data from AVA excluded . For the calculation of motor command dwell time distributions , the behaviors are assigned using the same behavioral assignment given by Kato et al . ( 2015 ) . For the predictions of behavioral switches presented in Figure 3—figure supplement 1 , the mapping from neuronal activity space to manifold space does not make use of the data from AVA . However , we can include AVA in the mapping from manifold space back to neuronal activity space to recover the expected activity of AVA even though AVA was not explicitly used at any point in the model construction . | How can we go about trying to understand an object as complex as the brain ? The traditional approach is to begin by studying its component parts , cells called neurons . Once we understand how individual neurons work , we can use computers to simulate the activity of networks of neurons . The result is a computer model of the brain . By comparing this model to data from real brains , we can try to make the model as similar to a real brain as possible . But whose brain should we try to reproduce ? The roundworm C . elegans , for example , has just 302 neurons in total . Advances in brain imaging mean it is now possible to identify each of these neurons and compare its activity across worms . But doing so reveals that the activity of any given neuron varies greatly between individuals . This is true even among genetically identical worms performing the same behavior . Researchers trying to model the roundworm brain have attempted to model the average activity of each neuron across many worms . They hoped they could use these averages to predict the behavior of other worms from their neuronal activity . But this approach did not to work . Even in roundworms , the coordinated activity of many neurons is required to generate even simple behaviors . Averaging the activity of neurons across worms thus scrambles the information that encodes each behavior . Brennan and Proekt have now overcome this problem by developing a more abstract model that treats the nervous system as a whole . The model takes into account changes in the activity of neurons , and in the worms’ behavior , over time . A model of this type built using one set of worms can predict the behavior of another set of worms . This approach may work because in evolution natural selection acts at the level of behaviors , and not at the level of individual neurons . The activity of individual neurons can thus vary between animals , even when those neurons encode the same behavior . This means it may also be possible to model the human brain without knowing the activity of each of its billions of neurons . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2019 | A quantitative model of conserved macroscopic dynamics predicts future motor commands |
Aging , obesity , hypertension , and physical inactivity are major risk factors for endothelial dysfunction and cardiovascular disease ( CVD ) . We applied fluorescence-activated cell sorting ( FACS ) , RNA sequencing , and bioinformatic methods to investigate the common effects of CVD risk factors in mouse cardiac endothelial cells ( ECs ) . Aging , obesity , and pressure overload all upregulated pathways related to TGF-β signaling and mesenchymal gene expression , inflammation , vascular permeability , oxidative stress , collagen synthesis , and cellular senescence , whereas exercise training attenuated most of the same pathways . We identified collagen chaperone Serpinh1 ( also called as Hsp47 ) to be significantly increased by aging and obesity and repressed by exercise training . Mechanistic studies demonstrated that increased SERPINH1 in human ECs induced mesenchymal properties , while its silencing inhibited collagen deposition . Our data demonstrate that CVD risk factors significantly remodel the transcriptomic landscape of cardiac ECs inducing inflammatory , senescence , and mesenchymal features . SERPINH1 was identified as a potential therapeutic target in ECs .
According to WHO , cardiovascular diseases ( CVDs ) account for 10% of the global disease burden and constitute the number 1 cause of death in the western world . CVDs are mainly caused by behavioral ( physical inactivity , unhealthy diet ) and metabolic ( obesity , hypertension , diabetes , high cholesterol ) risk factors ( Mendis et al . , 2011 ) . Aging , however , is by far the biggest contributor to CVD , and aging population is becoming an enormous challenge worldwide . The heart contains a dense vascular network , and endothelial cells ( ECs ) are indeed the most abundant cell population in the adult mouse heart ( Pinto et al . , 2016 ) . In addition to their transport function , ECs are defined to control vasomotor tone , maintain vascular homeostasis , regulate angiogenesis , and establish bidirectional communication with other cell types and organs via paracrine signaling mechanisms ( Aird , 2007; Aird , 2012; Kivelä et al . , 2019; Talman and Kivelä , 2018; Hemanthakumar and Kivelä , 2020 ) . ECs are found to be highly adaptive to physiological stimuli during normal growth and development ( White et al . , 1998; Bloor , 2005 ) , and the diversity of ECs in different tissues has now been acknowledged . ECs are also maladaptive to a spectrum of pathological events involving , for example , inflammation or oxidative stress ( Cines et al . , 1998; Gimbrone and García-Cardeña , 2016 ) , and the development of heart diseases is strongly linked to endothelial dysfunction and impaired vascular remodeling . However , the molecular cues , which cause maladaptation and dysfunction of ECs in the heart in response to pathological signals , remain elusive . Physical inactivity increases the incidence of several chronic diseases , whereas regular exercise training has positive effects on most of our tissues ( Hawley et al . , 2014 ) . Because microcirculation is present in every organ in the body , ECs have a unique ability to influence the homeostasis and function of different tissues , and they are potentially a major cell type mediating the positive effects of exercise throughout the body . Although the cardiac benefits of exercise are clear and there have been major advances in unraveling the molecular mechanisms , the understanding of how the molecular effects are linked to health benefits is still lacking ( Hawley et al . , 2014 ) . Especially , the effects of exercise on ECs have not been characterized . We hypothesized that the major CVD risk factors aging , obesity , and pressure overload will induce adverse remodeling of cardiac EC transcriptome ( Gimbrone and García-Cardeña , 2016; Ungvari et al . , 2018; Brandes , 2014 ) , whereas exercise training would provide beneficial effects ( White et al . , 1998; Bloor , 2005 ) . Both physiological and pathological stimuli significantly modified the cardiac EC transcriptome . Intriguingly , our results demonstrated that CVD risk factors promoted activation of transforming growth factor-β ( TGF-β ) signaling , inflammatory response , cellular senescence , and induced mesenchymal gene expression in cardiac EC , whereas exercise training promoted opposite protective effects .
To mimic the effect of the most common CVD risk factors ( aging , obesity , pressure overload/hypertension , and physical inactivity ) , we used adult C57BL/6J wild-type mice in the following experimental groups: aged ( 18 months ) vs . young ( 2 months ) mice , high-fat diet ( HFD ) induced obesity ( 14 weeks HFD ) vs . lean mice , transverse aortic constriction ( TAC ) vs . sham-operated mice , and exercise training ( progressive treadmill running for 6 weeks ) vs . sedentary mice ( Figure 1—figure supplement 1A , B ) . Exercise trained mice showed improved ejection fraction compared to the sedentary mice , whereas aging , HFD , and TAC resulted in impaired heart function ( Figure 1—figure supplement 1C–F and Figure 1—source data 2 ) . HFD also induced marked weight gain , increased fat mass , and impaired glucose tolerance ( Figure 1—figure supplement 1G–I ) . Left ventricular ( LV ) mass was increased in aged , HFD-treated , and TAC mice ( Figure 1—source data 2 ) . Exercise training also slightly increased LV mass , which reflects mild physiological hypertrophy often observed in endurance-trained athletes ( Arbab-Zadeh et al . , 2014; Figure 1—source data 2 ) . Exercise training significantly increased , whereas aging , HFD , and TAC decreased the percentage , count , and mean fluorescence intensity of the cardiac ECs ( CD31+CD140a-CD45-Ter119-DAPI- ) compared to the controls , when analyzed by fluorescence-activated cell sorting ( FACS; Figure 1A , B , Figure 1—figure supplement 2A–D ) . This was also demonstrated by immunohistochemistry for CD31-positive coronary vessels ( Figure 1C , D ) . The cardiac ECs were gated and sorted by FACS ( Figure 1—figure supplement 3A ) , and the isolated ECs were first analyzed by quantitative PCR analysis , which indicated significant enrichment of EC markers Cdh5 and Tie1 in the sorted fraction compared to whole heart or other cardiac mononuclear cells ( Figure 1—figure supplement 3B ) . In addition , isolation resulted in 87 . 4 ± 1 . 9% cell viability and RNA purification strategy yielded intact and stable RNA with average RNA integrity number ( RIN ) of 8 . 7 ( Figure 1—figure supplement 3C , D ) . RNA sequencing of isolated ECs was used to profile the expression pattern of cardiac EC transcripts in different experimental groups . Two-dimensional PCA of the EC transcriptomes exhibited significant proportion of variance in the gene expression pattern , which can be attributed to the treatment-induced changes in cardiac EC transcriptome ( Figure 2—figure supplement 1A–E ) . Notably , unsupervised hierarchical clustering of EC data sets for all experimental interventions ( sedentary , exercise trained , young , aged , sham , TAC ) revealed consistent clustering and high degree of similarity in the gene expression pattern ( Figure 2—figure supplement 1F–J ) . The analysis for differentially expressed genes ( DEGs ) showed a large number of up- and downregulated genes especially in aged , obese , and TAC-operated mice followed by a smaller number of affected genes in exercise trained mice . The number of significantly up- and downregulated genes with the false discovery rate ( FDR ) 0 . 05 for each treatment are shown in the MA plots and the top 50 DEGs for each treatment are presented by heat maps ( Figure 2A–E , F–J ) . To understand the biological functions of the DEGs , we used PANTHER classification analysis ( Figure 3A ) . The analysis revealed that genes related to EC development , adherence junction organization , IGFR signaling , adrenomedullin receptor signaling , and mitochondria were upregulated by exercise training . Furthermore , exercise training downregulated pathways related to cellular aging , vascular membrane permeability , negative regulation of angiogenesis , TGF-β1 production , collagen activated tyrosine kinase signaling , and ossification . In contrast , pathways related to TGF-β , IFNα , TNFα , oxidative stress , EC differentiation , vascular permeability , cell aging , collagen synthesis , SMAD signaling , and mesenchymal cell development were highly enriched in cardiac EC from both aged and obese mice . Downregulated pathways in these mice included tissue and lipid homeostasis , ECM assembly , tube morphogenesis , cell adhesion , cell number maintenance , EC proliferation , vasculature development , artery development , and NOTCH signaling . Pressure overload activated pathways such as cellular response to TGF-βR2 activation of fibrotic pathways , inactivation of cell survival pathways Erk1/2 and MAPK , and ossification process , whereas cellular homeostasis and vasculature development were repressed . Comparison of the GO biological terms , which were significantly affected by exercise training and the CVD risk factors , demonstrated clear opposite effects on the EC transcriptome . Aging and HFD promoted oxidative stress response , activation of inflammatory and fibrosis pathways ( Figure 3—figure supplement 1A–E ) and cellular aging , and inhibited pathways regulating cell number maintenance , proliferation , and lipid homeostasis . Exercise training , in turn , promoted EC homeostasis and vascular growth , and prevented vascular aging , inflammation , and pathological activation . In the cardiac ECs of HFD and TAC-treated mice , a significant upregulation of senescence-associated secretory phenotype ( SASP ) genes ( Figure 3—figure supplement 2C–E ) were observed . To validate the link between obesity and senescence in the heart , we performed SA-β-galactosidase staining in HFD- and chow-fed mouse heart sections . In HFD-fed mice , we observed several clusters of SA-β-galactosidase positive cells in the heart , which were not observed in the chow-fed animals ( Figure 3—figure supplement 2A ) . Quantification showed significant increase in these cells ( Figure 3—figure supplement 2B ) . Further studies are needed to identify these cells and their role in CVD development . Because the analyses indicated upregulation of genes and pathways associated with mesenchymal development and endothelial-to-mesenchymal transition ( EndMT ) by all of the CVD risk factors , we reviewed our DEG sets for the expression of selected endothelial and mesenchymal markers based on the previously published data sets ( Figure 3—source data 1 ) . We found significant upregulation of many mesenchymal markers and downregulation of EC genes in aged and obese mice ( Figure 3C , D ) . After 2 weeks of TAC , we also observed upregulation of several mesenchymal markers , whereas after 7 weeks of TAC , there was both up- and downregulation of the EC and mesenchymal markers , indicating possible reversal of the process ( Figure 3E , F ) . Strikingly , exercise training downregulated several EndMT genes ( Fscn1 , Cd93 , Vwa1 , Sparc , Tuba1a , Cd44 , Trp53 , Col4a2 , Mest , Cnn2 , Tnfaip1 , Lamb1 , Ltbp4 , and Unc5b ) , the angiogenesis inhibitor gene Vash1 , and the endothelial activation marker Apln and its receptor Aplnr , whereas it upregulated the expression of Malat1 , Mgp , Krit1 , and Calcrl ( Figure 3B ) . We validated the results using an expanded set of samples by qPCR for Apln , Vim , Tgfbr2 , Vash1 , Sparc , and Tgfb1 ( Figure 3—figure supplement 3A–F ) . To identify genes , which could mediate the negative effect of aging and obesity and the protective effects of exercise , we performed gene overlap analysis of DEGs from these three experimental interventions . We found four genes significantly affected by all treatments , of which two genes ( Serpinh1 and Vwa1 ) were upregulated by aging and HFD and downregulated by exercise training . The other two genes ( Mest and Fhl3 ) were upregulated by HFD and downregulated by exercise training and aging ( Figure 4A–C ) . We performed an in silico secretome analysis to characterize the properties of the identified genes using MetaSecKB database ( Figure 4D ) . Both Serpinh1 and Vwa1 contain a signal peptide for secretion , indicating they could act as angiocrines in autocrine and/or paracrine fashion . We focused on Serpinh1 , as it has a known role as a collagen chaperone and has been linked to fibrosis ( Ito and Nagata , 2019 ) , making it an attractive candidate . We validated the endothelial Serpinh1 expression by qPCR ( Figure 4C ) , and at single cell level using Tabula Muris database ( Tabula Muris Consortium et al . , 2018 ) and cardiac EC atlas from the Carmeliet lab ( Kalucka et al . , 2020 ) . The scRNAseq analysis revealed that Serpinh1 is expressed in variety of cell types within the mouse heart , including fibroblasts , myofibroblasts , smooth muscle cells , ECs , endocardial cells , and to lesser extent in cardiomyocytes ( Figure 4—figure supplement 1A–D ) . In ECs , Serpinh1 was found to be expressed throughout all EC clusters , with the highest expression in the apelin-high cluster marking activated ECs ( Figure 4—figure supplement 2A–F ) . Interestingly , the expression of mesenchymal markers such as Tagln2 , Vim , and Smtn was also high in this cluster . Next , we analyzed the expression of SERPINH1 ( also called as HSP47 ) in healthy human heart and in human cardiac ECs . Immunohistochemistry demonstrated SERPINH1 to be highly expressed throughout the coronary vasculature and in fibroblasts in human heart , and weak staining was also detected in cardiomyocytes ( Figure 4E–G , Figure 4—figure supplement 1D ) . Analysis using the EndoDB database ( E-GEOD-43475 ) showed that the expression of SERPINH1 is highly similar in both veins and arteries and in different tissues ( heart , lungs , liver , human cardiac arterial EC ( HCAEC ) , and human umbilical venous EC [HUVEC] ) ( Figure 4—figure supplement 3A ) . In human cardiac ECs , SERPINH1 was localized perinuclearly , similar to what has been demonstrated in other cells types , and consistent with the ER retention motif in its N-terminus ( Figure 4E; Masuda et al . , 1994; Razzaque et al . , 1998; Honzawa et al . , 2014 ) . We next tested , if exercise training can attenuate the expression of Serpinh1 , Vwa1 , and selected markers of TGF-β signaling/EndMT also in aged mice . Of the studied genes , mRNA expression of Serpinh1 and Vim were significantly repressed by exercise training , and there was a tendency also for Vwa1 ( Figure 4H–K ) . To study the effects of SERPINH1 in human ECs , we produced lentiviral vector encoding myc-tagged hSERPINH1 . Both HUVECs and HCAECs were analyzed . SERPINH1 protein was localized similar to the native protein ( Figure 5B ) , and the expression was verified by western blotting ( Figure 5—figure supplement 1A ) . Overexpression of SERPINH1 altered the cellular morphology characterized by impaired or discontinuous vascular endothelial cadherin junctions , increased stress fiber formation , and larger cell size ( Figure 5A , B ) . Furthermore , analysis of EC and mesenchymal cell-related transcripts demonstrated significant repression of EC markers ( CD31 , CDH5 , TIE1 , NRARP , and ID1 ) and induction of a proliferation gene CCND1 , and mesenchymal/EndMT markers ( TAGLN , aSMA , CD44 , VIM , NOTCH3 , ZEB2 , SLUG , FN1 , VCAM1 , and ICAM1 ) ( Figure 5C ) . VE-cadherin downregulation was also confirmed at protein level ( Figure 5D ) and increased aSMA expression by immunofluorescence staining ( Figure 5E ) . We also analyzed the effect of SERPINH1 on cellular senescence . SA-β-galactosidase staining showed increased number of cells undergoing senescence and there was a clear upregulation of senescence-associated genes ( Figure 5G , H ) . Transcriptomic changes pointed toward activated TGF-β signaling and oxidative stress in response to all of the CVD risk factors . Both are known to contribute to EC dysfunction and EndMT , and thus we tested if they act as upstream regulators of SERPINH1 . Indeed , our results show that TGF-β1-treatment of HCAECs significantly upregulated the expression of SERPINH1 together with other known EndMT markers , and there was also small but significant induction of SERPINH1 by hydrogen peroxide treatment ( Figure 5F ) . To investigate the significance of SERPINH1 depletion in human cardiac ECs , HCAECs were transduced with four independent shSERPINH1 lentiviral constructs . The constructs induced approximately 80% deletion of SERPINH1 mRNA ( Figure 6D ) . The cell morphology was not affected after 2 days ( Figure 6A ) , but 10 days of silencing significantly changed EC morphology and decreased the cell density in culture ( Figure 6B ) , suggesting that SERPINH1 might play a role in EC homeostasis and survival . SERPINH1 silencing significantly inhibited collagen fibril deposition , detected by immunohistochemistry for type 1 collagen ( Figure 6B , C ) . Only the cells transduced with the construct #1 could produce some extracellular collagen 1 , and these cells also survived better than the cells transduced with constructs #2 , #3 , or #4 ( Figure 6B , C ) . Next , we treated the cells with TGF-β1 and hydrogen peroxide for 5 days to induce EndMT features , as described previously ( Evrard et al . , 2016; Magenta et al . , 2011 ) . We used the shSERPINH1 ( #1 ) construct , because from the other silencing constructs not enough cells survived for the experiments . The results indicated that silencing of SERPINH1 prevented the appearance of Taglin-positive cells , a commonly used readout for EndMT , which were observed in the control cells ( Figure 6E ) . We also studied the effect of SERPINH1 on cell proliferation/migration . In the scratch wound assay , overexpression of SERPINH1 significantly promoted wound closure ( Figure 7A , B ) , whereas silencing of SERPINH1 for 2 days significantly decreased EC proliferation/migration ( Figure 7C , D ) . Cell proliferation was slightly increased by SERPINH1 overexpression , whereas silencing almost completely blocked proliferation , as determined by EdU incorporation ( Figure 7E–G ) .
Here we have used transcriptomic profiling to decipher how the major CVD risk factors aging , obesity , and pressure overload remodel cardiac ECs , and how the protective effects of exercise are mediated . The results demonstrate that the CVD risk factors activate transcriptional programs promoting cell aging , senescence , TGF-β activation , inflammation , and oxidative stress in cardiac ECs . Importantly , exercise attenuated these same pathways , even in healthy mice . Furthermore , we found that aging , obesity , and pressure overload induced mesenchymal gene programs in cardiac ECs , which can contribute to dysfunctional endothelium and CVD development . Analysis of potential disease-promoting genes identified Serpinh1 to be induced by aging and obesity , while its expression was significantly repressed by exercise , also in old mice . Mechanistically , SERPINH1 was induced by TGF-β and ROS , and the overexpression of SERPINH1 increased cell size and stress fiber formation , weakened cell–cell junctions , and promoted mesenchymal and senescence-associated gene expression in human cardiac ECs . Immunohistochemistry of human hearts showed that SERPINH1 is abundantly expressed throughout the cardiac vasculature . The largest dysregulation of the cardiac EC transcriptome was found in aged mice , followed by obesity and pressure overload . Exercise training affected a smaller number of transcripts , which can be accounted , at least partly , to the young and healthy control mice , which could move unrestrictedly in their home cages . Interestingly , however , most of the pathways activated by CVD risk factors were the same that were repressed by exercise training , highlighting the potential of physical activity to improve cardiovascular health via modulating EC phenotype and function . The positive effects of exercise on skeletal muscle and cardiac angiogenesis have been described previously ( Hudlicka et al . , 1992 ) , but exercise-induced molecular changes in ECs have not been characterized . It is important to note that here we studied the chronic adaptations to exercise training , and not the acute responses , which likely explains why more genes were found to be downregulated than upregulated in these mice . The effects of exercise training in cardiac ECs were associated with EC homeostasis and stabilization , with increased expression of genes related to establishment of EC barrier , polarity , and focal adhesion . Importantly , exercise induced repression of inflammatory , permeability , senescence , and mesenchymal gene networks . It also attenuated the expression of apelin , which is considered as a marker of activated ECs , and also its receptor Aplnr . This suggests that regular exercise training promotes stabilization and quiescence in cardiac ECs and prevents cellular aging . Aging and obesity , on the other hand , are known to contribute to capillary rarefaction and/or dysfunction ( Cines et al . , 1998; Gimbrone and García-Cardeña , 2016; López-Otín et al . , 2013; Ungvari et al . , 2010 ) , and another novel aspect in this study was the comparison of several CVD risk factors to identify common pathways and genes , which could drive the pathogenesis in cardiac disease , and could be considered as potential therapeutic targets . ECs would provide an attractive target for drug development , as they are the first cells to encounter drugs in the bloodstream . Dysfunctional endothelium likely contributes to more diseases than any other tissue in the body as it affects all organs . On the other hand , endothelium could act as an important mediator of the health-promoting effects of exercise in a variety of tissues . Our finding that aging , obesity , and pressure overload induce mesenchymal gene programs in cardiac ECs adds to the increasing evidence that activated endothelial TGF-β signaling and acquisition of mesenchymal features play an important role in the development of EC dysfunction and cardiac diseases ( Kovacic et al . , 2019; Chen et al . , 2015; Chen et al . , 2019; Xiong et al . , 2018 ) . Importantly , genes related to TGF-β production and cellular aging were repressed by exercise , highlighting the potential of exercise training in preventing and delaying the development of CVD . The molecular mechanisms of exercise-mediated repression of TGF-β signaling are not known . Nitric oxide ( NO ) has been previously shown to attenuate TGF-β/SMAD2 signaling in ECs , whereas mice lacking endothelial NO synthase activity presented increased TGF-β signaling and collagen 1 in their aortas ( Saura et al . , 2005 ) . Increased blood flow during exercise induces eNOS expression and NO production , which could repress TGF-β activity in the vasculature . Reduced TGF-β activation was also recently reported in whole heart lysates in exercised rats ( Lin et al . , 2020 ) , and enhanced TGF-β signaling was also suggested to be a negative regulator of exercise response in human skeletal muscle ( Böhm et al . , 2016 ) . The activation of TGF-β signaling pathway has been implicated as a driving force for EndMT ( Evrard et al . , 2016; Cooley et al . , 2014; James and Rafii , 2014; Bischoff , 2019; Bischoff et al . , 2016 ) . Several studies have recently suggested that EndMT could contribute to the development of various CVDs ( Zeisberg et al . , 2007; Kovacic et al . , 2019; Li et al . , 2018; Sánchez-Duffhues et al . , 2018 ) , but currently there is a lack of understanding of the causal relationships and mechanisms linking EndMT and CVD ( Kovacic et al . , 2019 ) . Furthermore , whether the transition from ECs to mesenchymal cells occurs completely in various CVDs is still actively debated in the literature . It has been suggested that pathological EC activation will result in acquired EndMT features for example expression of mesenchymal genes , without full transformation from one cell type to another ( Chen et al . , 2020 ) . This is in line with our findings , as only cells with high CD31 expression and with no expression of CD45 , CD140a , and Ter119 were included in our analyses . Thus , all the analyzed cells were ECs , but in the CVD risk factor groups they demonstrated increased mesenchymal marker expression . Long-term lineage tracing of ECs in response to CVD risk factors would provide further knowledge if and to what extent full transformation of ECs to mesenchymal cells occurs in cardiac vasculature . Our results , however , demonstrate that ECs acquire mesenchymal features due to CVD risk factors , which likely results in EC dysfunction even without full EndMT . It was not surprising that SASP-genes were induced in old mouse ECs , but the observation that this was also seen in the obese and pressure overloaded mice caught our attention . In ECs of obese mice , increased p53 expression has been reported , which led to reduced eNOS phosphorylation both in vitro and in vivo ( Yokoyama et al . , 2014 ) . Combined with our data , this could then result in increased TGF-β activity ( Saura et al . , 2005 ) , linking senescence and TGF-β signaling . We also observed significantly more SA-β-galactosidase positive cells in the hearts of obese mice compared to the lean mice . These cells were often found in clusters , and in addition to ECs , they could also be other non-myocytes . Thus the significance of these cells to cardiac vasculature remains to be further studied . Endothelial deletion of p53 has also been demonstrated to protect against pressure overload-induced cardiac dysfunction and fibrosis , suggesting that increased p53 and other senescence-associated genes are important mediators of EC dysfunction ( Gogiraju et al . , 2015 ) . To identify possible pathology-driving genes , which would be common for several risk factors , we performed gene overlap analysis using all data sets . Two genes , Serpinh1 and Vwa1 , were found to be significantly increased by both aging and obesity and decreased by exercise , suggesting that they could act as common mediators of EC dysfunction . We focused in this study on Serpinh1 , as it is a collagen chaperone and has been shown to contribute to tissue fibrosis ( Ito and Nagata , 2019; Khalil et al . , 2019 ) , an important feature of many cardiac diseases . Recently , it was demonstrated using Serpinh1 cell type-specific knockout mice that Serpinh1/Hsp47 in myofibroblasts is an important regulator of pathologic cardiac fibrosis ( Khalil et al . , 2019 ) . In line with our results , collagen 1 production was decreased also in the EC-specific Serpinh1 deficient hearts in TAC-operated mice ( Khalil et al . , 2019 ) . In human cardiac ECs , our results placed SERPINH1 downstream of TGF-β and ROS , and demonstrated that its overexpression promoted mesenchymal features and senescence . Furthermore , SERPINH1 was found to be important for extracellular collagen 1 deposition and EC proliferation and migration . Overexpression of SERPINH1 slightly increased proliferation , but the effect was more pronounced on migration , whereas silencing of SERPINH1 almost completely blocked proliferation . Silencing also prevented the TGF-β induced appearance of TAGLIN-positive cells in human cardiac EC , which is considered as a marker for EndMT ( Evrard et al . , 2016; Magenta et al . , 2011 ) . It is counterintuitive that SERPINH1 increased senescence markers that inhibit cell proliferation , but at the same time increased proliferation and migration . Proliferation was evaluated 48 hr after transduction , thus it is possible that at this time point , the induced mesenchymal properties override the senescence signals , which might take over at later time points , reflected as increased cell size and SA-beta galactosidase staining . Based on the publicly available single-cell RNA sequencing data and immunohistochemistry of the human heart samples , SERPINH1 is abundantly expressed in all cardiac endothelial populations , as well as in arterial and venous ECs in other tissues . Not much is known about the role of SERPINH1 in heart disease . In a study by Kato et al . , SERPINH1 was found to colocalize with several other EndMT markers in some of the ECs in left atrium in patients with atrial fibrillation , a disease which is often related to fibrosis ( Kato et al . , 2017 ) . To further the translational potential , the role of endothelial SERPINH1 in aged , obese , and hypertensive human hearts needs to be determined . In the in vivo experiments for endothelial RNA sequencing analyses , we have used male mice , which were age and gender matched in exercise , obesity , and TAC models . Thus , most of the results represent responses in male mice , whereas in the exercise training experiment in old mice , female mice were used . Some of the responses , for example the repression of Serpinh1 in exercised mice , were similar to those observed in young male mice; however , some of the changes were not significant in old female mice . It will be important to determine if the findings presented here in male mice are also valid in females , and even more interestingly , in humans . In conclusion , our data demonstrate that the major CVD risk factors significantly remodel the cardiac EC transcriptome promoting cell senescence , oxidative stress , TGF-β signaling , and mesenchymal gene features , whereas exercise training provided opposite and protective effects ( Figure 8 ) . SERPINH1 was identified as one of the downstream effectors of TGF-β , which could provide a novel therapeutic target in ECs .
All animal experiments were approved by the committee appointed by the District of Southern Finland . Male C57BL/6J wild-type mice were purchased from Janvier Labs and used in the following experimental set-ups: physical activity ( progressive exercise training vs sedentary ) , obesity ( high-fat fed for 14 weeks vs chow ) , aging ( 18 months vs 2 months ) , and pressure overload/heart failure ( transaortic constriction for 2 and 7 weeks vs sham ) . Female C57BL/6J wild-type mice of 19–24 months old were used for a separate exercise training experiment . The mice were housed in individually ventilated cages and acclimatized at least for 1 week in the animal facility before any experiments . The cohort size ( n ) for each experiment is indicated in the figures or figure legends . Ten-week-old C57BL/6J male mice ( used for RNAseq ) or 19–24 months old female mice ( used for qPCR analyses ) were trained on a treadmill ( LE 8710 , Bioseb ) . The mice were familiarized to the treadmill for three consecutive days with low speed ( 8–10 cm/s ) . Progressive training program consisted of 1–1 . 5 hr training bouts 5 days a week for a total of 6 weeks with increasing speed , inclination , and/or duration each week . The following parameters in the treadmill controller were opted , tread inclination: 0°−10°; minimum and maximum tread speed: 10–30 cm/s; shock grid intensity: 0 . 2 mA . The aged female mice were exercise-trained for 4 weeks and the same procedures were followed during the training program . Ten-week-old C57BL/6J male mice were fed with standard chow diet or HFD containing 60% kcal derived from fat ( Research Diets , D12492 ) for 4 or 14 weeks and used for immunohistochemistry or RNA-seq analysis , respectively . Ten-week-old C57BL/6J male mice were anesthetized with ketamine and xylazine . The mice were placed in supine position and intubated . The skin along the supra-sternal notch to mid sternum was incised to perform sternotomy to expose the aortic arch , right innominate , and left common carotid arteries together with the trachea . Ligation of the transverse aorta between the right innominate left common carotid arteries against blunted 27-gauge needle with a 7–0 suture was performed and the needle was gently removed . The sternum and skin were ligated with monofilament polypropylene suture . Mice were placed in a warm chamber to recover , treated with analgesics ( 0 . 05 mg/kg of Temgesic i . m . ) at the time of the surgery and twice a day for following 2 days . For the control group ( sham ) , all the steps in the surgical procedure were followed , except constricting the aorta . One group was killed 2 weeks and another group 7 weeks after the surgery . Echocardiography was performed once a week during the experiment . To analyze cardiac function and ventricle dimensions , two-dimensional echocardiography images were acquired ( Vevo 2100 Ultrasound , FUJIFILM Visual Sonics ) . The LV internal diameter , LV posterior wall thickness , and interventricular septum thickness at end-systole and end-diastole were measured in M-mode along the parasternal short axis view and analyzed by Simpson’s modified method ( Kivelä et al . , 2019 ) . The mice were anesthetized with ketamine and xylazine and the percentage of total body fat was measured using dual energy X-ray absorptiometry ( Lunar PIXImus , GE Medical systems ) . Mice were fasted for 4–5 hr before the experiment . Glucose ( 1 g/kg ) was administered by oral gavage to mice . Blood from the tail tip was used to measure glucose levels at the following time points ( 15 , 30 , 60 , and 90 min ) using blood glucose meter ( Contour , Bayer ) . Frozen mouse heart sections ( 10 μm ) were cut with cryomicrotome and stained as described previously ( Kivelä et al . , 2019 ) . The primary antibodies are listed in the Key Resource Table . Primary antibodies were detected with Alexa 488 , 594 or 647-conjugated secondary antibodies ( Molecular Probes , Invitrogen ) . The sections were mounted with Vectashield Hard Set mounting media with DAPI ( Vector Laboratories ) . The images were acquired with 20× , 40× air or 40× oil immersion objectives using AxioImager epifluorescent microscope ( Carl Zeiss ) . The stained micrographs were initially adjusted for threshold , and an area fraction tool was used to quantify the area percentage of the vessels and collagen ( Image J software , NIH ) . Human heart samples were obtained from four organ donor hearts , which could not be used for transplantation for example due to size or tissue-type mismatch . The collection was approved by institutional ethics committee and The National Authority for Medicolegal Affairs . The human paraffin heart sections ( 4 μm ) were cut , deparaffinized , and rehydrated with xylene , descending concentration series of ethanol ( 99% , 95% , 70% , and 50% ) and H2O , and incubated in high pH antigen retrieval buffer containing 10 mM Tris , 1 mM EDTA , 0 . 05% Tween 20 ( pH 9 . 0 ) . For HSP47 immunohistochemical analysis , VECTASTAIN Elite ABC kit ( PK-6100 ) and DAB substrate were used to label and amplify the antibody signal . The 20× or 63× images were acquired with light microscope ( Leica ) . For immunofluorescent staining , after the antigen retrieval step the sections were blocked with donkey immunomix ( 5% normal donkey serum , 0 . 2% BSA , 0 . 3% Triton X-100 in PBS ) , incubated overnight at 4°C with the primary antibodies for HSP47 and VE-Cadherin ( CDH5 ) and detected with Alexa 488 and 594 conjugated secondary antibodies ( Molecular probes , Invitrogen ) . The sections were mounted with Vectashield hardset with DAPI ( Vector labs ) and 40× images were acquired using AxioImager epifluorescent microscope ( Carl Zeiss ) . The harvested hearts were briefly rinsed in ice-cold Dulbecco’s phosphate-buffered saline ( DPBS , #14190–094 , Gibco ) supplemented with 0 . 3 mM calcium chloride ( CaCl2 ) , cut opened longitudinally into two halves to expose the cardiac chambers and minced longitudinally and transversely into small pieces . To enzymatically dissociate the heart , 4 ml of pre-warmed digestion media ( 1 mg/ml ) of each collagenase types ( type I [#17100–017] , type II [#17101–015] , and type IV [#17104–019] ) from Gibco were dissolved in DPBS containing 0 . 3 mM CaCl2 and added to the minced hearts , and incubated in water bath at 37°C for 25 min . During the digestion process , the samples were very gently mixed by vortexing for every 5 min . After incubation , the cell suspension was gently passed through T10 serological pipette 20 times . To neutralize the digestion , 10 ml of rinsing media ( Dulbecco’s modified eagle medium [#31053–028] supplemented with 10% heat inactivated FCS ) was added to the cell suspension and filtered through the 70 μm nylon cell strainer ( Corning , #352350 ) . Throughout the isolation process the cell suspensions were centrifuged for 5 min , 300 g , and 4°C between each rinsing step . The cell pellet was resuspended in 5 ml of ice-cold staining buffer ( DPBS containing 2% heat inactivated FCS and 1 mM EDTA ) . Before antibody staining , the cells were incubated with Fc receptor blocking antibody ( CD16/32 ) for 5 min . The cells were incubated with the CD31 , PDGFRa/CD140a , CD45 , and Ter119 antibodies for 30 min ( Key Resource Table for the antibody details ) . Prior to FACS , the cells were rinsed twice with the staining buffer and filtered through 5 ml cell strainer tubes ( Corning , #352235 ) . The cells were passed through a 100 μm nozzle . Multiple light scattering parameters for forward- and side-scatter properties of the cells were employed to gate , analyze , and sort live cardiac ECs . Initially , total cells were gated based on the forward and side-scatter area of the cells ( FSC-A and SSC-A ) . The single cells were selected depending on forward scatter parameters area , height , and width of the cells ( FSC-A , FSC-H , or FSC-W ) . DAPI was used to determine live and dead cells . To enrich and FACS sort pure and viable cardiac ECs , ECs were stained with CD31 , mesenchymal cells with PDGFRa/CD140a , leucocytes with CD45 , and red blood cells with Ter119 . The live cardiac ECs were defined as CD31+ CD45- Ter119- CD140a- DAPI- . Cells were sorted using FACS Aria II ( BD Biosciences ) , and the data was acquired with BD FACSDIVA v8 . 0 . 1 and further analyzed with FlowJo v10 . 1 ( FlowJo , LLC ) software . We verified the enrichment and purity of the FACS sorted Cardiac EC population ( CD31+ PDGFRa ( CD140a ) - CD45- Ter119- DAPI- ) by QPCR analysis for classical cardiac EC markers . Recently , we have used the same isolation method for single-cell RNAseq experiments , and these results show that there is about 3% contamination from other cells types , mainly pericytes and hemangioblasts . The sorted cardiac ECs were immediately suspended in lysis buffer ( 350 μl of RLT buffer plus 10 μl of β-mercaptoethanol ) , the cells were homogenized in QIAshredder ( #79654 , Qiagen ) , and the RNA was purified using RNeasy Plus Micro Kit ( #74034 , Qiagen ) according to the manufacturer’s instruction . The RNA integrity was analyzed with bioanalyzer ( Agilent Tape Station 4200 ) and the concentration was determined by Qubit fluorescence assay ( ThermoFisher ) . The cells from the post sort fractions were stained with propidium iodide ( PI ) and the viability of the cells were determined by Luna automated cell counter . The purity of the post sort fraction was determined by QPCR analysis for EC markers . Indexed cDNA library was synthesized using SMARTer Stranded Total RNA-Seq Kit V2 – Pico Input Mammalian ( Takara Bio , USA ) kit according to the manufacturer’s instructions . The library quality was determined using bioanalyzer and sequenced using illumina NextSeq 550 System with the following specifications: 1 × 75 bp , 50M single end reads were sequenced using NextSeq 500/550 High-Output v2 . 5 kit . The sequenced reads were analyzed with the following software packages embedded in the Chipster analysis platform ( Kallio et al . , 2011 ) ( v3 . 12 . 2; https://chipster . csc . fi ) . Trimmomatic tool ( Bolger et al . , 2014 ) ( https://chipster . csc . fi/manual/trimmomatic . html ) was used to preprocess Illumina single end reads . The HISAT2 package ( Kim et al . , 2015 ) ( https://chipster . csc . fi/manual/hisat2 . html ) was employed to align the reads to mouse genome GRCm38 . 90 and the HTSeq count tool ( Anders et al . , 2015 ) ( https://chipster . csc . fi/manual/htseq-count . html ) to quantify the aligned reads per gene . The raw read count table for genes generated utilizing the HTSeq count were used as an input to perform two-dimensional principal component analysis ( PCA ) and unsupervised hierarchical clustering analysis using DESeq2 Bioconductor package ( Love et al . , 2014 ) ( https://chipster . csc . fi/manual/deseq2-pca-heatmap . html ) . Next , to perform the differential gene expression ( DGE ) analysis , the DESeq2 Bioconductor package ( Love et al . , 2014 ) was used . The advantage of DEseq2 tool is sensitive and precise for analyzing the DEG in studies with few biological replicates . To reliably estimate the within group variance , Empirical Bayes shrinkage for dispersion estimation was used and a dispersion value for each gene was estimated through a model fit procedure ( refer to the Figure 2—figure supplement 2A , which illustrates the shrinkage estimation for the experimental conditions ) . The gene features obtained after the dispersion estimation were used to perform statistical testing . Next , negative binomial generalized linear model was fitted for each gene and Wald test ( raw p-value ) was calculated to test the significance . Finally , DEseq2 applies Benjamini–Hochberg correction test to control the FDR ( refer to the Figure 2—figure supplement 2B indicating the distribution of raw and FDR adjusted p-value for the experimental conditions ) . In our DEG analysis , we have set the FDR ( p adj . ) cutoff as less than or equal to 0 . 05 ( FDR/p-adj ≤ 0 . 05 ) for pathway analysis and gene overlap analysis . The RNA sequencing data is deposited in the GEO database , under the series accession number GSE145263 . The gene function and pathway analysis of the DGE were determined by performing statistical overrepresentation test using the PANTHER classification system ( Mi et al . , 2019 ) ( V . 14 . 1; http://www . pantherdb . org ) . The p<0 . 05 was considered for the further analysis and the data is presented as −log2 ( p-value ) . The differentially expressed up- and downregulated genes ( adjusted p-value 0 . 05 ) from the different experimental conditions were imported to VENNY 2 . 1 Venn-diagram analysis software ( BioinfoGP; https://bioinfogp . cnb . csic . es/tools/venny/ ) to identify genes which were significantly affected by several experimental conditions . The MetazSecKB knowledgebase ( Meinken et al . , 2015 ) ( http://proteomics . ysu . edu/secretomes/animal/index . php ) , TargetP2 . 0 server ( Almagro Armenteros et al . , 2019 ) ( http://www . cbs . dtu . dk/services/TargetP/index . php ) , and SecretomeP1 . 0 server ( Bendtsen et al . , 2004 ) ( http://www . cbs . dtu . dk/services/SecretomeP-1 . 0/ ) were used to characterize molecular functions , subcellular localizations , and possible secretion properties of the identified common genes . HUVEC and HCAEC were purchased from PromoCell ( cell lines were authenticated and tested for mycoplasma status by the vendor ) . Cell cultures in the lab are regularly checked for their mycoplasma status using Mycoalert mycoplasma detection kit ( LT07-218 , Lonza ) . Both HUVEC and HCAEC were cultured and maintained in EC growth Basal Medium MV ( C-22220 , PromoCell ) supplemented with Supplement Pack GM MV ( C-39220 , PromoCell ) and gentamycin . For both gene overexpression and silencing studies , 80% confluent monolayer culture of HUVECs and HCAECs was used . To overexpress SERPINH1 in EC , we cloned a lentiviral vector FUW-hSERPINH1-Myc ( map and plasmid available by request ) . A scrambled sequence in the same vector was used as a control . 293FT cells ( ATCC ) were cultured and maintained in DMEM supplemented with 10% FCS and L-glutamine , and co-transfected with the lentiviral packaging plasmid vectors CMVg , CMV∆8 . 9 , and the target plasmid . The supernatants were collected at 48 and 72 hr , and concentrated by ultracentrifugation as described previously ( Lois et al . , 2002 ) . For overexpression , HUVEC and HCAEC were transfected with lentivectors for 48 hr . For gene silencing studies , HCAEC was treated with lentivectors encoding for four independent clones of human shSERPINH1 for 24 hr . Subsequently , the cells were treated with puromycin ( 2 µg/ml ) for 48 hr to select the transduced cells . After selection , the cells were used for further analysis . The clone id and target sequence for human shSERPINH1 constructs are shown in the Key Resource Table . The SERPINH1 overexpressed or silenced HCAECs were seeded in the IncuCyte ImageLock 96-well microplate precoated with 0 . 1% gelatin and cultured in complete EC growth medium . To the confluent cell monolayers , 700–800 micron scratch wounds were introduced with IncuCyte WoundMaker , and the wells were briefly rinsed with and maintained in complete EC growth medium . The kinetics of the cell migration were recorded and 10× phase contrast time-lapse images were acquired using IncuCyte Live-Cell Analysis System . The wound closure region was measured by Edge-detection and thresholding method in Image J software ( NIH ) . The data is presented as wound closure ( % ) relative to time . The coverslips or 6-well plates were precoated with 0 . 1% gelatin for 20 min at 37°C , scrambled or SERPINH1 silenced HCAEC were seeded and cultured in complete EC growth medium . The cells were treated with or without 50 ng/ml of recombinant human TGF-β ( R and D Technologies ) and/or 200 μM hydrogen peroxide ( Acros organics ) for 5 days as described previously ( Evrard et al . , 2016; Magenta et al . , 2011 ) . The cells grown on the coverslips were fixed with 4% PFA in PBS for 15 min . Blocking was done using donkey immunomix and the cells were stained with primary antibodies and secondary antibodies as indicated in the Key resources table . DAPI was used to stain the nucleus , and the cells were mounted using Vectashield ( Vector labs ) . The amount of COL1 was quantified by adjusting 10× images for threshold and area fraction tool was used to quantify the area percentage of the collagen deposition ( Image J software , NIH ) . The SERPINH1 overexpressed HCAECs at passage 6 ( P6 ) were seeded on the coverslips coated with 0 . 1% gelatin . The cells were allowed to reach 80% confluence , rinsed twice with ice cold PBS , incubated in the fixative solution ( #11674 , Cell signaling technology ) at room temperature for 10 min , rinsed twice with ice cold PBS , and stored at 4°C . The senescence-associated beta-galactosidase ( SA-β-gal ) activity at pH 6 . 0 was detected with the SA-β-gal staining kit ( #9860 , Cell Signaling Technology ) according to the manufacturer’s instructions . The SA-β-gal+ cells were quantified using point tool ( Image J software , NIH ) and normalized to the total number of cells per field . The data presented as percentage of SA-β-gal+ cells of all cells . To detect SA-β-gal activity in the hearts of HFD- and chow-fed mice , 4 μm thick cryo sections were fixed with 1% PFA in PBS for 1 min at room temperature , and the sections were rinsed twice with PBS and incubated in the β-galactosidase staining solution , pH 6 . 0 ( #9860 , Cell signaling technology ) for 24 hr at 37°C . The slides were rinsed twice with PBS , post fixed with 1%PFA in PBS for 1 min , rinsed twice with PBS , counter stained with 0 . 1% eosin , rinsed twice with distilled water , and the sections were mounted with Immuno-mount ( Thermo scientific ) ( Cazin et al . , 2017 ) . The images were acquired using light microscope ( Leica ) and the SA-B-gal+ cells were scored using point tool ( Image J software , NIH ) per field . The data is presented as SA-β-gal+ cells/field . The SERPINH1 overexpressed and silenced HCAECs were seeded on the coverslips coated with 0 . 1% gelatin and cultured in complete EC growth medium overnight . The cells were allowed to reach 70% confluence , incubated with 10 μM of EdU labeling solution in EC complete growth media for 7 hr under normal culture conditions . The cells were rinsed twice with ice cold PBS and fixed with 4% PFA in PBS for 10 min . The manufacturer’s instructions in the Click-iT EdU Alexa Fluor 594 staining kit ( Thermo scientific ) were followed to detect the Edu+ proliferating cells and the nuclei were counterstained with Hoechst . The percentage of Edu+ cells were normalized to Hoechst+ nuclei using area fraction tool ( Image J software , NIH ) . The data is presented as percentage of EdU/Hoechst ( % ) . To analyze the expression of SASP genes in our cardiac EC RNA-seq data sets ( HFD and TAC models ) , we have reviewed and compared the DEGs in our data sets with FDR threshold of 0 . 05 using the following databases: SASP atlas ( http://www . saspatlas . com ) ( Basisty et al . , 2020 ) and SeneQuest ( https://senequest . net ) . Further , the endothelial expression of the identified genes was checked using Tabula Muris database ( https://tabula-muris . ds . czbiohub . org ) . RNA from the cultured cells was purified and isolated using NucleoSpin RNA II Kit according to the manufacturer’s protocol ( Macherey-Nagel ) . cDNA was synthesized with High-Capacity cDNA Reverse Transcription Kit ( Applied biosystems , #4368814 ) . SYBR green or TaqMan gene expression assays were performed using FastStart Universal SYBR green master mix ( Sigma-Aldrich , #04913914001 ) and TaqMan gene expression master mix ( Applied Biosystems , #4369016 ) , respectively . mRNA expression was analyzed using Bio-Rad C1000 thermal cycler according to standardized protocol of the qPCR master mix supplier . The average of the technical triplicates for each sample was normalized to the housekeeping gene HPRT1 . The mRNA expression levels were calculated and presented as fold change ( Ctrl = 1 ) . The primer sequences are listed in the Key Resource Table . The cells were harvested and homogenized in lysis buffer containing 0 . 5% NP-40 ( v/v ) and 0 . 5%Triton X-100 ( v/v ) in PBS , supplemented with protease and phosphatase inhibitors ( A32959 , Pierce , Thermo Scientific ) . Protein concentration was determined using a BCA protein assay kit ( Pierce , Thermo Scientific ) . Equal amounts of total protein were resolved in Mini-PROTEAN TGX Precast gels ( Bio-Rad ) and transferred to PVDF membrane ( immobilon-P , Millipore ) . 5% BSA ( wt/vol ) and 0 . 1% Tween 20 ( v/v ) in TBS were used to block the membranes followed by incubation with primary antibodies listed in the ( Key Resource Table ) overnight at 4°C . HRP-conjugated secondary antibodies ( DAKO ) were used , and HRP signals were developed with Super-Signal West Pico Chemiluminescent substrate or Femto Maximum sensitivity substrate ( Thermo Scientific ) . The blots were imaged with Odyssey imager ( Li-COR Biosciences ) or Chemi Doc imaging system ( Bio-Rad ) and quantified with Image Studio Lite Software ( Li-COR Biosciences ) . The data from the individual experiments were analyzed by Student’s t-test . p<0 . 05 value was considered statistically significant and p-values in the graphs are shown as *p<0 . 05 , **p<0 . 01 , and ***p<0 . 001 . The data is shown as mean ± SEM . The GraphPad Prism 7 software was used for statistical analysis . | Cardiovascular diseases are the number one cause of death in the western world . Endothelial cells that line the blood vessels of the heart play a central role in the development of these diseases . In addition to helping transport blood , these cells support the normal running of the heart , and help it to grow and regenerate . Over time as the body ages and experiences stress , endothelial cells start to deteriorate . This can cause the cells to undergo senescence and stop dividing , and lay down scar-like tissue via a process called fibrosis . As a result , the blood vessels start to stiffen and become less susceptible to repair . Ageing , obesity , high blood pressure , and inactivity all increase the risk of developing cardiovascular diseases , whereas regular exercise has a protective effect . But it was unclear how these different factors affect endothelial cells . To investigate this , Hemanthakumar et al . compared the gene activity of different sets of mice: old vs young , obese vs lean , heart problems vs healthy , and fit vs sedentary . All these risk factors – age , weight , inactivity and heart defects – caused the mice’s endothelial cells to activate mechanisms that lead to stress , senescence and fibrosis . Whereas exercise training had the opposite effect , and turned off the same genes and pathways . All of the at-risk groups also had high levels of a gene called SerpinH1 , which helps produce tissue fiber and collagen . Experiments increasing the levels of SerpinH1 in human endothelial cells grown in the laboratory recreated the effects seen in mice , and switched on markers of stress , senescence and fibrosis . According to the World Health Organization , cardiovascular disease now accounts for 10% of the disease burden worldwide . Revealing the affects it has on gene activity could help identify new targets for drug development , such as SerpinH1 . Understanding the molecular effects of exercise on blood vessels could also aid in the design of treatments that mimic exercise . This could help people who are unable to follow training programs to reduce their risk of cardiovascular disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"medicine"
] | 2021 | Cardiovascular disease risk factors induce mesenchymal features and senescence in mouse cardiac endothelial cells |
Trade-offs between protein stability and activity can restrict access to evolutionary trajectories , but widespread epistasis may facilitate indirect routes to adaptation . This may be enhanced by natural environmental variation , but in multicellular organisms this process is poorly understood . We investigated a paradoxical trajectory taken during the evolution of tetrapod dim-light vision , where in the rod visual pigment rhodopsin , E122 was fixed 350 million years ago , a residue associated with increased active-state ( MII ) stability but greatly diminished rod photosensitivity . Here , we demonstrate that high MII stability could have likely evolved without E122 , but instead , selection appears to have entrenched E122 in tetrapods via epistatic interactions with nearby coevolving sites . In fishes by contrast , selection may have exploited these epistatic effects to explore alternative trajectories , but via indirect routes with low MII stability . Our results suggest that within tetrapods , E122 and high MII stability cannot be sacrificed—not even for improvements to rod photosensitivity .
Nature-inspired strategies are increasingly recruited toward engineering objectives in protein design ( Khersonsky and Fleishman , 2016; Jacobs et al . , 2016; Goldenzweig and Fleishman , 2018 , a central challenge of which is to successfully manipulate backbone structure to modulate stability without introducing undesirable pleiotropic effects on protein activity ( Khersonsky and Fleishman , 2016; Goldenzweig and Fleishman , 2018; Starr and Thornton , 2017; Tokuriki and Tawfik , 2009 ) . Engineering protein stability and activity requires an understanding of a protein’s sequence-function relationship , or landscape ( Pál and Papp , 2017; Wu et al . , 2016; Starr et al . , 2017 ) , where billions of possible pair-wise and third-order interactions can exist between amino acids ( Starr and Thornton , 2017; Storz , 2016 ) , and only a limited number of amino acid combinations will confer the function of interest ( Wu et al . , 2016; Starr et al . , 2017; McMurrough et al . , 2014; Mateu and Fersht , 1999; Tarvin et al . , 2017 ) . To understand the context-dependence of amino acid functional effects ( also known as intramolecular epistasis [Starr et al . , 2017; Storz , 2016; Echave et al . , 2016] ) , approaches such as deep mutational scanning ( Wu et al . , 2016; Starr et al . , 2017; Sailer and Harms , 2017 ) can explore a subset of sequence-function space formed in response to a limited set of artificial selection pressures ( Starr and Thornton , 2017 ) . By contrast , natural protein sequence variation reflects the range of protein function that evolved in response to changing ecological variables ( Starr and Thornton , 2017; Pál and Papp , 2017; Ogbunugafor et al . , 2016 ) , where convergent ‘solutions’ for protein function and stability can be derived through the evolution of alternative protein sequences ( McMurrough et al . , 2014; Mateu and Fersht , 1999; Tarvin et al . , 2017 ) . This suggests that closer examination of natural sequence variation may reveal new blueprints for protein design . The dim-light visual pigment rhodopsin ( RH1/RHO ) is an excellent model for understanding how both ecological variables and biophysical pleiotropy may interact to determine the availability of functional evolutionary solutions for environmental challenges ( Kojima et al . , 2017; Gozem et al . , 2012; Dungan and Chang , 2017; Castiglione et al . , 2018 ) . Spectral tuning mutations that shift the RH1 wavelength of maximum absorbance ( λMAX ) can adapt dim-light vision to a remarkable range of spectral conditions across aquatic and terrestrial visual ecologies ( Hunt et al . , 2001; Hauser and Chang , 2017a; Dungan et al . , 2016 ) . Recently , λMAX was revealed to exist within a complex series of epistasis-mediated trade-offs with the non-spectral functional properties of RH1 long understood as adaptations for dim-light ( Gozem et al . , 2012; Dungan and Chang , 2017; Castiglione et al . , 2017; Hauser et al . , 2017b ) . These include an elevated barrier to spontaneous thermal-activation , which minimizes rod dark noise and is promoted by blue-shifts in λMAX ( Kojima et al . , 2017; Gozem et al . , 2012; Kefalov et al . , 2003; Yue et al . , 2017 ) ; and a slow decay of its light-activated conformation , which we refer to here as metarhodopsin-II ( MII ) for simplicity ( Imai et al . , 1997; Lamb et al . , 2016; Kojima et al . , 2014; Sommer et al . , 2012; Schafer et al . , 2016; Van Eps et al . , 2017 ) . The RH1 MII active conformation is associated with rapid and efficient activation of G-protein transducin ( Gt ) ( Kojima et al . , 2014; Sugawara et al . , 2010 ) , yet the reasons for its long-decay after Gt signaling remain unclear ( Kefalov et al . , 2003; Imai et al . , 1997; Imai et al . , 2007 ) . To sustain vision , all-trans retinal ( atRAL ) chromophore must be released from MII after Gt signaling ( Palczewski , 2006 ) —a process that depends on the conformational stability of the MII-active-state structure ( Schafer et al . , 2016; Schafer and Farrens , 2015 ) . Cone opsins have low MII stability and therefore rapidly release atRAL ( Imai et al . , 1997; Chen et al . , 2012a ) , where it is quickly recycled back into 11-cis retinal ( 11CR ) through the cone visual ( retinoid ) cycle , enabling rapid regeneration of cone pigments for bright-light vision ( Wang and Kefalov , 2011; Tsybovsky and Palczewski , 2015 ) . Rods , in contrast , regenerate thousands of times slower than cones after bright-light exposure ( Mata et al . , 2002 ) . Indeed , rod exposure to bright flashes of light leads to atRAL release that can outpace clearance by visual cycle enzymes ( Sommer et al . , 2014; Rózanowska and Sarna , 2005 ) , thus leading to accumulation ( Saari et al . , 1998; Lee et al . , 2010 ) and light-induced retinopathy through various modes of cellular toxicity involving oxidative stress ( Maeda et al . , 2009; Chen et al . , 2012b ) . Interestingly , recent biochemical evidence suggests MII may play a role in retinal photoprotection by complexing with arrestin after Gt signaling to re-uptake and thus provide a sink for toxic atRAL after rod photobleaching ( Sommer et al . , 2014 ) . This suggests the evolution of rhodopsin’s high conformational selectivity for toxic atRAL may be a functional specialization ( Schafer et al . , 2016; Schafer and Farrens , 2015 ) , which could in turn reflect differences in retinoid metabolism between rods vs . cones ( Wang and Kefalov , 2011; Tsybovsky and Palczewski , 2015; Imai et al . , 2005 ) . Consistent with the overlapping mechanisms of RH1 spectral and non-spectral functions via the highly constrained RH1 structure ( Gozem et al . , 2012; Yue et al . , 2017 ) , this biophysical pleiotropy likely necessitates costly trade-offs between the spectral and non-spectral functions of RH1 in natural systems ( Dungan and Chang , 2017; Luk et al . , 2016 ) . By comparison , directed evolution and synthetic biology approaches have successfully engineered either spectral , or non-spectral aspects of rhodopsin function , but did not address trade-offs arising from shifts in function . It has thus been possible to shift the spectral absorbance of archaea and bacterial rhodopsins close to the limit of the visible spectrum ( Herwig et al . , 2017; McIsaac et al . , 2014 ) , and to engineer tetrapod rhodopsins with high thermal stability ( Xie et al . , 2003 ) , constitutive activation ( Deupi et al . , 2012; Standfuss et al . , 2011 ) , and alternative chromophore-binding sites ( Devine et al . , 2013 ) . However , it has not been investigated whether rod visual pigments with novel combinations of spectral and non-spectral functional properties can be engineered by manipulating the biophysical pleiotropy of RH1 otherwise exploited by natural selection . Site 122 ( Bos taurus RH1 numbering ) is a molecular determinant of both the spectral and non-spectral functional properties of rhodopsin and the cone opsins ( Hunt et al . , 2001; Yue et al . , 2017; Imai et al . , 1997; Imai et al . , 2007; Yokoyama et al . , 1999 ) . Intriguingly , vertebrate visual pigment families show differences in which amino acid variants predominate at this site ( Figure 1A ) , with I122 strongly conserved in the most ancestrally diverging cone opsins such as the long-wave sensitive opsins ( LWS ) ( Lamb et al . , 2007 ) , whereas in the most derived opsin group , the rhodopsins ( RH1 ) , E122 predominates ( Figure 1B , C ) ( Imai et al . , 1997; Lamb et al . , 2007; Imai et al . , 2007; Carleton et al . , 2005 ) . E122 is a key component of an important hydrogen-bonding network with H211 that is known to stabilize the MII active-conformation ( Choe et al . , 2011 ) . This stability increase is so dramatic that E122 is considered a functional determinant distinguishing rhodopsin from cone opsins ( Figure 1B ) ( Imai et al . , 1997; Lamb et al . , 2016; Kojima et al . , 2014 ) . Paradoxically , by conferring this increase in MII stability , the evolution of E122 likely involved a costly fitness trade-off that diminished tetrapod rod photosensitivity ( Yue et al . , 2017 ) , which can affect visual performance in animals ( Kojima et al . , 2017; Aho et al . , 1988 ) . Indeed , it is possible to improve tetrapod rod photoreceptor sensitivity by decreasing rod dark noise in vivo by replacing E122 with a cone opsin amino acid variant ( COV; Figure 1A ) at site 122 , such as Q122 , which predominates in RH2 cone opsins ( Yue et al . , 2017; Lin et al . , 2017 ) . The strict conservation of E122 in all tetrapod rhodopsins ( Figure 1C , Table 1 , Supplementary file 1 ) therefore suggests that during the evolution of tetrapod dim-light vision , natural selection may have prioritized MII stability ( Figure 1B , D ) at the expense of rod sensitivity . This apparent evolutionary trade-off is perplexing given that the low spontaneous thermal activation of rhodopsin ( and therefore rod dark noise ) is a functional hallmark of rhodopsin divergence from the cone opsins ( Kojima et al . , 2017; Gozem et al . , 2012; Kefalov et al . , 2003; Lamb et al . , 2016 ) . Why has tetrapod RH1 been constrained to this paradoxical compromise at site 122 for the last 350 million years ? Interestingly , and in contrast to tetrapod rhodopsins , fish rhodopsins show variation at site 122 , such as in the Coelacanth ( Latimeria chalumnae ) , Lungfish ( Neoceratodus forsteri ) , and deep-sea fish lineages , where COV ( I , Q , M ) and other residues at site 122 ( V , D ) are found ( Figure 1C; Figure 1—figure supplement 1; Tables 1–2 , Supplementary file 2 ) ( Hunt et al . , 2001; Yokoyama et al . , 1999; Carleton et al . , 2005 ) . These substitutions have been shown to blue-shift λMAX by up to ~10 nm ( Hunt et al . , 2001; Yokoyama et al . , 1999 ) , and may improve dim-light sensitivity in poorly-lit aquatic environments ( Yue et al . , 2017 ) . Strikingly , one of the largest freshwater groups—the Characiphysi ( which includes piranhas , electric eels , and catfishes [Chen et al . , 2013] ) —has the COV I122 residue completely fixed ( Figure 1—figure supplement 1 , Supplementary file 3 ) . In tetrapods by contrast , the red-shifting E122 mutation is strictly maintained , increasing MII stability ( Imai et al . , 1997 ) but greatly decreasing rod sensitivity ( Yue et al . , 2017 ) . Why the strong constraints on high MII stability and E122 are relaxed only within certain aquatic visual ecologies , remains unknown . In light of these ecological patterns , we questioned whether it was possible to synthesize an evolutionary alternative: a tetrapod RH1 that never lost COV at site 122 but still developed high MII stability . We reasoned that relative to tetrapods , the diversity and complexity of fish visual ecologies ( Hunt et al . , 2001; Hauser and Chang , 2017a ) may have allowed selection the opportunity to explore the pleiotropic potential of site 122 through the evolution of novel structural interactions with nearby sites that could compensate for the destabilizing loss of the E122-H211 hydrogen bond . To identify these interactions , our goal was to use analyses of evolutionary rates to predict sites coevolving with site 122 , and to investigate the functional consequences of coevolving sites with experimental site-directed mutagenesis studies . Ultimately , we used our analyses of natural variation as a guide to artificially engineer a tetrapod rhodopsin with increased MII stability , but within a non-E122 sequence background . We demonstrated that this synthetic alternative is possible , even if evolution did not proceed down this mechanistic trajectory toward a dim-light adapted visual pigment .
To better understand the selection pressures that may be constraining E122 to fixation during tetrapod evolution , we constructed a large vertebrate rhodopsin phylogenetic dataset ( Figure 1—figure supplements 1 and 2 , Supplementary files 1–2 ) and investigated the evolutionary history of site 122 using ancestral reconstruction ( Materials and methods ) . We found that E122 ( codon GAA; Figure 2A ) has been fixed in tetrapod RH1 since the most recent common ancestor ~350 million years ago ( MYA ) ( Hedges et al . , 2015 ) , where it appears along the ancestral branch leading to tetrapods ( Figure 2A; Table 3 ) following the diversification from lungfishes ( I122 , codon ATA , Figure 2A; Supplementary file 1 ) and the coelacanth ( Q122 , codon CAA , Figure 2A; Supplementary file 1 ) . This transition period in vertebrate evolution is characterized by extensive morphological modifications for vision within terrestrial environments , and likely included large increases in environmental light irradiance ( MacIver et al . , 2017; Warrant and Johnsen , 2013 ) . Apart from the lungfishes and coelacanth , the high conservation of E122 in tetrapods is also reflected in other vertebrate rhodopsins ( Figure 1 , Figure 1—figure supplement 1; Tables 1–2; Supplementary files 1–2 ) , but there are important exceptions within certain lineages of teleost fishes , such as the Characiphysi . Within this group , the COV residue I122 was introduced likely through E122I ( codon ATC; Figure 2A ) , where I122 is now completely fixed across the extant Characiphysi ( Supplementary file 3 ) . Since fishes ( Teleosts ) , unlike tetrapods , display amino acid variation at site 122 ( Figure 1C ) , we hypothesized that compensatory mutations may be coevolving with site 122 across fish RH1 . To test this hypothesis , we investigated across the entire transmembrane domain of rhodopsin ( residues 53–302 ) for evidence of sites coevolving with site 122 within an alignment of Teleost RH1 ( Materials and methods; Supplementary file 2 ) . Using phylogenetically corrected mutual information ( MI ) analyses ( MISTIC; [ ( Simonetti et al . , 2013] ) with z-score cut-off determined by analyses of randomized datasets ( Ashenberg and Laub , 2013 ) , we found significant evidence of coevolution with site 122 at several RH1 positions , all of which clustered within 6 Å of E122 ( Figure 2B ) in the MII crystal structure ( Choe et al . , 2011 . This is within the range at which intramolecular forces such as Van der Waals and hydrophobic interactions between amino acids are thought to occur ( Ivankov et al . , 2014 ) . It is known , however , that there is a tendency of covariation analyses such as MI to identify coevolving sites proximal to each other , which may in turn overlook more distal coevolving sites potentially indirectly interacting with site 122 ( Talavera et al . , 2015 ) . Nevertheless , sites detected within this 6 Å radius ( sites 119 , 123 ) have been previously found capable of functionally compensating for human pathogenic mutations ( e . g . A164V ) disrupting the MII-stabilizing E122-H211 interaction ( Stojanovic et al . , 2003 ) , suggesting that natural variation at coevolving sites within this radius could compensate for the functional effects of COV at site 122 . We therefore decided to focus our investigations on identifying natural compensatory mutations at sites within this 6 Å radius . Relative to Teleost RH1 ( where site 122 varies ) , we found that sites within this radius displayed decreased amino acid variation in Tetrapod and Characiphysi RH1 , where E122 and I122 are fixed , respectively ( asterisks , Figure 2C ) . This observation is consistent with an intramolecular evolutionary process known as entrenchment ( Pollock et al . , 2012; Goldstein and Pollock , 2017; Shah et al . , 2015 ) , where functionally favourable amino acid residues compensating for an original mutation tend to become fixed , thus mutually entrenching favourable amino acids at each position within the coevolving network . We therefore reasoned that if residues at nearby positions are indeed compensatory , then these sites should display a relative decrease in amino acid variation specifically in those vertebrate lineages where an amino acid has been fixed at site 122-- such as E122 in tetrapods and I122 in the Characiphysi . Furthermore , we hypothesized that decreases in amino acid variation observed in these lineages would be driven by an increase in purifying selection on non-synonymous codons , ultimately reflecting the entrenchment of compensatory amino acid residues by natural selection . We therefore employed codon-based phylogenetic likelihood methods to test for a relative increase of purifying selection at RH1 sites within 6 Å of site 122 , within Tetrapod vs Teleosts , as well as in Characiphysi vs other Teleosts ( Yang , 2007 ) ( Materials and methods ) . Using likelihood ratio tests of alternative ( Clade model C [ ( Bielawski and Yang , 2004] ) and null ( M2a_REL ( [Weadick and Chang , 2012 ) ] ) model analyses of codon substitution rates ( dN/dS ) across the RH1 coding-sequence , we identified statistically significant evidence of gene-wide increases in purifying selection within Tetrapods ( Table 3 ) and Characiphysi ( Table 5 ) relative to teleosts ( ( p<0 . 001 ) ) . Sites estimated to be under this increase in purifying selection were those identified in the CmC divergent site class through a Bayes empirical Bayes analysis as previously described ( Castiglione et al . , 2017 ) . Consistent with the fixation of E122 and I122 in tetrapod and Characiphysi RH1 , respectively ( asterisks , Figure 2C ) , we detected a relative increase of purifying selection on site 122 codons in tetrapod and Characiphysi RH1 relative to that of teleosts ( Figure 2D; Tables 3–5 ) , suggesting that a corresponding increase of purifying selection may have occurred at putatively coevolving sites within the 6 Å radius ( Pollock et al . , 2012; Goldstein and Pollock , 2017; Shah et al . , 2015 ) . No evidence for this was detected at sites 126 and 211 , the other members of the TM3-TM5 domain stabilizing the MII active-state ( Table 4; [ ( Choe et al . , 2011 ) ] ) . Yet within this radius , we found significant evidence for a relative increase of purifying selection in tetrapods and the Characiphysi ( relative to teleosts ) at several RH1 sites ( 119 , 124 , 168; Figure 2D; Table 4 ) , some of which ( sites 119; 168 ) also displayed significant statistical evidence for covariation ( MI ) with site 122 in Teleost RH1 ( Figure 2B vs . 2D; Table 4 ) . Furthermore , one of these sites ( 119 ) also exceeded the significance threshold in our Bonferroni-corrected phylogenetic tests of correlated evolution with site 122 where p-values were calculated by performing Monte Carlo tests using data from simulations ( Pagel , 1994 ) ( Materials and methods; Table 4 ) . Taken together , these results provide evidence that an increase in purifying selection on non-synonymous codons drove the reduction in amino acid variation at positions coevolving with site 122 , and this likely accompanied the fixation of E122 and I122 in tetrapods and the Characiphysi , respectively . Due to the consistency of these findings with coevolutionary entrenchment , we hypothesized that we could identify fixed residues within this 6 Å radius in Characiphysi RH1 that may be functionally compensatory for the ancient E122I mutation that occurred in the ancestral Characiphysi ( Figure 2A ) . Of the RH1 sites displaying significant statistical evidence for covariation ( MI ) with site 122 in Teleost RH1 ( Figure 2B; 119 , 127 , 168 ) , as well as those displaying significant evidence for a relative increase of purifying selection in tetrapods and the Characiphysi ( relative to teleosts; 119 , 124 , 168; Figure 2D; Table 4 ) only sites 119 , 124 and 127 had fixed amino acid residues in Characiphysi RH1 relative to other Teleosts ( asterisks , Figure 2C ) , suggesting this strict conservation pattern may reflect entrenchment due to the fixation of I122 . In contrast , despite a statistically significant increase in purifying selection on non-synonymous codons relative to other Teleosts , site 168 nevertheless displayed amino acid variation in the Characiphysi ( T/V168; Figure 2C , D ) , suggesting it may not necessarily play a functionally compensatory role for the ancient E122I mutation , especially since T vs . V168 may be reasonably expected to have biochemically and/or structurally dissimilar effects on this region of the rhodopsin TM3-TM5 microdomain ( Choe et al . , 2011 ) . Conversely , although C127 has been fixed in Characiphysi RH1 relative to other Teleosts ( asterisks , Figure 2C ) and may therefore be functionally important , there was no increase in purifying selection on non-synonymous codons at site 127 relative to Teleosts ( Figure 2D ) , suggesting that the fixation of C127 in Characiphysi RH1 may be a historical contingency that does not necessarily reflect intramolecular entrenchment by the ancient E122I mutation ( Goldstein and Pollock , 2017 . Although this same logic ostensibly applies to site 123 , unlike C127—a residue shared with some tetrapods ( Figure 2C; Supplementary files 1–3 ) —we observed a striking fixation of a rare amino acid residue in Characiphysi RH1 ( N123 , asterisks , Figure 2C ) which is not , to our knowledge , observed within any vertebrate rhodopsin other than the Characiphysi where it is completely fixed ( Tables 1–2 , Supplementary files 1–3 ) , and located between coevolving sites 119 , 122 and 124 which are also fixed in the Characiphysi ( Figures 2C; 3A ) . This unique natural variation is particularly interesting as site 123 has been previously found capable of functionally compensating for human pathogenic mutations ( e . g . A164V ) disrupting the MII-stabilizing E122-H211 interaction ( Stojanovic et al . , 2003 ) . Therefore , we decided to focus on sites 119 , 123 , and 124 , two of which ( 119 , 123 ) are thought to have functional effects via the TM3-TM5 microdomain ( Stojanovic et al . , 2003 ) . Altogether , these sites are located in close proximity to several important structural regions known to affect MII stability , such as N302 of the NPxxY motif , the TM3-TM5 microdomain involving sites 122–211 , as well as the all-trans retinal binding pocket ( Choe et al . , 2011 ) ; Figure 3A ) , suggesting that residues at these positions may form novel structural interactions that could compensate for the destabilizing loss of the E122-H211 hydrogen bond ( Stojanovic et al . , 2003; Morrow et al . , 2017 ) . Consistent with the entrenchment of compensatory mutations at coevolving sites ( Talavera et al . , 2015; Pollock et al . , 2012; Shah et al . , 2015 ) , using ancestral reconstruction we found that sites 119 , 122 , 123 , and 124 are strongly conserved as the LxxEIA ( L119-E122-I123-A124; referred to as ‘LEIA’ ) and FxxINS motifs ( F119-I122-N123-S124; ‘FINS’ ) within tetrapod and Characiphysi RH1 , respectively , likely since the most recent common ancestor of each lineage , where LEIA is predicted as the ancestral Osteichthyes motif ( Figure 3B; Tables 1 , Supplementary files 1 , 3 ) . The maintenance of these two completely different amino acid motifs in Characiphysi and tetrapod RH1 strongly suggests that natural selection has constrained intramolecular interactions at these sites , which we hypothesized to be associated with modulating the pleiotropic functional consequences of sequence variation at site 122 . We therefore tested the ability of coevolving sites 119 , 123 and 124 to affect tetrapod rhodopsin function and the potential for natural variation at these sites to compensate for the destabilizing loss of the E122-H211 hydrogen bond . We conducted site-directed mutagenesis and in vitro expression of mutant rhodopsins using detergent micelles ( Materials and methods ) . This was followed by in vitro functional characterization using spectroscopic absorbance- and fluorescence-based measurements of both λMAX and the stability of the active-state conformation ( Figure 4; Figure 4—figure supplement 1; Table 6; Materials and methods ) , both of which can provide information on relative differences that exist within natural systems ( Schafer et al . , 2016; Van Eps et al . , 2017; Schott et al . , 2016a . Tetrapod RH1 with E122I ( Figure 4A ) and other FINS motif single substitutions to the coevolving sites ( L119F , I123N , A124S; Figure 4B ) displayed large shifts in rhodopsin λMAX and MII stability , with two single mutations ( L119F , A124S ) significantly increasing the stability of the active-conformation but producing opposite spectral tuning effects ( Figure 4B; Table 6 ) . Meanwhile , I123N destabilized the active-conformation almost as dramatically as E122I but produced no spectral tuning effect ( Figure 4B; Table 6 ) . This suggested that FINS substitutions at coevolving sites could functionally compensate for some of the pleiotropic effects of E122I on tetrapod rhodopsin . We created double and triple mutants representing partial replacements of the LEIA with the FINS motif , which tended to blue-shift λMAX ( Figure 4C–D; Table 6 ) . Yet , none of these intermediates were sufficient to restore WT-levels of MII stability within the COV I122 background ( Figure 4C–D; Table 6 ) . We therefore reasoned that the complete recapitulation of the FINS motif within tetrapod rhodopsin may be required for a full restoration of WT active-conformation stability . We found , incredibly , that the L119F/I123N/A124S triple mutation fully restored the MII stability of E122I tetrapod rhodopsin to WT levels , while even further blue-shifting λMAX relative to E122I ( Figure 4E; Table 6 ) . The LEIA and FINS motifs are therefore two configurations conferring convergent MII stabilities but different spectral sensitivities , with the blue-shifting I122-containing FINS motif likely also decreasing rod dark noise in vivo ( Gozem et al . , 2012; Yue et al . , 2017 ) . Our experiments demonstrate that N123 , which is not , to our knowledge , observed within any vertebrate rhodopsin other than the Characiphysi ( Tables 12 , Supplementary files 1–3 ) is nevertheless required for a complete rescue of MII stability within the LWS COV I122 background , where it has opposite functional effects depending on E vs . I122 backgrounds ( also known as sign-epistasis [Storz , 2016; Weinreich et al . , 2006] ) ( Figure 4D–E; Figure 5 ) . Structural analysis of a homology model of the MII active-state structure ( Materials and methods; Figure 5—figure supplement 1 ) suggests the conformation of the FINS motif mediates a series of context-dependent structural rearrangements promoting novel interactions ( F119 with W161; N123 with N78/T160; Figure 5—figure supplement 1 ) that can interact with existing GPCR hydrogen bond networks known to stabilize the MII active conformation ( S124 with D83-S298- N302; Figure 5—figure supplement 1; [Choe et al . , 2011] ) . These epistatic structural interactions produce correspondingly variable pleiotropic effects on RH1 spectral absorbance and MII stability ( Figure 4 ) , which were consistent with patterns of natural sequence variation at these positions across vertebrate rhodopsins . Using these patterns of naturally occurring sequence variation , we could successfully navigate a complex sequence-function landscape ( Figure 5 ) to engineer the spectral and non-spectral functions of rhodopsin simultaneously .
Did a physiological advantage related to high MII stability drive the fixation of the LEIA and FINS motifs ? Consistent with predictions of intramolecular coevolutionary theory ( Talavera et al . , 2015; Pollock et al . , 2012; Shah et al . , 2015 ) , evolutionary trajectories from the LEIA to FINS motifs must pass through sub-optimal sequence-function intermediates which include variants associated with active-state instability and human rhodopsin disease phenotypes ( e . g . A164V ) ( Stojanovic et al . , 2003 ) ( Figure 5 ) . Similar to E122I , disease variants such as A164V are likely pathogenic through disruption of the E122-H211 hydrogen bond , which has been shown to stabilize the active-state conformation but can be affected indirectly through mutations at nearby sites 119 and 123 ( Imai et al . , 1997; Stojanovic et al . , 2003; Morrow and Chang , 2015 . Although correlations between dark-state stability and active-state ( MII ) stability have been recently postulated ( Kojima et al . , 2017 ) , there exists substantial conformational differences in the TM3-TM5 region thought to stabilize both structures , including the reconfiguration of E122-H211 and E122-W126 hydrogen bonds upon light activation ( Choe et al . , 2011; Ahuja et al . , 2009; Okada et al . , 2004; Lin and Sakmar , 1996 ) . While it is unclear if such structural differences exist within the cone opsins , the lack of E122 ( e . g . Q122 in Rh2 ( except the lamprey [ ( Lin et al . , 2017; Davies et al . , 2007 ) ) , I122 in LWS ( Lamb et al . , 2007; Carleton et al . , 2005 ) ) strongly suggests that natural selection has prioritized dark-state stability over MII stability within the cone opsins , which may be related to mitigating the high noise of cone photoreceptors , especially in red-shifted LWS ( Gozem et al . , 2012; Kefalov et al . , 2003; Imai et al . , 1997; Chen et al . , 2012a; Kefalov et al . , 2005 ) . By contrast , in tetrapod rhodopsins , E122 predominates , increasing MII stability while red shifting spectral absorbance , and therefore also decreasing in vivo rod photosensitivity ( Gozem et al . , 2012; Yue et al . , 2017 ) . This suggests that selection has maintained E122 , and therefore the stability of the MII active-conformation , for reasons distinct from those maintaining the stability of the dark-state conformation , which modulates rod photosensitivity . Why has the increased rod photosensitivity conferred by COV at site 122 been sacrificed in all tetrapods ? In addition to setting the limit on rod photosensitivity ( Baylor et al . , 1980 ) , rhodopsin is also associated with light-induced photodamage ( Grimm et al . , 2000; Williams and Howell , 1983 ) , where retinal susceptibility strongly correlates with rhodopsin expression levels , and can be altered through ambient lighting conditions in some animals ( Rózanowska and Sarna , 2005; Organisciak and Vaughan , 2010 ) . Below we describe the mounting indirect evidence that the high stability and long decay of the rhodopsin MII active conformation may be a photoprotective mechanism against light-induced retinal damage ( Sommer et al . , 2014; Imai et al . , 2005 ) , and we postulate that this likely accompanied the evolution of dim-light vision . First , one of the most promising strategies to increase retinal resistance to photodamage is to slow the rate of rhodopsin regeneration , which can be achieved via mutations or molecules inhibiting the normal functioning of visual cycle proteins responsible for synthesizing 11-cis retinal ( 11CR ) ( Wenzel et al . , 2001; Saari et al . , 2001; Mandal et al . , 2011 ) . This can also be done through blocking rhodopsin regeneration and binding of 11CR ( Radu et al . , 2003; Sieving et al . , 2001 ) , reducing the light-dependent accumulation of atRAL condensation products such as diretinoid-pyridinium-ethanolamine ( A2E ) , which contributes to lipofuscin deposits in the retinal pigment epithelium associated with human retinal diseases ( Maeda et al . , 2009; Chen et al . , 2012b; Radu et al . , 2003; Sparrow , 2003 ) . Importantly , whether rhodopsin will bind available 11CR , or atRAL is dictated by the conformational selectivity of the rhodopsin dark- and active-state ( MII ) conformations , respectively ( Schafer et al . , 2016; Schafer and Farrens , 2015; Chen et al . , 2012a ) —a new finding consistent with previous observations that rhodopsins with high MII stability tend to also have slowed 11CR regeneration rates , which is a key distinguishing feature from the cone opsins ( Chen et al . , 2012a; Imai et al . , 2005 ) . These observations suggest that rhodopsin conformational selectivity may be an overlooked functional specialization of dim-light vision—one that may be associated with rates of regeneration and therefore photoprotection in the eye . Although multiple molecular mechanisms within the visual cycle appear to have evolved to prevent the accumulation of toxic atRAL ( Rózanowska and Sarna , 2005; Chen et al . , 2012c ) as well as excess 11CR ( Quazi and Molday , 2014 , a possible role for rhodopsin’s intrinsic conformational selectivity in sequestering these retinal ligands has been mostly overlooked . Recent advances in rhodopsin biochemistry suggest that such a photoprotective mechanism may indeed exist . In contrast to cones , which have an expanded retinoid recycling capacity ( Wang and Kefalov , 2011; Tsybovsky and Palczewski , 2015 ) , in rods atRAL clearance is limited by the activity of retinal dehydrogenases ( RDH ) ( Saari et al . , 1998; Chen et al . , 2012b; Chen et al . , 2012c ) , and can transiently accumulate to toxic levels ( Sommer et al . , 2014 causing light induced-retinopathy through a variety of mechanisms involving oxidative stress ( Maeda et al . , 2009; Chen et al . , 2012b . Unlike cone opsins , which have low active-conformation stability , the rhodopsin active-conformation is highly stable due in large part to the evolution of E122 ( Imai et al . , 1997 ) , and when phosphorylated and bound to rod arrestin , contains a binding affinity for atRAL sufficient for sequestration and reduction below toxic levels ( Sommer et al . , 2014; Rózanowska and Sarna , 2005 ) . Indeed , it is now known that elevated atRAL concentrations will increase atRAL re-uptake by active-conformation rhodopsin in vitro ( Schafer et al . , 2016 ) , and in bright light levels when the risk of photodamage and atRAL levels are highest ( Rózanowska and Sarna , 2005; Organisciak and Vaughan , 2010 , this process is further promoted by the constitutive binding of arrestin to the active rhodopsin conformation , which importantly does not block RDH access to atRAL ( Gurevich et al . , 2011; Sommer et al . , 2012 ) . This proposed survival mechanism not only precludes Gt signaling but also promotes re-uptake of atRAL by rhodopsin , therefore delaying MII decay and regeneration of the dark state , and enhancing atRAL sequestration ( Sommer et al . , 2012; Sommer and Farrens , 2006 ) . Although the other rhodopsin in the homodimer bound by arrestin is likely free to decay to the inactive conformation , permitting regeneration with 11CR ( Sommer et al . , 2012; Schafer and Farrens , 2015; Beyrière et al . , 2015 ) , a higher intrinsic stability of the active conformation would not only likely delay regeneration with 11CR , but also likely push the equilibrium toward atRAL re-uptake —a process that could be even further promoted if atRAL levels are high ( Sommer et al . , 2012; Schafer et al . , 2016 ) , such as within dark-adapted animals exposed to bright flashes ( Saari et al . , 1998; Lee et al . , 2010 , and within disease models where atRAL clearance is delayed ( Maeda et al . , 2009; Chen et al . , 2012c . The evolution of high conformational selectivity of active-state rhodopsin for atRAL—a distinguishing feature from the cone opsinsmay therefore play a key role within these putatively photoprotective ternary complexes , which have been previously proposed to provide an atRAL sink for rods in bright light ( Sommer et al . , 2014 ) . While only detailed experimental investigations can determine the relationships between rhodopsin regeneration rates , atRAL-associated photodamage , and the recently expanded ensemble of spectrally identical MII conformational substates ( Van Eps et al . , 2017 , a putative photoprotective role for the intrinsic stability of the rhodopsin MII active conformation would imply the presence of strong rhodopsin functional constraints in addition to those canonical constraints associated with rod photosensitivity . This model is consistent with the fact that high rod photosensitivity and susceptibility to photodamage appear to be a trade-off that accompanied the evolution of rhodopsin-mediated dim-light vision ( Grimm et al . , 2000; Williams and Howell , 198390 , 91 . Indeed , a trade-off model of rhodopsin evolution may clarify why the experimental relevance of long MII decay still remains unclear , as the focus has been mostly on mutational effects to photosensitivity , rather than photodamage ( Kojima et al . , 2014; Imai et al . , 2007 ) . Below , we outline a trade-off model of rhodopsin evolution , and describe in detail how it may help to unravel the paradoxical distribution of natural sequence variation at rhodopsin site 122 . As discussed above , tetrapod susceptibility to photodamage is a necessary side-effect of rhodopsin-mediated dim-light vision ( Grimm et al . , 2000; ) , yet it has been an often-overlooked possibility that the functional constraints governing rhodopsin evolution could have also been shaped by those associated with rhodopsin-mediated photodamage , which induces oxidative stress leading to retinal degenerative diseases via toxic atRAL ( Tsybovsky and Palczewski , 2015; Sommer et al . , 2014; Rózanowska and Sarna , 2005; Wenzel et al . , 2005 ) . By delaying both atRAL release and 11CR binding , high rhodopsin MII stability could provide an additional protective mechanism against rhodopsin-mediated photodamage outside the visual cycle—one which could be modulated parsimoniously in response to light conditions through mutations altering MII stability ( Dungan and Chang , 2017; Gutierrez et al . , 2018; Hauser et al . , 2017b . Our model would therefore predict that the evolution of rhodopsin after divergence from the cone opsins involved unique functional specializations for both photosensitivity and photoprotection . These photodamage-related constraints associated with the evolution of dim-light vision may have been especially relevant within dim-light adapted animals with high levels of rhodopsin and an increased susceptibility to photodamage ( Rózanowska and Sarna , 2005; Organisciak and Vaughan , 2010 where exposures to bright light flashes can dramatically increase toxic ATR accumulation levels , leading to photoreceptor degeneration ( Saari et al . , 1998; Chen et al . , 2012b ) . Interestingly , due to environmental differences , variation in the selective constraints associated with rhodopsin’s role in photosensitivity vs . photodamage may explain the paradoxical ecological patterns associated with natural variation at site 122 . Specifically , our results demonstrate that the only visual ecologies where the selective constraints on E122 are repeatedly relaxed across the phylogeny is within the constant darkness of deep-dwelling fish environments ( Hunt et al . , 2001; Yokoyama et al . , 1999 ) —the natural system where one might expect the fitness effects of the rhodopsin-mediated trade-off between photosensitivity and photoprotection to drastically shift , as there is likely little photodamage risk for fishes living below 1000 m in near-permanent darkness ( Denton , 1990 ) . All whales by contrast—some of which routinely dive into complete darkness at depths near 2000 m ( e . g . the sperm whale ) ( Denton , 1990; Watkins et al . , 1993 ) —strictly maintain E122 , thereby forgoing the photosensitivity increase conferred by COV at site 122 that would otherwise likely prove advantageous within these dark marine environments ( Hunt et al . , 2001; Yue et al . , 2017; Yokoyama et al . , 1999 ) . Yet , unlike deep dwelling fishes , all whales must resurface , suggesting that photoprotection-associated constraints may be maintaining E122 despite the cost of decreased photosensitivity: a prediction consistent with our model and with increases to MII stability as a key feature of whale evolution ( Dungan and Chang , 2017 . In contrast , in Characiphysi fishes , the FINS motif evolved—a novel molecular mechanism likely increasing rod photosensitivity without the consequent trade-off on MII stability . Although this may be related to mitigating the increased dark noise that can arise as consequence of spectral tuning to some red-shifted freshwater environments ( Gozem et al . , 2012; Van Nynatten et al . , 2015 ) , this remains unclear , as the ancestral condition is uncertain , and the distribution of characiphysian fishes occur in a wide range of environments ( Castiglione et al . , 2017; Chen et al . , 2013 ) . Whether other sequence motifs within this network represent different ‘tuning solutions’ for the visual system across different environments is unknown , yet the possible permutations appear to have been dramatically limited by a combination of natural selection , historical contingency and epistasis ( Tables 1–2 ) ( Storz , 2016 ) . This would be an interesting avenue of future investigation . In tetrapods , none of these coevolutionary motifs include COV at site 122 ( Tables 1 , Supplementary files 1–2 ) . This suggests that tetrapods have been confined to a local optimum ( E122 ) , which makes it tempting to speculate that this evolutionary constraint could only be maintained by the existence of a strongly detrimental pleiotropic effect , which we propose to be that of rhodopsin-mediated photodamage . Potential caveats to this theory include the existence of subterranean tetrapods maintaining E122 ( Table 1 ) —a system where one might expect the putative photodamage-associated constraints on rhodopsin to relax , as may have occurred within a variety of deep-sea fishes ( although it remains unclear if increases to photosensitivity would even be prioritized within these animals if the putative constraints on photoprotection were indeed relaxed ( Partha et al . , 2017 ) ) . Although speculative , our trade-off model of rhodopsin evolution , combined with fitness landscape theory ( Hartl , 2014 ) could potentially explain why the evolutionary trajectories between the LEIA and FINS motifs have been traversed by some freshwater fishes , but never by a tetrapod lineage . Evolutionary pathways often include compensatory mutations ( Talavera et al . , 2015; Pollock et al . , 2012; Shah et al . , 2015; Tokuriki et al . , 2008 ) , where adaptive mutations are permitted by non-adaptive neutral mutations ( Pál and Papp , 2017; Starr et al . , 2017; Tarvin et al . , 2017 ) ( also known as ‘pre-adaptations’ [Pál and Papp , 2017] , or ‘pre-adjustments’ [ ( Goldstein and Pollock , 2017 ) ) and this contingency opens new evolutionary paths by accommodating the subsequent mutational perturbations to protein activity and/or stability ( Tokuriki and Tawfik , 2009; Ivankov et al . , 2014; DePristo et al . , 2005 ) . Similarly , we find that a triple mutation ( L119F/I123N/A124S ) would be required to functionally compensate for the detrimental effects of E122I on tetrapod rhodopsin active-conformation stability . Yet , one of these 'pre-adaptations' ( I123N ) is as destabilizing as E122I itself , and displays strong sign-epistasis ( Figure 4 ) , which may be sufficient to close the evolutionary trajectory leading from the LEIA to FINS motifs ( Storz , 2016; Weinreich et al . , 2005; Poelwijk et al . , 2007 ) . Accordingly , a wide variety of indirect evolutionary paths may cut through these valleys to access fitness peaks ( Wu et al . , 2016; Starr et al . , 2017; Weinreich et al . , 2006 ) —a scenario which has not been extensively characterized in protein systems evolving within natural environments ( Pál and Papp , 2017; Hartl , 2014 ) . Detrimental intermediates and historical contingency ( Pál and Papp , 2017; Wu et al . , 2016; Starr et al . , 2017; Weinreich et al . , 2006; Palmer et al . , 2015 ) may ultimately explain why the road to the FINS motif was less travelled by in evolutionary history; although tetrapods could have in theory evolved both higher rod photosensitivity and high MII stability via the FINS motif ( Gozem et al . , 2012; Yue et al . , 2017; Imai et al . , 1997 ) , the sign-epistasis of site 123 ( Figure 4 ) may have constrained tetrapod RH1 to the local fitness optima of E122 and the LEIA motif ( Storz , 2016; Weinreich et al . , 2005; Poelwijk et al . , 2007 ) . E122 as a historical contingency may have promoted MII-mediated photoprotection within terrestrial environments and was therefore entrenched by purifying selection pressures and epistatic interactions with nearby sites ( Pollock et al . , 2012; Goldstein and Pollock , 2017; Shah et al . , 2015 ) . E122 may therefore be a ‘molecular springboard’ ( Pál and Papp , 2017 for reaching higher levels of MII stability and photoprotection that COV could not potentiate ( Dungan and Chang , 2017; Hauser et al . , 2017b; Gutierrez et al . , 2018 ) indeed , E122/S124 is nearly six-fold more stable than I122/S124 ( Figure 4 ) . By contrast , temporal and spatial variation in fish visual ecologies ( Hauser and Chang , 2017a; Bowmaker , 2008 may have opened up the indirect trajectories containing low MII stability ( Pál and Papp , 2017; Ogbunugafor et al . , 2016; Steinberg and Ostermeier , 2016 created by the epistasis of the coevolutionary network ( Wu et al . , 2016; Palmer et al . , 2015 ) , as evidenced by the fact that selection has allowed multiple fish lineages to innovate at this coevolving network through amino acid variation that may promote photosensitivity instead of MII stability ( Hunt et al . , 2001; Yue et al . , 2017; Yokoyama et al . , 1999 ) ; Table 2 , Supplementary file 2 ) . Although speculative , this implies that changing environmental constraints within the ancestral Characiphysi population ( Chen et al . , 2013 ) may have bridged the evolutionary valleys between the LEIA and FINS motifs , as has been observed in some experimental studies on bacterial enzymes mediating antibiotic resistance ( Ogbunugafor et al . , 2016; Steinberg and Ostermeier , 2016 ) . The specific environmental differences that may be responsible for opening and closing these alternative evolutionary trajectories within fishes remain to be identified and would be an interesting subject of future investigation . It is important to note that although the sequence-function landscape of site 122 is likely more complex than what we demonstrated here , recent studies from ours and other groups have begun unravelling important epistatic interactions among residues at four-site motifs ( Wu et al . , 2016; Starr et al . , 2017; Tarvin et al . , 2017 ) . We therefore present a powerful integrative approach for the exploration of inferred fitness landscapes using natural variation . This has generated multiple insights . First , our results strongly suggest that E122 was not necessary for the evolution of high MII stability , and therefore expands on previous work demonstrating site 122 as an evolutionary dynamic determinant of visual pigment spectral and non-spectral properties ( Hunt et al . , 2001; Yue et al . , 2017; Imai et al . , 1997; Yokoyama et al . , 1999; Imai et al . , 2007 ) . Second , this further argues that novel sequence-function solutions in proteins ( McMurrough et al . , 2014; Mateu and Fersht , 1999; Tarvin et al . , 2017 ) can be discovered by integrating genetic and ecological information to reveal ancient evolutionary trajectories ( Liebeskind et al . , 2015; McTavish et al . , 2013 ) . These evolutionary solutions may be otherwise unpredictable from biophysical perspectives ( Starr and Thornton , 2017; Sailer and Harms , 2017; Otwinowski and Plotkin , 2014 ) where the accuracy of computational models remains limited to describing changes in protein stability ( Goldenzweig and Fleishman , 2018; Echave et al . , 2016; Goldstein and Pollock , 2017 ) , rather than the adaptive shifts in protein function and trade-offs with stability—a scenario likely widespread in natural systems ( Pál and Papp , 2017; McMurrough et al . , 2014; Mateu and Fersht , 1999; Tarvin et al . , 2017; DePristo et al . , 2005 ) . Finally , our work suggests that even within biological systems as complex as that of animal vision , the existence of novel biophysical and ecological constraints can still be elucidated through comparative analyses of natural variation .
Rhodopsin-coding sequences ( rh1 ) originating from Teleost fishes , Tetrapods , and other vertebrate outgroups ( Supplementary file 1–3 ) were obtained from GenBank using BlastPhyMe ( Schott et al . , 2016b ) . Teleost fish rh1 sequences were sampled from all available phylogenetic orders denoted in Betancur et al . , 2013 . Tetrapod rh1 sequences were sampled from all major phylogenetic groupings ( Figure 1—figure supplement 1 ) ( Hedges et al . , 2015; Foley et al . , 2016; Prum et al . , 2015; Amemiya et al . , 2013 , as described previously ( Hauser et al . , 2016 ) . Rh1 alignments were generated using PRANK codon alignment ( Löytynoja and Goldman , 2008 . The final rh1 alignment encoded for rhodopsin amino acid residues 42 – 307 ( bovine RH1 numbering ) , inclusively , where for mutual information analyses gaps were trimmed from the beginning and end of the alignment , resulting in a shorter alignment ( residues 53– 302 ) . In both instances , the alignments used for bioinformatic analysis encompassed the entire seven-transmembrane domain of rhodopsin . Using this alignment , we constructed three separate rh1 datasets for phylogenetic analysis: ( 1 ) Tetrapods ( n = 86; Supplementary file 1 ) ; ( 2 ) Teleost fishes ( n = 119; Supplementary file 2 ) ; ( 3 ) Vertebrate ( n = 209 ) which included ( 1 ) and ( 2 ) in addition to outgroups . For each dataset , a species tree was constructed by reference to established relationships for Tetrapods ( Hedges et al . , 2015; Foley et al . , 2016; Prum et al . , 2015; Amemiya et al . , 2013 ) and Teleosts ( Betancur et al . , 2013 ) . The Vertebrate phylogeny was assembled by adding non-tetrapod Sarcopterygian outgroups to the Tetrapod phylogeny , combining this with the Teleost phylogeny , and then adding cartilaginous fish outgroups , all according to species relationships ( Figure 1—figure supplement 2 ) ( Betancur et al . , 2013; Amemiya et al . , 2013122 , 125 . These phylogenies were used in subsequent computational analyses . We also constructed a Characiphysi rh1 dataset , representing wide phylogenetic sampling ( Supplementary File 3 ) , with an rh1 alignment encoding for residues 42– 307 as that described above , and where a species tree ( Figure 2—figure supplement 1 ) was constructed by reference to established relationships ( Chen et al . , 2013 and references therein ) . We took a multifaceted approach toward detecting sites coevolving with site 122 , corroborating our phylogenetic tests of evolutionary rates ( dN/dS ) ( Yang , 2007 ) with phylogenetically corrected statistical tests of amino acid covariation ( Simonetti et al . , 2013 ) , and phylogenetic analyses of correlated evolutionary patterns in amino acid substitutions ( Pagel , 1994 . We used these three approaches to search for evidence of coevolution between rhodopsin site 122 and sites within a 6 Å radius within the MII active-conformation crystal structure ( Choe et al . , 2011 ) . We used codon models of molecular evolution from the PAML 4 . 7 software package ( Yang , 2007 ) to identify evidence of increased purifying selection in rhodopsin-coding sequences ( rh1 ) . First , we estimated the evolutionary rates ( dN/dS ) within each rh1 dataset ( Teleosts , Tetrapods , Vertebrates , Characiphysi ) using the random sites models ( M1 , M2 , M3 , M7 , M8 ) implemented in the CODEML program . This required pruning the outgroups from the Teleost and Tetrapod datasets . Site-specific evolutionary rates were obtained from M8 , which was the best fitting model in each dataset as assessed by differences in Akaike information criterion ( Tables 7–10 ) . Next , we employed PAML Clade models ( Bielawski and Yang , 2004 to explicitly test for long-term shifts in evolutionary rates ( dN/dS ) between foreground and background branches or clades within the rhodopsin datasets . In any partitioning scheme , all non-foreground data are present in the background partition . The foreground partition is listed after the underscore for the clade models ( e . g . CmC_foreground ) . CmC analyses tested for long-term shifts in purifying selection between: tetrapod and teleost clades within the Vertebrate dataset ( Table 3 ) ; the branch leading to the tetrapod clade within the Vertebrate dataset ( Table 3 ) ; and the Characiphysi clade and the branch leading to the clade within the Teleost dataset ( Table 5 ) . M2aREL was used as the null model ( Weadick and Chang , 2012 ) . For all PAML models , multiple runs with different starting priors were carried out to check for the convergence of parameter estimates . Significant differences in model fits we determined by likelihood ratio-tests . Statistical tests of covariation ( e . g . Mutual Information; MI ) are an approximate measure for identifying coevolving sites in alignments of homologous protein families , but can have high false-positive rates due to sampling bias and random background effects ( Ashenberg and Laub , 2013; Talavera et al . , 2015 ) , especially if there is a lack of phylogenetic correction ( Simonetti et al . , 2013; Dunn et al . , 2008 ) . Nevertheless , MI methods appear able to detect sites of functional importance that are close in proximity to each other ( Ashenberg and Laub , 2013; Talavera et al . , 2015 ) . Given all these factors , we decided to employ MI analyses within our dataset only as a qualitative guide to provide additional insight into the putative coevolutionary dynamics within Vertebrate RH1 , and to potentially corroborate our molecular evolution analyses since overlap between evolutionary rates and statistical covariation of amino acids has been described in detail ( Talavera et al . , 2015 . Since MI is usually employed within large protein family datasets , rather than intrafamily comparisons ( Ashenberg and Laub , 2013; Talavera et al . , 2015 ) we subjected phylogenetically corrected MI z-scores ( MISTIC; [ ( Simonetti et al . , 2013 ) ) to a significance threshold representing the top absolute z-score from all pairwise comparisons from across analyses of randomized datasets ( n = 150 ) , as previously described ( Ashenberg and Laub , 2013 . These MI calculations were conducted using MISTIC on the Teleost and Tetrapod RH1 amino acid alignments , separately , and phylogenetically corrected MI z-scores were reported for sites within a 6 Å radius of site 122 ( Table 4 ) . Lastly , to further corroborate our dN/dS analyses we investigated for evidence of correlated evolution between site 122 amino acid variation and variation at other sites within a 6 Å radius . This was done using an amino acid alignment of Teleost RH1 only; Tetrapod RH1 was not analyzed since site 122 is invariant . Consensus amino acid residues were determined for each site that fell within the 6 Å radius , where a consensus residue at a given position within a given taxa was represented as a ‘0’ , whereas a natural variant was numbered as ‘1’ . A phylogenetic method ( Pagel , 1994 ) was then used to test for correlated evolution in amino acid variation between a given site within a 6 Å radius of site 122 . The Teleost species phylogeny described above was used for these analyses within the MESQUITE software package ( Maddison and Maddison , 2017 , where p-values were calculated by performing Monte Carlo tests using data from simulations ( n > 1000 ) as previously described ( Pagel , 1994 ) . Significance was determined using p-values subjected to a Bonferroni-correction for multiple testing ( Table 4 ) . To reconstruct the evolutionary history of sites 119 , 122 , 123 and 124 at the origin of both Tetrapods and the Characiphysi , we used the Vertebrate rh1 alignment and phylogeny described above . This dataset was then used to implement codon-based marginal ancestral sequence reconstructions using the PAML 4 . 7 software package ( Yang , 2007 ) . Ancestral sequences were chosen from the best-fitting random sites model , which was M8 ( Table 7 ) . The likelihood-based reconstruction uses branch lengths and relative substitution rates between nucleotides , followed by empirical Bayesian reconstruction of ancestral codon states at ancestral nodes , where uncertainty is measured as posterior probabilities ( Yang , 2006 . To identify ancestral codons at the ancestral nodes ( Figure 2 ) , we consulted the full posterior probability distribution from the marginal reconstruction , where the character with the highest posterior probability is the best reconstruction ( Yang , 2006 . We verified the complete conservation of F119/I122/N123/S124 in Characiphysi RH1 by reference to an expanded Characiphysi RH1 amino acid alignment we assembled using a wide phylogenetic sampling of publicly available rh1 sequences ( Supplementary file 3 ) . The complete coding sequence of bovine ( Bos taurus ) rhodopsin in the pJET1 . 2 cloning vector ( ThermoFisher Scientfic ) , as described in a previous study was used here ( Castiglione et al . , 2017 ) . Site-directed mutagenesis primers were designed to induce single amino acid substitutions via PCR ( QuickChange II , Agilent ) . All sequences were verified using a 3730 DNA Analyzer ( Applied Biosystems ) at the Centre for Analysis of Genome Evolution and Function ( CAGEF ) at the University of Toronto . Wild type and mutant rhodopsin sequences were transferred to the pIRES-hrGFP II expression vector ( Stratagene ) for subsequent transient transfection of HEK293T cells ( 8 µg per 10 cm plate ) using Lipofectamine 2000 ( Invitrogen ) . HEK293T cells were obtained from David Hampson ( University of Toronto ) , were authenticated by STR profiling ( Centre for Applied Genomics , The Hospital for Sick Children ) and tested negative for mycoplasma contamination . Media was changed after 24 hr , and cells were harvested 48 hr post-transfection . Cells were washed twice with harvesting buffer ( PBS , 10 µg/mL aprotinin , 10 µg/mL leupeptin ) , and rhodopsins were regenerated for 2 hr in the dark with 5 µM 11-cis-retinal generously provided by Dr . Rosalie Crouch ( Medical University of South Carolina ) . After regeneration the samples were incubated at 4°C in solubilisation buffer ( 50 mM Tris pH 6 . 8 , 100 mM NaCl , 1 mM CaCl2 , 1% dodecylmaltoside , 0 . 1 mM PMSF ) for 2 hr and immunoaffinity purified overnight using the 1D4 monoclonal antibody coupled to the UltraLink Hydrazide Resin ( ThermoFisher Scientific ) . Resin was washed three times with wash buffer 1 ( 50 mM Tris pH 7 . 0 , 100 mM NaCl , 0 . 1% dodecylmaltoside ) and twice using wash buffer 2 ( 50 mM sodium phosphate , 0 . 1% dodecylmaltoside; pH 7 . 0 ) . Rhodopsins were eluted from the UltraLink resin using 5 mg/mL of a 1D4 peptide , consisting of the last nine amino acids of bovine rhodopsin ( TETSQVAPA ) . The UV-visible absorption spectra of purified rhodopsin samples ( Figure 4—figure supplement 1 ) were recorded in the dark at 25°C using a Cary 4000 double-beam absorbance spectrophotometer ( Agilent ) . All λMAX values were determined by fitting dark spectra to a standard template curve for A1 visual pigments ( Govardovskii et al . , 2000 . Rhodopsin samples were light-activated for 30 s using a fiber optic lamp ( Dolan-Jenner ) , resulting in a shift in λMAX to ~380 nm , characteristic of the biologically active metarhodopsin II intermediate ( Van Eps et al . , 2017 ) . Retinal release following rhodopsin photoactivation was monitored using a Cary Eclipse fluorescence spectrophotometer equipped with a Xenon flash lamp ( Agilent ) , according to a protocol modified from previous studies ( Schafer et al . , 2016; Farrens and Khorana , 1995 ) . Rhodopsin samples ( 0 . 1 – 0 . 2 μM ) were bleached for 30 s at 20°C with a fiber optic lamp ( Dolan-Jenner ) using a filter to restrict wavelengths of light below 475 nm to minimize heat . Fluorescence measurements were recorded at 30 s intervals with a 2 s integration time , using an excitation wavelength of 295 nm ( 1 . 5 nm slit width ) and an emission wavelength of 330 nm ( 10 nm slit width ) . There was no noticeable activation by the excitation beam prior to rhodopsin activation . This assay detected increasing fluorescence as a result of decreased quenching of intrinsic tryptophan fluorescence at W265 by the retinal chromophore ( Farrens and Khorana , 1995 , and is a reliable proxy for the tracking the decay of MII ( Schafer et al . , 2016 ) . Data was fit to a three variable , first-order exponential equation ( y = y0+a ( 1-e-bx ) ) , and half-life values were calculated using the rate constant b ( t1/2 = ln2/b ) . All curve fittings resulted in r2 values greater than 0 . 95 . Differences in retinal release half-life values were statistically assessed using a two-tailed t test with unequal variance . To better evaluate the potential for natural variants at sites 119 , 122 , 123 and 124 to disrupt nearby structural motifs of rhodopsin , the L119F/E122I/I123N/A124S quadruple mutant structure was computationally estimated from the 3D structure of MII ( PDB code: 3PQR ) ( Choe et al . , 2011 . A 3D structure of MII with all-trans-retinal bound was inferred via homology modelling by MODELLER ( Sali and Blundell , 1993; Eswar et al . , 2006133 , 134 ) . Minimizing the MODELLER objective function generated 100 separate models , and the run with the lowest discrete optimized protein energy ( DOPE ) score was assessed ( Shen and Sali , 2006 ) , with reference to the next four best fitting models serving as validation of structural changes . For each estimated structure , ProCheck was used to verify the high probability of bond angle and length stereochemical conformations , as indicated by positive overall G-factor ( Laskowski et al . , 1993 . Comparisons of each model’s total energy to that expected by random chance were examined using ProSA-web ( Wiederstein and Sippl , 2007 ) . Images of 3D structures were generated using the PyMOL molecular graphics system , version 1 . 3 ( Schrödinger , LLC ) . | People can see in dim light because of cells at the back of the eye known as rods . These cells contain two key components: molecules called retinal , which are bound to proteins called rhodopsin . When light hits a rod cell , it kicks off a cascade of reactions beginning with the retinal molecule changing into an activated shape and ending with a nerve impulse travelling to the brain . The activated form of retinal is toxic , and as long as it remains bound to the rhodopsin protein it will not damage the rod or surrounding cells . The toxic retinal also cannot respond to light . It must be released from the protein and converted back to its original shape to restore dim light vision . As with all proteins , rhodopsin’s structure comprises a chain of building blocks called amino acids . Every land animal with a backbone has the same amino acid at position 122 in its rhodopsin . This amino acid , named E122 , helps to stabilize the activated rhodopsin , slowing the release of the toxic retinal . Yet E122 also makes the rod cells less sensitive , resulting in poorer vision in dim light . In contrast , some fish do not have E122 but rather one of several different amino acids takes its place . What remains unclear is why all land animals have stuck with E122 , and whether there were other options that evolution could have explored to overcome the trade-off between sensitivity and stability . By looking at the make-up of rhodopsins from many animals , Castiglione and Chang found other sites in the protein where the amino acid changed whenever position 122 changed . The amino acids at these so-called “coevolving sites” were then swapped into the version of rhodopsin that is found in cows , which had also been engineered to lack E122 . These changes fully compensated for the destabilizing loss of E122 on activated rhodopsin but without sacrificing its sensitivity to light . Further experiments then confirmed that unless all amino acids were substituted at once , the activated rhodopsin was very unstable . Indeed , it was almost as unstable as mutated rhodopsins found in some human diseases . These findings suggest that , while there was in principle another solution available to land animals , the routes to it were closed off because they all came with an increased risk of eye disease . These findings highlight that rhodopsin likely plays a more important role in protecting humans and many other land animals against eye disease than previously assumed . More knowledge about this protective role may lead to new therapies for these conditions . Also , investigating similar evolutionary trade-offs could help to explain how and why different proteins work the way that they do today . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"biochemistry",
"and",
"chemical",
"biology"
] | 2018 | Functional trade-offs and environmental variation shaped ancient trajectories in the evolution of dim-light vision |
Nuclear entry of HIV-1 replication complexes through intact nuclear pore complexes is critical for successful infection . The host protein cleavage-and-polyadenylation-specificity-factor-6 ( CPSF6 ) has been implicated in different stages of early HIV-1 replication . Applying quantitative microscopy of HIV-1 reverse-transcription and pre-integration-complexes ( RTC/PIC ) , we show that CPSF6 is strongly recruited to nuclear replication complexes but absent from cytoplasmic RTC/PIC in primary human macrophages . Depletion of CPSF6 or lack of CPSF6 binding led to accumulation of HIV-1 subviral complexes at the nuclear envelope of macrophages and reduced infectivity . Two-color stimulated-emission-depletion microscopy indicated that under these circumstances HIV-1 complexes are retained inside the nuclear pore and undergo CA-multimer dependent CPSF6 clustering adjacent to the nuclear basket . We propose that nuclear entry of HIV-1 subviral complexes in macrophages is mediated by consecutive binding of Nup153 and CPSF6 to the hexameric CA lattice .
Reverse transcription of the viral single-stranded RNA genome into double-stranded cDNA , followed by integration into the host genome , are defining features of retrovirus replication ( Hu and Hughes , 2012 ) . These events occur in the cytoplasm and nucleus of the newly infected cell within ill-defined nucleoprotein complexes , termed reverse transcription ( RTC ) and pre-integration complex ( PIC ) , respectively . Accumulating evidence indicates that the viral capsid structure plays an essential role in early replication ( Campbell and Hope , 2015; Yamashita and Engelman , 2017 ) . Mutations in the HIV-1 CA ( capsid ) protein have been shown to cause defects in post-entry stages , and various cellular proteins binding CA exhibit positive or negative effects on early HIV-1 replication ( Hilditch and Towers , 2014; Ambrose and Aiken , 2014; Campbell and Hope , 2015; Yamashita and Engelman , 2017 ) . Accordingly , CA and/or the structure of the viral capsid have been implicated in reverse transcription , cytoplasmic trafficking , evasion from the cell-autonomous immune response and nuclear entry ( Ambrose and Aiken , 2014; Hilditch and Towers , 2014; Yamashita and Engelman , 2017; Campbell and Hope , 2015 ) . Neither the exact composition of the HIV-1 RTC/PIC nor the functional role of host cell dependency or restriction factors that interact with CA during early replication is currently well understood . Furthermore , the time of capsid uncoating during early HIV-1 replication is not well established , and may differ depending on the target cell type ( Arhel et al . , 2007; Zhou et al . , 2011; Hulme et al . , 2015; Chen et al . , 2016 ) . Individual viruses entering a host cell may follow different - productive and non-productive – pathways , and it is crucial to identify productive phenotypes correlating with infection . It is generally accepted that incoming capsids remain intact after cytoplasmic entry and during the initial stages of reverse transcription , while different reports suggest uncoating in the cytoplasm or at the nuclear pore complex ( NPC ) ( e . g . Xu et al . , 2013; Mamede et al . , 2017; Francis and Melikyan , 2018 ) . Nuclear HIV-1 PIC have been observed to contain at least some CA ( Zhou et al . , 2011; Peng et al . , 2014; Hulme et al . , 2015; Chin et al . , 2015; Chen et al . , 2016; Stultz et al . , 2017 ) , but the amount of residual CA and whether it maintains the lattice structure are unknown . Furthermore , the size of the conical viral capsid ( ~60 nm at the broad end ( Briggs et al . , 2003 ) ) is larger than the width of the NPC transport channel ( Beck et al . , 2004 ) , indicating that ( partial ) capsid disintegration or at least capsid or NPC remodeling is needed for HIV-1 nuclear translocation in non-dividing cells . Cleavage and polyadenylation specificity factor 6 ( CPSF6 ) was initially identified as an HIV-1 restriction factor when exogenously expressed in a truncated form . The truncated protein is enriched in the cytoplasm , interacts with HIV-1 CA and impairs HIV-1 replication prior to nuclear import ( Lee et al . , 2010 ) . In contrast , neither overexpression of full length CPSF6 , which is almost exclusively nuclear , nor knock-down of CPSF6 significantly affected HIV-1 infectivity in cell lines ( Lee et al . , 2010; Hori et al . , 2013; Fricke et al . , 2013; De Iaco et al . , 2013 ) . Structural analyses identified a CPSF6-interacting interface and an overlapping site interacting with the nucleoporin Nup153 on the hexameric form of CA , the basic building block of the viral capsid ( Price et al . , 2012; Price et al . , 2014; Bhattacharya et al . , 2014 ) . An HIV-1 derivative carrying a point mutation within this CPSF6-binding motif in CA ( N74D ) exhibited impaired CPSF6 binding in vitro and escaped restriction by truncated CPSF6 , but displayed full infectivity in reporter cell lines ( Lee et al . , 2010; Schaller et al . , 2011; Ambrose et al . , 2012 ) . In contrast , both CPSF6 knock-down and the N74D exchange impaired HIV-1 infection in post-mitotic primary human macrophages ( Schaller et al . , 2011; Ambrose et al . , 2012 ) , and this effect was attributed to induction of an interferon response triggered by viral DNA sensing ( Rasaiyaah et al . , 2013 ) . Based on these results , it was speculated that cytoplasmic CPSF6 binds incoming viral capsids and blocks reverse transcription during cytoplasmic transport to prevent recognition of newly synthesized cDNA by cytoplasmic DNA sensors ( Rasaiyaah et al . , 2013 ) . More recently , CPSF6 was suggested to affect HIV-1 nuclear entry in a HeLa-based reporter cell line ( Chin et al . , 2015 ) and to be an important factor in targeting HIV-1 integration in infected primary CD4+ T-cells and macrophages ( Sowd et al . , 2016; Rasheedi et al . , 2016; Achuthan et al . , 2018 ) . Strikingly , a later study described another point mutation in the CPSF6-binding interface of HIV-1 CA ( A77V ) , which also abolished interaction with CPSF6 but did not appear to affect replication in primary cells ( i . e . primary human macrophages and CD4+ T-cells ) , although it was negatively selected after passaging in vivo ( Saito et al . , 2016b ) . Taken together , it is clear that CPSF6 has an important , CA dependent function in early HIV-1 infection of macrophages , which may not be fully recapitulated in HeLa- or 293T-based reporter cell lines . This may reflect differences between cell lines and primary cells , but also between different cell types . The mechanism ( s ) of action of CPSF6 is not fully defined , and the relative contribution of cytoplasmic versus nuclear CPSF6 as well as the reason for the apparent cell-type specific differences are not resolved . We have previously established a microscopy-based approach for quantitative analysis of HIV-1 post entry events on a single particle level ( Peng et al . , 2014 ) . Labeling nascent RT products by incorporation of the thymidine analog EdU followed by click labeling , in conjunction with fluorescent labeling of the bona fide RTC/PIC component IN , identified reverse transcription competent HIV-1 RTC/PIC in the cytoplasm and nucleus of infected cells and enabled direct visualization of viral and cellular proteins associated with these complexes . Employing this system to investigate CPSF6 recruitment , we had observed weak or no CPSF6 signals on cytosolic RTC/PIC in model cell lines; pronounced-co-localization was only observed when transportin 3 ( TNPO3 ) , which is needed for CPSF6 nuclear import , was depleted ( Peng et al . , 2014 ) . We have now used this approach for a detailed analysis of CPSF6 recruitment and its role for HIV-1 nuclear import in primary human monocyte-derived macrophages ( MDM ) . CPSF6 was strongly enriched on nuclear complexes , and depletion of CPSF6 or the A77V mutation in CA reduced HIV-1 infectivity in MDM . RTC/PIC accumulated close to the nuclear envelope in these cases . Two-color Stimulated Emission Depletion ( STED ) microscopy revealed that CA-containing HIV-1 complexes directly co-localized with NPCs , and CPSF6 was associated with the nuclear basket at these sites in a CA-dependent manner . These results indicate that CPSF6 facilitates nuclear entry of HIV-1 in post-mitotic human macrophages in a CA–dependent manner at the level of the NPC .
The poor association of cytoplasmic RTC/PIC with CPSF6 observed in our previous study ( Peng et al . , 2014 ) argued against the model that CPSF6 regulates viral reverse transcription during cytoplasmic trafficking ( Rasaiyaah et al . , 2013 ) . Our experimental system allowed us to directly address this issue by correlating the presence of CPSF6 on cytosolic RTC/PIC with the EdU/click signal intensity as a measure of reverse transcription products . These experiments were performed in a HeLa-derived TNPO3 knock-down cell line which displays a high cytosolic level of CPSF6 ( Thys et al . , 2011 ) . Cells were infected with HIV-1 carrying IN . eGFP as RTC/PIC marker , subjected to EdU incorporation , and fixed and click-labeled 4 . 5 hr post infection . IN . eGFP/EdU positive objects were classified according to whether or not they were associated with CPSF6 immunofluorescence . In accordance with our previous results ( Peng et al . , 2014 ) , the majority of cytoplasmic RTC/PIC ( 95/121; 78 . 5% ) was positive for CPSF6 in this cell line with high cytoplasmic CPSF6 levels ( Figure 1—figure supplement 1A ) . EdU signal intensities on individual CPSF6-positive complexes were found to be significantly higher on average compared to those on CPSF6-negative , but IN . eGFP-positive objects ( Figure 1—figure supplement 1B ) , implying that CPSF6 association with cytoplasmic RTC/PIC did not inhibit reverse transcription . To analyze further whether CPSF6 affects reverse transcription , CPSF6 was depleted in MDM followed by infection and RTC/PIC quantification . The number of reverse-transcription competent RTC/PIC ( i . e . EdU positive signals co-localizing with IN . eGFP ) in CPSF6 depleted MDM was comparable with that of control cells ( Figure 1—figure supplement 1C ) . These results are inconsistent with an inhibitory effect of cytoplasmic CPSF6 on HIV-1 reverse transcription . We therefore focused on the characterization of CPSF6 in the nucleus and at the NPC in primary human MDM in the following experiments . Reverse transcription of HIV-1 in MDM is much slower than in reporter cell lines or activated T-cells , presumably due to the low dNTP levels in these non-dividing cells ( Diamond et al . , 2004 ) . We therefore performed inhibitor time-of-addition experiments to define the appropriate time window for RTC/PIC detection in this cell type . MDM were prepared from healthy blood donors and infected with HIV-1 carrying an R5-tropic Env protein derived from a primary human isolate ( Schnell et al . , 2011 ) . The non-nucleosidic reverse transcriptase ( RT ) inhibitor efavirenz ( EFV ) or DMSO was added at different time points post single-round infection ( p . i . ) , and the percentage of infected cells was determined by CA immunostaining 6d p . i . . Only minor resilience against EFV inhibition was observed at 24 h p . i . and this increased until 72 h p . i . , where EFV inhibition was lost ( Figure 1—figure supplement 2A , B ) . These results confirmed that completion of reverse transcription occurs late in MDM ( supporting a recent more detailed analysis of replication dynamics ( Bejarano et al . , 2018 ) ) and defined the time window for detection of HIV-1 RTC/PIC . MDM were infected with R5-tropic HIV-1 carrying IN . eGFP followed by detection of RTC/PIC via click-labeling of EdU incorporated into nascent viral DNA . The nuclear envelope was visualized by immunostaining of lamin A/C . At 48 h p . i . , mock-infected cells displayed diffuse , predominantly nuclear localization of CPSF6 ( Figure 1A ) , as previously reported ( Dettwiler et al . , 2004 ) . In contrast , CPSF6 was strongly enriched on almost all nuclear HIV-1 complexes ( 75/77; 97 . 4% ) in infected MDM , and strong punctate CPSF6 signals co-localizing with nuclear HIV-1 complexes were easily detected above the diffuse nuclear background ( Figure 1B ) . Since the signal intensities of HIV-1 associated CPSF6 punctae were clearly much higher than those of the small nuclear CPSF6 speckles observed in both non-infected and infected cells , we consider it likely that this reflects CPSF6 recruitment by subviral complexes rather than recruitment of HIV-1 derived structures to pre-existing CPSF6 clusters . CPSF6 signal intensities were much higher on nuclear HIV-1 subviral complexes compared to those on complexes localized near the nuclear envelope . CPSF6 levels on the latter structures were generally close to the detection level , and cytoplasmic RTC/PIC were almost always CPSF6-negative ( Figure 1F ) . Interestingly , cells transduced with lentiviral vectors expressing a non-targeted shRNA exhibited similar characteristic CPSF6 clusters in the nucleus ( Figure 1—figure supplement 2C ) , while this was not observed in mock-transduced cells . We attribute these signals to nuclear complexes of the lentiviral vector . These results clearly show that CPSF6 association with RTC/PIC occurs mainly in the nucleus in MDM and leads to strong CPSF6 clustering on the subviral complex . To assess the degree of CA retention on nuclear subviral complexes , MDM were infected with HIV-1 for 48 hr as above and subjected to CA immunostaining . In agreement with our previous findings ( Peng et al . , 2014 ) , nuclear HIV-1 RTC/PIC were strongly CA-positive , with CA signals clearly co-localizing with EdU , IN . eGFP and CPSF6 ( Figure 1C ) . Quantifying localization and CPSF6/CA association of HIV-1 complexes in a total of 92 infected MDM at 48 h p . i . , we observed a large majority of IN . eGFP positive structures outside the nucleus ( 7 , 904/8 , 403; 94% ) . These structures mostly lacked detectable EdU signals ( 7558/7904; 95 . 6% ) , in agreement with the assumption that the majority of particles in the cytosolic area represent virions taken up into endosomes ( inaccessible to EdUTP ) and non-productive particles . In contrast , the majority of IN . eGFP positive objects within the nucleus were EdU positive ( 328/499; 65 . 7% ) . The observation that a subset of complexes detected in the nucleus was EdU negative confirms that reverse transcription is not a prerequisite for nuclear import of HIV-1 complexes , as reported by ( Burdick et al . , 2017 ) . In a recent study on HIV-1 replication dynamics in macrophages , we observed higher EdU intensities associated with nuclear complexes compared to complexes near the nuclear envelope , however , suggesting that reverse transcription may promote efficiency of nuclear entry or that reverse transcription can even be completed in the nucleus in this cell type ( Bejarano et al . , 2018 ) . 325 of the 328 EdU-positive complexes in the nucleus ( 99% ) were associated with detectable CA signals , and a similar CA signal was also observed on 85% of the EdU-negative nuclear IN . eGFP-positive complexes ( 146/171 ) . Mean CA signal intensity of individual RTC/PIC differed only modestly between cytoplasmic and nuclear CA-positive HIV-1 structures , suggesting that nuclear RTC/PIC in HIV-1 infected MDM retain the majority of CA molecules . Interestingly , the mean CA signal intensity associated with RTC/PIC close to the nuclear envelope was lower than observed for either cytoplasmic or nuclear RTC/PIC ( Figure 1G ) . To test whether recruitment of CPSF6 to HIV-1 nuclear complexes affects CA retention , MDM were subjected to CPSF6 knock-down prior to HIV-1 infection . CPSF6 knock-down had no apparent effect on CA signals associated with HIV-1 RTC/PIC independent of the subcellular localization ( Figure 1—figure supplement 3A ) . Comparable results were obtained upon infection of MDM with CPSF6 binding-defective HIV-1 carrying the A77V exchange in CA ( Figure 1—figure supplement 3B ) . To test whether CPSF6 enrichment on nuclear subviral complexes requires HIV-1 reverse transcription or integration , MDM were infected and labeled in the presence of RT or IN inhibitors , respectively . Nuclear import of HIV-1 subviral complexes and CPSF6 recruitment to these structures was independent of reverse transcription ( Figure 1D; 102/105 nuclear complexes were CPSF6-positive; 97 . 1% ) and integration ( Figure 1E: 101/104 nuclear complexes were CPSF6-positive; 97 . 1% ) . To verify that the CPSF6-enriched nuclear structures contain reverse-transcribed HIV-1 genomes , we performed immunofluorescence combined with DNA in situ hybridization ( immuno-FISH ) on infected MDM at 72 h p . i . . Co-localization of viral DNA signals with CA immunostaining and CPSF6-enrichment ( Figure 2A ) confirmed that the EdU signal on nuclear complexes corresponds to HIV-1 DNA and ascertained that the identified objects represent HIV-1 replication complexes that have undergone reverse transcription . We then investigated whether a specific CPSF6 complex is recruited to nuclear HIV-1 RTC/PIC . CPSF6 is part of the cellular CF Im complex involved in pre-mRNA processing ( Gruber et al . , 2012; Hardy and Norbury , 2016 ) . Two forms of this tetrameric complex have been described . They are composed of two CPSF5 molecules , associated with two molecules of either CPSF6 or CPSF7 ( Gruber et al . , 2012 ) . MDM were infected with HIV-1 for 48 hr and co-immunostained with antibodies against CPSF6 and CPSF5 , or against CPSF7 and CPSF5 . We observed clear co-localization of CPSF5 and CPSF6 on almost all nuclear complexes analyzed ( 42/44 ) , while CPSF7 was only detected on one of 82 nuclear complexes analyzed ( Figure 2B ) . Thus , we conclude that the CPSF52-CPSF62 form of the CF Im complex is recruited to nuclear HIV-1 complexes . Given the role of the CF Im complex in cellular pre-mRNA processing , we investigated whether CPSF6 accumulation on nuclear HIV-1 complexes depends on their transcriptional activity . To this end , MDM were infected with a transcriptionally impaired HIV-1 variant carrying a deletion in the viral Tat transactivator ( HIV-1NL4-3ΔTat ) . Alternatively , wild-type HIV-1 infected MDM were treated at 96 h p . i . with the transcription inhibitor flavopiridol for 12 hr . HIV-1 transcription and infectivity were strongly impaired by either , Tat deletion or flavopiridol treatment ( Figure 2—figure supplement 1A and B ) . Impairment of HIV-1 RNA transcription did not affect CPSF6 recruitment to nuclear HIV-1 complexes , however . CPSF6 enrichment on nuclear replication complexes of HIV-1NL4-3ΔTat infected MDM ( 54/58; 93%; Figure 2C ) was comparable with wild-type HIV-1 , and this was also true for flavopiridol treatment ( 63/64; 98%; Figure 2D ) . CPSF6 recruitment to transcriptionally inactive nuclear HIV-1 complexes was confirmed by immunostaining in combination with RNA FISH at 108 h p . i . . This analysis revealed that CPSF6-positive nuclear punctae in wild-type HIV-1 infected MDM mostly co-localized with viral RNA signals ( 50/69; 72 . 5%; Figure 2E ) , indicating some transcriptional activity . On the other hand , nuclear CPSF6 punctae in MDM infected with the Tat-defective HIV-1 variant ( 72/98; 74%; Figure 2F ) or in the presence of Flavopiridol ( 29/34; 85%; Figure 2G ) mostly lacked detectable FISH signals for viral RNA . These results indicate that recruitment of CPSF6 to HIV-1 nuclear replication complexes does not require active transcription of the HIV-1 genome and does thus not result from CPSF6 function in pre-mRNA processing . The p75 isoform of the host cell factor Lens Epithelium-Derived Growth Factor ( LEDGF/p75 ) has been reported to be important for HIV-1 integration and to influence integration site selection ( Kvaratskhelia et al . , 2014; Debyser et al . , 2015 ) . We therefore analyzed nuclear HIV-1 complexes for the presence of LEDGF . Co-immunostaining of HIV-1 infected MDM with antibodies against LEDGF ( detecting both the p75 and p52 isoform ) and CPSF6 revealed all nuclear complexes to be CPSF6-positive in this experiment and ~80% to contain detectable LEDGF as well ( Figure 2—figure supplement 1C ) . In order to determine whether enrichment of CPSF6 on nuclear HIV-1 complexes is functionally relevant for HIV-1 replication , we performed CPSF6 knock-down experiments and employed a CPSF6 binding deficient virus . Mutation A77V in CA has previously been reported to impair CPSF6 interaction without affecting replication in MDM ( Saito et al . , 2016b ) , and thus to be more specific compared to the N74D mutation used in other studies . MDM from three donors each were transduced with either lentiviral or adeno-associated virus ( AAV ) based vectors expressing a combination of three shRNAs against CPSF6 or a non-targeted shRNA . CPSF6 signal intensities in the nucleus were quantitated to determine the level of knock-down ( Figure 3—figure supplement 1A ) . While both approaches yielded ca . 50–60% reduction of mean CPSF6 signal intensities ( Figure 3—figure supplement 1A , compare top panels to bottom panels ) , transduction with lentiviral vectors resulted in the appearance of nuclear CPSF6 punctae ( Figure 1—figure supplement 2C ) indistinguishable from those associated with HIV-1 derived complexes . We therefore employed AAV vectors for all imaging experiments , while lentiviral knock-down was used in some infection experiments as well . Transduced MDM were infected with non-labeled R5-tropic HIV-1 in a single-round infection at an MOI of 3 . 5 ( based on titration in HeLa-based reporter cells ) and scored for productive infection at day 6 p . i . . Knock-down of CPSF6 modestly , but significantly reduced wild-type HIV-1 infection ( Figure 3A ) . A similar decrease in infectivity was seen for the A77V variant without CPSF6 silencing , different from a previous report where this variant appeared to be unimpaired ( Saito et al . , 2016b ) . CPSF6 depletion had no additional effect on infection by the A77V variant , indicating that its phenotype is indeed due to loss of CPSF6 binding ( Figure 3A and Figure 3—figure supplement 1B , bottom panels show data for AAV mediated depletion ) . Since CPSF6 knock-down levels varied between individual cells , we analyzed the correlation between CPSF6 signal intensity and HIV-1 infection at the single-cell level ( Figure 3B , C ) . Stratification of cells into quartiles according to CPSF6 staining intensity revealed a correlation between HIV-1 infection and CPSF6 staining intensity for wild-type HIV-1 , but not for the A77V variant ( Figure 3B ) . Single-cell analysis further revealed that a certain threshold level of CPSF6 was apparently required for MDM to become productively infected ( Figure 3C ) . This phenotype was more obvious in CPSF6 depleted MDM , but exhibited significant donor-to-donor variation . Taken together , these observations suggest that a threshold level of CPSF6 is required for efficient infection of MDM with wild-type HIV-1 , and that CPSF6 facilitates HIV-1 infection in a CA-dependent manner . Having established that CPSF6 promotes HIV-1 infection of MDM , we next aimed to define the replication step affected . For this , we visualized viral particles in infected MDM at two time points after infection: 24 h p . i . , when reverse transcription was still ongoing , and 60 h p . i . , when reverse transcription was completed in most cells ( Figure 1—figure supplement 2A ) . MDM were transduced with AAV vectors for CPSF6 knock-down and subsequently infected with wild-type HIV-1 or the A77V variant , both carrying IN . eGFP at an MOI of 14 . 5 ( determined on HeLa-derived reporter cells ) . At 24 h p . i . , IN . eGFP-positive subviral complexes derived from wild-type HIV-1 were frequently detected near the nuclear envelope in cells displaying either normal ( NS control ) or low ( K/D ) CPSF6 levels ( 3 or 6 . 2% of IN . eGFP-positive objects , respectively; Figure 4A , Figure 4—figure supplement 1B ) . This was similar for cells infected with the A77V variant ( 5 . 8 or 4 . 4% , respectively; Figure 4—figure supplement 1A , B ) , indicating that CPSF6 plays no role in RTC/PIC trafficking to the nuclear envelope in MDM . A small fraction of IN . eGFP signals was already observed inside the nucleus for wild-type HIV-1 at this time ( <0 . 5% of IN . eGFP positive objects ) , while only a single nuclear IN . eGFP object was detected in HIV-1 A77V infected cells ( Figure 4—figure supplement 1B ) . As expected , a higher proportion of IN . eGFP-positive complexes had entered the nucleus at 60 h p . i . . In MDM transduced with non-targeted vector , 2 . 4% of IN . eGFP positive objects were detected in the nucleus at this time for wild-type HIV-1 infection ( Figure 4B , D ) . Cells depleted of CPSF6 displayed a significantly lower number of nuclear IN . eGFP signals ( 0 . 7%; Figure 4B , D , E ) , while the proportion of IN . eGFP-positive complexes close to the nuclear envelope was higher than in cells expressing normal levels of CPSF6 ( Figure 4D ) . The proportion of nuclear IN . eGFP-positive objects remained low for cells infected with the A77V variant at 60 h p . i . ( 0 . 2%; Figure 4C , D ) , and was further reduced upon CPSF6 knock-down ( Figure 4C , D ) . Interestingly , a higher proportion of HIV-1 complexes was observed close to the nuclear envelope for the A77V variant at this later time point compared to wild-type HIV-1 ( Figure 4C , D ) . The reduction of nuclear complexes for the A77V variant was independently confirmed by immuno-FISH analysis ( Figure 2—figure supplement 2A ) , where 15 of 100 randomly selected cells scored positive for nuclear HIV-1 proviral DNA for the A77V variant compared to 75 in the case of wild-type HIV-1 . Furthermore , the number of viral DNA signals per cell was also lower for the A77V variant ( Figure 2—figure supplement 2B ) , consistent with its reduced infectivity . Quantitation of CPSF6 signal intensities associated with individual RTC/PIC stratified for the subcellular localization revealed clear enrichment of CPSF6 on nuclear complexes with an intermediate signal for complexes at the nuclear envelope and no detectable CPSF6 on cytoplasmic complexes ( Figure 4F , Figure 4—figure supplement 1C ) . Interestingly , this pattern and the observed signal intensities were maintained after CPSF6 knock-down , consistent with our results showing that HIV-1 infection after CPSF6 knock-down preferentially occurs in cells that retain a threshold level of CPSF6 . Very low to non-detectable CPSF6 levels were observed on A77V derived complexes ( Figure 4F , Figure 4—figure supplement 1C ) , consistent with its defect in CPSF6 recruitment . Taken together , our data do not support involvement of CPSF6 in cytoplasmic transport of RTC/PIC to the NPC , but argue for a role of CPSF6 in nuclear import in terminally differentiated macrophages . The small molecule PF-3450074 ( PF74 ) has been shown to block HIV-1 infection by interaction with the viral CA protein ( Blair et al . , 2010; Shi et al . , 2011 ) . Structural studies revealed that the compound preferentially binds to the assembled CA hexamer , where it targets a pocket that serves as binding site for CPSF6 and for the nucleoporin Nup153 ( Price et al . , 2014; Bhattacharya et al . , 2014 ) . PF74 has been shown to block HIV-1 reverse transcription at concentrations > 10 µM ( Shi et al . , 2011; Rasaiyaah et al . , 2013; Peng et al . , 2014; Saito et al . , 2016a ) , most likely by disrupting the integrity of the incoming capsid ( Márquez et al . , 2018 ) . At lower concentrations , the drug has been reported to affect HIV-1 nuclear entry without inhibiting reverse transcription ( Peng et al . , 2014; Saito et al . , 2016a ) . To investigate the effects of PF74 in relation to CPSF6-binding of the viral capsid , we performed single-round infections of MDM with wild-type HIV-1 or the CPSF6-recruitment defective A77V variant in the presence or absence of 2 . 5 µM PF74 and scored infectivity 6d later . PF74 decreased wild-type HIV-1 infection ca . 10-fold compared to the DMSO control ( Figure 3D ) and showed a similar effect on the A77V variant ( 7-fold reduction; Figure 3D ) . These results suggest that PF74 may exhibit additional effects besides blocking the CA-CPSF6 interaction . Parallel infection experiments were performed with the prototypic CPSF6-binding defective CA variant N74D . In agreement with previous reports ( Schaller et al . , 2011; Ambrose et al . , 2012; Rasaiyaah et al . , 2013 ) , we observed a much stronger reduction in infectivity compared to the A77V variant ( Figure 3—figure supplement 2A ) , while PF74 had no additional effect on infection by the N74D variant ( Figure 3—figure supplement 2B ) . To determine whether PF74 affects nuclear entry of HIV-1 subviral complexes , we infected MDM with IN . eGFP-labeled wild-type HIV-1 or the A77V variant in the presence of 2 . 5 µM PF74 . PF74 treatment caused a strong reduction in the proportion of nuclear complexes in wild-type HIV-1 infected cells ( 0 . 01% corresponding to a single nuclear HIV-1 structure in the presence of PF74 compared to 3 . 6% nuclear structures in the control; Figure 4G , Figure 4—figure supplement 1D ) . Interestingly , we also observed a significant reduction in the proportion of IN . eGFP signals close to the nuclear envelope in the presence of PF74 ( 8 . 2% in control vs . 4% in PF74-treated cells; Figure 4G , Figure 4—figure supplement 1D ) . In the case of the A77V variant , the proportion of nuclear structures was very low even in the control , and we could thus not conclusively determine whether PF74 had an additional effect on nuclear entry in this case ( Figure 4G ) . However , the proportion of IN . eGFP signals close to the nuclear envelope was again higher for this variant compared to wild-type HIV-1 ( 13 , 9% vs 8 . 2% ) , and this number was also reduced by PF74 treatment ( Figure 4G , Figure 4—figure supplement 1E ) . These observations indicate that 2 . 5 µM PF74 exhibits additional effects besides affecting nuclear entry of subviral complexes and may explain why PF74 also inhibits MDM infection by the A77V variant . The observation that both , CPSF6 knock-down and the A77V mutation led to accumulation of HIV-1 replication complexes in close proximity to the nuclear envelope suggested that failure to interact with CPSF6 ( due to either lack of CPSF6 or lack of CPSF6 binding ) may arrest incoming subviral HIV-1 particles at the nuclear pore . To directly test this hypothesis , we performed two-color Stimulated Emission Depletion ( STED ) super resolution microscopy of complexes close to the nuclear envelope , achieving a lateral resolution of <50 nm . In order to maximize the number of complexes localized at the nuclear envelope , MDM were infected with HIV-1 ( A77V ) carrying IN . eGFP at an MOI of 14 . 5 and fixed at 60 hr . Cells were immunostained for FG-containing nucleoporins to identify NPCs and for HIV-1 CA . Confocal images identified CA-positive complexes close to the nuclear envelope ( Figure 5A ) and STED microscopy revealed that most of these CA-positive structures were directly associated with nuclear pores ( 34 of 38 CA-positive structures close to the nuclear envelope directly co-localized with NPCs; 89%; Figure 5B ) . In order to quantify the number of particles arrested at the NPC and obtain comprehensive insight into the relative position of subviral complexes with respect to the NPC , we conducted two-color 3D STED analysis . The nucleus of MDM has a diameter of 6–10 µm along the optical axis . Sampling the entire nuclear volume required the acquisition of super-resolved images in many optical sections , corresponding to 200–300 super-resolved images per nucleus . To minimize bleaching during the acquisition , we implemented a light dose management that specifically activates the STED depletion laser beam to switch off fluorophores in the vicinity of a fluorescent feature to be recorded ( Staudt et al . , 2011 ) . This procedure reduced the number of state transition cycles that a fluorescent molecule undergoes and permitted acquisition of hundreds of super-resolved images instead of a few dozen as typically acquired under standard conditions . This approach enabled us to reconstruct the entire nucleus of a MDM infected with HIV-1 ( A77V ) for 60 hr , achieving an almost isotropic final resolution of 100 nm x 100 nm x 150 nm ( xyz ) ( Video 1 ) . Analysis of these reconstructions revealed that 88% of CA-positive complexes at the nuclear envelope were associated with nuclear pores ( 408/465 , five cells from two independent experiments ) . To validate these results , we depleted CPSF6 and subsequently infected MDM with wild-type HIV-1 or the A77V variant carrying IN . eGFP . In addition to CA immunostaining , we performed immunostaining of the nuclear basket protein Nup153 , which binds the CA hexamer overlapping with the CPSF6 binding site and has been implicated in HIV-1 nuclear import ( Matreyek et al . , 2013 ) . Samples were analyzed by two-color STED microscopy . Depletion of CPSF6 induced clear accumulation of IN . eGFP and CA double-positive structures at the nuclear envelope of cells infected with wild-type HIV-1 ( Figure 6A ) , similar to the phenotype of the A77V variant ( Figures 5 , 6C ) . The vast majority of these structures co-localized with Nup153 ( Figure 6A , C; STED images ) , confirming arrest of the HIV-1 replication complex directly at the NPC . Interestingly , line profile analysis of individual wild-type HIV-1 ( with CPSF6 knock-down; Figure 6B ) or A77V subviral particles ( Figure 6D ) revealed partial co-localization of CA signals with the nuclear basket protein Nup153 . The main CA signal localized to the cytoplasmic side of the Nup153 signal with a distance between the peak intensities of approximately 70 nm . Signal intensities for CPSF6 were high for nuclear HIV-1 complexes , much lower for complexes adjacent to the nuclear envelope and undetectable for cytoplasmic structures ( Figure 4F ) . This observation would be consistent with initial recruitment of CPSF6 to the subviral structure at the nuclear basket of the NPC ( where Nup153 resides ) via interaction with the hexameric capsid lattice . To characterize the potential recruitment of CPSF6 to HIV-1 replication complexes at the NPC , we performed two-color STED microscopy of MDM following partial depletion of CPSF6 and infection with wild-type HIV-1 or the A77V variant ( carrying IN . eGFP ) . Fixed cells were immunostained for CPSF6 and Nup153 . Wild-type HIV-1 infected MDM without CPSF6 depletion exhibited intranuclear HIV-1 replication complexes positive for IN . eGFP and CPSF6 and clearly distant from the nuclear envelope ( Figure 7A ) . Subviral structures were arrested at the NPC in CPSF6-depleted cells for both wild-type HIV-1 and the A77V variant ( Figure 7B , C ) . Two-color STED microscopy for CPSF6 and Nup153 combined with diffraction-limited detection of IN . eGFP in a third channel revealed clear enrichment of CPSF6 at those Nup153-positive structures ( i . e . NPCs ) that contained an arrested IN . eGFP-positive HIV-1 replication complex ( 69 of 80 IN . eGFP-positive structures at NPCs co-localized with a strong CPSF6 signal , 86%; Figure 7C ) . In contrast , a CPSF6 signal was detected on only 16% of NPC-associated subviral structures for the A77V variant ( 16 of 99 particles; Figure 7B ) , and the CPSF6 signal intensity was much weaker in this case ( Figure 7B ) . These observations suggest that CPSF6 is recruited to HIV-1 subviral complexes in a CA-dependent manner when these complexes have reached the nuclear basket . Line profile analysis of individual CPSF6 and Nup153 signals associated with IN . eGFP positive structures further revealed that CPSF6 clusters localize at the nuclear side of the NPC ( Figure 7D ) .
HIV-1 infection of all non-dividing cells requires transport of subviral complexes through the intact nuclear pore , and this can also be relevant in dividing cells . Various viral and host cell factors have been implicated in this process ( Matreyek and Engelman , 2013; Bin Hamid et al . , 2016; Yamashita and Engelman , 2017 ) , but the exact mechanism of nuclear import remains unclear and may be cell-type dependent . We have analyzed nuclear entry of HIV-1 and productive infection in post-mitotic human MDM with a focus on the viral CA and the cellular CPSF6 protein . Applying quantitative microscopy of RTC/PIC , we detected CPSF6 primarily on nuclear complexes; cytoplasmic RTC/PIC were only found to be positive when CPSF6 was artificially targeted to the cytoplasm . No impairment of reverse transcription was observed in the latter case , arguing against the hypothesis that CPSF6 binding to the viral capsid arrests viral DNA synthesis to avoid recognition of viral DNA by DNA sensors ( Rasaiyaah et al . , 2013 ) . In contrast , CPSF6 strongly associated with nuclear HIV-1 complexes , and bright nuclear punctae of CPSF6 could be used to identify subviral complexes following HIV-1 infection or lentiviral vector transduction even against the abundant background of CPSF6 in the nucleus . In agreement with a recent study ( Rasheedi et al . , 2016 ) , we observed that CPSF6 is recruited to HIV-1 replication complexes as component of the CPSF52-CPSF62 CF Im complex . This complex is normally involved in pre-mRNA processing , but active transcription was not required for CPSF6 recruitment to HIV-1 PIC . We could also demonstrate that most nuclear subviral complexes co-localized with a signal from LEDGF immunostaining . Our results thus provide direct microscopic support for the model that LEDGF/p75 is recruited to the HIV-1 PIC ( reviewed in Engelman and Singh , 2018 ) , with the caveat that the antibody used did not allow us to discriminate between the two major isoforms of LEDGF . Imaging of RTC/PIC revealed that CPSF6 knock-down or the A77V mutation had little or no effect on trafficking of the RTC/PIC to the nuclear envelope , while nuclear import was severely impaired in both cases . HIV-1 RTC/PIC accumulated close to the nuclear envelope under these conditions , indicating that the viral replication complexes were impeded or arrested at this site . RTC/PIC accumulation at the nuclear envelope was also observed for WT HIV-1 infection with normal CPSF6 levels at early , but not at late time points p . i . , indicating that transfer across the nuclear envelope is a rate-limiting step as recently suggested ( Burdick et al . , 2017 ) . In agreement with several recent studies ( e . g . Burdick et al . , 2017 ) , nuclear import of the HIV-1 post-entry complex did not require reverse transcription , but was affected by depletion of CPSF6 or lack of CPSF6 binding ( Chin et al . , 2015; Ambrose et al . , 2012; Price et al . , 2012 ) . Few nuclear subviral complexes were observed in these cases and reverse transcription positive complexes were largely retained at nuclear pores . Infection of macrophages was only two- to threefold reduced under these conditions , however . This relatively modest effect may be due to the presence of residual CPSF6 , since HIV-1 infection preferentially occurred in cells retaining a threshold level of CPSF6 . Furthermore , infectivity was scored on the bulk population of macrophages with incomplete CPSF6 knock-down , while low CPSF6-expressing cells were selected in the imaging experiments . It should also be considered that HIV-1 RTC/PIC arrested at or close to the nuclear pore in the absence of sufficient CPSF6 recruitment may eventually integrate into chromosomal DNA even without release from the nuclear basket , thus contributing to the observed infection rate . Integration at the NPC and possibly without full release from the nuclear basket is supported by accumulation of HIV-1 proviral DNA in the nuclear periphery for HIV-1 variants with defective CPSF6 binding and for CPSF6 depletion ( Chin et al . , 2015; Achuthan et al . , 2018 ) . It may also explain the altered integration site profile observed for CPSF6 binding defective HIV-1 variants ( Schaller et al . , 2011; Saito et al . , 2016b ) and upon CPSF6 depletion ( Sowd et al . , 2016; Rasheedi et al . , 2016; Achuthan et al . , 2018 ) . Both Nup153 and CPSF6 bind preferentially to the CA hexamer ( Price et al . , 2014 ) , indicating that at least some capsid lattice is retained upon transfer through the NPC in MDM . This is consistent with our observation that a strong CA signal is observed on almost all nuclear HIV-1 complexes in these cells . Several recent studies have reported absent or low CA immunostaining on nuclear HIV-1 structures ( Zhou et al . , 2011; Hulme et al . , 2015; Mamede et al . , 2017 ) , and we were unable to detect a clear CA signal on nuclear complexes in HeLa-based reporter cell lines as well ( Peng et al . , 2014 ) . In contrast , CA was easily detected on almost all nuclear RTC/PIC of HIV-1 infected MDM , and signal intensities were similar or only slightly reduced when compared to cytoplasmic HIV-1 derived structures detected in the same cell . While this does not prove the presence of an intact HIV-1 capsid , the presence of CA on almost all nuclear complexes , the intensity of the CA signal and the known dependence of CPSF6 binding on the assembled CA hexamer suggest that nuclear HIV-1 replication complexes in infected MDM retain at least a large fraction of their CA coat and expose hexameric binding sites to interacting cellular factors . Whether this complex comprises the entire capsid , a partially uncoated capsid , or a capsid-derived structure stabilized by other factors needs to be investigated by ultrastructural analyses . Stabilization of an incomplete capsid by host factor binding is supported by the recent observation that PF74 , binding to the same interface as CPSF6 and Nup153 , can stabilize partially disassembled capsids in vitro ( Márquez et al . , 2018 ) . Using a slightly different imaging-based approach , Francis and Melikyan ( 2018 ) recently reported HIV-1 capsid uncoating at the nuclear envelope , preceding nuclear import of subviral complexes . However , there are pronounced differences in the detection of CA signals on nuclear HIV-1 replication complexes in different cell types ( Peng et al . , 2014 ) , and this apparent discrepancy may simply reflect cell-type specific differences in the mechanism of HIV-1 nuclear import . Using two-color 2D and 3D STED microscopy , we provide direct proof that HIV-1 RTC/PIC at the nuclear envelope are associated with NPC in almost all cases . Accordingly , lack of CPSF6 interaction arrests or delays the HIV-1 replication complex at or within the NPC . A recent report suggested that CPSF6-capsid interactions allow subviral complexes to penetrate deeper into the nucleus of HEK293T- and primary CD4+ T-cells for integration , while loss of CPSF6 interaction led to integration into transcriptionally repressed lamina-associated heterochromatin ( Achuthan et al . , 2018 ) . This may be due to integration of arrested PICs adjacent to the nuclear basket or may indicate an additional role of CPSF6 in intranuclear trafficking of viral PIC . The distance between the peak CA and Nup153 signals in our study was ca . 70 nm , with the CA signal oriented towards the cytoplasmic side . CPSF6 , on the other hand , accumulated adjacent to Nup153 , but clearly localized towards the nucleoplasm; this CPSF6 clustering was dependent on CA . Taken together , these results suggest that a strongly CA-positive HIV-1 replication complex retaining at least a partial hexameric lattice is arrested inside the pore of the NPC . The central pore is ca . 40–90 nm long , with filaments extending 50–75 nm on both sides ( Beck et al . , 2004 ) . Accordingly , a single HIV-1 capsid or capsid-derived structure of ca . 100 nm in length ( Briggs et al . , 2003 ) could span the pore . It can concomitantly present numerous CA epitopes at a peak distance of 70 nm from the nuclear basket towards the cytoplasm and induce strong CA hexamer dependent CPSF6 binding on the nucleoplasmic face of the nuclear basket . Tight association with the pore channel and the CPSF6 cluster may partially obscure antibody detection of CA , but the strong CA signal on nuclear complexes indicates that CA is largely retained after NPC passage . CPSF6 recruited to this capsid-derived structure at the nuclear basket might displace Nup153 , thereby freeing Nup153 molecules of this NPC to progressively bind to upstream CA hexamers and thus promote nuclear import . Accordingly , abundant nuclear CPSF6 competing Nup153 from the common interface results in the observed large CPSF6 clusters on HIV-1 complexes at the nuclear basket and may eventually cause their release into the nucleoplasm . This model thus proposes consecutive and competitive binding of Nup153 and CPSF6 to the CA lattice on HIV-1 replication complexes as driving force for their nuclear entry; a graphic depiction of the model is shown in Figure 8 . We realize that the width of 60 nm at the wide end of the cone shaped HIV-1 capsid ( Briggs et al . , 2003 ) exceeds the width of the NPC translocation channel of ca . 40 nm ( Beck et al . , 2004 ) , and the capsid-derived structure may be modulated to fit these size requirements . Further studies applying ( correlative ) cryo-electron microscopy will be needed to define the ultrastructure of this HIV-1 capsid-derived structure at the NPC and to determine whether the NPC structure changes during nuclear translocation of such a large cargo . The experimental system described in this report seems ideally suited to address these important questions , which are relevant not only for HIV-1 biology , but for understanding nuclear transport of large cargo and the flexibility of the NPC in general .
Human embryonic kidney 293 T cells ( HEK293T ) and TZM-bI indicator cells have been previously described ( Pear et al . , 1993; Wei et al . , 2002 ) . The HeLaP4-derived cell line stably transduced with shRNA targeting TNPO3 ( TNPO3-KD ) was kindly provided by Z . Debyser ( University of Leuven , Belgium ) ( Thys et al . , 2011 ) . Cells were cultured at 37°C and 5% CO2 in Dulbecco´s Modified Eagle Medium ( DMEM; Thermo Fisher Scientific , Waltham , USA ) , supplemented with 10% fetal calf serum ( FCS; Biochrom GmbH , Berlin , Germany ) , 100 U/mL penicillin , and 100 µg/mL streptomycin . For preparation of monocyte-derived macrophages ( MDM ) human peripheral blood mononuclear cells ( PBMC ) were isolated from buffy coats of healthy donors by Ficoll density gradient centrifugation . PBMC were seeded in RPMI 1640 medium ( Thermo Fisher Scientific ) supplemented with 10% heat inactivated FCS and antibiotics for 2 hr at 37°C . Subsequently , non-adherent cells were removed , adherent monocytes were washed , and further cultured in RPMI 1640 containing 10% heat inactivated FCS , antibiotics and 5% human AB serum ( Sigma Aldrich , St . Louis , USA ) or 50 ng/mL macrophage colony-stimulating factor ( M-CSF; Peprotech , Rocky Hill , USA; for experiments performed with AAVs ) for 10d until differentiation to macrophages . The HIV-1 proviral plasmid pNL4-3 has been described ( Adachi et al . , 1986 ) . Plasmid pNL4-3ΔEnv contains a 2 bp fill-in of an NdeI site in the env ORF resulting in a frameshift and premature stop codon . The A77V and N74D exchanges in the CA-coding region of gag were introduced into pNL4-3 through PCR directed mutagenesis , and transferred into pNL4-3ΔEnv through double digestion with BssHII and AgeI , followed by ligation with the corresponding fragment from pNL4-3-A77V or -N74D . Plasmid pNLC4-3ΔTat contains a 31 bp deletion in the first exon of tat and was kindly provided by Thorsten Müller ( University of Heidelberg , Germany ) . Plasmid pEnv-4059 expressing an R5-tropic Env protein from a primary HIV-1 isolate ( Schnell et al . , 2011 ) was kindly provided by R . Swanstrom ( University of North Carolina , USA ) . Plasmid pVpr . IN . eGFP ( Albanese et al . , 2008 ) encoding a Vpr . IN . eGFP fusion protein with an HIV-1 protease cleavage site between Vpr and IN . eGFP was kindly provided by A . Cereseto ( CIBIO , Mattareo , Italy ) . The adenoviral helper plasmid pVAE2AE4-5 has been described ( Matsushita et al . , 1998 ) . The AAV helper plasmid encoding the rep and cap genes was kindly provided by D . Grimm ( Heidelberg University , Germany ) . The vector for expression of triple short hairpin RNA ( shRNA ) in AAV was also kindly provided by D . Grimm ( Heidelberg University , Germany ) and will be reported elsewhere . Into this AAV vector , we inserted three CPSF6 targeting sequences ( CPSF6-1: 5´-GCGAAGAGTTCAACCAGGAA-3´; CPSF6-2: 5´-GCCAGAAGACCGAGATTACAT-3´; CPSF6-3: 5´-GGTGGACAACAGATGAAGA-3´ ) under the control of three different promoters or a single non-silencing ( 5´-TCGGCGCAGTCTAATTATA-3´ ) shRNA . The lentiviral vectors pHIVSIREN expressing shRNAs CPSF6-1 ( 5´-GCCAGAAGACCGAGATTACAT-3´ ) , CPSF6-2 ( 5´-GCGAAGAGTTCAACCAGGAA-3´ ) or non-silencing control ( 5´-TCGGCGCAGTCTAATTATA-3´ ) were kindly provided by G . Towers ( University College London , UK ) ( Rasaiyaah et al . , 2013 ) . The VSV-G-expressing envelope plasmid pMD2 . G and the packaging vector psPAX2 were generated in the lab of D . Trono ( EPFL , Lausanne , Switzerland ) and obtained through AddGene . Rabbit and sheep polyclonal antisera against HIV-1 CA were raised against purified recombinant proteins . Mouse monoclonal laminA/C antibody ( sc-7292 ) was purchased from Santa Cruz ( Heidelberg , Germany ) . Mouse monoclonal antibodies against Nup153 ( QE5 , Ab24700 ) and Nuclear Pore Complex proteins ( mAb414 ) were purchased from Abcam ( Cambridge , UK ) . Mouse monoclonal antibody against LEDGF ( 611714; recognizing an epitope present in the p52 and the p75 isoform of the protein ) was purchased from BD Biosciences ( Franklin Lakes , USA ) . Affinity purified antibodies against CPSF5 ( mouse , SAB1404890 ) , CPSF6 ( rabbit , HPA039974 ) and CPSF7 ( rabbit , HPA041094 ) were purchased from Sigma Aldrich . Alexa Fluor labeled secondary antibodies were purchased from Thermo Fisher Scientific . Secondary antibodies labeled with Atto or STAR RED dyes for STED were purchased from Sigma-Aldrich . 10 mM stock solutions of Efavirenz ( obtained through the AIDS Research and Reference Reagent Program , Division AIDS , NIAID ) , PF74 ( Sigma Aldrich ) or Maraviroc ( Sigma Aldrich ) were prepared in dimethyl sulfoxide and stored at −20°C . A 10 mM Stock solution of Raltegravir ( obtained through the AIDS Research and Reference Reagent Program , Division AIDS , NIAID ) was prepared in H2O and stored at −20°C . All chemicals and reagents for transfection and FISH were obtained from standard commercial sources , unless indicated otherwise . For production of HIV-1 particles and lentiviral vectors , HEK293T cells were transfected using a standard calcium phosphate transfection . For producing viral particles containing IN . eGFP , pNL4-3 was co-transfected with pVpr . IN . eGFP at a molar ratio of 4 . 5:1 . R5-tropic and R5-tropic tat-defective viral particles were produced by co-transfection of pNL4-3ΔEnv ( -WT or -A77V ) or pNLC4-3ΔTat , pEnv-4059 and pVpr . IN . eGFP at a molar ratio of 4 . 5:1:1 . Lentiviral vectors for CPSF6 knock-down were generated by co-transfection of pHIVSIREN , pMD2 . G and psPAX2 at a molar ratio of 2:1 . 4:1 . 4 . Supernatants containing HIV-1 or lentiviral particles were collected at 36 h p . t . , filtered through 0 . 45 µm nitrocellulose filters and concentrated by ultracentrifugation through a 20% ( w/w ) sucrose cushion . Subsequently , particles were resuspended in phosphate buffered saline ( PBS ) solution containing 10% heat inactivated FCS and 10 mM Hepes pH 7 . 5 , and stored in aliquots at −80°C . Virus titer was determined by titration on TZM-bI indicator cells followed by microscopic quantitation of beta-lactamase expressing cells at 48 h p . i . . Titration on this indicator cell line was used as a reference to determine the virus input for imaging and infectivity experiments in macrophages . Particle-associated RT activity was determined by SG-PERT ( SYBR Green based Product Enhanced Reverse Transcription assay ) ( Pizzato et al . , 2009 ) . The concentration of CA was measured by an in-house enzyme-linked immunosorbent assay ( Wiegers et al . , 1998 ) . Briefly , ELISA plates were coated with 50 ng of monoclonal anti-p24 antibody from hybridoma cell line 183 clone H12-5C ( obtained through the AIDS Research and Reference Reagent Program , Division AIDS , NIAID ) . Subsequently , wells were blocked with 10%FCS ( Biochrom ) in PBS , and samples of interest ( previously diluted in PBS/0 . 1% Tween20 ) were added . Antigen detection was done by addition of an in-house polyclonal rabbit antiserum against CA , followed by the addition of goat antiserum against rabbit immunoglobulin G conjugated to horseradish peroxidase ( Dianova , Hamburg , Germany ) , and detection of enzymatic activity obtained from absorbance readings after adding the substrate tetramethylbezidine ( TMB; Thermo Fisher Scientific ) . As standard , purified recombinant HIV-1 CA of known concentration was used . For production of AAV vectors , HEK293T cells were transfected with Turbofect ( Thermo Fisher Scientific ) . Cells were co-transfected with AAV helper plasmid encoding rep and AAV6- or DJP2- cap genes , AAV triple shRNA vector and adenoviral helper plasmid at a molar ratio of 1:1:1 . 72 h p . t . , cells were collected in PBS and lysed by freeze-thaw cycles in liquid nitrogen and subsequent sonification . Cell debris was removed by centrifugation ( 16000 x g , 10 min , at room temperature ) and supernatant containing the AAV particles was stored in aliquots at −80°C . For imaging of RTC/PIC , TNPO3-KD cells and MDM were seeded in 8-well LabTek ( #155411 , Thermo Fisher Scientific ) or in 8-well LabTek II chamber slides ( #155409 , Thermo Fisher Scientific ) for confocal and STED microscopy , respectively . TNPO3-KD cells were seeded in the presence of 6 µM aphidicolin ( Sigma Aldrich ) and infected on the following day with HIV-1 ( IN . eGFP ) at a multiplicity of infection ( m . o . i . ) of 25 in medium containing 10 µM EdU ( Thermo Fisher Scientific ) and 6 µM aphidicolin . Cells were pre-incubated at 16°C for 30 min and then shifted to 37°C . After 2 hr , medium was removed and replaced by fresh pre-warmed medium containing 10 µM EdU and incubation was continued at 37°C . To stop infection , cells were washed with PBS and fixed with 4% paraformaldehyde ( PFA; Electron Microscopy Sciences , Hatfield , USA ) for 30 min at room temperature . Subsequently , cells were washed and permeabilized with 0 . 5% ( vol/vol ) Triton X-100 for 15 min . Cells were washed and click-labeling was performed for 30 min at room temperature using the Click-iT EdU-Alexa Fluor 647 Imaging Kit ( Thermo Fisher Scientific ) following manufacturer´s instructions . For immunostaining , cells were blocked for 30 min with 3% bovine serum albumin ( BSA ) in PBS and incubated with the primary antibody in 0 . 5% BSA in PBS for 1 hr at room temperature . Cells were washed and incubated with the corresponding secondary antibody for 1 hr at room temperature in 0 . 5% BSA . For MDM , cells were infected with 100 ng CA of WT or A77V HIV-1 NL4-3 ( IN . eGFP ) pseudotyped with R5-tropic 4059 Env in the presence of 10 µM EdU . This amount of virus corresponds to an MOI of 14 . 5 , based on the infectious titer determined on TZM-bI indicator cells . Infections with unlabeled virus were performed at an MOI of 3 . 5 corresponding to ~15 ng CA , unless otherwise indicated . It should be noted that MDM are less efficiently infected than TZM-bl cells , thus yielding lower infection rates , and exhibited strong donor-dependent variability . For infection times longer than 24 hr , Maraviroc ( Sigma Aldrich ) was added to a final concentration of 5 µM at 24 h p . i . to prevent a second round of infection . For experiments with Efavirenz ( EFV ) and Raltegravir ( Ral ) , MDM were seeded in the same way and infected with HIV-1 in the presence of 5 µM EFV ( Sigma Aldrich ) or 5 µM Ral ( AIDS Research and Reference Reagent Program , Division AIDS , NIAID ) . For experiments with flavopiridol , 5 µM flavopiridol ( Sigma Aldrich ) was added to the medium at 96 h p . i . and cells were incubated for further 12 hr before fixation . Fixation , click-labeling and immunostaining were performed as described above . Detection of viral DNA with FISH was performed essentially as described ( Solovei and Cremer , 2010 ) . MDM were seeded in 24-well plates with glass cover slips and infected with 50 ng CA ( MOI 8 , based on titration on TZM-bI indicator cells ) of WT or A77V HIV-1 NL4-3 with R5-tropic 4059 Env . 72 h p . i . cells were fixed with 4% PFA in PBS for 10 min at room temperature , washed and permeabilized with PBS/0 . 5% Triton X-100 for 10 min . After permeabilization , cells were treated with ethylene glycol bis- ( succinimidyl succinate ) , permeabilized again with PBS/0 . 5% Triton X-100 , kept overnight in PBS/20% glycerol , and subjected to five freeze-thaw cycles . Subsequently , cells were treated with 0 . 1N HCl for 10 min and PBS/0 . 5% Triton X-100 for 5 min , followed by RNAse A ( 100 µg/mL in 2x SSC ) for 1 hr , washing and storage in 50% formamide/2x SSC overnight . 3 µg of an HIV-1 HXB2 proviral plasmid was labeled by nick translation in the presence of 16-dUTP biotin at 15°C for 3 hr using a nick translation kit ( Roche , Basel , Switzerland ) . The probe was precipitated with ethanol in the presence of Cot-1 and herring sperm DNA and resuspended in 20 µL of hybridization buffer ( 50% formamide , 10% dextran in 2x SSC buffer ) . 1 µL probe was used per cover slip , denatured at 95°C for 5 min and kept on ice until incubation with immunostained cells on glass cover slips . Samples were heat-denatured at 80°C for 6 min . Hybridization was performed for 48 hr at 37°C in a water bath . Biotin-labeled hybridized probes were detected using the TSA Plus system ( Perkin Elmer , Waltham , USA ) . For detection of viral RNA by FISH , MDM were seeded in 8-well LabTeks ( #155411 , Thermo Fisher Scientific ) and infected with 50 ng CA ( MOI 8 , based on titration on TZM-bI indicator cells ) of HIV-1 NL4-34059 or HIV-1 NL4-3ΔTat4059 . At 96 h p . i . cells were treated with DMSO or 5 µM flavopiridol . 12 hr after addition of the inhibitors , cells were fixed with 3 . 7% formaldehyde in PBS for 10 min at room temperature , washed and permeabilized with 70% ethanol at 4°C overnight . After permeabilization , cells were washed , and immunostained as described above . After immunostaining , cells were washed three times with 10% formamide in Stellaris RNA FISH wash buffer A ( Biosearch Technologies , Novato , USA; Cat# SMF-WA1-60 ) at room temperature for 5 min , followed by hybridization . For hybridization , probe was diluted to a final concentration of 125 nM in Stellaris RNA FISH Hybridization buffer ( Biosearch Tech . Cat# SMF-HB1-10 ) with 10% formamide , and left incubating overnight at 37°C in humid chamber . After hybridization , cells were washed twice for 2 min with Stellaris RNA FISH wash buffer A with 10% formamide , and subsequently washed three times for 2 min with Stellaris RNA FISH wash Buffer B ( Biosearch Tech . Cat#SMF-WB1-20-BS ) . Stellaris probe for RNA FISH was synthesized by Biocat GmbH ( Heidelberg , Germany ) using HIV-1 NL4-3 proviral plasmid and labeled with CAL Fluor Red 610 dye . For high throughput analysis of infectivity , MDM ( ca . 1 × 104 cells/well ) were seeded in 96-well plates ( Costar #3606 ) . CPSF6 knock-down was performed using either AAV vectors ( for imaging and infectivity experiments ) or lentiviral vectors ( for infectivity experiments ) . MDM were transduced once with lentiviral vectors ( 4d after induction of differentiation ) expressing a non-targeted shRNA ( NS control ) or two different shRNAs against CPSF6 ( K/D ) . A total of 300mU RT ( determined by quantitating RT activity ) of lentiviral vectors was used for transduction of each well . 24 hr later , medium was replaced , and cells were left standing for another 7d . For AAV transduction , cells were transduced three times ( 4 , 8 , 12d after induction of differentiation ) with equal amounts of AAV crude lysates expressing three shRNAs against CPSF6 or a non-targeted shRNA . Seven ( when using lentiviral vectors ) or 12d ( when using AAV ) after initial transduction , cells were infected in triplicate with 15 ng CA ( MOI 3 . 5 , based on titration on TZM-bI indicator cells ) of WT or A77V HIV-1 NL4-3 with R5-tropic 4059 Env . To block further entry events and prevent secondary infection , 5 µM Maraviroc ( Sigma Aldrich ) was added to the medium at 24 h p . i . After 6d , cells were fixed in 4% PFA for 90 min and immunostained with anti-CA ( sheep ) and anti-CPSF6 ( rabbit ) antisera , as described above . Additionally , cells were counterstained with Hoechst ( Thermo Fisher Scientific ) . Plates were imaged with a fully automated epifluorescence ScanR screening microscope equipped with the ScanR acquisition software ( Olympus Biosystems , Shinjuku , Japan ) . Images were acquired in the Hoechst , CA- and CPSF6-staining channels using the corresponding excitation and emission filters . The percentage of infected cells was quantified as previously described ( Börner et al . , 2010 ) . Mock-infected wells were used as a negative control to set the threshold . For infectivity experiments with PF74 , cells were infected as described above , and a final concentration of 2 . 5 µM PF74 ( Sigma Aldrich ) was added together with the virus . For time-of-addition experiments , MDM were seeded in 96-well plates and infected as described above . 5 µM Efavirenz or DMSO was added to the infection together with the virus or every 24 hr for 3 days . 5 µM Maraviroc ( Sigma Aldrich ) was also added to the medium at 24 h p . i . Fresh medium with inhibitors was supplied after 3d . 6d after infection , cells were fixed in 4% PFA for 90 min , immunostained with anti-CA ( rabbit ) antiserum and counterstained with Hoechst ( Thermo Fisher Scientific ) . Percentage of infected cells was quantified as described above . MDM were infected as described above in the presence or absence of 5 µM Raltegravir . 24 h p . i . 5 µM Maraviroc ( Sigma Aldrich ) was added to the medium . At 72 h p . i . medium was replaced by fresh medium with 5 µM Maraviroc ( Sigma Aldrich ) and 5 µM Ral , if needed . 96 hr after infection , medium was removed , and cells were washed twice with PBS before lysis . RNA was extracted using InviTrap Spin Universal RNA Mini Kit ( Stratec Biomedical , Birkenfeld , Germany ) following manufacturer´s instructions . cDNA synthesis was performed with the SuperScript III Reverse Transcriptase kit ( Thermo Fischer Scientific ) following manufacturer´s instructions , using 100 ng of RNA . cDNA was used as a template for detecting HIV-1 transcripts with TaqMan quantitative PCR . PCR conditions were as follows: 1X iQ Supermix ( BioRad , Hercules , USA ) , 900 nM primers and 200 nM probe . Cycling conditions were: 98°C for 3 min , 44 cycles of 98°C for 10 s and 60°C for 40 s , followed by 60 cycles with a ramp rate of 0 . 5°C/cycle for 5 s each starting at 65°C . Primers for detection of gag transcripts used were: Forward , 5´ ACATCAAGCAGCCATGCAAAA 3´ , Reverse , 5´ TGGATGCAATCTATCCCATTCTG 3´ , Probe , 5´-FAM- AAGAGACCATCAATGAGGAA-TAMRA 3´ . Primers and probe binding to eukaryotic 18S rRNA ( VIC/MGB , Thermo Fisher 4319413E ) were used in parallel as endogenous control for normalization . Multi-channel 3D image series were acquired with a Perkin Elmer Ultra VIEW VoX 3D spinning disk confocal microscope ( SDCM ) using a 100x oil immersion objective ( NA 1 . 4 ) ( Perkin Elmer ) , with a z-spacing of 200 nm . Images were recorded in the 405 , 488 , 561 and 640 nm channels . Images of RNA FISH samples were acquired with a Leica SP8 DLS laser scanning confocal using a 63x oil immersion objective ( NA 1 . 4 ) ( Leica , Wetzlar , Germany ) , with a z-spacing of 300 nm . Stimulated emission depletion ( STED ) imaging was performed with a λ = 775 nm STED system ( Abberior Instruments GmbH , Göttingen , Germany ) , using a 100x Olympus UPlanSApo ( NA 1 . 4 ) oil immersion objective . Images were acquired using the 590 and 640 nm excitation laser lines . Nominal STED laser power was set to 80% of the maximal power of 1250 mW with 20-30µs pixel dwell time and 20 nm pixel size . For 3D STED data 60% of the STED laser power was used for fluorescence depletion in the Z-axis and RESCue illumination scheme was used to minimize bleaching . Sampling frequency was 30 nm in all three axis ( xyz ) . All STED images shown ( except 3D STED images ) were linearly deconvolved with a Lorentzian function ( fwhm 50 nm ) using the software Imspector ( Abberior Instruments GmbH ) . To quantify the signal intensity from objects distributed throughout the entire volume of the cells , the data ( Z-image series ) were reconstructed in the 3D space using Imaris 8 . 1 ( Bitplane AG , Zürich , Switzerland ) . Acquired images were first deconvolved by Autoquant X3 ( Media Cybernetics , Rockville , USA ) using Constrained Maximal Likelihood Estimation ( CMLE ) algorithm with 10 iterations and SNR = 20 . Next , the xyz coordinates of objects in IN . eGFP channel were automatically identified using the ‘spot detection’ module in Imaris . Background was subtracted , and an estimated diameter of 300 nm was used for spot detection . Mean signal intensities from selected spots were measured in the EdU , IN . eGFP , laminA and CPSF6/CA channels . To determine CPSF6 and CA mean intensities from single cells in infectivity assays , images were processed using the Konstanz Information Miner ( KNIME , www . knime . org ) and the KNIME image processing plugins ( KNIP ) . A previously described workflow was modified ( Grosse et al . , 2017 ) . Briefly , cellular objects in the Hoechst , CA and CPSF6 channels were identified by background subtraction , automatic global thresholding and connected component analysis . For each individual cellular object with positive Hoechst signal , the mean signal intensities from CPSF6- and CA-staining ( in the nucleus and in the whole cell ) were calculated . Data analysis was performed using GraphPad Prism software ( GraphPad Software , Inc , La Jolla , USA ) . Statistical significance was only assessed for sample sizes n > 3 . Before assessing statistical significance , a Shapiro Wilk test ( α = 0 . 05 ) was performed to verify normality . Two-tailed non-paired Mann-Whitney test ( α = 0 . 05 ) was used to check statistical significance of non-parametric data . In analyses of the distribution of IN . eGFP objects , a non-paired two-tailed Z-test ( α = 0 . 05 ) was performed to assess statistical significance of two proportions . | Viruses are miniscule parasites that hijack the resources of a cell to make more of themselves . For many , this involves getting inside the nucleus , the fortress that protects the cell’s genetic information . To do so , viruses need to first find a way through a double-layered membrane called the nuclear envelope , which only opens up when a cell divides . Yet , the human immunodeficiency virus type 1 ( HIV-1 ) can infect cells that no longer divide , and in which the nucleus’ walls never come down . The virus cores then head for the nuclear pores , heavily guarded holes in the nuclear envelope that allow the cell's own molecules to go in and out of the nucleus . But HIV-1 is too big to fit through , as its genetic information is encased in a capsid , a coat made of a complex assembly of proteins . However , research shows that these capsid proteins can bind to host proteins at the pore or even inside the nucleus . For example , the capsid protein can recognize the pore protein Nup153 , or the nuclear protein CPSF6 . These interactions could help the virus make its way in , but how these events unfold is still unclear . To explore this , Bejarano , Peng et al . attached fluorescent labels to HIV-1 and watched as it infected non-dividing cells . Rather than completely get rid of their capsid before they crossed the pores , the virus particles hung on to a large part of their lattice . This remaining coat then attached to CPSF6; when this protein was missing or could not bind to capsid proteins , the viral complexes got stuck in the nuclear pores . This suggests that the capsid lattice could first interact with Nup153 inside the pores: then , CPSF6 would take over , knocking Nup153 away and pulling HIV-1 into the nucleus . Armed with this knowledge , virologists and drug developers could try to block HIV-1 from entering the cell’s nucleus; they could also start to dissect how drugs that target the HIV-1 capsid work . Ultimately , HIV-1 may serve as a model to unravel how large objects can pass the nuclear pore , which may help us understand how molecules are constantly trafficked in and out of the nucleus . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | HIV-1 nuclear import in macrophages is regulated by CPSF6-capsid interactions at the nuclear pore complex |
Tumor cells display features that are not found in healthy cells . How they become immortal and how their specific features can be exploited to combat tumorigenesis are key questions in tumor biology . Here we describe the long non-coding RNA cherub that is critically required for the development of brain tumors in Drosophila but is dispensable for normal development . In mitotic Drosophila neural stem cells , cherub localizes to the cell periphery and segregates into the differentiating daughter cell . During tumorigenesis , de-differentiation of cherub-high cells leads to the formation of tumorigenic stem cells that accumulate abnormally high cherub levels . We show that cherub establishes a molecular link between the RNA-binding proteins Staufen and Syncrip . As Syncrip is part of the molecular machinery specifying temporal identity in neural stem cells , we propose that tumor cells proliferate indefinitely , because cherub accumulation no longer allows them to complete their temporal neurogenesis program .
Throughout the animal kingdom , stem cells supply tissues with specialized cells . They can do this because they have the unique ability to both replicate themselves ( an ability termed self-renewal ( Smith , 2006 ) ) and to simultaneously generate other daughter cells with a more restricted developmental potential . Besides their role in tissue homeostasis , stem cells have also been linked to tumor formation ( Reya et al . , 2001 ) . They can turn into so-called tumor stem cells that sustain tumor growth indefinitely . The mechanisms that endow tumor stem cells with indefinite proliferation potential are not fully understood . The fruit fly Drosophila has emerged as a genetically tractable system to model tumors in a developmental context and adult tissues ( Gateff , 1978; Gonzalez , 2013 ) as well as to study naturally occurring tumors ( Salomon and Jackson , 2008; Siudeja et al . , 2015 ) . In the developing CNS , neural stem cells , called neuroblasts ( NBs ) give rise to most neurons and glial cells of the adult fly brain ( Truman and Bate , 1988 ) . For this , they repeatedly divide into one self-renewing and one differentiating daughter cell ( Kang and Reichert , 2015; Neumüller et al . , 2011 ) . Disrupting these asymmetric cell divisions can generate lethal , transplantable brain tumors ( Bello et al . , 2006; Betschinger et al . , 2006; Cabernard et al . , 2010; Janssens and Lee , 2014; Knoblich , 2010; Lee et al . , 2006; 2006c; 2006d ) . Importantly , the failure to divide asymmetrically has also been linked to tumorigenesis in mammals , particularly in breast cancer ( Cicalese et al . , 2009 ) , myeloid leukemia ( Ito et al . , 2010; Wu et al . , 2007; Zimdahl et al . , 2014 ) and gliomas ( Chen et al . , 2014 ) . Most Drosophila brain tumors originate from the so-called type II neuroblasts ( NBIIs ) ( Figure 1A ) . NBIIs divide asymmetrically into a larger cell that retains NB characteristics and a smaller intermediate neural progenitor ( INP ) . Newly formed immature INPs ( iINPs ) go through a defined set of maturation steps to become transit-amplifying mature INPs ( mINPs ) . After this , a mINP undergoes 3–6 divisions generating one mINP and one ganglion mother cell ( GMC ) that in turn divides into two terminally differentiating neurons or glial cells ( Bello et al . , 2008; Boone and Doe , 2008; Bowman et al . , 2008 ) . Similar to mammalian brain progenitors ( Kohwi and Doe , 2013 ) , Drosophila NBs exit proliferation once they complete a specified temporal program during which they generate different types of morphologically distinct progeny ( Homem et al . , 2014; Liu et al . , 2015; Maurange et al . , 2008; Ren et al . , 2017; Syed et al . , 2017 ) . It is thought that their correct temporal identity requires the RNA-binding proteins IGF-II mRNA-binding protein ( Imp ) and Syncrip ( Syp ) . During early larval stages , Imp levels are high and Syp levels are low . Over time , Imp expression gradually decreases while the amount of Syp increases . This leads to highly Syp-positive NBs with no detectable Imp at the end of larval development . Manipulating these opposing gradients changes the number and type of neurons made ( Liu et al . , 2015; Ren et al . , 2017; Syed et al . , 2017 ) . During each NBII division , a set of cell fate determinants is segregated into the INP ( Bello et al . , 2008; Boone and Doe , 2008; Bowman et al . , 2008 ) ( Figure 1A ) . Among those are the Notch inhibitor Numb and the TRIM-NHL protein Brain tumor ( Brat ) ( Bello et al . , 2006; Betschinger et al . , 2006; Knoblich et al . , 1995; Lee et al . , 2006d; Spana et al . , 1995 ) . Loss of these cell fate determinants ( Arama et al . , 2000; Bello et al . , 2008; Betschinger et al . , 2006; Gateff , 1978; Lee et al . , 2006d; Wang et al . , 2006 ) leads to the generation of ectopic NB-like cells at the expense of differentiated brain cells . Formation of malignant brain tumors has also been observed upon the depletion of downstream factors that normally maintain the INP fate ( Eroglu et al . , 2014; Janssens and Lee , 2014; Koe et al . , 2014; Weng et al . , 2010 ) . These features make Drosophila a model for the stepwise acquisition of tumor stem cell properties . When numb or brat are inactivated ( Figure 1B ) , the smaller NBII progeny fails to establish an INP fate ( Janssens et al . , 2014; Lee et al . , 2006d ) and initially enters a long transient cell cycle arrest ( Bowman et al . , 2008; Lee et al . , 2006d ) . Only after this lag period , the smaller cell regrows to a NB-sized cell that has acquired tumor stem cell properties and that we therefore refer to as tumor neuroblast ( tNB ) ( Bello et al . , 2008; Betschinger et al . , 2006; Bowman et al . , 2008; Lee et al . , 2006d ) . NBIIs and ectopic tNBs are indistinguishable in terms of markers . Both cell populations are characterized by the expression of self-renewal genes and lack differentiation markers ( Bello et al . , 2006; Betschinger et al . , 2006; Lee et al . , 2006d ) , but nevertheless behave differently . Shortly after entering pupal stages , NBs decrease their cell volumes successively with each NB division before they exit the cell cycle and differentiate ( Homem et al . , 2014; Maurange et al . , 2008 ) . However , tNBs do not shrink during metamorphosis ( Homem et al . , 2014 ) and continue to proliferate even in the adult fly brain ( Bello et al . , 2006; Loop et al . , 2004; Mukherjee et al . , 2016; Narbonne-Reveau et al . , 2016 ) . Moreover , in contrast to wild-type brains , the resulting tumor brains can be serially transplanted into host flies for years ( Caussinus and Gonzalez , 2005; Gateff , 1978 ) , indicating the immortality of these tumors . Similarly , mammalian homologues of numb ( Cicalese et al . , 2009; Colaluca et al . , 2008; Ito et al . , 2010; Pece et al . , 2004 ) and brat ( Chen et al . , 2014; Mukherjee et al . , 2016 ) have been shown to inhibit tumor growth . Furthermore , the human brat homologue TRIM3 is depleted in 24% of gliomas ( Boulay et al . , 2009; Chen et al . , 2014 ) and NUMB protein levels are markedly reduced in 55% of breast tumor cases ( Pece et al . , 2004 ) . Therefore , results obtained in these Drosophila tumor models are highly relevant . Here , we used the Drosophila brat tumor model to investigate how tNBs differ from their physiological counterparts , the NBIIs . Our results indicate that progression towards a malignant state is an intrinsic process in brat tNBs that does not correlate with stepwise acquisition of DNA alterations . Transcriptome profiling of larval NBs identified the previously uncharacterized long non-coding ( lnc ) RNA cherub as crucial for tumorigenesis , but largely dispensable for NB development . Our data show that cherub is the first identified lncRNA to be asymmetrically segregated during mitosis into INPs , where the initial high cherub levels decrease with time . Upon the loss of brat , the smaller cherub-high cell reverts into an ectopic tNBs resulting in tumors with high cortical cherub . Molecularly , cherub facilitates the binding between the RNA-binding protein Staufen and the late temporal identity factor Syp and consequently tethers Syp to the plasma membrane . Depleting cherub in brat tNBs leads to the release of Syp from the cortex into the cytoplasm and represses tumor growth . Our data provide insight into how defects in asymmetric cell division can contribute to the acquisition of tumorigenic traits without the need of DNA alterations .
Transplanted brat tumors have the ability to indefinitely grow in wild-type host flies in contrast to injected control brains ( Caussinus and Gonzalez , 2005 ) . However , it is unclear whether the proliferation potential of tNBs is an intrinsic feature or an altered response to signals from the brain niche . For this purpose , we transplanted NBs , isolated from other brain cells by Fluorescence-activated cell sorting ( FACS ) , into adult host flies . This procedure also allows the transplantation of equivalent NB numbers for each condition in contrast to the different NB quantities of wild-type compared to brat depleted brain pieces . Similar to the transplantation of tissue fragments , only brat RNAi tNBs formed tumors in host flies , whereas control NBs did not ( Figure 1C ) . Thus , brat RNAi tNBs require intrinsic mechanisms that render them immortal . It is generally assumed that tumorigenesis involves multiple DNA mutations affecting proliferation control pathways ( Hanahan and Weinberg , 2011 ) . Nonetheless , the events leading to malignant transformation in brat mutants could be genetic or ( post ) transcriptional . To distinguish between these possibilities , we analyzed the DNA content of cells from tumor and control brains by Hoechst staining and subsequent FACS analyses . In larval and adult stages , tumor brains , similar to control brains , showed two peaks corresponding to diploid ( G0/1 phase ) and tetraploid ( G2/M phase ) karyotypes ( Figure 2A ) . Although brat tumor cells showed an increase in tetraploid cells due to more dividing cells , we did not observe any aneuploid and polyploid karyotypes ( Figure 2A ) . To investigate potential tumor-specific DNA alterations in more detail , we sequenced the genome of three adult brain tumors from hypomorphic brat k06028 mutants and compared it to that of tumor-free abdominal tissue from the same flies . To identify large genomic aberrations we compared the coverage of tumor versus control samples across the genome . As exemplified for the chromosome arm 2L ( Figure 2B ) , several regions show differential coverage . However , all of these correspond to regions of the Drosophila genome that are known to be either amplified in follicle cells ( Claycomb and Orr-Weaver , 2005 ) or under-replicated in adult polytene cells ( Nordman et al . , 2011 ) , which are present in the abdomen but not in the brain . Thus , these copy number variations are not tumor-specific , but rather due to comparing different tissues ( Figure 2B , C ) . Besides that , tumor brains showed no recurrent amplifications or depletions of large chromosome regions , although the detection of afore mentioned biological phenomena shows the ability to detect gross chromosomal aberrations . Similarly , the few detected somatic point mutations did not overlap between the three tumor samples ( Figure 2D ) . Small insertions and deletions ( InDels ) were mainly restricted to introns or intergenic regions ( Figure 2E ) and we did not identify regions ( length 10000 bp ) with such InDels common to all three tumor samples . For example , the two InDels affecting exons were identified in separate tumor samples and did not affect the same gene ( Figure 2F ) . Our data of primary bratk06028 tumors did not reveal recurrent DNA alterations among several tumors nor could we detect abnormal karyotypes . These results suggest that malignant transformation is unlikely to be regulated through a stepwise acquisition of DNA mutations . Thus , the cell-autonomous events leading to highly proliferative tNBs seem to be of transcriptional or post-transcriptional nature . To identify genes involved in brain tumorigenesis , we determined the transcriptomes of FACS-sorted tNBs and NBIIs ( Figure 3A ) . As larval fly brains contain only 16 NBIIs , we developed a method that combines transposase fragmentation with molecular barcodes ( DigiTAG ) to derive high quality transcriptome data from low amounts of input RNA ( Figure 3B ) . In total , we found 1372 up- and 345 downregulated genes in tNBs compared to NBIIs ( FDR 0 . 01 , log2fold change ≤ −2 and ≥2 , respectively ) ( Figure 3C ) . As a quality control , we successfully verified transcriptional changes of brat and fifteen genes with a range of different expression levels by quantitative PCR ( qPCR ) ( Figure 3—figure supplement 1A–C ) . In contrast to previous whole brain expression datasets ( Carney et al . , 2012 ) , using defined cell populations in our transcriptome analysis allowed us to identify tNB-specific expression changes . The upregulated gene most highly expressed in tNB was CR43283 ( Figure 3C ) , which encodes three alternatively spliced polyadenylated transcripts without any long open reading frames . Peptides of short open reading frames found in CR43283 were not detected in the brat tumor proteome ( Jüschke et al . , 2013 ) ( data not shown ) and PhyloCSF analysis of CR43283 did not reveal any coding potential ( Figure 3D ) . Thus , CR43283 likely acts as a lncRNA . To indicate its brat antagonizing activity ( see below ) , we renamed CR43283 into cherub , the antonym for brat . To analyze cherub’s function , we inserted FRT sites at the 5’ and 3’ ends of the longest predicted transcript and generated a deletion by Flp recombination ( cherub DEL ) ( Figure 4—figure supplement 1A–C ) . Our phenotypic analysis also included a promoter deletion created using CRISPR-Cas9 ( cherub promDEL ) and an RNAi construct ( Figure 4—figure supplement 1D ) to exclude artefacts from deleting DNA elements . cherub mutants were viable and fertile , as numbers of eggs laid per female and of emerging adults were similar to control flies ( Figure 4A , B ) . Furthermore , cherub mutant flies showed normal behavior in a geotaxis assay ( Figure 4C ) . NBII lineages lacking cherub appeared normal ( Figure 4D ) and contained one large NBII expressing Deadpan ( Dpn ) but not Asense ( Ase ) surrounded by Dpn-Ase- immature and Dpn+Ase+ mature INPs ( Figure 4E , F ) . Nevertheless , in a cherub mutant background neither brat mutations nor brat RNAi resulted in the formation of large brain tumors ( Figure 5A , B , Figure 5—figure supplement 1A ) . Both , tumor volumes and numbers of Dpn+ tNBs were strongly reduced ( Figure 5—figure supplement 1B ) . The reduction in tumor growth resulted in increased viability ( Figure 5C , Figure 5—figure supplement 1C ) . A genomic rescue construct was able to again increase the number of Dpn+ tNBs in cherub brat depleted brains ( Figure 5—figure supplement 1D ) . Thus , cherub is required for tNBs to form large brain tumors . Since the subcellular localization of lncRNAs is often associated with their function ( Chen , 2016 ) , we visualized the expression of cherub in NBII lineages using a single-molecule FISH technique . cherub was expressed in all central brain and ventral nerve cord NB lineages ( Figure 6—figure supplement 1A ) . In NBIIs , FISH probes against cherub stained two nuclear dots ( Figure 6A ) representing nascent transcripts ( Levesque and Raj , 2013 ) . In addition , the lncRNA was enriched at the cell cortex and was only weakly detected in the cytoplasm ( Figure 6A ) . The highest levels of cherub were found in newborn INPs , where it was cytoplasmic ( Figure 6A , B ) . Over time , cherub expression in INPs decreased and the lncRNA was only weakly detected in GMCs and neurons ( Figure 6B ) . The majority of NBs cease their proliferation in early pupal stages and terminally differentiate ( Homem et al . , 2015 ) . Consequently , adult brains showed no expression of cherub ( Figure 6—figure supplement 1B ) . Notably , this expression pattern and the localization of cherub was conserved in other Drosophila species ( Figure 6—figure supplement 1C , D ) . Remarkably , the cortical pool of cherub localized asymmetrically in mitotic NBs , opposite to the apical marker aPKC and thus segregated into the INP upon cytokinesis ( Figure 6C ) . To avoid false positive cherub crescents due to surrounding cherub high daughter cells , we additionally confirmed the basal localization of cherub in isolated NBIIs in vitro ( Figure 6D , E , Video 1 ) . Among all three cherub isoforms , only isoform RC can be unambiguously detected . However , the combination of FISH probes against specifically RC and regions present in RA and RC or RB and RC confirmed that all isoforms showed the same localization pattern and were asymmetrically segregated in NBIIs during mitosis ( Figure 6—figure supplement 1E–K ) . Previous work has shown that during mitosis the apical aPKC-Par complex releases Lethal ( 2 ) giant larvae ( Lgl ) from the apical side of the cell , which then promotes the localization of factors to the basal cell pole ( Betschinger et al . , 2003; Lee et al . , 2006c ) . To determine whether this mechanism is involved in cherub polarization during cell division , components of the apical as well as of the basal domain were misexpressed or knocked down by RNAi . While expressing a constitutively active form of Lgl ( Lgl3A ) led to a uniform cortical localization of cherub , the overexpression of a constitutively active form of aPKC ( aPKC∆N ) resulted in cytoplasmic cherub ( Figure 6F ) . Additionally , cherub transcripts became enriched in the nearest daughter cells of the NBII in control brains , but upon aPKC∆N expression cherub concentrations were comparable between both cells ( Figure 6G ) , indicating symmetric distribution of cherub . From these results , we conclude that the canonical asymmetric cell division machinery establishes the unequal partitioning of cherub between daughter cells . Basal cherub polarity depended on Miranda ( Figure 6F , G ) , a known adaptor protein required to tether proteins to the basal part of a NB’s plasma membrane . As Miranda does not possess any RNA-binding domains , one of the proteins transported by Miranda might localize cherub . The Miranda-associated RNA-binding protein Staufen ( Matsuzaki et al . , 1998; Schuldt et al . , 1998; Shen et al . , 1998; Slack et al . , 2007 ) has been described to enrich at the basal cell pole of dividing embryonic and larval NBs ( Broadus et al . , 1998; Jia et al . , 2015; Li et al . , 1997 ) and indeed it also localized asymmetrically in dividing NBIIs ( Figure 6—figure supplement 2A ) . Consequently , Staufen was enriched in the most recently born iINPs and declined upon INP differentiation ( Figure 6—figure supplement 2B ) . In staufen depleted NBIIs , cherub was no longer cortical , failed to segregate asymmetrically and accumulated in the cytoplasm ( Figure 6H , I , Figure 6—figure supplement 2C ) . qPCR on RNA , isolated from anti-Staufen immunoprecipitates , detected cherub but not the highly expressed control RNA RpL32 ( Figure 6J ) . In-silico analysis of cherub transcripts showed limited sequence similarity across Drosophila species ( Figure 6—figure supplement 3A , B ) , but revealed four thermodynamically stable and evolutionary conserved secondary RNA structures ( Figure 6—figure supplement 3A , C ) . The identified structures resemble stem loops previously described for Staufen binding ( Ferrandon et al . , 1994; Laver et al . , 2013 ) and hence may constitute potential binding sites . Notably , in some species base-pairing nucleotides were mutated in such a way to still allow base-pairing ( Figure 6—figure supplement 3C ) . This indicates that the structure rather than the sequence is preserved , which is in accordance with the fact that Staufen binds double-stranded structures and not a specific sequence motif ( Ramos et al . , 2000 ) . Therefore , binding to Staufen is required for segregating cherub into INPs to prevent its accumulation in NBIIs over time ( Figure 6K ) . To understand how high levels of cherub emerge in brat tumors , we utilized the ability to detect both , nascent nuclear and cortical cherub by FISH . Intensity measurements of nuclear cherub dots , which represent transcribed RNA ( Figure 7—figure supplement 1A ) , did not reveal any significant increase in brat RNAi tNBs ( Figure 7A ) . This suggests that high levels are not due to transcriptional upregulation . Cortical cherub , in contrast , was strongly enhanced in brat tNBs ( Figure 7B , Figure 7—figure supplement 1B ) and still asymmetrically inherited by one daughter cell in the majority of tNBs ( Figure 7C , D ) . To understand how cherub high tNBs arise , we used a temperature sensitive system to induce brat RNAi for a defined time to follow cherub accumulation upon brat depletion over time . After 24 hr , brat RNAi NBIIs form iINPs that do not differentiate into mINPs and therefore do not re-enter the cell cycle ( Bowman et al . , 2008 ) . In these brat depleted INPs , cherub remained highly expressed and cortical , whereas it became cytoplasmic in control INPs ( Figure 7E ) . After 48 hr , brat RNAi iINPs revert into supernumerary Dpn+ , Ase- NBIIs and initiate tumor formation . Whereas control INPs started to downregulate cherub , these extra tNBs had strongly elevated cortical cherub ( Figure 7F , G ) , presumably because the lncRNA was not released into a differentiating cell in which cherub levels gradually decrease . Hence , cherub accumulates in tNBs , because these supernumerary NBs arise from the daughter cells that carry high levels of cherub . Over time , Drosophila NBs express distinct sets of genes , which confer distinct temporal identities . A temporal fate allows NBs to generate neurons with different axonal projection patterns during development and is important for their timely cell cycle exit ( Homem et al . , 2014; Kohwi and Doe , 2013; Liu et al . , 2015; Maurange et al . , 2008 ) . Moreover , the molecular NB clock has been shown to be important for generating malignant tumors . When tumors are induced early in development tumor cells are generated that drive tumor proliferation even at later developmental stages due to the failure to turn-off early growth-promoting genes ( Narbonne-Reveau et al . , 2016 ) . Although tNBs used for transcriptome analysis were collected at late larval stages , they still expressed genes usually restricted to early NBs ( 24–50 hr after larval hatching ) ( Liu et al . , 2015 ) , among them the early NB identity gene Imp ( Figure 8A ) . Temporal identity in NBIIs is controlled by opposing gradients of the RNA-binding proteins Imp and Syp ( Figure 8B ) ( Ren et al . , 2017; Syed et al . , 2017 ) . The transition from early Imp-positive to a late Syp-positive NB state is triggered by the nuclear receptor Seven-up ( Svp ) ( Narbonne-Reveau et al . , 2016; Ren et al . , 2017; Syed et al . , 2017 ) . To investigate whether tNBs might be defective in temporal NB identity , we analyzed the temporal gene expression signature of brat mutant tNBs . Indeed , antibody staining for the early temporal factor Imp confirmed that brat tumors contain tNBs of early temporal identity ( Figure 8C ) . Interestingly , the formation of tNBs with an early temporal identity was dependent on cherub as upon the depletion of cherub Imp-positive tNBs are no longer present ( Figure 8C ) . On the other hand , the late temporal identity factor Syp was also present in tNBs , but rather than being mainly cytoplasmic in late NBs as previously described ( Liu et al . , 2015 ) , Syp was enriched at the cell cortex ( Figure 8D ) . brat tumors depleted of cherub still expressed Syp , however Syp was no longer enriched at the cell cortex ( Figure 8D ) . Thus , brat mutant tNBs fail to fully progress to the late stages of temporal identity in a cherub-dependent manner . We hypothesized that if temporal identity defects contribute to tumor proliferation , ‘aging’ or ‘rejuvenating’ tNBs should have an effect on tumor growth and survival of tumor-bearing flies . Advancing tNBs’ temporal identity by svp overexpression ( Figure 8—figure supplement 1A–C ) or Imp depletion by RNAi ( Figure 8E , G , I ) significantly reduced tumor growth and increased the median survival time ( 50% survival rate ) of tumor-bearing flies . Conversely , shifting temporal identity of brat RNAi tNBs towards younger stages by svp misexpression ( Figure 8—figure supplement 1A ) or Syp RNAi ( Figure 8 - F , H , J ) significantly decreased the median survival rates indicating faster tumor growth . Thus , temporal identity defects are involved in brain tumor growth . Since the cortical localization of Syp is cherub-dependent in tNBs , it is tempting to speculate that cherub acts as an adapter to facilitate the binding between Staufen and Syp . This assumption implies that firstly Syp shows a similar localization pattern to Staufen and cherub and secondly Syp segregates asymmetrically via Staufen into iINPs . Indeed , in NBIIs Syp was localized at the cortex and also detected in the cytoplasm ( Figure 9D , E ) . Similar to Staufen and cherub , Syp segregated asymmetrically in mitosis ( Figure 9A–C ) , and became enriched in the cytoplasm of the most recently-born INPs ( Figure 9D , E ) . Furthermore , Syp colocalized with Staufen and cherub in mitotic NBIIs ( Figure 9—figure supplement 1A–G ) . In cherub mutants or upon staufen knockdown , Syp was exclusively cytoplasmic and no longer cortical ( Figure 9D , E ) . These results indicate that Syp is recruited to the cortex by the cherub-Staufen complex . To further confirm the role of cherub in Staufen-Syp complex formation , we performed immunoprecipitation experiments on brain lysates from brat tumors . Co-immunoprecipitation experiments demonstrated that Staufen and Syp are in a complex and binding between the two proteins was lost upon RNA digestion or cherub depletion ( Figure 9F , G ) . RIP-qPCR experiments further confirmed that cherub was enriched upon HA-tagged Syp pull-down ( Figure 9H ) . These data suggest that cherub functions as adapter between Syp and Staufen ( Figure 9I ) .
It is commonly assumed that cancer cells become malignant and gain replicative immortality by acquiring genetic lesions ( Hanahan and Weinberg , 2011 ) . Surprisingly , however , our data indicate that brat tumors do not require additional genetic lesions for the transition to an immortal state . This is not a general feature of Drosophila tumors as genomic instability alone can induce tumors in Drosophila epithelial cells ( Dekanty et al . , 2012 ) and intestinal stem cells ( Siudeja et al . , 2015 ) . However , our results are supported by previous experiments demonstrating that defects in genome integrity do not contribute to primary tumor formation in NBs ( Castellanos et al . , 2008 ) . Similarly , tumors induced by loss of epigenetic regulators in Drosophila wing discs do not display genome instability ( Sievers et al . , 2014 ) . In addition , the short time it takes from the inactivation of brat to the formation of a fully penetrant tumor phenotype would most likely be insufficient for the acquisition of tumor-promoting DNA alterations . More likely , the enormous self-renewal capacity and fast cell cycle of Drosophila NBs requires only minor alterations for the adoption of malignant growth . Interestingly , epigenetic tumorigenesis was described before in humans , where childhood brain tumors only harbor an extremely low mutation rate and very few recurrent DNA alterations ( Lee et al . , 2012; Mack et al . , 2014; Northcott et al . , 2012; Parsons et al . , 2011; Pugh et al . , 2012; Robinson et al . , 2012; Sausen et al . , 2013; Wu et al . , 2014; Zhang et al . , 2012 ) . Comparable observations have been made for leukemia ( Quesada et al . , 2011; Yan et al . , 2011 ) . Our results might help to understand mechanisms of epigenetic tumor formation , which are currently unclear in humans . cherub is the first lncRNA described to segregate asymmetrically during mitosis . Once cherub is allocated through binding to the RNA-binding protein Staufen into the cytoplasm of INPs , its levels decrease over time . Our results show that the inability to segregate cherub into differentiating cells leads to its accumulation in tNBs . The increasing amount of tumor transcriptome data indicates that a vast number of lncRNAs show increased expression levels in various tumor types ( Huarte , 2015 ) . Intriguingly , the mammalian homologue of cherub’s binding partner Staufen has been also described to asymmetrically localize RNA in dividing neural stem cells ( Kusek et al . , 2012; Vessey et al . , 2012 ) . Hence , besides transcriptional upregulation , asymmetric distribution of lncRNAs between sibling cells might play a role in the accumulation of such RNAs in mammalian tumors . Our data suggest a functional connection between cherub and proteins involved in temporal neural stem cell patterning . We show that tNBs retain the early temporal identity factor Imp even during late larval stages . However , Imp expression in brat mutants is heterogeneous and only a subpopulation of tNBs maintains young identity . Tumor heterogeneity has also been described for pros tumors , where only a subset of tNBs maintains expression of the early temporal factors Imp and Chinmo ( Narbonne-Reveau et al . , 2016 ) . Interestingly , it is this subpopulation that drives tumor growth in prospero tumors . Consistent with this , our genetic experiments show that ‘rejuvenating’ tNBs enhances tumor growth and consequently increases the survival of tumor bearing flies , whereas ‘aging’ tNBs identity has the opposite effect . Although mammalian counterparts of Imp have not yet being shown to act as temporal identity genes , their upregulated expression has been implied in various cancer types ( Dai et al . , 2017; Lederer et al . , 2014 ) . Therefore , temporal patterning of NBs has an essential role in brain tumor propagation in Drosophila . The subset of tNBs that retain early identity in tumors is lost in a cherub mutant background . This suggests that cherub itself might regulate temporal identity . In NBs and tNBs cherub regulates Syp localization by facilitating the binding of Syp to Staufen and thus recruiting it to the cell cortex . In tumors depleted of cherub , Syp localizes mainly to the cytoplasm and is no longer observed at the cortex . As the removal of Syp in tNBs leads to enhanced tumor growth and early lethality , those data suggest that cherub could control temporal NB identity by regulating the subcellular localization of Syp . How could cherub regulate the function of Syp ? The RNA-binding protein Syp is a translational regulator ( Duning et al . , 2008; McDermott et al . , 2012; 2014 ) and has been suggested to control mRNA stability ( Yang et al . , 2017b ) . As mammalian SYNCRIP/hnRNP Q interacts with a lncRNA that suppresses translation ( Duning et al . , 2008; Kondrashov et al . , 2005 ) , cherub might regulate Syp to inhibit or promote the translation of a subset of target mRNAs . In particular , in NBs Syp acts at two stages in NBs during development: Firstly , approximately 60 hr after larval hatching it represses early temporal NB factors , like Imp ( Syed et al . , 2017; Yang et al . , 2017a ) . Secondly , at the end of the NB lifespan Syp promotes levels of the differentiation factor prospero to facilitate the NB’s final cell cycle exit ( Yang et al . , 2017a; Yang et al . , 2017b ) . As cherub depletion in brat tumors leads to decreased tumor growth ( Figure 5A , B ) , it is possible that cherub inhibits the Syp-dependent repression of the early factor Imp , which we showed is required for optimal tumor growth ( Figure 8E , G ) . However , cherub mutant NBIIs do not show altered timing or expression of Imp during development ( data not shown ) . In accordance , brat tumors show high cortical cherub levels , but only a subset of NBs expresses Imp ( Figure 8C ) . Rather than rendering Syp completely inactive , we suggest that cherub decreases the ability of Syp to promote factors important to restrict NB proliferation . As prospero is not expressed in NBIIs , it remains to be investigated which Syp targets are affected by cherub . Remarkably , cherub mutants are viable , fertile and do not affect NBII lineages . Neurons generated by NBIIs predominantly integrate into the adult brain structure termed central complex , which is important for locomotor activity . As cherub mutants show normal geotaxis , function of the lncRNA seems dispensable for NBIIs to generate their neural descendants . Nevertheless , the conserved secondary RNA structures of cherub and its conserved expression pattern in other Drosophila species suggest that it has a functional role . There are several possibilities why we do not observe a phenotype upon the loss of cherub . In wild-type flies cherub might confer robustness . A similar scenario was observed in embryonic NBs , in which Staufen segregates prospero mRNA into GMCs ( Broadus et al . , 1998; Schuldt et al . , 1998 ) . The failure to segregate prospero mRNA does not result in a phenotype , but it enhances the hypomorphic prospero GMC phenotype ( Broadus et al . , 1998 ) . Thus segregation of prospero mRNA serves as support for Prospero protein to induce a GMC fate . Similarly , cherub could act as a backup to reliably establish correct Syp levels in NBIIs and in INPs . Alternatively , cherub might fine-tune the temporal patterning by regulating the cytoplasmic pool of Syp in the NBs . Increasing Syp levels have been suggested to determine distinct temporal windows , in which different INPs and ultimately neurons with various morphologies are sequentially born ( Ren et al . , 2017 ) . Therefore , we cannot exclude that changes in Syp levels lead to subtle alterations in the number of certain neuron classes produced by NBIIs that only reveal themselves in pathological conditions like tumorigenesis . Our study illustrates how a lncRNA can control the subcellular localization of temporal factors . In addition to temporal NB identity , Syp regulates synaptic transmission and maternal RNA localization ( McDermott et al . , 2012 ) . While cherub is not expressed in ovaries or adult heads , Staufen has been implicated in these processes , suggesting that other RNAs might act similarly to cherub . Interestingly , the mammalian Syp homolog hnRNP Q binds the noncoding RNA BC200 ( Duning et al . , 2008 ) , whose upregulation is used as a biomarker in ovarian , esophageal , breast and brain cancer ( De Leeneer and Claes , 2015; Perez et al . , 2008; Zhao et al . , 2016 ) ( De Leeneer and Claes , 2015; Perez et al . , 2008; Zhao et al . , 2016 ) . In the future , it will be interesting to investigate whether the mechanism we have identified in Drosophila is involved in mammalian tumorigenesis as well .
The following Drosophila stocks were used: UAS-brat RNAi ( VDRC , TID 105054 and 31333 ) , UAS-mira RNAi ( Betschinger et al . , 2006 ) , UAS-mCherry RNAi ( BL35785 ) , UAS-lgl 3A ( Betschinger et al . , 2003 ) , UAS-aPKCΔN ( Betschinger et al . , 2003 ) , UAS-staufen RNAi ( VDRC , TID106645 ) , UAS-Syp RNAi ( VDRC , TID33012 ) , UAS-Syp-RB-HA ( Liu et al . , 2015 ) , UAS-Imp RNAi ( BL34977 ) , UAS-svp RNAi ( VDRC , TID 37087 ) , UAS-svp ( gift from Y . Hiromi ) , brat k06028 ( Arama et al . , 2000 ) , UAS-dcr2; wor-GAL4 , ase-GAL80; UAS-CD8::GFP ( Neumüller et al . , 2011 ) , UAS-dcr2; insc-GAL4 , UAS-CD8::GFP ( Neumüller et al . , 2011 ) , UAS-stinger::RFP and UAS-stinger::GFP ( Barolo et al . , 2000 ) , tubulin-GAL80ts ( BL7017 ) , D . simulans ( Drosophila Species Stock Center , 14021–0251 . 265 ) , D . willistoni ( Drosophila Species Stock Center , 14030–0814 . 10 ) . Stocks generated in this study: UAS-cherub RNAi , cherub DEL , cherub promDEL , cherubgenomic rescue . For the latter , the BAC construct CH322-116G10 was integrated into the attP40 landing site via integrase-mediated transgenesis . If not stated otherwise , UAS transgenes were expressed at 29°C . For time-dependent induction of brat RNAi crosses were reared at 18°C for 6 days and then shifted at 29°C for 24 or 48 hr . For measuring cherub intensities upon aPKCΔN and miranda RNAi expression , crosses were left for 8 days at 18°C , then shifted to 29°C for 43 hr to be able to identify separate lineages and nearest daughter cells . For survival measurements , flies were raised , collected 0–3 days after eclosion and kept at 29°C . Surviving flies were counted and moved to fresh non-yeasted vials every 2–3 days . UAS-mCherry RNAi was used as control for experiments involving UAS-transgene expression , whereas w1118 was used as control for comparison with mutants . For temperature-induced RNAi knockdown , the driver line was crossed to tubulin-GAL80ts as control . To create a full deletion ( cherub DEL ) , FRT sites were inserted upstream and downstream of the locus using the CRISPR-Cas9 system ( Gokcezade et al . , 2014 ) and single-stranded oligonucleotide donors . The gRNAs used were TGGCGTCGGTTCGACCGATC and ATGAAAGTGTGAATCTTCCA . Single-stranded oligonucleotide donors were ATCCTGGCAGACAATGGACAAAGCTCTAGCATCCTGATTGCGATCGGATCGCTTGGCGTCGGTTCGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTCTGGCGGGTATATAAACTGCGGCTGCTGCGCAGAATCAATCAGTTTCATTTCAATCTTCAAACGCTGA and CTTTTACTTAACTGTGCTATTATTAAGTGAGGATATTTGGAAAAGGGATTCCAAATGAAAGTGTGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTCCAAGGGATATTTACGAAATCTGTAATAATGGTCACCACTTCTTCAAATGGTAAGAAAAAATTAA . The FRT-flanked locus was deleted by Flp-FRT recombination ( Golic and Golic , 1996 ) . Two gRNAs ( GGCGTCGGTTCGACCGATC , GCCTGGACATGGCGCTGCG ) were used to create a 180 bp deletion covering the promotor of cherub ( cherub promDEL ) . Mutants were verified by Sanger sequencing and PCR . Generation of a short hairpin RNA line was performed according to the Transgenic RNAi Project’s protocol ( Ni et al . , 2011 ) . Briefly , oligonucleotides were annealed and cloned into the Walium20 vector . The oligonucleotides used were CTAGCAGTAGACATATGGTTACTGCTCGATAGTTATATTCAAGCATATCGAGCAGTAACCATATGTCTGCG and AATTCGCAGACATATGGTTACTGCTCGATATGCTTGAATATAAC TATCGAGCAGTAACCATATGTCTACTG . To assess female fecundity , five-day-old female virgins were crossed to two-day-old males in cages to allow mating for two days . The apple juice plate was replaced and flies laid eggs for 6 hr ( 9-15:00 each day ) . Subsequently , eggs were counted . The procedure was repeated on day 3 and 4 post mating . For each genotype three independent crosses were tested . Flies were allowed to lay eggs for 4 hr on apple juice plates . For each replicate 100 eggs were collected , transferred into fresh food vials and nine days later eclosed adult flies were counted daily for five days . The climbing pass rate , which is the percentage of flies passing the 8 . 5 cm mark in 10 s , was assessed as described before ( Ali et al . , 2011 ) . Each replicate was measured 10 times with a 1 min rest period between measurements . For each genotype and gender , 10 biological replicates consisting each of 10 two-days old adult flies were measured . Larval brains were dissected in PBS , fixed for 20 min at room temperature in 5% paraformaldehyde in PBS and washed three times with 0 . 1% TritonX in PBS ( PBST ) . After 1 hr incubation in blocking solutions ( 1% normal goat serum in PBST ) , brains were incubated with primary antibodies in blocking solution overnight , then washed three times with PBST , incubated for 2 hr at room temperature with secondary antibodies ( 1:500 , goat Alexa Fluor , Invitrogen ) , washed with PBST and mounted in Vectashield Antifade Mounting Medium ( Vector Labs ) . Antibodies used in this study were: guinea pig anti-Deadpan ( 1:1000 , [Eroglu et al . , 2014] ) , rat anti-Asense ( 1:500 , [Eroglu et al . , 2014] ) , guinea pig anti-Syncrip ( 1:500 , [McDermott et al . , 2012] ) , rat anti-Staufen ( 1:100 , [Krauss et al . , 2009] ) , mouse anti-PH3 ( Ser10 ) ( 1:1000 , Cell Signalling Technologies ) , rat anti-Elav ( 1:200 , DSHB 7E8A10 ) , guinea pig anti-Miranda ( 1:500 , [Eroglu et al . , 2014] ) , rabbit anti-aPKC ( 1:500 , Santa Cruz Biotechnology ) , chicken anti-GFP ( 1:500 , Abcam ) , rabbit anti-Imp ( 1:500 , [Geng and Macdonald , 2006] ) . The protocol was adapted as previously described ( Raj et al . , 2008 ) with minor changes . Briefly , brains were fixed for 40 min . After removing the fixative with four washes of PBS , brains were permeabilized in 70% Ethanol overnight at 4°C . Ethanol was removed and 400 µl washing buffer ( 2x SSC , 10% Formamide ) were added . After 5 min at 37°C the wash buffer was substituted by 100 µl hybridization buffer ( 1 mg/ml E . coli t-RNA , 2 mM Vanadyl ribonucleoside complex , 200 nM BSA , 2x SSC , 10% Formamide , 100 mg/ml Dextran sulfate ) including the FISH probes . After overnight incubation at 37°C in the dark , brains were washed twice with washing buffer , each time for 30 min at 30°C . Brains were mounted in 2xSSC . The FISH probe sets against cherub were compromised of oligonucleotides labeled with a Quasar Dye ( Stellaris Biosearch Technologies ) and were designed using the Stellaris probe designer ( https://www . biosearchtech . com/stellarisdesigner ) . DNA staining was achieved by a 20 min incubation with Hoechst ( 1:1000 , Thermo Scientific Pierce Hoechst 33342 ) in washing buffer at 30°C . When antibody staining and FISH were combined , the standard immunohistochemistry protocol was performed first , followed by the FISH protocol without incubation in 70% Ethanol . FISH probe sets against cherub were compromised of the following oligonucleotide sequences: Against all isoforms ( labeled with Quasar 570 Dye ) : GTGTGGAGATGCTGCAACAGGGTTGATTTTGTTCTTCTGTCGACTGAATTTTGTTGGGCTTAATCTGACATTCGGCGGTATGGTACTTGGGTTTCTTTTTGCAAATTGTTCGTTTGGCTTTGCCTTGTTGTGTTTAGTATTTTTTGCTTAAACTTAGGCTTCGCTGACATTGATTTTTTTATGCTTTGATTTCATGTGTTCAAGATGGTATCGTGGTTGTGGTACTTTGGAATTTGGTTTTTGGATATCAGCTTTCCTTATTGGTTTCAGGCGAATTACAGCCAGATGTGTTGTTAACTTATTTGTCATCTCGTGCTATTTGGGGTTCCTGTAAGATATATAGCTTTCCTTTCTTCAGATATATGGTTACTGCTCGATCCGAGTTTGATCCGGATGCAAGACAATTTCGTTGATGGGGATGCAGTAGAGAATTATGAGCAGAATTTCGTTCTTGGCGAGACGATTTTGATAGCAGTTTGGTTTGGCTGGTTGGGAATTTCGCAGTTGCTTAAGACACTGGATTTTTTGCTTGGGGCACAGGATTTCGGTGTGATTTGCTGAATTTTGGTTTCCTTCCTTGTATTTGGTTTGCGCCTGGACCCCCAGCAAAAGTGTACTTGGAGTTCATCATCATCAGCTTTCAAATGCTCGATCCTTCTCCTCCTTTCTCTTGATTATCTCTGCTTAGATATTCAGCTGGTTTTGCTGCTTGGGGTAAATATGTGGGCTGTATGTGTGTGGTGTGTGCTTGTTTGTGAATTCGGAGAGTAAAAGGTGGGC Against isoform RA and RC ( labeled with Quasar 570 Dye ) : GGGGTTACTCAAATCTGTTTTTTTACTTAGTAAGGCTCCAGCTCGGTTTTAGGTTTTTGAAGTGAAATGAGGCTCACCTTGAAATTTTGCTTCTGGGCTGTGGATTTTGAGTTGTGTCGGGCCGTAGAAGAAGAGTTCGTTGGAAATGAAGCCTTCTGCGTTCAGGGATACTTTCTTTGCGGCGAACTCTTTGGTTGAATTATTGCATTGTGTGTGTGGAATTCGATTTGCGTTTTGGGATGACGAAACGCGAATTTGGACATTTGCTTGATTTTTGCATGCCATGCGAAACTGTCTTTTTGGTTGATTTGCGCCTAAAGTGAGTGTATGGGTGAGCATAAAGGGGTGTGAGTGTTTGATCTTTGCAGCCATTTTATTGCTGGGAATCGTAGAAGCAGCTGTCAGGCTATTTCTTTTCTTTTCCTATCGAACTTTTTGGGCCTATAACTCGCTTTCGTTGTAGTCCTTTCGCTAACCTATTAACTGGAGTTTGGTTTGGCAGCGCTGAGAAATGCTTGAGGCCTTTGCTTTCTTTAATGAGTGTGATTAAATAATGCCGCATTTCTATCAGCACTTCTGGTGTACTGTCGTTTGCTGTACCTTAAATGACTTCCTGGGGTTGAGTACTTAAGGGTTGGGGGTCAATGACCAGAAGACATTTTGGTCGGTCGAAGTTGTTTTCGTAATGAGGTTTGCGTTTGCTTGGCAGTATATTGATTTACCGGAATGTAGTATTTAGAAGCGTGGGTAAATAAACCAATGTGATATCGCGTTTTGGTACGATTTTTGCAATTTCGTGATTTGTTTGGCTTGATTTGGACTTCAGTGTGTTGTCGGAAAATTTTCATGATGCATAGCGATTTTGAATTTGGATGCGTGAGTGTGTGCATTATATTTTCCACCTTTTCGGATTTTGAATTGTCAATTTTATGCTCTTTGATGTGACTTTTAACGGTTTGA Against isoform RB and RC ( labeled with Quasar 570 Dye ) : AGTTTTGATTACCTGGTCTGCTGCTTTGTGTGATTTTGGACATTGGTTTGGTTGGACTTTACCGAAGGTTCCTCGAAATGCTCAATTTAATGGGTGGGCTTTTTGAATTTTTGGCGGCTCCACTTCTTTTTTTCGCTCTTTTCCGCAGTTTCTTGTAATTTTTTTGCTCCTGACATTTGTCAGAAAGTTGCCAGGAGTTTCCGCCATTTCTAAGATTTTGTTTCGCTGTGATTTATCTGCCACGAGAGCTGGGAATTGTTTCAGTTTCGGTTTCAGTTTTGCAACCCTCTTCTTTGAAATGGATCAAAGTGGTCATGGTCGGCGAAAGTAATGGGCTTGAATTTGTGATTTTTGTGCGGCTGGGTTAAGTTGCAGTTTGTTGCCTTCATTTTGAGTTTGATTAAAGTTGCGGTTCTGCTGTGACACTCGTGCCGAATATTTGCCTGGAGTTTTAAGTTTCCCAAGTCTGACCACAATTTCATGGTTCGAACTTTTGCCAATTGTTTGATGTGCTTTCAGCGCTATTGCCAGTAGTTTATAATGCCAGCCATATTTTATTTGGCAACTTTTGCCATGAATTACGGTTTTTGCAGTTTCTTGGGTGATTTCGAGGGATTTTCTTGGTTTTTTCTTTTTGGGCGGTTGCTTGAATTTGCTTGACTAGCCAAAGGTGGGAAGATGCCCGAAACTTTTTTGGTAAAGATTCCATTTGCTACTCTTAGCTTTCTTTGCTTTTCTTTTATACTCGCATACTACCTGTACCTTGTGTCCGAAATGTTATTGGGTCTAAAGAGGCCTAATTCCTGTTCTTTCTTCATTTGTGGTACAATTCGCGCTTAAGTTTTCCTTTCTTTAGCATTATTCCGCAATTCAATTGCGAAAATGCGGCAATCGTTTCGTCGCTTATCACATGTGTCGAATGGTTCTTTTTTTTGTCCTGTGTTTTTGAGGTTGTACGGT Against isoform RC ( labeled with Quasar 670 Dye ) : GAATTTCTATCGTGCAATCAACCTTTTCGGATTTTGAATTGTCAATTTTATGCTCTTTGATGTGACTTTTAACGGTTTGAGCAGTTTTTAATTTGGTAGTACGCGATAATTCATTTTGCTGCTGATATGTTTTTGTTTGCACATTGACATTTCTTTGCTTGTATTTCTTGGTTTGAGTTTACTGCGGTTTTTGATTGATTGTTGTTTTTGCTCTTTTTGAGTTCGTATAATGTCTTGCTCATTCGGTTCGAACTTTTGCTATCTGACATTTTACTTGGTGTTCCAATAGGGCTTTTTGTTATGTGTGTTTTGATTGTTCTGCTATCTTCATTTTGTTTCTGTTCATAGTTTTCAGTCTGAGCCTTTTGAATTGCAATCATTTTTCTCAGAACTGGTTTCATCCTGATGCATTAAGACTCTACGTTTCCATAATTGCATAAGTTCCGTTCTTTACCGGATATTTGATTTTTGTATTGCGGTAGACTGGAAGTAGTTACCTTGTAAAATGGTGGCACATGTTTTTGGTTTTCTAGACCATTTTTTTGAGATTCAAATCGGGTGTGTTATTTAAGGCATTTGGATTTGGTTTGTCTGGTTCTACAGAGGAGCCGAAAGTTGAAGGTTTTCGGTTTTCAAATTT For confirming RNA specificity of the probe set against cherub RNA , the above mentioned standard protocol was performed except that brains were incubated for 2 hr at 37°C with RNase A ( 100 μg/ml ) before the incubation step in 70% Ethanol or with RNase H ( 100 U/ml ) after hybridization in order to let RNA/DNA duplexes be formed first . Cells from 25 to 35 dissociated larval brains ( UAS-dcr2; wor-GAL4 , ase-GAL80; UAS-CD8::GFP ) were plated on cell culture dishes and incubated in complete Schneider’s medium ( Homem et al . , 2013 ) with Colcemid ( 25 µM ) for 5 hr at 25°C . After a 15 min fixation with 5% PFA in PBS at room temperature , cells were incubated with 3% normal goat serum in 0 . 1% TritonX in PBS overnight , one hour with primary antibodies and one hour with secondary antibodies . Between steps cells were washed with PBS . Cells were again fixed for 20 min , washed with FISH washing buffer and FISH was performed with a probe incubation time of 4 hr at 37°C . After one wash with washing buffer for 30 min at 37°C , cells were imaged in 2xSSC . NBIIs were unambiguously identified by a size >10 µm and GFP expression . Confocal images were acquired with Zeiss LSM 780 confocal microscopes . For intensity measurements of cherub foci , dot areas were marked on the plane of a z-stack with the largest diameter . Raw integrated density ( sum of grey values of all selected pixels ) was measured using FIJI . To determine intensity ratios in Figure 6G and I , cytoplasmic cherub of interphase NBIIs and closest daughter cells were measured as raw integrated intensities ( I ) using FIJI . Ratios were calculated as ( I daughter cell/area daughter cell ) / ( I NBII/area NBII ) . For intensity quantifications in cross sections of cells , FIJI’s plot profile function was used and the cross section length of each cell was scaled to 100% . Cortical cherub intensity in cells of 72 hr clones were measured by determining the cortex via cortical Miranda staining and measuring raw integrated intensities , which were normalized to the cortical area . For the 3D movie , Imaris was used to 3D reconstruct a z-stack of one NBII from an in vitro FISH/IF experiment . To measure the relative position of cherub/Syp and aPKC crescents , a perpendicular line was drawn through the center of each crescent , which was defined by bisecting the connecting line between the crescent’s ends ( see scheme in Figure 6E ) . The angle between the two bisecting lines was measured using FIJI . An angle of 180°C is measured when cherub/Syp and aPKC form crescents opposite of each other ( this is equivalent to a cherub/Syp crescent perpendicular to the mitotic spindle ) . For quantifications of tumor volumes , one brain lobe per brain was imaged and used for quantifications . To evaluate colocalization , z-stacks were recorded of in vitro mitotic NBIIs using 63x immersion oil objective with optimal pixel size and z-stack distance . Pearson’s coefficient and Li’s intensity correlation analysis ( ICA ) were calculated using the FIJI plugin JACoP ( Bolte and Cordelières , 2006 ) . The covariance of intensity is calculated as ( pixel intensity Ai – mean intensity A ) ( pixel intensity Bi – mean intensity B ) , A and B being the respective channels . ICA plots were made in R . Statistical analyses were performed with GraphPad Prism 7 . Unpaired two-tailed Student’s t-test was used to assess statistical significance between two genotypes/conditions and one-way ANOVA for comparison of multiple samples . No statistical methods were used to predetermine the sample size . Sample sizes for experiments were estimated based on previous experience with a similar setup that showed significance . Experiments were not randomized and investigator was not blinded . RIP was performed as previously described ( Gilbert and Svejstrup , 2006 ) with minor modifications . Shortly , dissected brains were cross-linked with 0 . 5% Formaldehyde , quenched with Glycin ( final 0 . 125 M ) and homogenized in RIP lysis buffer ( 50 mM Hepes pH 7 . 5 , 140 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 0 . 1% Sodium-deoxycholate , 1x protease inhibitor , 1 mM PMSF , 40 U/ml RNasin ) . After DNA removal using DNaseI , protein lysates were incubated overnight with antibodies ( goat anti-Staufen dN-16 antibody , Santa Cruz Biotechnology or mouse anti-HA antibody clone 12CA5 , in both cases 10x blocking peptide was used ) at 4°C and subsequently incubated with Protein G Dynabeads ( Thermo Scientific Fisher ) for 1 hr . Eluted Protein-RNA complexes were treated with Proteinase K . After RNA extraction , pull-down of RNAs was detected by qPCR . RNA was isolated by using TRIzol reagent ( Ambion ) or acidic Phenol/Chloroform ( 5:1 , pH4 . 5 , Ambion ) for RIP samples . Ethanol-precipitated RNA samples were then used as template for first-strand cDNA synthesis with random hexamer primers ( SuperScriptIII , Invitrogen ) . qPCR samples were prepared with Bio-Rad IQ SYBR Green Supermix and run on a Bio-Rad CFX96 cycler . Primers used were: cherub - all isoforms ( AGCAGCACCAGCAGCAGTAG , GCGGTGGATTTGGTTGATTT ) , cherub - RC isoform ( TCAAAAGGCGATGAAACCAGT , ATTGCGGTTTGTTCCGTTCT ) , brat ( CACAAGTTCGGGTGCTCTAA , CCGATTGTCGCTGATGAAGA ) , sle ( GAGTCCGTTGGCAGTAAAGATA , CTCGTCTTCGTTGTCCGATAC ) , CG42232 ( GAAGATGGCGGTGAAGTAGAA , GGCCTGTAGAGCTGGAATTAG ) , CycG ( CACTACACTCACCCTTGATTCC , CGAGTTGTACGAAACCCTCAA ) , Gbs-76A ( GTCCACATCTACGGTGAGATTAC , ACCAGAGGAAAGCAGGAATG ) , CG13185 ( GTCTGGAGTTCGATCAGGAAAG , GTCGGAAGCATCTGGTGTATAG ) , spen ( GAAGAGCGGCATCGACTAAA , GGCAAAGAAGGTGAGGTAGAA ) , Su ( var ) 2-HP2 ( TCCTTGGGATTCGGGAAATG , GAGGCTGCTACTGAGCTAATG ) , tai ( ACTACGGTGGCTTCAACTTC , TGGATTGCTACTGCTGCTATT ) , CG32479 ( GCAAGTCCCACAGCAACTAT , CTGCGGATTGGCTGATGAA ) , Su ( Tpl ) ( GGTACTCATCGTAGTCGCTTTC , CGCTACGACTTCAGCCAATA ) , Taf12L ( ACAGCGATAAATCGTCGGATAA , GGACAGACTGGCTCTCAATTAC ) , CG41128 ( TTAAAGGATGTGGAGGCGTAAT , TCCTATAAGCGATGCCCATTC ) , Hsp67Ba ( GCCAGCAATCTCCCACTATT , AATAATCTGCACGGGTAGGC ) , CG2021 ( CATGAGCGCGTCTTCTCTAC , AGTCGATGGTCTCGTCTATCA ) , CG11882 ( ACTAACAGCGTCAGCTTCTC , GAGCCTGATGAAGGGCTATT ) . Gene expression was normalized to Act5C ( AGTGGTGGAAGTTTGGAGTG , GATAATGATGATGGTGTGCAGG ) . Primers used for RIP-qPCR were: RpL32 ( GCCGCTTCAAGGGACAGTAT , TTCTGCATGAGCAGGACCTC ) , cherub - all isoforms ( AGCAGCACCAGCAGCAGTAG , GCGGTGGATTTGGTTGATTT ) . Primers for RT-PCR ( 30 cycles ) used to detect expression in different Drosophila species were: for GPDH: D . melanogaster ( AACTTCTGCGAAACGACAAT , CGTAACACGTCGTGATCAG ) , D . simulans ( AACTTCTGCGAAACGACAAT , CGTAACACGTCGTGATCAG ) , D . willistoni ( ATACCATGCGCCGTACTG , CATAACACGTCGTGATAAGATCC ) , for cherub D . melanogaster ( AGCAGCACCAGCAGCAGTAG , GCGGTGGATTTGGTTGATTT ) , D . simulans ( GAGTAGGAGCCGCACAGGAG , CGGTGTGGAGATGCTGCAAC ) , D . willistoni ( GGAAGGATCTATGCAGAGAGAGACA , CCCCAACCTTCTTGTGTCCG ) . Immunoprecipitation was performed according to the RIP protocol except omission of formaldehyde fixation , DNaseI treatment steps and instead of RNA isolation samples were boiled in 2x Laemmli buffer to elute protein complexes and loaded on 3–8% gradient Tris-Acetate gels ( NuPAGE , Invitrogen ) . After SDS-PAGE according to Invitrogen’s protocol , proteins were transferred to a Nitrocellulose membrane ( 0 . 22 µm , Odyssey LI-COR ) for 2 hr at 100V , blocked with 5% milk powder in blocking solution ( PBS with 0 . 2% Tween ) for 1 hr , overnight incubation with primary antibody in blocking solution at 4°C , 3x washed with washing solution ( PBS with 0 . 1% Tween ) and followed by 1 hr incubation with secondary antibody ( HRP-linked Whole Antibodies from GE Healthcare ) in blocking solution . After three washes with washing solution , horseradish peroxidase activity was detected with Pierce ECL Plus ( Thermo Fisher Scientific ) . Antibodies used were: goat anti-Staufen ( 1:2000 , dN-16 Santa Cruz Biotechnology ) , guinea pig anti-Syncrip ( 1:3000 , ( McDermott et al . , 2012 ) ) , mouse anti-HA ( 1:500 , clone 12CA5 ) . A Multiz alignment of 27 insecta ( 23 Drosophila species , house fly , Anopheles gambiae and mellifera , honey bee and Tribolium castaneum ) aligned to Drosophila genome dm6 in multiple alignment format ( MAF ) was downloaded from UCSC Genome Browser . The strand specific gene models of the mentioned genes ( FlyBase r6 . 09 ) were provided as BED files . Those inputs were used in Galaxy ( Blankenberg et al . , 2010 ) in ‘Stitch MAF blocks’ followed by ‘concatenate FASTA alignment by species’ functions to generate FASTA alignments for each gene in the 12 Drosophila specified by the PhyloCSF phylogeny . PhyloCSF ( Lin et al . , 2011 ) was run with the resulting FASTA file using the following parameters: ‘--orf=ATGStop --frames=3 removeRefGaps --aa’ . Crosses were set up at 29°C . Third-instar larval brains were collected after 5–6 days , NBs were isolated by FACS ( Harzer et al . , 2013 ) and transplantations of GFP+ NBs ( UAS-dcr2; insc-GAL4 , UAS-stinger::GFP ) or RFP+ tNB ( brat RNAi driven by UAS-dcr2; wor-GAL4 , ase-GAL80; UAS-stinger::RFP ) suspensions were performed as previously described ( Caussinus and Gonzalez , 2005 ) with minor modifications . Pictures of transplanted host flies were taken with a Sony Alpha NEX-5 compact camera . Dissected brains were enzymatically dissociated in Rinaldini solution as described previously ( Berger et al . , 2012 ) and incubated with Hoechst 33342 ( 20 µM ) for 1 hr at room temperature . Then samples were kept on ice until FACS sorting . Data plots were generated with FlowJo software . Genomic DNA was isolated from the dissected tumor brains and abdomens of the same adult female brat k06028 fly using standard Phenol-Chloroform extraction procedure including RNase and Proteinase K treatment . In total , three flies were sequenced . DNA was fragmented using a microtip sonicator ( Omni-Ruptor 250 , Omni International ) . Quality control was performed with Agilent High Sensitivity DNA Kit ( Agilent Technologies ) . DNA libraries were prepared using NEBNext Ultra DNA Library Prep Kit ( Illumina ) and 100 base pair paired-end sequencing was performed on a Hiseq2000 platform . After deduplication , an average sequencing depth of >170 x was achieved for each sample . Sequence data has been deposited at the short read archive ( https://www . ncbi . nlm . nih . gov/sra , SUB1954694 ) . Leading and trailing Ns of the paired reads were trimmed . Reads were aligned with BWA ( v0 . 6 . 2 ) ( Li and Durbin , 2009 ) to the genome ( FlyBase r5 ) with a maximum insert size of 1000 . Picard tools ( v1 . 82 , http://broadinstitute . github . io/picard ) were used to fix the alignment ( CleanSam ) and add read groups ( AddOrReplaceReadGroups ) . Duplicates ( MarkDuplicates ) were marked within all samples derived from the same fly . Reads of a fly were realigned with GATK ( v2 . 3 ) for each chromosome and remerged with Picard tools . Summary statistics were computed with GATK ( McKenna et al . , 2010 ) . Somatic point mutations were identified with MuTect ( v1 . 1 . 4 ) ( Cibulskis et al . , 2013 ) . InDels ( strand . bias = TRUE ) were identified with the SomaticIndelDetector of GATK . Variants were characterized with SnpEff ( v3 . 2a ) ( Cingolani et al . , 2012 ) . For coverage plots , reads were counted in genomic bins and normalized by the median . The foldchange and coverage were plotted with R . Published aCGH datasets were used to identify under-replicated ( Sher et al . , 2012 ) and amplified ( Kim et al . , 2011 ) regions . We devised a method that combines transposon-mediated library preparation with molecular barcoding to quantify the original library molecules rather than their amplicons ( Figure 3B ) . The NBII driver line UAS-dcr2; wor-GAL4 , ase-GAL80; UAS-stinger::RFP was used . Brains from wandering third instar larvae were dissected , dissociated and NBs were isolated by FACS ( Berger et al . , 2012; Harzer et al . , 2013 ) . For each condition , three samples were prepared . RNA from 300 cells per sample was used for library preparation . Total RNA isolated with TRIzol LS reagent ( Ambion ) was reverse transcribed into first strand cDNA using Superscript III Reverse Transcriptase ( Invitrogen ) with oligo- ( dT ) 20 primers . After second strand synthesis was performed , the sequencing library was prepared with the Nextera DNA Library Preparation Kit ( Illumina ) . In an enzymatic tagmentation reaction , cDNA was simultaneously fragmented and tagged with adapter sequences: 15 µl TDE1 Tagment DNA buffer , 0 . 2 µl TDE1 Tagment DNA enzyme was added to 15 µl cDNA and incubated for 5 min at 55°C . After purification ( Agencourt AMPure XP beads , Beckman Coulter ) , 19 . 5 µl tagmented DNA was PCR amplified using 25 µl Phusion HF 2x master mix ( Thermo Fisher Scientific ) , 2 . 5 µl 20x Eva Green ( Biotium ) , 1 µl Nextera primers mix ( 10 µM each ) , 1 µl Index two primers ( N501-N506 , for multiplexing ) and 1 µl modified Index one primers , which included random 8-mer tags for molecular barcoding . Cycling conditions according to the manufacturer ( Nextera DNA Library Preparation Kit , Illumina ) were used . Purified libraries ( Agencourt AMPure XP beads ) were subjected to 50 base pair Illumina single-end sequencing on a Hiseq2000 platform . The reads were screened for ribosomal RNA by aligning with BWA ( v0 . 6 . 1 ) ( Li and Durbin , 2009 ) against known rRNA sequences ( RefSeq ) . The rRNA subtracted reads were aligned with TopHat ( v1 . 4 . 1 ) ( Trapnell et al . , 2009 ) against the Drosophila melanogaster genome ( FlyBase r5 . 44 ) and a maximum of 6 mismatches . Introns between 20–150000 bp were allowed which is based on FlyBase statistics . Maximum multihits was set to one and InDels as well as Microexon-search was enabled . Additionally , a gene model was provided as GTF ( FlyBase r5 . 44 ) . snRNA , rRNA , tRNA , snoRNA and pseudogenes were masked for downstream analysis . Reads arising from duplication events were marked as such in the alignment ( SAM/BAM files ) as follows: The different tags were counted at each genomic position . Thereafter , the diversity of tags at each position was examined . First , tags were sorted descending by their count . If several tags had the same occurrence , they were further sorted alphanumerically . Reads sharing the same tag , were sorted by the average PHRED quality . Again if several reads had the same quality , they were further sorted alphanumerically . Now the tags were cycled through by their counts . Within one tag , the read with the highest average PHRED quality was the unique correct read and all subsequent reads with the same tag were marked as duplicates . Furthermore , all reads which had tags with one mismatch difference compared the pool of valid read tags were also marked as duplicates . The aligned and deduplicated reads were counted with HTSeq ( Anders et al . , 2015; Li et al . , 2009 ) and the polyA containing transcripts were subjected to differential expression analysis with DESeq ( v1 . 10 . 1 ) ( Anders and Huber , 2010 ) . Note that the basemean values of brat mRNA appear unchanged in brat RNAi NBII compared to control NBII . We attribute this to a large number of reads , unique to the brat RNAi condition , matching non-coding regions in the second intron of brat-RA , -RE , and the first intron of brat-RB , -RC . Conversely , reads matching the coding region of the gene are reproducibly lower in brat RNAi condition . Furthermore , brat knockdown was confirmed by qPCR targeting a coding region . Data has been deposited in the data depository Gene Expression Omnibus ( http://www . ncbi . nlm . nih . gov/geo/ , GEO serial accession number GSE87085 ) . To investigate the expression of early temporal NB identity genes ( Figure 8A ) in tNBs we made use of a previously published gene dataset from antenna lobe NBs ( Liu et al . , 2015 ) . The Multiz alignment of 27 insects was accessed for the genomic locus via the UCSC table browser . The MAF blocks are stiched in Galaxy ( Blankenberg et al . , 2011 ) . Percent of sequence identity defined as ‘100 * ( identical positions ) / ( aligned positions + internal gap positions ) ’ was calculated with Biostrings in R ( R package version 2 . 40 . 2 . ) . Drosophila suzukii was excluded due to a very small homology region . Thermodynamically stable and evolutionary conserved RNA structures were predicted using the RNAz Web server ( Gruber et al . , 2007 ) using step size 10 and a window size of 200 . Sequence repeats were identified by RepeatMasker ( Jurka , 2000 ) and conservation was assessed by PhastCons and PhyloP ( Siepel et al . , 2005 ) via the UCSC Genome Browser ( dm6 ) . | Many biological signals control how cells grow and divide . However , cancer cells do not obey these growth-restricting signals , and as a result large tumors may develop . Recent experiments have suggested that stem cells – the precursors to the different types of specialized cells found in the body – are particularly important for generating tumors . A stem cell normally divides unequally to form a self-renewing cell and a more specialized cell ( often a progenitor cell that will give rise to increasingly specialized cell types ) . The timing of when the specialization occurs can be key to guiding the ultimately produced cell progenies to their final identity . However , in a tumor cells can retain the ability to self-renew . Ultimately , the resulting ‘tumor stem cells’ become immortal and proliferate indefinitely . It is not fully understood why this uncontrolled proliferation occurs . Just like mammals ( including humans ) , fruit flies can develop tumors . Some of the DNA mutations responsible for tumor development were already identified in flies as early as in the 1970s . This has made fruit flies a well-studied model system for uncovering the principle defects that cause tumors to form . Landskron et al . have now studied the neural stem cells found in brain tumors in fruit flies . Additional DNA mutations were not responsible for these cells becoming immortal . Instead , certain RNA molecules – products that are ‘transcribed’ from the DNA – were present in different amounts in tumor cells . The RNA that showed the greatest increase in tumor cells is a so-called long non-coding RNA named cherub . This RNA molecule has no important role in normal fruit flies , but is critical for tumor formation . Landskron et al . found that during cell division cherub segregates from the neural stem cells to the newly formed progenitor cells , where it breaks down over time . Progenitor cells that contain high levels of cherub give rise to tumor-generating neural stem cells . At the molecular level , cherubhelps two proteins to interact with each other: one called Syncrip that makes the neural stem cells take on a older identity , and another one ( Staufen ) that tethers it to the cell membrane . By restricting Syncrip to a particular location in the cell , cherub alters the timing of stem cell specialization , which contributes to tumor formation . Overall , the results presented by Landskron et al . reveal a new role for long non-coding RNAs: controlling the localization of the proteins that determine the fate of the cell . They also highlight a critical link between the timing of stem cell development and the proliferation of the cells . Further work is now needed to test whether the same control mechanism works in species other than fruit flies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"cancer",
"biology"
] | 2018 | The asymmetrically segregating lncRNA cherub is required for transforming stem cells into malignant cells |
Peptidoglycans ( PGNs ) are immunogenic bacterial surface patterns that trigger immune activation in metazoans and plants . It is generally unknown how complex bacterial structures such as PGNs are perceived by plant pattern recognition receptors ( PRRs ) and whether host hydrolytic activities facilitate decomposition of bacterial matrices and generation of soluble PRR ligands . Here we show that Arabidopsis thaliana , upon bacterial infection or exposure to microbial patterns , produces a metazoan lysozyme-like hydrolase ( lysozyme 1 , LYS1 ) . LYS1 activity releases soluble PGN fragments from insoluble bacterial cell walls and cleavage products are able to trigger responses typically associated with plant immunity . Importantly , LYS1 mutant genotypes exhibit super-susceptibility to bacterial infections similar to that observed on PGN receptor mutants . We propose that plants employ hydrolytic activities for the decomposition of complex bacterial structures , and that soluble pattern generation might aid PRR-mediated immune activation in cell layers adjacent to infection sites .
Activation of antibacterial defenses in multicellular eukaryotic organisms requires recognition of bacterial surface patterns through host-encoded pattern recognition receptors ( PRRs ) ( Chisholm et al . , 2006; Jones and Dangl , 2006; Ishii et al . , 2008; Boller and Felix , 2009; Vance et al . , 2009; Segonzac and Zipfel , 2011; Monaghan and Zipfel , 2012; Broz and Monack , 2013; Stuart et al . , 2013 ) . Immunogenic microbial signatures are collectively referred to as pathogen- or microbe-associated molecular patterns ( PAMPs/MAMPs ) ( Janeway and Medzhitov , 2002 ) . Bacteria-derived PAMPs such as lipopolysaccharides ( LPS ) or flagellins possess immunity-stimulating activities in metazoans and plants , suggesting that the ability to sense bacterial surface structures and mount immunity is conserved across lineage borders ( Nürnberger et al . , 2004; Boller and Felix , 2009 ) . Likewise , peptidoglycans ( PGNs ) are major building blocks of the cell walls of Gram-positive and Gram-negative bacteria that have been shown to trigger host immune responses in mammalians , insects , and plants ( Dziarski and Gupta , 2005; Gust et al . , 2007; Erbs et al . , 2008; Kurata , 2014 ) . Structurally , PGNs are heteroglycan chains that are composed of polymeric alternating β ( 1 , 4 ) -linked N-acetylglucosamine ( GlcNAc ) and N-acetylmuramic acid ( MurNAc ) residues ( Schleifer and Kandler , 1972; Glauner et al . , 1988 ) . Such chains are interconnected by oligopeptide bridges which form a coordinate meshwork , thereby providing structural integrity to the bacterial envelope . Recognition of different PGN substructures in animal hosts is brought about by structurally diverse PRRs such as nucleotide-binding oligomerization domain-containing proteins ( NODs ) , peptidoglycan recognition proteins ( PGRPs/PGLYRPs ) , scavenger receptors , or Toll-like receptor TLR2 ( Strober et al . , 2006; Royet and Dziarski , 2007; Dziarski and Gupta , 2010; Müller-Anstett et al . , 2010; Magalhaes et al . , 2011; Kurata , 2014 ) . In plants , a tripartite PGN recognition system at the plasma membrane of Arabidopsis thaliana with shared functions in PGN sensing and transmembrane signaling was recently described ( Willmann et al . , 2011 ) . This system comprises Lysin motif ( LysM ) domain proteins LYM1 and LYM3 for PGN ligand binding and the transmembrane LysM receptor kinase CERK1 that is likely required for conveying the extracellular signal across the plasma membrane and for initiating intracellular signal transduction . All three proteins were shown to be indispensable for PGN sensitivity and to contribute to immunity to bacterial infection ( Willmann et al . , 2011 ) , which is in agreement with their proposed role as a PGN sensor system . More recently , a similar PGN perception system made of LysM domain proteins LYP4 and LYP6 has been reported from rice ( Liu et al . , 2012a ) . Microbial patterns such as bacterial PGN , LPS , flagellin , or fungal chitin harbor immunogenic epitopes that are parts of supramolecular structures building microbial surfaces ( Boller and Felix , 2009; Kumar et al . , 2013; Newman et al . , 2013; Pel and Pieterse , 2013 ) . It is therefore assumed that recognition by host PRRs most likely requires the presence of soluble , randomly structured ligands derived from a complex matrix . X-ray structure-based insight into the binding of bacterial flagellin to the Arabidopsis receptor complex FLS2/BAK1 or of fungal chitin to the Arabidopsis receptor CERK1 supports this view ( Willmann and Nürnberger , 2012; Liu et al . , 2012b; Sun et al . , 2013 ) . Moreover , the existence of fungal LysM effector proteins that scavenge soluble chitin fragments , thus preventing recognition by plant PRRs , suggests that mechanisms releasing these soluble fragments from fungal cell walls must exist ( de Jonge et al . , 2010 ) . Most often , however , it is an open question whether soluble ligand presentation to eukaryotic host PRRs is the result of spontaneous decomposition of the microbial extracellular matrix during infection or , alternatively , whether host-derived factors contribute to the generation of immunogenic ligands for PRR activation . For example , only monomers of bacterial flagellin induce immune responses through human TLR5 whereas filamentous flagella , in which the immunogenic flagellin structure is buried and thus is not accessible to TLR5 , do not ( Smith et al . , 2003 ) . It was proposed that a number of circumstances cause flagellin monomer release from intact flagella . For instance , Caulobacter crescentus deliberately ejects its flagellum once it is no longer required for the bacterial life cycle ( Jenal and Stephens , 2002 ) . Moreover , during infection , Pseudomonas aeruginosa produces rhamnolipids which act as surfactants and cause flagellin shedding from intact flagella , resulting in a more pronounced immune response ( Gerstel et al . , 2009 ) . Alternatively , host factors such as proteases or environmental conditions such as pH , temperature , or bile salts have been proposed to mediate shearing of flagella from bacterial surfaces ( Ramos et al . , 2004 ) . Likewise , recognition of PGN by intracellular receptors , such as mammalian NOD1 and NOD2 , or by plasma membrane receptors , such as mammalian TLR2 or plant LYM1 , LYM3 and CERK1 ( Müller-Anstett et al . , 2010; Sorbara and Philpott , 2011; Willmann et al . , 2011 ) , is facilitated by soluble ligands . Animal lysozymes have been implicated in PGN hydrolysis , bacterial lysis , and host immunity ( Callewaert and Michiels , 2010 ) , probably through partial PGN degradation and generation of soluble ligands for PGN sensors ( Cho et al . , 2005; Dziarski and Gupta , 2010; Davis et al . , 2011 ) . In plants , knowledge of the mode of release of immunogenic fragments from microbial extracellular structures and their contribution to plant immunity is lacking . We here describe a plant enzyme activity ( LYS1 ) that hydrolyzes β ( 1 , 4 ) linkages between N-acetylmuramic acid and N-acetylglucosamine residues in PGN and between N-acetylglucosamine residues in chitooligosaccharides , thus closely resembling metazoan lysozymes ( EC 3 . 2 . 1 . 17 ) . Importantly , PGN breakdown products produced by LYS1 are immunogenic in plants , and LYS1 mutant genotypes were immunocompromised upon bacterial infection . Our findings suggest that plant enzymatic activities , such as LYS1 , are capable of generating soluble PRR ligands that might contribute to the activation of immune responses in cells at and surrounding the site of their generation . We also infer that eukaryotic hosts more generally make concerted use of PGN hydrolytic activities and of PRRs in order to cope with bacterial infections .
Soluble oligomeric PGN fragments have previously been shown to stimulate plant immune responses in Arabidopsis ( Gust et al . , 2007; Erbs et al . , 2008; Willmann et al . , 2011 ) . As some metazoan PGRPs harbor PGN-degrading enzyme activities ( Gelius et al . , 2003; Wang et al . , 2003; Bischoff et al . , 2006; Dziarski and Gupta , 2010; Kurata , 2010 ) , we tested whether recombinant Arabidopsis PGN binding proteins LYM1 and LYM3 were able to catalyze PGN degradation . For this , we have employed a standard lysozyme assay ( Park et al . , 2002 ) that is based on reduced turbidity in suspensions of Gram-positive Micrococcus luteus cell wall preparations due to PGN degradation . PGN-degrading activity of hen egg-white lysozyme served as a positive control in these assays . As shown in Figure 1A , lysozyme , but not recombinant LYM1 or LYM3 , displayed cell wall-degrading lytic activity , suggesting that the latter are unable to release PGN fragments from bacterial cell walls . This is in agreement with a lack of sequence similarities between LYM1 or LYM3 and known metazoan PGN hydrolytic activities . We therefore conclude that LYM1 and LYM3 constitute plant PGN sensors that appear to be functionally related to non-enzymatic mammalian or Drosophila PGRPs ( Cho et al . , 2005; Bischoff et al . , 2006; Dziarski and Gupta , 2010; Kurata , 2010 ) . 10 . 7554/eLife . 01990 . 003Figure 1 . The Arabidopsis lysozyme 1 ( LYS1 ) gene is transcriptionally activated upon pathogen-infection . ( A ) LYM1 and LYM3 do not possess peptidoglycan ( PGN ) hydrolytic activity . Micrococcus luteus cell wall preparations were incubated with 20 μg affinity-purified His6-tagged LYM1 or LYM3 or 0 . 5 μg hen egg-white lysozyme and PGN hydrolytic activity was assayed in a turbidity assay at the indicated time points . As negative control ( nc ) , non-induced His6-tagged LYM3 bacterial lysates were used for affinity purification and eluates were subjected to turbidity assays . Means ± SD of three replicates per sample are given . Statistical significance compared with the negative control ( **p<0 . 001 , ***p<0 . 0001 , Student’s t test ) is indicated by asterisks . ( B ) Multiple sequence alignment of the 24 Arabidopsis chitinases using the ClustalW2 algorithm . Full length amino acid sequences were aligned and subgroups were classified according to Passarinho and de Vries ( 2002 ) . Arabidopsis lysozyme 1 ( LYS1 , At5g24090 ) represents the only member of class III . ( C ) The expression of LYS1 in transgenic pLYS1::GUS reporter plants . Leaf halves of transgenic pLYS1::GUS or pPR1::GUS reporter plants were infiltrated with the virulent Pseudomonas syringae pv . tomato ( Pto ) DC3000 , the type III secretion system-deficient Pto DC3000 hrcC- or the avirulent Pseudomonas syringae pv . phaseolicola ( Pph ) strain ( 108 cfu/ml ) or 10 mM MgCl2 as control . After 24 hr the leaves were harvested and stained for β-glucuronidase ( GUS ) activity . ( D ) Leaves of wild-type plants were treated for 3 or 24 hr with 1 µM flg22 , 100 µg/ml PGN from Pto or 100 µg/ml lipopolysaccharide ( LPS ) . Total RNA was subjected to RT-PCR using LYS1 or Flagellin-responsive kinase 1 ( FRK1 ) specific primers . EF1α transcript was used for normalization . All experiments shown in panels ( A ) , ( C ) and ( D ) were repeated once with similar results . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 003 Lysozymes ( EC 3 . 2 . 1 . 17 ) hydrolyze β ( 1 , 4 ) linkages between N-acetylmuramic acid and N-acetylglucosamine residues in PGNs and between N-acetylglucosamine residues in chitodextrins ( http://enzyme . expasy . org/EC/3 . 2 . 1 . 17 ) . Plant genomes do not encode lysozyme-like proteins , but many plant species produce lysozyme-like enzyme activities such as chitinases ( EC 3 . 2 . 1 . 14 ) ( Audy et al . , 1988; Sakthivel et al . , 2010 ) . Plant chitinases fall into five classes ( I–V , Figure 1B ) ( Passarinho and de Vries , 2002 ) and are grouped into structurally unrelated families 18 and 19 of glycosyl hydrolases , respectively ( Henrissat , 1991 ) . Chitinases belonging to family 18 of glycosyl hydrolases are ubiquitously found in all organisms whereas chitinases of glycosyl hydrolase family 19 are found almost exclusively in plants . Class III chitinases ( glycosyl hydrolase family 18 ) represent bifunctional plant enzymes with lysozyme-like activities . One such enzyme , hevamine from the rubber tree Hevea brasiliensis ( Beintema et al . , 1991 ) , has been shown to hydrolyze PGN and the structurally closely related β ( 1 , 4 ) -linked GlcNAc homopolymer chitin in vitro ( Bokma et al . , 1997 ) . To explore host-mediated PGN degradation and its possible implication in plant immune activation , we have addressed the only class III chitinase ( which we named LYS1 , At5g24090 ) encoded by the Arabidopsis genome ( Passarinho and de Vries , 2002; Figure 1B ) . Bacterial infection of Arabidopsis plants stably expressing a pLYS1::GUS construct revealed that LYS1 gene expression is enhanced upon infection with host non-adapted Pseudomonas syringae pv . phaseolicola ( Pph ) or disarmed host adapted P . syringae pv . tomato ( Pto ) DC3000 hrcC− . Likewise , expression of the immune response marker pathogenesis-related protein 1 ( PR1 ) was enhanced by the same treatment ( Figure 1C ) . Failure to detect LYS1 expression in plants infected with virulent host adapted Pto DC3000 suggests bacterial effector-mediated suppression that is reminiscent of that observed for PGN receptor proteins LYM1 and LYM3 ( Willmann et al . , 2011 ) as well as numerous other immunity-associated genes ( Kemmerling et al . , 2007; Postel et al . , 2010 ) . LYS1 gene expression is not only triggered upon bacterial infection , but was also observed upon treatment with different MAMPs including bacterial flagellin , LPS , or PGN preparations ( Figure 1D ) , similar to the immune marker gene Flagellin-responsive kinase 1 ( FRK1 ) . Altogether , infection-induced LYS1 transcriptional activation suggests that the LYS1 protein is implicated in immunity to bacterial infection . To analyze the enzymatic properties of LYS1 , recombinant protein production was attempted . Overexpression in Escherichia coli failed to produce active enzyme and LYS1 production in eukaryotic Pichia pastoris entirely failed to produce recombinant protein ( not shown ) . Therefore , we resorted to generate p35S::LYS1-GFP-overexpressing ( LYS1OE ) plants ( Figure 2A , B ) . Notably , LYS1-GFP was glycosylated ( Figure 2C ) , possibly explaining the failure to produce enzymatically active LYS1 protein in E . coli . Expression of the green fluorescent protein ( GFP ) fusion protein in Arabidopsis plants was accompanied by substantial proteolytic cleavage resulting in the predominant release of a protein with an approximate molecular mass of 35 kDa , most likely representing untagged LYS1 ( Figure 2B ) . Analysis of this major cleavage product by liquid chromatography-mass spectrometry ( LC-MS/MS ) after tryptic in-gel digestion and by peptide mass fingerprint not only confirmed the identity of LYS1 in this band but also yielded peptides spanning the whole protein sequence , except for the first 53 amino acids ( data not shown ) , thus indicating cleavage of the LYS1-GFP fusion protein between LYS1 and GFP . 10 . 7554/eLife . 01990 . 004Figure 2 . Analysis of LYS1 overexpression lines . ( A ) RT-qPCR analyses of transcript levels in mature leaves of two independent transgenic lines expressing p35S::LYS1-GFP ( LYS1OE-1 , LYS1OE-2 ) relative to expression levels in wild-type . EF1α transcript was used for normalization . Error bars , SD ( n = 3 ) . Statistical significance compared with wild-type ( ***p<0 . 001 , Student’s t test ) is indicated by asterisks . ( B ) Immunoblot analysis of protein extracts from leaves of two independent LYS1OE lines , a LYS1 knock-down line ( LYS1KD-1 , see Figure 3 ) and wild-type plants . Total leaf protein was separated by SDS-PAGE and blotted onto a nitrocellulose membrane . Immunodetection was carried out using α-tobacco class III chitinase ( α-Chit ) or green fluorescent protein ( α-GFP ) ( both from rabbit ) and an anti-rabbit HRP-coupled secondary antibody . Ponceau S red staining of the large subunit of RuBisCO served as loading control . ( C ) Total protein extracts from leaves of LYS1OE-1 plants were subjected to deglycosylation with a deglycosylation kit ( NEB ) . The negative control ( − ) was treated as the deglycosylation sample ( + ) but without addition of the deglycosylation enzyme mix . Immunoblot analysis was carried out as described in ( B ) . All experiments shown were repeated at least once . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 004 Three mutant lines with T-DNA insertions in the LYS1 gene were available from the Nottingham Arabidopsis Stock Centre . However , neither the insertion in the 5' untranslated region nor the insertions in the first intron and at the end of the last exon of the coding region abolished formation of the LYS1 transcript ( Figure 3—figure supplement 1 ) . As an alternative to knock-out lines , LYS1 knock-down lines ( LYS1KD ) were produced by artificial micro RNA technology ( Schwab et al . , 2006; Figure 3 ) . As proven by quantitative reverse transcriptase polymerase chain reaction ( RT-qPCR ) , we obtained two genetically independent LYS1KD lines with residual transcript levels not exceeding 10% of those detected in wild-type plants ( Figure 3C ) . In contrast , the transcription of potential off-target genes was not affected ( Figure 3C ) . Protein extracts derived from transgenic plants were tested for chitinolytic activity by employing 4-methylumbelliferyl β-D-N , N′ , N″-triacetylchitotriose ( 4-MUCT ) as substrate . Leaf protein extracts from LYS1OE plants exhibited significant chitinase activity when compared with a Streptomyces griseus chitinase control ( Figure 4A ) . In contrast , wild-type and LYS1KD plants exhibited only marginal chitinase activities . Likewise , using 4-MUCT in a gel electrophoretic separation-based chitinase assay produced a zymogram in which enzyme activity was solely detectable in protein extracts obtained from LYS1OE plants , but not in those from control plants expressing secreted GFP ( secGFP ) ( Figure 4B ) . Thus , LYS1 indeed harbors the predicted chitinase activity . As 4-MUCT is also a typical substrate for lysozymes ( Brunner et al . , 1998 ) , this was the first indication that LYS1 might also harbor lysozyme activity . Next , leaf protein extracts from LYS1OE plants were tested for their ability to solubilize complex PGN presented by intact Gram-positive M . luteus cells and to cleave preparations of complex , insoluble Bacillus subtilis PGN . Again , protein extracts from LYS1OE plants exhibited significant PGN-degrading activity whereas wild-type and LYS1KD plants showed basal activity levels only ( Figure 4C , D ) . Likewise , PGN-solubilizing activity profiles of protoplast suspensions derived from these transgenics confirmed significant PGN-degrading activity of LYS1OE plants ( Figure 4E ) . 10 . 7554/eLife . 01990 . 005Figure 3 . Analysis of LYS1 amiRNA lines . ( A ) Predicted LYS1 gene structure ( exons , black bars; introns , black lines; untranslated regions , gray ) . The region targeted by the amiRNA construct is indicated by an arrowhead . ( B ) Off-target genes for the LYS1-amiRNA construct were identified using the Web microRNA Designer ( WMD; http://wmd . weigelworld . org ) . The region targeted by the amiRNA is given for each gene , mismatches are indicated in red . Potential off targets either possess more than one mismatch at positions 2–12 or have mismatches at position 10 and/or 11 which will limit amiRNA function . ( C ) Transcript levels of the four top hits shown in ( B ) were determined by RT-qPCR in untreated seedlings of two independent transgenic LYS1-amiRNA knock-down lines ( LYS1KD-1 , LYS1KD-2 ) using gene-specific primers for LYS1 ( At5g24090 ) , At4g02540 , At1g05615 , At5g58780 , and At3g51010 . EF1α transcript was used for normalization . Error bars , SD ( n = 3 ) . Statistical significance compared with the wild-type control ( which was set to 1 for each primer set ) is indicated by asterisks ( ***p<0 . 001 , Student’s t test ) . The experiment was repeated once with similar results . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 00510 . 7554/eLife . 01990 . 006Figure 3—figure supplement 1 . Characterization of LYS1 T-DNA insertion lines . ( A ) Predicted LYS1 gene structure ( exons , black bars; introns , black lines; untranslated regions , gray ) . T-DNA insertion sites are indicated by triangles . ( B ) The T-DNA insertion lines ( each two samples ) and the corresponding wild-type accessions were genotyped using the following primer combinations: LP_N853931 and RP_N853931 ( WT-PCR , lys1-1 ) , Wisc-Lba and RP_853931 ( Lba-PCR , lys1-1 ) , LP_N595362 and RP_N595362 ( WT-PCR , lys1-2 ) , Salk-Lba and RP_N595362 ( Lba-PCR , lys1-2 ) , At5g24090F1 and At5g24090R1 ( WT-PCR , lys1-3 ) , and Ds5-1 and At5g24090R1 ( Lba-PCR , lys1-3 ) . ( C ) The LYS1 transcript analysis in mature leaves was done by semi-quantitative RT-PCR using the following primer combinations: At5g24090F and At5g24090R ( lys1-1 and lys1-2 ) and At5g24090F and At5g24090RP2 ( lys1-3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 00610 . 7554/eLife . 01990 . 007Figure 4 . LYS1 is a glucan hydrolase . ( A-D ) Protein extracts from adult wild-type or LYSOE-1 and LYSKD-1 homozygous lines were assayed for hydrolytic activity towards glycan substrates . Plants expressing secreted green fluorescent protein ( GFP ) ( secGFP ) served to control the effect of external GFP . ( A ) Leaf protein extracts from indicated transgenic plants were assayed for chitinolytic activity using the 4-methylumbelliferyl β-D-N , N′ , N″-triacetylchitotriose ( 4-MUCT ) substrate . Enzymatic activities 4 hr after treatment were calculated using Streptomyces griseus chitinase as positive control ( pc ) . ( B ) Protein extracts from LYS1OE-1 or secGFP plants were separated on a cetyltrimethylammonium bromide-polyacrylamide gel and hydrolytic activity was assayed by overlaying the gel with the substrate 4-MUCT . Fluorescent bands are indicative of substrate cleavage . The arrowhead indicates the position of LYS1 . ( C and D ) Micrococcus luteus cells ( C ) or Bacillus subtilis peptidoglycan ( PGN ) ( D ) were subjected to hydrolysis by leaf protein extracts and PGN hydrolytic activity was calculated after 4 hr using hen egg-white lysozyme as positive control ( pc ) . Significant differences compared with the buffer control are indicated by asterisks ( *p<0 . 05; Student’s t test; A , C , D ) . ( E ) Protoplasts of transgenic lines were pelleted and protein extracts of the protoplast ( PP ) pellet or medium supernatant were subjected to the PGN hydrolysis assay as described in ( C ) . As controls , buffer or protoplast medium ( PP medium ) was used . Means ± SD of two replicates per sample are given , bars with different letters are significantly different based on one-way ANOVA ( p<0 . 05 ) . ( F ) Lysis of M . luteus cells was determined in a turbidity assay with LYS1OE leaf protein extracts as described in ( C ) at the indicated pH . Means ± SD of two replicates per sample are given . All experiments shown were repeated at least once . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 00710 . 7554/eLife . 01990 . 008Figure 4—figure supplement 1 . LYS1 is located in the plant apoplast . ( A ) Apoplastic washes were prepared from leaves of wild-type Arabidopsis plants or the LYS1OE-1 and LYS1KD-1 lines . Apoplastic fluids ( concentrated tenfold ) or total leaf protein extracts were subjected to western blot analysis using antibodies raised against green fluorescent protein ( α-GFP ) , tobacco class III chitinase ( α-chit ) , or the cytoplasmic mitogen-activated protein kinase 3 ( MPK3 ) . ( B ) The p35S::LYS1-GFP and p35S::LYS1ΔSP-GFP constructs were transiently expressed in Nicotiana benthamiana leaves using Agrobacterium tumefaciens-mediated transformation . GFP fluorescence in the leaf epidermal cells was analyzed 3 days post infection . FM4-64 was used to stain the plasma membrane . Argon/krypton laser was used for excitation of GFP at 488 nm and the 543 nm line of helium/neon laser for the excitation of FM4-64 . Detection wavelengths of emitted light were 500–600 nm ( GFP ) and 560–615 nm ( FM4-64 ) . All experiments shown were repeated three times . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 00810 . 7554/eLife . 01990 . 009Figure 4—figure supplement 2 . LYS1 is devoid of cellulose hydrolytic activity . LYS1 was purified from 5-week-old LYS1OE plants and used for cellulase activity assays . The substrate 4-methylumbelliferyl-β-D-cellobioside was incubated for 1 hr with purified LYS1 , commercial reference cellulose , or buffer as control . Fluorescence was determined ( ex/em = 365 nm/455 nm ) after stopping the reaction with 0 . 2 M sodium carbonate . Means ± SD of three replicates per sample are given . Statistical significance compared with the buffer control ( ***p<0 . 001 , Student’s t test ) is indicated by asterisks . The experiment was repeated once with the same result . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 009 To determine specific enzyme activities , untagged LYS1 was purified from LYS1OE Arabidopsis lines by fast protein liquid chromatography ( FPLC ) and used for enzyme assays . The 4-MUCT assay yielded a Michaelis constant ( Km ) of 70 ± 14 µM and a Vmax of 378 ± 42 µM min−1 mg−1 for LYS1 , and a Km of 53 ± 27 µM and a Vmax of 397 ± 145 µM min−1 mg−1 for commercial S . griseus chitinase . Using the turbidity assay with M . luteus cell wall preparations , a Km of 18 . 2 ± 2 . 5 mg/ml and Vmax of 4 . 4 ± 0 . 6 mg mg−1 min−1 were obtained for LYS1 , and a Km of 8 . 4 ± 0 . 8 mg/ml and Vmax of 192 ± 120 mg mg−1 min−1 for commercial hen egg-white lysozyme . The Km values for LYS1 are thus comparable to the commercial enzymes . As shown in Figure 4E , the majority of LYS1 activity was found in the supernatant of the protoplasts , suggesting an apoplastic localization of LYS1 . To confirm this localization we prepared apoplastic washes from LYS1OE Arabidopsis lines . Both the LYS1-GFP fusion protein as well as free LYS1 was detectable in concentrated apoplastic fluids whereas the cytoplasmic mitogen-activated protein kinase MPK3 was only present in the total leaf protein samples ( Figure 4—figure supplement 1A ) . Moreover , transient expression in the heterologous plant system Nicotiana benthamiana of the p35S::LYS1-GFP construct resulted in labeling of the cell periphery , whereas expression of a construct lacking the LYS1 signal peptide-encoding sequence yielded labeling of intracellular structures ( Figure 4—figure supplement 1B ) . Use of the fluorescent dye FM4-64 , a plasma membrane and early endosome marker ( Bolte et al . , 2004 ) , revealed that LYS1 signals co-localized to a large extent with the plasma membrane ( Figure 4—figure supplement 1B ) . Thus , LYS1 likely operates in close vicinity of the plant surface . Indeed , previous identification within the Arabidopsis cell wall proteome ( Kwon et al . , 2005 ) suggests that LYS1 acts in the plant apoplast . Since the plant apoplast is an acidic compartment ( pH 5–6 ) ( Schulte et al . , 2006 ) , we investigated whether LYS1 is active at physiologically relevant pH conditions . For this , the M . luteus cell wall-degrading activity of an LYS1OE leaf extract was determined at different pH values . Although active at pH values ranging from 3 . 2 to 7 . 2 , a pronounced maximum of LYS1 activity was detected around pH 6 which coincided with the apoplastic pH of plant cells ( Figure 4F ) . To further confirm LYS1 glucan hydrolytic activity , an epitope-tagged LYS1 fusion construct was transiently expressed in N . benthamiana ( Figure 5A ) . Similar to the Arabidopsis LYS1OE leaf extracts , extracts from p35S::LYS1-myc expressing N . benthamiana leaves also displayed in-gel chitinolytic activity ( Figure 5B ) compared with extracts from control leaves expressing the viral silencing suppressor p19 only . Likewise , N . benthamiana protein extracts containing LYS1-myc were able to cleave preparations of complex insoluble B . subtilis PGN ( Figure 5C ) . 10 . 7554/eLife . 01990 . 010Figure 5 . LYS1 transiently expressed in Nicotiana benthamiana possesses hydrolytic activity . ( A ) Protein extracts from N . benthamiana leaves expressing LYS1 fused to the myc-epitope tag under control of the p35S promoter were separated on an SDS-polyacrylamide gel and analyzed by western blot using antibodies raised against the myc-epitope tag . As control , plants were infiltrated with agrobacteria harboring the p19 suppressor of silencing construct ( p19 ) . Protein sizes ( kDa ) are indicated on the left . ( B ) N . benthamiana protein extracts from leaves expressing LYS1myc or p19 were separated on a cetyltrimethylammonium bromide-polyacrylamide gel and hydrolytic activity was assayed by overlaying the gel with the substrate 4-methylumbelliferyl β-D-N , N′ , N″-triacetylchitotriose . Fluorescent bands are indicative of substrate cleavage . Arrowheads indicate the positions of epitope-tagged LYS1 . ( C ) Protein extracts from N . benthamiana leaves expressing LYS1myc or p19 were assayed for peptidoglycan ( PGN ) hydrolytic activity in a turbidity assay using Bacillus subtilis PGN . Relative activities ( 2 hr post treatment ) were calculated using hen egg-white lysozyme as standard . Statistical significance compared with the untreated control ( *p<0 . 05 , Student’s t test ) is indicated by asterisks . All experiments shown were repeated at least once . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 010 In sum , we provide biochemical evidence that LYS1 harbors hydrolytic activity for chitin as well as for PGN of the lysine-type ( M . luteus ) and diaminopimelic acid-type ( B . subtilis ) . Importantly , LYS1 failed to exhibit activity on cellobiose as a substrate , indicating it might have no cellulose activity ( Figure 4—figure supplement 2 ) . Thus , LYS1 resembles enzymatic activities reported for metazoan lysozymes and should be classified as lysozyme ( EC 3 . 2 . 1 . 17 ) instead of chitinase ( EC 3 . 2 . 1 . 14 ) . To analyze immunogenic activities of PGN cleavage products generated by LYS1 , untagged LYS1 was purified from LYS1OE Arabidopsis lines by FPLC and used for degradation of B . subtilis PGN . Solubilized PGN fragments found in the supernatant of LYS1-digested PGN were subsequently analyzed by high performance liquid chromatography ( HPLC ) ( Figure 6A ) . Only a few peaks could be detected in the supernatant of PGN incubated with a buffer control or with heat-inactivated LYS1 . In contrast , PGN digests produced by native LYS1 yielded several characteristic peaks that were also detectable in the supernatants of PGN preparations treated with mutanolysin , which has been shown to cleave O-glycosidic bonds between GlcNAc and MurNAc residues in complex PGN ( Yokogawa et al . , 1975 ) . LYS1-generated PGN fragments were subsequently tested for their ability to trigger plant immunity-associated responses ( Figure 6B–D ) . First , supernatants of PGN preparations treated with either native or heat-denatured LYS1 were used to trigger immune marker gene FRK1 expression in Arabidopsis seedlings . Importantly , only supernatants from PGN digests produced by native LYS1 or mutanolysin induced FRK1 expression whereas buffer controls or digests produced by heat-inactivated LYS1 did not release immunogenic soluble fragments from complex PGNs ( Figure 6B ) . Notably , activation of immune responses by LYS1-generated PGN fragments was dependent on Arabidopsis PGN receptor complex components LYM1 , LYM3 , and CERK1 as the respective mutant genotypes failed to respond to immunogenic PGN fragments ( Figure 6B ) . Second , we tested whether LYS1-generated PGN fragments were able to trigger an immunity-associated response , medium alkalinization , in rice cell suspensions . This plant was chosen for testing as a PGN receptor system very similar to that in Arabidopsis has recently been reported ( Liu et al . , 2012a ) . As shown in Figure 6C , LYS1-released PGN fragments triggered medium alkalinization in cultured rice cells , suggesting that immune defense stimulation by soluble PGN fragments is not restricted to Arabidopsis only . 10 . 7554/eLife . 01990 . 011Figure 6 . Purified LYS1 generates immunogenic peptidoglycan ( PGN ) fragments . LYS1 was purified from 5-week-old LYS1OE plants and used for PGN digestion . ( A ) 500 µg Bacillus subtilis PGN were digested for 7 hr with mutanolysin ( 50 µg/ml ) , native purified LYS1 ( 140 µg/ml ) , heat-denatured purified LYS1 ( 140 µg/ml ) , or the reaction buffer alone and subjected to high performance liquid chromatography fractionation . Shown are the peak profiles of representative runs . The signal intensity is given in milliabsorbance units ( mAU ) . ( B ) B . subtilis PGN was digested for 4 hr as described in ( A ) and Arabidopsis wild-type seedlings or the indicated mutant lines were treated for 6 hr with 25 µl/ml digest supernatant containing solubilized PGN fragments . Total seedling RNA was subjected to RT-qPCR using Flagellin responsive kinase ( FRK1 ) specific primers . EF1α transcript was used for normalization , water treatment served as control and was set to 1 . ( C ) Supernatants of digested PGN ( 25 µl/ml ) were added to cultured rice cells and medium alkalinization was determined 20 min post addition . Treatment with water or MES buffer served as control . All data represent triplicate samples ± SD , bars with different letters are significantly different based on one-way ANOVA ( p<0 . 05; B and C ) . ( D ) B . subtilis PGN was digested with native purified LYS1 for the indicated times or overnight ( o/n ) and digest supernatant was used to trigger medium alkalinization in rice cells as described in ( C ) . All data represent triplicate samples ± SD , asterisks indicate significant differences compared to the buffer control ( *p<0 . 05; **p<0 . 01; ***p<0 . 001; Student’s t test ) . All experiments shown were repeated at least once . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 011 We further investigated the kinetics of PGN fragment release from complex PGNs . As shown in Figure 6D , release of immunogenic PGN fragments into solution occurred rapidly within 10 min of incubation with native LYS1 . Incubation of complex PGNs with LYS1 yielded the highest immunogenic activity of the digest supernatant after 30 min , suggesting that at that time point the maximum amount of immunogenic PGN fragments was generated . However , prolonged incubation with LYS1 again resulted in a loss of activity with overnight digestion completely abolishing stimulatory activity of the PGN digest . We assume that LYS1 is capable of releasing immunogenic fragments from complex PGNs , but extensive or complete digest into PGN monomers or small PGN fragments appears to abolish the immunogenic activity of PGN fragments . This result is in accordance with our previous observations that prolonged digestion of PGN with mutanolysin diminishes its defense-inducing activity ( Gust et al . , 2007 ) . To examine the physiological role of LYS1 in plant immunity , LYS1OE and LYS1KD lines were subjected to infection with various phytopathogens . As LYS1 harbors chitinase activity ( Figures 4A , B and 5B ) and as LYS1 transcripts accumulate upon fungal infection ( Samac and Shah , 1991 ) , we first analyzed the role of LYS1 in immunity towards fungal infection . Leaves of transgenic LYS1OE or LYS1KD lines and wild-type plants were infected with the necrotrophic fungus Botrytis cinerea and disease symptoms were monitored 2–3 days post infection . Fungal hyphal growth and necrotic leaf lesions at infection sites were detectable in all plant lines tested and hyphal outgrowth or cell death lesion sizes revealed no differences between wild-type , LYS1OE or LYS1KD lines ( Figure 7 ) . Likewise , infection with the necrotrophic fungus Alternaria brassicicola resulted in indistinguishable necrotic lesions in LYS1OE and LYS1KD transgenics compared to those observed in wild-type control plants ( Figure 8 ) . Trypan blue staining and microscopic analysis of the infection sites did not reveal major differences in fungal hyphal growth among all lines tested ( Figure 8B , C ) . Although disease indices at day 11 after infection were slightly increased in LYS1KD lines ( Figure 8D ) , such subtle differences were not statistically significant . In conclusion , we failed to detect a role for LYS1 in immunity to fungal infection with B . cinerea and A . brassicicola under our experimental conditions . However , these results cannot be generalized and LYS1 might still have a role under infection regimes other than the ones used here or it might be important for defense against other fungal pathogens . 10 . 7554/eLife . 01990 . 012Figure 7 . LYS1 lines are not impaired in resistance towards infection with Botrytis cinerea . Five-week-old plants were infected with the necrotrophic fungus Botrytis cinerea . 5 μl spore suspension of 5 × 105 spores/ml was drop-inoculated on one half of the leaf; two leaves per plant were infected . The plants were analyzed for development of symptoms 2 and 3 days post infection ( dpi ) . ( A ) Trypan blue stain showing visible symptoms after 2 dpi . ( B ) Microscopic analysis of the infection site and fungal hyphae 2 dpi visualized by Trypan blue stain . ( C ) Measurement of lesion size 3 dpi . Shown are means and standard errors ( n = 16 ) . No significant differences were observed ( Student’s t test ) . The experiment was repeated once with the same result . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 01210 . 7554/eLife . 01990 . 013Figure 8 . LYS1 mutation does not impinge on resistance towards Alternaria brassicicola . Five-week-old plants were infected with the necrotrophic fungus Alternaria brassicicola . Six 5 μl droplets of a spore suspension of 5 × 105 spores/ml were inoculated on the leaf; two leaves per plant were infected . The plants were analyzed for symptom development 7 , 11 , and 14 days post infection ( dpi ) . ( A ) Visible symptoms of four independent leaves at 14 dpi . ( B ) Disease symptoms 14 dpi visualized by Trypan blue stain . ( C ) Microscopic analysis of the infection site and fungal hyphae 14 dpi visualized by Trypan blue stain . ( D ) Calculation of the disease index at 7 , 11 , and 14 dpi . Shown are means and standard errors ( n = 16 ) . No significant differences were observed ( Student’s t test ) . The experiment was repeated once with the same result . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 013 To examine the role of LYS1 in immunity to bacterial infection , we infected wild-type plants or LYS1KD and LYS1OE lines with virulent Pto DC3000 . Two independent LYS1KD lines exhibited hypersusceptibility to bacterial infection ( Figure 9A ) , suggesting that lack of PGN-degrading activity results in reduced plant immunity . Likewise , immunity to hypovirulent Pto DC3000 ΔAvrPto/PtoB was compromised in these lines ( Figure 9B ) . Moreover , expression of the immune marker gene FRK1 upon administration of complex PGNs was greatly impaired in the LYS1KD mutants ( Figure 9C ) . These findings suggest that the enzymatic activity of LYS1 on PGN contributes substantially to plant immunity against bacterial infection . 10 . 7554/eLife . 01990 . 014Figure 9 . Manipulation of LYS1 levels causes hypersusceptibility towards bacterial infection and loss of peptidoglycan ( PGN ) -triggered immune responses . ( A and B ) Transgenic LYS1 plants are hypersusceptible to bacterial infection . Growth of Pseudomonas syringae pv . tomato ( Pto ) DC3000 ( A ) or Pto DC3000 ΔAvrPto/AvrPtoB ( B ) was determined 2 or 4 days post infiltration of 104 colony forming units ml−1 ( cfu/ml ) . Data represent means ± SD of six replicate measurements/genotype/data point . Representative data of at least four independent experiments are shown . ( C ) Transgenic LYS1 plants are impaired in PGN-induced immune gene expression . Leaves of wild-type plants or transgenic LYS1 plants were treated for 6 hr with 100 µg Bacillus subtilis PGN and total RNA was subjected to RT-qPCR using Flagellin responsive kinase ( FRK1 ) specific primers . EF1α transcript was used for normalization . Data represent means ± SD of triplicate samples , and shown is the result of one of three independent experiments . Statistical significance compared with wild-type ( *p<0 . 05 , Student’s t test ) is indicated by asterisks . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 01410 . 7554/eLife . 01990 . 015Figure 9—figure supplement 1 . Impact of weak LYS1 overexpression . ( A ) Transcript levels of LYS1 and the peptidoglycan ( PGN ) receptors LYM1 , LYM3 , and CERK1 in the strong LYS1 overexpressor line LYS1OE-1 compared with the weak overexpressor line LYS1OE-3 . Total RNA from untreated seedlings ( top panel ) or mature leaves ( bottom panel ) was subjected to RT-qPCR using specific primers for LYS1 , LYM1 , LYM3 , or CERK1 . EF1α transcript was used for normalization . Data represent means ± SD of triplicate samples . For mature leaves , CERK1 protein levels were also determined using an anti-CERK1 antibody ( bottom panel , inset ) . Ponceau S red staining of the large subunit of RuBisCO served as loading control . ( B ) Immunoblot analysis of protein extracts from the leaves of two independent LYS1OE lines ( LYS1OE-1 and LYS1OE-3 ) and wild-type plants . Total leaf protein was subjected to western blot analysis using α-tobacco class III chitinase ( α-Chit ) or green fluorescent protein ( α-GFP ) ( both from rabbit ) and an anti-rabbit HRP-coupled secondary antibody . Ponceau S red staining of the large subunit of RuBisCO served as loading control . ( C ) Growth of Pto DC3000 was determined 2 days post infiltration of 104 colony forming units ml−1 ( cfu/ml ) . Data represent means ± SD of six replicate measurements/genotype/data point . Statistical significance compared with wild-type ( *p<0 . 05; **p<0 . 01 , Student’s t test ) is indicated by asterisks . All experiments shown were repeated at least once . DOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 015 Unexpectedly , bacterial growth on LYS1OE lines was also significantly enhanced compared with that observed on wild-type plants ( Figure 9A , B ) . FRK1 transcript accumulation upon administration of complex PGN was also strongly reduced in LYS1 overexpressors ( Figure 9C ) . To exclude a direct effect of LYS1 overexpression on PGN receptor abundance , we examined transcript levels of LYM1 , LYM3 , and CERK1 but found no effect on the transcription of these receptor genes in the LYS1OE lines ( Figure 9—figure supplement 1A ) . Also , CERK1 protein levels were unaltered in the LYS1OE lines , whereas there was no CERK1 protein detectable in the cerk1-2 mutant ( Figure 9—figure supplement 1A ) . Moreover , we included the LYS1OE-3 line with only moderately increased LYS1 transcript and protein levels in mature leaves ( Figure 9—figure supplement 1A , B ) . Susceptibility to Pseudomonas infection in the LYS1OE-3 line was only slightly but not significantly increased ( p=0 . 064 , Student’s t test ) . These results indicate that lowering LYS1 expression levels , accompanied by lower LYS1 hydrolytic activity on PGN , brings down these lines close to wild-type . Thus , massive LYS1 overexpression and loss-of-function mutations are phenocopies of each other , irrespective of the fact that LYS1KD and LYS1OE lines show dramatic differences in LYS1 enzymatic activities ( Figure 4 ) . Altogether , we propose that LYS1 contributes to plant immunity to bacterial infection by decomposition of bacterial PGNs and generation of soluble PGN-derived patterns that trigger immune activation in a LYM1-LYM3-CERK1 receptor-complex-dependent manner .
It is generally little understood whether and how microbial patterns derived from complex extracellular assemblies , such as bacterial cell walls , are accessible to host PRRs for host immune activation in eukaryotes . This holds true for bacterial PGNs , but also for other patterns including bacterial LPS , flagellin , or fungus-derived chitin or glucan structures , all of which have been ascribed triggers of innate immunity in metazoans and plants ( Boller and Felix , 2009; Kumar et al . , 2013; Newman et al . , 2013; Pel and Pieterse , 2013 ) . Limited insight into the 3D structure of ligand–PRR complexes , as well as knowledge on ligand structural requirements for plant immune activation , suggests that small ligand epitopes are crucial for binding to host PRRs ( Liu et al . , 2012b; Sun et al . , 2013 ) . It is thus generally assumed that soluble fragments derived from complex microbial matrices serve as ligands for host PRRs and subsequent immune activation in both lineages . Two possible scenarios as to how soluble PGN fragments might be generated from macromolecular assemblies of cross-linked PGNs are discussed . First , during bacterial multiplication and cell wall biogenesis , large portions of soluble PGN fragments are shed into the extracytoplasmic space from which only 50–90% are recycled ( Park and Uehara , 2008; Reith and Mayer , 2011; Johnson et al . , 2013 ) . This implies that imperfect recycling of bacterial walls might serve as a source of soluble ligands for host PRRs sensing PGN ( Boudreau et al . , 2012; Wyckoff et al . , 2012 ) . Indeed , muramylpeptides spontaneously shed by Shigella flexneri directly stimulate NOD1-dependent immune responses in mammalian immune cells , and bacterial mutants impaired in PGN recycling hyperactivate host immunity ( Nigro et al . , 2008 ) . Second , host lysozyme activity has been demonstrated to generate soluble PGN ligands for NOD2 receptor-mediated immune activation and clearance of Streptococcus pneumoniae colonization in mice ( Callewaert and Michiels , 2010; Clarke and Weiser , 2011; Davis et al . , 2011 ) . Importantly , Davis et al . ( 2011 ) established a role for host lysozymes in PGN release from bacteria in the absence of detectable bacterial lysis . Likewise , Drosophila Gram-negative bacteria-derived binding protein 1 ( GNBP1 ) was shown to possess PGN-hydrolyzing activity and to deliver fragmented PGN to the PGN sensor , PGRP-SA ( Filipe et al . , 2005; Wang et al . , 2006 ) . Thus , both passive and active mechanisms of PGN decomposition appear to occur simultaneously during host pathogen encounters and might not be mutually exclusive . We here report on a lysozyme-like enzyme ( LYS1 ) that is produced in infected Arabidopsis plants and is capable of generating soluble PGN fragments from complex bacterial PGNs . LYS1 has been demonstrated to hydrolyze β ( 1 , 4 ) linkages between N-acetylmuramic acid and N-acetylglucosamine residues in PGN and between N-acetylglucosamine residues in chitin oligomers , thus closely resembling metazoan lysozymes . LYS1-generated fragments trigger immunity-associated responses in a PGN receptor-dependent manner . Activation of defenses has been further shown to occur in the two plants ( Arabidopsis and rice ) for which PGN perception systems have been described to date ( Willmann et al . , 2011; Liu et al . , 2012a ) . Importantly , Arabidopsis plants with strongly reduced LYS1 expression were impaired in immunity to bacterial infection , suggesting strongly that LYS1 function is an important element of the immune system of this plant . Notably , immunocompromised phenotypes in LYS1KD plants were comparable to those observed in either lym1 lym3 or cerk1 PGN receptor mutant genotypes ( Willmann et al . , 2011 ) . We further found that plants overexpressing LYS1 were also susceptible to bacterial infections , suggesting that defined LYS1 levels in wild-type plants are required for LYS1 immune function . The most compelling explanation for this phenotype is that PGN hyperdegradation ( in LYS1OE plants ) or lack of PGN degradation ( in LYSKD mutants ) are equally disadvantageous to plant immunity and that immune activation in Arabidopsis requires oligomeric PGN fragments of a particular minimum degree of polymerization ( DP ) . This view is supported by our findings that prolonged digestion of PGN by LYS1 ( Figure 6D ) or by mutanolysin ( Gust et al . , 2007 ) abolished the immunogenic activity of PGN . Likewise , immunogenic activities of fungal chitin or oomycete glucans have been reported to require defined minimum ligand sizes with a minimum DP of >5 ( Cheong et al . , 1991; Zhang et al . , 2002 ) . We therefore propose that LYS1 overexpression might result in PGN fragments of insufficient size , thereby mimicking the physiological status in LYS1KD mutants lacking major PGN hydrolytic activities . Plants produce various carbohydrate-degrading hydrolytic enzyme activities , some of which have been implicated in plant immunity to microbial infection , such as glucanases and chitinases ( van Loon et al . , 2006 ) . While it is often not entirely clear how these enzymes contribute to plant immunity , it is widely assumed that this is due to microcidal activities of these proteins . In our study we have shown that Arabidopsis LYS1 cleaves O-glycosidic bonds formed between GlcNAc ( indicative of chitinolytic activity ) as well as those formed between GlcNAc and MurNAc ( indicative of peptidoglycanolytic activity ) . However , we have been unable to demonstrate any deleterious effect of LYS1 overexpression on fungal infections , suggesting that B . cinerea and A . brassicicola at least are not affected by LYS1 function . Likewise , we have been unable to demonstrate direct bactericidal activity of LYS1 to P . syringae ( not shown ) , suggesting that the positive role of LYS1 in plant immunity to bacterial infection is not due to its direct inhibitory effect on bacterial fitness . This view is further supported by the fact that LYS1OE plants with strongly enhanced PGN hydrolytic activity do not exhibit enhanced immunity to Pseudomonas infections but become hypersusceptible to infection ( Figure 9 ) . We cannot rule out at this point LYS1-mediated bacterial lysis , which would likely also result in the release of immunogenic PGN fragments . We would like to emphasize , however , that our findings are in agreement with a predominant role of LYS1 in the generation of PGN fragments that subsequently can trigger plant immunity via PRRs . Hence , plant LYS1 functionally resembles recently described mammalian lysozymes that were shown to generate soluble PGN fragments for PGN receptor NOD2 , thereby mediating immunity to S . pneumoniae infection in mice ( Davis et al . , 2011 ) . LYS1 gene expression is strongly enhanced upon PAMP administration or bacterial infection while expression levels in naive plants are low . It is conceivable that the low constitutive LYS1 levels are sufficient to generate soluble PGN fragments from bulk PGN-containing bacterial walls which are then perceived via the LYM1-LYM3-CERK1 receptor complex . It is possible that the pathogen-inducible later increase in LYS1 activity could have further roles for generating diffusible signals that might serve innate immune activation , not only in cells that are directly in contact with invading microbes but also in cell layers adjacent to infection sites . A role for plant glycosyl hydrolases in immunogenic PAMP generation and immune activation has been proposed previously ( Mithöfer et al . , 2000; Fliegmann et al . , 2004 ) . An extracellular soluble bipartite soybean glucan binding protein ( GBP ) was shown to harbor 1 , 3-β-glucanase activity and binding activity for glucan fragments of DP >6 derived from intact glucans . Complex glucans constitute major constituents of various Phytophthora species , many of which are plant pathogens ( Kroon et al . , 2011 ) . It was therefore suggested that , during infection , GBP endoglucanase activity produces soluble Phytophthora-derived oligoglucoside fragments as ligands for the high-affinity binding site within this protein ( Fliegmann et al . , 2004 ) . While this study supported the concept of plant hydrolases tailor-making ligands for plant PRRs , causal evidence for the involvement of the endoglucanase activity in plant immunity was not provided . Eukaryotic PGN recognition proteins ( PGRP , PGLYRP ) are conserved from insects to mammals , bind PGN , and function in antibacterial immunity ( Cho et al . , 2005; Bischoff et al . , 2006; Dziarski and Gupta , 2010; Kurata , 2010 , 2014 ) . Some PGRP family members are non-enzymatic PRRs ( NOD1 , NOD2 ) while others possess PGN-degrading activities ( Gelius et al . , 2003; Wang et al . , 2003; Bischoff et al . , 2006; Dziarski and Gupta , 2010; Kurata , 2010 ) . PGN hydrolytic enzyme activities such as lysozymes have been ascribed functions in direct bacterial killing ( Cho et al . , 2005 ) and in generating soluble PGN fragments as ligands for PRRs ( Wang et al . , 2006; Davis et al . , 2011 ) . LYS1 constitutes the first plant lysozyme-type activity for which a role in host immunity has been established . LYS1 is capable of generating immunogenic fragments from complex PGNs , which themselves serve as ligands for the LYM1-LYM3-CERK1-PGN recognition complex in Arabidopsis . It is noteworthy that LYM1 and LYM3 are PGN recognition proteins that lack apparent intrinsic PGN-degrading activity . We conclude that metazoans and plants employ hydrolytic activities for the decomposition of bacterial PGNs during host immune activation . In addition to the established role of PGNs in pattern-triggered immune activation , host-mediated degradation of bacterial PGNs constitutes another conserved feature of innate immunity in both lineages . However , as the molecular components involved differ structurally among phyla , both facets of PGN-mediated immunity might have evolved convergently .
A . thaliana Columbia-0 wild-type and N . benthamiana plants were grown on soil as previously described ( Brock et al . , 2010 ) . T-DNA insertion lines for LYS1 ( lys1-1 , WiscDsLox387C11; lys1-2 , SALK_095362; lys1-3 , CSHL_ET14179 ) were obtained from the Nottingham Arabidopsis Stock Centre . The transgenic pPR1::GUS and secGFP lines and the lym1 lym3 and cerk1-2 mutants have been described previously ( Shapiro and Zhang , 2001; Teh and Moore , 2007; Willmann et al . , 2011 ) . Rice ( Oryza sativa ) suspension cell cultures were grown in MS-medium ( 4 . 41 g/l MS salt , 6% [wt/vol] sucrose , 50 mg/l MES , 2 mg/l 2 , 4-D ) at 150 rpm and sub-cultured every week . Bacterial strains P . syringae pv . tomato DC3000 or Pto DC3000 ΔAvrPto/AvrPto , A . brassicicola isolate MUCL 20297 , and B . cinerea isolate BO5-10 were grown and used for infection assays on Arabidopsis leaves of 4–5-week-old plants as described previously ( Lin and Martin , 2005; Kemmerling et al . , 2007 ) . To visualize plant cell death and fungal growth on a cellular level , infected plants were stained with Trypan blue in lactophenol and ethanol as described elsewhere ( Kemmerling et al . , 2007 ) . Flg22 peptide has been described previously ( Felix et al . , 1999 ) . The purification of P . syringae pv . tomato PGN was performed as described previously ( Willmann et al . , 2011 ) . M . luteus cell wall preparations and B . subtilis PGN were purchased from Invivogen ( San Diego , California , United States ) , Cecolabs ( Tübingen , Germany ) , and Sigma-Aldrich ( Hamburg , Germany ) . PGNs and LPS ( from P . aeruginosa , Sigma-Aldrich ) were dissolved in water at a concentration of 10 mg/ml and stored at −20°C . Mutanolysin was purchased from Sigma-Aldrich . Recombinant His6-LYM1 and His6-LYM3 were expressed in E . coli and purified as previously described ( Willmann et al . , 2011 ) . As negative control , a protein purification using non-induced cultures harboring the His6-LYM3 construct was performed . For the p35S::LYS1 fusion constructs , a 903 bp fragment of the LYS1 coding sequence without STOP codon was cloned using the primers At5g24090gatF and At5g24090gatR ( Table 1 ) . In a second PCR , the recombination sites of the inserts were completed using the Gateway adaptor primers attB1 and attB2 ( Invitrogen , Darmstadt , Germany ) . The resulting fragments were then subcloned into pDONR201 ( Invitrogen ) by using the BP clonase reaction according to the manufacturer’s protocol ( Invitrogen ) and inserted into the binary expression vectors pK7FWG2 . 0 ( C-terminal GFP-tag ) ( Karimi et al . , 2002 , 2005 ) or pGWB17 ( C-terminal myc-tag ) ( Nakagawa et al . , 2007 ) by using the LR clonase reaction following the manufacturer’s protocol ( Invitrogen ) . For the pLYS1::GUS reporter construct , a 1948 bp fragment of the LYS1 promoter sequence was amplified from Arabidopsis Col-0 genomic DNA using the primers At5g24090gatF2 and At5g24090gatR2 ( Table 1 ) , extended in a second PCR with Gateway adaptor primers attB1 and attB2 and subcloned into pDONR207 ( Invitrogen ) before being inserted into the binary expression vector pBGWFS7 ( Karimi et al . , 2002 , 2005 ) . 10 . 7554/eLife . 01990 . 016Table 1 . Primers used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 01990 . 016AGIPrimer nameSequence 5′ → 3′At5g24090 ( LYS1 ) At5g24090F1CCAGAGGTGGCATAGCCATCAt5g24090R1CATCTGGTGGGATATAGCCACAt5g24090FATGACCAACATGACTCTTCGAt5g24090RTCACACACTAGCCAATATAGAt5g24090RP2TGATGCCACGAGACTGACLP_N853931TGACGAACCATGATAAATGGGRP_N853931CATAACCTCACACTGTGCTCGLP_N595362TAGTGCATGCATGTTAAACCGRP_N595362AGCTCCTCAATGTCCATTTCCSalk-LbaTGGTTCACGTAGTGGGCCATCGDs5-1GAAACGGTCGGGAAACTAGCTCTACWisc-Lba ( p745 ) AACGTCCGCAATGTGTTATTAAGTTGTCAt5g24090FqCACTTGCACCCATTTTGGCAt5g24090RqCCTCGACCCAATCGAGTAAt5g24090miR-sGATTTGACGTAAGCATACCGCCCTCTCTCTTTTGTATTCCAt5g24090miR-aGAGGGCGGTATGCTTACGTCAAATCAAAGAGAATCAATGAAt5g24090miR*sGAGGACGGTATGCTTTCGTCAATTCACAGGTCGTGATATGAt5g24090miR*sGAATTGACGAAAGCATACCGTCCTCTACATATATATTCCTAt5g24090gatFAAAAAGCAGGCTACATGACCAACATGACTCTTCGAt5g24090gatRAGAAAGCTGGGTACACACTAGCCAATATAGATGAt5g24090gatR-STOPAGAAAGCTGGGTATCACACACTAGCCAATATAGAt5g24090gatF2AAAAAGCAGGCTATGCCGTAGGCGAGTGTTTCAt5g24090gatR2AGAAAGCTGGGTGTTTTTGGTTAAAGATGTTTGAt1g07920/30/40 ( EF1α ) Ef1α-100-fGAGGCAGACTGTTGCAGTCGEf1α-100-rTCACTTCGCACCCTTCTTGAAt2g19190 ( FRK1 ) FRK1-FAAGAGTTTCGAGCAGAGGTTGACFRK1-RCCAACAAGAGAAGTCAGGTTCGTGAt4g02540At4g02540-qf1GTACCACGCCTATCTATTAt4g02540-qr1CTCATAGAAGAAACCAGCAAt1g05615At1g05615-qf1GGATTCCTATCTCTACCTAt1g05615-qr1TTCTTTACCCTCATCAACCAt5g58780At5g58780-qf1CTCTCTTCTCTTTTATCTCTCCAt5g58780-qr1CTCCTCCACTCCTACCACAAt3g51010At3g51010-qf1GCGTCGTGCTTTTATACTGAt3g51010-qr1TTCTTCCTCTTCGCCTCTAt1g21880 ( LYM1 ) Lym1-100-fTACAACGGTATAGCCAACGGCACTLym1-100-rGTGGAGCTAGAAGCGGCGCAAt1g77630 ( LYM3 ) Lym3-100-fACTTCGCAGCAGAGTAGCTCLym3-100-rAGCGGTGCTAATTGTTGCGGAt3g21630 ( CERK1 ) CERK1-100-fGGGCAAGGTGGTTTTGGGGCTCERK1-100-rCCGCCAAGAACTGTTTCGATGCCattB1GGGGACAACTTTGTACAAAAAAGCAGGCTattB2GGGGACCACTTTGTAC AAGAAAGCTGGGT For the generation of pLYS1::GUS and p35S::LYS1-GFP overexpression lines ( LYS1OE ) , wild-type Col-0 plants were transformed . Stable transgenic lines were generated using standard Agrobacterium tumefaciens-mediated gene transfer by the floral dip procedure ( Clough and Bent , 1998 ) . Expression of GFP fusion proteins was confirmed by immunoblot analysis using an anti-GFP antibody ( Acris Antibodies GmbH ) and anti-tobacco class III chitinase antibody ( kindly provided by Michel Legrand , IBMP Strasbourg , France ) . The histochemical detection of β-glucuronidase ( GUS ) enzyme activity in whole leaves of pLYS1::GUS or pPR-1::GUS transgenic Arabidopsis ( Shapiro and Zhang , 2001 ) was determined as described earlier ( Gust et al . , 2007 ) . Artificial microRNA-mediated gene silencing was used to specifically knock down LYS1 in the Col-0 background as mutant lines carrying T-DNA insertions in the LYS1 gene were unavailable . The Web microRNA Designer ( WMD; http://wmd . weigelworld . org ) was used to select the primers At5g24090miR-s , At5g24090miR-a , At5g24090miR*s , and At5g24090miR*s ( Table 1 ) for the generation of an artificial 21mer microRNA ( Schwab et al . , 2005 ) . The LYS1-specific amiRNA was then introduced into the vector miR319a pBSK ( pRS300 ) by directed mutagenesis . Knock-down of the LYS1 transcript level in stably transformed Col-0 plants ( LYS1 knock-down line , LYS1KD ) was determined by RT-qPCR using primers At5g24090Fq and At5g24090Rq listed in Table 1 . Off-target genes were identified using the Web microRNA Designer and transcript levels of the four top hits were determined by RT-qPCR using primers listed in Table 1 . A . tumefaciens-mediated transient transformation of N . benthamiana was performed as described previously ( Brock et al . , 2010 ) . The leaves were examined for expression of tagged fusion proteins 3–4 days post infection . Expression of fusion proteins was confirmed by immunoblot analysis using anti-myc antibodies ( Sigma-Aldrich ) and localization studies of GFP fusion proteins were carried out using a confocal laser-scanning microscope , as described elsewhere ( Willmann et al . , 2011 ) . From 5-week-old LYS1OE Arabidopsis plants , 100 g leaf tissue was frozen in liquid nitrogen and ground to fine powder . After addition of buffer A ( 20 mM sodium acetate , pH 5 . 2 , 0 . 01% [vol/vol] β-mercaptoethanol ) , the extract was incubated on ice overnight . After filtration through four layers of cheesecloth , the homogenate was centrifuged at 10 , 000× g for 30 min . The supernatant was loaded on a cation exchange column ( SP Sepharose , GE Healthcare , München , Germany ) equilibrated with buffer A . The column was washed with buffer A and proteins were eluted with a 0 to 1 M NaCl gradient in buffer A . The elution fractions were monitored for LYS1 activity with the 4-MUCT assay and protein purification was further confirmed by SDS-PAGE . 4-MUCT-active fractions were pooled and exchanged to buffer A using Vivaspin 3 kDa columns ( GE Healthcare ) . Protein concentration was determined using the Bradford assay . For LC-MS analysis , the Coomassie Blue-stained band of the major cleavage product of the purified LYS1-GFP sample was cut and in-gel digested with trypsin , as described elsewhere ( Borchert et al . , 2010 ) . LC-MS analyses of the peptides were done on an EasyLC nano-HPLC ( Proxeon Biosystems ) coupled to an LTQ Orbitrap Elite mass spectrometer ( Thermo Scientific ) as described elsewhere ( Conzelmann et al . , 2013 ) . MS data were processed using the software suite MaxQuant , version 1 . 2 . 2 . 9 ( Cox and Mann , 2008 ) and searched using Andromeda search engine ( Cox et al . , 2011 ) against a target-decoy A . thaliana database containing 33 , 351 forward protein sequences , the sequence of the LYS1-GFP fusion protein , and 248 frequently observed protein contaminants . MS data were processed twice , once considering only fully tryptic peptides and once considering only semi-tryptic peptides . In each case , two missed cleavage sites were allowed , carbamidomethylation of cysteine was set as the fixed modification , and N-terminal acetylation and methionine oxidation were set as variable modifications . Mass tolerance was set to 6 parts per million ( ppm ) at the precursor ion and 20 ppm at the fragment ion level . Identified peptide spectrum matches ( PSM ) were statistically scored by MaxQuant software by calculation of posterior error probabilities ( PEP ) ( Käll et al . , 2008 ) for each PSM . All PSMs having a PEP below 0 . 01 were considered as valid . For matrix-assisted laser desorption/ionization time-of-flight mass spectrometry ( MALDI-TOF-MS ) , protein digestion was performed as described elsewhere ( Maurer et al . , 2013; Amin et al . , 2014 ) . Briefly , the Coomassie Blue-stained band of the major cleavage product of the FPLC-purified LYS1-GFP sample was cut from the gel and destained with 30% ( vol/vol ) acetonitrile in 50 mM ammonium bicarbonate buffer . Disulfide bonds were reduced with 10 mM dithiothreitol ( DTT ) , 50 mM iodoacetamide was used to alkylate the cysteines followed by overnight protein digestion with mass spectrometry grade trypsin ( Promega , Manheim , Germany ) at 37°C . The digests were acidified by the addition of trifluoric acid ( TFA ) to a final concentration of 0 . 5% . Extracted peptides were desalted and mixed with an equal volume of 2 , 5-dihydroxybenzoic acid for Reflex-IV MALDI-TOF-MS ( Bruker Daltonics , Bremen , Germany ) measurements . Each spectrum was processed internally for trypsin autolysis before database search . The identity of protein was annotated using the SwissProt database ( 542782 sequences; 193019802 residues ) . To achieve the best possible results , the search parameters were as follows: one miscleavage was set for trypsin specificity and carbamidomethyl modification of cysteine and oxidation of methionine were selected as fixed and optional modifications , respectively . At a mass tolerance of 5 ppm , only protein scores greater than 70 ( p<0 . 05 ) were assigned significant with an expected value of 10−7 . Apoplastic washes were obtained from mature leaves of 4-week-old Arabidopsis plants by vacuum-infiltrating complete rosettes with 20 mM sodium acetate , pH 5 . 2 . Afterwards , leaf tissue was dipped dry on paper towels , placed in 50 ml Falcon tubes and spun at 1000× g for 5 min at 4°C . Collected fluids were concentrated tenfold using Vivaspin 500 columns with a 3 kDa cut-off ( GE Healthcare ) . Isolation of mesophyll protoplasts from leaves of 4–5-week-old Arabidopsis plants was performed according to a protocol described previously ( Yoo et al . , 2007 ) . Isolated protoplasts were resuspended in W5 solution ( 2 mM MES , pH 5 . 7 , 154 mM sodium chloride , 125 mM calcium chloride , 5 mM potassium chloride ) and incubated overnight at room temperature in the dark ( 2 × 105 protoplasts in 1 ml W5 solution ) . Subsequently , protoplasts were removed by centrifugation ( 20 s , 800 rpm , 4°C ) and secreted proteins in the medium were concentrated using Vivaspin 2 columns with a 10 kDa cut-off ( GE Healthcare ) . Total protein extracts from the harvested protoplast pellet of 4–5-week-old leaves of A . thaliana or N . benthamiana were prepared using 20 mM sodium acetate , pH 5 . 2 , supplemented with 15 mM β-mercaptoethanol and proteinase inhibitor cocktail ( Roche Applied Science , Mannheim , Germany ) . Approximately 40–60 µg total protein of the leaf extracts or 15 µg of the protoplast samples were added to the enzyme assays . For all in-tube enzyme assays described in the supplemental information , the reaction mix was incubated with shaking at 37°C in 20 mM sodium acetate , pH 5 . 2 . The 4-MUCT chitinase assay was performed as described ( Brunner et al . , 1998 ) . Briefly , the hydrolytic activity towards 4-MUCT ( Sigma-Aldrich ) was measured for 30 min and compared with that of 2 µg S . griseus chitinase ( Sigma-Aldrich ) . After enzyme incubation in 250 µl final volume of 0 . 05% ( wt/vol ) 4-MUCT , 20 µl of the reaction mixture were removed and added to 980 µl 0 . 2 M sodium carbonate solution . Free 4-MU ( Sigma-Aldrich ) was used for the generation of a standard curve . The intensity of the fluorescence was monitored with an MWG Sirius HT fluorescence microplate reader . For the zymogram , discontinuous cetyltrimethylammonium bromide ( CTAB ) polyacrylamide gel electrophoresis was performed using a 12% separating gel ( 43 mM potassium hydroxide [KOH] , 280 mM acetic acid , pH 4 . 0 , 12% [vol/vol] acrylamide bisacrylamide 37 . 5:1 , 8% [vol/vol] glycerol , 1 . 3% ammonium persulphate and 0 . 16% N , N , N , N-tetramethylethylene diamine [TEMED] ) overlaid by a 4% stacking gel ( 64 mM KOH , 94 mM acetic acid , pH 5 . 1 , 4% acrylamide , 1 . 25% ammonium persulphate and 0 . 125% TEMED ) . Prior to loading , the gel was pre-run using anode buffer ( 40 mM beta-alanine , 70 mM acetic acid , 0 . 1% CTAB , pH 4 . 0 ) and cathode buffer ( 50 mM KOH , 56 mM acetic acid , pH 5 . 7 , 0 . 1% CTAB ) for 1 hr at 250 Volts . Crude protein extracts were mixed with an equal volume of loading buffer ( 5 M urea , 25 mM potassium acetate , pH 6 . 8 , methylene blue ) and separated for 2 hr at 150 Volts and 4°C . After electrophoresis the CTAB gel was washed with 20 mM sodium acetate , then sprayed with 0 . 00625% ( wt/vol ) 4-MUCT in 20 mM sodium acetate , pH 5 . 2 , and incubated at 37°C for 30 min . Fluorescent bands were documented under UV light using the Infinity-3026WL/26MX gel imaging system ( PeqLab , Erlangen , Germany ) . The turbidity assay was done as described previously ( Park et al . , 2002 ) . Lytic activity towards M . luteus cell wall preparations or B . subtilis peptidoglycan ( Invivogen , Cecolabs ) was measured for 4 hr and compared with that of 1 µg hen egg-white lysozyme ( Sigma-Aldrich ) . 1 ml 0 . 02% ( wt/vol ) M . luteus cells or PGN suspension were incubated together with the enzyme and the decrease in absorbance at 570 nm of the suspension was measured with a spectrophotometer over time . The 4-MUC cellulase assay was performed using 4-methylumbelliferyl-β-D-cellobioside ( 4-MUC; Sigma-Aldrich ) as substrate . 1 mM 4-MUC was incubated in 20 mM sodium acetate ( pH 5 . 2 ) at 37°C for 1 hr in a 96 well plate with either 40 µg purified LYS1 or cellulase ( Duchefa , Haarlem , The Netherlands ) in a total volume of 100 µl . The reaction was stopped with 0 . 2 M sodium carbonate and the intensity of the fluorescence was monitored with an MWG Sirius HT fluorescence microplate reader using excitation and emission wavelengths of 365 nm and 455 nm , respectively . 500 µg/ml B . subtilis PGN was incubated with 140 µg LYS1 purified from LYS1OE plants or controls in 20 mM sodium acetate , pH 5 . 2 , at 37°C with shaking for 7 hr . After stopping the reaction by heating at 100°C for 10 min , the reaction was centrifuged and the supernatant analyzed by HPLC . The analyses were done by Cecolabs on an Agilent 1200 system with a Prontosil C18-RP column ( Bischoff Chromatography , Leonberg , Germany ) . The mobile phase was ( A ) 100 mM sodium phosphate , 5% ( vol/vol ) methanol and ( B ) 100 mM sodium phosphate , 30% ( vol/vol ) methanol . RNA isolation , semi-quantitative RT-PCR and RT-qPCR analysis were performed as described previously ( Kemmerling et al . , 2007; Willmann et al . , 2011 ) . For RT-qPCR , all quantifications were made in duplicate on RNA samples obtained from three independent experiments , each performed with a pool of 3–5 seedlings or two leaves . EF1α transcripts served normalization; corresponding water controls were set to 1 . The sequences of the primers used for PCR amplifications are given in Table 1 . The histochemical detection of β-glucuronidase ( GUS ) enzyme activity in whole leaves of pLYS1::GUS or pPR-1::GUS transgenic Arabidopsis ( Shapiro and Zhang , 2001 ) was determined as described earlier ( Gust et al . , 2007 ) . For the measurement of extracellular pH , 300 µl of cultured rice cells were transferred to 48 well plates and equilibrated at 150 rpm for 30 min . After addition of elicitors , the pH in the cell culture was monitored with an InLab Micro electrode ( Mettler Toledo , Gießen , Germany ) . For assays with LYS1-digested PGN , 100 µg/ml B . subtilis PGN was incubated with 40 µg LYS1 purified from LYS1OE plants or controls in 2 . 5 mM MES , pH 5 . 2 , at 37°C with shaking for 4 hr . After stopping the reaction by heating at 100°C for 10 min , the reaction was centrifuged and the supernatant used for triggering immune responses . Statistical significance between two groups has been checked using the Student’s t test . Asterisks represent significant differences ( *p<0 . 05; **p<0 . 01; ***p<0 . 001 ) . One-way analysis of variance ( ANOVA ) was performed for multiple comparisons combined with Duncan’s multiple range test indicating significant differences with different letters ( p<0 . 05 ) . | The immune response of plants and animals is triggered when cells detect small molecules that are present on the surface of the bacteria or fragments of peptidoglycans—the polymers that are a major component of the bacterial cell wall . The mechanisms by which small molecules trigger the immune response in plants have been widely studied in the model plant Arabidopsis thaliana , but less is known about the ways in which peptidoglycan fragments can initiate an immune response . Proteins called lysozymes are known to break peptidoglycans into smaller pieces in animals . Plants do not produce lysozymes , but they do produce other enzymes such as chitinases that have similar properties . Now Liu , Grabherr , et al . have shown that a chitinase called LYS1 acts as an enzyme that catalyzes the breakdown of peptidoglycans and has a central role in triggering the immune response of Arabidopsis . Plants that were genetically engineered to produce little or no LYS1 were highly susceptible to bacterial infection because there were no enzymes that could break the peptidoglycans into smaller fragments . However , plants that were engineered to produce very high levels of LYS1 also had a compromised immune response because the peptidoglycans were broken into fragments that were too small to be detected . The findings of Liu , Grabherr et al . demonstrate that animals and plants employ similar strategies to break down bacterial peptidoglycans to allow them to be detected by the immune system . However , as the enzymes responsible have different structures , they are likely to have evolved separately in plants and animals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2014 | Host-induced bacterial cell wall decomposition mediates pattern-triggered immunity in Arabidopsis |
To initiate DNA replication , cells first load an MCM helicase double hexamer at origins in a reaction requiring ORC , Cdc6 , and Cdt1 , also called pre-replicative complex ( pre-RC ) assembly . The essential mechanistic role of Cdc6 ATP hydrolysis in this reaction is still incompletely understood . Here , we show that although Cdc6 ATP hydrolysis is essential to initiate DNA replication , it is not essential for MCM loading . Using purified proteins , an ATPase-defective Cdc6 mutant ‘Cdc6-E224Q’ promoted MCM loading on DNA . Cdc6-E224Q also promoted MCM binding at origins in vivo but cells remained blocked in G1-phase . If after loading MCM , Cdc6-E224Q was degraded , cells entered an apparently normal S-phase and replicated DNA , a phenotype seen with two additional Cdc6 ATPase-defective mutants . Cdc6 ATP hydrolysis is therefore required for Cdc6 disengagement from the pre-RC after helicase loading to advance subsequent steps in helicase activation in vivo .
Eukaryotic cells allow pre-replicative complex ( pre-RC ) assembly at origins only once per cell cycle ( Li and Araki , 2013; Yardimci and Walter , 2014 ) to promote genome integrity ( Abbas et al . , 2013 ) . An inactive MCM ( mini-chromosome maintenance ) double hexamer is first loaded on double-stranded DNA ( dsDNA ) in late M- to early G1-phase and then activated in a subsequent step that requires conserved protein kinases ( Labib , 2010 ) . Cdc6 is a AAA+ ATPase family member that binds ORC ( origin recognition complex ) , and together these proteins function as an MCM ‘helicase loader’ ( Liang et al . , 1995; Cocker et al . , 1996; Neuwald et al . , 1999; Wendler et al . , 2012; Duderstadt and Berger , 2013 ) . Since the weakly hydrolyzed ATP-γ-S analogue traps an OCCM complex on DNA containing ORC-Cdc6-Cdt1 and a single hexamer of MCM ( all six MCM subunits are also AAA+ proteins ) , double hexamer loading requires ATP hydrolysis ( Sun et al . , 2013 ) . Following MCM complex loading , Cdt1 and Cdc6 disengage from the pre-RC to allow additional protein loading complexes to activate the MCM helicase ( Yardimci and Walter , 2014 ) . The AAA+ family shares multiple conserved motifs , including Walker A and B motifs that promote ATP binding and hydrolysis , as well as sensor 1 and 2 motifs that help couple changes in ATP/ADP occupancy with protein conformational changes ( Wendler et al . , 2012; Duderstadt and Berger , 2013 ) ( Figure 1A ) . Early mutational analysis of the Walker A motif agreed that budding yeast Cdc6 must bind ATP to execute its essential function ( Perkins and Diffley , 1998; Wang et al . , 1999; Weinreich et al . , 1999 ) . However , there have been conflicting reports on the role of Cdc6 ATPase activity in pre-RC assembly and yeast viability . Two different mutants altering the Walker B motif ( required for ATP hydrolysis ) gave opposite phenotypes: a Cdc6-E224G mutant was lethal and acted as a dominant negative mutant when overexpressed ( Perkins and Diffley , 1998 ) , but a double Cdc6-D223A , E224A mutant was viable suggesting the Cdc6 ATP hydrolysis might not be required for its essential function ( Weinreich et al . , 1999 ) . A Walker B DExx>AAxx mutant in Schizosaccharomyces pombe Cdc6 ( Cdc18 ) was also shown to be viable ( Liu et al . , 2000 ) . A polar residue in the sensor 1 motif at the tip of the β4-strand ( corresponding to N263 in budding yeast Cdc6 ) is also required for ATP hydrolysis in various AAA+ proteins ( Wendler et al . , 2012 ) . Mutation of Cdc6 N263 to A has been reported to result in normal growth at 24°C and no growth at 37°C in one study ( Schepers and Diffley , 2001 ) but was lethal at 30°C in another ( Takahashi et al . , 2002 ) . Alanine mutation of the Cdc6 sensor 2 arginine ( R332 ) had a modest effect on growth , suggesting it does not perform an essential role ( Schepers and Diffley , 2001 ) , but mutation of R332 to E was lethal ( Takahashi et al . , 2002 ) . Lastly , AAA+ members typically form multisubunit protein assemblies and an ‘R-finger’ residue in one subunit can stimulate ATP hydrolysis in an adjacent subunit . The R-finger is named for an analogous arginine first discovered in the Ras GTPase-activating protein , p120GAP ( Scheffzek et al . , 1997 ) . However , a triple alanine mutant spanning the budding yeast Cdc6 R-finger ( R274 ) gave no growth phenotype ( Schepers and Diffley , 2001 ) , suggesting that the Cdc6 R-finger is not required for stimulating either ORC or MCM ATPase activity . 10 . 7554/eLife . 05795 . 003Figure 1 . The Cdc6 Walker B catalytic residue is essential in yeast but is easily bypassed by intragenic suppressor mutations . ( A ) Structure of the Pyrobaculum aerophilum Cdc6 orthologue with bound ADP ( Liu et al . , 2000 ) highlighting the ATP binding pocket and key residues . The ‘DD’ Walker B residues are colored red and the sensor I residue is colored orange . ( B ) Multiple sequence alignment of Cdc6 Walker B motif . Uppercase letters in consensus are highly conserved . From top to bottom: Homo sapiens , Xenopus laevis , S . pombe , S . cerevisiae , Mus musculus , Drosophila melanogaster , C . elegans , and Pyrobaculum aerophilum Cdc6 homologues . ( C ) Viability analysis of Cdc6 Walker B motif mutants reveals only E224 is essential . M4466 ( cdc6Δ::ura3 pRS416-CDC6 ) was transformed with the indicated pMW71-derived plasmids ( listed in Supplementary file 2 ) and then restreaked onto SCM-Leu plates representing growth of CDC6/cdc6 or FOA plates , growth of cdc6 mutant only . ( D ) Growth of plasmid cdc6 derivatives transformed into wild-type yeast ( W303-1A ) reveals that multiple E224 substitutions have dominant growth affects . ( E ) Summary of Cdc6 mutational analysis . Shaded regions are non-viable mutants . ( F ) Flow cytometry profiles of strains M4759 ( K4055 × pRS415 , vector ) , M4758 ( K4055 × pMW71 , CDC6-WT ) , M4760 ( K4055 × pFJ21 , cdc6-EQ ) , and M4762 ( K4055 × pFJ230 , cdc6-NQ ) after addition of methionine to repress expression of wild type MET3p-CDC6 present in all strains . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 00310 . 7554/eLife . 05795 . 004Figure 1—figure supplement 1 . Growth properties of various cdc6 Walker B mutants . For ( A ) and ( B ) , M4466 ( W303-1A , cdc6Δ::ura3 ) containing only the indicated cdc6 alleles on pMW71 ( Supplementary file 2 ) were spotted onto YPD plates in a 10-fold dilution series and incubated at the indicated temperatures . ( A ) Alanine scanning mutants across the Cdc6 Walker B box . ( B ) Double and triple mutants within the ‘DEMD’ core region . The AAAD , AAMA , and AEAA mutants exhibited substantial temperature sensitivity . ( C ) The indicated Cdc6-E224 mutant alleles or wild-type CDC6 on pMW71 ( Supplementary file 2 ) were transformed into wild-type yeast ( W303-1A ) . These transformants were spotted using 10-fold serial dilutions onto SCM-Leu plates ( selecting for the pMW71-derivative ) to quantitatively measure their dominant negative growth phenotype . This indicates that all the E224 alleles exhibited some dominate growth effects over the wild type . Two separate codons for E224G were tested . ( D ) Diagram of pGAL-CDC6 plasmids . pFJ224 contains the GAL1 , 10 promoter 39 bp upstream of wild-type CDC6 coding sequence to give a GAL1p-CDC6 promoter fusion . This plasmid contains a 49-bp sequence ( CCGGGAATTTCCGGTGGTGGTGGTGGAATTCTAGACTCC ATG TCA GCT A ) predicted to form a stable multi-stem loop structure that overlaps the Cdc6 ATG , underlined . The 39 bp immediately preceding the Cdc6 ATG is derived from vector sequences . This pGAL-Cdc6 construct complements Cdc6 function when yeast is propagated on galactose but not on glucose media . ( E ) Overexpression of Cdc6-E224G or -E224Q mutant proteins from pFJ216 ( differing from pFJ224 only by a deletion of Cdc6 residues 2–49 , which significantly stabilizes Cdc6 protein ) or from pFJ235 ( differing from pJF224 only by a 36 bp deletion disrupting the stem loop ) on galactose causes complete growth inhibition of wild type yeast . In contrast , galactose-induced expression of these mutant proteins in the pFJ224 plasmid background causes only a mild growth inhibition , indicative of lower Cdc6 protein induction . The indicated plasmids ( Supplementary file 2 ) were transformed into wild-type yeast ( W303-1A ) , and 6 transformants each were streaked onto selective minimal media ( SCM-Ura ) containing galactose or glucose as the carbon source and incubated at 25°C for 3 days . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 004 An early in vitro MCM loading assay utilizing purified proteins together with crude extracts reported that Cdc6 ATP hydrolysis ( using the Cdc6-E224G mutant ) was required for MCM loading ( Randell et al . , 2006 ) . However , two recent reports using only purified proteins ( and using Cdc6-E224G or Cdc6-N263A mutants ) find that Cdc6 ATP hydrolysis is not required for efficient MCM loading in vitro ( Coster et al . , 2014; Kang et al . , 2014 ) . The Cdc6-N263A sensor 1 mutant was previously shown to be defective in ATP hydrolysis in vitro ( Speck and Stillman , 2007 ) . Coster et al . and Kang et al . have suggested that Cdc6 ATPase is required for a quality control step that ejects incomplete or non-functional MCM hexamers before loading . Surprisingly , no systematic mutational analysis of the Cdc6 Walker B residues has been reported . Here , we have combined yeast genetics and in vivo assays together with an in vitro MCM loading reaction to probe the role of Cdc6 Walker B motif residues for viability and growth , ATP hydrolysis , pre-RC assembly , and DNA replication in yeast . We find that Cdc6 ATP hydrolysis is required for yeast viability but not for MCM loading either in vitro or in vivo . Furthermore , since the requirement for Cdc6 ATP hydrolysis can be bypassed by degrading Cdc6 ATPase defective proteins after MCM loading , ATP hydrolysis is likely instead required to disengage Cdc6 from the pre-RC ( presumably an ORC-Cdc6-MCM intermediate ) to allow subsequent steps in MCM helicase activation .
The Walker B motif ( DExx ) contains a glutamate residue that is required for the hydrolysis of ATP to ADP ( Ogura et al . , 2004 ) and is highly conserved among Cdc6 orthologues ( Williams et al . , 1997 ) ( Figure 1A , B ) . An alanine scan across the yeast Cdc6 Walker B motif revealed that only the E224A mutation caused inviability ( Figure 1C , top panel ) . Mutation of E224 to G ( Perkins and Diffley , 1998 ) , L , or Q also caused inviability ( Figure 1C , middle panel ) . The other single alanine mutants with the Walker B motif exhibited nearly wild type growth over a wide temperature range ( Figure 1—figure supplement 1 ) . Expression of plasmid-borne cdc6-E224G , -A , -L , and -Q mutants from the Cdc6 promoter showed that all mutants were dominant negative for growth in otherwise wild-type yeast cells , with the E224Q mutant having the greatest effect ( Figure 1D; Figure 1—figure supplement 1 ) . Interestingly , all double and triple mutant combinations that changed the preceding aspartate D223 together with E224A or E224Q restored viability , including a triple mutant that removed all acidic residues from the Cdc6 DEMD Walker B motif ( Figure 1C , bottom panel ) . The growth properties of the double and triple mutants were surprisingly robust with the exception of the triple alanine mutants , ‘AAAD’ , ‘AAMA’ , and ‘AEAA’ , which were temperature sensitive at 37°C ( Figure 1—figure supplement 1 ) . The results of the mutational scan are summarized in Figure 1E . These data show that Cdc6 E224 is the only residue in the Walker B motif required for viability . Furthermore , the relative ease with which the requirement for the E224 catalytic residue could be bypassed by the intragenic suppressor mutations suggested that ATP hydrolysis might not be mechanistically coupled to MCM loading . To determine whether the Cdc6-E224Q mutant exhibited a cell cycle block , we propagated cells using a heterologous MET3p-CDC6 promoter fusion ( Piatti et al . , 1995 ) and then measured DNA content using CDC6 plasmid alleles expressed from the CDC6 promoter after shutting off MET3p-CDC6 expression with methionine addition . The CDC6 and cdc6-DE ( 223 , 224 ) NQ ( or ‘NQ’ ) strains proliferated normally after repressing wild-type MET3p-CDC6 expression ( Figure 1F ) . However , the cdc6-E224Q mutant strain had a larger fraction of G1- and S-phase cells even at t = 0 hr ( when both wild type and mutant Cdc6 were expressed ) and largely arrested in G1-phase by t = 6 hr after shutting off wild-type Cdc6 expression , as did the cells lacking any additional copy of CDC6 . This indicates that cdc6-E224Q blocks progression into S-phase as expected for a defect in DNA replication initiation . We purified Cdc6-WT , Cdc6-E224Q , and Cdc6-NQ proteins from bacteria ( Figure 2A ) and then measured their ATPase activity . Cdc6 has no ATPase activity when assayed alone but its ATPase activity is measurable in the context of an ORC-Cdc6-DNA complex as an increase in ATP hydrolysis seen by ORC-DNA alone ( Randell et al . , 2006 ) . ORC had a low intrinsic ATPase activity on DNA but the addition of wild-type Cdc6 increased the overall ATPase activity by 4–5 fold , as seen previously ( Randell et al . , 2006 ) ( Figure 2B ) . In contrast , the Cdc6-E224Q or -NQ mutant proteins had no ATPase activity ( Figure 2B ) because they did not increase the ORC–DNA ATPase activity but instead suppressed the minimal ORC ATPase activity . This analysis also reveals that the NQ mutant promotes yeast viability without restoring Cdc6 ATPase activity . 10 . 7554/eLife . 05795 . 005Figure 2 . Cdc6 ATPase activity is not required for MCM loading in vitro . ( A ) Silver-stained 10% SDS gel with molecular weight standards and 100 ng of the indicated purified Cdc6 proteins . ( B ) Cdc6 ATPase assays . ( C ) MCM loading assay using the purified proteins shown on left . ‘Low salt’ shows proteins associated with DNA , ‘High salt’ wash reveals loaded MCM protein . Arrow marks Cdt1 . ( D ) MCM loading assay as in ( C ) using a more stringent ( 0 . 3 M K acetate ) low salt wash indicates that Cdc6-E224Q is stabilized on DNA relative to WT Cdc6 and Cdc6-NQ protein . Arrow marks Cdt1 , arrowhead marks Cdc6 . ( E ) ATPase assays with Cdc6-WT-ORC-DNA , Cdc6-E224Q-ORC-DNA , or Cdc6-NQ-ORC-DNA complexes after MCM-Cdt1 addition . The first two lanes show soluble MCM and MCM-Cdt1 ATPase activities as controls . ( F ) The double hexamer of Mcm2-7 loaded by Cdc6-E224Q after high salt wash ( 1 M NaCl ) of loading reactions is indistinguishable from that loaded by wild-type Cdc6 . 528 raw cryo-EM particle images were used for 2D classification and averaging for the dhMCM protein loaded with ORC-Cdc6-E224Q . 3217 raw particles were used to generate the averaged views of the dhMCM loaded with wild-type ORC-Cdc6 . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 00510 . 7554/eLife . 05795 . 006Figure 2—figure supplement 1 . MCM loading by ORC-Cdc6-E224Q gives rise to a heterogeneous mixture of intermediates by EM . ( A ) Negative stain-EM and ( B ) cryo-EM of complexes formed with ORC , Cdc6-E224Q , and MCM-Cdt1 proteins revealed a heterogeneous mixture of OCCM ( ORC-Cdc6-Cdt1-MCM ) and OCM ( ORC-Cdc6-MCM ) complexes; averages from 3655 individual particle images and 1436 cryo-EM raw particle images , respectively . The red arrows point to Cdt1 density , and blue arrows indicate the absence of Cdt1 . The structure of Cdc6-E224Q-containing OCCM particles is similar to wild-type OCCM assembled in the presence of ATP-γ-S . However , the Cdt1 density is variable in its location in the mutant OCCM and is of course , absent in OCM complexes . Most of the particles were OCCM . We do not know the exact ratio between OCCM and OCM because the views without Cdt1 density are not necessarily of OCM; they could be OCCM particles at slightly different side views that prevented visibility of Cdt1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 006 We next examined the ability of wild type and mutant Cdc6 proteins to load MCM using an in vitro assay with purified proteins ( Evrin et al . , 2009 ) . ORC , Cdc6 ( wild type or mutant ) , Cdt1 , and MCM were incubated at 24°C with replication origin-containing plasmid DNA linked to magnetic beads and then washed under low salt conditions . This allows detection of proteins specifically bound to origin DNA . Wild type and both mutant Cdc6 proteins bound to ORC and promoted binding of MCM to DNA ( Figure 2C , low salt ) . Interestingly , the E224Q mutant had a higher amount of Cdt1 bound than the wild type or Cdc6-NQ complex , as reported for a Cdc6-E224G mutant ( Randell et al . , 2006; Fernandez-Cid et al . , 2013 ) , suggesting build up of an intermediate Cdt1-containing complex prior to dhMCM ( double hexamer MCM ) loading . ‘Loaded’ dhMCM that encircles dsDNA is resistant to a 0 . 5 M high salt wash but ORC , Cdc6 , Cdt1 and any MCM proteins that are merely ‘associated’ with DNA are not . Both mutant Cdc6 proteins were capable of loading MCM protein , albeit less efficiently than wild-type Cdc6 ( Figure 2C , high salt ) . The increased Cdt1 retention in low salt by Cdc6-E224Q was also seen in reactions treated with a more stringent ( 0 . 3 M K acetate ) low salt wash ( Figure 2D ) . In addition , Cdc6-E224Q protein was retained within the low salt complex but wild-type Cdc6 and the Cdc6-NQ protein were significantly less stable ( Figure 2D , Cdc6 is marked with an arrowhead ) . Cdc6-E224G is also defective in dissociating from the pre-RC relative to wild-type Cdc6 ( Coster et al . , 2014 ) . Cdc6 ATP binding is required for interaction with ORC and for pre-RC assembly ( Perkins and Diffley , 1998; Weinreich et al . , 1999; Speck et al . , 2005 ) and since Cdc6-E224Q and Cdc6-NQ proteins formed similar pre-RC complexes to wild-type Cdc6 ( Figure 2C , D , low salt ) , they are not defective in ATP binding . Although ORC-Cdc6-E224Q complexes exhibited no ATPase activity ( Figure 2B ) , substantial ATPase activity was evident after MCM and Cdt1 protein addition ( Figure 2E ) , at about two-thirds the level of the wild-type reaction ( Fernandez-Cid et al . , 2013 ) . Therefore , Cdc6-E224Q does not prevent MCM ATPase activity , which might be sufficient to support double hexamer loading . Recent experiments using multiple ATP binding and ATPase-defective MCM mutants in yeast support the model that MCM ATPase activity is necessary for efficient double hexamer loading ( Coster et al . , 2014; Kang et al . , 2014 ) . Using cryo-EM , the Mcm2-7 double hexamer loaded by ORC and mutant Cdc6 has the characteristic four-layered architecture in side views comprised of the two distal C-terminal AAA layers and the two middle N-terminal layers from two Mcm2-7 hexamers in their side views . This structural organization is identical to the double hexamer loaded by the wild-type ORC-Cdc6 ( Figure 2F ) . As mentioned previously , ATP-γ-S traps a normally transient ORC-Cdc6-Cdt1-MCM ( OCCM ) intermediate on the DNA with a single MCM hexamer encircling dsDNA but with a small gap in the hexamer at the Mcm2-Mcm5 interface ( Bochman and Schwacha , 2010; Costa et al . , 2011; Sun et al . , 2013; Samel et al . , 2014 ) . OCCM complexes are not seen with wild-type proteins in ATP , since ATP hydrolysis quickly promotes Cdt1 release to give a ‘OCM’ intermediate complex and then double hexamer loading ( Yardimci and Walter , 2014 ) . Both negative staining-EM and cryo-EM revealed a heterogeneous mixture of OCCM and OCM complexes formed with the Cdc6-E224Q protein in the presence of ATP ( Figure 2—figure supplement 1 ) that were not seen with wild-type Cdc6 . These data also suggest that Cdc6 ATP hydrolysis promotes Cdc6 and Cdt1 release from the pre-RC ( Speck et al . , 2005; Fernandez-Cid et al . , 2013 ) since these OCCM and OCM reaction intermediates accumulate with the Cdc6-E224Q mutant . Taken together , the biochemical data show that the Cdc6-E224Q protein lacks ATPase activity , is defective in Cdc6 release from the pre-RC , but can promote dhMCM loading onto DNA . We used MCM ChIP to test whether Cdc6-E224Q could promote MCM binding to ARS sequences in vivo . Since Cdc6-E224Q expression was dominant to the wild type , we arrested cells expressing only cdc6-1 ( a temperature-sensitive cdc6 allele ) in G2/M using nocodazole at 25°C , shifted cells to 37°C , and simultaneously turned on expression of an additional wild-type or mutant Cdc6 protein under the control of the Gal1 promoter . Cells were then released from the G2/M phase block at 37°C in the presence of galactose and harvested during a G1-phase pheromone block for MCM ChIP ( Figure 3A , B ) . Although MCM protein was not detected at ARS305 in the nocodazole arrest or in the absence of Cdc6 expression , all three mutant proteins ( E224Q , NQ , and NQMN ) promoted MCM binding to ARS305 , as did wild-type Cdc6 ( Figure 3C ) . 10 . 7554/eLife . 05795 . 007Figure 3 . Cdc6 ATPase mutants promote MCM origin binding in yeast . ( A ) Cell synchronization protocol for MCM ChIP . ( B ) PCR products amplified in base pairs surrounding ARS305 . ( C ) All the cdc6 ATPase mutants tested promote MCM binding to origins similar to the wild-type CDC6 regardless of whether they complement yeast viability . Strains used: M378 ( cdc6-1 ) , M4455 ( cdc6-1 GAL1p-CDC6::LEU2 ) , M4531 ( cdc6-1 GAL1p-cdc6-E224Q::LEU2 ) , M4513 ( cdc6-1 GAL1p-cdc6-NQ::LEU2 ) and M4464 ( cdc6-1 GAL1p-cdc6-NQMN::LEU2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 007 To test whether the MCM protein loaded by Cdc6-E224Q could promote DNA replication , that is , whether the dhMCM was functional , we expressed just Cdc6-E224Q ( or Cdc6-WT ) in a 2-hr window from M- to G1-phase and then monitored S-phase progression after switching off Cdc6 ( mutant or WT ) expression to allow Cdc6 removal . Cdc6 is an unstable protein ( t1/2 ≤ 5 min ) ( Piatti et al . , 1995 ) and so we expected that glucose addition ( which strongly represses any further Cdc6 expression ) would be followed by the disappearance of Cdc6 protein . Cells were synchronized similarly as in Figure 3 , except after release from the G2/M block in galactose at 37°C , glucose was added 90 min later to repress the Gal1 promoter ( Figure 4A ) . When no additional CDC6 allele was expressed after the mitotic release , cells arrested with a 1C DNA content for the duration of the experiment and at later time points some cells had a sub-G1 DNA content ( Figure 4A , leftmost time course ) , indicative of the reductional mitosis seen in the absence of CDC6 ( Piatti et al . , 1995 ) . When wild-type Cdc6 was expressed , cells progressed into G1 and S-phase quickly between the 90- and 120-min time points . When wild-type Cdc6 expression was repressed by glucose addition at t = 120 , Cdc6 protein gradually disappeared ( Figure 4B ) and cells cycled approximately once more during the experiment . In contrast , if Cdc6-E224Q was continually expressed , cells arrested in G1-phase after release from the G2/M block and remained arrested ( Figure 4A , asterisks ) . As a control , continual expression of wild-type Cdc6 or the Cdc6-NQ protein from the Gal1 promoter did not cause a G1 arrest ( Figure 4—figure supplement 1 ) as published previously for wild-type Cdc6 ( Perkins and Diffley , 1998; Schepers and Diffley , 2001 ) . Remarkably , if Cdc6-E224Q expression was instead repressed 90 min after release ( t = 120 ) , the protein was rapidly eliminated ( Figure 4B , bottom ) , and these cells entered a single round of DNA replication as indicated by the substantial S-phase population 30 min after glucose addition ( Figure 4A , rightmost time course ) . This could only occur if a large number of replication origins initiated DNA replication following Cdc6-E224Q removal . 10 . 7554/eLife . 05795 . 008Figure 4 . Cdc6 removal after MCM loading bypasses ATPase requirement and is sufficient to allow DNA replication . ( A ) Synchronization protocol ( left ) and flow cytometry profiles ( right ) of yeast strains M378 ( cdc6-1 ) , M4455 ( cdc6-1 GAL1p-CDC6::LEU2 ) , and M4531 ( cdc6-1 GAL1p-cdc6-E224Q::LEU2 ) . Asynchronous cells were arrested in G2/M for 3 hr with nocodazole , shifted up to 37°C for 30 min ( from t = 0 to 30 min ) , and then released into G1-phase at 37°C expressing no additional Cdc6 , GAL1p-CDC6 or GAL1p-cdc6-E224Q . GAL1 promoter-driven CDC6 expression was shut off at t = 120 min by the addition of glucose except where marked by an asterisk . Identical flow profiles to M4531 were seen using an independent cdc6-1 GAL1p-cdc6-E224Q integrant strain , M4530 . ( B ) Cdc6 Western blots ( top panels ) and total protein ( bottom panels ) by Ponceau S staining of the samples are shown in panel A . The addition of glucose at 120 min causes Cdc6 wild-type protein to disappear and this occurs more rapidly for the Cdc6-E224Q mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 00810 . 7554/eLife . 05795 . 009Figure 4—figure supplement 1 . Overproduction of wild-type or Cdc6-NQ proteins does not cause a G1-arrest . ( A ) Cell synchronization protocol ( right ) and flow cytometry profiles ( left ) of yeast strains M378 ( cdc6-1 ) , M4455 ( cdc6-1 GAL1p-CDC6::LEU2 ) , and M4513 ( cdc6-1 GAL1p-cdc6-NQ::LEU2 ) . Asynchronous cells were arrested in G2/M for 3 hr with nocodazole , shifted up to 37°C for 30 min ( from t = 0 to 30 min ) , and then released into G1-phase at 37°C expressing no additional Cdc6 , GAL1p-CDC6 , or GAL1p-cdc6-NQ . Cells overexpressing wild-type Cdc6 or Cdc6-NQ proteins proceed normally into S-phase . ( B ) Cdc6 Western blots ( top panels ) and total protein ( bottom panels ) by Ponceau S staining of the samples is shown in panel A . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 009 This phenotype was not unique to the Cdc6-E224Q mutant . Recently , Cdc6-E224G has also been shown to efficiently load MCM in vitro using only purified proteins ( Coster et al . , 2014; Kang et al . , 2014 ) and this allele is also defective in ATP hydrolysis ( Randell et al . , 2006; Speck and Stillman , 2007 ) . We integrated this mutant into yeast under the control of the Gal1 promoter and repeated the experiment shown in Figure 4 . When cells were released from the G2/M block continually expressing Cdc6-E224G ( profile marked with asterisks in Figure 5A ) , they entered G1 phase and remained blocked in G1-phase like the cdc6-1 control , indicating a failure to initiate DNA replication . Thus , cdc6-E224G behaves as a null mutant in agreement with the genetic analysis in Figure 1 and previous data ( Perkins and Diffley , 1998 ) . If Gal1p-cdc6-E224G was instead repressed 90 min after release from the nocodazole block ( t = 120 ) and the protein was then degraded ( Figure 5B bottom ) , cells entered S-phase and completely replicated DNA ( Figure 5A , rightmost time course ) . Thus , the Cdc6-E224G mutant must also load MCM proteins in vivo ( as it does in vitro ) , but subsequent steps in DNA replication initiation are blocked by its persistent expression . 10 . 7554/eLife . 05795 . 010Figure 5 . Cdc6-E224G ATPase mutant is also defective in G1 progression but its degradation promotes DNA replication . ( A ) Synchronization protocol ( left ) and flow cytometry profiles ( right ) of yeast strains M378 ( cdc6-1 ) and M4766 ( cdc6-1 GAL1p-cdc6-E224G::LEU2 ) . Asynchronous cells were arrested in G2/M for 3 hr with nocodazole , shifted up to 37°C for 30 min ( from t = 0 to 30 min ) , and then released into G1-phase at 37°C expressing no additional Cdc6 or GAL1p-cdc6-E224G as in Figure 4 . ( B ) Cdc6 Western blots ( top panels ) and total protein ( bottom panels ) by Ponceau S staining of the samples shown in panel A . ( C ) Model for the role of Cdc6 ATP hydrolysis in DNA replication initiation ( see text ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 01010 . 7554/eLife . 05795 . 011Figure 5—figure supplement 1 . Growth phenotypes of cdc6 mutants altering additional residues that potentially affect ATP hydrolysis or ATP sensing . Mutational summary of Box VI , sensor 1 , Box VII ( R-finger ) , Box VII' , and Box VIII ( sensor 2 ) regions of Cdc6 . Yellow highlighting indicates that the mutation gave a lethal phenotype in yeast at 25°C . Plasmids from this study or previous published references for each mutant are listed at the left , and growth properties are listed at the right . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 01110 . 7554/eLife . 05795 . 012Figure 5—figure supplement 2 . Expression of the Cdc6-N263A ATPase mutant protein is dominant negative for growth , causes a G1 block , but loads functional MCM that can promote DNA replication after Cdc6-N263A degradation . ( A ) pFJ235 ( pRS416 Gal1p-CDC6 ) and mutant derivatives pFJ237 ( Gal1p-cdc6-E224Q ) , pFJ404 ( Gal1p-cdc6-N263A ) , and pFJ412 ( Gal1p-cdc6-N263L ) ( Supplementary file 2 ) were transformed into wild-type yeast , W303-1A . Multiple transformants were streaked onto SCM-Ura plates containing glucose or galactose . ( B ) Cell synchronization protocol ( left ) and flow cytometry profiles ( right ) of yeast strains M378 ( cdc6-1 ) and M4763 ( cdc6-1 GAL1p-cdc6-N263A::LEU2 ) . Asynchronous cells were arrested in G2/M for 3 hr with nocodazole , shifted up to 37°C for 30 min ( from t = 0 to 30 min ) , and then released into G1-phase at 37°C expressing no additional Cdc6 or GAL1p-cdc6-N263A . GAL1 promoter-driven cdc6-N263A expression was shut off at t = 120 min by the addition of glucose except where marked by an asterisk . ( C ) Cdc6 Western blots ( top panels ) and total protein ( bottom panels ) by Ponceau S staining of the samples are shown in panel ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05795 . 012 We constructed additional mutations in Cdc6 residues implicated in ATP hydrolysis or sensing bound nucleotide to further explore this function ( Figure 5—figure supplement 1 ) , which also lists previously published data for comparison . Single alanine substitutions of Cdc6 R270 or R274 ( the conserved R-finger residue ) gave no noticeable growth phenotype . Charge reversal mutations at these residues ( R270E or R274E ) also gave no phenotype indicating that these residues are not functioning as R-fingers . Mutation of charged residues to alanine in the conserved box VIII of AAA+ proteins including the R332 sensor 2 residue also gave no noticeable growth phenotype . In contrast , mutation of the sensor 1 asparagine residue ( N263 ) to alanine or leucine gave a lethal phenotype and strong overproduction of the Cdc6 N263A or N263L mutant proteins from the Gal1 promoter gave a complete dominant negative growth phenotype ( Figure 5—figure supplement 2; [Schepers and Diffley , 2001] ) , similar to the E224G or E224Q alleles . In summary , alanine or leucine mutations at the N263 sensor 1 residue were unique among the sensor 1 , sensor 2 , and ‘R-finger’ Cdc6 residues we tested in giving a lethal phenotype and both N263 mutant proteins were dominant negative for growth when over-expressed in wild-type yeast . Since the N263A mutant is also defective for ATP hydrolysis ( Speck and Stillman , 2007 ) but is capable of loading MCM proteins in vitro ( Coster et al . , 2014; Kang et al . , 2014 ) , we integrated this allele under the control of the Gal1 promoter in yeast and tested its role in DNA replication . Similar to the cdc6-E224Q and cdc6-E224G alleles , expression of only cdc6-N263A gave a growth arrest in G1 phase , but when the gene was repressed and the Cdc6-N263A protein subsequently degraded , cells entered S-phase ( Figure 5—figure supplement 2 ) . Cdc6 ATPase activity is required for viability in budding yeast as seen by the phenotypes of multiple E224 and N263 substitution mutants ( Figure 1; Figure 5—figure supplements 1 , 2; and [Perkins and Diffley , 1998; Takahashi et al . , 2002] ) . Using purified proteins , Cdc6 ATPase activity is not essential for MCM loading in vitro ( Figure 2; and [Coster et al . , 2014; Kang et al . , 2014] ) or in vivo ( Figure 3 ) but is required for initiating DNA replication ( Figures 4 , 5 ) . Since degradation of the ATPase defective Cdc6-E224Q , Cdc6-E224G and Cdc6-N263A mutant proteins after MCM loading promotes DNA replication ( Figures 4 , 5 , and Figure 5—figure supplement 2 ) , Cdc6 ATPase activity is required for Cdc6 release from the pre-RC , that is , an ORC-Cdc6 intermediate complex engaged with MCM ( see model in Figure 5C ) . So , Cdc6 activity is not required after MCM loading other than for its removal from the helicase-loading complex . Since we did not trap an intermediate complex with ORC-Cdc6-Cdt1 and MCM double hexamer in vitro , we cannot conclude whether MCM hexamers are loaded in tandem , from two separate OCCM structures , or as a double hexamer . Since Cdc6-E224Q is defective for ATP hydrolysis and suppresses ORC ATP hydrolysis in the ORC-Cdc6-DNA context ( Figure 2B ) , it appears that MCM ATP hydrolysis ( Figure 2E ) can drive the reaction forward in vitro to give double hexamer formation , albeit less efficiently that with wild-type Cdc6 ( Figure 2C , D ) . However , since blocking Cdc6 ATP hydrolysis in vivo prevents subsequent MCM activation , this suggests that the in vitro reactions do not capture some essential aspect of this reaction . For instance , the dhMCM engaged with ORC-Cdc6-E224Q may be much more stable on origins within chromatin in yeast , which have a small 125-bp nucleosome-free region on average ( Eaton et al . , 2010 ) . We isolated multiple intragenic suppressors of the Cdc6-E224A or E224Q single mutants in the highly conserved ‘DEMD’ core region of the Walker B motif , that is , AAMD , AAAD , AAMA , NQMD , NQMA , and NQNM ( Figure 1 ) . All of these intragenic suppressors also mutate the first aspartic acid residue indicating that the presence of this D223 residue blocks Cdc6 activity in vivo when Cdc6 cannot hydrolyze ATP . The NQMD mutation did not restore Cdc6 ATPase activity ( Figure 2B ) . This simplest explanation for the bypass phenotype is that a conformational change within the Cdc6-NQ mutant compensated for the loss of ATPase activity . This is supported by the fact that the AAMD , AAMA , AAAD mutants grew poorly but the NQMD and NQMN mutants , which have quite different substitutions at the ‘DE’ residues , grew very robustly at all temperatures ( Figure 1—figure supplement 1 ) . It is well established that ATP hydrolysis by AAA+ family members leads to conformational changes within these proteins , which drive the reactions forward , and this conformational change is communicated by residues in close proximity to the ATP/ADP binding pocket ( Hanson and Whiteheart , 2005; Wendler et al . , 2012; Duderstadt and Berger , 2013 ) . Recent work has discovered that the Cdc6 ATPase performs a quality control function in vitro ( Coster et al . , 2014; Kang et al . , 2014 ) . We suggest this is not its primary role in vivo since blocking a quality control function for the Cdc6 ATPase ( i . e . , ejecting incompletely assembled MCM hexamers or MCM-Cdt1 heptamers that bind the ORC-Cdc6 loader before loading onto dsDNA ) should not give rise to a lethal phenotype . The hexameric replicative helicase structure and loading mechanisms have been honed over a long evolutionary history to be highly efficient ( Forsburg , 2004; Bae et al . , 2009; Slaymaker and Chen , 2012 ) and probably very few errors occur in vivo . Our data strongly suggest that the essential role of the Cdc6 ATPase is to disengage Cdc6 from the pre-RC after MCM is loaded . Interestingly , this role of the Cdc6 ATPase in initiating DNA replication is similar to bacterial helicase loaders: the AAA+ Escherichia coli helicase loader DnaC does not require ATP hydrolysis to load the hexameric DnaB helicase at oriC—but DnaC ATP hydrolysis leads to DnaC disassembly and release from a DnaC–DnaB complex that actually inhibits DnaB helicase activity ( Davey et al . , 2002; Arias-Palomo et al . , 2013 ) . Our data therefore reveal a broad conservation of replicative helicase loading mechanisms across kingdoms , which can hopefully be further elucidated by crystallographic or other biophysical methods using wild type and mutant helicase loading complexes .
Yeast strains and plasmids used in this study are listed in Supplementary files 1 , 2 . All strains were derivatives of W303-1A ( MATa ade2-1 trp1-1 can1-100 leu2-3 , -112 his3-11 , -15 ura3 ) . GAL1-CDC6 genes were integrated into the LEU2 locus as previously described ( Weinreich et al . , 1999 ) . Genomic DNA from LEU2+ transformants was isolated , digested with SalI , and then probed by Southern blotting to identify single or double integrants . The presence of individual mutations was confirmed by sequencing the integrated GAL1p-CDC6 allele . Single or double mutant integrant strains were chosen for further analysis based on which one gave the most similar amount of galactose-induced Cdc6 protein compared to the single wild-type CDC6::LEU2 integrant , M4455 . Point mutations and deletions within CDC6 were generated by site-directed mutagenesis using the QuikChange system ( Agilent Technologies , Santa Clara , CA ) . The wild-type GAL1 , 10 promoter was cloned on a ∼800-bp EcoR1-BamHI fragment into pRS416 , and the wild-type CDC6 coding sequence was cloned 39 bp downstream of the GAL1 , 10 promoter cassette to give pFJ224 . A potential 49-bp multi-stem loop structure that overlapped the first few codons of Cdc6 was removed by deleting 36 bp between the BamHI site on pFJ224 and the Cdc6 ATG , giving rise to pFJ235 ( Figure 1—figure supplement 1 ) . A SalI–SacI fragment of pFJ224 containing pGAL1-CDC6 wild type was cloned into pRS405 to give pFJ304 . The mutants pFJ305 ( E224Q ) , pFJ306 ( NQ ) , pFJ307 ( NQMN ) , pFJ418 ( N263A ) , and pFJ419 ( E224G ) were derived from pFJ304 by QuikChange . pGEX-CDC6 was described previously ( Speck et al . , 2005 ) . The E224Q and DE ( 223 , 224 ) NQ mutations were introduced into pGEX-CDC6 using QuikChange to yield pFJ259 and pFJ263 , respectively . Yeast cells were cultured in YPD ( 1% yeast extract , 2% peptone , 2% D-glucose ) or synthetic complete medium ( SCM ) as described ( Rose et al . , 1990 ) . The flow cytometry profiles of yeast cells were performed as described ( Haase and Reed , 2002 ) with some modifications . Approximately 1 × 107 cells were harvested , washed , and cell pellets were resuspended in 400 μl of water and fixed by adding 950 μl of 100% ethanol . Samples were kept at −20°C until further processed . Fixed cells were pelleted , washed with cold water , resuspended in 0 . 5 ml of 50 mM Tris pH 8 , 15 mM NaCl containing 2 μg/ml RNase A , and incubated at 37°C overnight . Proteinase K was added to the samples for 1-hr 50°C incubation . Treated samples were pelleted again , resuspended in 0 . 5 ml 50 mM Tris pH 7 . 5 , and sonicated to break-up cell clumps , if any . The samples were stained with SYTOX Green ( Thermo Fisher , Grand Island , NY ) at a final concentration of 2 μM and analyzed using a MoFlo Astrios ( Beckman Coulter , Miami , FL ) . The Mcm2 ChIP assays were performed essentially as described ( Pappas et al . , 2004 ) using strains M378 , M4455 , M4464 , M4513 , and M4531 . Overnight log cultures were diluted into fresh YP-Raf medium to a uniform OD and arrested with nocodazole for 3 . 5 hr . Galactose was added to a final 2% , and cells were shifted up to 37°C to inactivate cdc6-1 . Cells were collected and resuspended in 37°C YP-Gal medium containing 5 μg/ml alpha factor and then treated with formaldehyde at the indicated time points . Monoclonal antibodies against Mcm2 ( Mcm2-49 , a gift of Bruce Stillman , Cold Spring Harbor Laboratory ) were used for the ChIP assay after shearing chromatin to an average of ∼500 bp . The DNA primers used to amplify ARS305 ( 310 bp ) and surrounding regions are ‘305-350F’ GTCCCTGTAATTGGAAGAGC , ‘305-350R’ ACCACATAATGTGAAGCCTT , ‘305-310F’ ATGAGGTCTCTAGCAAAAAG , ‘305-310R’ TACTGTCCGGTGTGATTTAT , ‘305-239F’ TGAGCCTTCTAATAATAAAGGGGA , and ‘305-239R’ GTAACGTACCATTTTTGATCTTGG . The yeast strains M378 , M4455 , M4513 , M4530 , M4531 , M4763 , and M4766 were grown overnight in YP-Raffinose at 25°C and then treated with 15 μg/ml of nocodazole for 3 hr . 2% galactose ( final ) was then added , and cells were incubated at 37°C for a further 30 min . Nocodazole-arrested cells were washed once with water then released from the G2/M arrest in YP-Gal at 37°C . 2% glucose was added to the cultures at the indicated times unless stated otherwise . The pre-RC assay was performed as described with minor modifications ( Evrin et al . , 2009 ) . Here , a one-step reaction was used . 40 nM ORC , 80 nM ( wt or mutant ) Cdc6 , 40 nM Cdt1 , 40 nM MCM2-7 in buffer A ( 50 mM HEPES-KOH pH 7 . 5 , 100 mM KGlu , 10 mM MgAc , 50 μM ZnAc , 3 mM ATP , 5 mM DTT , 0 . 1% Triton X-100 , and 5% glycerol ) were added to 6 nM linear pUC19-ARS1 DNA coupled to magnetic beads for 15 min at 24°C . Beads were washed 2 times with buffer A containing 300 mM KGlu plus 1 mM EDTA , or 2× short washes with buffer A-1 ( 50 mM HEPES-KOH pH 7 . 5 , 1 mM EDTA , 300 mM KAc , 10% glycerol , 0 . 1% Triton X-100 , and 5 mM DTT ) , or 3× with buffer B ( 50 mM HEPES-KOH pH 7 . 5 , 1 mM EDTA , 500 mM NaCl , 10% glycerol , 0 . 1% Triton X-100 , and 5 mM DTT ) before digestion with 1 U of DNase I in buffer A plus 5 mM CaCl2 for 2 min at 24°C . The samples were separated by SDS-PAGE and analyzed by silver staining . ORC was expressed by using baculovirus-infected cells and purified as described ( Klemm et al . , 1997 ) . Cdc6 ( WT and mutants ) and Cdt1 were expressed in bacteria and purified as described ( Speck et al . , 2005; Evrin et al . , 2009 ) . Mcm2-7 were expressed in Saccharomyces cerevisiae and purified as described ( Evrin et al . , 2009 ) . The pre-RC were assembled in a one-step reaction: 40 nM ORC , 80 nM Cdc6 , 40 nM Cdt1 , and 40 nM MCM2–7 in buffer A ( 50 mM HEPES–KOH pH 7 . 5 , 100 mM KGlu , 10 mM MgAc , 50 mM ZnAc , 3 mM ATP , 5 mM DTT , 0 . 1% Triton X-100 , and 5% glycerol ) were added to 6 nM pUC19-ARS1 plasmid beads at 24°C ( Evrin et al . , 2009 ) . After 10 min ( for pre-RC intermediate analysis ) or 40 min ( for MCM2-7 double-hexamer analysis ) , the beads were washed 3 times with buffer A ( pre-RC intermediate ) or B ( MCM2-7 double-hexamer ) and 3 times with buffer C ( 50 mM HEPES-KOH [pH 7 . 5] , 100 mM potassium acetate , 5 mM magnesium acetate , 5 mM CaCl2 ) and eluted with 1 U DNase I in 5 µl buffer C . The ATPase assay was performed as described ( Speck and Stillman , 2007; Fernandez-Cid et al . , 2013 ) . Error bars represent the standard deviation from at least three independent experiments . The concentration of pre-RC intermediates and loaded dhMCM was low for cryo-EM . To get more particles in each image , we added a thin continuous carbon on top of the commercially available lacey carbon grids ( SPI #3830C-MB ) . We coated a very thin carbon layer on freshly cleaved mica using an Edwards Auto 306 evaporator , floated the thin carbon film off the mica surface onto water surface , then lowered the water level to deposit the carbon film on lacey grids . We glow-discharged the dried grids in a PELCO easiGlow before applying 3 µl sample onto the surface . After blotting for 5 s , we plunged the grid into liquefied ethane to get vitrified sample in the FEI Vitrobot , keeping the sample chamber temperature at 11°C , relative humidity 90% , and offset position of −1 mm . For negative stain EM grids preparation , we first glow-discharged the carbon-coated EM grids in PELCO easiGlow , applied 3 µl sample , blotted the grids with a piece of filter paper , applied one drop of 1% uranyl acetate solution , waited about 30 s , blotted and applied another drop of stain solution , waited about 1 min , then blotted to nearly but not completely dry , and left a thin layer of stain solution on the grid for air drying . JEM-2010F TEM with a Gatan 626 cryo-holder was used for both negative stain and cryo EM grids observation . Micrographs were recorded with an electron dose of 15 e−/Å2 at a magnification of 50 , 000 in a 4k × 4k Gatan Ultrascan CCD camera . EMAN2 . 1 package ( Tang et al . , 2007 ) was used for image processing . We first manually picked raw particle images with e2boxer , then did contrast transfer function correction following EMAN2 document . The phase flipped particles were pooled into one image stack , shrunk by a factor of 2 , and computationally classified and averaged . Each image class had at least 10 particles for negative-stained samples and 20 for cryo samples . The defocus range was −0 . 5 to −4 μm . The image pixel size was 4 . 23 Å . | Before a cell divides , it first creates copies of its DNA so that the two daughter cells both receive a complete copy of its genetic blueprint . The DNA is arranged in a double helix that is made of two single DNA strands that twist together . The process of copying the DNA requires a group or ‘complex’ of proteins called the MCM helicase complex that binds to this double-stranded DNA molecule . MCM then separates the two DNA strands to allow the production of new DNA strands in a process that uses the original strands as templates . After copying , the two resulting DNA double helices each have one of the original strands and one new strand . An enzyme called Cdc6 works together with several other proteins to help MCM bind to double-stranded DNA . Cdc6 uses energy to promote DNA copying , but it is not clear how this works . Here , Chang et al . studied the activity of yeast Cdc6 . A mutant form of Cdc6 that lacked its enzyme activity still promoted MCM binding to DNA . However , yeast cells with this mutant enzyme were unable to copy their DNA and did not divide . Next , Chang et al . used a technique called ‘single particle electron microscopy’ to investigate how the MCM complex , DNA and Cdc6 interact with each other . These experiments show that normal Cdc6 enzymes detach from the MCM complex after the energy is used to allow DNA copying and cell division to proceed . However , the mutant Cdc6 enzymes remain stuck to the complex , which blocks DNA copying . In cells , if the mutant Cdc6 enzymes are deliberately destroyed after the MCM complex binds to DNA , DNA copying proceeds normally . This implies that Cdc6 inhibits MCM activity as long it remains bound to the complex . A similar sequence of steps occurs when helicases bind to DNA in bacteria , which suggests that this important process has been maintained during billions of years of evolution . The next steps will be to understand how Cdc6 is able to inhibit the MCM complex , and how Cdc6's enzyme activity enables it to detach from the complex later on . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"short",
"report"
] | 2015 | Cdc6 ATPase activity disengages Cdc6 from the pre-replicative complex to promote DNA replication |
Retinitis pigmentosa ( RP ) is an inherited retinal disease affecting >20 million people worldwide . Loss of daylight vision typically occurs due to the dysfunction/loss of cone photoreceptors , the cell type that initiates our color and high-acuity vision . Currently , there is no effective treatment for RP , other than gene therapy for a limited number of specific disease genes . To develop a disease gene-agnostic therapy , we screened 20 genes for their ability to prolong cone photoreceptor survival in vivo . Here , we report an adeno-associated virus vector expressing Txnip , which prolongs the survival of cone photoreceptors and improves visual acuity in RP mouse models . A Txnip allele , C247S , which blocks the association of Txnip with thioredoxin , provides an even greater benefit . Additionally , the rescue effect of Txnip depends on lactate dehydrogenase b ( Ldhb ) and correlates with the presence of healthier mitochondria , suggesting that Txnip saves RP cones by enhancing their lactate catabolism .
Retinitis pigmentosa ( RP ) is one of the most prevalent types of inherited retinal diseases affecting approximately 1 in ~4 , 000 people ( Hartong et al . , 2006 ) . In RP , the rod photoreceptors , which initiate night vision , are primarily affected by the disease genes and degenerate first . The degeneration of cones , the photoreceptors that initiate daylight , color , and high-acuity vision , then follows , which greatly impacts the quality of life . Currently , one therapy that holds great promise for RP is gene therapy using adeno-associated virus ( AAV ) ( Maguire et al . , 2019 ) . This approach has proven successful for a small number of genes affecting a few disease families ( Cehajic-Kapetanovic et al . , 2020 ) . However , due to the number and functional heterogeneity of RP disease genes ( ≈100 genes that primarily affect rods , https://sph . uth . edu/retnet/ ) , gene therapy for each RP gene will be logistically and financially difficult . In addition , a considerable number of RP patients do not have an identified disease gene . A disease gene-agnostic treatment aimed at prolonging cone function/survival in the majority of RP patients could thus benefit many more patients . Given that the disease gene is typically not expressed in cones , and thus their death is due to non-autonomous mechanisms that may be in common across affected families , answers to the question of why cones die may provide an avenue to a widely applicable therapy for RP . To date , the suggested mechanisms of cone death include oxidative damage ( Komeima et al . , 2006; Wellard et al . , 2005; Xiong et al . , 2015 ) , inflammation ( Wang et al . , 2020; Wang et al . , 2019; Zhao et al . , 2015 ) , and a shortage of nutrients ( Aït-Ali et al . , 2015; Kanow et al . , 2017; Punzo et al . , 2012; Punzo et al . , 2009; Wang et al . , 2016 ) . In 2009 , we surveyed gene expression changes that occurred during retinal degeneration in four mouse models of RP ( Punzo et al . , 2009 ) . Those data led us to suggest a model wherein cones starve and die due to a shortage of glucose , which is typically used for energy and anabolic needs in photoreceptors via glycolysis . Evidence of this ‘glucose shortage hypothesis’ was subsequently provided by orthogonal approaches from other groups ( Aït-Ali et al . , 2015; Wang et al . , 2016 ) . These studies have inspired us to test 20 genes that might affect the uptake and/or utilization of glucose by cones in vivo in three mouse models of RP ( Figure 1—source data 1 ) . Only one gene , Txnip , had a beneficial effect , prolonging cone survival and visual acuity in these models . Txnip encodes an α-arrestin family member protein with multiple functions , including binding to thioredoxin ( Junn et al . , 2000; Nishiyama et al . , 1999 ) , facilitating removal of the glucose transporter 1 ( GLUT1 ) from the cell membrane ( Wu et al . , 2013 ) , and promoting the use of non-glucose fuels ( DeBalsi et al . , 2014 ) . Because α-arrestins are structurally distinct from the visual or β-arrestins , such as ARR3 , Txnip is unlikely to bind to opsins or to participate in phototransduction ( Hwang et al . , 2014; Kang et al . , 2015; Puca and Brou , 2014 ) . We tested a number of Txnip alleles and found that one allele , C247S , which blocks the association of Txnip with thioredoxin ( Patwari et al . , 2006 ) , provided the greatest benefit . Investigation of the mechanism of Txnip rescue revealed that it required lactate dehydrogenase b ( Ldhb ) , which catalyzes the conversion of lactate to pyruvate . Imaging of metabolic reporters demonstrated an enhanced intracellular ATP:ADP ratio when the retina was placed in lactate medium . Moreover , by several measures , mitochondria appeared to be healthier as a result of Txnip addition , but this improvement was not sufficient for cone rescue . The above observations led to a model wherein Txnip shifts cones from their normal reliance on glucose to enhanced utilization of lactate , as well as marked improvement in mitochondrial structure and function . Analysis of the rescue activity of several additional genes predicted to affect glycolysis provided support for this model . Finally , as our goal is to rescue cones that suffer not only from metabolic challenges , but also from inflammation and oxidative damage , we tested Txnip in combination with anti-inflammatory and anti-oxidative damage genes , and found additive benefits for cones . These treatments may benefit cones not only in RP , but also in other ocular diseases where similar environmental stresses are present , such as in age-related macular degeneration ( AMD ) .
We delivered genes that might address a glucose shortage and/or mismanagement of metabolism in a potentially glucose-limited environment . To this end , 12 AAV vectors were constructed to test genes singly or in combination for an initial screen ( Figure 1—figure supplement 1E ) . Subsequently , an additional set of AAV vectors were made based upon the initial screen results , as well as other rationales , to total 20 genes tested in all ( Figure 1—source data 1 ) . Most of these vectors carried genes to augment the utilization of glucose , such as hexokinases , phosphofructokinase , and pyruvate kinase . Each AAV vector used a cone-specific promoter , which was previously found to be non-toxic at the doses used in this study ( Xiong et al . , 2019 ) . An initial screen was carried out in rd1 mice , which harbor a null allele in the rod-specific gene , Pde6b . This strain has a rapid loss of rods , followed by cone death . The vectors were subretinally injected into the eyes of neonatal rd1 mice , in combination with a vector using the human red opsin ( RedO ) promoter , to express a histone 2B-GFP fusion protein ( AAV-RedO-H2BGFP ) . The H2BGFP provides a very bright cone-specific nuclear labeling , enabling automated quantification . As a control , eyes were injected with AAV-RedO-H2BGFP alone . Rd1 cones begin to die at ≈postnatal day 20 ( P20 ) after almost all rods have died ( Figure 1—figure supplement 1A , Figure 1—figure supplement 2A ) . The number of rd1 cones was quantified by counting the H2BGFP+ cells using a custom-made MATLAB program ( Figure 1A , Figure 1—figure supplement 1C ) . Because ~11 , 000 rd1 cones were counted in the central ½ radius of retina before their death at P20 ( Figure 1—figure supplement 1E ) , we estimated ~20% H2BGFP labeling efficiency using data for wildtype mice for comparison ( i . e . , ~50 , 000 cones within ½ radius of wildtype retina ) ( Jeon et al . , 1998 ) , with this injection dose . Only cones within the central ½ radius region of the retina were counted since RP cones in the periphery die much later ( Hartong et al . , 2006; Punzo et al . , 2009 ) . Among the vectors with individual or combinations of genes , only Txnip preserved rd1 cones at P50 ( Figure 1A , B , Figure 1—figure supplement 1E ) . The effects were likely on cone survival as it did not change the number of cones at P20 prior to their death , but did provide survival benefit by ~P30 ( Figure 1A , B , Figure 1—figure supplement 2A–C ) . The level of Txnip rescue in P50 rd1 cones was comparable to that seen using AAV with a cytomegalovirus ( CMV ) promoter to express a transcription factor , Nrf2 , that regulates anti-oxidation pathways and reduces inflammation , as we found previously ( Xiong et al . , 2015; Figure 1—figure supplement 1E ) . One combination led to a reduction in cone survival , that of Hk1 plus Pfkm ( Figure 1—figure supplement 1E ) . Our initial screen used the RedO promoter to drive Txnip expression . To evaluate a different cone-specific promoter , Txnip also was tested using a newly described cone-specific promoter , SynPVI ( Jüttner et al . , 2019 ) . This promoter also led to prolonged cone survival ( Figure 1—figure supplement 1E ) . To explore whether Txnip gene therapy is effective beyond rd1 , it was tested in rd10 mice , which carry a missense Pde6b mutation , and in Rho-/- mice , which carry a null allele in a rod-specific gene , rhodopsin . Cone survival was evaluated after the majority of central cones had died , with different ages for different strains , based upon our previous work ( Punzo et al . , 2009; Wang et al . , 2019; Xiong et al . , 2015 ) . Both rd10 and Rho-/- mice showed improved cone survival ( Figure 1A , B ) . The rescue effect did not persist long term , however , as by P240 in the Rho-/- strain it was not significant ( Figure 1—figure supplement 2D ) . To determine if Txnip-transduced mice sustained greater visual performance than control RP mice , an optomotor assay was used to measure maximal visual threshold for spatial frequency ( i . e . , visual acuity ) ( Prusky et al . , 2004 ) . Under conditions that simulated daylight , Txnip-transduced eyes showed enhanced visual acuity compared to the control contralateral eyes in rd10 and Rho-/- mice ( Figure 1C ) . The rd1 strain degenerates so quickly that it could not be evaluated in this assay . To determine if there was an improvement in overall cone phototransduction , summed across all cones , electroretinography ( ERG ) was carried out . No effect was observed in rd10 mice transduced with Txnip ( Figure 1—figure supplement 2E ) . Txnip also was evaluated for effects on cones in wildtype ( WT ) mice using peanut agglutinin ( PNA ) staining , which stains the cone-specific extracellular matrix and reflects cone health . No effect was seen on PNA staining ( Figure 1—figure supplement 1D ) . In addition , retinas from both WT and P21 rd1 mice were stained using anti-ARR3 , which stains the entire cone . At P31 , the approximate number and morphology of Txnip-transduced cones in WT retinas was similar to uninfected WT retinas ( Figure 1—figure supplement 2A ) . At P21 and P30 , immunohistochemistry ( IHC ) for ARR3 in rd1 retinas did not show an obvious rescue of cone outer segments by Txnip ( Figure 1—figure supplement 2A ) . Previous studies of Txnip provided a number of alleles that could potentially lead to a more effective cone rescue by Txnip and/or provide some insight into which of the Txnip functions are required for enhancing cone survival . A C247S mutation has been shown to block Txnip’s inhibitory interaction with thioredoxin ( Patwari et al . , 2009; Patwari et al . , 2006 ) , which is an important component of a cell's ability to fight oxidative damage via thiol groups ( Junn et al . , 2000; Nishinaka et al . , 2001; Nishiyama et al . , 1999 ) . If cone rescue by Txnip required this function , the C247S allele should be less potent for cone rescue . Alternatively , if loss of thioredoxin binding freed Txnip for its other functions and made more thioredoxin available for oxidative damage control , this allele might more effectively promote cone survival . The C247S clearly provided more robust cone rescue than WT Txnip in all three RP mouse strains ( Figure 2 , Figure 2—figure supplement 1A , B ) . These results indicate that the therapeutic effect of Txnip does not require an inhibitory interaction with thioredoxins . This finding is in keeping with previous work , which showed that anti-oxidation strategies promoted cone survival in RP mice ( Komeima et al . , 2006; Wu et al . , 2021; Xiong et al . , 2015 ) . An additional mutation , S308A , which loses an AMPK/Akt-phosphorylation site on Txnip ( Waldhart et al . , 2017; Wu et al . , 2013 ) , was tested in the context of WT Txnip and in the context of the C247S allele . The S308A change did not benefit cone survival in either context ( Figure 2 ) . In addition , the S308A allele was assayed for negative effects on cones by an assessment of rd1 cone number prior to P20 , that is , before the onset of cone death ( Figure 2—figure supplement 1C ) . It did not reduce the cone number at this early timepoint , indicating that Txnip . S308A was not toxic to cones . This finding suggests that the S308 residue is critical for the therapeutic function of Txnip through an unclear mechanism . One additional set of amino acid changes , LL351 and 352AA , was tested in the context of C247S . This allele eliminates a clathrin-binding site , and thus hampers Txnip’s ability to remove GLUT1 from cell surface through clathrin-coated pits ( Wu et al . , 2013 ) . Txnip . C247S . LL351 and 352AA could still delay RP cone death compared to the control ( Figure 2B ) , suggesting that the therapeutic effect of Txnip was unlikely to be only through the removal of GLUT1 from the cell surface . To further explore the role of GLUT1 , an shRNA to Slc2a1 , which encodes GLUT1 , was tested . It did not prolong RP cone survival ( Figure 2—figure supplement 1D ) . The slight decrease of Txnip . C247S . LL351 and 352AA in cone rescue compared to Txnip . C247S might be due to other , currently unknown effects of LL351 and 352 , or a less specific effect , for example , a protein conformational change . People with Txnip null mutations present with lactic acidosis ( Katsu-Jiménez et al . , 2019 ) , suggesting that Txnip deficiency might compromise lactate catabolism . A metabolomic study of muscle using a targeted knockout of Txnip suggested that Txnip increases the catabolism of non-glucose fuels , such as lactate , ketone bodies , and lipids ( DeBalsi et al . , 2014 ) . This switch in fuel preference was proposed to benefit the mitochondrial tricarboxylic acid cycle ( TCA cycle ) , leading to a greater production of ATP . As presented earlier , a problem for cones in the RP environment might be a shortage of glucose ( Aït-Ali et al . , 2015; Punzo et al . , 2009; Wang et al . , 2016 ) . A benefit of Txnip might then be to enable and/or force cells to switch from a preference for glucose to one or more alternative fuels . To test this hypothesis , we co-injected AAV-Txnip with shRNAs targeting the mRNAs for the rate-limiting enzymes for the catalysis of lactate , ketones , or lipids . Ldhb , encoded by the Ldhb gene , is the enzyme that converts lactate to pyruvate to potentially fuel the TCA cycle , and lactate dehydrogenase a ( Ldha , encoded by Ldha gene ) converts pyruvate to lactate ( Eventoff et al . , 1977 ) . We found that Txnip rescue was significantly decreased by any one of three Ldhb shRNAs ( siLdhb ) or by overexpression of Ldha ( Figure 3A , B , Figure 3—figure supplement 1B–E ) . We also tested the rescue effect of Txnip plus an shRNA against Oxct1 ( siOxct1 ) , a critical enzyme for ketolysis ( Zhang and Xie , 2017 ) , or against Cpt1a ( siCpt1a ) , a component for lipid transporter that is rate limiting for β-oxidation ( Shriver and Manchester , 2011 ) . These shRNAs , tested singly or in combination , did not reduce the effectiveness of Txnip rescue ( Figure 3C ) . Taken together , these data support the use of lactate , but not ketones or lipids , as a critical alternative fuel for cones when Txnip is overexpressed . If the improved survival of cones following Txnip overexpression is due to improved utilization of non-glucose fuels , cones might show improved mitochondrial metabolism . To begin to examine the metabolism of cones , we first attempted to perform metabolomics of cones with and without Txnip . However , so few cones are present in these retinas that we were unable to achieve reproducible results . An alternative assay was conducted to measure the ratio of ATP to ADP using a genetically encoded fluorescent sensor ( GEFS ) . AAV was used to deliver PercevalHR , an ATP:ADP GEFS ( Tantama et al . , 2013 ) , to rd1 cones with and without AAV-Txnip . The infected P20 rd1 retinas were explanted and imaged in three different types of media to measure the cone intracellular ratio of ATP:ADP . Txnip increased the ATP:ADP ratio ( i . e . , higher FPercevalHR488:405 ) of rd1 cones in lactate-only medium . This was also seen in pyruvate-only medium , perhaps due to improved mitochondrial health ( i . e . , greater oxidative phosphorylation [OXPHOS] activity ) . Consistent with the role of Txnip in removing GLUT1 from the plasma membrane , Txnip-transduced cones had a lower ATP:ADP ratio ( i . e . , lower FPercevalHR488:405 ) in high-glucose medium ( Figure 4A , B ) . To further probe whether intracellular glucose was reduced after overexpression of Txnip ( Wu et al . , 2013 ) , a glucose sensor iGlucoSnFR was used ( Keller et al . , 2019 ) . This sensor showed reduced intracellular glucose in Txnip-transduced cones ( Figure 4—figure supplement 1A , B ) . Because the fluorescence of GEFS may also be subject to environmental pH , we used a pH sensor , pHRed ( Tantama et al . , 2011 ) , to determine if the changes of PercevalHR and iGlucoseSnFR were due to a change in pH , and found no significant pH change ( Figure 4—figure supplement 1C , D ) . We also found that lactate , but not pyruvate , utilization by Txnip-transduced cones was critically dependent upon Ldhb for ATP production as introduction of siLdhb abrogated the increase in ATP:ADP in Txnip-transduced cones ( Figure 4C ) . Furthermore , in correlation with improved cone survival by Txnip . C247S compared to WT Txnip ( Figure 2B ) , cones had a higher ATP:ADP ratio in lactate medium when Txnip . C247S was used relative to WT Txnip ( Figure 4D , E ) . Similarly , in correlation with no survival benefit when transduced with Txnip . S308A ( Figure 2B ) , there was no difference in the ATP:ADP ratio when Txnip . S308A was used , relative to control , in lactate medium ( Figure 4D , E ) . To further probe the mechanism ( s ) of Txnip rescue , we first tested if all of the benefits of Txnip were due to Txnip's effects on Ldhb . Ldhb was thus overexpressed alone or with Txnip . Ldhb alone did not prolong cone survival , nor did it increase the Txnip rescue ( Figure 8—figure supplement 1D ) . An additional experiment was carried out to investigate if there might be a shortage of the mitochondrial pyruvate carrier , which could limit the uptake of pyruvate into the mitochondria of photoreceptors for ATP synthesis ( Grenell et al . , 2019 ) . The pyruvate carrier , which is a dimer encoded by Mpc1 and Mpc2 genes , was overexpressed , but did not prolong rd1 cone survival ( Figure 7—figure supplement 1C ) . To take a less biased approach , the transcriptomic differences between Txnip-transduced and control RP cones were characterized . H2BGFP-labeled RP cones were isolated by FACS sorting at an age when cones were beginning to die , and RNA sequencing was performed ( Figure 5—figure supplement 1A ) . Data were obtained from two RP strains , rd1 and Rho-/- . By comparing the differentially expressed genes in common between the two strains , relative to control , 7 genes were seen to be upregulated and 17 were downregulated ( Figure 5—source data 1 ) . Three of the seven upregulated genes were mitochondrial electron transport chain ( ETC ) genes . The upregulation of these three ETC genes in Txnip-transduced rd1 cones was confirmed by ddPCR ( Figure 5—figure supplement 1B ) . Similarly , we also looked for transcriptomic differences induced by Txnip in WT cones using C57BL/6J and BALB/c mice , and found only Txnip mRNA upregulation in common ( Figure 5—figure supplement 2A , Figure 5—source data 2 ) . Interestingly , there was almost no Txnip mRNA detected by RNA-seq in the WT or RP control cones , but there was high number of Txnip transcripts following addition of RedO-Txnip in all strains ( Figure 5—source data 3 ) . The finding of upregulated ETC genes in Txnip-transduced RP cones , but not in WT cones , suggested effects of Txnip on mitochondria during cone degeneration . Murakami et al . previously showed that cone mitochondria were swollen and deteriorated in rd10 mouse retinas ( Murakami et al . , 2012 ) . The morphology of Txnip-transduced mitochondria in RP cones and uninjected WT cones was examined by transmission electron microscopy ( TEM ) ( Figure 5—figure supplement 2B–D ) . There was an increase in the size of mitochondrial cross sections by Txnip transduction , with a greater increase following transduction with Txnip . C247S in P20 rd1 cones ( Figure 5A , B ) . Mitochondrial membrane potential ( ΔΨm ) , a reflection of mitochondrial ETC function , was also examined using JC-1 dye staining of freshly explanted Txnip-transduced P20 rd1 retinas ( Reers et al . , 1995 ) . Both Txnip and Txnip . C247S increased the ratio of J-aggregates:JC1-monomers ( Figure 5C , D ) , indicating an increased ΔΨm and/or a greater number/size of mitochondria with a high ΔΨm following Txnip overexpression . This finding was further investigated in vivo using infection by an AAV encoding mitoRFP , which only accumulates in mitochondria with a high ΔΨm ( Brodier et al . , 2020; Hood et al . , 2003 ) . Compared to the control cones without Txnip transduction , the intensity of mitoRFP was higher in P20 rd1 cones transduced with Txnip ( Figure 5—figure supplement 1C , D ) . A previous study identified 15 proteins that interact with Txnip . C247S ( Forred et al . , 2016 ) . Among these interactors was Parp1 , which can negatively affect mitochondria through deleterious effects on the mitochondrial genome ( Hocsak et al . , 2017; Szczesny et al . , 2014 ) , as well as effects on inflammation and other cellular pathways ( Fehr et al . , 2020 ) . Due to the similarities between the effects of Txnip addition and of Parp1 inhibition on mitochondria , Parp1 was tested for a potential role in Txnip-mediated rescue . Parp1 expression was first examined by IHC and found to be enriched in cone inner segments , which are packed with mitochondria ( Hoang et al . , 2002 ) , as well as in cone nuclei ( Figure 5—figure supplement 1G ) . Interestingly , when a GFP-Txnip fusion protein was expressed in cones , it also was found in these regions ( Figure 1—figure supplement 1B ) . To test for a role of Parp1 , Parp1-/- mice were bred to rd1 mice , and their cone mitochondria were examined by TEM and mitoRFP . Parp1-/- rd1 cones possessed larger mitochondria ( Figure 5—figure supplement 1H , I ) and higher mitoRFP signals than cones from Parp1+/+ rd1 controls ( Figure 5E , F ) . Addition of Txnip . C247S to Parp1-/- rd1 cones did not alter the mitoRFP signals ( Figure 5E , F ) . When Txnip . C247S was added to Parp1-/- rd1 retinas , cone survival was similar to that of Txnip . C247S-transduced Parp1+/+ rd1 retinas , showing that Txnip-mediated survival does not require Parp1 ( Figure 5G , H ) . Parp1-/- rd1 cone survival also was similar to the Parp1+/+ rd1 cones ( Figure 5G , H ) . The rd1 cone degeneration seemed to be faster in these Parp1+/+ and Parp1-/- mice ( on 129S background ) for unknown reason ( s ) . The discordance between improved mitochondria and cone survival in these experiments suggested that mitochondrial improvement alone is not sufficient to prolong cone survival . This is consistent with the observations from transduction with Txnip . S308A , as well as Txnip + siLdhb , both of which failed to prolong rd1 cone survival despite improvements in mitochondria ( Figure 2A , B , Figure 5A–D , Figure 5—figure supplement 1C–F ) . To test if improved cone survival requires enhanced lactate catabolism in addition to mitochondrial improvement , we delivered Ldhb to Parp1-/- rd1 cones . Unlike on Parp1+/+ background ( Figure 8—figure supplement 1D ) , a small but significant improvement in cone survival was observed ( Figure 5I , J ) . The results above suggest that Txnip may prolong RP cone survival by enhancing lactate catabolism via Ldhb , which may lead to greater ATP production by the OXPHOS pathway . Cone photoreceptors are known to require high levels of ATP to maintain their membrane potential , relying primarily upon the Na+/K+ ATPase pump ( Ingram et al . , 2020 ) . To investigate whether Txnip affects the function of the Na+/K+ pump in RP cones , freshly explanted P20 rd1 retinas were stained with RH421 , a fluorescent small-molecule probe for Na+/K+ pump function ( Fedosova et al . , 1995 ) . Addition of Txnip improved Na+/K+ pump function of these cones in lactate medium as reflected by an increase in RH421 fluorescence ( Figure 6A , B ) , consistent with Txnip enabling greater utilization of lactate as fuel . In RP cones , it is also known that protein expression of cone opsin is downregulated , postulated to be due to insufficient energy supply ( Punzo et al . , 2009 ) . Compared to control , greater anti-opsin staining was observed in Txnip-transduced rd1 cones at P50 ( Figure 6C ) , further supporting the idea that Txnip improves the energy supply to RP cones . If improved lactate catabolism and OXPHOS are at least part of the mechanism of Txnip rescue , RP cone survival might be promoted by other molecules serving similar functions . HIF1α can upregulate the transcription of glycolytic genes ( Majmundar et al . , 2010 ) . Increased glycolytic enzyme levels might push RP cones to further rely on glucose , rather than lactate , to their detriment if glucose is limited . To investigate whether HIF1α might play a role in cone survival , a WT and a dominant-negative HIF1α ( dnHIF1α ) allele ( Jiang et al . , 1996 ) were delivered to rd1 retinas using AAV . A target gene of HIF1α , Vegfa , which might improve blood flow and thus nutrient delivery ( but might also increase inflammation ) , also was tested . The dnHIF1α increased rd1 cone survival , while WT Hif1a and Vegf each decreased cone survival ( Figure 7 , Figure 7—figure supplement 1D , E ) . To determine if retention of glucose by the RPE might underlie a glucose shortage for cones ( Kanow et al . , 2017; Wang et al . , 2016 ) , we attempted to reprogram RPE metabolism to a more ‘OXPHOS’ and less ‘glycolytic’ status by overexpressing Txnip or dnHIF1α with an RPE-specific promoter , the Best1 promoter ( Esumi et al . , 2009 ) . The goal was to increase lactate consumption in the RPE , thus freeing up more glucose for delivery to cones . However , no RP cone rescue was observed ( Figure 7—figure supplement 1B ) , possibly due to a clearance of GLUT1 by Txnip from the surface of RPE cells , which would create a glucose shortage for both the RPE and the cones ( Swarup et al . , 2019; Figure 7—figure supplement 1A ) . To examine the level of GLUT1 in the RPE following introduction of WT Txnip , or Txnip . C247S . LL351 and 352AA , which should prevent efficient removal of GLUT1 ( see background in previous section ) , IHC for GLUT1 was carried out . This assay showed that AAV-Best1-Txnip . LL351 and 352AA did result in less clearance of GLUT1 from the surface of the RPE relative to WT Txnip ( Figure 7—figure supplement 1A ) . Best1-Txnip . C247S . LL351 and 352AA was then tested for rd1 cone rescue , where it was found to improve cone survival ( Figure 7 ) , in keeping with the model that the RPE retains glucose to the detriment of cones in RP . Finally , as our goal is to provide effective , generic gene therapy for RP , and potentially other diseases that affect photoreceptor survival , we used combinations of AAVs that encode genes that we have previously shown prolong RP cone survival and vision . The combination of Txnip . C247S expression in cones , with expression of Nrf2 , a gene with anti-oxidative damage and anti-inflammatory activity , in the RPE , provided an additive effect on cone survival relative to either gene alone ( Figure 8A , B ) . This combination also preserved a structure resembling cone outer segments . In WT cones , opsin protein is localized to the outer segment , where photon detection and phototransduction take place . During degeneration in RP , the cone outer segments collapse , and opsin is mislocalized to the plasma membrane ( Figure 8C , Figure 8—figure supplement 1A ) . The combination of RedO-Txnip . C247S and Best1-Nrf2 led to the localization of opsin protein to the outer segment-like structure , rather than to the plasma membrane . An additional morphological phenotype that is especially prominent in the FVB rd1 strain is that of ‘craters’ in the photoreceptor layer . These are circumscribed areas without cones that are obvious when the retina is viewed as a flat-mount . AAV-Best1-Nrf2 alone suppressed the formation of these craters ( Figure 8A; Wu et al . , 2021 ) , while AAV-RedO-Txnip did not , despite the fact that AAV-RedO-Txnip . C247S provided the most robust RP cone rescue ( Figure 2A , Figure 8A , D ) . It was also noted that dnHIF1α decreased , and Best1-Txnip . C247S . LL351 and 352AA increased , the FVB-specific retinal craters , while both vectors prolonged RP cone survival ( Figure 7A ) . The significance of these craters is thus unclear , as is the mechanism of their formation , though these data point to an origin within the RPE . An additional combination that was tested was AAV-RedO-Txnip . C247S with AAV-RedO-Tgfb1 , an anti-inflammatory gene that our previous studies showed could eliminate the craters on its own ( Wang et al . , 2020 ) . This combination did not improve cone survival beyond that of Txnip alone , but almost completely eliminated the craters in an additive fashion with Txnip ( Figure 8D , E ) . In addition , other genes ( Hk2 , Ldhb , and Cx3cl1 ) and Nrf2 were expressed specifically in cones in combination with WT Txnip , but did not provide any obvious improvement over Txnip alone ( Figure 8—figure supplement 1 ) .
Photoreceptors have been characterized as being highly glycolytic , even under aerobic conditions , as originally described by Warburg , 1925 . Glucose appears to be supplied primarily from the circulation , via the RPE , which has a high level of GLUT1 ( Gospe et al . , 2010 ) . Photoreceptors , at least rods , carry out glycolysis to support anabolism , to replace their outer segments ( Chinchore et al . , 2017 ) , and contribute ATP , to run their ion pumps ( Okawa et al . , 2008 ) . If glucose becomes limited , as has been proposed to occur in RP , cones may have insufficient fuel for their needs . To explore whether we could develop a therapy to address some of these metabolic shortcomings in RP , we delivered many different types of genes that might alter metabolic programming . From these , Txnip had the strongest improvement on cone survival and vision ( Figure 1 , Figure 1—figure supplements 1 and 2 ) . This was surprising as Txnip has been shown to inhibit glucose uptake by binding to and aiding in the removal of GLUT1 from plasma membrane ( Wu et al . , 2013 ) . Moreover , it inhibits the anti-oxidation proteins , the thioredoxins , again by direct binding ( Junn et al . , 2000; Nishinaka et al . , 2001; Nishiyama et al . , 1999 ) , so would have been predicted to increase oxidative damage and therefore decrease cone survival . The results with Txnip in its WT form , and from the study of several mutant alleles , provide some insight into how it might benefit cones . The Txnip . C247S allele prevents binding to thioredoxins and gave enhanced cone survival relative to WT Txnip ( Figure 2 , Figure 2—figure supplement 1 ) . We speculate that , by being free of this interaction , the C247S mutant protein may be more available for other Txnip-mediated activities . In addition , thioredoxin may be made more available for its role in fighting oxidative damage following transduction of the C247S mutant , rather than the WT Txnip allele . The mechanisms by which Txnip might benefit cones are not fully known , but a study of Txnip’s function in skeletal muscle suggested that it plays a role in fuel selection ( DeBalsi et al . , 2014 ) . If glucose is limited in RP , then cones may need to switch from a reliance on glucose and glycolysis to an alternative fuel ( s ) , such as ketones , fatty acids , amino acids , or lactate . Cones express oOxct1 mRNA ( Shekhar et al . , 2016 ) , which encodes a critical enzyme for ketone catabolism , suggesting that cones are capable of ketolysis . In addition , a previous study showed that lipids might be an alternative energy source for cones by β-oxidation ( Joyal et al . , 2016 ) . It is likely that cones can use these alternative fuels to meet their intense energy demands ( Ingram et al . , 2020; Figure 6 ) . However , the Txnip rescue did not depend on ketolysis or β-oxidation ( Figure 3 ) . Due to the diversity of amino acid catabolic pathways , we did not study whether these pathways were required for Txnip’s rescue effect . However , we did discover that Ldhb , which converts lactate to pyruvate , was required . This is an interesting switch as photoreceptors normally have high levels of Ldha ( Figure 5—source data 4 ) and produce lactate ( Chinchore et al . , 2017; Kanow et al . , 2017 ) . An important factor in the reliance on Ldhb could be the availability of lactate , which is highly available from serum ( Hui et al . , 2017 ) . Lactate could be transported via the RPE and/or Müller glia , and/or the internal retinal vasculature that comes in closer proximity to cones after rod death . Ketones are usually only available during fasting , and lipids are hydrophobic molecules that are slow to be transported across the plasma membranes . Moreover , lipids are required to rebuild the membrane-rich outer segments , and thus might be somewhat limited . Ldhb is not sufficient , however , to delay RP cone degeneration as its overexpression did not promote RP cone survival . Txnip-transduced RP cones also had larger mitochondria with a greater membrane potential and likely were able to use the pyruvate produced by Ldhb for greater ATP production via OXPHOS . Indeed , Txnip-transduced cones had an enhanced ATP:ADP ratio ( Figure 4 ) . However , healthier mitochondria were not sufficient to prolong RP cone survival . Txnip . S308A led to larger mitochondria than control mitochondria as seen by TEM , as well as brighter JC-1 staining and mitoRFP signals , which are indicators of better mitochondrial health , but this allele did not induce greater cone survival ( Figure 5 , Figure 5—figure supplement 1 ) . Moreover , as Txnip has been shown to interact with Parp1 , which can negatively affect mitochondria , we investigated if Parp1 knockout mice might have cones that survive longer in RP . Indeed , the Parp1 knockout mitochondria appeared to be healthier , but Parp1 knockout retinas did not have better RP cone survival than Parp1-WT rd1 retinas ( Figure 5 , Figure 5—figure supplement 1 ) . In addition , cone rescue by Txnip was not changed in the Parp1 knockout retinas ( Figure 5 ) . The well-described effects of Txnip on the removal of GLUT1 from the cell membrane might seem at odds with the promotion of cone survival . It could be that removal of GLUT1 from the plasma membrane of cones forces the cones to choose an alternative fuel , such as lactate , and perhaps others too . Interestingly , as GLUT1 knockdown was not sufficient for cone survival , Txnip must not only lead to a reduction in membrane-localized GLUT1 , but also potentiate a fuel switch , via an unknown mechanism ( s ) that at least involves an increase of Ldhb activity . A reduction in glycolysis might also lead to a fuel switch . Introduction of dnHIF1α , which should reduce expression of glycolytic enzymes , also benefited cones , while introduction of WT HIF1α did not ( Figure 7 ) . HIF1α has many target genes and may alter pathways in addition to that of glycolysis , thus also potentiating a fuel switch once glycolysis is downregulated . An additional finding supporting the notion that the level of glycolysis is important for cone survival was the observation that AAV-Pfkm plus AAV-Hk1 led to a reduction in cone survival ( Figure 1—figure supplement 1E ) . Phosphorylation of glucose by Hk1 followed by phosphorylation of fructose-6-phosphate by the Pfkm complex commits glucose to glycolysis at the cost of ATP . These AAVs may have promoted the flux of glucose through glycolysis , which may have inhibited a fuel switch , and/or depleted the ATP pool , for example , if downstream glycolytic intermediates were used for anabolic needs so that ATP production by glycolysis did not occur , and pyruvate was not produced for conversion to acetyl CoA and ATP production by the mitochondria . The observations described above suggest that at least two different pathways are required for the promotion of cone survival by Txnip ( Figure 9 ) . One pathway requires lactate utilization via Ldhb , but as Ldhb was not sufficient , another pathway is also required . As greater mitochondrial health was observed following Txnip transduction , a second pathway may include the effects on mitochondria . This notion is supported by the observation that the addition of Ldhb to Parp1-/- rd1 cones , which have healthier mitochondria , led to improved cone survival ( Figure 5 ) . Txnip alone may be able to promote cone health by impacting both lactate catabolism and mitochondrial health . There may be additional pathways required as well . The effects of Txnip alleles expressed only in the RPE provide some support for the hypothesis that the RPE transports glucose to cones for their use , while primarily using lactate for its own needs ( Kanow et al . , 2017; Swarup et al . , 2019 ) . Lactate is normally produced at high levels by photoreceptors in the healthy retina . When rods , which are 97% of the photoreceptors , die , lactate production goes down dramatically . The RPE might then need to retain glucose for its own needs . Introduction of an allele of Txnip , C247S . LL351 and 352AA , to the RPE provided a rescue effect for cones , while introduction of the WT allele of Txnip to the RPE did not . The LL351 and 352AA mutations lead to a loss of efficiency of the removal of GLUT1 from the plasma membrane , while the C247S mutation might create a less glycolytic RPE . The combination of these mutations might then allow more glucose to flow to cones . The untreated RP cones seem to be able to use glucose at a high concentration for ATP production , at least in freshly explanted retinas ( Figure 4A ) . These findings are also consistent with the reported mechanism for cone survival promoted by RdCVF , a factor that is proposed to improve glucose uptake by RP cones , which might be important if glucose is present in low concentration due to retention by the RPE ( Aït-Ali et al . , 2015; Byrne et al . , 2015 ) . As cones face multiple challenges in the degenerating RP retina , we tested Txnip in combination with genes that we have found to promote cone survival via other mechanisms . The combination of Txnip with vectors fighting oxidative stress and inflammation ( AAV-Best1-Nrf2 ) supported greater cone survival than any of these treatments alone . These combinations utilize cell type-specific promoters , reducing the chances of side effects from global expression of these genes . Of note , the Nrf2 expression was limited to the RPE , which we recently showed leads to much improved RPE survival and morphology in RP mice ( Wu et al . , 2021 ) . Best1-Nrf2 was additive to RedO-Txnip for cone survival . This finding is in keeping with the interdependence of photoreceptors and the RPE , which is undoubtedly important not only in a healthy retina , but in disease as well .
Detailed information of all AAVs used in this study is listed in Figure 1—source data 1 , along with the authentication information . cDNAs of mouse Txnip , Hif1a , Hk2 , Ldha , Ldhb , Slc2a1 , Bsg1 , Cpt1a , Oxct1 , Mpc1 , and Mpc2 , and human Nrf2 , were purchased from GeneCopoeia ( Rockville , MD ) . Mouse Vegf164 cDNA ( Robinson and Stringer , 2001 ) was synthesized through Integrated DNA Technologies ( Coralville , IA ) . We obtained the following plasmids as gifts from various depositors through Addgene ( Watertown , MA ) : Hk1 , Pfkm , and Pkm2 ( William Hahn and David Root; #23730 , #23728 , #23757 ) , Pkm1 ( Lewis Cantley and Matthew Vander Heiden; #44241 ) , H2BGFP ( Geoff Wahl; #11680 ) , mitoRFP ( i . e . , DsRed2-mito , Michael Davidson; #55838 ) , GFP-Txnip ( Clark Distelhorst; #18758 ) , W3SL ( Bong-Kiun Kaang; #61463 ) , 3xFLAG ( Thorsten Mascher , #55180 ) , and PercevalHR and pHRed ( Gary Yellen; #49082 , #31473 ) . The cDNA of mouse RdCVF was a gift from Leah Byrne and John Flannery ( UC Berkeley , CA ) . iGlucoSnFR was provided under a Material Transfer Agreement by Jacob Keller and Loren Looger ( Janelia Research Campus , VA ) . RedO promoter was provided as a gift , and SynPVI ( also known as ProA7 ) and SynP136 ( also known as ProA1 ) promoters were provided under a Material Transfer Agreement , from Botond Roska ( IOB , Switzerland ) . The Best1 promoter was synthesized by a lab member , Wenjun Xiong , using Integrated DNA Technologies based on the literature ( Esumi et al . , 2009 ) . Mutated Txnip , dominant-negative HIF1α ( Jiang et al . , 1996 ) , and RO1 . 7 promoter ( Ye et al . , 2016 ) were created from the Hif1a and RedO plasmids correspondingly in house using Gibson assembly . Of note , we found that the RedO promoter is stronger than SynP136 or SynPVI promoters , but less specific . RedO has a low level of expression in some rods . SynP136 and SynPVI drive expression that is exclusive to cones , that is , no rod expression in keeping with the observation of Roska Lab ( Jüttner et al . , 2019 ) . SynPVI ( 0 . 5 kb ) is shorter than SynP136 ( 2 kb ) , and is thus better for packaging insert genes that have a large size . All of the new constructs in this study were cloned using Gibson assembly . For example , AAV-RedO-Txnip was cloned by replacing the EGFP sequence of AAV-RedO-EGFP at the NotI/HindIII sites , with the Txnip sequence , which was PCR-amplified from the cDNA vector adding two 20 bp overlapping sequences at the 5′- and 3′-ends . All of the AAV plasmids were amplified using Stbl3 Escherichia coli ( Thermo Fisher Scientific ) . The sequences of all AAV plasmids were verified with directed sequencing and restriction enzyme digestion . The key plasmids were verified with Next-Generation complete plasmid sequencing ( MGH CCIB DNA Core ) , which is able to capture the full sequence of the ITR regions . The genome sequence of critical AAVs ( i . e . , AAV8-RedO-Txnip . C247S and AAV8-RedO-Txnip . S308A ) was verified with PCR and directed sequencing . All of the vectors were packaged in recombinant AAV8 capsids using 293 T cells and purified with iodixanol gradient as previously described ( Grieger et al . , 2006; Xiong et al . , 2015 ) . The titer of each AAV batch was determined using protein gels , comparing virion protein band intensities with a previously established AAV standard . The concentration of our AAV production usually ranged from 2 × 1012 to 3 × 1013 gc/mL . Multiple batches of key AAV vectors ( e . g . , four batches of AAV8-RedO-Txnip and three batches of AAV-RedO-siLdhb ( #2 ) ) were made and tested in vivo to avoid any potential batch effects . The shRNA plasmids of Ldhb , Slc2a1 , Oxct1 , and Cpt1a were purchased from GeneCopoeia , provided as three or four distinct sequences for each gene , driven by the H1 or U6 promoter . The knockdown efficiency of these candidate shRNA sequences was tested by co-transfecting with CAG-TargetGene-IRES-d2GFP vector in 293 T cells as previously described ( Matsuda and Cepko , 2007; Wang et al . , 2014 ) . The GFP fluorescence intensity served as a fast and direct read out of the knockdown efficiency of these shRNAs . Using this method , we selected the following sense strand sequences to knock down the targeted genes ( Figure 2—figure supplement 2 , Figure 3—figure supplements 2–4 ) : siLdhb ( #2 ) 5′-CCATCATCGTGGTTTCCAACC-3′; siLdhb ( #1 ) 5′-GCAGAGAAATGTCAACGTGTT-3′; siLdhb ( #3 ) 5′-GCCGATAAAGATTACTCTGTG-3′; siSlc2a1 ( #a ) 5′-GGTTATTGAGGAGTTCTACAA-3′; siOxct1 ( #c ) 5′-GGAAACAGTTACTGTTCTCCC-3′; siCpt1a ( #c ) 5′-GCATAAACGCAGAGCATTCCT-3′; and siNC ( control sequence that does not target any known gene ) 5′-GCTTCGCGCCGTAGTCTTA-3′ . We cloned the entire hairpin sequence ( including a 6 bp 5′-end lead sequence 5′-gatccg-3′ , a 7 bp loop sequence 5′-TCAAGAG-3′ between sense and antisense strands , and a >7 bp 3′-end sequence 5′-ttttttg-3′ ) and packaged them into AAV8-RedO-shRNA using Gibson assembly as described above . To maximize the knockdown efficacy using a Pol II promoter in AAV ( Giering et al . , 2008 ) , no extra base pairs were retained between the RedO promoter and the 5′-end sequence of shRNAs . Due to the lack of an adequate Ldhb antibody , we confirmed the in vivo Ldhb knockdown efficiency of all three AAV8-RedO-siLdhb vectors by co-injection with an AAV8-Ldhb-3xFLAG vector into WT mouse eyes with detection using FLAG immunofluorescence as described in the Histology section ( Figure 3—figure supplement 1A ) . On the day of birth ( P0 ) , AAVs were injected into the eyes of pups as previously described ( Matsuda and Cepko , 2007; Xiong et al . , 2015 ) . For all experiments in which cones were quantified , and to provide a means to trace infection ( e . g . , for IHC ) , 2 . 5 × 108 vg/eye of AAV8-RedO-H2BGFP was co-injected with the experimental AAVs , or alone as a control . The dose of each experimental AAV was ≈1 × 109 vg/eye , individually or when in combination with other experimental AAV’s , and the injection volume was the same for each injection . For all other experiments , such as FACS sorting and ex vivo live imaging , 1 × 109 vg/eye of AAV8-SynP136-H2BGFP , which provides better cone specificity but lower expression level than RedO-H2BGFP , was co-injected . All of the control groups in this study refer to AAV reporter ( e . g . , H2BGFP or PercevalHR ) injection alone . The photopic optomotor response of mice was measured using the OptoMotry System ( CerebralMechanics ) at a background light of ≈70 cd/m2 as previously described ( Xiong et al . , 2019 ) . The contrast of the grates was set to be 100% , and temporal frequency was 1 . 5 Hz . The threshold of mouse visual acuity ( i . e . , maximal spatial frequency ) was tested by an examiner without knowledge of the control vs . experimental groups . During each test , the direction of movement of the grates ( i . e . , clockwise or counterclockwise ) was randomized , and the spatial frequency of each testing episode was determined by the software . Without knowing the spatial frequency of the moving grates , the examiner reported either ‘yes’ or ‘no’ to the system until the threshold of acuity was determined by the software . Optomotor tests were conducted on rd10 and Rho-/- mice , but not with rd1 strain , which loses vision at a very early age , before any meaningful test could be performed . To probe rd10 cone function , photopic ERGs were measured in anesthetized mouse eyes in vivo as previously described using an Espion E3 System ( Diagonsys LLC ) ( Xiong et al . , 2019; Xue et al . , 2020 ) . Multiple white flashes were applied to elicit ERG responses at 1 ( peak ) , 10 ( peak ) , 100 ( xenon ) , and 1000 ( xenon ) cd s/m2 intensities with a white light background of 30 cd/m2 . A ketamine/xylazine cocktail ( 100/10 mg/kg ) was introduced via intraperitoneal injection to mice for anesthesia . Tropicamide 1% eye solution ( Bausch + Lomb ) was used to dilate the pupil . Mice were euthanized with CO2 and cervical dislocation , and the eyes were enucleated . For flat-mounts , retinas were separated from the rest of the eye using a dissecting microscope and were fixed in 4% paraformaldehyde solution for 30 min . The retinas were then flat-mounted on a glass slide and coverslip . For H2BGFP-labeled cone imaging , we used a Keyence fluorescence microscope with a ×10 objective ( Plan Apo Lamda 10x/0 . 45 Air DIC N1 ) and GFP filter box ( OP66836 ) . For cone opsin antibody staining in whole-mount retinas , after fixation , retinas were blocked for 1 hr in PBS with 5% normal donkey serum and 0 . 3% Triton X-100 at room temperature . After blocking , retinas were incubated with a mixture of 1:200 anti-s-opsin ( OPN1MW ) antibody ( AB5407 , EMD Millipore ) and 1:600 anti-m-opsin ( OPN1SW ) antibody ( AB5405 , EMD Millipore ) in the same blocking solution overnight at 4°C , followed by secondary donkey-anti-rabbit antibody staining ( 1:1000 , Alexa Fluor 594 ) at room temperature for 2 hr , then flat-mounted on a glass slide and coverslip . For frozen sections , whole eyes were fixed in 4% paraformaldehyde solution for 2 hr at room temperature , followed by removal of the cornea , lens , and iris . The eye cups then went through a 15% and 30% sucrose solution to dehydrate at room temperature , followed by overnight incubation in 1:1 30% sucrose and Tissue-Tek O . C . T . solution at 4°C . Eye cups were embedded in a plastic mold , frozen in a −80°C freezer , and cut into 20 or 12 μm thin radial cross sections that were placed on glass slides . Antibody staining was done similarly to whole-mounts as described above and previously ( Wang et al . , 2014 ) . PBS with 0 . 1% Triton X-100 , 5% normal donkey serum , and 1% bovine serum albumin ( BSA ) was used as the blocking solution , except for FLAG detection ( 10% donkey serum and 3% BSA ) . GLUT1 ( encoded by Slc2a1 gene ) antibody ( GT11-A , Alpha Diagnostics ) was used at 1:300 dilution , PARP1 antibody ( ab227244 , Abcam ) was used at 1:300 dilution , GFP antibody ( ab13970 , Abcam ) was used at 1:1000 dilution to detect GFP-Txnip , ARR3 antibody ( AB15282 , Millipore Sigma ) was used at 1:1000 dilution to detect cone arrestin ( encoded ay the Arr3 gene ) , and FLAG antibody ( ab1257 , Abcam ) was used at 1:2000 based on a previous study ( Ferrando et al . , 2015 ) . If applicable , 1:1000 PNA ( CY5 or rhodamine labeled ) for cone extracellular matrix labeling and 1:1000 DAPI were used to co-stain with secondary antibodies . Stained sections were imaged with a confocal microscope ( LSM710 , Zeiss ) using ×20 or ×63 objectives ( Plan Apo 20x/0 . 8 Air DIC II , or Plan Apo 63X/1 . 4 Oil DIC III ) . The cone-H2BGFP images of entire flat-mounted retinas were first analyzed in ImageJ to acquire the diameter and the center parameters of the sample . We used a custom MATLAB script to automatically count the number of H2BGFP-positive cones in the central ½ radius of the retina since RP cones degenerate faster in the central than the peripheral retina . The algorithm was based on a Gaussian model to identify the centers of labeled cells ( Wu et al . , 2021 ) . The threshold of peak intensity and the variance of distribution were initially determined using visual inspection , and a comparison to the number of manually counted cones from six retinas . The threshold of intensity and variance thus determined were then set at fixed values for all the experiments that used cone quantification . The background intensity did not interfere with the accurate counting on the raw images by this MATLAB script despite the observation that the intensity of some images looked different at low magnification . This method of cone counting was further verified using FACS analysis of cones in this study ( Figure 1—figure supplement 2B ) and an independent study ( Wang et al . , 2020 ) . For JC-1 mitochondrial dye staining , the retina was quickly dissected in a solution of 50% Ham's F-12 Nutrient Mix ( 11765054 , Thermo Fisher Scientific ) and 50% Dulbecco's Modified Eagle Medium ( DMEM; 11995065 , Thermo Fisher Scientific ) at room temperature . They were then incubated in a culture medium containing 50% Fluorobrite DMEM ( A1896701 , Thermo Fisher Scientific ) , 25% heat-inactivated horse serum ( 26050088 , Thermo Fisher Scientific ) , and 25% Hanks' Balanced Salt Solution ( 14065056 , Thermo Fisher Scientific ) with 2 μM JC-1 dye ( M34152 , Thermo Fisher Scientific ) at 37°C in a 5% CO2 incubator for 20 min . The retinas were washed in 37°C in this medium without JC-1 three times , transferred to a glass-bottom culture dish ( MatTek P50G-1 . 5-30F ) with culture medium , and imaged using a confocal microscope ( LSM710 Zeiss ) , which was equipped with a chamber pre-heated to 37°C with pre-filled 5% CO2 . Right before imaging , a cover slip ( VWR 89015-725 ) was gently applied to flatten the retina . Regions of interest ( with H2BGFP as an indicator of successful AAV infection and to set the correct focal plane on the cone layer ) were selected under the eyepiece with a ×63 objective ( Plan Apo 63X/1 . 4 Oil DIC III ) . Fluorescent images from the same region of interest were obtained with the excitation wavelength of 561 nm ( for J-aggregates ) , 514 nm ( for JC-1 monomer ) , and 488 nm ( for H2BGFP ) . Four different regions of interest from the central part of the same retina were imaged before moving to the next retina . For RH421 ( Na+/K+ ATPase dye ) staining , similar steps were taken as for JC-1 staining , with the following modifications: ( 1 ) 0 . 83 μM RH421 dye ( 61017 , Biotium ) was added to the glass-bottom culture dishes just before imaging , but not during incubation in the incubator , due to the fast action of RH421 . ( 2 ) Five regions of interest were imaged per retina from the central area . ( 3 ) The dissection and culture medium were lactate-only medium ( see below ) . ( 4 ) Excitation wavelengths: 561 nm ( RH421 ) and 488 nm ( H2BGFP ) . For imaging genetically encoded metabolic sensors ( PercevalHR , iGlucoSnFR , and pHRed ) , retinas were placed in the incubator for 12 min and then taken to confocal imaging without any staining . For the high-glucose condition , the culture medium described above contains ≈15 mM glucose without lactate or pyruvate . For the lactate-only condition , the culture and dissection media were both glucose-pyruvate-free DMEM ( A144300 , Thermo Fisher Scientific ) and were supplemented with 20 mM sodium L-lactate ( 71718 , Sigma-Aldrich ) . For the pyruvate-only condition , the culture and dissection media were both glucose-pyruvate-free DMEM plus 10 or 20 mM sodium pyruvate ( P2256 , Sigma-Aldrich ) . No AAV-H2BGFP was co-injected with these sensors since the sensors themselves could be used to trace the area of infection . The excitation wavelengths for sensors were 488 and 405 nm ( PercevalHR , ratiometric high and low ATP:ADP ) , 488 and 561 nm ( iGlucoSnFR , glucose-sensing GFP and normalization mRuby ) , and 561 and 458 nm ( pHRed , ratiometric low and high pH ) . The fluorescent intensity of all acquired images was measured by ImageJ . The ratio of sensors/dye was normalized to averaged control results taken at the same condition . All flow cytometry and cell sorting were performed on MoFlo Astrios EQ equipment . Retinas were freshly dissected and dissociated using cysteine-activated papain followed by gentle pipetting ( Shekhar et al . , 2016 ) . Before sorting , all samples were passed through a 35 μm filter with buffer containing Fluorobrite DMEM ( A1896701 , Thermo Fisher Scientific ) and 0 . 4% BSA . Cones labeled with AAV8-SynP136-H2BGFP ( highly cone-specific ) were sorted into the appropriate buffer for either ddPCR or RNA sequencing . RNA sequencing was done as previously described ( Wang et al . , 2019 ) . 1000 H2BGFP-positive cones per retina were sorted into 10 μL of Buffer TCL ( QIAGEN ) containing 1% β-mercaptoethanol and immediately frozen in −80°C . On the day of sample submission , the frozen cone lysates were thawed on ice and loaded into a 96-well plate for cDNA library synthesis and sequencing . A modified Smart-Seq2 protocol was performed on samples by the Broad Institute Genomics Platform with ∼6 million reads per sample ( Picelli et al . , 2013 ) . The reads were mapped to the GRCm38 . p6 reference genome after quality control measures . Reads assigned to each gene were quantified using featureCounts ( Liao et al . , 2014 ) . Count data were analyzed using DESeq2 to identify differentially expressed genes , with an adjusted p value less than 0 . 05 considered significant ( Anders and Huber , 2010 ) . The raw results have been deposited to Gene Expression Omnibus ( accession number GSE161622 for RP cones and GSE168503 for WT cones ) . RNA was isolated from 20 , 000 sorted cones per retina using the RNeasy Micro Kit ( QIAGEN ) as previously described ( Wang et al . , 2020 ) and converted to cDNA using the SuperScript III First-Strand Synthesis System ( Invitrogen ) . cDNA from each sample was packaged in droplets for Droplet Digital PCR ( ddPCR ) using QX200 EvaGreen Supermix ( #1864034 ) . The reads of expression were normalized to the housekeeping gene Hprt . Sequences for RT-PCR primers were designed using the IDT online RealTime qPCR primer design tool . The following primers were selected for the genes of interest: Txnip ( forward 5′-ACATTATCTCAGGGACTTGCG-3′; reverse 5′-AAGGATGACTTTCTTGGAGCC-3′ ) , Hprt ( forward 5′-TCAGTCAACGGGGGACATAAA-3′; reverse 5′-GGGGCTGTACTGCTTAACCAG-3′ ) , mt-Nd4 ( forward 5′-AGCTCAATCTGCTTACGCCA-3′; reverse 5′-TGTGAGGCCATGTGCGATTA-3′ ) , mt-Cytb ( forward 5′-ATTCTACGCTCAATCCCCAAT-3′; reverse 5′-TATGAGATGGAGGCTAGTTGGC-3′ ) , mt-Co1 ( forward 5′-TCTGTTCTGATTCTTTGGGCACC-3′; reverse 5′-CTACTGTGAATATGTGGTGGGCT-3′ ) , Acsl3 ( forward 5′- AACCACGTATCTTCAACACCATC-3′; reverse 5′- AGTCCGGTTTGGAACTGACAG-3′ ) , and Ftl1 ( forward 5′- CCATCTGACCAACCTCCGC-3′; reverse 5′- CGCTCAAAGAGATACTCGCC-3′ ) . Intracardial perfusion ( 4% PFA + 1% glutaraldehyde ) was performed on ketamine/xylazine ( 100/10 mg/kg ) anesthetized mice before the removal of eyes . The cornea was sliced open and the eye was fixed with a fixative buffer ( 1 . 25% formaldehyde + 2 . 5% glutaraldehyde + 0 . 03% picric acid in 0 . 1 M sodium cacodylate buffer , pH 7 . 4 ) overnight at 4°C . The cornea , lens , and retina were removed before resin embedding , ultrathin sectioning , and negative staining at Harvard Medical School Electron Microscopy Core . The detailed methods can be found on the core’s website ( https://electron-microscopy . hms . harvard . edu/methods ) . The stained thin sections were imaged on a conventional TEM ( JEOL 1200EX ) with an AMT 2k CCD camera . For the comparison of two sample groups , two-tailed unpaired Student’s t-test was used to test for the significance of difference , except for P140 Rho-/- optomotor assay ( paired two-tail t-test ) . For comparison of more than two sample groups , ANOVA and Dunnett's multiple comparison test were performed in Prism 8 software to determine the significance . A p-value of <0 . 05 was considered statistically significant . All error bars are presented as mean ± standard deviation , except for the rd10 optomotor assays , ERG , FACS cell % , and RNA-seq raw reads ( mean ± SEM ) . All animal experiments were approved by the IACUC of Harvard University in accordance with institutional guidelines . | Retinitis pigmentosa is an inherited eye disease affecting around one in every 4 , 000 people . It results from genetic defects in light sensitive cells of the retina , called photoreceptor cells , which line the back of the eye . Though vision loss can occur from birth , retinitis pigmentosa usually involves a gradual loss of vision , sometimes leading to blindness . Rod photoreceptors , which are responsible for vision in low light , are impacted first . The disease then affects cone photoreceptors , the cells that detect light during the day , providing both color and sharp vision . Around 100 mutated genes associated with retinitis pigmentosa have been identified , but only a handful of families with one of these mutant genes have been treated with a gene therapy specific for their mutated gene . There are currently no therapies available to treat the vast number of people with this disease . The mutations that cause retinitis pigmentosa directly affect the rod cells that detect dim light , leading to loss of night vision . There is also an indirect effect that causes cone photoreceptors to stop working and die . One theory to explain this two-step disease process relates to the fact that cone photoreceptors are very active cells , requiring a high level of energy , nutrients and oxygen . If surrounding rod cells die , cone photoreceptors may be deprived of some essential supplies , leading to cone cell death and daylight vision loss . To examine this theory , Xue et al . tested a new gene therapy designed to alleviate the potential shortfall in nutrients . The experiments used three different strains of mice that had the same genetic mutations as humans with retinitis pigmentosa . The gene therapy used a virus , called adeno-associated virus ( AAV ) , to deliver 20 different genes to cone cells . Each of the 20 genes tested plays a different role in cells’ processing of nutrients to provide energy . After administering the treatment , Xue et al . monitored the mice to see whether or not their vision was affected , and how cone cells responded . Only one of the 20 genes , Txnip , delivered using gene therapy , had a beneficial effect , prolonging cone cell survival in all three mouse strains . The mice that received Txnip also retained their ability to discern moving stripes on vision tests . Further investigations demonstrated that activating Txnip forced the cones to start using a molecule called lactate as an energy source , which could be more available to them than glucose , their usual fuel . These cells also had healthier mitochondria – the compartments inside cells that produce and manage energy supplies . This dual effect on fuel use and mitochondrial health is thought to be the basis for the extended cone survival and function . These experiments by Xue et al . have identified a good gene therapy candidate for treating retinitis pigmentosa independently of which genes are causing the disease . Further research will be required to test the safety of the gene therapy , and whether its beneficial effects translate to humans with retinitis pigmentosa , and potentially other diseases with unhealthy photoreceptors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"neuroscience"
] | 2021 | AAV-Txnip prolongs cone survival and vision in mouse models of retinitis pigmentosa |
The role of specific phospholipids ( PLs ) in lipid transport has been difficult to assess due to an inability to selectively manipulate membrane composition in vivo . Here we show that the phospholipid remodeling enzyme lysophosphatidylcholine acyltransferase 3 ( Lpcat3 ) is a critical determinant of triglyceride ( TG ) secretion due to its unique ability to catalyze the incorporation of arachidonate into membranes . Mice lacking Lpcat3 in the intestine fail to thrive during weaning and exhibit enterocyte lipid accumulation and reduced plasma TGs . Mice lacking Lpcat3 in the liver show reduced plasma TGs , hepatosteatosis , and secrete lipid-poor very low-density lipoprotein ( VLDL ) lacking arachidonoyl PLs . Mechanistic studies indicate that Lpcat3 activity impacts membrane lipid mobility in living cells , suggesting a biophysical basis for the requirement of arachidonoyl PLs in lipidating lipoprotein particles . These data identify Lpcat3 as a key factor in lipoprotein production and illustrate how manipulation of membrane composition can be used as a regulatory mechanism to control metabolic pathways .
Phospholipids ( PLs ) are important components of biological membranes and also serve as precursors for the generation of diverse signaling molecules ( Spector and Yorek , 1985 ) . In mammalian cells PLs synthesized de novo undergo further remodeling through deacylation by phospholipases and the subsequent and reacylation by lysophospholipid acyltransferases ( Lpcats ) . Membrane PLs ultimately reach an equilibrium in which the majority of PL species contain a saturated acyl chain at the sn-1 position and an unsaturated chain at the sn-2 position . The Lpcat-dependent remodeling process is essential for the diversity and asymmetric distribution of acyl chains because the de novo PL synthesis pathway has little substrate specificity ( Yamashita et al . , 2014 ) . We previously identified the sterol-activated nuclear receptor LXR as an integrator of cellular lipid levels and membrane PL composition . LXR controls the expression of Lpcat3 , which is the most abundant Lpcat family member in liver and intestine . Cell-based assays suggest that Lpcat3 preferentially catalyzes the synthesis of phosphatidylcholine ( PC ) species containing an unsaturated fatty acyl chain at the sn-2 position ( Hishikawa et al . , 2008; Zhao et al . , 2008 ) , but the importance of Lpcat3 activity for membrane PL composition in vivo remains to be established . We found that acute overexpression or knockdown of Lpcat3 in cultured cells or mouse liver altered the distribution of PL species , particularly those containing unsaturated fatty acyl chains ( Rong et al . , 2013 ) . Moreover , we showed that the ability of the LXR-Lpcat3 pathway to promote unsaturation of membrane lipids was protective against ER stress and inflammation in the setting of cellular lipid excess . However , the in vivo relevance of LXR-dependent modulation of endogenous PL composition for systemic lipid homeostasis was unclear . Moreover , the larger topic of the regulatory potential of dynamic phospholipid remodeling has been largely unexplored . It has been reported that modifications of PL composition influence a range of cellular processes ( Holzer et al . , 2011; Pinot et al . , 2014 ) . However , most analyses of the consequences of altered PL fatty acyl composition have been performed in purified membrane systems or have involved treating cells with high levels of exogenous lipids . Such experimental manipulations are unlikely to accurately model physiologic perturbations in membrane composition . The recognition that Lpcat3 activity can be regulated by cellular lipid status through LXRs raises the possibility that Lpcat3 activity could contribute to some of the well-documented effects of LXR on systemic lipid homeostasis . The rate of lipoprotein production has been linked with the availability of PC , but the mechanisms underlying this link are not clear ( Vance , 2008; Abumrad and Davidson , 2012 ) . PC is the major PL component of lipoproteins ( Ågren et al . , 2005 ) . It has been reported that active PC de novo synthesis is necessary for the biogenesis and/or secretion of very low-density lipoprotein ( VLDL ) from hepatocytes ( Vance , 2008 ) . However , impaired de novo PC synthesis impacts all the PC species and reduces the total PC content of cellular membranes . Whether a specific PC species is selectively required for hepatocyte lipoprotein production is unclear . In the intestine , it has been proposed that PC may enhance lipid uptake by enterocytes and/or promote chylomicron assembly and secretion ( Tso et al . , 1977 , 1978; Voshol et al . , 2000 ) . Luminal PC is not absorbed intact , but is hydrolyzed into lysophosphatidylcholine ( Lyso-PC ) in the lumen and subsequently re-acylated within enterocytes by Lpcat enzymes ( Nilsson , 1968; Parthasarathy et al . , 1974 ) . The mechanistic role of specific PL species in lipid transport and lipoprotein production has been difficult to address due to an inability to selectively manipulate membrane PL composition in vivo . Interestingly , the LXR pathway has been reported to promote hepatic triglyceride ( TG ) secretion by the liver ( Okazaki et al . , 2010 ) . The possibility that PL remodeling contributes to this effect has not been tested . We demonstrate here that Lpcat3 is uniquely required for the incorporation of arachidonic acid into membranes in vivo , and that an absence of arachidonoyl PLs profoundly affects lipid transport and lipoprotein production . Biophysical , electron microscopy ( EM ) and biochemical studies indicate that Lpcat3-dependent production of arachidonoyl PLs is important for lipid movement within membranes and for the efficient lipidation of apoB–containing lipoproteins . We also show that induction of Lpcat3 activity is required for the ability of LXRs to promote hepatic VLDL production . These data identify Lpcat3-dependent phospholipid remodeling as a critical , LXR-regulated step in TG secretion , and suggest that this step might be further explored as a strategy to treat hyperlipidemias .
To examine the consequence of Lpcat3 deficiency in vivo , we generated Lpcat3-deficient mice from targeted ES cells ( Figure 1A ) . The targeted allele was ‘conditional-ready’ , making it possible to create both global and tissue-specific knockout mice . The global knockout mice ( i . e . , homozygous for the targeted allele ) showed markedly reduced levels of Lpcat3 transcripts in liver and intestine ( Figure 1B , C ) . Global Lpcat3−/− mice on a C57BL/6 background were born at the expected Mendelian frequency , and their weights were indistinguishable from WT mice at birth ( Figure 1D , Table 1 ) . However , the blood glucose levels of Lpcat3−/− mice were very low at birth , and none survived beyond day 2 ( Figure 1D , E ) . Lpcat3−/− pups survived for up to 6 days when given subcutaneous glucose injections , but the pups did not thrive and invariably died ( Figure 1E , F ) . Analysis of gene expression in liver and small intestines of the pups revealed changes in a number of genes linked to lipid metabolism , some of which were common in both tissues ( Figure 1G , H ) . Although these changes appeared consistent with a role for Lpcat3 in lipid metabolism , it was impossible to exclude the possibility that these gene-expression alterations were simply due to the extremely poor health of the mice . We therefore turned our attention to tissue-selective knockout mice , with the hope that we could obtain viable mice and decipher the function of Lpcat3 . 10 . 7554/eLife . 06557 . 003Figure 1 . Generation and analysis of global Lpcat3 knockout mice . ( A ) Strategy for generating global Lpcat3 knockout mice . A ‘knock-out first/conditional-ready’ gene-targeting vector was used to generate targeted cells . A gene-trap cassette is located between the two FRT sites . LacZ , β-galactosidase; neo , neomycin phosphotransferase II . ( B ) Genotyping of Lpcat3+/+ ( WT ) , Lpcat3+/− ( Het ) , and Lpcat3−/− ( KO ) mice . Genomic DNA was prepared from tail biopsies , and PCR products were separated on a 1% agarose gel . ( C ) Expression of Lpcat3 in liver and small intestine of newborn Lpcat3−/− and Lpcat3+/+ pups . Gene expression was quantified by real-time PCR ( n ≥ 5/group ) . Values are means ± SEM . ( D ) The body weight and blood glucose of Lpcat3+/+ ( WT ) , Lpcat3−/+ ( Het ) , and Lpcat3−/− ( KO ) newborn pups ( n ≥ 8/group ) . Values are means ± SEM . ( E ) Kaplan–Meier survival curve of Lpcat3+/+ ( WT ) , Lpcat3−/+ ( Het ) and Lpcat3−/− ( KO ) pups after birth ( n ≥ 20 mice/group ) . The neonatal lethality can be delayed by injection of 50 μl 10% glucose solution once per day after born . 5 Lpcat3+/+ ( WT ) , 9 Lpcat3−/+ ( Het ) and 6 Lpcat3−/− ( KO ) mice were used in the rescue experiment . ( F ) Representative photograph of Lpcat3+/+ ( WT ) and Lpcat3−/− ( KO ) pups after 5 days of glucose injections . ( G–H ) Gene expression in livers ( G ) and small intestines ( H ) of Lpcat3+/+ ( WT ) and Lpcat3−/− ( Lpcat3 KO ) newborn pups . Gene expression was quantified by real-time PCR ( n ≥ 5/group ) . Values are means ± SEM . Statistical analysis was performed using Student's t-test ( C , G and H ) and one-way ANOVA with Bonferroni post-hoc tests ( D ) . *p < 0 . 05; **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 00310 . 7554/eLife . 06557 . 004Table 1 . Breeding data for global Lpcat3-deficient miceDOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 004GenotypeNumber of pups/miceObserved %Expected %TimeWT252825At birthHet434850KO212325WT503525At weaningHet916450KO0025Genotypic ratio of newborns and weanlings obtained from Lpcat3 heterozygote intercrosses . We generated a conditional knockout allele ( Lpcat3fl ) by breeding the global heterozygous knockout mice with mice expressing FLPe recombinase ( Rodriguez et al . , 2000 ) . Mice harboring the ‘floxed’ Lpcat3 allele were then crossed with albumin-Cre transgenic mice to create liver-specific Lpcat3 knockout mice ( here designated ‘L-Lpcat3 KO’; Figure 2A ) . In contrast to the global Lpcat3 knockout mice , L-Lpcat3 mice were born at the expected Mendelian frequency , survived to adulthood , and appeared ( at least by external inspection ) to be indistinguishable from control ( homozygous floxed , Cre-negative ) mice ( Table 2 and data not shown ) . Expression of Lpcat3 transcripts in whole liver from L-Lpcat3 KO mice was markedly reduced ( Figure 2B ) . The residual expression of Lpcat3 mRNA in the liver of Lpcat3 KO mice was likely due to persistent expression of Lpcat3 in cell types that do not express the albumin-Cre transgene ( Kupffer cells , endothelial cells ) . Consistent with that idea , Lpcat3 expression was reduced by >90% in primary hepatocytes from L-Lpcat3 KO mice ( Figure 2B ) . Unfortunately , we were unable to measure levels of Lpcat3 protein because specific antibodies are not currently available . We observed no compensatory upregulation of Lpcat1 or Lpcat2 in livers of L-Lpcat3 KO mice ( Figure 2B ) . Lpcat4 expression was undetectable in the liver . 10 . 7554/eLife . 06557 . 005Figure 2 . Altered triglyceride ( TG ) metabolism in liver-specific Lpcat3 knockout mice . ( A ) Strategy for generating tissue-specific Lpcat knockout mice . Lpcat3−/+ mice carrying the conditional-ready knockout allele were mated with Flpe transgenic mice to generate the Lpcat3fl/fl mice . Lpcat3fl/fl mice were bred with tissue-specific Cre transgenic mice to generate tissue-specific Lpcat3 knockout mice . ( B ) Expression of Lpcat family members in liver of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed a chow diet . Gene expression was quantified by real-time PCR ( n ≥ 6/group ) ( left panel ) . Primary hepatocytes from Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice were treated overnight with 1 μM LXR agonist GW3965 ( GW ) . Lpcat3 expression was quantified by real-time PCR ( right panel ) . Ct values in Lpcat3fl/fl liver samples were shown . Values are means ± SEM . ( C ) Plasma lipid levels in chow diet-fed Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice under fed ( ad libitum ) or overnight ( o/n ) fasting ( n ≥ 8/group ) . Plasma levels of cholesterol and NEFA were measured after an overnight fast . Values are means ± SEM . ( D ) Plasma was harvested from chow diet-fed Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice after an overnight fast . ApoB protein was analyzed by western blotting and quantified ( Figure 2—figure supplement 2B ) . ( E ) Plasma from Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3-KO ) mice fasted overnight was pooled ( n = 5 ) . Lipoprotein profiles were analyzed by fast protein liquid chromatography ( FPLC ) ( upper panel ) . ApoB protein in each corresponding fraction was analyzed by western blot ( lower panel ) and quantified ( Figure 2—figure supplement 2A ) . ( F ) Lipid contents in livers of chow diet-fed Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fasted overnight ( n ≥ 8/group ) . Values are means ± SEM . ( G ) Hematoxylin and eosin staining of liver sections from chow-fed Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice after an overnight fast . ( H ) Liver expression of Lpcat3 ( left panel ) and plasma TG levels ( right panel ) in Lpcat3fl/fl mice 3 weeks after being injected with an adenovirus encoding GFP or Cre . Mice were sacrificed under fed conditions ( ad libitum ) or after overnight fasting ( o/n ) ( n ≥ 6/group ) . Gene expression was measured by real-time . Values are means ± SEM . Statistical analysis was performed with a Student's t-test ( B , C , F , H ) . *p < 0 . 05; **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 00510 . 7554/eLife . 06557 . 006Figure 2—figure supplement 1 . Blood glucose levels and body weight in control and liver-specific Lpcat3 KO mice ( LKO ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 00610 . 7554/eLife . 06557 . 007Figure 2—figure supplement 2 . Quantification of western blots . Immunoblot results of apoB-100 and apoB-48 proteins in ( A ) Figure 2E and ( B ) Figure 2D were quantified by densitometry . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 00710 . 7554/eLife . 06557 . 008Table 2 . Breeding data for liver-specific Lpcat3-deficient miceDOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 008GenotypeNumber of miceObserved %Expected %Flox/Flox265350Flox/Floxalbumin-cre234650Genotypic ratio of adult mice obtained from Lpcat3fl/fl and Lpcat3fl/fl Albumin-Cre intercrosses . Analysis of plasma lipid levels revealed lower plasma TG levels following an overnight fast in L-Lpcat3 KO mice compared to controls ( Figure 2C ) . Levels of plasma total cholesterol and non-esterified free fatty acids ( NEFA ) were not different between groups . Body weight and fasting blood glucose levels were also not different between groups ( Figure 2—figure supplement 1 ) . Although total levels of plasma apolipoprotein B ( apoB ) were similar between groups ( Figure 2D , Figure 2—figure supplement 2B ) , fractionation of plasma lipoproteins revealed lower levels of apoB in the VLDL fraction in L-Lpcat3 KO mice ( Figure 2E , Figure 2—figure supplement 2A ) . Moreover , TG levels in the VLDL fraction were markedly reduced . We also observed a trend towards TG stores in the liver of L-Lpcat3 KO mice , along with histological evidence of increased lipid accumulation ( Figure 2F , G ) . As a complement to our analysis of L-Lpcat3 KO mice , which lack Lpcat3 expression in their livers from birth , we acutely deleted Lpcat3 in the liver of adult Lpcat3fl/fl mice with a Cre-expressing adenoviral vector . Interestingly , acute inactivation of Lpcat3 resulted in a more prominent decrease in fasting plasma TG levels compared to developmental deletion ( Figure 2H ) . Furthermore , acute deletion uncovered a decrease in ad-lib plasma TG levels that was not observed with developmental deletion ( Figure 2C , H ) . This finding suggests that there may be partial compensation for the chronic loss of Lpcat3 in the L-Lpcat3 KO mice . Collectively , the data of Figure 2 are consistent with a potential role for Lpcat3 in hepatic TG metabolism . To further explore a role for Lpcat3 in TG secretion , we challenged control and L-Lpcat3 KO mice with western diet ( 40% high fat and 0 . 2% cholesterol ) for 9 weeks . Mice lacking hepatic Lpcat3 again showed lower total plasma TG levels ( Figure 3A ) and a striking loss of TG in the VLDL plasma lipoprotein fraction ( Figure 3B , Figure 3—figure supplement 1A ) . In addition , these mice had increased levels of plasma apoB-100 compared to controls ( Figure 3C , Figure 3—figure supplement 1B ) . Analysis of tissue lipids revealed prominent accumulation of hepatic TG and cholesterol in the liver of L-Lpcat3 KO mice ( Figure 3D , E ) . We also challenged mice with a high-sucrose diet , which strongly stimulates hepatic lipid synthesis and secretion . On the high-sucrose diet , L-Lpcat3 KO mice exhibited hepatic TG accumulation ( Figure 3F , G ) . The accumulation of hepatic TG in L-Lpcat3 KO mice in the setting of reduced plasma VLDL TG implied that Lpcat3 activity might be crucial for the assembly and secretion of TG-rich lipoproteins from the liver . 10 . 7554/eLife . 06557 . 009Figure 3 . Dietary challenge accentuates metabolic phenotypes in liver-specific Lpcat3 knockout mice . ( A ) Plasma lipids of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed on a western diet for 9 weeks . Plasma was collected from mice fasted for 6 hr ( n ≥ 8/group ) . Values are means ± SEM . ( B ) Plasma samples same as in ( A ) were pooled ( n = 5 ) . Lipoprotein profile was analyzed by FPLC ( upper panel ) . ApoB protein in each fraction was analyzed by western blot ( lower panel ) and quantified ( Figure 3—figure supplement 1A ) . ( C ) ApoB protein in plasma samples same as in ( A ) was analyzed by western blot and quantified ( Figure 3—figure supplement 1B ) . ( D–E ) Hematoxylin and eosin staining ( D ) and lipid contents ( E ) of livers from western diet-fed Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice . Values are means ± SEM . ( F–G ) Hematoxylin and eosin staining ( F ) and lipid contents ( G ) of livers from high sucrose diet-fed Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice . Mice were fed on diet for 3 weeks and sacrificed after 6 hr fasting . Values are means ± SEM . Statistical analysis was performed using Student's t-test ( A , E and G ) . *p < 0 . 05; **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 00910 . 7554/eLife . 06557 . 010Figure 3—figure supplement 1 . Quantification of western blots . Immunoblot blot results of apoB-100 and apoB-48 proteins in ( A ) Figure 3B and ( B ) Figure 3C were quantified by densitometry . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 010 Lpcat3 is expressed at high levels in intestine as well as in the liver . We showed previously that hepatic Lpcat3 expression is regulated by the sterol-activated nuclear receptor LXR ( Rong et al . , 2013 ) . Here , we showed that intestinal Lpcat3 expression is strongly responsive to the administration of a synthetic LXR-agonist , GW3965 ( Figure 4A ) . To address whether Lpcat3 activity may also be important for TG metabolism in intestinal enterocytes , we generated intestine-specific Lpcat3 KO mice ( I-Lpcat3 KO ) by crossing the floxed mice to villin-Cre transgenics . I-Lpcat3 KO mice were born at the predicted Mendelian frequency , and their body weights at birth were comparable to controls ( Table 3 , Figure 4B ) . However , even though the pups suckled , they failed to thrive and showed severe growth retardation by 1 week of age ( Figure 4C ) . Expression of Lpcat3 was reduced more than 90% in duodenum of I-Lpcat3 KO mice as expected , and there was no compensatory increase in expression of Lpcat1 , Lpcat2 or Lpcat4 ( Figure 4D ) . Blood glucose levels in 1-week-old I-Lpcat3 pups were very low ( Figure 4E ) , consistent with results obtained with global knockouts ( Figure 1 ) . Plasma insulin levels were also correspondingly reduced . Plasma TG levels were lower and total cholesterol and NEFA levels were unchanged in I-Lpcat3 KO pups ( Figure 4E ) . Histological analysis of intestines from I-Lpcat3 KO pups revealed a dramatic accumulation of cytosolic lipid droplets in intestinal enterocytes ( Figure 4F ) , a phenotype reminiscent of intestinal apoB-deficient mice . Analysis of intestinal gene expression in I-Lpcat3 KO mice revealed reduced expression of several genes linked to intestinal TG metabolism , including Apob , Cd36 , Dgat2 , and Mogat2 ( Figure 4G ) . Given the massive enterocyte lipid accumulation in enterocytes , it is conceivable that some of those gene-expression changes were due , at least in part , to poor nutrition or cell toxicity . Nevertheless , these data were consistent with a role for Lpcat3 in TG mobilization and secretion–in the intestine as well as in the liver . 10 . 7554/eLife . 06557 . 011Figure 4 . Altered TG metabolism in intestine-specific Lpcat3 knockout mice . ( A ) Induction of Lpcat3 mRNA expression in duodenum of mice treated with 40 mg/kg/day GW3956 by oral gavage for 3 days ( n = 5/group ) . Gene expression was measured by real-time PCR . Values are means ± SEM . ( B ) Representative photograph and body weight of newborn Lpcat3fl/fl Villin-cre ( IKO ) and control Lpcat3fl/fl ( F/F ) pups ( n = 5/group for body weight measurement ) . Values are means ± SEM . ( C ) Representative photograph and body weight of 1 week-old Lpcat3fl/fl Villin-cre ( IKO ) and control Lpcat3fl/fl ( F/F ) pups ( n ≥ 6/group for body weight measurement ) . Values are means ± SEM . ( D ) Expression of Lpcat family members in 1 week-old Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Villin-cre ( IKO ) duodenum measured by real-time PCR ( n ≥ 6/group ) . Ct values of F/F samples were shown . Values are means ± SEM . ( E ) Blood glucose , plasma lipids and insulin levels in 1 week-old Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Villin-cre ( IKO ) pups ( n ≥ 6/group ) . Values are means ± SEM . ( F ) Hematoxylin and eosin staining of intestines from 1 week-old Lpcat3fl/fl ( WT ) and Lpcat3fl/fl Villin-cre ( IKO ) pups . ( G ) Expression of genes in duodenum of 1 week-old Lpcat3fl/fl ( WT ) and Lpcat3fl/fl Villin-cre ( IKO ) pups . Gene expression was measured by real-time PCR ( n ≥ 6/group ) . Values are means ± SEM . Statistical analysis was performed using Student's t-test ( A , B , D , E and F ) . *p < 0 . 05; **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 01110 . 7554/eLife . 06557 . 012Table 3 . Breeding data for intestine-specific Lpcat3-deficient miceDOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 012GenotypeNumber of pups/miceObserved %Expected %TimeWT252825At birthHet434850KO212325GenotypeNumber of pupsObserved %Expected %TimeF/F , Cre+202425At birthF/F , Cre−242825F/+ , Cre+192225F/+ , Cre−202425F/F , Cre+1216251 week oldF/F , Cre−243225F/+ , Cre+192525F/+ , Cre−202625Genotypic ratio of newborns and 1 week-old pups obtained from Lpcat3fl/fl and Lpcat3+/F Villin-Cre intercrosses . To gain insight into how the enzymatic activity of Lpcat3 was linked to these phenotypes , we performed lipidomic analyses . Previous studies using in vitro systems have profiled the substrate specificities of the four mammalian Lpcat family members ( Hishikawa et al . , 2008; Zhao et al . , 2008 ) . These studies suggested that Lpcat3 exhibited a preference for LysoPC and polyunsaturated fatty acyl-CoAs as substrates . Consistent with those results , we previously observed subtle changes in the levels of PC species containing polyunsaturated acyl in response to acute Lpcat3 knockdown in hepatocytes and macrophages ( Rong et al . , 2013 ) . However , the consequences of a genetic inactivation of Lpcat3 for membrane composition in vivo have never been addressed . Unexpectedly , analysis of phospholipid species in whole liver extracts from L-Lpcat3 KO mice by ESI-MS/MS revealed that Lpcat3 activity is uniquely required for the incorporation of arachidonate chains into PLs . Thus , despite the fact that Lpcat3 can catalyze the esterification of multiple unsaturated acyl chains into PC in vitro , and despite the fact that Lpcat3 is by far the most abundant Lpcat family member in the liver , the consequences of loss of Lpcat3 for PL composition were remarkably selective . The total level of PC was not different between L-Lpcat3 KO and control livers ( Figure 5A ) . However , there were striking reductions ( ∼70% ) in the abundance of 16:0 , 20:4 PC and 18:0 , 20:4 PC ( nomenclature: a:b , c:d; where a and c are the number of carbons and b and d are the number of double bonds of the sn-1 and sn-2 aliphatic groups , respectively ) , two of the most abundant arachidonoyl PC species in liver membranes ( Figure 5A ) . However , the actual reduction of 16:0 , 20:4 PC and 18:0 , 20:4 PCs in hepatocyte membranes was almost certainly much greater , given that a significant fraction of membranes in total liver extracts originate from other cell types ( where Lpcat3 activity is preserved ) . Furthermore , we observed compensatory increases in the abundance of other PC species in L-Lpcat3 KO livers , notably those containing monounsaturated chains ( e . g . , 16:0 , 18:1 PC and 18:0 , 18:1 PC; Figure 5A ) . 10 . 7554/eLife . 06557 . 013Figure 5 . Lpcat3 is required for the incorporation of arachidonate into phosphatidylcholine ( PC ) in mouse liver and very low-density lipoprotein ( VLDL ) . ( A–B ) ESI-MS/MS analysis of the abundance of PC species in livers from Lpcat3fl/fl ( Flox/Flox ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a chow diet ( A ) and a western diet ( B ) ( n ≥ 5/group ) . ( C ) ESI-MS/MS analysis of the abundance of PC species in plasma VLDL fraction from Lpcat3fl/fl ( fl/fl ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a chow diet . Plasma was harvested from mice after overnight fasting . VLDL fractions were pooled from 5 mice/group . ( D ) GC-FID analysis of the abundance of arachidonoyl acyl chain in triglyceride ( TAG ) in livers from Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed on a chow diet . Statistical analysis was performed using Student's t-test . Values are means ± SEM . *p < 0 . 05; **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 01310 . 7554/eLife . 06557 . 014Figure 5—figure supplement 1 . Lpcat3 is required for the incorporation of arachidonate into phosphatidylethanolamine in mouse liver . ( A–B ) ESI-MS/MS analysis of the abundance of phosphatidlyethanolamine ( PE ) species in livers from Lpcat3fl/fl ( Flox/Flox ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a chow diet ( A ) and a western diet ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 014 We also analyzed the effect of western diet on hepatic phospholipid composition in the presence and absence of Lpcat3 . The diet increased the abundance of certain PC species , such as 16:0 , 18:1 PC , presumably reflecting the abundance of oleate in the western diet . However , we observed the same prominent deficits in 16:0 , 20:4 PC and 18:0 , 20:4 PC on western diet as we observed on chow diet , with no change in total levels of PC ( Figure 5B ) . In addition , there were reductions in 16:1 , 18:2 PC and 18:1 , 20:4 PC in western diet-fed Lpcat3 KO livers compared to controls . Severe reductions of phosphatidlyethanolamine ( PE ) species containing arachidonate chains were also observed in L-Lpcat3 KO mice on both chow and western diets ( Figure 5—figure supplement 1 ) . Interestingly , 16:0 , 20:4 PE and 18:0 , 20:4 PE are particularly abundant PE species in liver on western diet , and reductions in their levels was sufficient to reduce total PE levels . Since unsaturated PC species are abundant in plasma lipoproteins , and since mice lacking Lpcat3 in liver or intestine show reduced plasma TG levels , we tested whether loss of Lpcat3 expression in liver would alter the phospholipid composition of VLDL secreted from the liver . After fasting L-Lpcat3 KO and control mice overnight , the VLDL was isolated from pooled plasma ( n = 5 mice/group ) . Analysis of PC species in the pooled VLDL fractions by ESI-MS/MS revealed highly selective reductions in 16:0 , 20:4 PC and 18:0 , 20:4 PC ( Figure 5C ) . Thus , the phospholipid deficits of L-Lpcat3 KO hepatic membranes are passed on to the VLDL particles that they generate , strongly suggesting that loss of these PC species is mechanistically related to the altered TG content of the particles . Interestingly , the severe loss of arachidonate in livers of Lpcat3 KO mice on chow diet was observed in PLs , but not in TG ( Figure 5D ) nor in cholesterol esters ( Figure 6A ) . There was an increase in total cholesterol ester and a number of cholesterol ester species in L-Lpcat3 KO mice , consistent with the histologic evidence of increased neutral lipid content in the liver ( Figure 2 ) . There was no accumulation of the major lipid substrates of Lpcat3 ( 16:0 lysoPC and 18:0 lysoPC ) in L-Lpcat3 KO mice ( Figure 6B ) , suggesting these precursors are efficiently shuttled into alternative pathways in the absence of Lpcat3 . Broadly similar results were obtained from mice fed western diet , although we did observe modestly lower levels of 20:4 lysoPC ( Figure 6C , D ) . 10 . 7554/eLife . 06557 . 015Figure 6 . Analysis of cholesteryl ester and lysoPC species in L-Lpcat3 KO mice . ( A–B ) ESI-MS/MS analysis of the abundance of cholesteryl ester ( A ) and lysophosphatidylcholine ( LysoPC ) ( B ) species in livers of Lpcat3fl/fl ( Flox/Flox ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a chow diet ( n ≥ 5/group ) . ( C–D ) ESI-MS/MS analysis of the abundance of cholesteryl ester ( C ) and lysophosphatidylcholine ( LysoPC ) ( D ) species in livers of Lpcat3fl/fl ( Flox/Flox ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a western diet ( n ≥ 5/group ) . Statistical analysis was performed using Student's t-test . Values are means ± SEM . *p < 0 . 05; **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 015 To determine the consequence of loss of Lpcat3 activity for hepatic gene expression we performed transcriptional profiling . Despite the marked changes in membrane composition in L-Lpcat3 KO livers , the effect on gene expression was surprisingly limited ( Figure 7A ) . On chow diet , only a handful of genes were altered more than twofold . Among the highest changes were increased expression of the lipoprotein remodeling enzyme Pltp and the lipid transport protein Cd36 . A number of additional genes involved in lipid metabolism were induced to a more modest degree , including Txnip and Fabp2 . Interestingly , many of the genes that were induced are known targets for the lipid-activated nuclear receptor PPARα , suggesting a compensatory gene expression response to increased cellular lipid levels in the absence of Lpcat3 ( Figures 2 , 3 ) . 10 . 7554/eLife . 06557 . 016Figure 7 . Altered gene expression linked to lipid metabolism and inflammation in liver-specific Lpcat3 knockout mice . ( A ) Gene expression in livers of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed on a chow diet was analyzed by Affymetrix arrays . Select genes under the gene ontology ( GO ) term ‘lipid metabolism’ are presented by heatmap . Samples from 5 mice/group were pool for analysis . ( B–C ) Gene expression in livers of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed on a western diet for 9 weeks was analyzed by Affymetrix arrays . Select genes under the GO term ‘lipid metabolism’ ( B ) and ‘immune system process’ ( C ) are presented by heatmap . Samples from 5 mice/group were pool for analysis . ( D ) Expression of selective lipid metabolism genes in livers of Lpcat3fl/fl ( Fl/Fl ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a chow diet was analyzed by real-time PCR ( n ≥ 5/group ) . Values are means ± SEM . ( E–F ) Expression of selective lipid metabolism ( E ) and inflammation ( F ) genes in livers of Lpcat3fl/fl ( Fl/Fl ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a western diet was analyzed by real-time PCR ( n ≥ 5/group ) . Values are means ± SEM . Statistical analysis was performed using Student's t-test ( D , E and F ) . **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 01610 . 7554/eLife . 06557 . 017Figure 7—figure supplement 1 . Chronic deletion of Lpcat3 does not induce unfolded protein response downstream gene expression . ( A ) Gene expression in livers of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed on a chow diet was analyzed by Affymetrix arrays . Select ER stress marker genes are presented by heatmap . Samples from 5 mice/group were pool for analysis . ( B ) Gene expression in livers of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( L-KO ) mice fed on a western diet was analyzed by Affymetrix arrays . Select ER stress marker genes are presented by heatmap . Samples from 5 mice/group were pool for analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 017 More prominent changes in gene expression in L-Lpcat3 KO mice were observed in the setting of western diet feeding ( Figure 7B ) . Multiple genes involved in lipid metabolism and transport and in lipid droplet formation were upregulated , again likely reflecting a response to increased hepatic TG and cholesterol content . Interestingly , the putative transporter Mfsd2a ( Nguyen et al . , 2014 ) was induced in Lpcat3-deficient livers on both chow and western diet , perhaps as compensation for the loss of arachidonoyl PLs . In addition , there was increased expression of a number of genes linked to inflammation and inflammatory cell recruitment ( Figure 7C ) . These data are consistent with our earlier study showing that adenoviral expression of Lpcat3 is protective against hepatic inflammation in the setting of lipid excess ( Rong et al . , 2013 ) . We validated the microarray results for a number of genes by real-time PCR and also assessed the expression of other genes relevant to lipoprotein production ( Figure 7D , E ) . Importantly , there was no change in the expression of mRNAs encoding apoB or of the critical lipid transfer protein MTTP in livers of Lpcat3 KO mice , indicating that altered plasma TG levels cannot be attributed to changes in expression of these factors . We previously reported that acute knockdown of Lpcat3 expression in livers of genetically obese mice exacerbated lipid-induced ER stress ( Rong et al . , 2013 ) . Genetic deletion of Lpcat3 from liver did not lead to increased mRNA expression of ER stress markers in mice fed chow or western diet ( Figure 7—figure supplement 1 ) . These observations suggest that there may be compensatory responses in membrane composition that prevent induction of the ER stress response in the setting of chronic Lpcat3 deletion . In support of this idea , we observed a prominent increase in the abundance of oleoyl-PC species in L-Lpcat3 mice ( Figure 5 ) . We cannot exclude the possibility that ER stress may be increased in L-Lpcat3 mice in the setting of genetic obesity or other causes of severe lipotoxicity . We next investigated the etiology of reduced VLDL TG levels in mice lacking Lpcat3 . Protein levels of apoB in liver were not different between control and L-Lpcat3 KO mice ( Figure 8A ) . Together with the preserved levels of apoB in plasma ( Figures 2 , 3 ) , these data suggest that a defect in production or secretion of apoB alone cannot explain the change in plasma VLDL . We directly assessed hepatic TG secretion by treating mice with the lipase inhibitor Tyloxapol and measuring TG accumulation in fasted mice . L-Lpact3 KO mice showed a markedly reduced rate of TG secretion ( Figure 8B ) , suggesting that reduced incorporation of TGs into lipoproteins was the likely cause of the reduced VLDL TG levels . Consistent with this hypothesis , negative staining of plasma VLDL fractions by EM revealed markedly smaller VLDL particles in L-Lpcat3 KO mice compared to controls ( Figure 8C ) . 10 . 7554/eLife . 06557 . 018Figure 8 . Loss of Lpcat3 from liver impairs hepatic TG secretion . ( A ) ApoB protein in livers from Lpcat3fl/fl ( Flox/Flox ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a chow diet was analyzed by western blot ( n = 3 ) . ( B ) VLDL-TG secretion in Lpcat3fl/fl mice transduced with adenoviral expressed Cre ( Ad-Cre ) , compared to control GPF ( Ad-GFP ) for 4 weeks . Mice were fasted for 6 hr followed by intravenous injection of tyloxapol . Plasma TG was measured at indicated durations after tyloxapol injection . Values are means ± SEM . ( C ) Plasma VLDL particle size in Lpcat3fl/fl Albumin-Cre mice . Isolated VLDL particles from Lpcat3fl/fl ( Fl/Fl ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice fed on a western diet were stained with 2 . 0% uranyl acetate and visualized by electron microscopy ( EM ) ( left ) and their size quantified ( right ) . Statistical analysis was performed using two-way ANOVA with Bonferroni post hoc tests ( B ) . **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 018 To obtain further insight into the nature of the lipoprotein production defect in L-Lpcat3 KO mice , we examined liver samples by EM . Nascent lipoproteins , ranging between 0 . 05 and 0 . 11 microns in diameter , were easily visualized in the Golgi apparatus and secretory vesicles of control mice ( Figure 9 ) . Lipoprotein particles were also present in Golgi and secretory vesicles of L-Lpcat3 KO mice , however they were markedly smaller , ranging between 0 . 03 to 0 . 08 microns in diameter ( Figure 9 ) . We also observed small lipoprotein particles in the ER in mice of both genotypes ( Figure 9—figure supplement 1 ) . We did not find differences in the morphology of Golgi , ER or mitochondria between control and L-Lpcat3 KO livers , suggesting that loss of arachidonoyl PLs does not dramatically alter membrane structure in these organelles ( Figure 9—figure supplement 1 ) . 10 . 7554/eLife . 06557 . 019Figure 9 . Reduced nascent lipoprotein particle size in the lumen of Golgi and secretory vesicles in Lpcat3-deficient liver . EM of imidazole-stained liver sections . Lpcat3fl/fl ( Fl/Fl ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice were fasted for 6 hr . Samples were fixed and processed as described in the methods . Arrowheads indicate nascent lipoprotein particles . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 01910 . 7554/eLife . 06557 . 020Figure 9—figure supplement 1 . Loss of Lpcat3 in liver does not alter membrane structure in ER and mitochondria . ( A ) EM of imidazole-stained liver sections . Arrowheads indicate nascent lipoprotein particles in smooth ER lumen . ( B ) EM of imidazole-stained liver sections . Mitochondria are shown . ( C ) EM of imidazole-stained liver sections . Rough ER is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 020 To further investigate the hypothesis that defective apoB lipidation was responsible for the phenotype of L-Lpcat3 KO mice , we isolated the Golgi apparatus from livers , fractionated the luminal contents by density gradient centrifugation , and analyzed the distribution of ApoB in the different fractions . There was reduced apoB in the most buoyant lipoprotein fractions , consistent with reduced lipidation of apoB particles in the absence of Lpcat3 ( Figure 10A ) . In line with this finding , we found reduced TG levels in Golgi membrane fractions isolated by density gradient centrifugation from L-Lpcat3 KO mice compared to controls in two independent purifications ( Figure 10B ) . 10 . 7554/eLife . 06557 . 021Figure 10 . Lpcat3 activity regulates membrane lipid mobility and apoB lipidation in hepatocytes . ( A ) The distribution of apoB-containing lipoproteins in the Golgi fractions isolated from livers of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( LKO ) mice . Golgi luminal contents were subject to a density gradient ultracentrifugation . 12 fractions from the gradient were harvested from top to bottom and analyzed by western blot . ( B ) Triglyceride contents of the Golgi fractions isolated from livers of Lpcat3fl/fl ( F/F ) and Lpcat3fl/fl Albumin-Cre ( LKO ) mice . Results from two representative experiments were shown . ( C ) Live primary hepatocytes from Lpcat3fl/fl ( Flox/Flox ) and Lpcat3fl/fl Albumin-Cre ( L-Lpcat3 KO ) mice were stained with laurdan . The laurdan emission spectrum was captured by a 2-photon laser-scan microscope . Generalized polarization ( GP ) was calculated from the emission intensities obtained from images . Higher GP value indicates that membranes are more ordered and less dynamic . The GP value of each pixel was used to generate a pseudocolor GP image . ( D ) The binary histograms of the GP distribution of the GP images ( n = 4 ) . The size of the GP binary is 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 021 To understand how changes in membrane phospholipid composition in Lpcat3-deficient mice might lead to reduced mobilization and secretion of TGs , we performed biophysical studies on the impact of Lpcat3 deficiency on lipid movement in living cells . Prior studies have employed the lipophilic dye laurdan to interrogate lipid dynamics in membranes ( Parasassi and Gratton , 1995; Vest et al . , 2006; Golfetto et al . , 2013 ) . Laurdan is a fluorescent lipophilic molecule that can be used to detect changes in membrane dynamics due to its sensitivity to the polarity of the membrane environment . Changes in membrane dynamics shift the laurdan emission spectrum , which can be quantified by the generalized polarization ( GP ) calculated from the spectrum shifts ( Parasassi et al . , 1990 ) . We isolated primary hepatocytes from L-Lpcat3 KO mice and controls , stained with laurdan , and monitored fluorescence in live cells by microscopy . A visual representation of the calculated GP intensity is presented in Figure 10C , and the data are quantitated in Figure 10D . The prominent increase in GP ( as shown by the yellow/orange pseudocolor signal ) in cells lacking Lpcat3 is indicative of areas of cellular membrane with reduced dynamics . These data show that membranes are less dynamic , and their component lipids move less readily , in the absence of Lpcat3 , an observation that is consistent with the reduced amount of arachidonate groups in membrane PLs . We propose that efficient transfer of bulk TG to nascent VLDL particles is enabled by the presence of arachidonoyl PLs in the ER/Golgi membrane due to its unique ability to facilitate optimal lipid movement and membrane dynamics . Finally , as LXR agonists have previously been reported to promote hepatic TG secretion ( Schultz et al . , 2000 ) , we tested the requirement for Lpcat3 induction in this effect . Control and L-Lpcat3 KO mice were treated for 2 days with the LXR agonist GW3965 ( 20 mpk/day ) by oral gavage . The TG levels in fractionated plasma were then determined after 6 hr fast . In line with prior studies , treatment of control mice with LXR agonist led to a prominent ( 95% ) increase in plasma VLDL TG levels ( Figure 11 ) . By contrast , treatment of L-Lpcat3 KO mice with LXR agonist had a much more modest effect ( 45% increase ) . These data suggest that induction of Lpcat3 expression plays an important role in the control of hepatic VLDL production by LXRs . 10 . 7554/eLife . 06557 . 022Figure 11 . Lpcat3 is required for LXR-dependent VLDL TG secretion from liver . Plasma lipoprotein profile from flox/flox and L-Lpcat3 mice gavaged with LXR agonist GW 3965 ( 20 mg/kg ) for 2 days . Pooled plasma ( 5 mice/group ) was analyzed by FPLC . Increases in VLDL TG were calculated based on the area under the curve . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 022
It has long been appreciated that phospholipid availability can influence the production of lipoproteins ( Vance , 2008; Abumrad and Davidson , 2012 ) , but the molecular basis of these connections , beyond the obvious need for sufficient PLs to coat the surface of lipoprotein particles , has been unclear . We have shown here that Lpcat3 is uniquely required for the incorporation of arachidonic acid into membranes , and that selective reduction in the abundance of these lipids impairs TG mobilization and lipoprotein production . These studies highlight a previously unrecognized requirement for a specific membrane lipid class in lipoprotein metabolism . Earlier studies using in vitro systems have demonstrated that changes in polyunsaturated PL can affect membrane-associated biological processes , including the assembly of signalsomes on membranes and endocytosis ( Koeberle et al . , 2013; Pinot et al . , 2014 ) . However , assessing the physiological impact of changing the abundance of individual PL species in animals has been a difficult experimental problem . Our study identifies Lpcat3 as a critical determinant of the abundance of arachidonoyl PLs in mice . The largely preserved levels of other PL species containing polyunsaturated chains in Lpcat3-deficient livers suggests that other Lpcats may be able to catalyze the incorporation of certain polyunsaturated fatty acids into membranes in the absence of Lpcat3 . However , Lpcat3 is apparently unique in its ability to catalyze arachidonoyl PL synthesis . Lpcat3-deficient mouse models therefore provide an unprecedented opportunity to study the physiological and pathophysiological consequences of manipulation of membrane phospholipid composition in vivo . Previous research has elucidated a requirement for de novo PC biogenesis in the production and secretion of VLDL from liver . Feeding mice a choline-deficient diet or deleting genes involved in de novo PC synthesis ( e . g . , Pemt , CT-α ) impairs VLDL secretion and induces hepatosteatosis ( Noga and Vance , 2003; Jacobs et al . , 2004 ) . These models exhibit reduced total apoB protein in plasma as well as lower plasma TGs , suggesting that adequate PC biosynthesis is important for both apoB secretion and VLDL lipidation . In our study , mice lacking Lpcat3 retain the ability to secrete apoB and are capable of producing small poorly-lipidated VLDL particles . However , they are unable to transfer TG to lipoproteins at an appropriate rate in the setting of increased metabolic demand . Our results suggest that the reduced availability of arachidonoyl PLs becomes especially problematic when there is a need to mobilize a large bolus of TG into lipoproteins . For example , when L-Lpcat3 KO mice are challenged with the lipogenic sucrose diet , they are unable to efficiently mobilize the newly synthesized TG into plasma lipoproteins and it instead accumulates in cytosolic lipid droplets in the liver . Similarly , mice lacking Lpcat3 in the intestine are unable to handle the high TG load of milk during suckling and accumulate large amounts of lipid in enterocytes . Given that Lpcat3 is the major Lpcat enzyme in enterocytes ( Figure 4D ) , loss of Lpcat3 would be expected to impair the re-esterification of LysoPC to PC in enterocytes , which has been demonstrated to be important for lipid absorption . Unfortunately , the early lethality of the I-Lpcat3 KO pups presents a challenge to the analysis of lipid absorption . An inducible knockout system may prove a better model for studying the impact of Lpcat3 on intestinal lipid transport in adult mice . Lpcat3 is a integral membrane protein of the ER and is therefore ideally positioned to produce arachidonoyl PC at the site of lipoprotein biogenesis ( Fisher and Ginsberg , 2002; Zhao et al . , 2008 ) . Our observations suggest that Lpcat3 and its lipid products are likely not essential for the cotranslational lipidation of apoB and the generation of primordial VLDL particles ( Fisher and Ginsberg , 2002 ) . However , the small size of plasma VLDL particles , together with the reduced TG-rich apoB-containing particles in the Golgi fraction of L-Lpcat3 KO livers , strongly suggest that Lpcat3 impacts the second step of VLDL assembly–bulk TG addition to lipid-poor apoB particles and the generation of mature VLDL . TG-rich VLDL assembly is highly dependent on the efficient trafficking of stored cytosolic lipid from lipid droplets or newly synthesized lipid to primordial VLDL particles . One factor that could affect this transfer is the membrane environment where the bulk lipidation occurs . Although the precise mechanism that links membrane PL composition and VLDL lipidation is not clear , biophysical studies suggested that greater lipid transport is generally observed with more fluid and highly curved membrane surfaces ( Lev , 2012 ) . Results from our laurdan staining experiments support the notion that the presence of arachidonoyl PLs in intracellular membranes promotes the dynamics of the membranes . However , detailing these biophysical changes and elucidating how they influence lipid movement will require further study with specialized tools and systems . Nevertheless , our data suggest that membrane dynamics may play an important role in VLDL lipidation in vivo . Interestingly , a recent proteomic study identified Lpcat3 as a component of the VLDL transport vesicle , indicating that Lpcat3 travels with primordial VLDL particles as they bud from the ER and move to the Golgi ( Rahim et al . , 2012 ) . Together with prior work , our data favor a model in which Lpcat3 modifies the arachidonoyl-PC composition of both membranes and lipoprotein particles during VLDL assembly , thereby generating a local membrane environment that facilitates lipid transport and bulk lipidation . Our prior studies showed that acute shRNA-mediated knockdown of Lpcat3 expression in liver exacerbated ER in the setting of obesity and hepatic steatosis . However , genetic deletion of Lpcat3 expression in liver did not lead to overt ER stress pathway activation , suggesting that increased abundance of monounsaturated PC species ( e . g . , 16:0 , 18:1 PC , Figure 5 ) may partially compensate to maintain ER membrane homeostasis . It remains to be determined whether stronger lipotoxic stimuli , such as the hepatosteatosis observed in ob/ob mice , may yet provoke increased ER stress responses in the genetic absence of Lpcat3 . This work illustrates how manipulation of membrane composition can be used as regulatory mechanism to control metabolic pathways . A model for Lpcat3 action is presented in Figure 12 . Lpcat3 is not only a required component of lipoprotein production; it is also a regulated one . The fact that Lpcat3 expression is dynamically regulated by LXRs in liver and intestine strongly suggests that the level of Lpcat3 enzymatic activity has evolved to respond to dietary and metabolic demands . LXR is transcriptionally activated as cellular cholesterol levels rise . Thus , the LXR-Lpcat3 pathway provides a mechanism to integrate sterol metabolism and membrane PL composition . We speculate that Lpcat3 expression is induced in response to lipid loading in order to increase ER/Golgi membrane flexibility and to facilitate efficient TG secretion ( thereby unloading cells of excess lipids ) . In support of this idea , we found that the ability of pharmacologic LXR agonist to stimulate hepatic VLDL production was impaired in the genetic absence of Lpcat3 . 10 . 7554/eLife . 06557 . 023Figure 12 . Schematic illustration of the role of the LXR-Lpcat3 pathway in VLDL lipidation . Activation of LXRs promotes the incorporation of arachidonate into intracellular membranes through the induction of Lpcat3 expression . This change in membrane composition creates a dynamic membrane environment that facilitates the transfer of TG synthesized on ER and/or in cytosolic LD to nascent apoB–containing lipoproteins particles , leading to the efficient lipidation of apoB–containing lipoproteins . DOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 023 Finally , these studies open the door to new strategies for pharmacologic intervention in systemic lipid metabolism . Abnormal phospholipid metabolism has been associated with several metabolic diseases . For example , lower amounts of polyunsaturated PLs have been observed in liver biopsies from nonalcoholic steatohepatitis patients ( Puri et al . , 2007 ) , and cell membranes from patients with atherosclerotic disease tend to show decreased membrane fluidity ( Chen et al . , 1995 ) . However , it has heretofore been difficult to draw a causal link between aberrant PL composition and the pathogenesis of metabolic diseases . Our studies provide direct evidence that alterations in arachidonoyl PC abundance impair lipoprotein metabolism in liver and intestine in vivo . Recently , a human GWAS reported a highly significant association between LPCAT3 and the phospholipid composition of red blood cell membranes , indicating that Lpcat3 is also a key regulator of phospholipid composition in humans ( Tintle et al . , 2014 ) . Therefore , an improved understanding of the mechanisms by which Lpcat3 and arachidonoyl PCs regulate lipid homeostasis could lead to new approaches to metabolic diseases . We speculate that small molecule inhibitors of Lpcat3 could be of potential utility for lowering plasma lipid levels in hyperlipidemic individuals . In pursuing such a strategy , one would need to monitor the impact of Lpcat3 inhibition on liver TG stores , hepatic inflammation and lipid malabsorption , in addition to the plasma lipids . However , modest increases in liver TG stores , even if they occurred in humans , might not doom such a strategy . Inhibiting apoB synthesis with antisense compounds has found a place in the management of some cases of hyperlipidemia , despite modest increases in liver TG stores .
Total RNA was isolated from cells and tissues with Trizol ( Invitrogen , Carlsbad , CA ) . cDNA was synthesized , and gene expression was quantified by real-time PCR with SYBR Green ( Diagenode , Denville , NJ ) and an ABI 7900 . Gene expression levels were normalized to 36B4 or GAPDH . Primer sequences are listed in Table 4 . For microarray experiments , RNA was pooled from n = 5 biological replicates and processed in the UCLA Microarray Core Facility with Gene-Chip Mouse Gene 430 . 2 Arrays ( Affymetrix , Santa Clara , CA ) . Data analysis was performed with GenespringGX ( Agilent , Santa Clara , CA ) . The GEO accession numbers for the microarray data are GSE65352 and GSE65353 . 10 . 7554/eLife . 06557 . 024Table 4 . Quantitative real-time PCR primer sequencesDOI: http://dx . doi . org/10 . 7554/eLife . 06557 . 024Murine qPCR primersForwardReverseCD36TTGAAAAGTCTCGGACATTGAGTCAGATCCGAACACAGCGTAPLPTGTCTAAAATGAATATGGCCTTCGCCAGAAGTGATGAACGTGGADGAT1TTCCGCCTCTGGGCATTGCCCACAATCCAGGCCADGAT2GGCGCTACTTCCGAGACTACTGGTCAGCAGGTTGTGTGTCAPOA4ACCCAGCTAAGCAACAATGCTGTCCTGGAAGAGGGTACTGAAPOEGACTTGTTTCGGAAGGAGCTGCCACTCGAGCTGATCTGTCAAOPBCTGAACATCAAGAGGGGCATCGGTAACCTGAGTTGAGCAGTTTMTTPGGCAGTGCTTTTTCTCTGCTTGAGAGGCCAGTTGTGTGACPDK4ACCGAAGAACCTGGCGAAGTGATCCCGTAAAATGTCAGGCCD68GACCTACATCAGAGCCCGAGTCGCCATGAATGTCCACTGCCL6TCTTTATCCTTGTGGCTGTCCTGGAGGGTTATAGCGACGATACSL3TCTAGGAGTGAAGGCCAACGGCAATATCTGAGGGCAGTGGACSL5AACCAGTCTGTGGGGATTGAGCGTCTTGGCGTCTGAGAAGTAMOGAT2TCTTCCAGTACAGCTTTGGCCTCATGATATAGCGCTGATGAAGCCGGTFABP1AGTACCAATTGCAGAGCCAGGAGAGACAATGTCGCCCAATGTCATGGTFABP2AGAGGAAGCTTGGAGCTCATGACATCGCTTGGCCTCAACTCCTTCATAFATP1CGCTTTCTGCGTATCGTCTGGATGCACGGGATCGTGTCTLPCAT3GGCCTCTCAATTGCTTATTTCAAGCACGACACATAGCAAGGALPCAT1GTGCACGAGCTGCGACTGCTGCTCTGGCTCCTTATCALPCAT2TGTACTAATCGCTCCTGTTTGATTCACTGGAACTCCTGGGATGLPCAT4TTCGGTTTCAGAGGATACGACAAAATGTCTGGATTGTCGGACTGAA36B4AGATGCAGCAGATCCGCATGTTCTTGCCCATCAGCACCGAPDHTGTGTCCGTCGTGGATCTGACCTGCTTCACCACCTTCTTGAT Cells and tissue lysates were prepared by homogenization in RIPA buffer ( 50 mM Tris–HCl , pH 7 . 4; 150 mM NaCl , 1% NP-40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS ) supplemented with protease and phosphatase inhibitors ( Roche Molecular Biochemicals ) . Lysates were cleared by centrifugation . Plasma samples were diluted with RIPA buffer . Protein lysate were then mixed with NuPAGE LDS Sample Buffer , size-fractionated on 4–12% Bis-Tris Gels ( Invitrogen ) , transferred to hybond ECL membrane ( GE Healthcare , Piscataway , NJ ) , and incubated with an apoB antibody ( Abcam , Cambridge , MA ) . Primary antibody binding was detected with a goat anti-rabbit secondary antibody and visualized with chemiluminescence ( ECL , Amersham Pharmacia Biotech ) . A conditional knockout allele for Lpcat3 was generated with a sequence replacement ‘knock-out first/conditional-ready’ gene-targeting vector . The vector was electroporated into JM8A1 . N3 ES cell line from C57BL/6N mice . Positive clones were identified by genotyping and long-range PCR at both the 5′ and 3′ ends . Two targeted ES cell clones were injected into C57BL/6 blastocysts to generate chimeric mice . High-percentage male chimeras were obtained , and resulting chimeras bred with female C57BL/6 mice to obtain heterozygous knockout mice ( Lpcat3+/− ) . Lpcat3+/− mice were intercrossed to produce homozygous knockout mice ( Lpcat3−/− ) . To create a conditional knockout allele , Lpcat3+/− mice were mated with mice expressing a Flpe recombinase deleter transgene ( Jackson Laboratory , Bar Harbor , Maine ) . That transgene excises the gene-trapping cassette in intron 2 of Lpcat3 , producing a conditional knockout allele containing loxP sites in intron 2 and intron 3 . Lpcat3Fl/Fl mice were crossed with albumin-Cre or villin-Cre transgenic mice from The Jackson Laboratory ( Bar Harbor , ME ) . Mice were fasted for 4 hr and then injected intravenously with 500 mg/kg body weight of tyloxapol ( Fisher , Pittsbugh , PA ) . Blood was subsequently collected , and the plasma separated by centrifugation for the measurement of TGs . Lpcat3−/− , Lpcat3fl/fl , Lpcat3fl/fl; Albumin-Cre , and Lpcat3fl/fl; Villin-Cre mice were generated as described above . All the mice were housed under pathogen-free conditions in a temperature-controlled room with a 12-hr light/dark cycle . Mice were fed a chow diet , a western diet ( Research Diets #D12079B ) , or a high-sucrose diet ( Research Diets #D07042201 ) . Liver tissues were collected and frozen in liquid nitrogen and stored at −80°C or fixed in 10% formalin . Blood was collected by retro-orbital bleeding , and the plasma was separated by centrifugation . Small intestines were excised and cut into three segments with length ratios of 1:3:2 ( corresponding to duodenum , jejunum and ileum ) . For adenoviral infections , VQAd-CMV Cre/eGFP and VQAd-Empty eGFP were purchased from Viraquest . 8- to 10-week-old male Lpcat3fl/fl mice were injected ( via the tail vein ) with 2 × 109 plaque-forming units ( pfu ) . Mice were sacrificed or used for TG secretion studies 3–4 weeks after the adenovirus injection . Blood was collected from mice by cardiac puncture . Plasma lipids were measured with the Wako L-Type TG M kit , the Wako Cholesterol E kit; and the Wako HR series NEFA-HR ( 2 ) kit . Tissue lipids were extracted with Bligh-Dyer lipid extraction ( Bligh and Dyer , 1959 ) and measured with the same enzymatic kits . Plasma fast protein liquid chromatography ( FPLC ) lipoprotein profiles were performed in the Lipoprotein Analysis Laboratory of Wake Forest University School of Medicine . Tissue histology was performed in the UCLA Translational Pathology Core Laboratory . Animal experiments were conducted in accordance with the UCLA Animal Research Committee . Liver tissue and plasma were snap frozen at the temperature of liquid nitrogen . Liver was homogenized on ice in phosphate buffered saline . Plasma or liver homogenates were subsequently subjected to a modified Bligh-Dyer lipid extraction ( Bligh and Dyer , 1959 ) in the presence of lipid class internal standards including eicosanoic acid , 1-0-heptadecanoyl-sn-glycero-3-phosphocholine , 1 , 2-dieicosanoyl-sn-glycero-3-phosphocholine , cholesteryl heptadecanoate , and 1 , 2-ditetradecanoyl-sn-glycero-3-phosphoethanolamine ( Demarco et al . , 2013 ) . Lipid extracts were diluted in methanol/chloroform ( 4/1 , vol/vol ) and molecular species were quantified using electrospray ionization mass spectrometry on a triple quadrupole instrument ( Themo Fisher Quantum Ultra ) employing shotgun lipidomics methodologies ( Han and Gross , 2005 ) . Lysophosphatidylcholine molecular species were quantified as sodiated adducts in the positive ion mode using neutral loss scanning for 59 . 1 amu ( collision energy = −28 eV ) . PC molecular species were quantified as chlorinated adducts in the negative ion mode using neutral loss scanning for 50 amu ( collision energy = 24 eV ) . Phosphatidylethanolamine molecular species were first converted to fMOC derivatives and then quantified in the negative ion mode using neutral loss scanning for 222 . 2 amu ( collision energy = 30 eV ) . Cholesteryl ester molecular species were quantified as sodiated adducts in the positive ion mode using neutral loss scanning of 368 . 5 amu as previously described ( Bligh and Dyer , 1959; Bowden et al . , 2011 ) . Individual molecular species were quantified by comparing the ion intensities of the individual molecular species to that of the lipid class internal standard with additional corrections for type I and type II 13C isotope effects ( Han and Gross , 2005 ) . Triacylglycerol fatty acid composition was determined as follows . Samples were subjected to Bligh-Dyer extraction in the presence of the internal standard , triheptadecenoin , followed by thin layer chromatographic purification of triacylglycerol using silica gel G plates with petroleum ether/ethyl ether/acetic acid; 80/20/1 as mobile phase . Purified triacylglycerol was then subjected to fatty acid methanolysis , and fatty acid methyl esters were then determined using gas chromatography with flame ionization detection as previously described ( Ford and Gross , 1988 , 1989 ) . Membrane dynamics was analyzed as described ( Golfetto et al . , 2013 ) . Briefly , primary hepatocytes from Lpcat3fl/fl and Lpcat3fl/fl; Albumin-Cre mice were isolated as described ( Rong et al . , 2013 ) . Cells were incubated with 1 . 8 mM Laurdan ( 6-dodecanoyl-2-dimethylaminonaphthalene; Invitrogen ) at 37°C for 30 min . Cells were rinsed with phosphate-buffered saline ( PBS ) , and fresh culture medium was added . Spectral data were acquired with a Zeiss LSM710 META laser scanning microscope coupled to a 2-photon Ti:Sapphire laser ( Mai Tai , Spectra Physics , Newport Beach , CA ) producing 80-fs pulses at a repetition of 80 MHz with two different filters: 460/80 nm for the blue channel and 540/50 nm for the green channel . Spectral data were processed by the SimFCS software ( Laboratory for Fluorescence Dynamics ) . The GP value was calculated for each pixel using the two Laurdan intensity images ( 460/80 nm and 540/50 nm ) . The GP value of each pixel was used to generate the pseudocolored GP image . GP distributions were obtained from the histograms of the GP images . The Golgi fraction was isolated as described ( Vance and Vance , 1988 ) . Briefly , fresh liver tissue was homogenized with a motor driven Potter-Elvehjem homogenizer in homogenization buffer ( 37 . 5 mM TRIS-maleate , pH 7 . 4; 0 . 5 M sucrose; 1% dextran; and 5 mM MgCl2 ) . After an initial centrifugation at 5000×g for 15 min , most of the supernatant was removed and the yellow-brown portion ( approximately the upper one-third ) of the pellet was suspended in homogenization buffer , layered over 2 . 7 ml of 1 . 2 M sucrose and spin for 30 min at 100 , 000×g in a swinging bucket rotor ( SW50 . 1 ) . The Golgi fraction was collected at the buffer/1 . 2 M sucrose interface . The Golgi was suspended in distilled water and pelleted by centrifugation at 5500×g for 20 min . Aliquots of the Golgi fraction were used to measure TG and protein concentration . Golgi apoB was analyzed as described ( Gusarova et al . , 2003; Li et al . , 2012 ) . Briefly , the luminal contents were released from Golgi fraction by treatment with 0 . 1 M sodium carbonate ( pH 11 ) and deoxycholic acid ( 0 . 025% ) for 30 min at room temperature . Bovine serum albumin ( BSA ) was added to a final concentration of 5 mg/ml , followed by centrifugation ( 50 , 000 rpm in a Beckman SW60 rotor for 1 hr at 4°C ) to remove membranes . The luminal contents ( supernatant ) were adjusted to pH 7 . 4 with acetic acid , adjusted to a sucrose concentration of 12 . 5% ( wt/vol ) , and placed on the top of a step gradient consisting of 1 . 9 ml of 49% sucrose and 1 . 9 ml of 20% sucrose . Next , 2 . 8 ml of PBS was layered on the top of the supernatants . All solutions contained protease inhibitors . After centrifugation at 35 , 000 rpm for 65 hr at 10°C in a Beckman SW41 rotor , 12 fractions with densities ranging from 1 . 0 to 1 . 125 g/ml were collected from the top of the tube . ApoB protein from each fraction was analyzed by western blotting . For VLDL isolation , 400 μl plasma from fasted mice was put into the bottom of a 1-ml thick-walled polycarbonate tube , overlaid with 600 μl of 1 . 006 g/ml KBr solution , and centrifuged at 100 , 000 rpm at 16°C for 2 hr in a TLA 120 . 2 rotor . 300 μl from the top of the tube was collected as the VLDL fraction . For EM , 5 μl of the VLDL preparation was applied to carbon-coated copper grids and stained with 2 . 0% uranyl acetate for 15 min . Grids were visualized with a JEOL 100CX transmission electron microscope . For EM of liver samples , animals were perfused through the left ventricle of the heart with a fixative of 1 . 5% glutaraldehyde , 4% polyvinylpyrrolidone , and 0 . 05% calcium chloride in 0 . 1 M sodium cacodylate buffer pH 7 . 4 , after an initial flush with 0 . 1 M sodium cacodylate buffer , pH 7 . 4 . The imidazole-buffered osmium tetroxide procedure described by Angermuller and Fahimi ( Angermuller and Fahimi , 1982 ) was used to stain for lipids . Tissue was en block stained in aqueous uranyl acetate , dehydrated , infiltrated and embedded in LX-112 resin ( Ladd Research Industries , Burlington , VT ) . Samples were ultrathin sectioned on the Reichert Ultracut S ultramicrotome and counter stained with 0 . 8% lead citrate . Grids were examined on a JEOL JEM-1230 electron microscope ( JEOL USA , Inc . , Peabody , MA ) and photographed using the Gatan Ultrascan 1000 digital camera ( Gatan Inc . , Warrendale , PA ) . | Living cells are surrounded by a membrane that forms a barrier between the cell and its external environment . This membrane is largely made up of a variety of molecules known as lipids . The particular lipid molecules found in a cell membrane strongly influence its mobility , flexibility and other physical properties . The liver and intestine can package lipids gained from the diet into molecules called lipoproteins , which are released into the bloodstream for use by the body . An enzyme known as Lpcat3 is found in high levels in the liver and intestine and it appears to be involved in the production of lipoproteins . Altering the amount of Lpcat3 in cells can change the types of lipids found in the cell membranes , connected to the production of lipoproteins . Rong et al . studied newborn mice that were missing the Lpcat3 protein in either the liver or intestine . Mice lacking Lpcat3 in the intestine had higher levels of lipids inside their intestine cells and grew more slowly than normal mice . Mice lacking Lpcat3 in the liver also accumulated lipids in their cells and their bloodstream had lower levels of lipids that contain a molecule called arachidonic acid than normal mice . Further experiments showed that the loss of Lpcat3 reduces the ability of lipids to move within the cell membrane . The experiments show that Lpcat3 plays a key role in attaching arachidonic acid to membrane lipids to promote the release of lipoproteins into the bloodstream . Rong et al . 's findings reveal that changing the type of lipids in the cell membrane plays an important role in regulating the levels of lipids in the blood . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | Lpcat3-dependent production of arachidonoyl phospholipids is a key determinant of triglyceride secretion |
When starved , the Gram-positive bacterium Bacillus subtilis forms durable spores for survival . Sporulation initiates with an asymmetric cell division , creating a large mother cell and a small forespore . Subsequently , the mother cell membrane engulfs the forespore in a phagocytosis-like process . However , the force generation mechanism for forward membrane movement remains unknown . Here , we show that membrane migration is driven by cell wall remodeling at the leading edge of the engulfing membrane , with peptidoglycan synthesis and degradation mediated by penicillin binding proteins in the forespore and a cell wall degradation protein complex in the mother cell . We propose a simple model for engulfment in which the junction between the septum and the lateral cell wall moves around the forespore by a mechanism resembling the ‘template model’ . Hence , we establish a biophysical mechanism for the creation of a force for engulfment based on the coordination between cell wall synthesis and degradation .
To survive starvation , the Gram-positive bacterium Bacillus subtilis forms durable endospores ( Tan and Ramamurthi , 2014 ) . The initial step of sporulation is the formation of an asymmetrically positioned septum ( polar septation ) , which produces a larger mother cell and a smaller forespore ( Figure 1A ) . After division , the mother cell engulfs the forespore in a phagocytosis-like manner . Engulfment entails a dramatic reorganization of the sporangium , from two cells that lie side by side to a forespore contained within the cytoplasm of the mother cell . The internalized forespore matures and is ultimately released to the environment upon mother cell lysis . After engulfment , the forespore is surrounded by two membranes within the mother cell cytoplasm , sandwiching a thin layer of peptidoglycan ( PG ) ( Tocheva et al . , 2013 ) . While a number of molecular players for engulfment have been identified , the mechanism of force generation to push or pull the mother cell membrane around the forespore remains unknown ( Higgins and Dworkin , 2012 ) . 10 . 7554/eLife . 18657 . 003Figure 1 . Peptidoglycan ( PG ) synthesis is essential for leading-edge ( LE ) migration . ( A ) Morphological changes during spore formation . Peptidoglycan shown in grey , membrane in red . ( 1 ) Vegetative cell . ( 2 ) The first morphological step in sporulation is asymmetric cell division , producing a smaller forespore and a larger mother cell . ( 3 ) The septum curves and protrudes towards the mother cell . ( 4 ) The mother cell membrane migrates towards the forespore pole . The different modules contributing to membrane migration are shown in the inset ( see Introduction for details ) . During engulfment , the septal PG is extended around the forespore ( Tocheva et al . , 2013 ) . ( 5 ) Fully engulfed forespore surrounded by two membranes sandwiching a thin layer of PG . ( B ) Snapshots of engulfing sporangia from time-lapse movies in the absence of antibiotics , or in the presence of cephalexin or bacitracin . Cells were stained with fluorescent membrane dye FM 4–64 and imaged in medial focal plane . In the absence of antibiotics ( top ) the septum curves and grows towards the mother cell without significant forward movement of the engulfing membrane for ∼20 min . After that , the LE of the engulfing membrane starts migrating and reaches the forespore pole in ∼1 hr . When PG precursor delivery system is blocked with bacitracin ( 50 μg/ml ) : ( I ) LE migration is stopped or ( II ) engulfment proceeds asymmetrically . Similar results are obtained when cells are treated with cephalexin ( 50 μg/ml ) . However , in this case the asymmetric engulfment phenotype observed at later time points is due to rotation of the engulfment cup ( C ) rather than to asymmetric movement forward of the engulfing membrane ( D ) . ( E ) FM 4–64 average kymograph of n = 24 engulfing cells ( see Materials and methods , Appendix 1 ) . Average fluorescent intensity along forespore contour vs time in the mother-forespore reference frame as shown in top inset . All cells are aligned in time based on time 0’ ( 0 min ) . Time 0’ is assigned to the onset of curving septum ( Figure 1—figure supplement 3 ) . Bottom inset is average kymograph represented as heat map . ( F–G ) Average kymograph for cells treated with cephalexin ( n = 18 ) ( F ) or bacitracin ( n = 26 ) . ( G ) When drug was added analyzed cells had ( 55 ± 5 ) % engulfment ( red arrow ) . The percentage of engulfment is calculated as total angle of forespore covered with mother membrane divided by full angle . All cells had fully curved septum . Non-engulfed part of the forespore is represented as the black regions in kymographs . ( H ) In untreated sporangia , gap starts to close ∼20 min after onset of membrane curving . In antibiotic-treated cells gap does not close . Sample size as in ( F–G ) . Red arrow points when drug is added . Average ± SEM . Scale bar 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 00310 . 7554/eLife . 18657 . 004Figure 1—figure supplement 1 . Sporulation minimal inhibitory concentration . ( A ) Microscopy pictures of cells sporulating before antibiotic treatment ( t2 ) , or 2 hr later ( t4 ) after treatment with antibiotics blocking different steps on the PG biosynthetic pathway: synthesis of cytoplasmic PG intermediates ( D-cycloserine ) , recycling of undecaprenyl-P ( bacitracin ) , cross-linking of the glycan strands ( vancomycin ) , or PBP activity ( amoxicillin , cephalexin , cloxacillin , oxacillin and penicillin V ) . Cells were stained with Mitotracker Green ( green , membrane permeable ) and FM 4–64 ( red , membrane impermeable ) to visualize membranes . When engulfment is completed , the forespore membranes are only stained by Mitotracker green , but not by FM 4–64 ( Sharp and Pogliano , 1999 ) . ( B ) Graphs showing the percentage of cells that have undergone polar septation ( % sporangia ) and the percentage of sporangia that have completed engulfment ( % engulfed sporangia ) at different time points after sporulation induction , in cultures treated with different antibiotics that block PG synthesis . Antibiotics were added 2 hr after sporulation induction ( red arrows ) . Samples were taken every hour for 5 hr , stained with MTG and FM 4–64 and visualized under the microscope . More than 300 cells were quantified per time point and antibiotic concentration . ( C ) Table showing the Minimal Inhibitory Concentration ( MIC ) of antibiotics blocking PG synthesis during vegetative growth ( Vegetative MIC ) , and the estimated MIC during sporulation ( Sporulation MIC ) . The Sporulation MIC was defined as the concentration or concentration interval that block the formation of new polar septa , and was inferred from the graphs in B . Scale bar 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 00410 . 7554/eLife . 18657 . 005Figure 1—figure supplement 2 . Quantification of cell division events in timelapse movies . Fraction of cell division events per cell observed during the first 90 min and 150 min of imaging in timelapse movies of sporulating cultures treated with bacitracin ( 50 μg/ml ) , cephalexin ( 50 μg/ml ) , or untreated . At least 296 vegetative cells were tracked over time for every condition . The total number of division events observed after 90 min or 150 min was divided by the number of cells tracked in each case . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 00510 . 7554/eLife . 18657 . 006Figure 1—figure supplement 3 . Image analysis of non-treated cells . ( A ) Time course of septum curvature . The horizontal dashed grey line corresponds to inverse cell-wall radius ( FM 4–64 ) measured at the cell middle ( 1/R= ( 2 . 3±0 . 4 ) μm−1 , n=14 ) . ( B ) Time course of mother-cell area . ( C–D ) FM 4–64 kymographs of partially engulfed forespores ( n = 6 with ( 55 ± 5 ) % of engulfment;= 7 with ( 70 ± 5 ) % of engulfment , respectively ) . This is a control analysis of non-treated cells for the experiment when partially engulfed cells treated with drugs stop engulfment ( see Figure 1F–G ) . Average ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 006 The cellular machinery for engulfment is complex , presumably to add robustness for survival ( Figure 1A , inset ) . First , the forespore protein SpoIIQ and the mother cell protein SpoIIIAH interact in a zipper-like manner across the septum ( Blaylock et al . , 2004 ) , and mediate the fast engulfment observed in the absence of cell wall ( Broder and Pogliano , 2006; Ojkic et al . , 2014 ) . This complex is static and is proposed to act as a Brownian ratchet to prevent backwards movement of the engulfing membrane , contributing to the robustness of engulfment in intact cells ( Sun et al . , 2000; Broder and Pogliano , 2006 ) . Second , the SpoIID , SpoIIM and SpoIIP complex ( DMP ) localizes at the leading edge ( LE ) of the mother cell engulfing membrane and is essential and rate limiting for membrane migration ( Abanes-De Mello et al . , 2002; Gutierrez et al . , 2010 ) . The complex contains two enzymes that degrade PG in a processive manner: SpoIIP removes stem peptides , and SpoIID degrades the resulting denuded glycan strands ( Abanes-De Mello et al . , 2002; Chastanet and Losick , 2007; Morlot et al . , 2010; Gutierrez et al . , 2010 ) . Mutants with reduced SpoIID or SpoIIP activity or protein levels engulf asymmetrically , with the engulfing membrane migrating faster on one side of the forespore ( Abanes-De Mello et al . , 2002; Gutierrez et al . , 2010 ) . Third , blocking PG precursor synthesis with antibiotics impairs membrane migration in mutants lacking the Q-AH zipper , suggesting that PG synthesis at the LE of the engulfing membrane contributes to engulfment ( Meyer et al . , 2010; Tocheva et al . , 2013 ) . However , the mechanistic details of membrane migration and for the coordination between PG synthesis and degradation remain unclear . The biophysical principles of cell wall remodeling in Gram-positive bacteria are not well understood . In Bacillus subtilis , the cell wall is about 20–40 nm thick , and is likely organized into multiple ( 20–30 ) PG layers ( Morlot et al . , 2010; Reith and Mayer , 2011; Lee and Huang , 2013; Misra et al . , 2013; Dover et al . , 2015 ) . In contrast , cryo-electron tomography has demonstrated that a thin PG layer is present between the septal membranes throughout engulfment , appearing to form continuous attachments with the old cell wall ( Tocheva et al . , 2011 , 2013 ) . The outer cell wall of Gram-positive bacteria also contains a significant amount of teichoic acids , important for cell morphology , phosphates , and antibiotic resistance ( Grant , 1979; Brown et al . , 2013 ) but largely absent in spores ( Chin et al . , 1968; Johnstone et al . , 1982 ) . Engulfment entails extensive cell wall remodeling , with peptidoglycan precursors , newly synthesized PG and the sporulation specific PG degradation machinery localizing at the LE of the engulfing membrane ( Meyer et al . , 2010; Tocheva et al . , 2013; Abanes-De Mello et al . , 2002 ) . However , since engulfment occurs at high turgor pressure within the cramped confines of the thick outer cell wall , we expect that membrane movement is severely reduced by steric hindrance ( Lizunov and Zimmerberg , 2006 ) . Hence , we anticipate that peptidoglycan remodeling is a critical step in engulfment , which may either act as a force generator or simply create room for engulfment by the mother cell membrane . Here , we provide a biophysical mechanism for engulfment in which PG synthesis and degradation move the junction between the septal PG and the lateral cell wall around the forespore , making room for the engulfing membrane to move by entropic forces . Using antibiotics that block different steps in PG synthesis , we demonstrate that PG synthesis is essential for membrane migration in all conditions and contributes to the localization of SpoIIDMP at the LE . We also show that components of the PG biosynthetic machinery , including several penicillin binding proteins ( PBPs ) and the actin-like proteins MreB , Mbl and MreBH track the LE of the engulfing membrane when produced in the forespore , but not when produced in the mother cell . We implement a biophysical model for PG remodeling at the LE of the engulfing membrane , based on the ‘template mechanism’ of vegetative cell growth and implemented by stochastic Langevin simulations . These simulations reproduce experimentally observed engulfment dynamics , forespore morphological changes , and asymmetric engulfment when PG synthesis or degradation is perturbed . Taken together , our results suggest that engulfment entails coordination of PG synthesis and degradation between the two compartments of the sporangium , with forespore-associated PBPs synthesizing PG ahead of the LE and the mother-cell DMP complex degrading this PG to drive membrane migration .
In contrast to previous studies ( Meyer et al . , 2010 ) , we attempted to find conditions that completely blocked PG synthesis in sporulating cultures ( Figure 1—figure supplement 1 ) . To estimate the sporulation minimal inhibitory concentration ( sMIC ) of antibiotics , we monitored the percentage of cells that had undergone polar septation over time in batch cultures . Polar septation depends on PG synthesis and is easy to track visually ( Pogliano et al . , 1999 ) , which makes it a good indicator for efficient inhibition . We assayed nine antibiotics inhibiting different steps in the PG biosynthesis pathway , and found concentrations that blocked the formation of new polar septa for seven of them ( Figure 1—figure supplement 1B , C ) . In most cases , the antibiotic concentration that blocked polar septation also inhibited completion of engulfment ( Figure 1—figure supplement 1B ) . Only two drugs , fosfomycin and D-cycloserine , failed to completely block polar cell division . These drugs inhibit production of PG precursors that , during starvation conditions , might be obtained by recycling rather than de novo synthesis ( Reith and Mayer , 2011 ) , potentially from cells that lyse during sporulation , as has been observed in studies of B . subtilis cannibalism ( González-Pastor et al . , 2003; Straight and Kolter , 2009; Lamsa et al . , 2012 ) , or from cells that lyse due to antibiotic treatment ( Lamsa et al . , 2016 ) . These results demonstrate that the later stages in PG synthesis are essential for engulfment in wild type sporangia . To investigate the role played by PG synthesis , we selected two antibiotics for further characterization: cephalexin , which inhibits PBP activity , and bacitracin , which blocks cell-wall precursor delivery ( recycling of undecaprenyl phosphate ) . Using time-lapse microscopy ( see Materials and methods for details ) , we monitored membrane dynamics during engulfment in the medial focal plane using the fluorescent membrane dye FM 4–64 ( Figure 1B , Video 1 ) . In these 2–5 hour-long movies we observed occasional cell division events occurred with bacitracin ( 0 . 08 division events/cell after 90 min , compared to 0 . 28 division events/cell in untreated cultures , Figure 1—figure supplement 2 ) , indicating that PG synthesis was not completely blocked under these conditions . However , negligible cell divisions occurred with cephalexin , indicating that PG synthesis was indeed completely blocked ( Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 18657 . 007Video 1 . Timelapse microscopy of sporulating B . subtilis stained with the membrane dye FM 4–64 . The left panel shows untreated cells , the middle panel cephalexin-treated cells ( 50 μg/ml ) , and the right panel bacitracin-treated cells ( 50 μg/ml ) . Cells were imaged in agarose pads supplemented with the appropriate antibiotics ( see Materials and methods for details ) . Pictures were taken every 5 min . Total time 2 . 5 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 007 To better monitor LE dynamics we used two image analysis approaches ( see Materials and methods for details ) . First , we created kymographs along forespore membranes ( Figure 1E–G ) . The angular position of forespore pixels was calculated relative to the mother-forespore frame of reference ( Figure 1E , inset ) . All cells were aligned in time based on the onset of septum curving ( Figure 1—figure supplement 3 ) , and for a given angle , the average fluorescence of different cells was calculated and plotted over time . Second , we measured the decrease in the distance between the two LEs of the engulfing membrane in the focal plane ( the gap arc length ) , in order to directly assess movement of the LE around the forespore ( Figure 1H ) . These analyses demonstrated that in untreated sporangia ( Figure 1B , top row ) , the septum curves and the forespore grows into the mother cell without significant forward movement of the LE for ∼20 min after polar septation ( at 30°C , Figure 1H ) . Subsequently , the LE of the engulfing membrane moves towards the forespore pole and engulfment completes within ∼60 min ( Figure 1E , H ) . In sporangia treated with cephalexin ( Figure 1B , middle row I ) , the septum curves and extends towards the mother cell , but there is no forward membrane migration ( Figure 1F , H ) . Sometimes the LE retracted on one side while advancing slightly on the other ( typically occurred after 90 min of imaging; Figure 1B , middle row II ) , which appears to be the rotation of the ‘cup’ formed by the engulfing membranes relative to the lateral cell wall ( Figure 1C ) . Similar to cephalexin , in most sporangia treated with bacitracin ( Figure 1B , bottom row I ) , the forespore extended into the mother cell without significant membrane migration ( Figure 1G , H ) . However , in ∼20% of the sporangia , the engulfing membrane migrated asymmetrically , with one side moving faster than the other , although usually it failed to completely surround the forespore ( Figure 1B , bottom row II; Figure 1D ) . The continued engulfment under bacitracin treatment might be related to the fact that PG synthesis is not completely blocked in bacitracin-treated cells under time-lapse conditions ( Figure 1—figure supplement 2 ) . Taken together , these results suggest that PG synthesis is not only essential for the final stage of engulfment ( membrane fission ) in wild type cells ( Meyer et al . , 2010 ) , but also for migration of the LE of the engulfing membrane around the forespore . It has been previously shown that there is an accumulation of membrane-bound PG precursors at the LE of the engulfing membrane ( Meyer et al . , 2010 ) . Furthermore , staining with fluorescent D-amino acids has demonstrated that new PG is synthesized at or close to the LE ( Tocheva et al . , 2013 ) . To investigate if there is a concomitant accumulation of PBPs at the LE , we stained sporangia with BOCILLIN-FL , a commercially available derivative of penicillin V that has a broad affinity for multiple PBPs in B . subtilis ( Lakaye et al . , 1994; Zhao et al . , 1999; Kocaoglu et al . , 2012 ) . We observed continuous fluorescent signal around the mother cell membrane that was enriched at the LE ( Figure 2A ) . To better monitor localization of PBPs during engulfment , we plotted fluorescence intensities along the forespores for the membrane and BOCILLIN-FL fluorescent signals as a function of the engulfment stage ( Figure 2B ) . Clearly , the LE is always enriched with PBPs throughout membrane migration . 10 . 7554/eLife . 18657 . 008Figure 2 . PG synthesis at the LE of the engulfing membrane by forespore PBPs contribute to proper localization of the DMP complex . ( A ) Sporulating cells stained with a green fluorescent derivative of penicillin V ( BOCILLIN-FL ) . Bright foci are observed at the LE of the engulfing membrane . Membranes were stained with FM 4–64 ( red ) . ( B ) Average BOCILLIN-FL ( green ) and FM 4–64 ( red ) fluorescence intensities along forespore contours plotted as a function of the degree of engulfment . Cells are binned according to percentage of engulfment . BOCILLIN-FL signal is enriched at the LE throughout engulfment ( n = 125 ) . ( C ) Cell-specific localization of the peptidoglycan biosynthetic machinery . GFP tagged versions of different B . subtilis PBPs and actin-like proteins ( ALPs ) were produced from mother cell- ( MC ) or forespore- ( FS ) specific promoters . ( D ) Six different localization patterns were observed upon cell-specific localization of PBPs and ALPs . For each pair of images , left panel shows overlay of membrane and GFP fluorescence , while the right panel only shows GFP fluorescence . Pictures of representative cells displaying the different patterns are shown ( top , GFP fusion proteins transcribed from spoIIR promoter for forespore-specific expression , and from spoIID promoter for mother cell-specific expression ) . The six different patterns are depicted in the bottom cartoon and proteins assigned to each one are indicated . Membranes were stained with FM 4–64 . See Figure 2—figure supplement 1 for cropped fields of all PBPs we assayed . Transglycosylase ( TG ) , transpetidase ( TP ) , carboxipetidase ( CP ) , endopeptidase ( EP ) , actin-like protein ( ALP ) . ( E ) TIRF microscopy of forespore-produced GFP-MreB in four different forespores ( i to iv ) . In every case , the leftmost picture is an overlay of the forespore membranes ( shown in white ) and the tracks followed by individual TIRF images of GFP-MreB ( color encodes time , from blue to red ) . Sporangia are oriented with the forespores up . For the first sporangia ( i ) , snapshots from TIRF timelapse experiments taken 8 s apart are shown . Arrows indicate GFP-MreB foci and are color coded to match the trace shown in the left panel . Rightmost panel for each forespore shows a kymograph representing the fluorescence intensity along the line joining the leading edges of the engulfing membrane over time ( from top to bottom; total time 100 s ) . Average focus speed ( n = 14 ) is indicated at the bottom . Timelapse movies of the examples presented here and additional sporangia are shown in Video 2 . ( F ) Localizaiton of GFP-SpoIIP in untreated sporangia , or in sporangia treated with bacitracin ( 50 μg/ml ) or cephalexin ( 50 μg/ml ) . ( G ) Fraction of GFP-SpoIIP fluorescence at LE of the engulfing membrane . Bars represent the average and standard error of 85 untreated sporangia , 38 sporangia treated with bacitracin ( 50 μg/ml ) , and 67 sporangia treated with cephalexin ( 50 μg/ml ) . ( H ) Model for PG synthesis and degradation at the LE of the engulfing membrane . New PG is synthesized ahead of the LE of the engulfing membrane by forespore-associated PG biosynthetic machinery , and is subsequently degraded but the mother-cell DMP complex . We propose that DMP has specificity for the peptide cross-links that join the newly synthesized PG with the lateral cell wall ( orange ) , which leads to the extension of the septal PG around the forespore . Scale bars 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 00810 . 7554/eLife . 18657 . 009Figure 2—figure supplement 1 . Cell-specific localization of PBPs and actin-like proteins . GFP was fused to the N-terminus of PBPs and actin-like proteins . The fusion proteins where produced in the forespore or in the mother cell after polar septation by placing the fusion genes under the control of either the forespore specific promoters ( PspoIIQ or PspoIIR , for stronger or weaker expression , respectively ) or the mother-cell specific promoter PspoIID . With the exception of GFP-PbpE , all the fusions localize to the membrane . GFP-MreB and GFP-Mbl associate to the membrane when produced in the forespore , while GFP-MreBH only shows a week membrane association . When produced in the mother cell , GFP-Mbl and GFP-MreBH remain mostly cytoplasmic , and GFP-MreBH forms some foci distributed around the membrane . Membranes were stained with FM 4–64 . The different localization patterns are summarized in Figure 2D . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 00910 . 7554/eLife . 18657 . 010Figure 2—figure supplement 2 . Localization of forespore GFP-PonA and GFP-PbpA in different mutant backgrounds . GFP-PonA and GFP-PbpA were produced specifically in the forespore after polar septation by placing the fusion genes under the control of PspoIIR . The localization of both proteins was determined in wild-type background and in different mutants lacking specific sporulation proteins . GFP-PonA and GFP-PbpA still track the leading edge of the engulfing membrane or localize to the interception between the septal peptidoglycan and the lateral cell wall in all the mutant backgrounds tested . Membranes were stained with FM 4–64 . Scale bar , 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01010 . 7554/eLife . 18657 . 011Figure 2—figure supplement 3 . SpoIIDMP localization upon treatment with different antibiotics blocking PG synthesis . ( A ) Localizaiton of GFP-SpoIIP in untreated sporangia , or in sporangia treated with bacitracin ( 50 μg/ml ) , amoxicillin ( 500 μg/ml ) , cephalexin ( 50 μg/ml ) , cloxacillin ( 500 μg/ml ) , oxacillin ( 50 μg/ml ) , or penicillin V ( 500 μg/ml ) . Membranes were stained with FM 4–64 . ( B ) Fraction of GFP-SpoIIP fluorescence at LE of the engulfing membrane . Bars represent the average and standard error of 85 untreated sporangia , 38 sporangia treated with bacitracin ( 50 μg/ml ) , 37 treated with amoxicillin ( 500 μg/ml ) , 67 treated with cephalexin ( 50 μg/ml ) , 43 treated with cloxacillin ( 500 μg/ml ) , 36 treated with oxacillin ( 50 μg/ml ) , and 39 treated with penicillin V ( 500 μg/ml ) . ( C , D ) Localization of GFP-SpoIID ( C ) and GFP-SpoIIM ( D ) in untreated sporangia or in sporangia treated with bacitracin ( μg/ml ) or cephalexin ( 50 μg/ml ) . Membranes were stained with FM 4–64 . ( E , F ) Fraction of GFP-SpoIID ( E ) or GFP-SpoIIM ( F ) at LE . Bars represent the average and standard error of 106 untreated sporangia , 110 bacitracin-treated sporangia and 126 cephalexin-treated sporangia for GFP-SpoIID ( E ) , and 86 untreated , 79 bacitracin-treated and 63 cephalexin-treated sporangia for GFP-SpoIIM ( F ) . Scale bars , 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 011 One possible explanation for the requirement of PG synthesis for engulfment is that PG polymerization by PBPs associated with the LE of the engulfing membrane creates force to pull the engulfing membrane around the forespore . If so , we would expect the PBPs to be located in the mother cell membrane as they polymerize PG . To test this possibility , we assessed the localization of components of the PG biosynthetic machinery in the mother cell or forespore by producing GFP-tagged fusion proteins from promoters that are only active in the mother cell ( PspoIID ) or in the forespore ( the stronger PspoIIQ and the weaker PspoIIR ) after polar septation ( Figure 2C , D , Figure 2—figure supplement 1 ) . One prior study tested the localisation of several PBPs during sporulation ( Scheffers , 2005 ) , but most of them were produced before polar separation and it was not possible to determine which cell compartment they were in . We successfully determined the cell-specific localization of 16 proteins involved in PG synthesis ( Figure 2—figure supplement 1 ) , including all class A and four class B high-molecular-weight ( HMW ) PBPs , five low-molecular-weight ( LMW ) PBPs ( four endopeptidases and one carboxipeptidase ) , and all three MreB paralogues ( actin-like proteins , ALPs ) . Surprisingly , only PonA ( PBP1a/b ) showed a weak enrichment at the LE of the engulfing membrane when produced in the mother cell ( Figure 2D ) . However , ten PBPs , including PonA and all the class B HMW PBPs and LMW PBPs tested , and all the MreB paralogues were able to track the LE only when produced in the forespore ( Figure 2D , Figure 2—figure supplement 1 ) . To follow the dynamics of the forespore PG biosynthetic machinery at the LE , we monitored the movement of GFP-MreB using TIRF microscopy ( Garner et al . , 2011; Domínguez-Escobar et al . , 2011 ) . Forespore GFP-MreB foci rotate around the forespore , coincident with the leading edge of the engulfing membrane , with speeds consistent with those previously reported ( Figure 2E , Video 2 ) . 10 . 7554/eLife . 18657 . 012Video 2 . Circumferential movement of forespore GFP-MreB . The movie shows the movement forespore GFP-MreB in eight different sporangia , determined by TIRF microscopy . A static membrane picture is shown to the left , and the TIRF microscopy of the corresponding GFP-MreB is shown immediately to the right . TIRF pictures were taken every 4 s , and the total duration of the movie is 100 s . The first four sporangia correspond to the examples ( i ) to ( iv ) shown in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 012 It is unclear how the PBPs recognize the LE , as localization of forespore produced GFP-PonA and GFP-PbpA did not depend on candidate proteins SpoIIB , SpoIID , SpoIIM , SpoIIP , SpoIIQ , SpoIIIAH , SpoIVFAB , or GerM ( Aung et al . , 2007; Abanes-De Mello et al . , 2002; Chastanet and Losick , 2007; Blaylock et al . , 2004; Rodrigues et al . , 2016 ) ( Figure 2—figure supplement 2 ) . However , these results indicate that the forespore plays a critical role in PG synthesis , and point to an engulfment mechanism that does not depend on pulling the engulfing membrane by mother cell-directed peptidoglycan synthesis . The observation that multiple PBPs can track the LE of the engulfing membrane from the forespore opens the possibility that PG synthesis happens ahead of the LE , preceding PG degradation by the mother cell DMP complex . In this context , PG synthesis might be required for proper activity and/or localization of the DMP complex , which is the only other essential engulfment module described so far . The DMP complex localizes at the LE throughout engulfment ( Gutierrez et al . , 2010 ) . To determine if PG synthesis is required for proper localization of DMP , we studied the localization of a GFP-SpoIIP fusion protein when PG synthesis was inhibited by different antibiotics ( Figure 2F , G ) . GFP-SpoIIP shows a well-defined localization at the LE , with ∼70% of the total GFP fluorescence at LE in native conditions ( Figure 2F , G ) . However , when PG biosynthesis is inhibited , there is a delocalization of GFP-SpoIIP , which is almost total in cells treated with bacitracin and partial when antibiotics targeting later stages of PG synthesis are used ( Figure 2F , G; Figure 2—figure supplement 3 ) . Equivalent results were obtained with GFP-SpoIID and GFP-SpoIIM fusions ( Figure 2—figure supplement 3 ) . These results are consistent with a model in which PG is synthesized ahead of the LE by forespore-associated PBPs specify the site of PG degradation by the DMP complex ( Figure 2H ) . Our data indicate that engulfment proceeds through coordinated PG synthesis and degradation at the LE . To explain how this coordination leads to engulfment , we propose a minimal biophysical mechanism based on the ‘template mechanism’ of vegetative cell growth assuming that glycans are oriented perpendicular to the long axis of the cell ( Figure 3A ) ( Koch and Doyle , 1985; Höltje , 1998; Domínguez-Escobar et al . , 2011; Garner et al . , 2011; Beeby et al . , 2013; Dover et al . , 2015 ) , without requiring any further assumptions about the outer cell wall structure of Gram-positive bacteria , which is still unclear ( Hayhurst et al . , 2008; Beeby et al . , 2013; Dover et al . , 2015 ) . In this mechanism , a new glycan strand is inserted using both the septal glycan and leading forespore-proximal glycan strand of the lateral wall as template strands to which the new PG strand is cross linked . Subsequently , peptide cross-links between the two template strands are removed from the mother-cell proximal side by the DMP complex . Specifically , in this complex SpoIIP has well documented endopeptidase activity ( Morlot et al . , 2010 ) . Note , similar ‘make-before-break’ mechanisms were proposed to allow vegetative cell wall growth without reducing cell wall integrity ( Koch and Doyle , 1985; Höltje , 1998 ) . A more detailed mechanism that requires the insertion of multiple new glycan strands to account for glycan removal by SpoIID is shown in Figure 3—figure supplement 1 . In either model , synthesis of new PG at the LE likely occurs before degradation , thereby naturally preventing cell lysis during engulfment . 10 . 7554/eLife . 18657 . 013Figure 3 . Template model for leading edge ( LE ) movement . ( A ) Cell cross-section with glycan strands in the plane perpendicular to the long axis of the cell . One strand from old cell wall ( blue ) and one strand from newly synthesized germ-cell wall ( green ) are used as a template for new glycan insertion . Coordination between glycan insertion ( orange arrow ) and peptide cross-link degradation ( black cross ) drives LE forward . ( B ) 3D model of stochastic glycan insertion by insertion-degradation complex ( IDC ) with transpeptidase and transglycosylase activity . Probability of IDC to start inserting new glycan from old glycan end and repair end defect is prep . ( C ) New inserted glycan shown in dark green . Probability of IDC to continue glycan insertion when it encounters gap in old cell wall is probability of processivity ppro . ( Inset ) Horizontal ( between old and new glycan strands ) and vertical ( between new glycan strands ) peptide links are shown in red . In our coarse-grained model glycans are simulated as semi-flexible filaments consisting of beads ( green ) connected with springs ( green ) . Peptides are simulated as springs ( red ) connecting neighboring glycan beads . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01310 . 7554/eLife . 18657 . 014Figure 3—figure supplement 1 . Extended models that account for glycan-strand degradation . Here we further explore possible mechanisms considering the fact that SpoIID protein of DMP complex shows transglycosylase activity ( Morlot et al . , 2010 ) . ( A ) In the two-for-one mechanism two new glycan strands are added and the newly inserted glycan strand at the LE is degraded ( Höltje , 1998 ) . Similarly , the three-for-one mechanism would also work ( Scheffers and Pinho , 2005 ) . ( B ) One new glycan strand is added and the innermost cell-wall glycan of the thick old cell wall is degraded . Similar to images of electron microscopy ( Tocheva et al . , 2013 ) . However , in these models cell-wall degradation without high level of coordination could affect cell-wall integrity and induce cell lysis . All these models share the ’make-before-break’ strategy promoting robustness of the remodeling process ( Koch and Doyle , 1985 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 014 The coordination between PG insertion from the forespore and removal by DMP in the mother cell could lead to movement of the junction between the septal peptidoglycan and the lateral peptidoglycan around the forespore to mediate successful engulfment . Based on this proposed mechanism , we created a model whereby insertion and degradation happens , for simplicity , simultaneously by an insertion-degradation complex ( IDC ) , also reflecting the high degree of coordination suggested by the template mechanism . In this model IDC recognizes the leading edge and inserts glycan polymers perpendicular to the long axis of the cell ( Figure 3B ) . Additionally , the model proposes that IDC can recognize glycan ends and initiate glycan polymerization from the end defect with probability of repair prep . During glycan insertion , when an IDC encounters a gap in the outer cell wall strands , it continues polymerization with probability of processivity ppro ( Figure 3C ) . A systematic exploration of the above model parameters showed that intact spores form for prep and ppro>> 0 . 8 with a marginal dependence on the number of IDCs ( Figure 4G , Figure 4—figure supplement 1 ) . However , to compare the model with microscopy data we require a 3D dynamic implementation of this model that reflects the stochasticity of underlying molecular events . 10 . 7554/eLife . 18657 . 015Figure 4 . Template model reproduces experimentally observed phenotypes . ( A ) Effective spring constants in our model represent coarse-grained PG network . Here the angle between neighboring stem peptides that belong to a single glycan is assumed to be 90° . Therefore , every other stem peptide is in plane with glycan sheet ( Nguyen et al . , 2015 , Huang et al . , 2008 ) . The role of effective glycan persistence length on engulfment is negligible ( see Figure 4—figure supplement 3 ) . ( B ) Simulations for different values of effective peptide kpep and glycan kgly spring constants are compared with experimentally measured forespore surface area , volume and engulfment using mutual χ2 statistics ( Equation 2 ) . Arrows point to effective literature kpep and kgly ( Nguyen et al . , 2015 ) . Dark blue region corresponds to simulation parameters that best fit experimental data ( Figure 4—figure supplement 4 , Video 3 ) . For large enough kgly > 200 pN/nm mutual χ2 is almost independent of kgly . ( C ) Snapshots of WT simulations for parameters ( kgly = 200 pN/nm , kpep = 25 pN/nm , NIDC = 5 ) marked with ’×’ in panel ( B ) ( Video 2 ) . The thick septum is treated as outer cell wall , and is assumed degraded once IDCs move along . ( D–E ) Time traces of experimentally measured engulfment , forespore surface area and forespore volume ( green ) in comparison with results from a single simulation ( orange ) . Parameters used in simulation are marked with ’×’ in panel ( B ) . For all other parameters see Appendix 2 , Appendix-table 1 . ( F ) Snapshots of fully engulfed forespores for various peptidoglycan elastic constants . ( G ) For various values of independent parameters prep and ppro roughness of the LE is calculated at the end of stochastic simulations ( see Figure 4—figure supplement 1 , and Video 4 ) . Here 0 roughness correspond to perfectly symmetric LE; for high enough prep=ppro > 0 . 8 LE forms symmetric profiles . ( H ) Simulation for asymmetric engulfment is obtained for same parameter as WT except prep=ppro = 0 . 7 ( marked with ’×’ in panel ( G ) ) . Average ± SD . Scale bars 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01510 . 7554/eLife . 18657 . 016Figure 4—figure supplement 1 . Simulation of the stochastic model of insertion at the leading edge ( LE ) . ( A–D ) Stochastic insertion at the LE of discretized cell circumference with 1570 segments . The details are explained in the Materials and Methodes SI section ( 2 . 1 ) . Simulations are run until the LE reaches 500 glycans in height . For obtained LE profiles roughness and their widths are calculated . For each set of independent parameters prep , ppro and NIDC we run 100 simulations and plot the average roughness and width . Parameters prep and ppro are varied in steps of 0 . 1 . ( A , C ) For NIDC = 10 smooth LEs are obtained for prep and ppro> 0 . 80 . For such parameters changing NIDC by an order of magnitude marginally affects LE width while keeping LE roughness within 10% . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01610 . 7554/eLife . 18657 . 017Figure 4—figure supplement 2 . In simulations majority of peptide extensions are in the linear elastic regime . ( A ) Histogram of all peptide link lengths during one engulfment ( kpep = 25 pN/nm , kgly = 200 pN/nm , Δp = 86 . 31 kPa ) . Black arrow points to the linear extension regime ( i . e . where each peptide is extended <1 nm or <50% of its equilibrium length of 2 nm ) ( Nguyen et al . , 2015 ) . ( B ) Percentage of peptide links in simulations that are extended in linear regime as a function of time during the process of engulfment . Dashed vertical line is same as in Figure 4D , E . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01710 . 7554/eLife . 18657 . 018Figure 4—figure supplement 3 . Engulfment is unaffected by glycan persistence length . ( A ) χ2 ( defined in Materials and methods ) is used to quantify the impact of effective glycan persistence length ( lp ) on engulfment dynamics . In weakly crosslinked bundles lp=nlp0 , where n is the number of glycans in the bundle and lp0 is the persistence length of a single glycan; in strongly cross-linked bundles lp=n2lp0 ( Claessens et al . , 2006; Piechocka et al . , 2010 ) . Since our simulated filaments represent bundles of seven glycans ( Figure 4B ) , the effective persistence length can reach ∼2 μm ( lp0 = 40 nm ) . ( B–C ) Engulfment , forespore surface area and forespore volume are not affected even for high values of effective glycan persistence length ( lp=4μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01810 . 7554/eLife . 18657 . 019Figure 4—figure supplement 4 . Simulations with different peptidoglycan ( PG ) elastic constants . ( A–C ) Simulation snapshots for three different sets of PG elastic constants marked with ’×’ in panel B ( A: kpep= kgly = 50 pN/nm; B: kpep = 25 pN/nm , kgly = 200 pN/nm C; kpep = 25 pN/nm , kgly = 5 570 pN/nm ) . Elastic constants in C are obtained from molecular dynamic simulations ( Nguyen et al . , 2015 ) . ΔT = 0 . 28 hr; scale bar 1 μm . ( D ) Same as Figure 4B , repeated here for clarity . ( E ) Relative forespore curvature at the end of engulfment where κ0 is the curvature of spherical cap . At the end of engulfment curvature was experimentally measured with σ ( κ ) /κ∼0 . 15 , where σ ( κ ) is the standard deviation ( see Figure 1—figure supplement 3A ) . Therefore , curvatures in , B , and C are within the experimentally measured standard deviation . ( F ) Snapshots of fully engulfed forespores for various PG elastic constants . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 01910 . 7554/eLife . 18657 . 020Figure 4—figure supplement 5 . Simulations with decoupled synthesis and degradation . ( A ) Simulation snapshots for different values of time delay τdelay . Newly inserted glycans are separated from the old cell wall by cutting connecting peptides with typical τdelay . Double arrow shows distance between synthesis and membrane leading edge . ( B ) Euclidian distance between insertion and degradation ( ID separation ) vs time for different values of τdelay . Average over five insertion complexes is plotted vs time . ( C ) Exploration of delay model when degradation erroneously cuts vertical peptide bonds with probability ppcut . ( D ) For relatively small ppcut= 0 . 1 , an irregular peptidoglycan meshwork is formed . ( E–F ) Exploration of role of random peptide degradation when synthesis is stopped . ( E ) Simulation snapshots for various random peptide degradation rates prpep = 2 . 2 , 22 , and 33 min−1 . ( F ) Forespore volume vs time for different peptide degradation rates after synthesis is stopped . Scale bars 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 02010 . 7554/eLife . 18657 . 021Video 3 . Simulations of WT ( left ) and asymmetric engulfment ( right ) . Parameters are the same ( kpep = 25 pN/nm , kgly = 200 pN/nm , NIDC = 5 ) except for WT engulfment prep =ppro = 1 and for asymmetric engulfment prep=ppro = 0 . 7 . For full exploration of stochastic insertion parameters see Video 4 and Figure 4—figure supplement 1 . Front opening of the forespore is not shown for clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 02110 . 7554/eLife . 18657 . 022Video 4 . Simulations for different values of elastic peptidoglycan ( PG ) parameters kpep and kgly . PG spring constants drastically affect forespore morphologies . By decreasing kpep forespores elongate , while by increasing kpep forespores shrink , as measured along the long axis of the cell . Changing kgly has only minor effects on volume and surface area . The main effect is on forespore curvature ( see Figure 4—figure supplement 4 ) : high kgly increases the curvature of forespore ends ( making them more pointy ) , while low kgly decreases the curvature of the forespore ends . Septum is not shown for clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 022 To simulate stochastic insertion at the leading edge we used Langevin dynamics of a coarse-grained PG meshwork ( see Materials and methods ) . Briefly , glycan strands are modeled as semi-flexible filaments consisting of beads connected with springs , while peptide bridges are modeled as springs connecting glycan beads ( Figure 3C ) ( Laporte et al . , 2012; Tang et al . , 2014; Huang et al . , 2008 ) . Typical length of inserted glycan polymer is ∼1 μm ( ∼1/3 cell circumference ) ( Hayhurst et al . , 2008 ) and in our model the peptide bridges between newly inserted glycan strands are in a relaxed state . Glycan beads experience forces due to glycan elastic springs ( kgly ) , glycan persistence length ( lp ) , elastic peptide links ( kpep ) , stochastic thermal fluctuations , and pressure difference ( Δp ) between forespore and mother cell ( see Equation 1 and Appendix 2 ) . Glycan strands in the PG layer are connected with neighboring glycans by stem peptides ( Figure 4A ) . In our model , the angle between neighboring stem peptides that belong to the same glycan strand is assumed to be 90° ( Nguyen et al . , 2015; Huang et al . , 2008 ) . Therefore , every other stem peptide is in plane with the glycan sheet . In our model Δp originates from the packing of the B . subtilis chromosome ( ∼4 . 2 Mbp ) in the small forespore compartment ( Errington , 1993; Perez et al . , 2000; Bath et al . , 2000; Yen Shin et al . , 2015 ) . To systematically explore the peptidoglycan parameters , we compared our simulations with actual changes in forespore volume , forespore surface area , and percentage of engulfment extracted from time-lapse movies , using χ2 fitting ( Figure 4B , Equation 2 , Materials and methods ) . Parameters that best fit experimental measurements belong to dark blue region in agreement with molecular dynamic simulations ( Nguyen et al . , 2015 ) . For a single peptide bond , the linear elasticity regime is valid for extensions that are less than 1 nm ( Nguyen et al . , 2015 ) and this elastic regime is maintained in the regions with low χ2 ( Figure 4—figure supplement 2 ) . For large enough glycan stiffness ( kgly>> 300 pN/nm ) χ2 becomes independent of kgly ( Figure 4B ) . A typical simulation shown in Figure 4C matches experimental measurements of time-dependent engulfment , volume , and surface area ( Figure 4D , E ) . PG spring constants drastically affect forespore morphologies . By decreasing kpep forespores elongate , while by increasing kpep forespores shrink , as measured along the long axis of the cell . Changing kgly has only minor effects on volume and surface area . However , the main effect is on forespore curvature ( see Figure 4—figure supplement 4 ) : high kgly increases the curvature of forespore ends ( making them more pointy ) , while low kgly decreases the curvature of the forespore ends . Our simulations successfully reproduce asymmetric engulfment ( Figure 4F , G; Video 5 ) . For prep and ppro⩽0 . 8 we obtained asymmetric engulfment that reproduces the phenotypes observed when PG synthesis or degradation is partially blocked . When defects in the peptidoglycan meshwork are not repaired , different parts of the leading edge extend in an uncoordinated manner , producing asymmetric engulfment . 10 . 7554/eLife . 18657 . 023Video 5 . Simulations for different values of stochastic parameters prep and ppro . Decreasing prep and ppro below 0 . 8 results in asymmetric engulfment . For full exploration of stochastic insertion parameter see Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 023 Since our simulations correctly reproduced engulfment dynamics we used simulation parameters to estimate glycan insertion velocities VIDC of IDC ( see Appendix 2 ) . Using this method we estimated a lower bound on product NIDC⋅VIDC ∼ 110 nm/s , where NIDC is the number of insertion complexes . Similarly , by estimating the total amount of newly inserted material in the forespore within ∼0 . 8 hr without any pausing we obtain NIDC⋅VIDC ∼ 117 nm/s . For circumferentially processive PBPs ( PbpA and PbpH ) , the absolute velocity measured using TIRF microscopy is ∼20–40 nm/s during vegetative cell growth ( Domínguez-Escobar et al . , 2011; Garner et al . , 2011 ) , which is in agreement with the speed of forespore GFP-MreB determined from our TIRF experiments ( ( 28 ± 8 ) nm/s , n = 14; Figure 2E ) . Using this estimate for VIDC , we obtain a lower bound 3–6 on the number of active , highly processive PBP molecules . However , the actual number of proteins could be higher for other nonprocessive PBPs ( Domínguez-Escobar et al . , 2011; Garner et al . , 2011 ) .
The results presented here suggest that engulfment involves coordinated PG synthesis and degradation processes that are segregated between different cell types: first , PG is synthesized in front of the LE of the engulfing membrane by a forespore-associated PG biosynthetic machinery that rotates following the LE of the engulfing membrane . Then this new PG is targeted for degradation by the mother cell-associated PG degradation machinery comprised of the DMP complex ( Figure 2H ) . The delocalization of DMP when PG synthesis is inhibited with antibiotics ( Figure 2 , Figure 2—figure supplement 3 ) indicates that the DMP either forms an actual complex with the PG biosynthetic machinery across the septal PG ( to form a single insertion degradation complex ( IDC ) , as shown in Figure 3 ) or that DMP targets the new PG synthesized at the LE of the engulfing membrane . In the latter , DMP might specifically target the cross-links that attach the old lateral cell wall to the new PG synthesized at the LE of the engulfing membrane ( Figure 2H , orange ) . Since those cross-links join old , modified PG from the lateral cell wall to newly synthesized PG at the LE , those peptide bridges might have a unique chemical composition or structural arrangement that could be specifically recognized by DMP . Hence , either approach provides a safety mechanism during engulfment , since it would prevent DMP from degrading the old PG of the lateral cell wall , which could lead to cell lysis . We have conceptualized these results in a biophysical model in which a PG insertion-degradation complex ( IDC ) , representing PBPs for PG synthesis and DMP proteins for PG degradation , catalyzes PG remodeling at the LE of the engulfing membrane . Specifically , we propose that new glycan strands are inserted ahead of the LE of the engulfing membrane and PG is degraded on the mother cell proximal side to create space for forward movement of the LE ( Figure 3 ) . This is similar to the ‘make-before-break’ model of vegetative cell-wall growth , which postulates that the vegetative cell wall is elongated by inserting new PG strands prior to degrading old strands ( Koch and Doyle , 1985 ) ( although bacteria can also make a de novo cell wall ( Ranjit and Young , 2013 , Kawai et al . , 2014 ) . The make-before-break mechanism also accounts for the directional movement of the LE towards the forespore pole , since the substrate for DMP is new PG synthesized by forespore PBPs , which is always ahead of the LE of the engulfing membrane . Using Langevin simulations we successfully reproduced the dynamics of engulfment , forespore volume , and surface area . Additionally , our model correctly reproduced asymmetric engulfment observed with reduced IDC activity , and we estimated that with only a handful of highly processive PBP molecules are necessary to reproduce the observed LE dynamics . A more general model without strong coupling between the PG biosynthetic and PG degradation machineries also leads to successful engulfment ( Appendix 2 , Figure 4—figure supplement 5 , Video 6 ) . However , DMP has to be guided to degrade only the peptide cross-links between old and new glycan strands , and should also prevent detachment of the septal peptidoglycan from the old cell wall . 10 . 7554/eLife . 18657 . 024Video 6 . Simulations with decoupled synthesis and degradation . New glycans are released from the old cell wall with typical delay time τdelay . Simulations for four different values of τdelay= 0 , 0 . 9 , 9 , and 18 min ( from left to right ) . For longer τdelay the larger is separation between synthesis and membrane leading edge that is shown as red cylinder . DOI: http://dx . doi . org/10 . 7554/eLife . 18657 . 024 Since our simple mechanism in Figure 3A entails hydrolysis of certain peptide bonds but no glycan degradation , we explored additional mechanisms since the SpoIID protein of the DMP complex shows transglycosylase activity ( Morlot et al . , 2010 ) . First , it is possible that engulfment entails a two-for-one mechanism , with two new glycan strands are added and the newly inserted glycan strand at the LE is degraded ( Höltje , 1998 ) ( Figure 3—figure supplement 1A ) . Similarly , the three-for-one mechanism would also work ( Scheffers and Pinho , 2005 ) . Second , one new glycan strand might be added and the inner most cell-wall glycan of the thick , lateral cell wall degraded ( Figure 3—figure supplement 1B ) . This would make the lateral cell wall thinner as the engulfing membrane moves forward ( Tocheva et al . , 2013 ) . Finally , it is possible that insertion and degradation are not intimately coupled , and that there is simply a broad region in which PG is inserted ahead of the engulfing membrane , to create multiple links between the septal PG and the lateral cell wall ( as shown in Figure 2H ) , and that the DMP complex has a preference for newly synthesized PG . All of these models require the spatial coordination between cell wall degradation and synthesis to avoid compromising cell wall integrity and inducing cell lysis , and all share a common ‘make-before-break’ strategy to promote robustness of the otherwise risky PG remodeling process ( Koch and Doyle , 1985 ) . In order to waste as little energy as possible , a more stringent ‘make-just-before-break’ strategy may even apply , motivating more intimate coupling between the PG biosynthetic and degradation machineries . Our simple biophysical mechanism postulates that engulfment does not rely on pulling or pushing forces for membrane migration . Instead , cell wall remodeling makes room for the mother cell membrane to expand around the forespore by entropic forces . During engulfment the mother-cell surface area increases by ∼2 μm2 ( ∼25% , see Figure 1—figure supplement 3 ) , and this excess of membrane could simply be accommodated around the forespore by remodeling the PG at the LE . However , our model does not include all potential contributors to engulfment . For instance , the SpoIIQ-AH zipper , which is dispensable for engulfment in native conditions ( Broder and Pogliano , 2006 ) , might prevent membrane backward movement , and might also help localize the IDC components toward the LE . Interestingly , SpoIIQ-AH interaction is essential for engulfment in Clostridium difficile where the SpoIIQ ortholog posseses endopeptidase activity ( Crawshaw et al . , 2014; Serrano et al . , 2016; Fimlaid et al . , 2015 ) . The model also does not consider the impact of the tethering of the LE of the engulfing membrane to the forespore via interactions between the mother cell membrane anchored DMP complex at the LE and forespore synthesized PG . Future experiments and modeling should address the role of these and other potential contributors to LE migration , which will allow us to refine our biophysical model and obtain a comprehensive view of membrane dynamics during engulfment . Furthermore , understanding the cooperation between PBPs and DMP will provide valuable clues about the structure of the cell wall in Gram-positive bacteria .
All the strains used in this study are derivatives of B . subtilis PY79 . Complete lists of strains , plasmids , and oligonucleotides see Appendix 3 . Detailed descriptions of plasmid construction are provided in Supplementary file 1 . For each experiment we had at least two biological replicas , and each one contains at least three technical replicas . Averages of individual cells , but not the averages of different replicas are reported . Sporulation was induced by resuspension ( Sterlini and Mandelstam , 1969 ) , except that the bacteria were grown in 25% LB prior to resuspension , rather than CH medium . Cultures were grown at 37°C for batch culture experiments , and at 30°C for timelapse experiments . Cells were visualized on an Applied Precision DV Elite optical sectioning microscope equipped with a Photometrics CoolSNAP-HQ2 camera and deconvolved using SoftWoRx v5 . 5 . 1 ( Applied Precision ) . When appropriate , membranes were stained with 0 . 5 μg/ml FM 4–64 ( Life Technologies , Waltham , Massachusetts ) or 1 μg/ml Mitotracker green ( Life Technologies ) . Cells were transferred to 1 . 2% agarose pads for imaging . The median focal plane is shown . Sporulation was induced at 30°C . 1 . 5 hr after sporulation induction , 0 . 5 μg/ml FM 4–64 was added to the culture and incubation continued for another 1 . 5 hr . Seven μl samples were taken 3 hr after resuspension and transferred to agarose pads prepared as follows: 2/3 vol of supernatant from the sporulation culture; 1/3 vol 3 . 6% agarose in fresh A+B sporulation medium; 0 . 17 μg/ml FM 4–64 . When appropriated , cephalexin ( 50 μg/ml ) or bacitracin ( 50 μg/ml ) was added to the pad . Pads were partially dried , covered with a glass slide and sealed with petroleum jelly to avoid dehydration during timelapse imaging . Petroleum jelly is not toxic and cannot be metabolized by B . subtilis , which poses an advantage over other commonly used sealing compounds , such as glycerol , which can be used as a carbon source and inhibit the initiation of sporulation . Pictures were taken in an environmental chamber at 30°C every 5 min for 5 hr . Excitation/emission filters were TRITC/CY5 . Excitation light transmission was set to 5% to minimize phototoxicity . Exposure time was 0 . 1 s . MreB tracking experiments were performed using the strain JLG2411 , which produced GFP-MreB in the forespore after polar septation from spoIIQ promoter . Sporulation and agarose pads were done as described in Timelapse fluorescent microscopy , except that FM 4–64 was only added to the agarose pads and not to the sporulating cultures . A static membrane picture was taken at the beginning of the experiment , and was used as a reference to determine the position of the GFP-MreB foci . GFP-MreB motion at the cell surface was determined by TIRF microscopy ( Garner et al . , 2011; Domínguez-Escobar et al . , 2011 ) , taking pictures every 4 s for 100 s . Imaging was performed at 30°C . We used two different microscopes to perform TIRF microscopy: ( i ) An Applied Precision Spectris optical sectioning microscope system equipped with an Olympus IX70 microscope , a Photometrics CoolSNAP HQ digital camera and a 488 nm argon laser . To perform TIRF in this microscope , we used an Olympus 1003 1 . 65 Apo objective , immersion oil n = 1 . 78 ( Cargille Laboratories ) , and sapphire coverslips ( Olympus ) . Laser power was set to 15% , and exposure time was 200 ms . ( ii ) An Applied Precision OMX Structured Illumination microscopy equipped with a Ring-TIRF system and a UApoN 1 . 49NA objective , immersion oil n = 1 . 518 . Exposure time was 150 ms . Images were analyzed using the ImageJ-based FIJI package . Sporangia were aligned vertically using the rotation function in FIJI . GFP-MreB foci were tracked using the TrackMate pluging ( Tinevez et al . , 2016 ) , using the LoG detector , estimated blob diameter of 300 nm , simple LAP tracked and linking max distance of 300 nm . Only tracks that contained more than four points were used to determine the MreB foci speed . We used the semi-automated active contour software JFilament available as ImageJ plugin to extract fluorescently labeled membrane position over time ( Smith et al . , 2010 ) . Membrane position obtained from the medial focal plane is used in custom built Mathematica software to calculate 3D volume and surface area by assuming rotational symmetry around the axis connecting the center of masses of mother cell and forespore . For available code and example see Supplementary file 2 . Kymographs as in Figure 1E were created by collecting intensities along the forespore contours . Subsequently , pixel angles were determined using pixel position relative to the mother-forespore frame as defined in inset of Figure 1E . Forespore fluorescent intensities along angles are normalized and interpolated using third-order polynomials . For a given angle the population intensity average of different cells is calculated and plotted over time . Time 0’ is the onset of septum curving . Antibiotics were added 2 hr after resuspension , and samples were taken one hour later for imaging . Exposure times and image adjustments were kept constant throughout the experiment . To determine the fraction of GFP signal at the LE , GFP pixel intensities of seven optical sections covering a total thickness of 0 . 9 μm were summed . GFP intensities at the LE ( ILE ) and in the rest of the mother cell ( IMC ) were determined separately by drawing polygons encompassing the LE or the MC . After subtraction of the average background intensity , the fraction of GFP fluorescence at LE ( ILEILE+IMC ) was determined for each sporangium . The Langevin dynamic equation of the ith bead at position 𝐫i is given by: ( 1 ) ζid𝐫idt=𝐅ispr+𝐅ibend+𝐅ipep+𝐅istoch+𝐅iΔp+𝐅iwall , where the left-hand side depends on the drag coefficient ζi≈4πηmedl0 ( Howard , 2001 ) , with ηmed is the medium viscosity and l0 equilibrium distance between neighbouring beads ( see Appendix 1 ) . On the right-hand side of Equation 5 we have contributions of glycan elastic spring , glycan bending , peptide elastic links , stochastic thermal fluctuations , pressure difference Δp between forespore and mother , and excluded volume from the old cell wall , respectively . To compare simulations with experiments we measured forespore volume ( Vi ) , forespore surface area ( Si ) and engulfment ( Ei ) and constructed a quality-of-fit function as: ( 2 ) χ2=∑i[ ( Viexp-Visim ) 2σ2 ( Viexp ) + ( Siexp-Sisim ) 2σ2 ( Siexp ) + ( Eiexp-Eisim ) 2σ2 ( Eiexp ) ] , where index i corresponds to ith time point , and σ is the standard deviation ( Spitzer et al . , 2006 ) . | Some bacteria , such as Bacillus subtilis , form spores when starved of food , which enables them to lie dormant for years and wait for conditions to improve . To make a spore , the bacterial cell divides to make a larger mother cell and a smaller forespore cell . Then the membrane that surrounds the mother cell moves to surround the forespore and engulf it . For this process to take place , a rigid mesh-like layer called the cell wall , which lies outside the cell membrane , needs to be remodelled . This happens once a partition in the cell wall , called a septum , has formed , separating mother and daughter cells . However , it is not clear how the mother cell can generate the physical force required to engulf the forespore under the cramped conditions imposed by the cell wall . To address this question , Ojkic , López-Garrido et al . used microscopy to investigate how B . subtilis makes spores . The experiments show that , in order to engulf the forespore , the mother cell must produce new cell wall and destroy cell wall that is no longer needed . Running a simple biophysical model on a computer showed that coordinating these two processes could generate enough force for a mother cell to engulf a forespore . Ojkic , López-Garrido et al . propose that the junction between the septum and the cell wall moves around the forespore to make room for the mother cell’s membrane for expansion . Other spore-forming bacteria that threaten human health – such as Clostridium difficile , which causes bowel infections , and Bacillus anthracis , which causes anthrax – might form their spores in the same way , but this remains to be tested . More work will also be needed to understand exactly how bacterial cells coordinate the cell wall synthesis and cell wall degradation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Cell-wall remodeling drives engulfment during Bacillus subtilis sporulation |
In central nervous system ( CNS ) synapses , action potential-evoked neurotransmitter release is principally mediated by CaV2 . 1 calcium channels ( CaV2 . 1 ) and is highly dependent on the physical distance between CaV2 . 1 and synaptic vesicles ( coupling ) . Although various active zone proteins are proposed to control coupling and abundance of CaV2 . 1 through direct interactions with the CaV2 . 1 α1 subunit C-terminus at the active zone , the role of these interaction partners is controversial . To define the intrinsic motifs that regulate coupling , we expressed mutant CaV2 . 1 α1 subunits on a CaV2 . 1 null background at the calyx of Held presynaptic terminal . Our results identified a region that directly controlled fast synaptic vesicle release and vesicle docking at the active zone independent of CaV2 . 1 abundance . In addition , proposed individual direct interactions with active zone proteins are insufficient for CaV2 . 1 abundance and coupling . Therefore , our work advances our molecular understanding of CaV2 . 1 regulation of neurotransmitter release in mammalian CNS synapses .
A critical determinant in regulating synaptic vesicle ( SV ) release probability and kinetics is coupling , the physical distance of SVs and voltage-gated calcium channels ( VGCCs ) at the presynaptic terminal ( Neher and Sakaba , 2008 ) . Differences in coupling distances between CaV2 VGCCs subtypes underpin the differences in CaV2 VGCC subtype effectiveness in eliciting AP evoked release and define the SV release mode in response to APs ( Eggermann et al . , 2011 ) . They are: nanodomain , a few tightly coupled VGCCs ( <30 nm ) , and microdomain , many loosely coupled VGCCs ( ~100 nm ) trigger SV release ( Baur et al . , 2015; Eggermann et al . , 2011; Fedchyshyn and Wang , 2005 ) . In the majority of central nervous system synapses , CaV2 . 1 VGCCs ( CaV2 . 1 ) are the principal CaV subtype that supports AP mediated neurotransmitter release , and CaV2 . 1 channels are thought to exist in closest proximity to SVs compared to other CaV subtypes ( Eggermann et al . , 2011 ) . The CaV2 . 1 α1 subunit cytoplasmic C-terminus is mutated in a class of CaV2 channelopathies ( Pietrobon , 2010 ) and contains many motifs implicated to directly interact with key active zone ( AZ ) proteins to control CaV2 . 1 coupling and abundance in the presynaptic terminal ( Simms and Zamponi , 2014 ) . Nevertheless , the necessity and mechanism of action of these motifs are highly controversial due to disparate results from different model systems and from knockout mouse models of AZ proteins ( Acuna et al . , 2015; Atasoy et al . , 2007; Butz et al . , 1998; Cao et al . , 2004; Das , 2016; Davydova et al . , 2014; Ho et al . , 2006; Hu et al . , 2005; Kaeser et al . , 2011; Wong et al . , 2014; Wong and Stanley , 2010 ) . In addition , it is unclear whether the mechanisms that control coupling and abundance are interrelated or separable . To address these questions , we utilized the calyx of Held/Medial Nucleus of the Trapezoid Body ( MNTB ) synapse , a large glutamatergic axosomatic synapse , in which: ( 1 ) individual AZs ultrastructure and ( 2 ) CaV2 subtype abundance and proximity to SVs controlling SV release at the calyx of Held is similar to many other synapses ( Borst and Soria van Hoeve , 2012 ) . Furthermore , due to its unparalleled experimental accessibility , molecular manipulations can be made exclusively in the presynaptic terminals ( Wimmer et al . , 2004; Young and Neher , 2009 ) , and presynaptic Ca2+ currents can be recorded and correlated with synaptic vesicle release rates ( Neher and Sakaba , 2001b ) , which allows for well-controlled measurements not achievable in other model systems . By directly manipulating the CaV2 . 1 α1 subunit in a native neuronal circuit , we were able to overcome the previous limitations in prior studies ( Acuna et al . , 2015; Atasoy et al . , 2007; Butz et al . , 1998; Cao et al . , 2004; Davydova et al . , 2014; Ho et al . , 2006; Hu et al . , 2005; Kaeser et al . , 2011; Wong et al . , 2014; Wong and Stanley , 2010 ) . Thus , we were able to identify a novel intrinsic motif in the C-terminus that regulates coupling and demonstrate that coupling and abundance are separable . Finally , we found that this novel C-terminal region in the CaV2 . 1 α1 subunit also regulates SV docking at the AZ . Therefore , our work provides new molecular insights into CaV2 . 1 α1 subunit regulation of SV release from presynaptic terminals in CNS synapses .
To manipulate CaV2 . 1 at the calyx , Helper-Dependent Adenoviral vectors ( HdAd ) ( Palmer and Ng , 2005 ) were utilized in conjunction with a Cacna1a conditional knock-out ( CKO ) mouse line ( Todorov et al . , 2006 ) . HdAds can package large amounts of foreign DNA , which is critical as the CaV2 . 1 α1 subunit cDNA is larger than commonly used viral vectors ( Lentz et al . , 2012 ) . To modify CaV2 . 1 expression at the calyx , we used stereotactic surgery to deliver our HdAd viral vectors expressing Cre recombinase ( HdAd Cre ) to create a Cacna1a null background ( CaV2 . 1−/− ) and the full transcript of CaV2 . 1 α1 subunit , ( HdAd CaV2 . 1 FT ) into the cochlear nucleus ( Chen et al . , 2013 ) ( Figure 1 ) . CaV2 . 1 full transcript ( FT ) is the longest CaV2 . 1 α1 subunit cDNA ( Mus musculus NP_031604 . 3 ) . By testing Ca2+ current sensitivity to CaV2 subtype-specific blockers ω-Agatoxin IVA ( Aga , CaV2 . 1-selective ) and ω-Conotoxin GVIA ( Cono , CaV2 . 2-selective ) , we confirmed that we could ablate CaV2 . 1 and subsequently rescue CaV2 . 1 abundance ( Figure 1 ) . Since we could manipulate CaV2 . 1 at the calyx , we tested whether various previously proposed direct binding sites are necessary for regulating CaV2 . 1 localization and abundance at the presynaptic membrane . This includes binding sites for RIM1/2 ( Kaeser et al . , 2011 ) , MINT1 ( Maximov et al . , 1999 ) , Rim Binding Proteins ( RBP ) ( Hibino et al . , 2002 ) , and CASK proteins ( Maximov et al . , 1999 ) , a secondary CaVβ4 interaction site ( Walker et al . , 1998 ) as well as PXXP motifs ( Davydova et al . , 2014 ) . To do so we generated HdAd vectors with mutations in CaV2 . 1 α1 subunits in which we deleted these interaction sites ( Figures 1A–B , 2 and 3A ) . We expressed them at the CaV2 . 1−/− calyx and carried out whole-cell patch clamp recordings of the presynaptic Ca2+ currents ( Figure 2—figure supplement 2 and 1 and Table 1 ) ; CaV2 . 1Δ2365–2368 deletes the α1 subunit DDWC motif that is implicated to bind directly to RIM1/2 and MINT ( Kaeser et al . , 2011 ) . CaV2 . 1Δ2213–2368 corresponds to a CaV2 . 1 α1 subunit splice variant which removes the RIM1/2 , MINT1 , RBP , and part of the CASK binding site and majority of PXXP motifs ( Soong et al . , 2002 ) . CaV2 . 1Δ2016–2368 removes the complete CASK binding site , a proposed secondary CaVβ4 interaction site , and two remaining PXXP motifs in the α1 subunit ( Figure 1 , Figure 2—figure supplement 1 ) . Analysis of the Ca2+ current as a function of voltage ( I ( V ) ) and tail currents revealed that expression of mutants lacking motifs located within the last 350 amino acids revealed no significant difference in Ca2+ current amplitudes or voltage dependent activation compared to CaV2 . 1 FT rescue ( Figure 2 , Figure 2—figure supplement 1 and Table 1 ) . Although there appeared to be a slight reduction in maximal Ca2+ current amplitudes compared to FT rescue , there was no statistically significant difference among mutants and control . Thus , the MINT1 , RIM1/2 , RBP , CASK proteins and the secondary CaVβ4 binding sites within the CaV2 . 1 α1 subunit C-terminus are not necessary for CaV2 . 1 localization to the presynaptic membrane . 10 . 7554/eLife . 28412 . 003Figure 1 . CaV2 . 1 can be selectively ablated and functionally rescued at the calyx of Held . ( A ) Cartoon depicting CaV2 . 1 α1 subunit distal C-terminal interaction partners . ( B ) Amino acid sequence of the distal Cav2 . 1 C-terminus indicating interaction sites and truncation mutants . ( C ) left: Schematic view of stereotactic surgery to inject/coinject HdAd vectors expressing Cre + eGFP and CaV2 . 1 constructs + mCherry into the aVCN at age P1 . Right: top: Experimental timeline from virus injection at P1 to electrophysiological recordings at P9-P11 prior to the onset of hearing ( P12 ) . Middle and bottom: schematic view of the viral constructs used , expressing either Cre + eGFP or CaV2 . 1 constructs + mCherry , respectively , driven by individual promotors . ( D ) Calyx of Held terminals transduced with Cre + eGFP ( top ) and CaV2 . 1 + mCherry ( middle ) . eGFP and mCherry signals overlap with those of a calyx of Held loaded with Lucifer Yellow via a patch pipette ( bottom ) . ( E ) Pharmacological isolation of presynaptic CaV2 isoforms in wildtype , CKO and CaV2 . 1 full transcript rescue calyxes . Traces in absence of any blockers ( black ) , after blocking CaV2 . 1 fraction with 200 nM ω-AgaIVA ( brown ) , after blocking CaV2 . 2 fraction with 2 µM ω-GVIA ( blue ) and after blocking all CaV2 channels with 50 µM Cd2+ ( gray ) . ( F ) Relative CaV2 current fractions in wildtype , CKO and CaV2 . 1 full transcript rescue calyxes ( n = 3 for each condition ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 00310 . 7554/eLife . 28412 . 004Figure 2 . C-terminal deletions in CaV2 . 1 do not affect CaV2 abundance at the presynaptic terminal . ( A ) Cartoons depicting CaV2 . 1 full transcript or mutants ( top ) with corresponding exemplary Ca2+ currents ( bottom ) triggered by 10 ms voltage steps from d -50 mV to 50 mV in 5mV steps . ( B–C ) Current-voltage relationship of absolute Ca2+ currents ( B ) and normalized current-voltage relationships ( I/Imax; C ) . ( D ) Mean absolute Ca2+ currents . ( E–F ) Absolute tail currents ( E ) and normalized ( I/Imax; F ) tail currents as a function of voltage . ( G ) Mean tail Ca2+ currents at +40 mV . For CaV2 . 1 α1 CKO ( n = 11 ) , wildtype ( n = 12 ) , CaV2 . 1 α1full transcript ( n = 10 ) , Δ2365–2368 ( n = 10 ) , Δ2213–2368 ( n = 10 ) and Δ2016–2368 ( n = 10 ) . All data are depicted as mean ± SEM . Detailed values can be derived from Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 00410 . 7554/eLife . 28412 . 005Figure 2—figure supplement 1 . Cav2 . 1 rescue after CKO does not affect biophysical properties of the Ca2+ current at the calyx of Held . ( A ) top: Raw peak Ca2+ current amplitudes and , bottom: normalized peak currents ( I/Imasx ) as a function of voltage . Continuous curves represent fits according to Equation 1 . ( B ) top: Averaged raw and , bottom: normalized tail currents ( I/Imax ) ( bottom ) as a function of voltage . Continuous curves represent Boltzmann fits according to Equation 2 . ( C ) top: Quantification of maximal Ca2+ peak currents and , bottom: cell capacitance . All data are depicted as mean ± SEM with wildtype ( n = 12 ) , full transcript ( n = 10 ) and CKO ( n = 11 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 00510 . 7554/eLife . 28412 . 006Figure 3 . A novel role for a C-terminal region between amino acids 2042 and 2061 that regulates fast release independent of CaV2 . 1 abundance . ( A ) Cartoon depicting truncated regions in our CaV2 . 1 α1 deletion mutants including the binding sites for CaVβ4 , CASK , RBP , PXXP , RIM1/2 and Mint-1 along with the effects of C-terminal truncations on ICa . ( B ) Averaged traces of RRP and total releasable pool measurements from mice expressing Cre + full transcript CaV2 . 1 rescue ( grey ) , Δ2365–2368 ( cyan ) , Δ2213–2368 ( yellow ) , Δ2061–2368 ( purple ) , Δ2042–2368 ( green ) or Δ2016–2368 ( blue ) . ICa ( top ) and EPSCs ( bottom ) triggered by 3 ms and 30 ms pulses , plotted on top of each other ( n = 10 for each group , except for Δ2212–2368: n = 8 ) . ( C–H ) Quantification of ICa charge ( 3 ms: C; and 30 ms: F ) , max . EPSC amplitudes ( 3 ms: D; 30 ms: G ) and the 10–90% rise of the EPSCs ( 3 ms: E; 30 ms: H ) . All data are depicted as mean ± SEM . Detailed values can be derived from Table 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 00610 . 7554/eLife . 28412 . 007Figure 3—figure supplement 1 . Cav2 . 1 Full transcript rescue does not affect synaptic transmission at the calyx of Held/MNTB synapse . ( A–B ) Averaged traces of SV pool measurements in noninjected control mice ( black; a ) and mice expressing Cre + Full length rescue construct ( grey; B ) . Top: stimulation protocol driving SV release by depolarization from −80 mV to 70 mV for 2 ms , followed by 0 mV for 3 ms or 30 ms . Below: Traces depicting the resulting presynaptic ICa ( middle ) with corresponding EPSCs ( bottom ) . ( C–D ) Summary graphs of SV release rates ( c ) and the cumulative synaptic vesicle release rates ( D ) after 3 ms and 30 ms stimulation . ( E–F ) Summary graphs depicting normalized SV release after 3 ms and 30 ms stimulation ( E ) and normalized cumulative SV release after 30 ms stimulation ( F ) . ( G–L ) Quantification of ICa amplitude ( G ) , charge of Ca2+ currents ( H I ) , EPSC amplitude ( J ) , 10–90% rise time of the EPSC ( K ) after 3 ms and 30 ms stimulation , respectively , as well as the number of SVs released by 30 ms stimulation ( L ) . Data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 00710 . 7554/eLife . 28412 . 008Table 1 . Electrophysiological parameters of IV relations of Ca2+ currents . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 008ParameterMean ± SEM ( n ) OW-ANOVA Dunnett’s TestMax . Ca2+ current amplitude Imax ( pA ) Wild type911 ± 63 ( 12 ) p=0 . 9998 ( n . s . ) Full transcript925 ± 99 ( 10 ) control groupCKO464 ± 59 ( 11 ) p<0 . 0001 ( **** ) △2365–2368863 ± 53 ( 10 ) p=0 . 9513 ( n . s . ) △2213–2368741 ± 56 ( 10 ) p=0 . 2261 ( n . s . ) △2016–2368744 ± 67 ( 10 ) p=0 . 2383 ( n . s . ) Membrane capacitance Cslow ( pF ) Wild type20 . 9 ± 1 . 6 ( 12 ) p=0 . 5071 ( n . s . ) Full transcript18 . 2 ± 1 . 2 ( 10 ) control groupCKO18 . 1 ± 1 . 8 ( 11 ) p>0 . 9999 ( n . s . ) △2365–236816 . 9 ± 1 . 1 ( 10 ) p=0 . 9593 ( n . s . ) △2213–236819 . 8 ± 1 . 4 ( 10 ) p=0 . 9933 ( n . s . ) △2016–236817 . 2 ± 1 . 8 ( 10 ) p=0 . 8986 ( n . s . ) IV fit: Half-maximal activation voltage Vm ( mV ) Wild type−24 . 6 ± 1 . 3 ( 12 ) p=0 . 9986 ( n . s . ) Full transcript−25 . 1 ± 1 . 3 ( 10 ) control groupCKO−22 . 3 ± 1 . 3 ( 11 ) p=0 . 3604 ( n . s . ) △2365–2368−23 . 1 ± 1 . 1 ( 10 ) p=0 . 6831 ( n . s . ) △2213–2368−23 . 8 ± 1 . 5 ( 10 ) p=0 . 9298 ( n . s . ) △2016–2368−23 . 3 ± 0 . 9 ( 10 ) p=0 . 7924 ( n . s . ) IV fit: Voltage-dependence of activation km ( mV ) Wild type8 . 0 ± 0 . 5 ( 12 ) p=0 . 9524 ( n . s . ) Full transcript7 . 4 ± 0 . 5 ( 10 ) control groupCKO12 . 6 ± 1 . 3 ( 11 ) p<0 . 0001 ( **** ) △2365–23687 . 6 ± 0 . 3 ( 10 ) p=0 . 9997 ( n . s . ) △2213–23688 . 4 ± 0 . 4 ( 10 ) p=0 . 7154 ( n . s . ) △2016–23688 . 2 ± 0 . 3 ( 10 ) p=0 . 8481 ( n . s . ) Boltzmann fit: Half-maximal activation voltage V0 . 5 ( mV ) Wild type−10 . 6 ± 1 . 4 ( 12 ) p=0 . 9997 ( n . s . ) Full transcript−11 ± 0 . 1 ( 10 ) control groupCKO−1 . 7 ± 1 . 1 ( 11 ) p<0 . 0001 ( **** ) △2365–2368−8 . 7 ± 1 . 1 ( 10 ) p=0 . 6311 ( n . s . ) △2213–2368−8 . 2 ± 1 . 8 ( 10 ) p=0 . 4343 ( n . s . ) △2016–2368−8 . 9 ± 1 . 1 ( 10 ) p=0 . 7130 ( n . s . ) Boltzmann fit: Voltage-dependence k ( mV ) Wild type8 . 3 ± 0 . 5 ( 12 ) p=0 . 9111 ( n . s . ) Full transcript8 . 9 ± 0 . 7 ( 10 ) control groupCKO10 . 3 ± 0 . 6 ( 11 ) p=0 . 3117 ( n . s . ) △2365–23687 . 7 ± 0 . 7 ( 10 ) p=0 . 4951 ( n . s . ) △2213–23688 . 9 ± 0 . 4 ( 10 ) p=0 . 9947 ( n . s . ) △2016–23687 . 2 ± 0 . 3 ( 10 ) p=0 . 1578 ( n . s . ) *One-Way ANOVA with a Dunnett’s Test with condition knockout as reference group was performed to calculate statistical significance . To determine the intrinsic motif ( s ) involved in the regulation of coupling , we performed paired whole cell voltage clamp recordings on the pre- and postsynaptic compartments of the calyx of Held/MNTB synapse with these deletion mutants ( Neher and Sakaba , 2001a , 2001b ) . For finer mapping we generated two additional deletion constructs ( Figure 3A ) . CaV2 . 1Δ2061–2368 deletes up to the secondary CaVβ4 interaction in the α1 subunit and CaV2 . 1Δ2042–2368 deletes an additional arginine rich stretch in the CaV2 . 1 α1 subunit , not found in CaV2 . 2 and CaV2 . 3 and the final shared PXPP motif . Conotoxin was included to block possible CaV2 . 2 channel contributions . First we applied either a 3 ms step depolarization pulse ( Figure 3—figure supplement 1A ) to the calyx which selectively depletes SVs within ~50–80 nm of CaV2 VGCCs ( Chen et al . , 2015 ) which participate in synchronous transmitter release ( fast pool ) ( Chen et al . , 2015; Lee et al . , 2012 ) . The fast pool is the relevant SV pool that supports AP-mediated release and thus considered the readily-releasable pool ( RRP ) ( Figures 3–4 , Figure 3—figure supplement 1 and Table 2 ) ( Sakaba , 2006 ) . Then we applied a 30 ms step depolarization ( Figure 3—figure supplement 1 ) which measures the entire pool of fusion competent SVs , all within ~200 nm of CaV2 and considered the total releasable pool ( Chen et al . , 2015; Lee et al . , 2012 ) ( Figures 3–4 , Figure 3—figure supplement 1 and Table 2 ) . To validate our approach we compared the effects of SV release between calyces expressing Cre + CaV2 . 1 FT construct and wild-type calyces . We found no differences in SV release between the CaV2 . 1 FT and wild-type calyces ( Figure 3 , Figure 3—figure supplement 1 and Table 2 ) , indicating that exogenous expression of the CaV2 . 1 α1 subunit did not alter calyx/MNTB synaptic transmission . 10 . 7554/eLife . 28412 . 009Table 2 . Summary of currents from synaptic vesicle pool measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 009Parameter3 ms ( mean ± SEM ( n ) OW-ANOVA Dunnett’s Test30 ms ( mean ± SEM ( n ) OW-ANOVA Dunnett’s TestCa2+ current amplitude ( nA ) Wild type0 . 99 ± 0 . 07 ( 10 ) p=0 . 2671 ( n . s . ) 0 . 93 ± 0 . 07 ( 10 ) p=0 . 3557 ( n . s . ) Full transcript1 . 24 ± 0 . 12 ( 10 ) control group1 . 18 ± 0 . 12 ( 10 ) control group△2365–23681 . 14 ± 0 . 84 ( 10 ) p=0 . 9320 ( n . s . ) 1 . 01 ± 0 . 12 ( 10 ) p=0 . 9451 ( n . s . ) △2213–23681 . 04 ± 0 . 09 ( 8 ) p=0 . 5366 ( n . s . ) 0 . 9 ± 0 . 1 ( 8 ) p=0 . 2766 ( n . s . ) △2061–23681 . 23 ± 0 . 08 ( 10 ) p=0 . 9999 ( n . s . ) 1 . 13 ± 0 . 08 ( 10 ) p=0 . 9993 ( n . s . ) △2042–23681 . 13 ± 0 . 15 ( 10 ) p=0 . 9121 ( n . s . ) 1 . 03 ± 0 . 13 ( 10 ) p=0 . 8410 ( n . s . ) △2016–23680 . 93 ± 0 . 38 ( 10 ) p=0 . 1091 ( n . s . ) 0 . 86 ± 0 . 46 ( 10 ) p=0 . 1175 ( n . s . ) Ca2+ influx charge ( pC ) Wild type2 . 95 ± 0 . 3 ( 10 ) p=0 . 5713 ( n . s . ) 25 . 93 ± 2 . 22 ( 10 ) p=0 . 4932 ( n . s . ) Full transcript3 . 62 ± 0 . 36 ( 10 ) control group31 . 71 ± 2 . 87 ( 10 ) control group△2365–23683 . 89 ± 0 . 31 ( 10 ) p=0 . 9838 ( n . s . ) 33 . 52 ± 2 . 65 ( 10 ) p=0 . 9943 ( n . s . ) △2213–23683 . 48 ± 0 . 35 ( 8 ) p=0 . 9996 ( n . s . ) 28 . 09 ± 2 . 7 ( 8 ) p=0 . 8924 ( n . s . ) △2061–23683 . 87 ± 0 . 33 ( 10 ) p=0 . 9893 ( n . s . ) 34 . 94 ± 2 . 93 ( 10 ) p=0 . 9132 ( n . s . ) △2042–23683 . 8 ± 0 . 54 ( 10 ) p=0 . 9977 ( n . s . ) 33 . 71 ± 4 . 22 ( 10 ) p=0 . 9910 ( n . s . ) △2016–23683 . 12 ± 0 . 14 ( 10 ) p=0 . 8063 ( n . s . ) 27 . 81 ± 1 . 14 ( 10 ) p=0 . 8255 ( n . s . ) EPSC amplitude ( nA ) Wild type8 . 99 ± 1 . 12 ( 10 ) p=0 . 9995 ( n . s . ) 9 . 77 ± 0 . 87 ( 10 ) p=0 . 8167 ( n . s . ) Full transcript8 . 54 ± 0 . 9 ( 10 ) control group8 . 31 ± 0 . 79 ( 10 ) control group△2365–23686 . 88 ± 0 . 78 ( 10 ) p=0 . 7024 ( n . s . ) 7 . 45 ± 0 . 73 ( 10 ) p=0 . 9789 ( n . s . ) △2213–23686 . 05 ± 1 . 2 ( 8 ) p=0 . 3672 ( n . s . ) 6 . 7 ± 0 . 94 ( 8 ) p=0 . 7880 ( n . s . ) △2061–23687 . 05 ± 1 . 25 ( 10 ) p=0 . 7836 ( n . s . ) 7 . 72 ± 1 . 44 ( 10 ) p=0 . 9963 ( n . s . ) △2042–23683 . 38 ± 0 . 89 ( 10 ) p=0 . 0028 ( ** ) 5 . 28 ± 1 . 31 ( 10 ) p=0 . 1689 ( n . s . ) △2016–23682 . 49 ± 0 . 89 ( 10 ) p=0 . 0004 ( *** ) 3 . 61 ± 0 . 91 ( 10 ) p=0 . 0098 ( ** ) EPSC 10–90% rise time ( ms ) Wild type1 . 46 ± 0 . 14 ( 10 ) p=0 . 9908 ( n . s . ) 1 . 71 ± 0 . 31 ( 10 ) p=0 . 9995 ( n . s . ) Full transcript1 . 25 ± 0 . 13 ( 10 ) control group1 . 41 ± 0 . 18 ( 10 ) control group△2365–23681 . 79 ± 0 . 09 ( 10 ) p=0 . 5934 ( n . s . ) 2 . 18 ± 0 . 27 ( 10 ) p=0 . 9097 ( n . s . ) △2213–23681 . 88 ± 0 . 06 ( 8 ) p=0 . 4956 ( n . s . ) 2 . 43 ± 0 . 3 ( 8 ) p=0 . 7964 ( n . s . ) △2061–23681 . 8 ± 0 . 12 ( 10 ) p=0 . 5792 ( n . s . ) 2 . 07 ± 0 . 21 ( 10 ) p=0 . 9529 ( n . s . ) △2042–23682 . 46 ± 0 . 42 ( 10 ) p=0 . 0200 ( * ) 4 . 67 ± 0 . 72 ( 10 ) p=0 . 0046 ( ** ) △2016–23683 . 2 ± 0 . 57 ( 10 ) p<0 . 0001 ( **** ) 6 . 99 ± 1 . 45 ( 10 ) p<0 . 0001 ( **** ) Synaptic delay ( ms ) Wild type1 . 82 ± 0 . 13 ( 10 ) p=0 . 9569 ( n . s . ) 1 . 92 ± 0 . 2 ( 10 ) p=0 . 9977 ( n . s . ) Full transcript1 . 7 ± 0 . 12 ( 10 ) control group1 . 73 ± 0 . 11 ( 10 ) control group△2365–23682 . 11 ± 0 . 1 ( 10 ) p=0 . 1170 ( n . s . ) 2 . 22 ± 0 . 16 ( 10 ) p=0 . 8464 ( n . s . ) △2213–23682 . 19 ± 0 . 1 ( 8 ) p=0 . 0615 ( n . s . ) 2 . 33 ± 0 . 19 ( 8 ) p=0 . 7507 ( n . s . ) △2061–23682 . 14 ± 0 . 14 ( 10 ) p=0 . 0748 ( n . s . ) 2 . 28 ± 0 . 15 ( 10 ) p=0 . 7714 ( n . s . ) △2042–23682 . 8 ± 0 . 07 ( 10 ) p<0 . 0001 ( **** ) 3 . 68 ± 0 . 3 ( 10 ) p=0 . 0019 ( ** ) △2016–23682 . 55 ± 0 . 2 ( 10 ) p<0 . 0001 ( **** ) 4 . 78 ± 0 . 83 ( 10 ) p<0 . 0001 ( **** ) *One-Way ANOVA with a Dunnett’s Test with full transcript as a control group was performed to calculate statistical significance . 10 . 7554/eLife . 28412 . 010Table 3 . Summary of 3ms /30ms EPSC ratios . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 010EPSC ratio ( 3 ms/30 ms ) Wild type0 . 89 ± 0 . 05 ( 10 ) p=0 . 5456 ( n . s . ) Full transcript1 . 03 ± 0 . 03 ( 10 ) control group△2365–23680 . 92 ± 0 . 04 ( 10 ) p=0 . 6887 ( n . s . ) △2213–23680 . 87 ± 0 . 05 ( 8 ) p=0 . 4631 ( n . s . ) △2061–23680 . 93 ± 0 . 04 ( 10 ) p=0 . 7858 ( n . s . ) △2042–23680 . 61 ± 0 . 09 ( 10 ) p=0 . 0004 ( *** ) △2016–23680 . 50 ± 0 . 12 ( 10 ) p<0 . 0001 ( **** ) 10 . 7554/eLife . 28412 . 011Figure 4 . The novel C-terminal region between amino acids 2042 and 2061 regulates size of the fast and total releasable pool and synaptic vesicle release kinetics . ( A–B ) Average release rate trace after 3 ms ( A ) or 30 ms stimulation ( B ) from calyces expressing either Cre + full transcript rescue ( grey ) , Δ2042–2368 ( green ) or Δ2016–2368 ( blue ) ; n = 10 for each group; ( C ) Averaged cumulative release after 3 ms and 30 ms stimulation . ( D ) Normalized cumulative release of the total releasable pool triggered by 30 ms stimulation . Inset presents a magnified view of the area encircled by the dashed box . ( E–G ) Quantification of SV numbers released by 3 ms ( E ) and 30 ms ( F ) as well as the ratio of SVs released by 3 ms and 30 ms stimulation ( G ) . All data are depicted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 011 In response to 3 ms and 30 ms presynaptic depolarizations , we found no difference in the presynaptic Ca2+ currents in all mutants , thus confirming our results depicted in Figure 2 ( Figure 3 and Table 2 ) . Since the 3 ms peak EPSC amplitude directly correlates to those SVs that are tightly coupled to CaV2 channels at the P9-11 calyx ( Chen et al . , 2015; Lee et al . , 2012 ) , we measured the peak 3 ms EPSC peak amplitudes in all our deletion mutants . Thus , if these intrinsic motifs were essential for SV to CaV2 . 1 coupling , we should see a dramatic reduction in the 3 ms peak EPSC amplitude , and if they were not essential there should be no change . Analysis of the 3 ms peak EPSC amplitudes revealed no change in the peak amplitudes with deletions from amino acid 2265 and beyond ( CaV2 . 1Δ2365–2368 , CaV2 . 1Δ2213–2368 and CaV2 . 1Δ2061–2368 ) , when compared to control ( CaV2 . 1 FT; Figure 3 and Table 2 ) . However , we saw a dramatic reduction in the EPSC amplitudes with CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 ( FT: 8 . 54 ± 0 . 9 nA; Δ2042–2368: 3 . 38 ± 0 . 89 nA ( p<0 . 01 ) ; Δ2016–2368: 2 . 47 ± 0 . 89 nA ( p<0 . 001 ) ; Figure 3 and Table 2 ) . In addition , only the CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 mutants showed a significant slowdown in the 10–90 rise time compared to FT and a significant increases in the synaptic delay time ( Figure 2 and Table 2 ) . In response to the 30 ms step pulse , we found no significant change in the 10–90 rise time or EPSC amplitudes with CaV2 . 1Δ2365–2368 , CaV2 . 1Δ2213–2368 and CaV2 . 1Δ2061–2368 compared to control ( Figure 2 , Table 2 ) . It is important to note that unlike the 3 ms peak EPSC amplitude , the 30 ms 10–90 peak EPSC rise time is an inaccurate measure of coupling of all SVs in the total releasable pool , as the 30 ms peak amplitude does not accurately measure the total releasable pool size ( Chen et al . , 2015; Lee et al . , 2012 ) . We found a significant increase in the 10–90 rise time with CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 ( FT: 1 . 41 ± 0 . 18 ms; Δ2042–2368: 4 . 67 ± 0 . 72 ms ( p<0 . 01 ) ; Δ2016–2368: 6 . 99 ± 1 . 45 ms ( p<0 . 0001 ) ) , with reduced EPSC amplitudes ( Figure 3 and Table 2 ) . In all cases , there was no difference between CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 indicating that the further deletion did not lead to more severe reductions in the ESPC amplitude or 10–90 rise time . Although , there appeared to be a slight slowing in the 10–90 rise time with CaV2 . 1Δ2365–2368 , CaV2 . 1Δ2213–2368 and CaV2 . 1Δ2061–2368 compared to control in both the 3 ms and 30 ms EPSC ( Figure 3 , Table 2 ) , this change was not statistically significant and was very minor compared to the dramatic deceleration in in the 10–90 rise times found in CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 . Based on these results , we conclude that a novel C-terminal region including at least the amino acids 2042–2061 is critical for fast release . To understand how the CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 truncations impacted the size and release kinetics of the fast pool ( AP-evoked release ) and the total releasable pool ( Figure 3 and Figure 3—figure supplement 1 ) , we used a deconvolution analysis routine to calculate release rates ( Neher and Sakaba , 2001a , 2001b ) . We found that both CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 lead to a dramatic reduction in peak vesicle release rates ( Figure 4 ) with a significant increase in the delayed release ( slow pool component ) and slower time to peak EPSC release rates compared to control ( Figure 4 ) . Integration of the release rates for both the 3 ms and 30 ms pulses revealed a dramatic reduction in both mutants of both the fast pool and the total releasable pool . ( RRP: FT: 1505 ± 245 SVs;Δ2042–2368: 430 ± 116 SVs ( p<0 . 001 ) ; Δ2016–2368: 357 ± 134 SVs ( p<0 . 001 ) ; total releasable pool: FT: 2152 ± 263 SVs; Δ2042–2368: 1292 ± 221 SVs ( p<0 . 05 ) ; Δ2016–2368: 1061 ± 258 SVs ( p<0 . 01 ) ) ( Figure 4 ) . To test how the kinetics of release were affected by the fast component of release , the cumulative release rates were normalized to their respective total number of vesicles released during the 30 ms depolarizing pulse ( Figure 4D ) . This clearly demonstrates that both mutants had a significantly decreased fast component . In all cases , there were no differences between CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 . Comparison of the ratio of the RPP size to the total releasable pool size revealed a significant reduction in the contribution of the RRP to the total releasable pool size in the mutants . ( Figure 4G ) ( Table 3 ) . Thus , based on our results we can conclude that the region between 2016 and 2042 is essential for both regulating the total number of releasable vesicles , as well as the relative contributions of fast and slow SV pool components . Since docked synaptic vesicles at the AZ are the morphological correlates of the RRP ( Schikorski and Stevens , 2001 ) , we next assessed how CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 affected presynaptic ultrastructure . To do so , we acquired and analyzed electron microscopy ( EM ) images from the uninfected contralateral in slice control , CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 expressing calyces to examine whether SV docking and distribution or AZ length were altered . Analysis of EM images revealed that AZ lengths were unchanged ( CaV2 . 1Δ2042–2368: 267 . 1 ± 8 . 1 nm vs . in slice control: 280 . 6 ± 8 . 2 nm ( n = 120 ) ; CaV2 . 1Δ2016–2368:268 . ± 7 . 9 nm vs . in slice control: 292 . 5 ± 8 . 1 nm ( n = 100 ) ) , but revealed a specific reduction in only those SVs within 5 nm of the plasma membrane , in both CaV2 . 1Δ2042–2368 and CaV2 . 1Δ2016–2368 ( CaV2 . 1Δ2042–2368: 0 . 77 ± 0 . 08 nmvs . in-slice control: 1 . 39 ± 0 . 1 ( n = 120; p<0 . 0001 ) ; CaV2 . 1Δ2016–2368: 0 . 63 ± 0 . 08 nm vs . in-slice control:1 . 58 ± 0 . 12 nm ( n = 100; p<0 . 0001 ) ( Figure 5 ) . Thus , morphological analysis revealed that the region between 2042 and 2061 in CaV2 . 1 α1 subunit regulates SV docking , and its deletion results in a reduced fast pool ( RRP ) size and total releasable pool size . 10 . 7554/eLife . 28412 . 012Figure 5 . The novel C-terminal region between amino acids 2042 and 2061 regulates the number of docked synaptic vesicles at the active zones . ( A–C ) Representative EM images showing AZs from nontransduced calyces ( A ) , and calyces transduced with Δ2042–2368 ( B ) or Δ2016–2368 ( C ) . Transduced cells were identified by pre-embedded nanogold immunolabelling for eGFP ( black dots in B and C ) . ( D–G ) Quantification of mean AZ length and docked SVs ( within 5 nm of the membrane ) of calyces transduced with Δ2042–2368 ( n = 120; D E ) or Δ2016–2368 ( n = 100; F G ) , compared to AZs from the nontransduced contralateral MNTB , respectively . ( H–I ) Quantification of the mean distribution of SVs up to 200 nm distant from AZs for calyces expressing Δ2042–2368 ( n = 120; H ) or Δ2016–2368 ( n = 100; I ) . Insets show SVs in closest proximity to the membrane ( up to 20 nm ) . All data are depicted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 28412 . 012
In our study , we specifically ablated the CaV2 . 1 α1 subunit in the calyx of Held using a flox mouse line of the CaV2 . 1 α1 subunit ( Todorov et al . , 2006 ) which circumvents potential artifacts due to global loss of CaV2 . 1 in the brain and lethality issues in the Cacna1a KO mouse line ( Jun et al . , 1999 ) . A major road block to studying CaV2 . 1 channel function in native mammalian neuronal circuits has been difficulties with the ability to make routine presynaptic molecular manipulations of CaV2 . 1 α1 subunit . The CaV2 . 1 α1 subunit ( >7 kb cDNA ) , ( Catterall , 2011 ) is larger than common viral vectors such as recombinant Adeno-associated virus ( rAAV ) or lentiviral vectors ( rLVV ) , 5 kb and 9 kb maximum packaging capacity ( Lentz et al . , 2012 ) . To overcome these challenges we utilized HdAd vectors which supplant earlier versions of recombinant Ad technology , permit packaging of up to 37 kb of foreign DNA , overcome the limitations of rAAV and rLVV , and do not impact neuronal viability ( Montesinos et al . , 2016; Muhammad et al . , 2012; Palmer and Ng , 2003; Palmer and Ng , 2005 ) . We used a small modified 470 bp human synapsin promoter ( hSyn ) ( Kügler et al . , 2003 ) that is in widespread use throughout the neuroscience field . This promoter is a relatively weak promoter and does not lead to massive overexpression as seen with CMV or CBA promoters ( Glover et al . , 2002 ) . Our rescue experiments showed CaV2 . 1 α1 subunit expression with our HdAd vectors in the CaV2 . 1−/− background lead to similar Ca2+ current amplitudes , CaV2 . 1 subtype levels and similar SV release rates as wild-type , thereby validating our experimental approach ( Figure 1 , Figure 2—figure supplement 1 ) . Thus , our HdAd vectors in conjunction with the Cacna1a CKO mouse line and our stereotactic surgery techniques ( Chen et al . , 2013 ) will be a useful platform technology to help decipher CaV2 . 1 function in native neuronal circuits . By deleting multiple AZ protein binding sites in the CaV2 . 1 α1 subunit and making direct presynaptic recordings at the calyx , we demonstrated that these binding sites and other motifs in the last 350 amino acids of the CaV2 . 1 α1 subunit are not necessary for CaV2 . 1 localization to the presynaptic membrane . Based on our results ( Figures 2 and 3 ) which demonstrated no significant changes in Ca2+ currents , we can rule out that proposed direct interactions with either RIM1/2 ( Kaeser et al . , 2011 ) , MINT1 , CASK ( Maximov and Bezprozvanny , 2002; Maximov et al . , 1999 ) and RBP ( Davydova et al . , 2014; Hibino et al . , 2002 ) proteins are essential . Our results are similar to those studies ( Cao and Tsien , 2010; Hu et al . , 2005 ) that expressed the CaV2 . 1 α1 subunit splice variant lacking the RIM1/2 , RBP , or MINT1 binding sites ( Soong et al . , 2002 ) . This splice variant is localized to the presynaptic terminal ( Hu et al . , 2005 ) and could rescue the CaV2 . 1 channel contribution to AP evoked release in Cacna1a KO primary hippocampal neurons ( Cao and Tsien , 2010 ) . Furthermore , our data is in line with studies from Cask KO ( Atasoy et al . , 2007 ) and X11α KO ( Ho et al . , 2003 ) which had no impact on basal AP-evoked release kinetics and Rim-bp1/ Rim-bp2 cKO ( Acuna et al . , 2015 ) animals which demonstrated no loss of CaV2 . 1 current density . Although our results cannot rule out that Bassoon controls CaV2 . 1 abundance , our results do not support the model in which Bassoon regulates CaV2 . 1 abundance at the presynaptic terminal through direct RBP interaction with the CaV2 . 1 α1 subunit ( Davydova et al . , 2014 ) . RIM proteins have been demonstrated to be important for regulating CaV2 channel current density and abundance at both invertebrate and vertebrate presynaptic terminals , as knock out RIM proteins lead to dramatic reductions in Ca2+ currents ( Graf et al . , 2012; Han et al . , 2011; Kaeser et al . , 2011 ) . However , deletion of the DDWC motif in the CaV2 . 1 α1 subunit which interacts both with MINT1 and RIM1/2 did not lead to any changes in CaV2 . 1 currents ( Figures 2 and 3 ) indicating this motif is not necessary for CaV2 . 1 targeting to the presynaptic membrane . Although the RIM1/2 PDZ domain interaction with the CaV2 . 1 α1 subunit DDWC is proposed to be critical for CaV2 . 1/2 . 2 abundance and localization to the presynaptic terminal , the necessity of this interaction was not directly demonstrated in vivo ( Kaeser et al . , 2011 ) . Furthermore , other biochemical assays have failed to detect this direct interaction ( Wong et al . , 2013 , 2014; Wong and Stanley , 2010 ) . Previous studies demonstrated that the RIM PDZ domain interacts with the CAST/ELKS proteins with an affinity of 200 nM ( Lu et al . , 2005 ) , while the RIM PDZ domain interaction with the CaV2 . 1/2 . 2 α1 subunit has an affinity of 20 µM ( Kaeser et al . , 2011 ) . Thus , an alternative interpretation is that RIM1/2 PDZ interacts with CAST/ELKS proteins to form a macromolecular complex that controls CaV2 . 1/2 . 2 abundance ( Hida and Ohtsuka , 2010 ) . What then could be the motifs that are essential for CaV2 . 1 localization/abundance at the presynaptic terminal ? In addition to the C-terminal region , the synprint region which binds to SNARE proteins has been proposed to be an integral motif for CaV2 . 1 incorporation into the presynaptic terminal ( Catterall , 2011 ) . However , syntaxin 1A binding sites in CaV2 . 2 are dispensable for synaptic targeting and AP-evoked release ( Szabo et al . , 2006 ) . Interestingly , synprint domains lacking syntaxin 1A binding sites have reduced levels of incorporation into the neuroendocrine cell membranes ( Rajapaksha et al . , 2008 ) . Another potential motif is the AID domain in the CaV2 . 1 α1 subunit which interacts with the CaVβ subunit to regulate CaV2 . 1 localization to the presynaptic membrane via a RIM1/2-dependent mechanism ( Kiyonaka et al . , 2007 ) . Since the CaV2 . 1 α1 subunit is heavily spliced depending on the neuronal cell-type ( Simms and Zamponi , 2014 ) , it is possible that no single motif is responsible for CaV2 . 1 localization and incorporation in the presynaptic membrane , but that various motifs may act in concert or independently to ensure CaV2 . 1 abundance . In addition , the necessity of these motifs may also vary at different synapses and the developmental state of the neuronal circuit in which the synapses are embedded . CaV2 . 1 channels are not randomly distributed in the presynaptic membrane but cluster within AZs ( Holderith et al . , 2012; Nakamura et al . , 2015 ) . Since the calyx of Held is a large presynaptic terminal that contains many AZs , the conclusions based our presynaptic recordings are limited to CaV2 . 1 localization to the presynaptic terminal . Thus , we cannot rule out that the C-terminal domains contained within amino acids 2016–2368 which are not essential for localization to the presynaptic membrane , are critical for CaV2 . 1 clustering/organization within individual AZs . To determine if the mechanisms that control CaV2 . 1 clustering and presynaptic membrane localization are independently regulated , morphological studies will need to be carried out . Although exons 44–47 of CaV2 . 1 can impact voltage dependent activation and inactivation in HEK293 cells ( Hirano et al . , 2017 ) , our direct recordings did not detect differences in the overall biophysical parameters of these mutants compared to terminals rescued with full length CaV2 . 1 α1 subunits or in a wild-type background . We did not block CaV2 . 2 or CaV2 . 3 currents which are present in prehearing calyces ( Doughty et al . , 1998; Iwasaki and Takahashi , 1998 ) . Therefore , the presence of these currents could obscure possible changes in CaV2 . 1 current with our mutants . In addition , we did not test for these regions role in the regulation of Calcium-dependent activation or facilitation . Thus , depending on the synapse and its developmental state , it is possible that these regions are critical for modulation of CaV2 . 1 function in response to high frequency stimulation . Although many proteins interact with the CaV2 . 1 channel complex ( Müller et al . , 2010 ) , it is largely unknown whether they interact with the CaV2 . 1 α1 subunit to regulate coupling in its native environment . Our paired recordings revealed a dramatic reduction in the 3 ms peak EPSC amplitude which measures the RRP , while release kinetics with both the 3 ms and 30 ms EPSCs were dramatically slowed down in the CaV2 . 1Δ2042-2368 mutant but not Δ2061-2368 mutant . Thus , our data demonstrates a motif or motif ( s ) in amino acids 2042–2061 in the CaV2 . 1 α1 subunit regulates RRP size and coupling ( Figure 6 ) . Prior studies using step depolarizations at the prehearing calyx have shown that loss of RIM and RBPs resulted in a 2–3 fold slowdown in release kinetics respectively and indicating that these proteins are involved in pathways that regulate the RRP size , coupling ( Acuna et al . , 2015; Han et al . , 2011 ) . In contrast , we did not see any dramatic slowdown in release or changes in RRP size that mimicked these phenotypes previously seen with RIM or RBP KO animals when these direct binding motifs in the CaV2 . 1 α1 subunit were deleted . Although we observed a slight slowing of the 10–90 rise times , ~20–30% , which was similar for the Δ2365–2368 , Δ2213–2368 , Δ2061–2368 mutants , this was not statistically significant . We did not directly measure AP-evoked release in this study , but it has been previously demonstrated that a CaV2 . 1 splice variants lacking RIM or RBP binding sites rescued AP-evoked release ( Cao and Tsien , 2010 ) . Finally , it has been demonstrated in PC12 cells that CaVβ interactions with RIM1/2 are critical for anchoring SVs to CaV2 calcium channels to control coupling ( Uriu et al . , 2010 ) . Thus , our results strongly support that proposed individual direct interactions in between the CaV2 . 1 α1 subunit with RIM1/2 ( Han et al . , 2011; Kaeser et al . , 2011 ) or RBP proteins ( Acuna et al . , 2015; Hibino et al . , 2002 ) at most play a minor role in regulating RRP size and coupling . In addition to the reduction in RRP size , we observed a reduction in the total releasable pool size , indicating a loss in the number of fusion competent vesicles . Our morphological analysis revealed that this corresponded to a 50% reduction in the number of docked SVs compared to wild-type . Recent work proposed that SV docking corresponds to priming ( Imig et al . , 2014 ) , therefore we propose that this region in the CaV2 . 1 α1 subunit is involved in priming . Close inspection of the amino acid sequence reveals little homology to CaV2 . 2 and CaV2 . 3 α1 subunits ( ClustalW-Gonnet series algorithm ) , however BLAST homology search reveals no known binding motifs ( data not shown ) . It is possible that this region may bind to known or unknown AZ proteins that organize/cluster CaV2 . 1 channels in the AZ , which in turn couple to SVs and promote docking of SVs in the AZ . An alternative explanation is that removal of the 2061 to 2041 region results in a misfolding of this specific region so that CaV2 . 1 channels cannot interact with other proteins through other regions and cannot cluster CaV2 . 1 channels in the AZ . In both cases , SV docking would rely on CaV2 . 1 clustering through another protein to promote SV docking . However , to delineate the molecular mechanisms of how this motif regulates coupling , RRP and total releasable pool size , future experiments to identify potential binding partner ( s ) or solving of the CaV2 . 1α1 subunit structure and these mutants will need to be performed . Despite containing CaV2 . 1 , some presynaptic terminals transition from microdomain to nanodomain during maturation of neuronal circuits that encode temporal fidelity at high firing rates ( Baur et al . , 2015; Fedchyshyn and Wang , 2005 ) . Therefore , these release states are not specific to individual CaV2 subtypes , but instead the intrinsic motifs within the CaV2 . 1 α1 subunit are differentially utilized based on the developmental state . Our results presented here focused solely on the regulation of fast release at the prehearing calyx P9-11 which utilizes microdomain release mode ( Borst and Sakmann , 1996; Fedchyshyn and Wang , 2005 ) . Since the calyx transitions from microdomain to nanodomain release after the onset of hearing it is highly possible that intrinsic motifs in the CaV2 . 1 α1 subunit dispensable for microdomain release are necessary to support nanodomain release . Finally , since proteome composition at the AZ may vary in different various presynaptic terminals , different CaV2 . 1 motifs may or may not be essential to support coupling . Taken together , these findings counter the prevailing views that ( 1 ) individual direct interactions between the CaV2 . 1 α1 subunit and RIM1/2 , MINT , RBP proteins are crucial for controlling CaV2 . 1 abundance and coupling to SVs ( Südhof , 2013 ) and ( 2 ) that PXXP motifs are involved in capturing and coupling SVs to calcium channels ( Wong et al . , 2014 ) . Thus , our results suggest that the mechanisms of action by which known AZ proteins regulate SV coupling and CaV2 . 1 channel abundance involve indirect interactions with protein ( s ) that bind directly to the CaV2 . 1 α1 subunit .
All procedures were performed in accordance with the animal welfare laws of the Max Planck Florida Institute for Neuroscience Institutional Animals Care and Use Committee ( IACUC ) . Stereotactic surgery was performed as described previously ( Chen et al . , 2013; Montesinos et al . , 2015 ) . In brief , Cacna1afl/fl ( floxed ) mice ( Todorov et al . , 2006 ) at P1 were anesthetized by hypothermia . Subsequently , 1–2 µl HdAd ( 1 µl/min ) in storage buffer ( in mM: 10 HEPES , 250 sucrose , 1 MgCl2 at pH: 7 . 4% and 6 . 6% mannitol ) was injected into the aVCN using pulled glass pipettes with a 20 µm opening ( Blaubrand IntraMARK , Wertheim , Germany ) . Two viral vectors , one expressing Cre + eGFP and the other vector one of our CaV2 . 1 α1 constructs + mCherry were co-injected ( Figure 1 ) . The amount of virus injected did not exceed a total of 2*109 viral particles as higher amounts of viral particles have been reported to cause neuronal cell loss ( Muhammad et al . , 2012 ) . To dissipate pressure after injection , the needle was slowly removed after the injection . After full recovery under an infrared lamp at ~37°C , pups were returned to their respective cages with their mother . Acute brainstem slices were prepared as previously described ( Chen et al . , 2013 ) . Briefly , after decapitation of P9-P11 mice of either sex , the brains were immersed in ice-cold low Ca2+artificial cerebrospinal fluid ( aCSF ) containing ( in mM ) : 125 NaCl , 2 . 5 KCl , 3 MgCl2 , 0 . 1 CaCl2 , 10 glucose , 25 NaHCO3 , 1 . 25 Na2HPO4 , 0 . 4 L-ascorbic acid , 3 myo-inositol , and 2 Na-pyruvate , pH 7 . 3–7 . 4 ( 310 mosmol/l ) . Coronal 200 µm slices of the brainstem containing MNTB were obtained using a vibrating tissue slicer ( Campden 7000 smz , Campden Instruments LTD , Loughborough , England ) or Leica VT1200 ( Leica Biosystems , Wetzlar , Germany ) . Slices were immediately transferred to standard aCSF ( 37°C , continuously bubbled with 95% O2 – 5% CO2 ) containing the same as the cutting buffer except that it contained 1 mM MgCl2 and 1–2 mM CaCl2 . After 45 min incubation , slices were transferred to a recording chamber with the same extracellular buffer at room temperature ( RT: 25°C ) . During all experiments , slices were continuously perfused with aCSF and visualized by an upright microscope ( BX51WI , Olympus ) through a 60x water-immersion objective ( LUMPlanFL N , Olympus , Tokyo , Japan ) and a CCD camera ( QI-Click , QImaging , Surrey , BC , Canada ) or a EMCCD camera ( LucaEM S , Andor Technology , Belfast , UK ) . Patch-clamp recordings were performed by using an EPC 10/2 patch-clamp amplifier ( HEKA , Lambrecht , Germany ) , operated by PatchMaster version 2×80 ( Harvard Instruments , Holliston , MA , USA ) . Data were low-pass filtered at 6 kHz and sampled with a rate of 50 kHz . Calyces transduced with HdAd expressing CaV2 . 1 α1 were identified visually with two coexpressed eGFP and mCherry markers . To visualize eGFP and mCherry , slices were illuminated with light of 470 nm or 560 nm , respectively , using a Lumen 200 metal arc lamp ( Prior Scientific , Rockland , MA , USA ) or a Polychrome V xenon bulb monochromator ( TILL Photonics , Gräfelfing , Germany ) . To isolate presynaptic Ca2+ currents , aCSF was supplemented with 1 µM tetrodotoxin ( TTX , Alomone labs , Jerusalem , Israel ) , 100 µM 4-aminopyridin ( 4-AP , Tocris , Bristol , UK ) and 20 mM tetraethylammonium chloride ( TEA , Sigma Aldrich , Darmstadt , Germany ) to block Na+ and K+ conductance . Calyxes were whole-cell voltage-clamped at −80 mV . Current-voltage relationships were recorded in the presence of 1 mM CaCl2 , pharmacological isolation of VGCC subtypes was performed in 2 mM CaCl2 . We used 200 nM ω-agatoxin IVA ( Alomone labs ) to selectively block CaV2 . 1 and 2 µM ω-conotoxin GVIA ( Alomone labs ) for CaV2 . 2 VGCCs . Remaining current was blocked by 50 µM CdCl2 . All experiments to isolate CaV2 subtypes were conducted in presence of cytochrome c ( 0 . 1 mg/ml ) . Presynaptic patch pipettes with open tip diameters 4–6 MΩ resistance were pulled from 2 . 0 mm thin-walled borosilicate glass ( Hilgenberg , Malsfeld , Germany ) and were filled with the following ( in mM ) : 145 Cs-gluconate , 20 TEA-Cl , 10 HEPES , 2 Na2-phosphocreatine , 4 MgATP , 0 . 3 NaGTP , and 0 . 5 EGTA , pH 7 . 2 , 325–340 mOsm ) . Pipettes were coated with Sylgard . Presynaptic series resistance was between 6 and 20 MΩ ( usually between 10–15 MΩ ) and was compensated online to 6 MΩ . Leak and capacitive currents were subtracted online with a P/5 routine . Cells with series resistance >20 MΩ and leak currents >100 pA were excluded from the analysis . For paired recordings , calyx of Held terminals and principal neurons of MNTB were simultaneously whole-cell voltage-clamped at −80 mV and −60 mV , respectively . Patch pipettes were pulled to open tip diameters of 3 . 5–6 MΩ for presynaptic and to 2 . 5–4 MΩ for postsynaptic recordings . Both pipettes were filled with the following: ( in mM ) : 145 Cs-gluconate , 20 TEA-Cl , 10 HEPES , 2 Na2-phosphocreatine , 4 MgATP , 0 . 3 NaGTP , pH 7 . 2 , 325–340 mOsm . To separate the fast and slow release components in the prehearing calyx , 0 . 5 mM EGTA were added in the presynaptic recording pipette ( Sakaba and Neher , 2001 ) . EGTA concentration in the postsynaptic pipette solution was 5 mM . Presynaptic series resistance was between 8 and 25 MΩ ( usually between 10–15 MΩ ) and was compensated online to 8 MΩ . Postsynaptic Rs ( <8 MΩ ) was online compensated to Rs <3 MΩ and remaining Rs was further compensated offline to 0 MΩ for all EPSCs , with a custom routine ( Traynelis , 1998 ) and can be found at ( http://www3 . mpibpc . mpg . de/groups/neher/index . php ? page=software ) . Recordings were performed in aCSF supplemented with 1 mM MgCl2 and 2 mM CaCl2 , cytochrome c ( 0 . 1 mg/ml; Sigma Aldrich ) , 100 µM 4-AP , 1 µm TTX , 50 µM D-AP5 and 20 mM TEA-Cl to isolate presynaptic Ca2+ currents and postsynaptic AMPA receptor-mediated EPSCs . Furthermore , 2 mM kynurenic acid ( Tocris ) and 100 µM Cyclothiazide ( CTZ , Tocris ) were added to prevent saturation and desensitization of AMPA receptors and CaV2 . 1-mediated ICa were isolated by 2 µM ω-conotoxin GVIA ( Alomone labs ) . Cells with series resistance >20 MΩ ( pre ) or >10 MΩ ( post ) and leak currents >100 pA ( pre ) or >200 pA ( post ) were excluded from the analysis . All data was analyzed offline with FitMaster version 2 × 80 ( Harvard Instruments ) , and custom routines written in Igor Pro ( version 6 . 37 , Wavemetrics , Portland , OR , USA ) . Voltage dependence of channel activation was described by both peak and tail currents as functions of voltage and in FitMaster . Peak currents were fitted according to a Hodgkin-Huxley formalism with four independent gates assuming a Goldman-Hodgkin-Katz ( GHK ) open-channel conductance Γ: ( 1 ) I ( V ) =Γ∗1−e−V−Erev25 mV1−e−V25 mV∗ ( 1−eV−VmKm ) −4 with Erev as reversal potential , Vm as half-maximal activation voltage per gate , and km as the voltage-dependence of activation . Tail currents were measured as peaks minus baseline and fitted with a Boltzmann function: ( 2 ) Itail=Ibase+Imin1+e−V−V1/2k where V1/2 represents the half-maximal voltage and k the corresponding slope factor . For EPSC analysis , EPSC amplitudes were measured as peak minus baseline . Synaptic delays in response to step depolarization ( step ) were defined as the duration between the onset of the ICa and the time at which the EPSCs were 50% of their maximum . 10–90 rise times were measured by subtracting the time at 10% of EPSC from 90% of peak amplitude . To estimate the presynaptic ICa charge the presynaptic ICa was integrated . The Ca2+ charges were measured from the onset of the Ca2+ influx to the point where 10% of the peak ICa remained . An established deconvolution approach for the calyx of Held/MNTB synapse was used to estimate quantal release rates and to measure the size of fast and slow vesicle pools ( 30 ms step depolarization ) and fast pool contribution ( 3 ms step depolarization ) ( http://www3 . mpibpc . mpg . de/groups/neher/index . php ? page=software ) ( Neher and Sakaba , 2001a , 2001b; Sakaba and Neher , 2001 ) . This method compensates for residual current , caused by delayed glutamate clearance in the synaptic cleft . After subtracting the estimated residual current , it deconvolves the remaining EPSCs . We determined quantal release rates and time constants of decay by using an empirically generated template miniature EPSC ( mEPSC ) waveform and by further offline analysis in IgorPro ( Wavemetrics ) . Quantal release rates were subsequently integrated to obtain the cumulative release . The fast pool was defined as the cumulative release at 3 ms . For the 30 ms long depolarization step to determine the total releasable pool , the cumulative release rates were further corrected for the refilling of the SV pools during the stimulation , assuming an average refilling rate assumed to be 10 SVs/ms . cDNAs , codon-optimized for expression in mouse ( GeneArt , Regensburg , Germany ) were used for Cre recombinase and CaV2 . 1 α1 subunit cDNA ( Mus musculus , Accession No . : NP_031604 . 3 ) . A series of mutants with deletions increasing in size from the end of the CaV2 . 1 α1 subunit cDNA C-terminal were generated to remove previously described protein interaction sites ( Butz et al . , 1998; Davydova et al . , 2014; Hibino et al . , 2002; Kaeser et al . , 2011; Maximov et al . , 1999; Wong et al . , 2013 , 2014 ) . Subsequently , each CaV2 . 1 expression cassette was cloned into the AscI site of a modified version of pdelta 28E4 , gift from Dr . Philip Ng ( Palmer and Ng , 2003 ) using InFusion ( Clontech , Takara Bioscience , Mountain View , CA , USA ) . This version of pdelta28E4 has been altered by removal of 5 kb stuffer sequence and the addition of a separate neurospecific mCherry expression cassette that is driven by the 470 bp hSyn promoter . The final HdAd plasmid allows for expression of CaV2 . 1 independently of mCherry as a dual expression recombinant Ad vectors similar to the strategy used with second generation rAd ( Montesinos et al . , 2015 , 2016; Young and Neher , 2009 ) . For HdAd Cre , the Cre recombinase cDNA was cloned into the AscI site of a different version of pdelta28E4 that has been modified to also contain a separate neurospecific EGFP expression cassette that is driven by the 470 bp hSyn promoter and the final HdAd plasmid allows for expression of Cre independently of EGFP . Production of HdAd was carried out as previously described ( Montesinos et al . , 2016; Palmer and Ng , 2003; Palmer and Ng , 2011 ) . Briefly , pHAD plasmid was linearized with PmeI and then transfected ( Profection Mammalian Transfection System , Promega , Madison , WI , USA ) into 116 producer cells , a derivative of 293N3S , developed for specifically for large scale HdAd production ( Palmer and Ng , 2003 ) . We did not test for mycoplasma contamintation . Helper virus ( HV ) was added the following day . Forty-eight hours post infection , after cytopathic effects have taken place , cells were subjected to three freeze/thaw cycles for lysis and release of the viral particles . To increase the HdAd titer , this lysate was amplified in a total of five serial coinfections of HdAd and HV from 3 × 6 cm tissue culture dishes followed by a 15 cm dish and finally 30 × 15 cm dishes of 116 cells ( confluence ~90% ) . HdAd was purified by CsCl ultracentrifugation . HdAd was stored at −80°C in storage buffer ( 10 mM Hepes , 1 mM MgCl2 , 250 mM sucrose , pH 7 . 4 ) . Mice ( P9-11 ) were anesthetized with Avertin ( 250 mg/kg of body weight , i . p . ) and perfused transcardially with warm phosphate-buffered saline ( PBS , in mM: 150 NaCl , 25 Na2HPO4 , 25 NaH2PO4 , pH 7 . 4 ) followed by warm fixative solution for 7–9 min containing 4% paraformaldehyde ( PFA ) , 0 . 5% glutaraldehyde , and 0 . 2% picric acid solved in phosphate buffer ( PB , in mM: 100 Na2HPO4 , 100 NaH2PO4 , pH 7 . 4 ) . Brains were postfixed with 4% PFA in PB for overnight and 50 µm coronal sections of the brainstem were obtained on a vibratome ( Leica VT1200 ) . Expression of EGFP at the calyx of Held was visualized using an epifluorescence inverted microscope ( CKX41 , Olympus ) equipped with XCite Series 120Q lamp ( Excelitas technologies , Wiesbaden , Germany ) and only those samples showing EGFP were further processed as follows . After washing with PB several times , sections were cryoprotected with 10% and 20% sucrose in PB for 1 hr each , followed by 30% sucrose in PB for 2 hr and submersed into liquid nitrogen for 1 min , then thawed at room temperature . Afterwards , sections were incubated in a blocking solution containing 10% normal goat serum ( NGS ) , 1% fish skin gelatin ( FSG ) , 0 . 05% Na3N in 50 mM Tris-buffered saline ( TBS , in mM: 150 NaCl , 50 Tris , pH 7 . 3 ) for 1 hr , and incubated with an anti-GFP antibody ( 0 . 1 µg/ml , ab6556 , Abcam , Cambridge , UK ) diluted in TBS containing 1% NGS and 0 . 1% FSG at 4°C for 48 hr . After washing with TBS , sections were incubated for overnight in nanogold-conjugated goat anti-rabbit IgG ( 1:100 , Cat . No . 2003 , Nanoprobes , Yaphank , NY , USA ) diluted in TBS containing 1% NGS and 0 . 1% FSG . Immuno-labeled sections were washed in PBS , briefly fixed with 1% glutaraldehyde in PBS , washed in PBS followed by MilliQ-H2O , and silver intensified for 6–8 min using HQ silver intensification kit ( Nanoprobe ) . After washing with PB , sections were briefly rinsed with H2O and treated with 0 . 5% OsO4 in 0 . 1M PB for 20 min , en-bloc stained with 1% uranyl acetate for 25 min , dehydrated in a graded series of ethanol , acetone , and propylene oxide , and flat embedded in Durcupan resin ( Sigma-Aldrich ) . After trimming out the MNTB region , ultrathin sections were prepared with 40 nm-thickness using an ultramicrotome ( EM UC7 , Leica ) . Sections were counterstained with uranyl acetate and lead citrate , and examined in a Tecnai G2 Spirit BioTwin transmission electron microscope ( FEI ) at 100 kV acceleration voltage . Images were taken with a Veleta CCD camera ( Olympus ) operated by TIA software ( FEI ) . Images used for quantification were taken at 60 , 000x magnification . All TEM data were analyzed using Fiji imaging analysis software ( http://fiji . sc/Fiji ) ( Schindelin et al . , 2012 ) . Positive calyces were identified by the existence of gold particles , and compared to contralateral nontransduced calyces . Each presynaptic AZ was defined as the membrane directly opposing postsynaptic density , and the length of each one was measured . Vesicles within 200 nm from each AZ were manually selected and their distances relative to the AZ were calculated using a 32-bit Euclidean distance map generated from the AZ . For data analysis , vesicle distances were binned every 5 nm and counted . Vesicles less than 5 nm from the AZ were considered ‘docked’ ( Taschenberger et al . , 2002; Yang et al . , 2010 ) and their numbers were averaged per animal . Three animals for each condition were analyzed . For both CaV2 . 1Δ2016–2368 and CaV2 . 1Δ2042–2368 , at least 100 individual AZs were analyzed and compared with the same number of AZs of respective calyces from the contralateral MNTB ( nontransduced in-slice control AZ ) . All statistical tests were conducted in Prism 6 ( GraphPad Software ) . Sample sizes for all experiments were chosen based on assuming a population with a normal distribution , a sample size of seven is sufficient to invoke the Central Limit Theorem . All data were tested for normal distribution by performing a Shapiro-Wilk test for normality and variances of all data were estimated and compared using Bartlett’s test . Electrophysiological data were compared with one-way analysis of variance ( ANOVA ) with a post hoc Dunnett’s test , always using Full transcript rescue as control dataset . Patch clamp recordings lacking proper clamp quality and with high leak were excluded from data sets . EM-data were compared using an unpaired t-test . Statistical significance was accepted at *p<0 . 05; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . In Figures and Tables , data are reported as mean ± SEM , unless otherwise stated . | The points of contact between nerve cells are called synapses , and nerve cells communicate across synapses via chemicals known as neurotransmitters . These chemical messengers are initially stored within bubble-like packages called synaptic vesicles that are released after they fuse with the membrane of the nerve cell at a specialized site referred to as the “active zone” . Calcium ions are one of the major factors that lead to the release of synaptic vesicles . Ion channel proteins in the membrane of the nerve cell control the flow of calcium ions into the cell . There are often many different ion channels at a synapse , but one type called CaV2 . 1 most effectively triggers the release of synaptic vesicles when a nerve impulse reaches the synapse . Various proteins at the active zone can bind directly to parts of the CaV2 . 1 channel that are identified by a short sequence of amino acids – the building blocks of all proteins . Several researchers have proposed that the interactions with some of these short sequences , which are also known as motifs , control how much of this ion channel is in the synapse and how it interacts with synaptic vesicles to regulate the release of neurotransmitters . However , other researchers do not agree with this proposed explanation . Lübbert , Goral et al . set out to determine which parts in a specific part of the CaV2 . 1 channel ( called the “α1 subunit C-terminus” ) are critical for its interaction with synaptic vesicles . The experiments revealed a new motif that regulates how many synaptic vesicles could be released in response to electrical impulses travelling along nerve cells from mice . The same motif also regulates the total number of synaptic vesicles at the active zone . Lübbert , Goral et al . went on to show that binding to known active proteins at most played a minor role in controlling the abundance of the CaV2 . 1 channels and how close they were to the synaptic vesicles . As such , these findings counter prevailing views of the roles of certain motifs in the α1 subunit of the CaV2 . 1 channel . Thus , it may be necessary to re-think how the CaV2 . 1 channel regulates the release of synaptic vesicles . Ion channels are vital to the activity of all nerve cells , and working out how the numbers and organization of CaV2 . 1 and related ion channels are regulated will be fundamental to understanding how information is encoded in brain . In addition , problems with these kinds of ion channel may result in disorders such as migraines and epilepsy . Therefore , the new findings may help to guide further studies investigating possible ways to treat these disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"neuroscience"
] | 2017 | A novel region in the CaV2.1 α1 subunit C-terminus regulates fast synaptic vesicle fusion and vesicle docking at the mammalian presynaptic active zone |
Social impairments are a hallmark of Autism Spectrum Disorders ( ASD ) , but empirical evidence for early brain network alterations in response to social stimuli is scant in ASD . We recorded the gaze patterns and brain activity of toddlers with ASD and their typically developing peers while they explored dynamic social scenes . Directed functional connectivity analyses based on electrical source imaging revealed frequency specific network atypicalities in the theta and alpha frequency bands , manifesting as alterations in both the driving and the connections from key nodes of the social brain associated with autism . Analyses of brain-behavioural relationships within the ASD group suggested that compensatory mechanisms from dorsomedial frontal , inferior temporal and insular cortical regions were associated with less atypical gaze patterns and lower clinical impairment . Our results provide strong evidence that directed functional connectivity alterations of social brain networks is a core component of atypical brain development at early stages of ASD .
Early preferential attention to social cues is a fundamental mechanism that facilitates interactions with other human beings . During the third trimester of gestation , the human foetus is already sensitive to both voices ( DeCasper and Spence , 1986 ) and face-like stimuli ( Reid et al . , 2017 ) . Newborns orient to biological motion ( Simion et al . , 2008 ) and prefer their mothers’ voices to those of other females ( DeCasper and Fifer , 1980 ) . Infants as young as two weeks imitate faces and human gestures ( Meltzoff and Moore , 1977 ) . The orientation to and interaction with social cues during infancy drives the later acquisition of social communication skills during toddler and preschool years . As function of experience , the repeated exposure leads to the progressive emergence of adaptive interactions with conspecifics . Alongside , the brain develops a network of cerebral regions specialized in understanding the social behaviours of others . This network includes the orbitofrontal and medial prefrontal cortices , the superior temporal cortex , the temporal poles , the amygdala , the precuneus , the temporo-parietal junction , the anterior cingulate cortex ( ACC ) and the insula ( Brothers , 1990; Frith and Frith , 2010; Adolphs , 2009; Blakemore , 2008 ) . Collectively , these areas form the social brain and are all implicated to some extent in processing social cues and encoding human social behaviours ( Brothers , 1990; Frith and Frith , 2010; Adolphs , 2009; Blakemore , 2008 ) . Autism is a life-long lasting , highly prevalent neurodevelopmental disorder that affects core areas of cognitive and adaptive function , communication and social interactions ( Christensen et al . , 2016 ) . A common observation in infants later diagnosed with ASD is the presence of less sensitivity and diminished preferential attention to social cues during the first year of life ( Osterling and Dawson , 1994 ) . Toddlers with ASD orient preferentially to non-social contingencies ( Klin et al . , 2009 ) . Indifference to voices ( Sperdin and Schaer , 2016 ) and faces ( Grelotti et al . , 2002 ) in ASD leads to deficits in the development of adapted social interactions with others and to difficulties in understanding human behaviours . It is not established why children with ASD show insensitivity to stimuli with social contingencies at early developmental stages , but this apparent indifference to social cues ultimately hinders the normal development of the social brain network or parts thereof ( Pelphrey et al . , 2011; Gotts et al . , 2012 ) . Some authors propose that deficits in the development of social cognition ( such as learning to attribute mental states to others , 'theory of mind' [Frith , 1989] ) and/or in sensory processing ( Dinstein et al . , 2012 ) prevent children with ASD to actively and appropriately engage with social stimuli . Another hypothesis suggests that they have difficulties building up stimulus-reward contingencies for social stimuli , due to a reduced motivation to attend and engage with them . Regardless of the reasons behind reduced social orienting , diminished interaction and exposure to social stimuli may in turn impede the development of the social brain at early developmental stages in ASD ( Chevallier et al . , 2012; Dawson et al . , 2004 ) . Evidence remains limited for brain network alterations in response to socially meaningful stimuli in ASD during the period spanning the toddler to preschool years , partly because the acquisition of data during that age period is extremely challenging ( Raschle et al . , 2012 ) . However , studying very young children with ASD , closer to their diagnosis , is all the more important when recent findings suggest the presence of major developmental changes in large-scale brain networks between adults and younger individuals with ASD ( Nomi and Uddin , 2015 ) . Currently , it remains unclear how autism affects the development of the functional brain networks implicated in the processing of socially meaningful information at early developmental stages . A better delineation of the timing and nature of the neurodevelopmental alterations associated with core social deficits in autism may in turn help to improve therapeutic strategies . Electroencephalography ( EEG ) is as a powerful non-invasive method to study atypical brain responses to social stimuli in clinical paediatric populations with ASD . For example , surface-based experiments have reported aberrant evoked potentials in response to dynamic eye gaze in infants at high-risk for ASD ( Elsabbagh et al . , 2012 ) or to speech stimuli in toddlers with ASD ( Kuhl et al . , 2013 ) with differences in resting EEG power in infants at high-risk for ASD ( Tierney et al . , 2012 ) . Whilst useful , most of the EEG experiments performed on very young children with ASD ( younger than four years ) have been done with few electrodes only and the analysis restricted to the sensor space . Therefore , hypothetical alterations in the functional brain networks underlying the processing of social stimuli remain to be determined for that age period in ASD . Here , we recorded high-density EEG and high resolution eye-tracking in toddlers and preschoolers with ASD and their TD peers as they watched naturalistic and ecologically valid dynamic social movies . Using data-driven methods , we first investigated whether the visual exploration behaviour was atypical in toddlers and preschoolers with ASD using kernel density distribution estimations . Then , we explored whether their ongoing source-space directed functional connectivity was altered compared to their TD peers using Granger-causal modelling applied to the EEG source signals . This method estimates brain connectivity in the frequency domain . It identifies which brain regions are the key drivers of information flow in a brain network and directional relationships between brain regions that belong to a network ( Coito et al . , 2016b ) . This approach has been applied to study connectivity alterations using intra-cranial recordings ( Wilke et al . , 2009; van Mierlo et al . , 2013; van Mierlo et al . , 2011 ) as well as source imaging based on EEG recordings in clinical populations ( Ding et al . , 2007; Coito et al . , 2016a; Coito et al . , 2015; Coito et al . , 2016b ) and in healthy human participants ( Astolfi et al . , 2007; Hu et al . , 2012; Plomp et al . , 2015b ) . Finally , we looked for relationships between directed functional connectivity measures , visual exploration behaviour and clinical phenotype . As toddlers with ASD have less preferential attention for social cues , we hypothesized that they would show both a different visual exploration behaviour of the dynamic social images and altered directed functional connectivity patterns in brain regions involved in processing social information compared to their TD peers .
The summed outflow ( i . e . - the amount of information transfer ) is a measure that reflects the importance ( i . e . -the amount of driving ) of a given region of interest ( ROI ) in the network ( see Materials and methods section ) . To understand the functional wiring and the dynamic flow underlying the processing of the dynamic social stimuli , we used a data-driven method to explore in which frequency band the highest summed outflow occurred in 82 ROIs across the whole brain . A ROI with a strong summed outflow has a key role in directing the activity towards other ROIs in the network . The strongest summed outflow across the whole brain occurred in the theta band ( 4–7 Hz ) in both groups . The summed outflow of the largest drivers across frequencies is illustrated for each group in Figure 1a . As can be seen , the largest peaks of activity are present in the theta band range ( 4–7 Hz ) in both groups followed by peaks of activity in the alpha band range ( 8–12 Hz ) . The global driving in the theta and alpha bands did not differ between groups ( theta : df = 34 , t = 0 . 6201 , p = 0 . 536; alpha : df = 34 , t = 0 . 1736 , p = 0 . 8632 ) . Driving in the theta band was higher compared to the driving in the alpha band in both groups ( ASD: df = 17 , t = 11 . 86 , p < 0 . 0001; TD:df = 17 , t = 8 . 025 , p < 0 . 0001 ) . Several regions common to both groups showed a large driving ( summed outflow ) in both frequency bands , and notably the bilateral medial frontal and superior orbitofrontal regions , the bilateral hippocampi , the bilateral ACC and the right amygdala ( Figure 1b ) . Thereafter , we characterized the differences in the summed outflow across all brain regions between the groups in the theta band and in the alpha band , separately . For the theta band , we identified six ROIs that showed a statistically higher driving ( stronger summed outflow ) in the ASD group in comparison to the TD group ( Mann-Whitney-Wilcoxon test , two-tailed , p<0 . 05 ) : the right orbital part of the superior frontal gyrus ( Ws = 267 , z = −2 . 088 . p=0 . 037 , r = −0 . 348 ) , the bilateral orbital parts of the middle frontal gyri ( Left: Ws = 259 , z = −2 . 341 , p=0 . 019 , r = −0 . 39; Right: Ws = 252 , z = −2 . 563 , p=0 . 01 , r = −0 . 427 ) , the right middle cingulate gyrus ( Ws = 259 , z = −2 . 341 , p=0 . 019 , r = −0 . 390 ) , the left superior occipital gyrus ( Ws = 270 , z = - 1 . 993 , p=0 . 047 , r = −0 . 332 ) , and the left superior temporal gyrus ( STG ) ( Ws = 255 , z = −2 . 468 , p=0 . 013 , r = −0 . 411 ) ( Figure 2a ) . This indicates the presence of a stronger driving from these regions in the toddlers and preschoolers with ASD compared to their TD peers when viewing the dynamic social images . For the alpha band , we identified three ROIs that had a different driving in the ASD group in comparison to the TD group ( Mann-Whitney-Wilcoxon test , two-tailed , p<0 . 05 ) . The the right orbital part of the middle frontal gyrus ( Ws = 262 , z = −2 . 246 , p=0 . 024 , r = −0 . 374 ) and the left cuneus ( Ws = 265 , z = −2 . 151 , p=0 . 031 , r = −0 . 358 ) had a higher driving and the right STG had a weaker driving ( Ws = 265 , z = −2 . 151 , p=0 . 031 , r = −0 . 358 ) ( Figure 2a ) . The boxplots with the summed outflow values for each group and for each of the significant ROIs are displayed in Figure 2b for theta and Figure 2c for alpha . We looked for differences in the region-to region directed functional connectivity using Granger-causal modelling ( see Materials and methods section ) from each of the six nodes for the theta band , and from each of the three nodes in the alpha band separately in both groups . In the theta band , all the connections from the six ROIs in the toddlers and preschoolers with ASD were stronger than the strongest connections in the TD participants ( Mann − Whitney − Wilcoxon , two − tailed , p < 0 . 05 , Benjamini − Hochberg = 0 . 05 ) . This suggests the presence of hyper-connectivity in the toddlers and preschoolers with ASD in theta . The region-to-region directed functional connectivity from the six ROIs in theta is illustrated in Figure 3 . The estimation of the region-to-region directed connectivity ( i . e . to which other ROIs the activity was directed ) also revealed different network patterns for all the six ROIs in the toddlers and preschoolers with ASD compared to their TD peers . The boxplots of the outflow values of the connections from the right orbital part of the superior frontal gyrus seed region are provided in Figure 3—figure supplement 2 for the ASD group and Figure 3—figure supplement 2 for the TD group . In the alpha band , the region-to-region directed functional connectivity analysis revealed stronger connections from the right orbital part of the middle frontal gyrus and the left cuneus whereas the right STG had weaker connection in the toddlers and preschoolers with ASD compared to their TD peers ( Mann-Whitney-Wilcoxon , two-tailed , p<0 . 05 , Benjamini-Hochberg = 0 . 05 ) . Similarly to what we found in the theta band , all three significant ROIs in the alpha band had different network patterns in the toddlers and preschoolers with ASD compared to their TD peers ( Figure 4 ) . We further explored associations between the summed outflow in the theta and alpha bands from the ROIs and clinical and behavioural phenotypes ( Spearman − rho , two − tailed , p <0 . 05 , Benjamini − Hochberg = 0 . 05 ) . None of the correlations between the summed outflow and ADOS-2 severity scores survived False discovery rate ( FDR ) correction for either bands ( Benjamini − Hochberg = 0 . 05 ) . For the summed outflow in the theta band , we found strong positive correlations between the summed outflow in the right lingual area and standard scores from the socialization domain ( rs = 0 . 751 , N = 18 , p=0 . 0003 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) as well as with standard scores from the leisure and play skills subdomain of the VABS-II ( rs = 0 . 802 , N = 18 , p=0 . 0001 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) . None of the correlations between the summed outflow and VABS-II standard scores survived FDR correction for the alpha band ( Benjamini-Hochberg = 0 . 05 ) . Higher summed outflow within the left Heschl area near the posterior convolutions of the insula and the left rolandic operculum near the circular sulcus of the insula rostrally were positively related to better fine ( rs = 0 . 745 , N = 18 , p=0 . 0004 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) and gross motor skills ( rs = 0 . 744 , N = 18 , p=0 . 0004 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) as measured by the PEP-3 . For the alpha band , higher summed outflow within the left hippocampus and the left rolandic operculum were positively related to better fine ( rs = 0 . 736 , N = 18 , p=0 . 0005 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) and gross motor skills ( rs = 0 . 737 , N = 18 , p=0 . 0005 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) as measured by the PEP-3 . The toddlers and preschoolers with ASD with a gaze pattern similar to their TD peers showed an increased driving in theta within the left middle cingulate cortex ( rs = 0 . 726 , N = 18 , p=0 . 0007 , two-tailed , <0 . 05; Benjamini-Hochberg = 0 . 05 ) and the right paracentral lobule ( rs = 0 . 738 , p = 0 . 0005 , two − tailed , <0 . 05; Benjamini − Hochberg = 0 . 05 ) . There was no significant relationship between the Proximity Index ( see Materials and methods section ) and the summed outflow in the alpha band after FDR correction . The significant correlations between the summed outflows and the Proximity Index , VABS-II standard scores and PEP-3 standard scores for each frequency band are displayed in Figure 5 . Finally , we explored associations between gaze performance with developmental scores obtained from the PEP-3 and with adaptive scores obtained from the VABS-II ( D'Agostino-Pearson omnibus normality test , K2 , p<0 . 05; Pearson’s r , two-tailed , p<0 . 05 ) . We didn't find any significant correlations between the Proximity Index and the global level of autistic severity as measured with the calibrated ADOS-2 severity score . However , we found that the toddlers and preschoolers with ASD with a better gaze performance had better global adaptive functioning as measured by the VABS-II ( K2 = 3 . 339 , p=0 . 1883; r = 0 . 578 , p=0 . 012 ) , which was driven by better global ( K2 = 0 . 8179 , p=0 . 6643; r = 0 . 575 , p=0 . 013 ) and fine ( K2 = 0 . 5438 , p=0 . 7619; r = 0 . 509 , p=0 . 031 ) motor skills , and better development of interpersonal relationships ( K2 = 5 . 308 , p=0 . 0704; r = 0 . 581 , p=0 . 011 ) . We also found that these children with a better gaze performance had better visual motor imitation skills ( K2 = 2 . 671 , p=0 . 263; r = 0 . 534 , p=0 . 022 ) as measured by the PEP-3 .
Abnormal processing of social cues is a hallmark of ASD ( Chevallier et al . , 2012; Dawson et al . , 2004; Dichter et al . , 2009; Elsabbagh et al . , 2012; Gotts et al . , 2012; Greene et al . , 2011; Klin et al . , 2009; Pelphrey et al . , 2011 ) . However , evidence for alterations of social brain networks at early stages of ASD is scant . Using data-driven methods , we observed aberrant gaze patterns together with frequency specific alterations in the directed functional connectivity in toddlers and preschoolers with ASD when exploring dynamic social stimuli compared to their TD peers . These differences manifested as increased driving and hyper-connectivity in the theta frequency band from nodes that include the right orbital part of the superior frontal gyrus , the bilateral orbital parts of the middle frontal gyri , the right middle cingulate gyrus , the left superior occipital gyrus and the left STG . For the alpha band , we found increased driving from the right orbital part of the middle frontal gyrus and the left cuneus and decreased connectivity from the right STG . To the best of our knowledge , this is the first evidence indicating concomitant alterations in the visual exploration of dynamic social images and in the directed functional connectivity involving key nodes of the social brain ( Brothers , 1990; Frith and Frith , 2010; Adolphs , 2009; Blakemore , 2008 ) at early stages of ASD . The results indicate that the highest information transfer ( i . e . summed outflow ) occurs at the global brain level in the theta band ( 4–7 Hz ) followed by the alpha frequency band ( 8–12 Hz ) . As such , our data-driven approach reveals high information transfer in physiologically relevant frequency bands given the young age of our participants . These are , namely , prominent brain rhythms during infancy and toddlerhood ( Saby and Marshall , 2012; Orekhova et al . , 2006 ) . Throughout development , slow waves modulate attentional brain states , encode specific information and ease communication between neuronal populations ( Lopes da Silva , 2013 ) . Theta and alpha bands are thought to underlie different cerebral functions , but are closely related ( Klimesch , 1999 ) . During infancy and early childhood , modulations in alpha band activity have been associated with the progressive development of visual attentional networks and inhibition of task-irrelevant brain areas ( Orekhova et al . , 2001; Stroganova et al . , 1999 ) , while theta is thought to play a functional role in memory formation , emotional and cognitive functioning ( Saby and Marshall , 2012; Orekhova et al . , 2006 ) . In our experiment , the highest information transfer occurred in the theta band . In very young children , theta modulations have been related to the development of the social brain . For example , surface-based EEG experiments in TD infants report enhanced theta power to social versus non-social stimuli at 12 months ( Jones et al . , 2015 ) . Theta increases during attention to social stimulation in infants and preschool aged children ( Orekhova et al . , 2006 ) . Hence , social contingencies modulate theta band activity . Similarly , our results show high information transfer in theta when toddlers and preschoolers visually explore dynamic social stimuli . The development of attentional processes in young children has been associated with modulations in alpha band activity ( Orekhova et al . , 2001 ) . Conversely , our young participants had to deploy their attentional focus to the dynamic social stimuli . This would explain why high information transfer was also found in the alpha frequency band . EEG experiments in individuals with ASD show a reduction or an increase in coherence patterns in the theta and/or alpha frequency bands compared to their TD peers at various ages and under different experimental conditions ( Schwartz et al . , 2017 ) . However , most of the available EEG experiments performed on very young children with ASD and analysis thereof were so far restricted to the scalp surface . As a result , information remains limited regarding the presence of frequency specific alterations within brain regions when young children are exposed to social stimuli . Here , our data-driven source-space approach revealed not only high information transfer in the theta and alpha frequency bands , but also , the involvement of the bilateral medial and the superior orbital frontal regions , the bilateral hippocampi , the bilateral ACC and right amygdala . These areas are implicated in processing social cues and encoding human social behaviours ( Brothers , 1990; Frith and Frith , 2010; Adolphs , 2009; Blakemore , 2008 ) . Our results further indicate the presence of frequency specific alterations in the driving from several brain areas in the toddlers and preschoolers with ASD compared to their TD peers . In the theta band , we found an overall dominant higher driving within several frontal and the cingulate regions . In TD individuals , theta generates within the frontal cortices or the ACC ( Asada et al . , 1999 ) . In comparison to their TD peers , both these areas develop differently in young toddlers later diagnosed with ASD ( Schumann et al . , 2010 ) . Accordingly , our results raise the possibility that the brain regions generating theta also follow a different development in the toddlers and preschoolers with ASD . For the alpha band , alteration in the driving was evident from the right orbital part of the middle frontal gyrus . Although experiments during the first three years of life are currently sparse , increased alpha-range EEG connectivity over frontal and central electrodes has recently been reported in high-risk infants who were diagnosed with ASD at 36 month ( Orekhova et al . , 2014 ) . A magnetoencephalography ( MEG ) study performed at rest using a source-space approach reported lower coherence in the theta and alpha bands within parietal and occipital regions but their ASD group only included adolescents ( Ye et al . , 2014 ) . The differences between this specific study and our could stem from either variations in the age groups ( adolescents versus toddlers and preschoolers ) , the stimuli used ( grey cross inside a white circle versus dynamic biological visual stimuli ) or the methods . More generally , several factors explain discrepancies in brain connectivity results between studies . The type of connectivity measures applied , the approach ( surface versus source based ) , the brain regions analysed and frequency bands examined are variables that influence the results or the age of the participants ( Mohammad-Rezazadeh et al . , 2016 ) . Frontal and cingulate areas have been implicated in various complex aspects of social cognition , social reward , social perception and social behaviour ( Jonker et al . , 2015; Apps et al . , 2013 ) . Metabolic changes within the medial prefrontal cortex and the cingulate cortex are correlated with social interaction impairments in childhood ASD ( Ohnishi et al . , 2000 ) . Several experiments report structural ( Patriquin et al . , 2016 ) and functional ( Gotts et al . , 2012; Patriquin et al . , 2016; Greene et al . , 2011; Cheng et al . , 2015 ) alterations within these brain areas in school aged children , adolescents and adults with ASD when compared to their TD peers . A recent study described hyper-connectivity within the ACC and bilateral insular cortices in a sample including children aged between seven to 12 years ( Uddin et al . , 2013a ) . Some authors propose that the two together form the salience network , whose role is to direct attention to behaviourally-relevant stimuli ( Menon and Uddin , 2010 ) . Although we didn’t find differences in the driving from the insula compared to the TD peers , there is an increasing number of evidence showing an abnormal development of the salience network or components thereof in ASD ( Uddin , 2015 ) , which may partially explain the limited interest for and engagement with social stimuli that is often observed in individuals with ASD and that constitutes a hallmark of the disorder ( Klin et al . , 2009; Pelphrey et al . , 2011 ) . Accordingly , the toddlers and preschoolers with ASD had a different visual exploration behaviour of the dynamic social stimuli raising the possibility of a reduced interest to visually engage with them . Alternatively , alterations in driving from these regions could partially reflect a reduced motivation to attend and engage with the dynamic social stimuli ( Chevallier et al . , 2012; Dawson et al . , 2004 ) . The alterations in the driving in the alpha band we found here , could also be related to the presence of developmental impairments in attentional networks and/or inhibitory functions ( Keehn et al . , 2013 ) . We found higher driving in theta from the left superior temporal and occipital gyri . In the alpha band , we found alterations in the driving from a node in the right STG and the left cuneus in the occipital lobe . Those brain areas are implicated in the processing of biological motion , in analysing the intentions of other people’s actions and self-reflection ( Pelphrey and Carter , 2008; Pelphrey et al . , 2005; Pelphrey and Morris , 2006; Pelphrey et al . , 2004 ) . Our result would suggest that the exploration of the dynamic social visual stimuli that contained biological movements led to altered driving from these brain areas in both frequency bands in the toddlers and preschoolers with ASD compared to their TD peers . Overall hyper-connectivity seems prevalent in younger populations ( that is , infants at high-risk for ASD , toddlers and preschoolers with ASD ) while hypo-connectivity is more observed during adolescence and adulthood in ASD ( Uddin et al . , 2013b ) . Conversely , a developmental shift occurs in brain growth with an initial period of early brain overgrowth followed by normalization sometime during adolescence ( Courchesne et al . , 2011 ) . Accordingly , structural white matter connectivity studies also highlight this shift from higher structural connectivity in very young children with ASD to lower connectivity in older children with ASD ( Hoppenbrouwers et al . , 2014; Conti et al . , 2015 ) . Therefore , a global higher-driving and hyper-connectivity from key nodes of the social brain in the theta frequency band in our ASD group is consistent with reports in the literature when considering the very young age of our participants ( around 3 years of age on average ) . For the alpha frequency band , we found alterations in the driving manifesting as both hyper-connectivity from a frontal and an occipital area and under-connectivity from a node in the superior temporal pole . Hence , we found frequency-specific network alterations with distinct patterns of directed functional connectivity in the toddlers and preschoolers with ASD compared to their TD peers . This result is in line with recent experiments indicating the presence of distinct patterns of hyper- and hypo-connectivity between brain regions functionally defined by neural oscillatory activity in children and adolescents with ASD ( Ye et al . , 2014; Kitzbichler et al . , 2015; Datko et al . , 2016 ) . We further explored associations between the driving in the nodes of the network ( that is , the summed outflow ) and clinical and behavioural phenotypes for each frequency band . We didn’t observe any significant relationships between summed outflow and the ADOS severity scores after FDR correction in either frequency bands . In theta , we observed an increased driving from the median cingulate cortex and the paracentral lobule in the toddlers and preschoolers with ASD who had a more similar visual exploration pattern to their TD peers . Thus , an improved visual exploration pattern of the dynamic social images was related to increased summed outflow in theta from these regions . Moreover , higher summed outflow from the right lingual area was related to better socialization behaviour and leisure and play skills as measured by the VABS-II . Higher summed outflow from the left Heschl's area near the posterior convolutions of the insula and the left rolandic operculum near the circular sulcus of the insula rostrally were positively related to better fine and gross motor skills as measured by the PEP-3 . Finally , for the alpha band , we found that higher driving within the left hippocampal area and the left rolandic operculum were positively related to better fine and gross motor skills as measured by the PEP-3 . As such , overall increased activity in the theta band within dorsomedial frontal , inferior temporal and insular cortical regions were associated with lower clinical impairment and less atypical gaze patterns whereas increased driving in the alpha band was selectively associated with better motor performance . The presence of hyperactivity within relevant brain region has been interpreted as a possible compensatory mechanism when performing a social target detection task , in adults with ASD at least ( Dichter et al . , 2009 ) . While to the best of our knowledge , there is currently no other relevant experimental data that addresses this question , we speculate that the overall hyper-driving from these relevant brain regions might be a mechanism to compensate for atypical development of the brain's circuitry over time as higher directed functional connectivity was related improved socialization , motor behaviours and better visual exploration of dynamic social images . However , longitudinal measurements are necessary to fully confirm this interpretation . De facto , the toddlers and preschoolers with ASD who had better gaze performance had better adaptive behaviour , improved global and fine motor skills and enriched interpersonal relationships as measured by the VABS-II and better visual motor imitation skills as measured by the PEP-3 . They were also those who had higher summed outflow in several relevant brain areas . Beyond functional and structural brain alterations reported elsewhere in older children and adults with ASD , our results suggest for the first time , the presence of frequency specific alterations in the driving of information flow from brain areas implicated in social information processing during the viewing of naturalistic dynamic social images in toddlers and preschool with ASD . Furthermore , we show that these frequency specific directed functional connectivity network alterations within regions of the social brain are present at early stages of ASD , justifying further investigation into how early therapeutic interventions targeting social orienting skills may help to remediate social brain development during this critical age period when plasticity is still possible . Longitudinal experiments on very young children with ASD are critically needed to better delineate modulations in brain patterns at the time of diagnosis , and how these alterations are influenced by therapeutic intervention . The present experiment is a first step towards that direction .
Recruitment of toddlers and preschoolers with ASD was achieved via clinical centres specialized in ASD and French-speaking parent associations . TD toddlers and preschoolers were recruited via announcements in the Geneva community . Prior to the experiments , all the procedures were approved by the Ethics Committee of the Faculty of Medicine of the University of Geneva Hospital in accordance with the ethical standards proclaimed in the Declaration of Helsinki . For all participants , an interview over the phone and a medical developmental history questionnaire were completed before their initial visit . All participants' parents gave their informed consent prior to inclusion in the study . 120 participants were recruited for the experiment . We did not manage to put the EEG cap on the head of 23 ASD and 7 TD participants . We managed to put the cap on 90 participants . Out of those , we excluded 28 ASD and 26 TD participants because of too many movement-related artefacts , unrepairable noisy signal , lack of interest , or insufficient amounts of epochs available for subsequent analysis . This was to be expected given the extremely sensitive population at study here . As a result , 36 participants were included: 18 young children with ASD ( 2 females; mean age 3 . 1 years ± 0 . 8 , range 2 . 2–4 . 4 ) and 18 age matched ( df = 34 , t = 2 . 72 , p=0 . 852 ) TD peers ( 5 females; mean age 3 . 1 years ± 0 . 9 , range 2 . 0–4 . 8 ) . All participants with ASD included in the study received a clinical diagnosis prior to their inclusion in the research protocol . Diagnosis of ASD was rigorously verified and confirmed with either the Autism Diagnosis Observation Schedule-Generic ( Lord et al . , 2000 ) or the Autism Diagnosis Observation Schedule , second edition ( ADOS-2 ) ( Luyster et al . , 2009 ) . The latter contains a toddler module that defines concern for ASD . ADOS assessments were administered and scored by experienced clinicians working at the institution and specialized in ASD identification . In order to compare scores from different modules , we transformed the ADOS-G scores into Calibrated Severity Scores ( ADOS-CSS ) ( Gotham et al . , 2009 ) . For the participants that underwent the ADOS-2-toddler module , we calibrated the scores into Severity Scores ( Esler et al . , 2015 ) . Five children under 30 months of age performed the toddler module of the ADOS-2 . All scored in the moderate to severe range of concern for ASD . For all the participants younger than 3 years of age ( n = 10 ) at the EEG acquisition , clinical diagnosis was confirmed after one year by a clinician specialized in ASD identification using the ADOS-G or ADOS-2 . The mean global ADOS-CSS for the entire group of patients with ASD was 7 . 9 ( SD = 1 . 6 ) . The assessment of the participants with ASD also included the administration of additional clinical standardized tests . Adaptive behaviour was assessed using the Vineland Adaptive Behaviour Scale-II ( VABS-II ) ( Sparrow et al . , 2005 ) , a standardized parent report interview . Developmental level was assessed with the Psycho-educational Profile Third Edition ( PEP-3 ) ( Lansing et al . , 2005 ) . See Table 1 for characteristics of study participants . Prior to their inclusion in our research protocol , potential TD participants were initially screened for neurological/psychiatric problems and learning disabilities using a medical and developmental history questionnaire before their visit . Moreover , they underwent ADOS-G or ADOS-2 evaluations to exclude any ASD symptomatology . Fourteen controls were tested with Modules 1 or 2 and four underwent the toddler module of the ADOS-2 . All TD participants had a minimal severity score of 1 , except one child who had a score of 3 . Stimuli consisted of two video sequences of dynamic social images without audio information of approximatively two minutes each . These videos included ecologically valid and complex naturalistic dynamic images where young children practised yoga alone , imitated animal-like behaviours ( behaving like a monkey or jumping like a frog ) , waived their arms , struck a pose , jumped , made faces or whistled ( Yoga Kids 3 ; Gaiam , Boulder , Colorado , http://www . gaiam . com , created by Marsha Wenig , http://yogakids . com/ ) . Presentation and timing of stimuli were controlled by Tobii Studio software ( Sweden , http://www . tobii . com ) . The experiment was conducted in a lit room at the office Médico-Pédagogique in Geneva . To familiarize the child with the procedure , the families received a kit containing a custom-made EEG replica cap and pictures illustrating the protocol two weeks prior to their first visit . Participants were seated on their parents lap in order to make them feel as secure as possible and to minimize head and body movements or alone . Once seated , the experimenter measured the circumference of the head and placed the corresponding cap on the participant's head . A couple of minutes were taken in order to allow the participants to settle into the experiment's environment and get used to the cap before starting the experiment . Following this , a five point eye-tracking calibration procedure was initiated using the Tobii system ( Sweden , http://www . tobii . com ) . An attractive colourful object ( either a kitten , a bus , a duck , a dog or a toy ) was presented together with its corresponding sound on a white background and the participants had to follow the object visually . The recording and presentation of the visual stimuli started when a minimum of four calibration points were acquired for each eye . To best capture the child’s attention , we first showed them an age-appropriate animated cartoon , followed by some fractals and another animated cartoon . The block ended with a film containing dynamic social images , the condition of interest in the present experiment . All participants were presented with the same visual stimuli in the same order . Following the first block , impedances were rechecked and electrodes were readjusted where needed to maintain them below 40 kOhm . A second block was then acquired ( animated cartoon; animated fractals; animated cartoon; second condition of interest: dynamic social images ) . The experimenter continuously monitored the eye-tracking to ensure children were looking at the screen . The whole experiment lasted about half an hour . We used stringent criteria and only participants with the highest data quality were kept for subsequent analysis . Eye-tracking data were recorded with the TX300 Tobii eye-tracking system ( sampling rate resolution of 300 Hz ) . In order to analyse and quantify differences in visual exploration between our groups , we developed a data-driven method to define dynamic norms of the exploration of the visual scenes ( Kojovic et al . , in preparation ) . First , we applied a kernel density distribution estimation ( Botev et al . , 2010 ) on the eye-tracking data recorded from the TD group at each time frame of the films containing dynamic complex social images to compute a normative gaze distribution pattern . Then , for each of the participants with ASD individually , we computed a deviation index from this normative gaze distribution , and this , for each single time frame separately ( Figure 6 ) . We averaged these values across the two films to obtain a mean Proximity Index ( PI ) value . This index describes for a given ASD participant , his distance from the normative gaze distribution pattern calculated on the TD group . A high index value indicates a visual behaviour approaching the visual exploration of the TD participants ( more similarity ) , while a low index indicates a visual behaviour deviating from the TD group ( more dissimilarity ) . The EEG was acquired with a Hydrocel Geodesic Sensor Net ( HCGSN , Electrical Geodesics , USA ) with 129 scalp electrodes at a sampling frequency of 1000Hz . On-line recording was band-pass filtered at 0−100Hz using the vertex as reference . Data pre-processing was done using Matlab ( Natick , MA ) and Cartool ( http://sites . google . com/site/cartoolcommunity/ ) . We down-sampled the montage to a 111-channel electrode array to exclude electrodes on the cheek and the neck since those are often contaminated with artefacts . Data were filtered between 1 and 40Hz ( using non-causal filtering ) and a 50Hz notch filter was applied . Each file was then visually inspected by one of the three EEG experts ( HFS , TAR , and RKJ ) to exclude periods of movements artefacts . Periods where subjects were not looking at the screen were excluded . Independent component analysis ( ICA ) was performed on the data to identify and remove the components related to eye movement artefacts ( eye blinks , saccades ) . Subsequently , channels with substantial noise were interpolated using spherical spline interpolation for each recording . Finally , the cleaned data were down-sampled to 125Hz , recalculated against the average reference and inspected by two EEG experts ( HFS and AC ) to ensure that no artefacts had been missed . One hundred and twenty artefact-free epochs of 1 second per participant were included for further analysis and were considered as a minimum to ensure high enough data quality . The general analysis strategy is summarized in Figure 7 . Electrical source imaging ( ESI ) was performed to reconstruct the sources of brain activity that gave rise to the scalp EEG field . For this , we used a toddler template head model ( 33–44 month ) ( using the Montreal Neurological Institute ( MNI ) brain ) with consideration of skull thickness ( Locally Spherical Model with Anatomical Constraints , LSMAC ) . 4159 solution points were equally distributed in the grey matter . We used a distributed linear inverse solution ( Low Resolution Electromagnetic Tomography , LORETA [Pascual-Marqui et al . , 1994] ) to compute the 3-dimensional ( 3D ) current source densities . We then projected this 3D dipole time-series onto the predominant dipole direction of each region of interest ( ROI ) across time and epochs , therefore obtaining a scalar time-series ( Coito et al . , 2016a; Coito et al . , 2015; Plomp et al . , 2015a; Coito et al . , 2016b ) . We parcelled the grey matter in 82 ROIs based on the automated anatomical labelling ( AAL ) digital atlas ( Tzourio-Mazoyer et al . , 2002 ) , after normalization to the MNI space using SPM8 ( Wellcome Trust Centre for Neuroimaging , University College London , UK , www . fil . ion . ucl . ac . uk/spm ) . In order to reduce the dimensionality of the solution space , we considered the solution point closest to the centroid of each ROI as representative of the source activity in that ROI for further analysis . This allowed to obtain the source activity across time of 82 solution points , representative of 82 ROIs ( Coito et al . , 2016b ) . Directed functional connectivity estimates the influence that one signal exerts onto another , facilitating the study of directional relationships between brain regions . It is commonly assessed using the concept of Granger-causality: given two signals in a process , if the knowledge of the past of one allows a better prediction of the presence of the other signal in the process , then the former signal is said to Granger-cause the latter signal ( Granger , 1969 ) . Granger-causal modelling is a well-validated statistical method ( Bressler and Seth , 2011 ) that has been successfully applied to estimate the strength of directed interactions between brain regions in rats using epicranial EEG ( Plomp et al . , 2014 ) and in non-human primates using intracranial recordings ( Brovelli et al . , 2004; Saalmann et al . , 2012 ) . It has also efficiently been used to study connectivity patterns in healthy humans with source imaging using EEG ( Astolfi et al . , 2007; Hu et al . , 2012; Plomp et al . , 2015b ) and MEG ( Michalareas et al . , 2016 ) . This approach has also effectively been applied in clinical populations to study network alterations in patients with focal epilepsy using intra-cranial recordings ( Wilke et al . , 2009; van Mierlo et al . , 2013; van Mierlo et al . , 2011 ) as well as electrical source imaging ( Ding et al . , 2007; Coito et al . , 2016a; Coito et al . , 2015; Coito et al . , 2016b ) . In order to have interpretable results , Granger-causality analysis should be performed using electrical source imaging rather than on electrodes measured at the scalp surface ( Schoffelen and Gross , 2009; Bastos et al . , 2015 ) . Therefore , in order to estimate the directional relationships in our data , we computed the weighted Partial Directed Coherence ( wPDC ) ( Baccalá and Sameshima , 2001; Astolfi et al . , 2006; Plomp et al . , 2014 ) using the 82 source signals . PDC is a multivariate approach , which considers all signals simultaneously in the same model and estimates brain connectivity in the frequency domain . It is computed using multivariate autoregressive models of a certain model order . Here , we used a model order of 5 , corresponding to 40ms . The wPDC was computed for each subject and epoch and then , the average of the PDC values within subjects was taken ( Coito et al . , 2016b ) . The average PDC was subsequently scaled ( 0 − 1 ) across ROIs and frequencies ( 1 − 40 Hz ) by subtracting the minimum power and dividing by the range . In order to weight the PDC by the spectral power ( SP ) of each source signal , while avoiding frequency doubling , we computed the Fast Fourier Transform ( FFT ) for each electrode , applied ESI to the real and imaginary part of the FFT separately and then combined them ( Coito et al . , 2016a; Coito et al . , 2015; Plomp et al . , 2015a; Yuan et al . , 2008 ) . The mean SP was obtained for each subject and scaled ( that is 0–1 , in the same way as PDC ) . For further details on the methodological approach to compute directed functional connectivity from electrical-source imaging signals , we refer the reader to ( Coito et al . , 2016b ) . For each subject , we obtained a 3D connectivity matrix ( ROIs x ROIs x frequency ) , representing the outflow from one ROI to another for each frequency . For further analysis , we reduced the connectivity matrix to 3 frequency bands: theta ( 4 − 7Hz ) , alpha ( 8 − 12Hz ) and beta ( 13 − 30Hz ) , by calculating the mean connectivity value in each band . For each subject and frequency band , we computed the summed outflow as the sum of wPDC values from a given ROI to all the others . This reflects the driving importance of this ROI in the network: ROIs with high summed outflow strongly drive the activity of other ROIs . We identified the highest information transfer ( summed outflow ) in the theta band followed by the alpha band . Therefore , we focused our subsequent analysis on these two frequency bands . We carried out statistical comparisons of the summed outflows between subject groups using a non-parametrical statistical test ( Mann − Whitney − Wilcoxon , two − tailed , p < 0 . 05 ) . We then investigated the outflows from the ROIs that showed statistically significant summed outflow between groups to the whole brain ( remaining 81 ROIs ) and carried out a statistical comparison of these outflows between groups ( Mann − Whitney − Wilcoxon , two − tailed , p < 0 . 05 , Benjamini − Hochberg = 0 . 05 ) . We correlated ( Spearman − rho , two − tailed , p < 0 . 05 ) the summed outflow results obtained in each of the 82 ROIs with ADOS-CSS scores , with developmental scores obtained from the PEP-3 , with adaptive scores obtained from VABS-II and with the PI values obtained from the eye-tracking data . In all cases , correlation p-values were Benjamini-Hochberg-corrected for multiple testing with p = 0 . 05 . Connectivity computations were performed in Matlab . Figures 1 , 2 and 3 , Figure 3—figure supplement 2 , Figure 3—figure supplement 2 , Figures 4 and 5 were produced using the BrainNet Viewer toolbox ( Xia et al . , 2013 ) . | Newborns are attracted to voices , faces and social gestures . Paying attention to these social cues in everyday life helps infants and young children learn how to interact with others . During this period of development , a network of connections forms between different parts of the brain that helps children to understand other people’s social behaviors . During their first year of life , infants who later develop autism spectrum disorders ( ASD ) pay less attention to social cues . This early indifference to these important signals leads to social deficits in children with ASD . They are less able to understand other people’s behaviors or engage in typical social interactions . It’s not yet clear why children with ASD are less attuned to social cues . But is likely that the development of brain networks essential for understanding social behavior suffers as a result . Studying how such networks develop in typical very young children and those with ASD may help scientist learn more . Now , Sperdin et al . confirm there are differences in the social brain-networks of very young children with ASD compared with their typical peers . In the experiment , 3-year-old children with ASD and without watched videos of other children playing , while Sperdin et al . recorded what they looked at and what happened in their brains . Eyemovements were measured with a tracker , and the brain activity was recorded using an electroencephalogram ( EEG ) , which uses sensors placed on the scalp to measure electrical signals . What children with ASD looked at was different than their typical peers , and these differences corresponded with alterations in the brain networks that process social information . Children with ASD who had less severe symptoms had stronger activity in these brain networks . What they looked at also was more similar to typical children . This suggests less severely affected children with ASD may be able to compensate that way . Identifying ASD-like behaviors and brain differences early in life may help scientists to better understand what causes the condition . It may also help clinicians provide more individualized therapies early in life when the brain is most adaptable . Long-term studies of these brain-network differences in children with ASD are necessary to better understand how therapies can influence these changes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Early alterations of social brain networks in young children with autism |
Dynamin is a mechanochemical GTPase essential for membrane fission during clathrin-mediated endocytosis . Dynamin forms helical complexes at the neck of clathrin-coated pits and their structural changes coupled with GTP hydrolysis drive membrane fission . Dynamin and its binding protein amphiphysin cooperatively regulate membrane remodeling during the fission , but its precise mechanism remains elusive . In this study , we analyzed structural changes of dynamin-amphiphysin complexes during the membrane fission using electron microscopy ( EM ) and high-speed atomic force microscopy ( HS-AFM ) . Interestingly , HS-AFM analyses show that the dynamin-amphiphysin helices are rearranged to form clusters upon GTP hydrolysis and membrane constriction occurs at protein-uncoated regions flanking the clusters . We also show a novel function of amphiphysin in size control of the clusters to enhance biogenesis of endocytic vesicles . Our approaches using combination of EM and HS-AFM clearly demonstrate new mechanistic insights into the dynamics of dynamin-amphiphysin complexes during membrane fission .
Clathrin-mediated endocytosis ( CME ) is the best characterized endocytic pathway by which cells incorporate extracellular molecules into cells as cargoes of clathrin-coated vesicles ( Kirchhausen et al . , 2014; McMahon and Boucrot , 2011 ) . CME is required for various essential processes including neuronal transmission , signal transduction and other cell membrane activities such as cell adhesion and migration . For precise progression of membrane invagination and fission during CME , various proteins need to be assembled in a temporally and spatially coordinated manner at the site of endocytosis . One of those endocytic proteins , dynamin , is a GTPase essential for membrane fission in CME ( Antonny et al . , 2016; Ferguson and De Camilli , 2012; Schmid and Frolov , 2011 ) . There are three dynamin isoforms in mammals: dynamin 1 and dynamin 3 , two tissue-specific isoforms which are highly expressed in neurons , and dynamin 2 , an ubiquitously expressed isoform ( Cao et al . , 1998; Cook et al . , 1996 , 1994 ) . Structural studies from several groups demonstrated that dynamin consists of five structurally distinct domains: a GTPase domain , a bundle signaling element ( BSE ) , a stalk , a pleckstrin homology ( PH ) domain and a proline-rich domain ( PRD ) from N-terminus to C-terminus ( Faelber et al . , 2011; Ford et al . , 2011; Reubold et al . , 2015 ) . The GTPase domain is responsible for hydrolysis of GTP ( guanosine triphosphate ) and the PH domain is required for membrane association by binding to negatively charged phospholipids such as PI ( 4 , 5 ) P2 ( phosphatidylinositol 4 , 5-bisphosphate ) . The stalk structure serves as a binding interface for dimerization and oligomerization of dynamin . The BSE , which is located between the stalk and GTPase domain , functions as a flexible hinge required for structural changes of dynamin upon GTP hydrolysis . Dynamin forms helical oligomers which was first observed in presynaptic terminals of shibire mutant flies at restrictive temperature ( Koenig and Ikeda , 1989 ) . Dynamin also assembles into helices at the neck of endocytic pits in the isolated presynaptic nerve terminals treated with slowly hydrolyzable GTP analogue GTPγS ( guanosine 5'-O-[gamma-thio]triphosphate ) ( Takei et al . , 1995 ) . Similar dynamin helices were reconstituted in vitro either with liposomes ( Sweitzer and Hinshaw , 1998; Takei et al . , 1998 ) or without liposomes in a low-salt condition ( Hinshaw and Schmid , 1995 ) . There is a consensus view about the dynamin-mediated membrane constriction and fission which is well supported by previous studies from different groups: membrane constriction is required , but not sufficient , for fission ( Antonny et al . , 2016; Faelber et al . , 2012; Schmid and Frolov , 2011 ) . However , it is still controversial how constriction is achieved , and what GTP energy is used for . For example , membrane constriction could be achieved by assembly into the highly constricted state when dynamin is bound to GTP ( Chen et al . , 2004; Mattila et al . , 2015; Mears et al . , 2007; Zhang and Hinshaw , 2001 ) . Alternatively , membrane constriction could be achieved by hydrolysis of GTP that induces a conformational change leading to constriction ( Cocucci et al . , 2014; Marks et al . , 2001; Roux et al . , 2006 ) . However , precise mechanisms involved in dynamin-mediated membrane constriction and fission remain unclear . Amphiphysin is a BAR domain protein required for membrane invagination in CME ( Wigge et al . , 1997 ) . Amphiphysin has a lipid interacting BAR ( Bin–Amphiphysin–Rvs ) domain in its N-terminal , a medial clathrin/AP-2 binding ( CLAP ) domain and C-terminal Src homology 3 ( SH3 ) domain . The BAR domain of amphiphysin forms crescent-shaped dimer and its concave surface serves as a platform for bending membrane or sensing membrane curvature ( Peter et al . , 2004 ) . The CLAP domain binds to clathrin and AP-2 , major components of clathrin-coated pits , and helps to recruit amphiphysin to the sites of CME . In addition , the C-terminal SH3 domain of amphiphysin binds directly to the PRD of dynamin 1 ( David et al . , 1996; Takei et al . , 1999 ) and enhances dynamin’s GTPase activity in the presence of liposomes ( Takei et al . , 1999; Yoshida et al . , 2004 ) . Amphiphysin copolymerizes with dynamin 1 into helical complexes , which form membrane tubules in vitro ( Takei et al . , 1999; Yoshida et al . , 2004 ) similar to those formed from synaptic plasma membranes ( Takei et al . , 1995 ) . Furthermore , injection of specific antibodies against amphiphysin into the giant synapse in lampreys ( Evergren et al . , 2004 ) or amphiphysin KO in mice ( Di Paolo et al . , 2002 ) causes suppressed endocytosis in synaptic vesicle recycling . These results suggest that dynamin mediates membrane fission in CME in collaboration with amphiphysin in vivo . However , the precise contribution of amphiphysin in the dynamin-mediated membrane fission remains elusive . In this study , we analyzed dynamics of dynamin-amphiphysin helical complexes using an approach combining electron microscopy ( EM ) and high-speed atomic force microscopy ( HS-AFM ) . Firstly , we show that the dynamin-amphiphysin helices are rearranged to form clusters upon GTP hydrolysis , and membrane constriction occurs at protein-uncoated regions between the clusters . Secondly , we reveal that GTP hydrolysis is required and sufficient for the cluster formation by dynamin-amphiphysin complexes by EM analyses . Finally , we show a novel function of amphiphysin in controlling cluster size , which in turn regulates biogenesis of endocytic vesicles . These findings provide new insights into the mechanism of membrane constriction and fission by dynamin-amphiphysin complexes .
To elucidate the mechanisms of dynamin-mediated membrane fission , we reconstituted the minimum system in vitro and analyzed the time course of its structural changes using EM . Human dynamin 1 and amphiphysin were purified ( Figure 1—figure supplement 1A ) and their activity to form ring-shaped complexes in a buffer of physiological ionic strength and pH condition ( Figure 1—figure supplement 1B ) at different stoichiometry of dynamin and amphiphysin ( Figure 1—figure supplement 2A ) were confirmed . As previously described ( Sweitzer and Hinshaw , 1998; Takei et al . , 1999 ) , the dynamin-amphiphysin complexes induced tubulation of large unilamellar vesicles ( LUVs ) in the absence of GTP ( Figure 1A , No GTP ) and it is not stoichiometry dependent ( Figure 1—figure supplement 2B ) . Immediately after the addition of 1 mM GTP , the appearance of lipid tubules was not affected ( Figure 1A , GTP 1 s ) , but they started to form multiple constriction sites over time ( Figure 1A , GTP 5 s , 10 s and 30 s ) and membrane fission occurred finally and numerous vesicles were generated within 1 min ( Figure 1A , GTP 1 min ) . The membrane fission activity by dynamin-amphiphysin complexes was stoichiometry sensitive: the membrane fission occurred efficiently at 1:0 . 5 or 1:1 molar ratio of dynamin and amphiphysin ( Figure 1-figure supplement 2C , 1:0 . 5 and 1:1 ) , while the fission activity was less efficient when dynamin and amphiphysin were mixed at 1:2 ratio ( Figure 1-figure supplement 2C , 1:2 ) , probably due to lower stimulatory effect of amphiphysin on dynamin GTPase activity at the higher molecular ratio ( Yoshida et al . , 2004 ) . Next , we tried to clarify how the membrane constriction and fission by dynamin-amphiphysin helices are correlated with guanine nucleotide conditions during GTP hydrolysis . The appearance of lipid tubules ( Figure 1B , No GTP ) was not affected in the presence of either slowly hydrolyzable GTP analogue GTPγS ( Figure 1B , GTPγS ) or non-hydrolyzable GTP analogue GMP-PNP ( guanosine 5′-[β , γ-imido]triphosphate ) ( Figure 1B , GMP-PNP ) . In contrast , in the presence of GDP ( guanosine diphosphate ) and vanadate , the complex which mimics the GDP⋅Pi transition state , lipid tubules were constricted at multiple sites ( Figure 1B , GDP + vanadate ) . Addition of only GDP did not cause membrane constriction or fission , but membrane tubules were deformed ( Figure 1B , GDP ) . Finally , numerous vesicles were generated 10 min after the addition of 1 mM GTP , in which multiple rounds of GTP hydrolysis were likely to have taken place ( Figure 1B , GTP ) . Taken these results together , GTP hydrolysis is essential for both membrane constriction and fission by the dynamin-amphiphysin complexes , but subsequent dissociation of GTP hydrolytic products ( GDP and/or phosphate ) is required for completing membrane fission . Although we determined the requirement of GTP hydrolysis in membrane constriction and fission by the dynamin-amphiphysin complexes , structural changes of the complexes were not clearly resolved in the in vitro assay system using LUVs . To improve the resolution , we used rigid lipid nanotubes containing glycolipid galactosylceramide ( GalCer ) ( Wilson-Kubalek et al . , 1998 ) , instead of using LUVs in the in vitro assay system . Lipid nanotubes are rod-shaped liposomes and similar in size to the unconstricted necks of clathrin-coated pits observed in vivo ( Figure 2A , Nanotube ) . Dynamin-amphiphysin complexes assembled into helices on the lipid nanotubes ( Figure 2A , No GTP ) , which is similar to those formed by dynamin alone ( Stowell et al . , 1999 ) . Interestingly , the dynamin-amphiphysin helices transiently formed clusters after the addition of GTP ( Figure 2A , GTP 1 s and 20 s , brackets ) . The dynamin-amphiphysin clusters were disorganized over time and partially dissociated from the nanotubes ( Figure 2A , GTP 30 s and 1 min ) . To correlate the dynamics of dynamin-amphiphysin complexes with GTP hydrolysis , we examined structural changes of the complexes on lipid nanotubes at different transition states of GTP hydrolysis . The appearance of dynamin-amphiphysin helices was unchanged even in the presence of GTPγS or GMP-PNP ( Figure 2B , No GTP , GTPγS and GMP-PNP ) . Interestingly , addition of GDP and vanadate induced rearrangement of the dynamin-amphiphysin helical complexes to form clusters similar to those observed after the addition of GTP ( Figure 2B , GDP + vanadate ) . The average pitch of helices in the clusters were shorter ( 15 . 0 ± 0 . 3 nm , mean pitch ± s . e . m . ) compared to the average pitch of the helical complexes in No GTP control ( 20 . 0 ± 0 . 5 nm , mean pitch ± s . e . m . ) ( Figure 2—source data 1 ) . Furthermore , unlike membrane fission activity ( Figure 1—figure supplement 2C ) , clustering behavior of the dynamin-amphiphysin complexes on the lipid nanotubes were not stoichiometry dependent ( Figure 2—figure supplement 1 , 1:0 . 5 , 1:1 and 1:2 , white brackets ) . In contrast , GDP alone did not affect the distribution of dynamin-amphiphysin helices ( Figure 2B , GDP ) . Finally , the dynamin-amphiphysin helical complexes were disorganized and eventually dissociated from the lipid nanotubes 10 min after the addition of 1 mM GTP ( Figure 2B , GTP ) . Taken these results together , the dynamin-amphiphysin helical complexes transiently form clusters in the GTP hydrolysis transition state of GDP⋅Pi during which membrane tubules are constricted . To elucidate the dynamics of dynamin-amphiphysin helical complexes during the membrane constriction and fission , we analyzed the clustering process of the complexes using HS-AFM ( Ando et al . , 2013 ) . LUVs were stably immobilized on the carbon-coated and glow-discharged mica substrate ( Figure 3—figure supplement 1A; Video 1 ) , and they were successfully tubulated in the presence of dynamin and amphiphysin ( Figure 3—figure supplement 1B; Video 2 ) . The dynamin-amphiphysin helices on the lipid tubules were aligned with an almost regular pitch ( 22 . 0 ± 0 . 7 nm , mean pitch ± s . e . m . ) and they were immobile before GTP addition ( Figure 3A , 0 s and 21 s; Video 3; Figure 3—source data 1 ) . Interestingly , the dynamin-amphiphysin helices became mobile after GTP addition and eventually formed clusters consisting of a few helices with shorter pitch ( 15 . 7 ± 0 . 3 nm , mean pitch ± s . e . m . ) ( Figure 3A , from 42 s to 131 s; Video 3; Figure 3—source data 1 ) . Particle tracking analyses of the individual dynamin-amphiphysin helices showed that the dynamin-amphiphysin complexes were static before GTP addition ( Figure 3B , 5-21 s; Video 4 ) , but addition of 1 mM GTP stimulated longitudinal movement of the helical complexes , leading to the cluster formation ( Figure 3B , 38-54 s , 38–86 s and 38–118 s; Video 5 ) . Although membrane fission was not observed in this sample probably due to a strong attachment of the lipid tubule to the substrate , the helices had a tendency to constrict during the cluster formation ( Figure 3C; Figure 3—figure supplement 2; Figure 3—source data 2; Figure 3—figure supplement 2—source data 1 ) . These results suggest that dynamin-amphiphysin helical complexes undergo two modes of structural changes , longitudinal clustering and radial constriction , during GTP hydrolysis . We next tried to correlate the cluster formation of dynamin-amphiphysin helical complexes with membrane constriction and fission . In the representative sample in which membrane constriction and fission occurred , a few dynamin-amphiphysin helices merged to form a cluster over time after GTP addition ( Figure 4A , 0 s , 125 . 3 s , 185 . 5 s and 227 . 5 s; Video 6; Figure 4—source data 1 ) . Interestingly , membrane constriction occurred at flanking regions of the cluster where membrane was bare of dynamin-amphiphysin complexes ( Figure 4A , fission point ( FP ) . 1 and FP . 2 ) . The heights at sites marked with FP . 1 and FP . 2 were not changed before constriction ( Figure 4B , before constriction; Figure 4—source data 2 ) , but they became lower in a stepwise manner from a pre-constriction height of around 30 nm down to 20–25 nm or below ( Figure 4B , after constriction; Figure 4—source data 2 ) . Similar longitudinal redistribution of the dynamin-amphiphysin helices before membrane constriction was also observed in another sample , in which constriction occurred at one end of clustered dynamin-amphiphysin complexes ( Figure 4c , arrow; Video 8 ) . These results strongly suggest that membrane constriction and fission occur at the protein-uncoated regions created as a result of the clustering of dynamin-amphiphysin helical complexes . We previously demonstrated that amphiphysin stimulates the GTPase activity of dynamin and thus enhances vesicle biogenesis ( Yoshida et al . , 2004 ) . In this study , we also noticed that the average size of vesicles formed by dynamin-amphiphysin complexes ( 70 . 0 ± 2 . 9 nm , mean diameter ± s . e . m . ) was significantly smaller compared to those formed by dynamin alone ( 204 . 6 ± 12 . 3 nm , mean diameter ± s . e . m . ) after GTP addition ( Figure 5A; Figure 5—source data 1 ) . Consistently , dynamin-amphiphysin complex formed constriction sites with shorter intervals ( 150 . 3 ± 9 . 8 nm , mean intervals ± s . e . m . ) compared to those formed by dynamin alone ( 193 . 5 ± 15 . 8 nm , mean intervals ± s . e . m . ) in the presence of GDP and vanadate ( Figure 5B; Figure 5—source data 2 ) . To further elucidate roles of amphiphysin in the membrane constriction and fission , the cluster formation by dynamin alone was compared to that by dynamin-amphiphysin complexes , using lipid nanotubes . As already described , dynamin-amphiphysin complexes formed clusters with a few helices in the presence of GDP and vanadate ( 34 . 2 ± 1 . 7 nm , mean cluster size ± s . e . m . ) ( Figure 5C , Dynamin + Amphiphysin; Figure 5—source data 3 ) . In contrast , dynamin alone formed larger-sized clusters consist of more helical complexes ( 59 . 3 ± 4 . 7 nm , mean cluster size ± s . e . m . ) ( Figure 5C , Dynamin; Figure 5—source data 3 ) . These results suggest that amphiphysin contributes to the effective generation of properly sized vesicles by controlling the cluster formation of dynamin-amphiphysin helical complexes .
In this study , we analyzed dynamics of dynamin-amphiphysin helical complexes during membrane constriction and fission using EM and HS-AFM . EM analyses showed that GTP hydrolysis is required for both membrane constriction and fission , but dissociation of hydrolytic products ( GDP and/or phosphate ) seems necessary for the completion of membrane fission ( Figure 1 ) . In the presence of GTP or GDP and vanadate , dynamin-amphiphysin helical complexes are reorganized , resulting in the formation of clusters consisting of a few dynamin-amphiphysin helices ( Figure 2 ) . HS-AFM analyses directly demonstrated that GTP hydrolysis induces dynamic longitudinal movement of the dynamin-amphiphysin helices as well as constriction during the cluster formation ( Figure 3 ) . Interestingly , HS-AFM analyses also demonstrated that membrane constriction and fission occur at the ‘protein-uncoated’ regions created as a result of cluster formation of dynamin-amphiphysin complexes ( Figure 4 ) . Finally , we found that amphiphysin contributes to effective biogenesis of endocytic vesicles by regulating size of the clusters formed by dynamin-amphiphysin helical complexes ( Figure 5 ) . There is a consensus view about the requirement of GTP hydrolysis in membrane fission , but the requirement of GTP hydrolysis in membrane constriction is still controversial ( Antonny et al . , 2016 ) . Membrane tubules are constricted in the presence of non-hydrolyzable GTP analogue ( Chen et al . , 2004; Mears et al . , 2007; Zhang and Hinshaw , 2001 ) and more constricted with a GTP-loaded GTPase defective K44A mutant ( Sundborger et al . , 2014 ) . In both cases , membrane tubules are evenly constricted and periodical membrane constriction sites which lead to membrane fission is not created . In the present study , we showed that membrane constriction sites are created in the presence of GDP and vanadate , which mimicked a transition state of GTP hydrolysis ( GDP⋅Pi ) , suggesting that complete hydrolysis of GTP is required for the formation of constriction sites leading to membrane fission ( Figure 1B ) . Membrane fission has never been observed in the presence of GDP and vanadate , suggesting that release of GTP hydrolytic products ( GDP and/or phosphate ) is a prerequisite for membrane fission . Further analyses will more precisely reveal which intermediate state in the GTPase reaction is responsible for the membrane fission or how many GTPase cycles are required for it . In this study , we revealed that dynamin-amphiphysin helical complexes are rearranged to form their clusters upon GTP hydrolysis ( Figure 2 and Figure 3 ) and membrane fission occurs at the flanking ‘protein-uncoated’ membrane regions ( Figure 4 ) . In the ‘constrictase’ model , dynamin constricts membrane until the membrane neck reaches to the hemi-fission state , which leads to spontaneous membrane fission ( Chen et al . , 2004; Hinshaw and Schmid , 1995; Mears et al . , 2007 ) . However , several lines of evidences are apparently inconsistent with this simple model . For instance , the super-constricted state of dynamin does not constrict the membrane sufficiently enough to reach the hemi-fission state ( Sundborger et al . , 2014 ) and membrane tension and/or torsion is required to overcome the energy barrier to fission ( Bashkirov et al . , 2008; Morlot et al . , 2012; Roux et al . , 2006 ) . In this study , we showed that GTP hydrolysis induces constriction of the dynamin-amphiphysin helices as well as clustering ( Figure 3 ) . These radial and longitudinal remodeling of the dynamin-amphiphysin helices may give local tension and/or torsion to the membrane tube at the edge of the clusters to drive membrane fission . Alternatively , the dynamin-amphiphysin clusters may serve as a lipid diffusion barrier that causes friction leading to membrane scission ( Simunovic et al . , 2017 ) . Longitudinal rearrangement upon GTP hydrolysis similar to the cluster formation by the dynamin-amphiphysin complexes was also observed in an EM study on the dynamics of dynamin with lipid nanotubes ( Stowell et al . , 1999 ) and more recently by HS-AFM analyses on dynamics of ΔPRD dynamin ( Colom et al . , 2017 ) , suggesting that the longitudinal rearrangement is an intrinsic property of dynamin during membrane fission . In our previous studies , we showed that amphiphysin enhances dynamin’s GTPase activity in the presence of liposomes ( Takei et al . , 1999; Yoshida et al . , 2004 ) . In this study , we revealed that amphiphysin may also contributes to effective vesicle biogenesis by controlling the number of constriction sites via cluster formation of dynamin-amphiphysin helices in a long membrane tubule formed in vitro ( Figure 5 ) . Although precise mechanisms of the cluster size control by amphiphysin remains unclear , amphiphysin could have roles either in determining the number of dynamin-amphiphysin helices comprising the clusters , or in positioning of breakage points in dynamin-amphiphysin helices to induce clustering . Tubular structures have long been known to be present in various synapses , and they are described as ‘membrane tubules’ ( Heuser and Miledi , 1971 ) , ‘cisternae’ ( Heuser and Reese , 1973 ) , ‘synaptic tubules’ ( Samorajski et al . , 1966 ) or ‘anastomosing tubules’ ( Ekström von Lubitz , 1981 ) . The tubules are enriched in endocytic proteins including dynamin , synaptojanin , amphiphysin , and endophilin ( Fuchs et al . , 2014; Takei et al . , 1998 ) , and the presence of the tubules becomes more prominent when synapses are stimulated ( Fuchs et al . , 2014; Takei et al . , 1998 ) , or when membrane fission is blocked in dynamin 1 K . O . mice ( Ferguson et al . , 2007 ) . These findings strongly suggest that the tubular structures represent endocytic intermediate at which dynamin-amphiphysin-dependent synergic vesicle formation takes place in the synapse . Besides amphiphysin , other BAR domain proteins , endophilin and syndapin , are also implicated in synaptic vesicle recycling ( Dittman and Ryan , 2009; Koch et al . , 2011; Milosevic et al . , 2011 ) . Interestingly , recent study showed that endophilin potently inhibits the dynamin-mediated membrane fission by intercalating dynamin rungs and preventing their trans-interactions required for membrane fission ( Hohendahl et al . , 2017 ) . Although the dynamin-mediated membrane fission is also inhibited when an excess of amphiphysin co-assembles with dynamin ( Figure 1—figure supplement 2C ) , it is rather stimulatory when the molar ratio of dynamin to amphiphsyin is around 1:1 ( Yoshida et al . , 2004 ) . One of the important future goals of dynamin study would be to clarify regulatory mechanisms by which dynamin alters its interactions with various BAR domain proteins in physiological contexts such as synaptic vesicle recycling . In conclusion , live imaging analyses using HS-AFM in this study and a study from another group ( Colom et al . , 2017 ) gave new mechanistic insights into the dynamin-mediated membrane fission . Combinatory approaches using high temporal resolution imaging with HS-AFM and high spatial resolution structural analyses with X-ray crystallography or Cryo-EM will be the most powerful approach in resolving various dynamic membrane remodeling processes in the future .
Human dynamin1 was purified using the method of Warnock et al . with some modification ( Warnock et al . , 1996 ) . Sf9 cells grown in 600 ml of SF-900II SFM ( Thermo Fisher Scientific , Waltham , MA ) to the cell density of 1 × 106 cells/ml and the cells were infected with baculoviruses expressing dynamin1 . After cultivation of cells at 28°C for 69 hr , the infected Sf9 cells were harvested by centrifugation at 500 × g for 10 min . The cell pellet was resuspended by 1/20 of the culture volume ( 30 ml ) of HCB ( Hepes column buffer ) 100 ( 20 mM Hepes , 100 mM NaCl , 2 mM EGTA , 1 mM MgCl2 , 1 mM DTT , 1 mM PMSF , 1 μg/ml Pepstatin A , 40 μM ALLN , pH 7 . 2 ) and cells were sonicated using a sonicator ( Advanced-Digital SONIFIER model 250 , BRANSON ) . The cell lysate was mixed with equal volume of HCB0 ( 20 mM Hepes , 2 mM EGTA , 1 mM MgCl2 , 1 mM DTT , 1 mM PMSF , 1 μg/ml Pepstatin A , 40 μM ALLN , pH 7 . 2 ) to make HCB50 ( 20 mM Hepes , 50 mM NaCl , 2 mM EGTA , 1 mM MgCl2 , 1 mM DTT , 1 mM PMSF , 1 μg/ml Pepstatin A , 40 μM ALLN , pH7 . 2 ) and centrifuged at 210 , 000 × g for 1 hr at 4°C . Ammonium sulfate was added to the cleared lysate to the 30% saturation and incubated at 4°C for 30 min and centrifuged at 10 , 000 × g for 10 min to recover the dynamin1 containing fraction in the pellet . The dynamin1 pellet was resuspended with 20 ml of HCB50 and dialyzed against 2L of HCB50 for total 4 hr ( 2 hr , 2 times ) using dialysis membrane ( Spectra/Por Dialysis Membrane MWCO: 3500 ) . The dialyzed dynamin1 fraction was applied to Mono Q5/50 GL column ( GE healthcare ) and bound proteins were eluted stepwise using HCB50 , HCB100 , HCB250 and HCB1000 buffers . Purified dynamin1 was recovered in HCB250 fraction and purity was determined by SDS-PAGE ( Figure 1—figure supplement 1A , Dynamin ) . Human amphiphysin was purified following manufacture’s instruction ( GE Healthcare ) with slight modifications . Host bacteria BL21 ( DE3 ) transformed with an expression construct for GST fusions of human amphiphysin ( pGEX6P2-HsAMPH ) were grown in 1 L of LB medium to the cell density of 0 . 6–0 . 8 ( OD 600 nm ) at 37°C and then protein expression was induced at 18°C for 12 hr in the presence of 0 . 1 mM IPTG . The bacterial cells were harvested by centrifugation at 7000 × g for 10 min and cell pellet was resuspended by 1/10 culture volume ( 100 ml ) of Elution/Wash 300 buffer ( 50 mM Tris-HCl , pH 8 . 0 , 300 mM NaCl ) . The resuspended cells were sonicated using Advanced-Digital SONIFIER model 250D ( Branson ) and centrifuged at 261 , 000 × g for 30 min at 4°C and cleared lysate was recovered in supernatant . To the cleared lysate , 1/100 culture volume ( 1 ml in bed volume ) of Glutathione Sepharose 4B Beads ( GE Healthcare ) was added and they are mixed using rotating mixer for 1 hr at 4°C . The beads were washed with the Elution/Wash 300 buffer for 5 times in a repeated cycle of centrifugation at 420 × g for 5 min at 4°C followed by mixing with rotator for 5 min at 4°C . The beads with purified GST fusions of amphiphsyin were equilibrated with PreScission Buffer ( 50 mM Tris-HCl , 150 mM NaCl , 1 mM EDTA , 1 mM DTT , pH 7 . 0 ) and GST-tag was removed by PreScission Protease ( GE Healthcare ) by incubating for 12 hr at 4°C . The purified amphiphysin was recovered by centrifuge ( 12 , 000 × g , 5 min at 4°C ) using spin column ( Ultrafree-Mc , GV 0 . 22 μm , Millipore ) and purity was determined by SDS-PAGE ( Figure 1—figure supplement 1A , Amphiphysin ) . Large unilamellar vesicles ( LUVs ) and lipid nanotubes were prepared as previously described ( Takei et al . , 2001 ) . For LUVs , 70% PS ( Cat . No 840032C , Avanti ) , 10% biotinPE ( Avanti ) and 20% cholesterol ( Avanti ) were mixed and , for lipid nanotubes 40% NFA Galactocerebrosides ( Sigma C1516 ) , 40% PC ( Avanti ) , 10% PI ( 4 , 5 ) P2 ( Calbiochem ) and 10% cholesterol ( Avanti ) were mixed in 250 μl of chloroform in a glass vial ( Mighty Vial No . 01 4 ml , Maruemu Cat . No 5-115-03 ) . Then chloroform was evaporated using slow-flow nitrogen gas to produce lipid a lipid film on the glass and then completely dried in a vacuum desiccator for 30 min . The dried lipid was rehydrated by water-saturated nitrogen gas followed by addition of 250 μl of filtered 0 . 3M sucrose for 2 hr at 37°C . The resultant LUVs and lipid nanotubes were passed through 0 . 4 μm- and 0 . 2μm- polycarbonate filters respectively 11 times using Avanti Mini extruder . The LUVs and lipid nanotubes ( 1 mg/ml of final concentration ) were stored in dark at 4°C avoiding photooxidation . LUVs and lipid nanotubes were diluted to 0 . 17 mg/ml in cytosolic buffer ( 25 mM Hepes-KOH , pH 7 . 2 , 25 mM KCl , 2 . 5 mM Magnesium acetate , 0 . 1 M K-glutamate , pH 7 . 4 ) . Dynamin-amphiphysin complexes ( 1:1 in molar ratio ) were diluted to 2 . 3 μM in the cytosolic buffer . Formvar filmed EM grids were carbon-coated , then glow-discharged . Droplets of the diluted lipids ( 10 μl each ) were prepared on Parafilm and adsorbed on EM grids for 5 min at room temperature . Then the EM grids with lipids were transferred to other droplets of the diluted dynamin-amphiphysin complexes and incubated for 30 min at room temperature in a humid chamber . To see the temporal effect of GTP hydrolysis , the EM grids were transferred to 1 mM of GTP and incubated for various time periods ( from 1 s to 10 min ) . The reaction was terminated by quick removal of the GTP solution by filter paper at room temperature . Alternatively , the EM grids were incubated either GTP , GTPγS , GMP-PNP , GDP plus Vanadate and GDP to analyze GTP hydrolysis transition state structures . The EM grids were negatively stained with filtered 2% uranyl acetate and observed with transmission electron microscope ( HITACHI H-7650 ) . All AFM images shown in this article were capture by a laboratory-built HS-AFM in which the amplitude-modulation mode was used . For the HS-AFM imaging , a small cantilever with dimensions of 7 μm long , 2 μm wide , and 90 nm thick was used ( Olympus ) . Its nominal spring constant and resonant frequency were ~0 . 2 N/m and ~800 kHz in an aqueous solution , respectively . To obtain a sharp tip , an amorphous carbon pillar was grown on the original bird-beak tip of the cantilever by electron beam deposition ( EBD ) and then sharpened by a plasma etching in an argon environment . The typical radius of the EBD tip was approximately 2 nm after sharpening . For the amplitude-modulation imaging , the cantilever was oscillated with amplitude less than 10 nm under free oscillation condition and the set-point was set at ~90% of the free oscillation amplitude . For HS-AFM imaging of liposomes and dynamin-amphiphysin complexes with lipid tubules or nanotubes , we used mica covered with carbon film . After coating a freshly cleaved mica surface with carbon film , hydrophilic treatment was carried out by a grow discharge . The liposomes ( 0 . 17 mg/ml ) were deposited on the hydrophilic mica surface and incubated for 5 min at room temperature followed by deposition of proteins ( 0 . 6 μM of dynamin1 and amphiphysin ) for 30 min at room temperature . After the incubation , the sample was thoroughly washed by cytosolic buffer to remove excess liposomes and proteins . After the washing , the cantilever tip was approached and the imaging was performed under the buffer . The EM and HS-AFM images were randomly captured to avoid data manipulation and representative images were shown in all the figures . The average pitch between the dynamin-amphiphysin helices in EM images ( Figure 2 ) and HS-AFM images ( Figure 3 ) , diameter of vesicles , intervals between constriction sites and size of clusters generated by either dynamin-amphiphysin complex or dynamin ( Figure 5 ) , were all measured by FIJI ( Schindelin et al . , 2012 ) . Experimental data were statistically analyzed using Excel ( Microsoft ) or Prism 7 ( GraphPad software ) . | The nerve cells that make up a nervous system connect at junctions known as synapses . When a nerve impulse reaches the end of the cell , membrane-bound packages called vesicles fuse with the surface membrane and release their contents to the outside . The contents , namely chemicals called neurotransmitters , then travels across the synapse , relaying the signal to the next cell . Nerve cells can fire many times per second . The membrane from fused vesicles must be retrieved from the surface membrane and recycled to make new vesicles , ready to transmit more signals across the synapse . Many proteins at these sites are involved in folding the fused membrane back into the cell , constricting the opening , and eventually pinching off the new vesicles – a process known as endocytosis . Two proteins named dynamin and amphiphysin cooperate in this process , but their precise mechanism remained elusive . Dynamin is a protein that acts like a motor; it breaks down a molecule called GTP to release energy . Previous studies have seen that dynamin-amphiphysin complexes join end to end to form long helical structures . Takeda et al . have now looked at how the structure of the helices changes during endocytosis . This revealed that the dynamin-amphiphysin helices rearrange to form clusters when the GTP is broken down . Further analysis showed that the folded membrane becomes constricted at regions that are not coated with the clusters of dynamin-amphiphysin helices . Takeda et al . also discovered that amphiphysin controls the size of the clusters to help make the new vesicles more uniform . The gene for dynamin is altered in a number of disorders affecting the nervous system and muscles , including epileptic encephalopathy , Charcot-Marie-Tooth disease and congenital myopathy . Moreover , a neurological disorder characterized by muscle stiffness ( known as Stiff-person syndrome ) occurs when an individual’s immune system mistakenly attacks the amphiphysin protein . As such , these new findings will not only help scientists to better understand the process of endocytosis , but they will also give new insight into a number of human diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Dynamic clustering of dynamin-amphiphysin helices regulates membrane constriction and fission coupled with GTP hydrolysis |
Methylation of cytosines ( 5meC ) is a widespread heritable DNA modification . During mammalian development , two global demethylation events are followed by waves of de novo DNA methylation . In vivo mechanisms of DNA methylation establishment are largely uncharacterized . Here , we use Saccharomyces cerevisiae as a system lacking DNA methylation to define the chromatin features influencing the activity of the murine DNMT3B . Our data demonstrate that DNMT3B and H3K4 methylation are mutually exclusive and that DNMT3B is co-localized with H3K36 methylated regions . In support of this observation , DNA methylation analysis in yeast strains without Set1 and Set2 shows an increase of relative 5meC levels at the transcription start site and a decrease in the gene-body , respectively . We extend our observation to the murine male germline , where H3K4me3 is strongly anti-correlated while H3K36me3 correlates with accelerated DNA methylation . These results show the importance of H3K36 methylation for gene-body DNA methylation in vivo .
In multicellular organisms , every cell type possesses the same genetic information , but manifests a different phenotype . Chromatin plays a fundamental role in both the establishment and maintenance of each cell's state . Many players contribute to chromatin states , including nucleosome organization , histone post-translational modifications , and non-coding RNAs ( Chen and Dent , 2014; Maze et al . , 2014; Quinodoz and Guttman , 2014 ) . Another mechanism for maintaining the state of a cell through cell division is the methylation of cytosines at position 5 ( 5meC ) , a widespread heritable DNA modification found in prokaryotes , plants , several fungi , and animals ( Iyer et al . , 2011 ) . In mammals , DNA methylation plays a fundamental role in processes such as imprinting , X-chromosome inactivation , transposon inactivation , and gene expression regulation ( Smith and Meissner , 2013 ) . Dysregulation of DNA methylation is a common feature in cancer ( Eden et al . , 2003; You and Jones , 2012 ) and a variety of human diseases are caused by defective imprinting ( Peters , 2014 ) . Methylation is mainly found at symmetric CpG dinucleotides , where it is introduced by the de novo DNA methyltransferases ( DNMT3a and DNMT3b ) and can be copied faithfully during DNA replication by the activity of a ‘maintenance’ DNA methyltransferase , DNMT1 ( Law and Jacobsen , 2010 ) . However , DNA methylation is not static throughout mammalian development . In fact , 5meC can either be lost by a passive mechanism , such as the failure to maintain DNA methylation through cell division or by an active mechanism such as the removal of methylcytosine , typically via an oxidized intermediate ( Pastor et al . , 2013 ) . Demethylation and de novo methylation can occur in a locus-specific manner , typically in concert with the activation or silencing of promoters or enhancers . However , global demethylation and de novo methylation events can also occur during development ( Pastor et al . , 2013; Seisenberger et al . , 2013 ) . For example , most DNA methylation is progressively lost between fertilization and the formation of the blastula and global de novo DNA methylation then occurs coincidently with implantation of the embryo . This de novo methylation event largely shapes the methylation pattern of the animal , with additional changes occurring in somatic tissues , which contribute to cellular identity . In the germline however , a second reprogramming event occurs . After specification of the germ cells , most DNA methylation is lost during early primordial germ cell ( PGCs ) development . Unlike in early embryogenesis , imprints are erased during this period . Genome-wide de novo methylation then occurs before birth in the male germline and upon oocyte maturation in females ( Smallwood et al . , 2011 ) . This de novo methylation event establishes the imprints that are inherited in the next generation . Considering the importance of local and global de novo methylation events in imprinting , gene regulation and cellular identity , it is important to understand how the de novo DNA methyltransferases are targeted to the correct genomic regions . DNMT3 proteins do not have strong sequence preferences beyond CpG dinucleotides ( Dodge et al . , 2002 ) . We therefore sought to determine which factors are critical for the targeting of de novo DNA methyltransferases . Active de novo DNA methyltransferases possess three different domains: the catalytic domain , found at the C-terminus of the protein , an ADD domain and a PWWP domain ( Figure 1A ) ( Law and Jacobsen , 2010 ) . In contrast , the inactive DNMT3L possesses only a functional ADD domain . The ADD domains of all three DNMTs have been shown to preferentially bind histone 3 tails that lack methylation at lysine 4 ( H3K4me0 ) ( Ooi et al . , 2007; Zhang et al . , 2010 ) , and this binding has been recently shown to relieve DNMT3a auto-inhibition ( Guo et al . , 2015 ) . This is consistent with the observation that genomic regions bearing H3K4 methylation are generally depleted of 5meC ( Singh et al . , 2013 ) . The PWWP domain of several proteins has been shown to bind H3K36 methylation ( Vermeulen et al . , 2010 ) , and indeed the DNMT3a-PWWP domain has also been shown to interact with the tri-methylated lysine 36 of histone H3 ( H3K36me3 ) in vitro ( Dhayalan et al . , 2010 ) . The importance of these histone-binding domains in targeting DNA methyltransferase activity in vivo is still unclear . It is also possible that the PWWP domain's primary function is to bind DNA and not nucleosomes ( Dhayalan et al . , 2010 ) . Recently , it has been reported that the PWWP domain is important in specifying the localization of DNMT3b in mouse embryonic stem cells ( Baubec et al . , 2015 ) . 10 . 7554/eLife . 06205 . 003Figure 1 . Distribution of induced DNA methylation in Saccharomyces cerevisiae . ( A ) Murine DNMT3 proteins with known domains: PWWP , ADD ( ATRX–DNMT3–DNMT3L ) , and C-5 methyltransferase domain ( not functional in DNMT3L ) . Accession numbers: DNMT3a = O88508; DNMT3b = O88509; DNMT3L = Q9CWR8 . ( B ) Constructs used in this study . The empty vector ( EV ) is pYES2 ( Life Technologies ) . DNMT3b expression is controlled by the GAL1 promoter . ( C ) Levels of 5meC in different dinucleotide contexts . The gray dotted line represents the unconversion rate . ( D ) Metagene plot of CpG methylation in cells expressing DNMT3b during logarithmic and stationary phase . EV ( strain not expressing DNMT3b ) . Exponential and stationary strains 1–6 are derived from the W303 strain , while stationary strains 7 and 8 are in a BY4741 background . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 00310 . 7554/eLife . 06205 . 004Figure 1—figure supplement 1 . Chromosome-wide view of DNA methylation and genomic features . Distribution of DNA methylation on chromosome XII of S . cerevisiae ( A and B ) . In ( B ) the density of other genomic features is shown ( arbitrary units ) . Averages for DNA methylation and genomic features are calculated on 4 Kb bins . Areas of repetitive sequences ( such as rRNA and transposable elements ) show very little to no coverage . Gene-rich bins also correspond to peaks in DNA methylation levels . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 00410 . 7554/eLife . 06205 . 005Figure 1—figure supplement 2 . Distribution of 5meC around TSS and TTS . CpG methylation levels around ( TSSs—blue ) and ( TTSs—green ) of yeast genes . Periodic peaks of DNA methylation are evident at the TSS , where nucleosomes form a well positioned array . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 005 While there has been extensive characterization of DNMT3 in vitro , a comprehensive analysis of the mechanisms guiding the activity of a de novo DNMT in vivo is still incomplete . To address this question , we introduced DNMT3b into an organism that has no endogenous DNA methylation machinery , the budding yeast Saccharomyces cerevisiae , to study the chromatin components affecting the activity of a mammalian de novo DNA methyltransferase . This system has several advantages over the study of DNA methylation in mammalian cells . Yeast has conserved histone sequences and many residues are modified at the same sites as those found in higher eukaryotes . However , unlike mammalian cells , yeast cells can be easily manipulated and the small size of their genome reduces costs associated with next-generation sequencing-based approaches . Moreover , yeast has already been used to show the importance of the N-terminus of histone H3 in targeting the DNA methylation complex ( Hu et al . , 2009 ) . Our data show that the chromatin template guides the activity of DNMT3b . DNMT3b preferentially deposits methylation in linker DNA compared to nucleosomal DNA . Also , DNMT3b activity correlates positively with H3K36me3 and negatively with H3K4me3 . In fact , mutation of the H3K36 methyltransferase Set2 decreases DNA methylation over regions that would normally contain H3K36me3 . Thus the marks themselves , as opposed to genomic features that correlate with these marks , are responsible for targeting DNA methylation . We also demonstrate that the pattern of H3K4 and H3K36 methylation in embryonic male germ cells accurately predicts which regions undergo de novo methylation , indicating that the mechanism observed in yeast is conserved in mammals .
S . cerevisiae does not have any endogenous cytosine DNA methyltransferases , and its DNA is therefore unmethylated . To study the activity of a de novo methyltransferase in this organism , we introduced the murine DNMT3b under the control of the inducible GAL1 promoter ( Figure 1B ) . We measured the levels of 5-methylcytosine ( 5meC ) in these strains using whole genome bisulfite sequencing ( WGBS ) ( Supplementary file 1A ) . We observed significant levels of 5meC of DNA extracted from the exponentially growing and stationary phases of the same strain culture ( Figure 1C and Supplementary file 2A ) , with higher methylation levels observed in stationary phase . CpG dinucleotides were preferentially methylated , as expected from the previously characterized activity of mammalian DNMT3 . The methylation levels of CpG dinucleotides range from 3 . 3 to 7 . 7% , depending on the yeast strain analyzed . These levels are about 10–20 times higher than the average of other dinucleotides levels ( Supplementary file 2A ) , and well above the bisulfite non-conversion rate of 0 . 27% , as estimated from an unmethylated lambda DNA spike-in . Despite some level of variability , we observe methylation across the entire yeast genome ( Figure 1—figure supplement 1A , B ) . When mapping reads to the genome we only retain those that map to a single position . As a result we do not obtain methylation estimates for regions that contain repetitive sequences , such as the rRNA containing regions in chromosome XII . We also observed a striking methylation distribution within genes ( Figure 1D ) , with low levels at the transcription start site ( TSS ) and increasing methylation in the gene body , reaching a maximum close to the transcription termination site ( TTS ) . The same pattern is found in mammals ( Lister et al . , 2009; Chodavarapu et al . , 2010 ) , suggesting that equivalent mechanisms regulating DNMT3 activity in mammalian genes might also be present in yeast . In yeast , nucleosomes are well positioned at the beginning of a gene , with nucleosome-free regions ( NFRs ) immediately upstream of the TSS and downstream of the TTS ( Brogaard et al . , 2012 ) . When average levels of 5meC are calculated around the TSS , we observed a periodicity of about 170 bp ( Figure 1—figure supplement 2 ) . A similar periodicity is also observed at the TTS . This suggested that nucleosomes might influence the activity of de novo DNMTs . To address this question , we measured nucleosome positioning genome-wide using micrococcal nuclease-digested chromatin and deep-sequencing ( MNase-seq ) ( Supplementary file 1B and Supplementary file 3A , B ) . We profiled the distribution of methylated cytosines at the TSS ( Figure 2A ) , TTS ( Figure 2B ) , and around each nucleosome center ( Figure 2C ) . 10 . 7554/eLife . 06205 . 006Figure 2 . Influence of nucleosome positioning on DNA methylation . Average distribution of nucleosomes and DNA methylation ( CpG context ) around ( A ) Transcriptional Start Site ( TSS ) , ( B ) Transcriptional Termination Site ( TTS ) , and ( C ) nucleosome centers . ( D ) Meta-nucleosome plot of CpG methylation . a . u . = Arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 00610 . 7554/eLife . 06205 . 007Figure 2—figure supplement 1 . Differences in nucleosome occupancy between DNMT3b-expressing and non-expressing yeast strains . Ratio of nucleosome occupancy between DNMT3b-expressing ( 3b ) and non-expressing ( EV ) yeast strains at TSS ( A ) and TTS ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 007 From these analyses , it is evident that DNMT3b preferentially methylates non-nucleosomal DNA . We observe a 50% increase in the methylation of linker DNA compared to nucleosome bound DNA ( Figure 2C ) . We also observe a slight 10 bp periodicity of methylated CpG ( Figure 2D ) , another feature shown in higher eukaryotes that reflects the periodicity of the DNA helix ( Klug and Lutter , 1981 ) . We considered the possibility that introducing 5meC would alter nucleosome distribution or gene expression in yeast . However , a comparison of DNMT3b-expressing and non-expressing strains showed no detectable change in nucleosome positioning by MNase treatment near the TSS , TTS ( Figure 2—figure supplement 1A , B and Supplementary file 3C ) , or elsewhere in the genome . RNA-seq analysis identified some differentially expressed genes ( about 5% of the total number of genes , with an equal number of up- and down-regulated transcripts ) between the strain expressing and non-expressing DNMT3b grown to stationary phase ( Figure 3 and Supplementary file 1C and Supplementary file 4A ) . The down-regulated genes showed enrichment for branched-chain aminoacid biosynthesis genes , while the up-regulated ones were enriched in ribosomal biogenesis genes ( Supplementary file 4B–F ) . However , these changes are likely due to stress response pathways that are triggered by the overexpression of MmDNMT3b , rather than by the changes in DNA methylation itself . In support of this view , when the levels of CpG , CpHpG , and CpHpH methylation in the up- and down-regulated genes were compared , no significant difference was evident ( Figure 3—figure supplement 1 ) . Moreover , the methylation levels of the differentially transcribed genes were not different from that of other members of the same Gene Ontology ( GO ) term ( Figure 3—figure supplement 2 ) . Since DNA methylation machinery is not native in yeast , it is likely that proteins able to recognize and mediate 5meC effects are also absent . 10 . 7554/eLife . 06205 . 008Figure 3 . Differences in RNA expression between DNMT3b-expressing and non-expressing yeast strains . The expression difference in RNA expression between DNMT3b and EV strains is plotted on the x axis , and false discovery rate ( FDR ) -adjusted significance is plotted on the y-axis ( –log2 scale ) . Upregulated and downregulated RNAs shown in red and green , respectively . Significantly expressed RNAs have a fold change bigger than two with a FDR smaller than 0 . 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 00810 . 7554/eLife . 06205 . 009Figure 3—figure supplement 1 . DNA Methylation in up- and down-regulated genes . Metagene plot of 5meC in different contexts ( CpG , CpHpG , CpHpH ) of upregulated ( red ) and downregulated ( green ) genes . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 00910 . 7554/eLife . 06205 . 010Figure 3—figure supplement 2 . DNA Methylation in ribosomal biogenesis genes . Metagene plot of 5meC in different contexts ( CpG , CpHpG , CpHpH ) of upregulated ribosomal biogenesis genes ( red ) compared to all the genes of the same class ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 010 We next sought to test whether the observed levels of 5meC could be explained by the underlying distribution of specific histone tail modifications . To address this , we mapped the distribution of DNMT3b and of specific histone residue modifications via ChIP-seq in both the DNMT3b-expressing and wild type ( wt ) ( non-expressing ) strains ( Supplementary file 1D ) . We found that , as expected , DNMT3b co-localizes with methylated regions ( Figure 4A ) . The distribution of DNMT3b is also consistent with the distribution of DNA methylation across the gene body ( Figure 4—figure supplement 1 ) . We also observed that DNMT3b and 5meC are strongly anti-correlated with H3K4me3 and positively correlated with H3K36me3 ( Figure 4B and Figure 4—figure supplements 2–4 ) . By examining the distribution of histone marks across gene bodies , we found that H3K4me3 is concentrated at the promoter while H3K36me3 levels peak near the 3′ end of the gene ( Figure 4—figure supplement 1 ) . These observations suggest that the ADD and PWWP domains of DNMT3B play a role in targeting the activity of the enzyme . H3K4me1 shows a weak positive correlation with both 5meC levels and DNMT3b . This might be due to the specific distribution of H3K4me1 within the gene body , partially overlapping to the H3K36me3 modification ( Figure 4A and Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 06205 . 011Figure 4 . Correlation between histone marks and DNA methylation . ( A ) Genome-wide distribution of nucleosome , 5meC , DNMT3b , H3K36me3 , H3K4me1 , H3K4me3 , and RNA polymerase II . ( B ) Spearman correlation coefficients between 5meC , histone marks , RNA polymerase II , DNMT3b and mRNA average levels for protein coding genes . ( C ) Prediction of DNMT3b levels using DNA methylation , H3K4 and H3K36 trimethylation , RNA polymerase II and nucleosome distribution as predictors . The y-axis shows the adjusted R2 value between the predicted linear model and observed values . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 01110 . 7554/eLife . 06205 . 012Figure 4—figure supplement 1 . Metagene plot of ChIP sequencing in a DNMT3b-expressing strain . ChIP-seq reads average intensity across yeast genes and 1 Kb upstream and downstream . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 01210 . 7554/eLife . 06205 . 013Figure 4—figure supplement 2 . Relationship between transcription and 5meC or histone marks levels . Average ChIP-seq intensity ( A–E ) or 5meC levels ( F ) across yeast genes divided in deciles based on RNA values ( RPKM ) . a . u . = Arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 01310 . 7554/eLife . 06205 . 014Figure 4—figure supplement 3 . Relationship between DNA methylation and histone marks levels . Average ChIP-seq ( A–E ) or DNA methylation ( F ) distribution across yeast genes divided in deciles based on average 5meCpG intragenic levels . a . u . = Arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 01410 . 7554/eLife . 06205 . 015Figure 4—figure supplement 4 . Relationship between H3K4me3 and 5meC or histone marks levels . Average ChIP-seq intensity ( A–E ) or 5meC levels ( F ) across yeast genes divided in deciles based on H3K4me3 average in the last third of each gene . a . u . = Arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 01510 . 7554/eLife . 06205 . 016Figure 4—figure supplement 5 . 5meC levels prediction using chromatin marks . Prediction of 5meC levels across the genome divided in 200-bp bins with a linear multivariate regression method using several combinations of chromatin marks . On the y-axis the adjusted R2 value is reported . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 016 5meC and DNMT3b distribution are also inversely correlated with gene transcription and Pol II abundance ( Figure 4B , Figure 4—figure supplements 2E–F , 3A ) . Both H3K4me3 and H3K36me3 correlate positively with transcription ( Figure 4B , Figure 4—figure supplement 2C–D ) . Since yeast genes are very small relative to mammalian genes , H3K4 methylation can spread well into the gene body ( Figure 4—figure supplement 2B–C ) and limit the deposition of 5meC in highly transcribed genes . In support of this observation , we find that a higher level of H3K4me3 in the last third of the gene , is associated with a lower level of DNMT3b or 5meC ( Figure 4—figure supplement 4 ) . To determine whether the methylation of H3K4 and H3K36 is sufficient to explain the observed DNA methylation of our DNMT3b strains , we constructed a simple linear model of DNA methylation based on our ChIP-seq data . We used linear multivariate regression to model whether the distribution of one or a few histone marks , nucleosome positioning or RNA polymerase II occupancy could predict the levels of DNMT3b or 5meC ( Figure 4C and Figure 4—figure supplement 5 ) . Strikingly , we found that H3K4me3 and H3K36me3 levels are sufficient to predict the distribution of both DNMT3b and 5meC with very high accuracy . The prediction could only be slightly improved by using additional data , suggesting that H3K4me3 and H3K36me3 are the key factors in determining the targeting of DNA methylation ( Supplementary file 5 ) . To determine whether H3K36me3 has a direct role in the recruitment/activity of DNMT3b in vivo , we measured the DNA methylation distribution in three mutant strains: set1Δ , set2Δ , and dot1Δ ( Supplementary file 1E ) . In yeast , Set1 is responsible for the methylation of H3K4 , Set2 is the methyltransferase for H3K36 , and Dot1 catalyzes the methylation of H3K79 . We included the dot1Δ strain as a control , since we do not expect its activity to influence the binding of DNMT3b . If the modification of H3K36 plays a role in DNMT3b activity we would expect a reduction in DNA methylation levels in gene bodies , which are the primary H3K36me3 positive regions . Due to an impact of the set mutations on global transcription , the levels of the induced DNMT3b and the resulting DNA methylation were lower in deletion strains than the wt . Nonetheless , the resulting 5meC levels were still significantly higher than background levels found in the wt strains ( Figure 5A and Supplementary file 2B ) . To account for the variations in global methylation levels we adopted two types of normalization: the first normalized by the total amount of DNA methylation in the sample and the second was based on the expression of DNMT3b measured via RT-qPCR ( Figure 5B and Figure 5—figure supplement 1 ) . Both strategies gave similar results ( data not shown ) . 10 . 7554/eLife . 06205 . 017Figure 5 . Effect of histone lysine methyltransferase deletions on the distribution of DNA methylation . ( A ) Metagene plot of CpG methylation in set1Δ and set2Δ cells expressing DNMT3b . Differently from Figure 5B , 5meC levels are not normalized . Replicates of the same strain are represented as dotted lines . Data from BY4741-derived strains . BY4741 = Wild type ( wt ) ; EV = Empty vector . ( B ) Metagene plot of CpG methylation in set1Δ , set2Δ , and dot1Δ cells expressing DNMT3b . set1Δ , set2Δ are in a BY4741 background , while dot1Δ is in a W303 background . 5meC levels are normalized by DNMT3b expression measured by RT-qPCR . Two replicates for each strain are shown ( solid and dotted line ) . ( C ) Metagene plots of CpG methylation ratio between the mutant and its wt counterpart . Two replicates for each mutant strain are shown ( solid and dotted line ) . Wt ratios ( =1 ) are represented by the horizontal dashed line ( green or blue ) . ( D ) Boxplots showing levels of DNA methylation in the wt ( left ) and set1Δ strain ( right ) of 200-bp genome bins sorted into deciles by H3K4me3 level . ( E ) Boxplots showing levels of DNA methylation in the wt ( left ) and set2Δ strain ( right ) of 200-bp genome bins sorted into deciles by H3K36me3 level . The dashed red line represents background levels of DNA methylation due to incomplete bisulfite conversion ( >99 . 7% ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 01710 . 7554/eLife . 06205 . 018Figure 5—figure supplement 1 . DNMT3b transcript levels in different yeast strains . Expression levels of DNMT3b of the wild-type strain compared to yeast mutants ( set1Δ , set2Δ , and dot1Δ ) . The relative levels were calculated using the ΔΔCt method ( Schmittgen and Livak , 2008 ) using TDH1 gene as reference . DNA methylation levels of each mutant used to produce Figure 5B , C were linearly scaled according to the reported averages of RT-qPCR replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 018 As expected , we see no significant differences in 5meC distribution in dot1Δ strains compared to wt ( Figure 5B ) . In contrast , in the set1Δ strain , we found that regions close to the TSS , with high H3K4me3 and low DNA methylation in a wt strain , contain methylation levels that are not significantly different from other regions outside of the gene ( Figure 5B , C ) . This suggests that H3K4 methylation plays an active role in suppressing DNA methylation in the wt , and that this effect disappears in the set1Δ strain ( Figure 5D ) . In a set2Δ strain , 5meC levels are reduced over gene bodies compared to wt strains ( Figure 5C ) . Moreover , in this strain maximum levels of DNA methylation peak outside of the gene , where H3K36me3 is not present ( Figure 5B ) . Thus , in this mutant strain DNA methylation is redistributed from gene bodies ( H3K36me3-rich regions ) to intergenic regions compared to the wt , suggesting that H3K36me3 is responsible for recruitment of DNMT3B ( Figure 5C , E ) . To extend our findings in yeast , we sought evidence to determine whether H3K36me3 also promotes de novo DNA methylation in mammals . The mouse germline is an excellent model for such studies . The mouse germline is specified from the epiblast at E7 . 25 and then progressively loses DNA methylation through subsequent rounds of cell division . By E13 . 5 , almost all DNA methylation has been lost ( Popp et al . , 2010; Seisenberger et al . , 2012 ) . In male germ cells , cell division halts , and the de novo methyltransferases and their co-factor DNMT3L are expressed between E13 . 5 and birth , when the genome undergoes global de novo DNA methylation ( Seisenberger et al . , 2012; Kobayashi et al . , 2013 ) . Thus in this setting , DNA methyltransferases are introduced into hypomethylated cells , and are therefore an ideal model to study the targeting of de novo DNA methylation . We mapped DNA methylation in the male germline at E16 . 5 , P2 . 5 , and P10 . 5 ( Supplementary file 1F ) ( Pastor et al . , 2014 ) , and obtained E13 . 5 DNA methylation data from published sources ( Seisenberger et al . , 2012 ) . Consistent with previous observations about the timing of de novo DNA methylation in the developing mouse germline , global CpG methylation rises from 7% at E13 . 5 to 55% at E16 . 5 and reaches at 75% by P2 . 5 ( Figure 6A ) . Previous studies have shown that the entire male germline genome is methylated by default , except for regions of H3K4 methylation such as TSSs which antagonize de novo DNA methylation ( Singh et al . , 2013 ) . However , charting the progression of DNA methylation over time , it is apparent that there exist ‘early methylating’ regions that reach their final methylation state by E16 . 5 and ‘late methylating’ regions that undergo substantial DNA methylation between E16 . 5 and P2 . 5 . We observed that heavily transcribed regions of chromosomes showed much higher DNA methylation at E16 . 5 than less transcribed regions ( Figure 6B ) . Furthermore , while the TSS of active genes was unmethylated , gene bodies of actively transcribed genes were typically early-methylators ( Figure 6B , D ) . Thus , transcriptional initiation correlates negatively with de novo DNA methylation while transcriptional elongation correlates positively with de novo methylation . 10 . 7554/eLife . 06205 . 019Figure 6 . H3K4me3 and H3K36me3 distribution predicts de novo DNA methylation pattern in male germline . ( A ) Genome-wide CG methylation levels during murine development as measured by bisulfite sequencing . ( B ) RNA-seq , and ChIP read abundance and relative DNA methylation levels are plotted across chromosome 14 . Note the correspondence between RNA-seq and H3K36me3 ChIP levels and rapid DNA methylation between E13 . 5 and E16 . 5 . ( C ) Boxplots showing the difference of DNA methylation levels between E13 . 5 and E16 . 5 of 1 Mb genome bins sorted into deciles by H3K36me3 level . ( D ) RNA-seq and ChIP read abundance and DNA methylation levels are plotted relative to transcriptionally active genes . The gene promoters contain high H3K4me3 and are not methylated , while the gene bodies contain high H3K36me3 and are methylated rapidly . ( E ) Metaplots showing DNA methylation level ±1000 bp relative to the TSS of genes sorted into deciles by H3K4me3 level . ( F ) Metagene plots showing DNA methylation across gene bodies sorted into deciles by H3K36me3 level . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 019 In light of the data from yeast , we considered that the trends noted above could be caused by the underlying chromatin environment , with H3K4me3 antagonizing and H3K36me3 promoting de novo DNA methylation . Since transcriptional elongation causes H3K36me3 deposition , we asked whether the association of transcriptional read-through with DNA methylation could explain the observed phenomenon . To test this hypothesis , we analyzed published H3K4me3 ChIP-seq data ( Lesch et al . , 2013 ) and performed H3K36me3 ChIP-seq on sorted germ cells of pooled E13 . 5 testis ( Supplementary file 1G ) . H3K4me3 at E13 . 5 correlates with low DNA methylation at all subsequent time points ( Figure 6D , E ) . Genes with high H3K36me3 levels at E13 . 5 showed significantly elevated gene-body DNA methylation at E16 . 5 , consistent with H3K36me3 accelerating DNA methylation ( Figure 6B , C , D , F ) . This trend was still apparent at P2 . 5 ( Figure 6F ) . Thus , H3K36me3 appears to direct DNA methylation in mammalian cells .
Our study aimed to identify chromatin features that affect the activity of mammalian de novo DNMTs in the establishment of DNA methylation . The expression of the murine DNMT3b in a host with no detectable levels of 5meC led to the methylation of CpG dinucleotides at different levels depending on the specific chromatin context . The presence of the H3K4me3 mark inhibits the activity of DNMT3b , while H3K36me3 promotes DNA methylation . This suggests that the activity of DNMT3B is guided by the interactions of the ADD and PWWP domains with histone tails . It has been recently shown ( Baubec et al . , 2015 ) that in embryonic stem cells the PWWP domain is responsible for the targeting of DNMT3b to regions enriched for the H3K36me3 . Similarly to our finding in yeast , the reintroduction of DNMT3b into methylation deficient DNMT1/DNMT3A/DNMT3B triple KO ( TKO ) ES cells partially restores 5meC levels . Methylation levels are higher at H3K36me3 sites , a trend eliminated by the ablation of the H3K36me3 methyltransferase Setd2 ( Baubec et al . , 2015 ) . Our findings are in agreement with the Baubec et al . observations , both in a system where other factors guiding DNA methylation are absent ( yeast ) , and during a period of biologically important de novo DNA methylation ( germ cells ) . In our yeast system , we detected an anti-correlation between transcript levels and DNA methylation , while we found a positive correlation in germ cells as was shown in ES cells ( Baubec et al . , 2015 ) . According to our findings , the levels of DNA methylation are guided by the presence of two transcription-dependent marks: H3K4 and H3K36 methylation . The discrepancy between the findings in yeast and germ cells can be explained by the difference in the length of their genes . Yeast genes are relatively small compared to genes in higher eukaryotes so , H3K4 methylation can spread within the body of the gene , thus preventing the binding of the DNMT3-ADD domain to the N-terminus of histone H3 and reducing its activity . In contrast , in mammals , H3K4me is localized to the start of the gene , and does not spread significantly within the gene body . Hence , highly transcribed genes in mammals show a strong enrichment of H3K4me3 around the TSS and H3K36me3 into the gene-body , shaping their intragenic DNA methylation distribution . The observation that transcriptional elongation is linked to DNA methylation has been noted in many contexts in addition to male germ cells . In mature oocytes , which have intermediate global levels of CpG methylation ( ∼50% ) , similar to male E16 . 5 PGCs , actively transcribed gene bodies have far higher levels of DNA methylation than less transcribed genes and intergenic regions ( Smallwood et al . , 2011; Kobayashi et al . , 2012 ) . Also , in oocytes , intragenic CpG islands show far higher DNA methylation than other CpG islands ( Smallwood et al . , 2011 ) . Transcriptional read-through is a common feature of maternally imprinted loci ( Weaver and Bartolomei , 2014 ) and ablation of an upstream promoter prevents proper methylation of the imprinted Gnas locus ( Chotalia et al . , 2009 ) . In mammalian soma , inactive X-chromosome shows higher promoter methylation , consistent with its silent state , but markedly lower intragenic methylation ( Hellman and Chess , 2007 ) . Transcriptional elongation is also correlated with DNA methylation in tumor cells ( Jin et al . , 2012 ) . It has been suggested that transcriptional read-through could ‘open’ chromatin for DNMTs , or that heterochromatin is physically inaccessible to DNMTs . We suggest however that direct recruitment of DNMTs by H3K36me3 is the most likely mechanism for the correlation between transcriptional read-through and DNA methylation . H3K36me3 functions both to suppress intragenic transcriptional initiation through recruitment of histone deacetylases , and to promote DNA methylation . These marks likely cooperate to induce lasting silencing of transcriptional initiation at target loci ( Figure 7 , Figure 7—figure supplement 1 ) . Intragenic TSSs originating at transposons have the potential to generate truncated or transposon/gene hybrid transcripts that could be deleterious to cell survival . H3K36me3 and DNA methylation could cooperate to silence these transposons in the germline and other periods of de novo methylation , and to maintain silencing through development . Moreover , where multiple TSSs exist for a gene , as in many imprinted loci , H3K36me3-mediated DNA methylation may serve to ensure the dominance of one promoter in a given cell type . 10 . 7554/eLife . 06205 . 020Figure 7 . Proposed model for de novo DNA methylation establishment . Model proposed for the targeting of DNMT3 during events of de novo 5meC establishment after genome-wide erasure of DNA methylation . Our model suggests that the presence of transcription-dependent histone modifications , such as H3K4me3 and H3K36me3 , determines the activity of DNMT3b in vivo . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 02010 . 7554/eLife . 06205 . 021Figure 7—figure supplement 1 . Factors affecting DNA methylation deposition . ( Top ) DNA methylation is negatively affected by H3K4me3 at the TSS . ( Bottom ) H3K36me3 promotes DNA methylation by ( A ) direct recruitment of DNMT3b and by ( B ) preventing the methylation of H3K4 , which antagonizes 5meC deposition . DOI: http://dx . doi . org/10 . 7554/eLife . 06205 . 021 A number of H3K36 methyltransferases exist in mammals but only one , SETD2 , can catalyze the conversion of H3K36me2 to me3 ( Wagner and Carpenter , 2012 ) . Setd2−/− mice exhibit profound vascular defects and die at E10 . 5–E11 . 5 ( Hu et al . , 2010 ) , while Setd2−/− are defective for differentiation toward endoderm ( Zhang et al . , 2014 ) . Setd2 is also a tumor suppressor mutated frequently in leukemia ( Zhu et al . , 2014 ) . It will be important to determine how loss of Setd2 affects the distribution of DNA methylation in the germline and soma , and whether loss of Setd2 contributes to aberrant methylation in cancer . More broadly , targeting of DNMT enzymes by association with H3K36me3 could explain methylation distribution across plants and animals . All catalytically active DNMT3-family methyltransferases in animals contain PWWP domains , and accordingly , gene body DNA methylation is observed in all animals that have retained DNMT3 enzymes . Preferential methylation of gene bodies over intergenic regions is observed for invertebrates such as honey bees ( Apis mellifera ) , sea squirts ( Ciona intestinalis ) , sea anemones ( Nematostella vectensis ) ( Zemach and Grafi , 2003; Feng et al . , 2010 ) . While the relationship between relative gene expression and gene-body methylation varies across these species , there is a strong correlation between gene-body H3K36me3 in Drosophila melanogaster genes and DNA methylation of homologous gene bodies in other invertebrates ( Nanty et al . , 2011 ) . DNA methylation is also associated with gene bodies in zebrafish ( Danio rerio ) ( Zemach and Grafi , 2003 ) and in mammalian contexts as discussed above . Finally , some chlorophyte algae have a ‘chlorophyte-type cytosine methylase’ , which evolved independently of DNMT3-family methyltransferases , which is fused to two PWWP domains ( Iyer et al . , 2011 ) . Thus , H3K36me3 could be relevant to DNA methylation targeting throughout the plant and animal kingdoms . | In animals and other multicellular organisms , there are many different types of cells that each perform particular roles in the body . This is possible because the genetic information—which is the same in all cells—is controlled so that only a subset of all the genes within an individual cell are ‘switched on’ at a particular time . Genetic information is contained within molecules of DNA , which are wrapped around proteins called histones . The genes in regions of DNA where these histones are packed tightly together tend to be switched off , while genes in regions of DNA that are loosely packed tend to be switched on . The level of packaging is controlled by the addition of ‘methyl’ tags to the histone proteins . These tags can also be added directly to the DNA in a process called DNA methylation . Enzymes called methyltransferases add the tags to the DNA , which tends to switch off the gene . The locations of the methyl tags can be copied when the DNA replicates before the cell divides so that the pattern of DNA methylation can be passed on to its daughter cells . However , it is not clear how the methyltransferases are able to target particular regions for methylation . To address this question , Morselli et al . introduced a methyltransferase called DNMT3b into yeast , a single-celled organism that does not normally add methyl tags to its DNA . The experiments show that the activity of the enzyme is affected by the presence of methyl tags on certain histone proteins . For example , a methyl tag at one particular site on a histone , called H3K4 , prevents the DNMT3b enzyme from adding methyl tags to DNA . However , a methyl tag at another site called H3K36 promotes DNA methylation . Morselli et al . found that these two histone sites had similar effects on DNA methylation in mouse sperm cells . Morselli et al . 's findings may be useful in the future development of treatments for cancer and other diseases that are caused by defects in DNA methylation . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"developmental",
"biology",
"genetics",
"and",
"genomics"
] | 2015 | In vivo targeting of de novo DNA methylation by histone modifications in yeast and mouse |
Chemical synaptic transmission relies on the Ca2+-induced fusion of transmitter-laden vesicles whose coupling distance to Ca2+ channels determines synaptic release probability and short-term plasticity , the facilitation or depression of repetitive responses . Here , using electron- and super-resolution microscopy at the Drosophila neuromuscular junction we quantitatively map vesicle:Ca2+ channel coupling distances . These are very heterogeneous , resulting in a broad spectrum of vesicular release probabilities within synapses . Stochastic simulations of transmitter release from vesicles placed according to this distribution revealed strong constraints on short-term plasticity; particularly facilitation was difficult to achieve . We show that postulated facilitation mechanisms operating via activity-dependent changes of vesicular release probability ( e . g . by a facilitation fusion sensor ) generate too little facilitation and too much variance . In contrast , Ca2+-dependent mechanisms rapidly increasing the number of releasable vesicles reliably reproduce short-term plasticity and variance of synaptic responses . We propose activity-dependent inhibition of vesicle un-priming or release site activation as novel facilitation mechanisms .
At chemical synapses , neurotransmitters ( NTs ) are released from presynaptic neurons and subsequently activate postsynaptic receptors to transfer information . At the presynapse , incoming action potentials ( APs ) trigger the opening of voltage gated Ca2+ channels , leading to Ca2+ influx . This local Ca2+ signal induces the rapid fusion of NT-containing synaptic vesicles ( SVs ) at active zones ( AZs ) ( Südhof , 2012 ) . In preparation for fusion , SVs localize ( dock ) to the AZ plasma membrane and undergo functional maturation ( priming ) into a readily releasable pool ( RRP ) ( Kaeser and Regehr , 2017; Verhage and Sørensen , 2008 ) . These reactions are mediated by an evolutionarily highly conserved machinery . The SV protein VAMP2/Synaptobrevin and the plasma membrane proteins Syntaxin-1 and SNAP25 are essential for docking and priming and the assembly of these proteins into the ternary SNARE complex provides the energy for SV fusion ( Jahn and Fasshauer , 2012 ) . The SNARE interacting proteins ( M ) Unc18s and ( M ) Unc13s ( where ‘M’ indicates mammalian ) are also essential for SV docking , priming and NT release ( Rizo and Südhof , 2012; Südhof and Rothman , 2009 ) , while Ca2+ triggering of SV fusion depends on vesicular Ca2+ sensors of the Synaptotagmin family ( Littleton and Bellen , 1995; Südhof , 2013; Walter et al . , 2011; Yoshihara et al . , 2003 ) . Cooperative binding of multiple Ca2+ ions to the SV fusion machinery increases the probability of SV fusion ( pVr ) in a non-linear manner ( Bollmann et al . , 2000; Dodge and Rahamimoff , 1967; Schneggenburger and Neher , 2000 ) . A distinguishing feature of synapses is their activity profile upon repeated AP activation , where responses deviate between successive stimuli , resulting in either short-term facilitation ( STF ) or short-term depression ( STD ) . This short-term plasticity ( STP ) fulfils essential temporal computational tasks ( Abbott and Regehr , 2004 ) . Postsynaptic STP mechanisms can involve altered responsiveness of receptors to NT binding , while presynaptic mechanisms can involve alterations in Ca2+ signalling and –sensitivity of SV fusion ( von Gersdorff and Borst , 2002; Zucker and Regehr , 2002 ) . Presynaptic STD is often attributed to high pVr synapses , where a single AP causes significant depletion of the RRP . In contrast , presynaptic STF has often been attributed to synapses with low initial pVr and a rapid pVr increase during successive APs . This was often linked to changes in Ca2+ signalling , for instance by rapid regulation of Ca2+ channels ( Borst and Sakmann , 1998; Nanou and Catterall , 2018 ) , saturation of local Ca2+ buffers ( Eggermann et al . , 2012; Felmy et al . , 2003; Matveev et al . , 2004 ) , or the accumulation of intracellular Ca2+ which may increase pVr either directly or via ‘facilitation sensors’ ( Jackman and Regehr , 2017; Katz and Miledi , 1968 ) . Alternatively , fast mechanisms increasing the RRP were proposed ( Fioravante and Regehr , 2011; Gustafsson et al . , 2019; Pan and Zucker , 2009; Pulido and Marty , 2017 ) . The coupling distance between Ca2+ channels and primed SVs is an important factor governing pVr ( Böhme et al . , 2018; Eggermann et al . , 2012; Stanley , 2016 ) . Previous mathematical models describing SV fusion rates from simulated intracellular Ca2+ transients have in many cases relied on the assumption of uniform ( or near uniform ) distances between SV release sites surrounding a cluster of Ca2+ channels and such conditions were shown to generate STF ( Böhme et al . , 2016; Meinrenken et al . , 2002; Nakamura et al . , 2015; Vyleta and Jonas , 2014 ) . However , alternative SV release site:Ca2+ channel topologies have been proposed , including two distinct perimeter distances , tight , one-to-one connections of SVs and channels , or random placement of either the channels , the SVs , or both ( He et al . , 2019; Böhme et al . , 2016; Chen et al . , 2015; Guerrier and Holcman , 2018; Keller et al . , 2015; Shahrezaei et al . , 2006; Stanley , 2016; Wong et al . , 2014 ) . So far , the precise relationship between SV release sites and voltage gated Ca2+ channels on the nanometre scale is unknown for most synapses , primarily owing to technical difficulties to reliably map their precise spatial distribution . However , ( M ) Unc13 proteins were recently identified as a molecular marker of SV release sites ( Reddy-Alla et al . , 2017; Sakamoto et al . , 2018 ) and super-resolution ( STED ) microscopy revealed that these sites surround a cluster of voltage gated Ca2+ channels in the center of AZs of the glutamatergic Drosophila melanogaster neuromuscular junction ( NMJ ) ( Böhme et al . , 2016; Böhme et al . , 2019 ) . Here , by relying on the unique advantage of being able to precisely map SV release site:Ca2+ channel topology we study its consequence for short-term plasticity at the Drosophila NMJ . Topologies were measured using electron microscopy ( EM ) following high pressure freeze fixation ( HPF ) or STED microscopy of Unc13 which both revealed a broad distribution of Ca2+ channel coupling distances . Stochastic simulations were key to identify facilitation mechanisms in the light of heterogenous SV release site:Ca2+ channel distances . Contrasting these simulations to physiological data revealed that models explaining STF through gradual increase in pVr ( from now on called ‘pVr-based models’ ) are inconsistent with the experiment while models of activity-dependent regulation of the RRP account for STP profiles and synaptic variance .
We first set out to quantify the SV release site:Ca2+ channel topology . For this we analysed EM micrographs of AZ cross-sections and quantified the distance between docked SVs ( i . e . SVs touching the plasma membrane ) and the centre of electron dense ‘T-bars’ ( where the voltage gated Ca2+ channels are located Fouquet et al . ( 2009 ) ; Kawasaki et al . ( 2004 ) ; Figure 1A ) . In wildtype animals , this leads to a broad distribution of distances ( ‘EM dataset wildtype’ , Figure 1—figure supplement 1A; Böhme et al . , 2016; Bruckner et al . , 2017 ) . At the Drosophila NMJ , the two isoforms Unc13A and –B confer SV docking and priming , but the vast majority ( ~95% ) of neurotransmitter release and docking of SVs with short coupling distances is mediated by Unc13A ( Böhme et al . , 2016 ) . We therefore investigated the docked SV distribution in flies expressing only the dominant Unc13A isoform ( Unc13A rescue , see Materials and methods for exact genotypes ) which showed a very similar , broad distribution of distances as wildtype animals ( ‘EM-dataset Unc13A rescue’ ) ( Reddy-Alla et al . , 2017; Figure 1A , B ) . In both cases , distance distributions were well described by a Rayleigh distribution ( Figure 1B , Figure 1—figure supplement 1A , solid green lines ) . The EM micrographs studied here are a cut cross-section of a three-dimensional synapse . To derive the relevant coupling distance distribution for all release sites ( including the ones outside the cross-section ) , the Rayleigh distribution was integrated around a circle ( Figure 1C ) , resulting in the following probability density function ( pdf , see Materials and methods for derivation ) :gx= 2π⋅σ3⋅x2⋅e-x2/ ( 2σ2 ) These pdfs were more symmetrical than the ones from the cross-sections and peaked at larger distances ( as expected from the increase in AZ area with increasing radius ) ( Figure 1D ) . The estimation of this pdf was very robust , resulting in near identical curves for the two EM datasets ( Figure 1—figure supplement 1B ) . We also used an independent approach to investigate the distribution of docked SV:Ca2+ channel coupling distances without relying on the integration of docked SV observations from cross-sections: since ( M ) Unc13 was recently described as a molecular marker of SV release sites ( Reddy-Alla et al . , 2017; Sakamoto et al . , 2018 ) we investigated AZ images of wildtype NMJs stained against Unc13A ( Böhme et al . , 2019 ) . Hundreds of individual AZ STED images ( lateral resolution of approx . 40 nm ) were aligned and averaged to obtain an average image of the AZ ( Figure 1E ) , which revealed a ring-like distribution of the Unc13A fluorescence . In previous works we had established that the voltage gated Ca2+ channels reside in the center of this ring ( Böhme et al . , 2016 ) . As this average image already reflects the distribution throughout the AZ area ( unlike for the EM data above where an integration was necessary ) the distribution of coupling distances can directly be computed based on pixel intensities and their distance to the AZ centre . Two independent datasets where analysed , resulting in very similar average images and distance distributions ( ‘wildtype STED dataset 1 and 2’ , Figure 1—figure supplement 1 ) . Remarkably , although the two approaches ( EM and STED microscopy ) were completely independent , the distributions of coupling distances quantified by either method coincided very well ( Figure 1F , Figure 1—figure supplement 1D; note that the integrated Rayleigh distributions were determined from EM micrographs and integration; they were NOT fit to the Unc13A distribution ) , supporting the accuracy of this realistic release site topology . The compliance between SV docking positions and Unc13A distribution further indicates that SVs dock to the plasma membrane where priming proteins are available , and therefore the entire distribution of docked SVs is potentially available for synaptic release ( Imig et al . , 2014 ) . Having identified the high degree of heterogeneity in the docked SV:Ca2+ channel coupling distances , we became interested in how this affected synaptic function . We therefore characterized synaptic transmission at control NMJs ( Ok6-GAL4 crossed to w[1118] ) in two electrode voltage clamp experiments . A common method to quantitatively evaluate synaptic responses and their STP behaviour is to vary the Ca2+ concentration of the extracellular solution which affects AP-induced Ca2+ influx ( see below ) . We used this approach and investigated responses evoked by repetitive ( paired-pulse ) AP stimulations ( 10 ms interval ) . In line with classical studies ( Dodge and Rahamimoff , 1967 ) , our results display an increase of the evoked Excitatory Junctional Current ( eEJC ) responses to the first AP ( eEJC1 amplitudes ) with increasing extracellular Ca2+ ( Figure 2A , B ) . STP was assessed by determining the paired-pulse ratio ( PPR ) : the amplitude of the second response divided by first . The eEJC2-amplitude was determined taking the decay of eEJC1 into account ( see insert in Figure 2C , Figure 2—figure supplement 1A ) . At low extracellular Ca2+ ( 0 . 75 mM ) , we observed strong STF ( with an average PPR value of 1 . 80 ) , which shifted towards depression ( PPR < 1 ) with increasing Ca2+ concentrations ( Figure 2C , D ) . Thus , the same NMJ displays both facilitation and depression depending on the extracellular Ca2+ concentration , making this a suitable model synapse to investigate STP behaviour . In panels B and D the mean eEJC1 amplitudes and PPRs from six animals are shown and the error bars indicate standard deviation , SD ( across all animals ) . We also examined the variation of repeated AP-evoked responses at the same NMJ between trials ( 10 s apart ) at different extracellular Ca2+ concentrations ( Figure 2E , F ) . At low concentrations ( 0 . 75 mM ) , the probability of transmitter release is low , resulting in a low mean eEJC1 amplitude with little variation ( Figure 2E , F , Figure 2—figure supplement 2 ) . With increasing extracellular Ca2+ , the likelihood of SV fusion increased and initially so did the variance ( e . g . at 1 . 5 mM extracellular Ca2+ ) . However , further increase in extracellular Ca2+ ( 3 mM , 6 mM , 10 mM ) led to a drop in variance ( Figure 2E , Figure 2—figure supplement 2 ) . Figure 2F depicts this average ‘variance-mean’ relationship from 6 cells ( means of cell means and means of cell variances , error bars indicate SEM ) . When assuming a binomial model , this approach has often been used to estimate the number of release sites nsites and the size of the postsynaptic response elicited by a single SV ( q ) ( Clements and Silver , 2000 ) . In agreement with previous studies of the NMJ this relationship was well described by a parabola with forced intercept at y = 0 and nsites = 164 and q = 0 . 64 nA ( Figure 2F , Figure 2—figure supplement 2; Matkovic et al . , 2013; Müller et al . , 2012; Weyhersmüller et al . , 2011 ) . Having determined the distribution of coupling distances ( Figure 1 ) and the physiological properties of the NMJ synapse ( Figure 2 ) , we next sought to compare how the one affected the other . There are two things two consider here . First of all , the SV release probability steeply depends on the 4th to 5th power of the local Ca2+ concentration ( Neher and Sakaba , 2008 ) . Secondly , because of the strong buffering of Ca2+ signals at the synapse , the magnitude of the AP-evoked Ca2+ transients dramatically declines with distance from the Ca2+ channel ( Böhme et al . , 2018; Eggermann et al . , 2012 ) . These two phenomena together make the vesicular release probability extremely sensitive to the coupling distance to the Ca2+ channels . Because we find that this distance is highly heterogeneous among SVs within the same NMJ , the question arises how these two properties ( heterogeneity of distances combined with a strong distance dependence of pVr ) functionally impact on synaptic transmission . Indeed , approaches by several labs to map the activity of individual NMJ AZs revealed highly heterogeneous activity profiles ( Akbergenova et al . , 2018; Gratz et al . , 2019; Muhammad et al . , 2015; Peled and Isacoff , 2011 ) . To quantitatively investigate the functional impact of heterogeneous SV placement , we wanted to use mathematical modelling to predict AP-induced fusion events of docked SVs placed according to the found distribution . A prerequisite for this is to first faithfully simulate local , AP-induced Ca2+ signals throughout the AZ ( such that the local transients at each docking site are known ) . We first determined the relevant AZ dimensions at the Drosophila NMJ , which , similarly to the murine Calyx of Held , is characterized by many AZs operating in parallel . We therefore followed previous suggestions from the Calyx using a box with reflective boundaries containing a cluster of Ca2+ channels in the base centre ( Meinrenken et al . , 2002 ) . The base dimensions ( length = width ) were determined as the mean inter-AZ distance of all AZs to their four closest neighbours ( because of the 4-fold symmetry ) from NMJs stained against the AZ-marker BRP ( Kittel et al . , 2006; Wagh et al . , 2006; Figure 3A ) . To save computation time , we further simplified to a cylindrical simulation ( where the distance to the Ca2+ channel is the only relevant parameter ) covering the same AZ area ( Figure 3B , Table 1 ) . To simulate the electrophysiological experiments above , where the extracellular Ca2+ concentration was varied ( Figure 2 ) , it was important to establish how the extracellular Ca2+ concentration influenced AP-induced Ca2+ influx . In particular , it is known that Ca2+ currents saturate at high extracellular Ca2+ concentrations ( Church and Stanley , 1996 ) . Unlike other systems , the presynaptic NMJ terminals are not accessible to electrophysiological recordings , so we could not measure the currents directly . We therefore used a fluorescence-based approach as a proxy . AP-evoked Ca2+ influx was assessed in flies presynaptically expressing the Ca2+-dependent fluorescence reporter GCaMP6m ( ;P{y[+t7 . 7] w[+mC]=20XUAS-IVS-GCaMP6m}attP40/Ok6-GAL4 ) . Fluorescence increase was monitored upon stimulation with 20 APs ( at 20 Hz ) while varying the extracellular Ca2+ concentration and showed saturation behaviour for high concentrations ( Figure 3—figure supplement 1 ) . This is consistent with a previously described Michaelis-Menten type saturation of fluorescence responses of a Ca2+-sensitive dye upon single AP stimulation at varying extracellular Ca2+ concentrations at the Calyx of Held , where half-maximal Ca2+ influx was observed at 2 . 6 mM extracellular Ca2+ ( Schneggenburger et al . , 1999 ) . This relationship was successfully used in the past to predict Ca2+ influx in modeling approaches Trommershäuser et al . ( 2003 ) . In our measurements , we determined a half maximal fluorescence response at a very similar concentration of 2 . 68 mM extracellular Ca2+ and therefore used this value as KM , current in a Michaelis-Menten equation ( Materials and methods , Equation 5 ) to calculate AP-induced presynaptic Ca2+ influx . The second parameter of the Michaelis-Menten equation , ( the maximal Ca2+ current charge , Qmax ) was optimized for each model ( Figure 3—figure supplement 2 , for parameter explanations and best fit parameters see Table 2 ) . We furthermore assumed that basal , intracellular Ca2+ concentrations at rest were also slightly dependent on the extracellular Ca2+ levels in a Michaelis-Menten relationship with the same dependency ( KM , current ) and a maximal resting Ca2+ concentration of 190 nM ( resulting in 68 nM presynaptic basal Ca2+ concentration at 1 . 5 mM external Ca2+ ) . With these and further parameters taken from the literature on Ca2+ diffusion and buffering ( see Table 1 ) the temporal profile of Ca2+ signals in response to paired AP stimulation ( 10 ms interval ) could be calculated at all AZ locations using the software CalC ( Matveev et al . , 2002; Figure 3C , Figure 3—figure supplement 2 ) . This enabled us to perform simulations of NT release from vesicles placed according to the distribution described above . In the past , we and others have often relied on deterministic simulations based on numerical integration of kinetic reaction schemes ( ordinary differential equations , ODEs ) . These are computationally effective and fully reproducible , making them well-behaved and ideal for the optimisation of parameters ( a property that was also used here for initial parameter searches , see Materials and methods ) . However , NT release is quantal and relies on only a few ( hundred ) SVs , indicating that stochasticity plays a large role ( Gillespie , 2007 ) . Moreover , deterministic simulations always predict identical output making it impossible to analyse the synaptic variance between successive stimulations , which is a fundamental hallmark of synaptic transmission and an important physiological parameter ( Figure 2F; Scheuss and Neher , 2001; Vere-Jones , 1966; Zucker , 1973 ) . Stochastic simulations allow a prediction of variance which can help identify adequate models that will not only capture the mean of the data , but also its variance . To compare this , data points are now shown with error bars indicating the square root of the average variance between stimulations within a cell ( Figure 4C , E , 6E , G and 7E , G ) . This is the relevant parameter since the model is designed to resemble an ‘average’ NMJ’ and therefore cannot predict inter-animal variance . Finally , as we show here deterministic simulations cannot be compared to experimentally determined PPR values because of Jensen’s inequality ( full proof in Materials and methods , see Figure 4—figure supplement 1 ) . Thus , stochastic simulations are necessary to account for SV pool sizes , realistic release site distributions , synaptic variance and STP . We thus implemented stochastic models of SV positions ( drawn randomly from the distribution above ) and SV Ca2+ binding states based on inhomogeneous , continuous time Markov models with transition rates governing reaction probabilities ( see Materials and methods for details ) . We also needed to consider where new SVs would ( re ) dock once SVs had fused and implemented the simplest scenario of re-docking in the same positions . This ensures a stable distribution over time and agrees with the notion that vesicles prime into pre-defined release sites , which are stable over much longer time than a single priming/unpriming event ( Reddy-Alla et al . , 2017 ) . The first model we tested was the single-sensor model proposed by Lou et al . ( 2005 ) , where an SV binds up to 5 Ca2+ ions , with each ion increasing its fusion rate or probability ( Figure 4A , Table 3 ) . Release sites were placed according to the distance distribution in Figure 1D and all sites were occupied by a primed SV prior to stimulation ( i . e . the number of release sites equals the number of vesicles in the RRP ) . Sites becoming available following SV fusion were replenished from an unlimited vesicle pool , making the model identical to the one described by Wölfel et al . ( 2007 ) . Ca2+ ( un ) binding kinetics were taken from Wölfel et al . ( 2007 ) Table 3 , the values of the maximal Ca2+ current charge ( Qmax ) , the SV replenishment rate ( krep ) and the number of release sites ( nsites ) were free parameters optimized to match the experimental data ( see Materials and methods for details , best fit parameters in Table 2 ) . To be able to compare the output of this and all subsequent models to experimental data as depicted in Figure 2 ( postsynaptic eEJC measurements ) , the predicted fusion events were convolved with a typical postsynaptic response to the fusion of a single SV ( mEJC , Figure 2—figure supplement 1B , see Materials and methods for more details ) . From the stochastic simulations ( 1000 runs each ) , we calculated the mean and variance of eEJC1 amplitudes , and the mean and variance of PPRs at various extracellular Ca2+ concentrations and contrasted these with the experimental data . This single-sensor model was able to reproduce the eEJC1 values ( Figure 4B , C ) . Moreover , the model accounted for the STD typically observed at high extracellular Ca2+ concentrations in the presence of rapid replenishment ( Hallermann et al . , 2010; Miki et al . , 2016 ) ( our best fit yielded τ ≈ 6 ms and reducing this rate led to unnaturally strong depression , Figure 4E , green curve+area ) . However , even despite rapid replenishment this model failed to reproduce the STF observed at low extracellular Ca2+ ( Figure 4D , E ) and the variances predicted by this model were much larger than found experimentally ( Figure 4F , G ) . The observation that eEJC1 amplitudes were well accounted for , but STPs were not , may relate to the fact that this model was originally constructed to account for a single Ca2+-triggered release event ( Lou et al . , 2005 ) . As this model lacks a specialized mechanism to induce facilitation , residual Ca2+ binding to the Ca2+ sensor is the only facilitation method which appears to be insufficient ( Jackman and Regehr , 2017; Ma et al . , 2015; Matveev et al . , 2002 ) . This result differs from our previous study using this model where we had placed all SVs at the same distance to Ca2+ channels which reliably produced STF ( Böhme et al . , 2016 ) . So why does the same model fail to produce STF with this broad distribution of distances ? To understand this we investigated the spatial distribution of SV release in simulations of the paired-pulse experiment at 0 . 75 mM extracellular Ca2+ ( Figure 5 ) . Figure 5A depicts two examples of synapses – seen from above – with SVs randomly placed according to the distance distribution in Figure 1D/5B . The synapse is shown immediately before AP1 , immediately after AP1 , immediately before AP2 ( i . e . after refilling ) and immediately after AP2 ( the external Ca2+ concentration was 0 . 75 mM ) . From this analysis it becomes clear that the pVr1 caused by AP1 essentially falls to zero around the middle of the SV distribution ( Figure 5B , top panel ) . This means that only SVs close to the synapse center fuse , and these high-pVr SVs are depleted by AP1 . SV replenishment refills the majority ( but not all ) of those sites and thus AP2/pVr2 essentially draws on the same part of the distribution ( Figure 5B , bottom panel ) . Because of this , and because the refilling is incomplete , this causes STD . Even with faster replenishment ( which would be incompatible with the low PPR values at high extracellular Ca2+ , Figure 4E ) this scenario would only lead to a modest increase of the PPR to values around 1 . Therefore , our analysis reveals that large variation in Ca2+ channel distances results in a specific problem to generate STF . Our analysis further indicates that with the best fit parameters of the single sensor model , the majority of SVs ( those further away ) is not utilized at all . The single-sensor model failed to reproduce the experimentally observed STF at low extracellular Ca2+ concentrations because of the dominating depletion of SVs close to Ca2+ channels , and the inability to draw on SVs further away . However , this situation may be improved by a second Ca2+ sensor optimized to enhance the pVr2 in response to AP2 . Indeed , in the absence of the primary Ca2+ sensor for fusion , Ca2+ sensitivity of synaptic transmission persists , which was explained by a dual sensor model ( Sun et al . , 2007 ) . It was recently suggested that syt-7 functions alongside syt-1 as a Ca2+ sensor for release ( Jackman et al . , 2016 ) , and deterministic mathematical dual fusion-sensor model assuming homogeneous release probabilities ( which implies homogeneous SV release site:Ca2+ channel distances ) was shown to generate facilitation ( Jackman and Regehr , 2017 ) . Similarly , stochastic modelling of NT release at the frog NMJ also showed a beneficial effect of a second fusion sensor for STF ( Ma et al . , 2015 ) . We therefore explored whether a dual fusion sensor model could account for synaptic facilitation from realistic release site topologies . The central idea of this dual fusion-sensor model is that while syt-1 is optimized to detect the rapid , AP-induced Ca2+ transients ( because of its fast Ca2+ ( un ) binding rates , but fairly low Ca2+ affinity ) , the cooperating Ca2+ sensor is optimized to sense the residual Ca2+ after this rapid transient ( Figure 3C ) ( with slow Ca2+ ( un ) binding , but high Ca2+ affinity ) . The activation of this second sensor after ( but not during ) AP1 could then enhance the release probability of the remaining SVs for AP2 ( Figure 6A , B ) . This is illustrated in Figure 6B , where k2 ( the on-rate of Ca2+ binding to the slow sensor ) is varied resulting in different time courses and amounts of Ca2+ binding to the second sensor . Increasing the release probability is equivalent to lowering the energy barrier for SV fusion ( Schotten et al . , 2015 ) . In this model both sensors regulate pVr and therefore additively lower the fusion barrier with each associated Ca2+ ion ( Figure 6A ) , resulting in multiplicative effects on the SV fusion rate . While the fast fusion reaction appears to have a 5-fold Ca2+ cooperativity ( Bollmann et al . , 2000; Burgalossi et al . , 2010; Schneggenburger and Neher , 2000 ) , it is less clear what the Ca2+ cooperativity of a second Ca2+ sensor may be , although the fact that the cooperativity is reduced in the absence of the fast sensor ( Burgalossi et al . , 2010; Kochubey and Schneggenburger , 2011; Sun et al . , 2007 ) could be taken as evidence for a Ca2+ cooperativity < 5 . We explored cooperativities 2 , 3 , 4 , and 5 ( cooperativities 2 and 5 are displayed in Figure 6 and Figure 6—figure supplement 1 ) . It is furthermore not clear whether such a sensor would be targeted to the SV ( like syt-1 /-2 ) , or whether it is present at the plasma membrane . Both scenarios are functionally possible and it was indeed reported that syt-7 is predominantly or partly localized to the plasma membrane ( Sugita et al . , 2001; Weber et al . , 2014 ) . A facilitation sensor on the plasma membrane would be more effective , which our simulations confirmed ( not shown ) , because it would not be consumed by SV fusion , allowing the sensor to remain activated . We therefore present this version of the model here . We used a second sensor with a Ca2+ affinity of KD = 1 . 5 μM ( Brandt et al . , 2012; Jackman and Regehr , 2017 ) . Like for the single-sensor model , all release sites were occupied with releasable vesicles ( nsites equals the number of RRP vesicles ) and their locations determined by drawing random numbers from the pdf . When fitting this model five parameters were varied: Qmax , krep , and nsites ( like in the single-sensor model ) together with k2 ( Ca2+ association rate constant to the second sensor ) and s ( the factor describing the effect of the slow sensor on the energy barrier for fusion ) ( see Table 2 for best fit parameters ) . The choice of k2 had an effect on the PPR in simulations , confirming that the second sensor was able to improve the release following AP2 ( Figure 6C ) . Figure 6D–I show that the dual fusion-sensor model could fit the eEJC1 amplitudes and the model slightly improved the higher PPR values at the low- and the lower PPR values at high extracellular Ca2+ concentrations compared to the single sensor model ( compare Figures 4E and 6G ) . However , the model failed to produce the STF observed experimentally ( the PPR values at 0 . 75 mM Ca2+were ~ 1 . 08 in the simulation compared to ~ 1 . 80 in the experiments ) . Another problem of the dual fusion-sensor model was that release became more asynchronous than observed experimentally ( Figure 6D ) , which was due to the triggering of SV fusion in-between APs . Finally , predicted variances were much larger than the experimental values ( Figure 6I ) . In addition to the optimization , we systematically investigated a large region of the parameter space ( Figure 6J , K ) , but found no combination of parameters that would be able to generate the experimentally observed STF . Lowering the Ca2+ influx ( by decreasing Qmax ) yielded a modest increase in PPR values ( Figure 6J ) , but required a large number of release sites ( nsites ) to match the eEJC1 amplitudes ( Figure 6K ) . Changing s had the largest effect when k2 was close to the best fit value and moving away from this value decreased the PPRs , either by increasing the effect of the second sensor on AP1 ( when increasing k2 ) or by decreasing the effect on AP2 ( when decreasing k2 ) , which both counteracts STF ( Figure 6B , J ) . Fitting the dual fusion-sensor model with a Ca2+ cooperativity of 5 did not improve the situation ( Figure 6—figure supplement 1 , best fit parameters in Table 2 ) : Although slightly more facilitation was observed , this model suffered from even larger variance overshoots ( Figure 6—figure supplement 1E ) and excessive asynchronous release ( Figure 6—figure supplement 1A , C ) . We explored different KD values between 0 . 5 and 2 μM at cooperativities 2–5 in separate optimizations , but found no satisfactory fit of the data ( results not shown ) . Thus , a dual fusion-sensor model is unlikely to account for STF observed from the realistic SV release site topology at the Drosophila NMJ . Note that this finding does not rule out that syt-7 functions in STF , but argues against a role in cooperating alongside syt-1 in a pVr-based facilitation mechanism . Since dual fusion-sensor models and other models depending on changes in pVr ( see Discussion ) are unlikely to be sufficient , we next investigated mechanisms involving an activity-dependent regulation of the number of participating release sites . For this we extended the single-sensor model by a single unpriming reaction ( compare Figures 4A and 7A ) . The consequence of reversible priming is that the initial release site occupation can be less than 100% ( in which cases nsites can exceed the number of RRP vesicles ) . This enables an increase ( ‘overfilling’ ) of the RRP ( /increase in site occupancy ) during the inter-stimulus interval ( consistent with reports in other systems Dinkelacker et al . , 2000; Gustafsson et al . , 2019; Pulido et al . , 2015; Smith et al . , 1998; Trigo et al . , 2012 ) . We assumed that Ca2+ would stabilize the RRP/release site occupation by slowing down unpriming ( Figure 7A ) . This made the steady-state RRP size dependent on the resting Ca2+ concentration and the modest dependence of this on the extracellular Ca2+ resulted in RRP enlargement with increasing extracellular Ca2+ ( Figure 7B ) , in agreement with recent findings on central synapses ( Malagon et al . , 2020 ) . This model ( like the dual fusion-sensor models depicted in Figure 6 and Figure 6—figure supplement 1 ) includes two different Ca2+ sensors , but the major difference is that these Ca2+ sensors operate to regulate two separate sequential steps ( priming and fusion ) . Indeed , this scenario aligns with reports of a syt-7 function upstream of SV fusion ( Liu et al . , 2014; Schonn et al . , 2008 ) . Figure 7C shows how the number of RRP vesicles develops over time in this model during a paired-pulse experiment for low and high extracellular Ca2+ concentrations . In all cases , SV priming was in equilibrium prior to the first stimulus , indicated by the horizontal lines ( 0–2 ms , Figure 7C ) . Note that prior to AP1 priming is submaximal ( ~41% ) for 0 . 75 mM extracellular Ca2+ , but near complete ( ~99% ) at 10 mM extracellular Ca2+ . At low extracellular Ca2+ the elevation of Ca2+ caused by AP1 results in a sizable inhibition of unpriming , leading to an increase ( ‘overfilling’ ) of the RRP during the inter-stimulus interval . With this , more primed SVs are available for AP2 , causing facilitation ( green line in Figure 7C ) . In contrast , at high extracellular Ca2+ concentrations , the rate of unpriming is already low at steady state and the RRP close to maximal capacity ( grey line in Figure 7C ) . At this high extracellular Ca2+ concentration , AP1 induces a larger Ca2+ current ( higher pVr ) , resulting in strong RRP depletion , of which only a fraction recovers between APs ( as in the other models , replenishment commences with a Ca2+ independent rate krep ) . Because Ca2+ acts in RRP stabilization , not in stimulating forward priming , this model ( unlike the dual fusion-sensor models in Figure 6 and Figure 6—figure supplement 1 ) did not yield asynchronous release in-between APs ( Figure 7D ) . Thus , the two most important features of this model are the submaximal site occupation and an inhibition of unpriming by intracellular Ca2+ . In this model we assumed a Ca2+ cooperativity of n = 5 for the unpriming mechanism ( we also explored n = 2 , see Figure 7—figure supplement 1 ) . The following parameters were optimized: Qmax , nsites and krep ( like in the single- and dual fusion-sensor models ) , together with KM , prim , the Ca2+ affinity of the priming sensor , and u , its Ca2+ cooperativity . These values together define the Ca2+-dependent unpriming rate ( see Table 2 for best fit parameters ) . The total number of fitted parameters ( 5 ) was the same as for the dual fusion-sensor models ( Figure 6 and Figure 6—figure supplement 1 ) . Figure 7D–I present the results . It is clear that both eEJC1 amplitudes and PPR values were described very well with this model at all extracellular Ca2+ concentrations . In addition , the variance-mean relationship of the eEJC1 was reproduced satisfactorily , except for a small variance overshoot for the highest extracellular Ca2+ concentrations ( Figure 7I , see Discussion ) . Fitting of the unpriming model with a Ca2+ cooperativity of 2 also led to a good fit ( Figure 7—figure supplement 1 ) , although the variance overshoot was somewhat larger . We also explored the time-dependence of the facilitation by simulating PPR values for various inter-stimulus intervals at different extracellular Ca2+ concentrations which could be investigated experimentally in the future to further refine parameters ( Figure 7—figure supplement 2 ) . Different facilitating synapses exhibit a large range of PPR values , some larger than observed at the Drosophila NMJ ( Jackman et al . , 2016 ) . Therefore , if this were a general mechanism to produce facilitation , we would expect it to be flexible enough to increase the PPR much more than observed here . To investigate the model’s flexibility we systematically explored the parameter space by varying Qmax , KM , prim , and u ( Figure 7J , K ) . Similar to Figure 6J , K , the colors of the balls represent the PPR value and the number of release sites needed to fit the eEJC1 amplitudes . Consistent with a very large dynamic range of this mechanism , PPR values ranged from 0 . 85 to 3 . 90 ( Figure 7J , K ) and unlike the dual fusion-sensor model , PPR values were fairly robust to changes in Ca2+ influx ( note the different scales on Figure 7J , K and Figure 6J , K ) . Moreover , because this mechanism does not affect the Ca2+ sensitivity of SV fusion , facilitation was achieved without inducing asynchronous release ( Figure 7D ) . We also investigated an alternative model based on Ca2+-dependent release site activation . In this model , all sites are occupied by a vesicle , but some sites are inactive and fusion is only possible from activated sites . We assumed that site activation was Ca2+-dependent . In order to avoid site activation during AP1 , which would again hinder STF and could contribute to asynchronous release , we implemented an intermediate delay state ( Figure 7—figure supplement 3A–B ) from which sites were activated in a Ca2+-independent reaction . This could mean that priming occurs in two-steps , with the first step being Ca2+-dependent . Similar to the unpriming model presented above , the modest increase of intracellular Ca2+ with extracellular Ca2+ yielded an RRP increase ( /increase in active sites ) ( Figure 7—figure supplement 3I ) . This model agreed similarly well with the data as the unpriming model ( Figure 7—figure supplement 3C–H ) . Thus , both mechanisms which modulate the RRP rather than pVr are fully capable of reproducing the experimentally observed Ca2+-dependent eEJC1 amplitudes , STF , release synchrony and variance . The unpriming model was preferred since it had fewer parameters and performed slightly better in optimisations than the site activation model . Why do nsite/priming-based mechanisms ( Figure 7 , Figure 7—figure supplement 1 , Figure 7—figure supplement 3 ) account for STF from the broad distribution of SV release site:Ca2+ channel coupling distances , while the pVr-based models ( Figures 4 and 6 , Figure 6—figure supplement 1 ) cannot ? To gain insight into this , we analysed the spatial dependence of transmitter release in the unpriming model during the paired-pulse experiment ( 0 . 75 mM extracellular Ca2+ ) in greater detail ( Figure 8 ) . Panel 8A , similarly to Figure 5A , shows example stochastic simulations ( at external Ca2+ concentration 0 . 75 mM , to illustrate facilitation ) . The best fit parameters of the unpriming model predicted a larger Ca2+ influx ( 1 . 64-fold and 3 . 05-fold larger Qmax value ) than the single- and dual fusion-sensor models ( Table 2 ) . The larger Ca2+ influx compensated for the submaximal priming of SVs ( reduced release site occupancy ) prior to the first stimulus by expanding the region where SVs are fused ( Figure 8B ) . Comparing to Figure 5B , a much larger part of the SV distribution is utilized during the first stimulus . Following AP1 , vesicles prime into empty sites across the entire distribution , allowing AP2 to draw again from the entire distribution . During this time , the increased residual Ca2+ causes overfilling of the RRP , that is more release sites are now occupied , giving rise to more release during AP2 . Notably , the AP2-induced release again draws from the entire distribution . Thus , the unpriming model not only reproduces STF and synaptic variance , but also utilizes docked SVs more efficiently from the entire distribution compared to the single- and dual fusion-sensor model .
We here described a broad distribution of SV release site:Ca2+ channel coupling distances in the Drosophila NMJ and compared physiological measurements with stochastic simulations of four different release models ( single-sensor , dual fusion-sensor , Ca2+-dependent unpriming and site activation model ) . We showed that the two first models ( single-sensor and dual fusion-sensor ) , where residual Ca2+ acts on the energy barrier for fusion and results in an increase in pVr , failed to reproduce facilitation . The two latter models involve a Ca2+-dependent regulation of participating release sites and reproduced release amplitudes , variances and PPRs . Therefore , the Ca2+-dependent accumulation of releasable SVs is a plausible mechanism for paired-pulse facilitation at the Drosophila NMJ , and possibly in central synapses as well . In more detail , our insights are as follows: In our model , all primed vesicles have identical properties , and only deviate in their distance to the Ca2+ channel cluster ( positional priming , Neher and Brose , 2018 ) . Alternatively , several vesicle pools with different properties ( molecular priming ) could be considered , which might involve either vesicles with alternative priming machineries , or vesicles being in different transient states along the same ( slow ) priming pathway ( Walter et al . , 2013 ) . In principle , if different primed SV states are distributed heterogenously such that more distant vesicles are more primed/releasable , such an arrangement might counteract the effects of a broad distance distribution , although this is speculative . Without such a peripheral distribution , the existence of vesicles in a highly primed/releasable state ( such as the ‘super-primed’ vesicles reported at the Calyx of Held synapse ) , would result in pronounced STD , and counteract STF , which indeed has been observed ( Lee et al . , 2013; Taschenberger et al . , 2016 ) . In this study electrophysiological recordings were performed on muscle 6 of the Drosophila larva which receives input from morphologically distinct NMJs containing big ( Ib ) and small ( Is ) synaptic boutons , which have been shown to differ in their physiological properties ( Atwood et al . , 1993; He et al . , 2009; Newman et al . , 2017 ) . This could add another layer of functional heterogeneity in the postsynaptic responses analysed here ( the EM and STED analyses shown here were focused on Ib inputs ) . Because our model does not distinguish between Is and Ib inputs , the estimated parameters represent a compound behaviour of all types of synaptic input to this muscle . Future investigations to isolate the contribution of the different input types ( e . g . by genetically targeting Is/Ib-specific motoneurons using recently described GAL4 lines; Pérez-Moreno and O'Kane , 2019 ) could help distinguish between inputs and possibly further refine the model to identify parameter differences between these input types . Figure 9 summarizes the results for the single-sensor , dual fusion-sensor and unpriming models . Facilitation in single and dual fusion-sensor models depend on the increase in release probability from the first AP to the next ( compare colored rings representing 25% release probability between row 2 and 4 ) . However , the increase is very small , even for the dual fusion-sensor model , and to nevertheless produce some facilitation , optimisation finds a small Ca2+ influx , which leads to an ineffective use of the broad vesicle distribution ( and a too-high estimate of nsites ) . In the unpriming model a higher fitted Ca2+ influx ( QMax ) leads to a more effective use of the entire SV distribution , and facilitation results from the combination of incomplete occupancy of release sites before the first AP ( row 1 ) , combined with ‘overshooting’ priming into empty sites between APs ( row 3 ) . Molecularly , syt-7 was linked to STF behaviour ( Jackman et al . , 2016 ) , and our data does not rule out that syt-7 is essential for STF at the Drosophila NMJ . However , we show clearly that a pVr-based facilitation mechanism ( dual fusion-sensor model ) cannot account for STF in synapses with heterogeneous distances between release sites and Ca2+ channels . Interestingly , syt-7 was also reported to function in vesicle priming and RRP replenishment ( Liu et al . , 2014; Schonn et al . , 2008 ) . Thus , future work will be necessary to investigate whether the function of syt-7 in STF might take place by Ca2+-dependent inhibition of vesicle unpriming or release site activation . Similar suggestions that facilitation results from a build-up of primed SVs during stimulus trains were made for the crayfish NMJ and mammalian synapses ( Gustafsson et al . , 2019; Pan and Zucker , 2009; Pulido and Marty , 2018 ) . This is in line with our results , with facilitation arising from modulation of the number of primed SVs rather than pVr . Our models are conceptually simple ( e . g . all SVs are equally primed and distinguished only by distance to Ca2+ channels , sometimes referred to as ‘positional priming’ Neher and Sakaba , 2008 ) , and we improved conceptually on previous work by using estimated SV release site:Ca2+ channel distributions , stochastic simulations and comparison to variance-mean relationships and we performed a systematic comparison of pVr- and priming-based models . It has not been clear whether increases in primed SVs are also required for paired-pulse facilitation , or only become relevant in the case of ‘tonic’ synapses that build up release during longer stimulus trains ( frequency facilitation Neher and Brose , 2018 ) . Paired-pulse facilitation is a more wide-spread phenomenon in synapses than frequency facilitation , and we show here for the case of Drosophila NMJ that it also seems to require priming-based mechanisms . Thus , Ca2+-dependent increases of the RRP during STP might be a general feature of chemical synapses .
Flies were kept under standard laboratory conditions as described previously ( Sigrist et al . , 2003 ) and reared on semi-defined medium ( Bloomington recipe ) at 25°C , except for GCaMP6m and synapGCaMP6f flies which were kept at room temperature , and Ok6-GAL4/+ ( Figure 2 , Figure 2—figure supplement 1 , Figure 4 panel B-E and G , Figure 6 panel D-G and I , Figure 6—figure supplement 1 , Figure 7D–G and I , Figure 7—figure supplement 1 , Figure 7—figure supplement 3C–F , H ) which were kept at 29°C ( for detailed genotypes see below ) . For experiments both male and female 3rd instar larvae were used . The following genotypes were used: Figure 7 Ok6-GAL4/+ ( Ok6-Gal4/II crossed to w[1118]; panel D-G , ( I ) . Figure 7—figure supplement 1: Ok6-GAL4/+ ( Ok6-Gal4/II crossed to w[1118] ) . Figure 7—figure supplement 3: Ok6-GAL4/+ ( Ok6-Gal4/II crossed to w[1118]; panel C-F , ( H ) . The following stocks were used: Ok6-GAL4/II ( Aberle et al . , 2002 ) , UAS-Unc13A-GFP/III ( Böhme et al . , 2016 ) , elav-Gal4/I ( Lin and Goodman , 1994 ) . The following stock were obtained from the Bloomington Drosophila Stock Center: P{w[+mC]=Mhc-SynapGCaMP6f}3–5/III ( Newman et al . , 2017 ) and w[1118]; P{y[+t7 . 7] w[+mC]=20XUAS-IVS-GCaMP6m}attP40 . The following stock was obtained from Kyoto Stock Center: P84200/IV . Sample preparation , EM image acquisition and the quantification of docked SV distances to the AZ center ( center of the electron dense ‘T-bar’ ) are described in Böhme et al . ( 2016 ) ; Reddy-Alla et al . ( 2017 ) . The Rayleigh distributions were fit to the distances of docked SVs to the T-bar pedestal center , which had been collected in two EM datasets; analyses of these datasets were published in two previous studies , ( Reddy-Alla et al . , 2017 ) for the histogram of distances depicted in Figure 1A and ( Böhme et al . , 2016 ) for the histogram of distances depicted in Figure 1—figure supplement 1A . The distances between Ca2+ channels and docked SVs in Drosophila NMJ obtained by EM was found to follow a Rayleigh distribution with best fit scale parameter σ = 76 . 51 nm ( EM dataset 1 ) and σ = 74 . 07 nm ( EM dataset 2 ) . The fitting was performed with a MATLAB ( MathWorks , version R2018b ) function , raylfit , which uses maximum likelihood estimation . As these distances are found by EM of a cross-section of the active zone , we integrate this distribution around a circle to obtain the two-dimensional distribution of SVs in the circular space around the active zone . The Rayleigh distribution has the following probability density function ( pdf ) :f ( x ) = xσ2e−x2/2σ2 , x>0 The pdf of the SV distribution will then be a scaling of the following function ( 1 ) g^ ( x ) =2πxf ( x ) =2πxxσ2e−x2/2σ2 In order to find the pdf of the 2D SV distribution , we integrate g^ to find the normalizing constant . By integration by parts we get∫0∞g^ ( x ) dx=∫0∞2πx1σ2xe−x22σ2 dx=2π ( [−xe−x22σ2]0∞+∫0∞e−x22σ2 dx ) =2π∫0∞e−x22σ2 dx=2π 12σ2πwhere the standard normal distribution was used in the last equality . Normalising ( 1 ) by this constant , we get the pdf of the distance distribution on a circular area in the active zone:gx=2π⋅σ3⋅x2⋅e-x2/2σ2 In order to use the above SV distribution in simulations , we need to determine probabilities . g ( x ) is a generalized gamma distribution with a=2⋅σ , p=2 , d=3 . The generalized gamma distribution with a>0 , p>0 , d>0 has the following pdf:hx;a , d , p= pad⋅xd-1⋅e-xapΓd/pand cumulative density function ( cdf ) :Hx;a , d , p= γd/p , x/apΓd/pwhere γ is the lower incomplete gamma function , and Γ is the ( regular ) gamma function . Both of these functions are implemented in MATLAB ( MathWorks , version R2018b ) , which easily allows us to draw numbers from them . Thus , the SV distribution has the following cdf: G ( x ) =γ ( 1 . 5 , ( x2/2σ2 ) ) Γ ( 1 . 5 ) That is , given a uniformly distributed variable q∈ ( 0 , 1 ) , we can use inbuilt MATLAB functions to sample SV distances , d: ( 2 ) d=G−1 ( q ) =γ−1 ( 1 . 5 , q⋅Γ ( 1 . 5 ) ) ⋅2σ2 The implementation is as follows: Note that in MATLAB the inverse incomplete gamma function with parameter s is scaled by Γ ( s ) , which is why we input q and not q/Γ ( 1 . 5 ) . Sample preparation , Unc13A antibody staining , STED image acquisition and the isolation of single AZ images are described in Böhme et al . ( 2019 ) and in the following . Third-instar w[1118] larvae were put on a dissection plate with both ends fixed by fine pins . Larvae were then covered by 50 µl of ice-cold hemolymph-like saline solution ( HL3 , pH adjusted to 7 . 2 [Stewart et al . , 1994]: 70 mM NaCl , 5 mM KCl , 20 mM MgCl2 , 10 mM NaHCO3 , 5 mM Trehalose , 115 mM D-Saccharose , 5 mM HEPES ) . Using dissection scissors a small cut at the dorsal , posterior midline of the larva was made from where on the larvae was cut completely open along the dorsal midline until its anterior end . Subsequently , the epidermis was pinned down and slightly stretched and the internal organs and tissues removed . For the ‘STED dataset 2’ shown in Figure 1—figure supplement 1C , D , animals were then incubated in a HL3 solution containing 0 . 5% DMSO for 10 min ( this served as a mock control for another experiment not shown in this paper using a pharmacological agent diluted in DMSO ) . The dissected samples were washed 3x with ice-cold HL3 and then fixed for 5 min with ice-cold methanol . After fixation , samples were briefly rinsed with HL3 and then blocked for 1 hr in 5% native goat serum ( NGS; Sigma-Aldrich , MO , USA , S2007 ) diluted in phosphate buffered saline ( Carl Roth Germany ) with 0 . 05% Triton-X100 ( PBT ) . Subsequently dissected samples were incubated with primary antibodies ( guinea-pig Unc13A 1:500; Böhme et al . , 2016 ) diluted in 5% NGS in PBT overnight . Afterwards samples were washed 5x for 30 min with PBT and then incubated for 4 hr with fluorescence-labeled secondary antibodies ( goat anti-guinea pig STAR635 ( 1:100 ) diluted in 5% NGS in PBT . For secondary antibody production STAR635 fluorophore ( Abberior , Germany ) was coupled to respective IgGs ( Dianova , Germany ) . Samples were then washed overnight in PBT and subsequently mounted in Mowiol ( Max-Planck Institute for Biophysical Chemistry , Group of Stefan Hell ) on high-precision glass coverslips ( Roth , Germany , LH24 . 1 ) . Two-color STED images were recorded on a custom-built STED-microscope ( Göttfert et al . , 2017 ) , which combined two pairs of excitation laser beams of 595 nm and 635 nm with one STED fiber laser beam at 775 nm . All STED images were acquired using Imspector Software ( Max Planck Innovation GmbH , Germany ) . STED images were processed using a linear deconvolution function integrated into Imspector Software ( Max Planck Innovation GmbH , Germany ) . Regularization parameter was 1e−11 . The point spread function ( PSF for deconvolution was generated using a 2D Lorentz function with its half-width and half-length fitted to the half-width and half-length of each individual image . Single AZ images of ‘STED dataset 1’ ( Figure 1E , F , Figure 1—figure supplement 1C , D ) had previously been used for a different type of analysis defining AZ Unc13A cluster numbers; Wild-type in supplementary Figure 2a of Böhme et al . ( 2019 ) . In this study here , we wanted to obtain the average Unc13A distribution from all AZs ( no distinction of AZ types ) . To get an average image of the Unc13A AZ distribution , we used a set of hundreds of 51 × 51 pixel images with a pixel size of 10 × 10 nm . We identified Unc13A clusters in each image using the fluorescence peak detection procedure described in Böhme et al . ( 2019 ) using MATLAB ( version 2016b ) . Peak detection was performed as follows: In each deconvolved 51 × 51 pixel image of an Unc13A-stained AZ , a threshold of 25 gray values was applied below which no pixels were considered . Then , local maxima values were found by finding slope changes corresponding to peaks along pixel columns using the function diff . The same was done along rows for all column positions where peaks were found . The function intersect was then used to determine all pixel positions common in both columns and rows . A minimum distance of 50 nm between neighboring peaks was used to exclude the repeated detection of the same peak , and an edge of 10 nm around the image was excluded to prevent the detection of neighboring AZs . The center of mass of all peak x , y-coordinates found in a single image was then calculated as follows:Px=n-1*∑1nxobsnPy=n-1*∑1nyobs ( n ) Here , n is the number of detected peaks , ( Px , Py ) represents the center of mass ( x , y ) -coordinate , and xobs ( n ) and yobs ( n ) are the coordinates of the n-th detected peak . The image was then shifted such that this position ( Px , Py ) would fall into the center pixel of the 51 × 51 AZ image . For this , we calculated the required shift ( dx and dy ) :dx=imgsizex2-Pxdy=imgsizey2-Py Here , imgsize ( x , y ) refers to the pixel dimensions of the image in both x and y dimensions . The required shift dx , y was then applied to the image using imtranslate , which directly takes these shift values as an input . All shifted images were then averaged into a single compound average image of all AZs by taking the average of each individual pixel and linearly scaling the result in a range between 0 and 255 . This resulted in a circular cloudy structure depicted in Figure 1E , Figure 1—figure supplement 1C . To obtain the distribution of fluorescence as a function of distance to the AZ center in the average picture , we determined the distance between the center of the image and the center of the pixel together with the fluorescence intensity in each pixel . The fluorescence intensity in each pixel was obtained by using the inbuilt MATLAB function ‘imread’ , which outputs the intensities in a matrix with indexes corresponding to the pixel location in the picture . From the indexes ( xp , yp ) of each pixel ( of size 10 nm ) , the distance ( in nm ) to the center was calculated by the following formula:d ( p ) = ( ( xp−26 ) 2+ ( yp−26 ) 2 ) ⋅10 nm We subtracted 26 from the pixel number , since the center pixel is the 26th pixel in x- and y-direction . These distances together with the intensity at each pixel provided the data for the histograms in Figure 1F and Figure 1—figure supplement 1D . The intensity values were normalized to the total amount of intensity making the y-axis of the histogram show percentage of the total amount of intensity . Stage L3 larvae ( n = 17; genotype: w[1118]; P{w[+mC]=Mhc-SynapGCaMP6f}3–5 , Bloomington #67739 ) were fixed in ice-cold Methanol for 7 min and IHC-stained for BRP ( mouse anti-Nc82 , 1:1000; secondary AB: goat anti-mouse Cy5 1:500 ) . Confocal images of the preparations were taken and processed as described in Reddy-Alla et al . ( 2017 ) for a different set of experiments not shown in this paper . Subsequently , the BRP channel was used to identify local fluorescence intensity maxima using the ImageJ-function ‘Find Maxima’ with a threshold setting between 10 and 20 . The locations of maxima for each cell were then loaded into MATLAB ( version 2016b ) and the distances of each x , y-coordinate to all others were determined using the MATLAB function pdist2 , resulting in a square matrix containing all possible inter-AZ distances . Each column of this matrix was then sorted in ascending order , and ( as the distance of one AZ to itself is always 0 ) the mean of the 2nd to 5th smallest values across all AZs was determined and depicted as 1-NND through 4-NND in Figure 3A . The mean distance of the four nearest neighbouring AZs ( 1–4-NND ) was calculated in each AZ ( gray circles in Figure 3A bottom right ) and the mean across AZs was used for quantification of the simulation volume ( see below ) . For both eEJC and mEJC ( spontaneous release events , ”miniature Excitatory Junctional Currents’ ) recordings , two electrode voltage clamp ( TEVC ) recordings were performed from muscle 6 NMJs of abdominal segments A2 and A3 as reported previously ( Qin et al . , 2005 ) . Prior to recordings , the larvae were dissected in haemolymph-like solution without Ca2+ ( HL3 , pH adjusted to 7 . 2 Stewart et al . , 1994: 70 mM NaCl , 5 mM KCl , 20 mM MgCl2 , 10 mM NaHCO3 , 5 mM Trehalose , 115 mM D-Saccharose , 5 mM HEPES ) on Sylgard ( 184 , Dow Corning , Midland , MI , USA ) and transferred into the recording chamber containing 2 ml of HL3 with CaCl2 ( concentrations used in individual experiments described below ) . TEVC recordings were conducted at 21°C using sharp electrodes ( borosilicate glass with filament , 0 . 86×1 . 5×80 nm , Science Products , Hofheim , Germany ) with pipette resistances between 20–30 MΩ , which were pulled with a P-97 micropipette puller ( Sutter Instrument , CA , USA ) and filled with 3 mM KCl . Signals were low-pass filtered at 5 KHz and sampled at 20 KHz . Data was obtained using a Digidata 1440A digitizer ( Molecular devices , Sunnyvale , CA , USA ) , Clampex software ( v10 . 6 ) and an Axoclamp 900A amplifier ( Axon instruments , Union City , CA , USA ) using Axoclamp software . Only cells with a resting membrane potential Vm below −50 mV , membrane resistances Rm above 4 MΩ and an absolute leak currents of less than 10 nA were included in the dataset . Because the presynaptic terminals of the Drosophila larval NMJ are not readily accessible to electrical recordings of Ca2+ currents , the saturation behaviour of Ca2+ influx as a function of extracellular Ca2+ concentrations was measured . We did so by engaging the fluorescent Ca2+ indicator GCaMP6m ( Genotype: w[1118]; P{y[+t7 . 7] w[+mC]=20XUAS-IVS-GCaMP6m}attP40 , Flybase ID: FBti0151346 ) , which we expressed presynaptically using OK6-Gal4 as a motoneuron-specific driver . Third instar larvae heterozygously expressing the indicator were used in experiments as follows . Dissection took place in Ca2+-free , standard hemolymph-like solution HL-3 ( in mM: NaCl 70 , KCl 5 , MgCl2 20 , NaHCO3 10 , Trehalose 5 , Sucrose 115 , HEPES 5 , pH adjusted to 7 . 2 ) ( Stewart et al . , 1994 ) . After dissection on a Sylgard-184 ( Dow-Corning ) block , larvae were transferred to the recording chamber containing HL-3 at varying CaCl2 concentrations ( see below ) . The efferent motoneuron axons were sucked into a polished glass electrode containing a chlorided silver-wire , which could be controlled via a mechanical micromanipulator ( Narishige NMN25 ) and was connected to a pipette holder ( PPH-1P-BNC , NPI electronics ) via a patch electrode holder ( NPI electronics ) , and connected to an S48 stimulator ( Grass Technologies ) . Larvae were then recorded using a white-light source ( Sutter DG-4 , Sutter Instruments ) and a GFP filter set with a Hamamatsu OrcaFlash 4 . 0v2 sCMOS ( Hahamatsu Photonics ) with a framerate of 20 Hz ( 50 ms exposure ) controlled by µManager software ( version 1 . 4 . 20 , https://micro-manager . org ) on an upright microscope ( Olympus BX51WI ) with a 60x water-immersion objective ( Olympus LUMFL 60 × 1 . 10 w ) . Muscle 4 1b NMJs in abdominal segments 2 to 4 were used for imaging . Imaging was conducted over 10 s , and at 5 s , 20 stimuli were applied to the nerve at 20 Hz in 300µs 7V depolarization steps . This procedure was begun in the lowest Ca2+ concentration ( 0 . 75 mM ) and then repeated in the same larva at increasing Ca2+ concentrations ( in mM 1 . 5 , 3 , 6 ) by exchanging the extracellular solution . To achieve a situation with no Ca2+ influx , a final recording was conducted where the bath contained HL-3 without CaCl2 and instead 8 . 3 mM EGTA ( this solution was made by diluting 2 . 5 ml of a 50 mM stock solution in H2O in 12 . 5 ml of HL3 , resulting in a pH of 8 . 0 ) . Because this results in a slight dilution ( 16% ) of the components in the HL3 , the same dilution was performed for the above described Ca2+-containing solutions by adding 2 . 5 ml H2O to 12 . 5 ml of HL3 before CaCl2 was added at above mentioned concentrations . Analysis of 5 Drosophila 3rd instar Larvae was done after automated stabilization of x , y-movement in the recordings ( 8-bit multipage . TIF-stacks , converted from 16 bit ) as described previously ( Reddy-Alla et al . , 2017 ) , manually selecting a ROI around the basal fluorescent GCaMP signal , and reading out the integrated density ( the sum of all pixel grey values ) of the whole region over time . Background fluorescence was measured in a region of the same size and shape outside of the NMJ and subtracted ( frame-wise ) from the signal , separately for each single recording . The quantification was then performed individually for each Ca2+ concentration , by subtracting the fluorescence 250 ms before the stimulation ( Ft=4 . 75s ) from the maximum fluorescence of the trace ( Fmax ) , yielding the change in fluorescence dF:dFCa2+=Fmax-Ft=4 , 75s This was repeated for each cell and a Hill fit was performed on the individual values using Prism ( version 6 . 07 , GraphPad Software Inc ) : ( 3 ) FCa2+ext=Fend*Ca2+extmKM , fluom+Ca2+extm+C In the above equation , Fend is the asymptotic plateau of the fluorescence increase . Furthermore , [Ca2+]ext is the extracellular Ca2+ concentration . KM , fluo ( best fit value: 2 . 679 mM ) is the concentration of extracellular Ca2+ at which fluorescence was half of Fend . The exponent m indicates a cooperative effect of the extracellular Ca2+ concentration on the fluorescence increase , which was constrained to a value of 2 . 43 ( unitless ) based on the described Ca2+ cooperativity of GCaMP6m ( Barnett et al . , 2017 ) . However , constraining this value only had a modest effect on the estimate of KM , fluo as leaving it as a free parameter yielded similar values for KM , fluo ( 3 . 054 mM ) and m ( 1 . 887 ) . The constant C added at the end of Equation 3 allowed the baseline fluorescence to be different from zero . Results and best fit are summarized in ( Figure 3—figure supplement 1 ) . We here prove that stochastic simulations of neurotransmitter release provide a different average PPR value than the PPR value estimated in deterministic simulations . In the following , the stochastic variables A1 and A2 represent the amplitudes of the first and second release , respectively , capital ‘E’ denotes the mean of a stochastic variable ( e . g . EA1 ) , and a1 and a2 represent the amplitudes of the first and second release in the deterministic simulations . In all cases of parameter sets that we tried , the average amplitudes from the stochastic simulations with 1000 repetitions differed < 0 . 5 nA from the deterministically determined amplitudes . Thus , we can assume that EA1 = a1 and likewise for the second release . In deterministic simulations , the estimate of the PPR isPPR-=a2a1=EA2EA1 On the other hand , stochastic simulations yield a sample of different PPR values , since repetitions of the simulation routine yield release varying from trial to trial . In that case , the estimated PPR is ( 4 ) PPR∼=E ( A2A1 ) This resembles the way the PPR is estimated in experiments . Using Jensen’s Inequality and the fact that the function f ( x ) =1/x is strictly convex , we get1EA1<E ( 1A1 ) =E ( A1−1 ) Applying this to ( 4 ) we getPPR∼=E ( A2A1 ) = E ( A1−1A2 ) = Cov ( A1−1 , A2 ) +E ( A1−1 ) E ( A2 ) >Cov ( A1−1 , A2 ) + EA2EA1=Cov ( A1−1 , A2 ) + PPR− Thus , the average stochastically simulated PPR do not necessarily converge to the deterministic estimate with increasing repetitions ( note that in general it is true that the mean of a non-linear function of two random variables is not equal to the non-linear function evaluated in the means ) . An example is shown in Figure 4—figure supplement 1 , where the single-sensor model was simulated with varying amounts of Ca2+ influx ( by varying Qmax ) . The most left blue point , for example , is significantly higher than the deterministic estimate ( p=4e-16 , one-sample t-test ) . This motivates the use of stochastic simulations for correct estimation of the PPR . All MATLAB procedures for simulation of the models can be found in Source code 1 . All simulations ( deterministic and stochastic , see below ) consisted of the same four basic steps , which we describe in detail here . For each new set of parameters , steps 1–4 were repeated . For stochastic simulations , steps 2–4 were repeated 1000 times except for the parameter exploration in Figures 6J–K and 7J–K , where we ran 200 repetitions per parameter set . The many repetitions allowed a good estimate of both mean and variance of the models . In all cases , the mean amplitudes from the stochastic simulations with 1000 repetitions differed < 0 . 5 nA from the deterministically determined amplitudes . Simulation of Ca2+ signals in the presynapse was performed with the program CalC version 6 . 8 . 6 developed and maintained by Victor Matveev ( Matveev et al . , 2002 ) . After this work was initiated , a bug affecting simulations of multiple Ca2+ channels in the same topology was found and a new version of CalC was released . This update had no effect on the simulations used in this study . Intracellular Ca2+ concentrations were simulated in space and time in a cylinder-shaped volume . The cylinder allowed us to assume spatial symmetry which reduced simulation time significantly . Borders of the simulation volume were assumed to be reflective to mimic diffusion of Ca2+ from adjacent AZs ( Meinrenken et al . , 2002 ) and a volume-distributed uptake mechanism was assumed . From measurements of the distance between an AZ and its four nearest neighbors ( Figure 3A ) we estimated the distance between centers of active zones to be 1 . 106 µm , leading to the assumption that the AZ spans a square on the membrane with area of 1 . 223 µm2 . In order for the cylindrical simulation volume to cover an area of the same size , the radius was set to 0 . 624 µm . The height of the simulation volume was set to 1 µm making the simulation volume 1 . 223 µm3 . Increasing the height further had no effect on the Ca2+ transients . The total amount of charge flowing into the cell was assumed to relate to extracellular Ca2+ in a Michaelis-Menten-like way ( as previously described by Schneggenburger et al . , 1999; Trommershäuser et al . , 2003 ) such that ( 5 ) Q= Qmax⋅[Ca]extKM , current+[Ca]ext KM , current was set to the value of 2 . 679 mM as determined for KM , fluo in the GCaMP6m experiments ( see above ) . Qmax was fitted during the optimizations of the models . We simulated a 10 ms paired pulse stimulus initiated after 0 . 5 ms of simulation . The Ca2+ currents for the two stimuli were simulated for 3 ms each and assumed to be Gaussian with FWHM = 360 µs and peak 1 . 5 ms after initiation . That is:ICa= {Q⋅1σ⋅2πe− ( t−2 ) 22σ2 , for t∈[0 . 5 , 3 . 5]Q⋅1σ⋅2πe− ( t−12 ) 22σ2 , for t∈[10 . 5 , 13 . 5]0 , elsewith σ= 0 . 36022⋅ln ( 2 ) = 0 . 153 . The CalC simulation output were data files that contained the spatio-temporal intracellular Ca2+ profile at the height of 10 nm from the plasma membrane . In exocytosis simulations , these concentrations were interpolated at the SV distances in the x , y-plane and at time points with MATLAB’s built-in interpolate functions when computing the reaction rates of the system at a given time point . The resting Ca2+ concentration was assumed to relate to the extracellular Ca2+ concentration in a similar way as during stimulation , such that ( 6 ) [Ca2+]basal=[Ca2+]max⋅[Ca2+]extKM , current+[Ca2+]extwith Ca2+max=190 nM For designation and value of Ca2+ parameters , see Table 1 . In all simulations we had to determine where to place release site . This was done by using the cdf of the SV distance distribution derived above ( Equation 2 ) . For deterministic simulations , which were used in the fitting routine of the models ( see below ) , the unit interval was divided into 180 bins of the formk-1180 , k180 , k=1 , 2…180 . The midpoints were the percentiles giving rise to distances at which we read the Ca2+ simulation . This approach provided an approximation of the SV distribution . In accordance with our assumption that the AZs work in parallel the 180 distances gave rise to 180 independent different systems of ODEs with 1/180 of the total amount of SVs in each system . The results were then added together as a good approximation of the mean of the stochastic simulations with random SV distance drawings . In each run of the stochastic simulations , we drew n random numbers from the unit interval , n being the number of SVs , and computed the distances based on the formula derived above . The models are summarized in Figures 4A , 6A and 7A , and Figure 7—figure supplement 3A , B . In the following equations the single-sensor , dual fusion-sensor , and unpriming models are all described . The site activation model is a combination of the equations for the single-sensor model and the site activation equations described below . The red text denotes terms that are unique to the dual fusion-sensor model , blue text indicates unpriming , which is unique to the unpriming model . Parameters are described below . For designation and value of parameters , see Tables 2 , 3 . Rate equations of the single-sensor model , dual fusion-sensor model and unpriming model:d[R ( 0 , 0 ) ]dt=krep[P0]− ( r⋅u+ 5[Ca2+]k1+2[Ca2+]k2+L+ ) [R ( 0 , 0 ) ]+k−1[R ( 1 , 0 ) ]+ k−2[R ( 0 , 1 ) ]d[R ( 1 , 0 ) ]dt=− ( 4[Ca2+]k1+k−1+2[Ca2+]k2+L+f ) [R ( 1 , 0 ) ]+5[Ca2+]k1[R ( 0 , 0 ) ]+ 2bfk−1[R ( 2 , 0 ) ]+ k−2[R ( 1 , 1 ) ]d[R ( 2 , 0 ) ]dt=− ( 3[Ca2+]k1+2bfk−1+2[Ca2+]k2+L+f2 ) [R ( 2 , 0 ) ]+4[Ca2+]k1[R ( 1 , 0 ) ]+ 3bf2k−1⋅[R ( 3 , 0 ) ]+k−2⋅[R ( 2 , 1 ) ]d[R ( 3 , 0 ) ]dt=− ( 2[Ca2+]k1+3bf2 k−1+2[Ca2+]k2+L+f3 ) [R ( 3 , 0 ) ]+3[Ca2+]k1[R ( 2 , 0 ) ]+ 4bf3k−1⋅[R ( 4 , 0 ) ]+k−2⋅[R ( 3 , 1 ) ]dR[ ( 4 , 0 ) ]dt=− ( [Ca2+]k1+4bf3k−1+2[Ca2+]k2+L+f4 ) [R ( 4 , 0 ) ]+2[Ca2+]k1[R ( 3 , 0 ) ]+ 5bf4k−1⋅[R ( 5 , 0 ) ]+k−2⋅[R ( 4 , 1 ) ]d[R ( 5 , 0 ) ]dt=− ( 2[Ca2+]k2+5bf4k−1+L+f5 ) [R ( 5 , 0 ) ]+[Ca2+]k1[R ( 4 , 0 ) ]+k−2⋅[R ( 4 , 1 ) ]d[R ( 0 , 1 ) ]dt=krep[P1]− ( 5[Ca2+]k1+[Ca2+]k2+k−2+L+s ) [R ( 0 , 1 ) ]+k−1⋅[R ( 1 , 1 ) ]+2[Ca2+]k2[R ( 0 , 0 ) ]+2bsk−2⋅[R ( 0 , 2 ) ]d[R ( 1 , 1 ) ]dt=− ( 4[Ca2+]k1+k−1+[Ca2+]k2+k−2+L+fs ) [R ( 1 , 1 ) ]+5[Ca2+]k1[R ( 0 , 1 ) ]+2bfk−1⋅[R ( 2 , 0 ) ]+ 2[Ca2+]k2[R ( 1 , 0 ) ]+ 2bsk−2⋅[R ( 1 , 2 ) ]d[R ( 2 , 1 ) ]dt=− ( 3[Ca2+]k1+2bfk−1+[Ca2+]k2+k−2+L+f2s ) [R ( 2 , 1 ) ]+4[Ca2+]k1[R ( 1 , 1 ) ]+3⋅bf2⋅k−1[R ( 3 , 1 ) ]+ 2[Ca2+]k2[R ( 2 , 0 ) ]+ 2bsk−2⋅[R ( 2 , 2 ) ]d[R ( 3 , 1 ) ]dt=− ( 2[Ca2+]k1+3bf2 k−1+[Ca2+]k2+k−2+L+f3s ) [R ( 3 , 1 ) ]+3[Ca2+]k1[R ( 2 , 1 ) ]+ 4bf3k−1⋅[R ( 4 , 1 ) ]+ 2[Ca2+]k2[R ( 3 , 0 ) ]+ 2bsk−2⋅[R ( 3 , 2 ) ]d[R ( 4 , 1 ) ]dt=− ( [Ca2+]k1+4bf3k−1+[Ca2+]k2+k−2+L+f4s ) [R ( 4 , 1 ) ]+2[Ca2+]k1[R ( 3 , 1 ) ]+5bf3k−1⋅[R ( 5 , 1 ) ]+ 2[Ca2+]k2[R ( 4 , 0 ) ]+ 2bsk−2[R ( 4 , 2 ) ]d[R ( 5 , 1 ) ]dt=− ( 5bf4k−1+[Ca2+]k2+k−2+L+f5s ) [R ( 5 , 1 ) ]+[Ca2+]k1[R ( 4 , 1 ) ]+ 2[Ca2+]k2[R ( 5 , 0 ) ]+ 2bsk−2⋅[R ( 5 , 2 ) ]d[R ( 0 , 2 ) ]dt=krep[P2]− ( 5[Ca2+]k1+2bsk−2+L+s2 ) [R ( 0 , 2 ) ]+k−1[R ( 1 , 2 ) ]+[Ca2+]k2[R ( 0 , 1 ) ]d[R ( 1 , 2 ) ]dt=− ( 4[Ca2+]k1+k−1+2bsk−2+L+fs2 ) [R ( 1 , 2 ) ]+5[Ca2+]k1[R ( 0 , 2 ) ]+ 2bfk−1[R ( 2 , 2 ) ]+ [Ca2+]k2[R ( 1 , 1 ) ]d[R ( 2 , 2 ) ]dt=− ( 3[Ca2+]k1+2bfk−1+2bsk−2+L+f2s2 ) [R ( 2 , 0 ) ]+4[Ca2+]k1[R ( 1 , 2 ) ]+ 3bf2k−1[R ( 3 , 0 ) ]+ [Ca2+]k2[R ( 2 , 1 ) ]d[R ( 3 , 2 ) ]dt=− ( 2[Ca2+]k1+3bf2 k−1+2bsk−2+[Ca2+]k2+L+f3s2 ) [R ( 3 , 2 ) ]+3[Ca2+]k1[R ( 2 , 2 ) ]+ 4bf3k−1⋅[R ( 4 , 2 ) ]+ [Ca2+]k2[R ( 3 , 1 ) ]d[R ( 4 , 2 ) ]dt=− ( [Ca2+]k1+4bf3k−1+2bsk−2+[Ca2+]k2+L+f4s2 ) [R ( 4 , 2 ) ]+2[Ca2+]k1[R ( 3 , 2 ) ]+ 5bf3k−1⋅[R ( 5 , 2 ) ]+ [Ca2+]k2[R ( 4 , 1 ) ]d[R ( 5 , 2 ) ]dt=− ( 5bf4k−1+2bsk−2+L+f5s2 ) [R ( 5 , 2 ) ]+[Ca2+]k1[R ( 4 , 2 ) ]+ [Ca2+]k2[R ( 5 , 1 ) ]d[F]dt=L+ ( [R ( 0 , 0 ) ]+f[R ( 1 , 0 ) ]+f2[R ( 2 , 0 ) ]+f3[R ( 3 , 0 ) ]+f4[R ( 4 , 0 ) ]+f5[R ( 5 , 0 ) ]+[sR ( 0 , 1 ) ]+fs[R ( 1 , 1 ) ]+f2s[R ( 2 , 1 ) ]+f3s[R ( 3 , 1 ) ]+f4s[R ( 4 , 1 ) ]+f5s[R ( 5 , 1 ) ]+[s2R ( 0 , 2 ) ]+fs2[R ( 1 , 1 ) ]+f2s2[R ( 2 , 1 ) ]+f3s2[R ( 3 , 1 ) ]+f4s2[R ( 4 , 1 ) ]+f5s2[R ( 5 , 1 ) ] ) d[P0]dt=L+ ( [R ( 0 , 0 ) ]+ f[R ( 1 , 0 ) ]+f2[R ( 2 , 0 ) ]+f3[R ( 3 , 0 ) ]+f4[R ( 4 , 0 ) ]+f5[R ( 5 , 0 ) ] ) +k−2[P1]−2k2[Ca2+][P0]−krep[R ( 0 , 0 ) ]+r⋅u[R ( 0 , 0 ) ]d[P1]dt=L+ ( [sR ( 0 , 1 ) ]+fs[R ( 1 , 1 ) ]+f2s[R ( 2 , 1 ) ]+f3s[R ( 3 , 1 ) ]+f4s[R ( 4 , 1 ) ]+f5s[R ( 5 , 1 ) ] ) −k−2[P1]+2k2[Ca2+][P0]+2bsk−2⋅[P2]−krep[R ( 0 , 1 ) ]d[P2]dt=L+ ( [R ( 0 , 0 ) ]+f[R ( 1 , 0 ) ]+f2[R ( 2 , 0 ) ]+f3[R ( 3 , 0 ) ]+f4[R ( 4 , 0 ) ]+f5[R ( 5 , 0 ) ]+[sR ( 0 , 1 ) ]+ fs[R ( 1 , 1 ) ]+f2s[R ( 2 , 1 ) ]+f3s[R ( 3 , 1 ) ]+f4s[R ( 4 , 1 ) ]+f5s[R ( 5 , 1 ) ]+[s2R ( 0 , 2 ) ]+ fs2[R ( 1 , 1 ) ]+f2s2[R ( 2 , 1 ) ]+f3s2[R ( 3 , 1 ) ]+f4s2[R ( 4 , 1 ) ]+f5s2[R ( 5 , 1 ) ] ) +2k2[P1]−2bsk−2[Ca2+][P2]−krep[R ( 0 , 2 ) ]r=1−[Ca2+]n[Ca2+]n+KM , unprimn In the single-sensor and site activation models , k2 = k-2=u = 0 , and s = 1 . This excludes all reactions exclusive for the dual fusion-sensor and unpriming models . Similarly , u = 0 in the dual fusion-sensor model and k2 = k-2=0 and s = 1 in the unpriming model . [R ( n , m ) ] denotes the Ca2+ binding state of a SV with n Ca2+ ions bound to the first sensor and m Ca2+ ions bound to the second fusion sensor . Note that in the single-sensor , site activation and unpriming models , m is always zero ( since there is no second fusion sensor ) , and the states are denoted with a single number in Figures 4A and 6A and Figure 7—figure supplement 3 . [F] counts the cumulative number of fused SVs . [P0] is not shown in the figures , but are part of the equations denoting the number of empty sites . That is , in the single-sensor and unpriming models r=1-Ca2+nCa2+n+KM , unprimn has a positive part equal to dP0dt and a negative part equal to the rate of replenishment . In the dual fusion-sensor model , there are three states of empty sites , [P0] , [P1] , [P2] . These corresponded to the different states of Ca2+ binding to the second fusion sensor of the empty sites since we assumed the second sensor to be located on the plasma membrane . Note that these equations describe the second sensor with cooperativity 2 , which is described in Results . We also optimized cooperativities 3 , 4 , and 5 . The equations can easily be extended to these cases , since the rate equations of the second fusion sensor are of the same form as for the first sensor . In the unpriming model ( Figure 7A ) we assumed unpriming to take place from state [R ( 0 ) ] with a Ca2+-dependent rate . For the individual reactions , we can express the rates of Ca2+ ( un ) binding , fusion , and replenishment of a single SV in a more general form . This is useful in the stochastic simulation method introduced later . In the following , we denote the general form of the rate for each possible reaction in the models described above . The expressions in brackets denote the states involved in the reaction . ( 7 ) [R ( n , m ) ]→[R ( n−1 , m ) ]:nk−1bn−1[R ( n , m ) ]→[R ( n+1 , m ) ]: ( nmax−n ) [Ca2+]k1[R ( n , m ) ]→[R ( n , m−1 ) ]:mk−2bm−1[R ( n , m ) ]→[R ( n , m+1 ) ]: ( mmax−m ) [Ca2+]k2[R ( n , m ) ]→[F]: L+smfn[P0]→[R ( 0 , 0 ) ]: krepwith nmax and mmax denoting the cooperativity of the first and second fusion sensors , respectively . Equations in line 3 and 4 in ( 7 ) were only non-zero in the dual fusion-sensor model . In the site activation model ( Figure 7—figure supplement 3 ) , all reactions regarding Ca2+ ( un ) binding and replenishment was as in the one-sensor model . In addition we assumed a mechanism acting on the release sites independently of the Ca2+ binding of the SV . All sites regardless of the SV status were either activated ( A state ) or not ( D or I states ) . This mechanism is proposed as a facilitation mechanism , which necessitates its primary effect to be on the second stimulus rather than the first . We were therefore forced to implement the D state , which is a temporary ‘delay’ state making sure the mechanism does not increase first release . The changing of [A] and [I] states at 0 . 75 and 10 mM extracellular Ca2+ are shown in ( Figure 7—figure supplement 3I ) . The site activation mechanism has the following rate equations:d[A]dt=−δ[A]+γ[D]d[D]dt=− ( β+γ ) [D]+α[Ca2+]n[I]+δ[A]d[I]dt=−α[Ca2+]n[I]+β[D]where α , β , δ , γ>0 are rate parameters . The deterministic implementation of the site activation model included 3 sets of ODEs , one for each state in the site activation model . Each set consisted of the equations of the one-sensor model as well as transitions between states of equal Ca2+ binding in the 3 sets of ODEs ( e . g . from R ( 0 , D ) to R ( 0 , A ) ) ( Figure 7—figure supplement 3B ) . In the stochastic simulations the site activation rates were included in the propensity vector like any other reaction . Whenever a site activation reaction occurred , a release site vector consisting of nsites elements was updated . For each site , the fusion rate was multiplied by 0 , when the site state was I or D . Prior to simulation , the Ca2+ binding states of all SVs were assumed to be in equilibrium . We can determine the steady state iteratively by settingd[I]dt=-αCa2+nI+β[D]α , β , δ , γ>0R0 , 0init=1 This can be reduced to the non-iterative expression:[R ( n , m ) ]init= ( ∏i=1n ( nmax+1−i ) ) ⋅[Ca2+]n⋅k1nn ! ⋅b∑j=1n ( j−1 ) ⋅k−1n⋅ ( ∏i=1m ( mmax+1−i ) ) ⋅[Ca2+]m⋅k2mm ! ⋅b∑j=1m ( j−1 ) ⋅k−2m= ( nmax ! ( nmax−n ) ! ⋅[Ca2+]nk1nn ! ⋅bn ( n−1 ) 2⋅k−1n ) ⋅ ( mmax ! ( mmax−m ) ! ⋅[Ca2+]mk2mm ! ⋅bm ( m−1 ) 2⋅k−2m ) Note that for n = 0 , the first parenthesis is 1 , while m = 0 implies that the second parenthesis is 1 , making this solution valid also in the absence of a second fusion-sensor . We ignored the very small fusion rate . In the steady-state of the unpriming model , the number of SVs in [R ( 0 , 0 ) ] must furthermore be in equilibrium with the number of empty states:Rn , minit= ( ∏i=1n ( nmax+1-i ) ) ⋅[Ca2+]n⋅k1nn ! ⋅b∑j=1nj-1⋅k-1n⋅ ( ∏i=1m ( mmax+1-i ) ) ⋅[Ca2+]m⋅k2mm ! ⋅b∑j=1mj-1⋅k-2m After finding this steady-state , the solution is scaled to match the desired number of SVs by multiplying all states with a constant , such that the sum of all [R ( n , m ) ] and [P] equals the number of SVs . The steady-state of the site activation was determined before simulation by calculating the fraction of states being in [A] , [D] , or [I] . This was done by calculating=nmax ! ( nmax-n ) ! ⋅Ca2+nk1nn ! ⋅bnn-12⋅k-1n ⋅mmax ! ( mmax-m ) ! ⋅Ca2+mk2mm ! ⋅bmm-12⋅k-2mand normalizing to sum to 1 . This determined the steady state fraction of activation of sites . In the stochastic simulations , the SVs were randomly assigned initial states according to the probabilities of the different states in the steady-state . All deterministic exocytosis simulations of the above equations were carried out with the inbuilt MATLAB ODE solver ode15s . All stochastic exocytosis simulations as well as simulation data handling were carried out in MATLAB with custom-written scripts ( included in Source code 1 ) . For the simulation itself we used a modified version of the Gillespie Algorithm ( Gillespie , 2007 ) , which included a minimal time step since reaction rates change quickly with the changing intracellular Ca2+ concentration . The minimal step was µ = 1e-6 s . In the algorithm , the time from the current simulation time point , t , until the next reaction , τ , is determined , the reaction is carried out and the new simulation time point is set to t+τ . Whenever the simulation yielded τ>µ , the simulation time point was set to t+µ , no reaction was carried out and the propensities of the model were updated at the new time point . This is a valid method of obtaining a better estimate because the waiting time until next reaction is exponentially distributed . The implementation of the algorithm takes advantage of the general form of the rate equations in ( 7 ) . Instead of calculating matrices of states and reaction rates , we have a vector , V , of length nsites , where each element represents the status of one SV/site . The SV state of a docked SV on the kth site in state [R ( n , m ) ] is denoted by the two-digit numberP=r⋅ukrep+r⋅u⋅[R0 , 0] If the site was empty ( due to initial submaximal priming or SV fusion ) we assigned Vk=100 . Using Equation 7 , the rates of any primed SV arerk= ( m⋅k−2⋅bm−1n⋅k−1⋅bn−1L+fnsm ( nmax−n ) ⋅[Ca2+]⋅k1 ( mmax−m ) ⋅[Ca2+]⋅k2r⋅b ) The sum of these rates of all SVs yield the summed propensities of the system , a0 , which is the basis of the calculation of τ , whereas the cumulative sum is used for determination of which SV undergoes a reaction ( Gillespie , 2007 ) . When a SV undergoes a reaction , we find the index of the reaction occurring , j , by using the cumulative sum of rk in the same way as in the standard implementation of the Gillespie Algorithm ( Gillespie , 2007 ) . Putting j^=j−3 allows us to easily update the status of the SV , sinceVk=Vk+1 ( j^≠3 ) ⋅sign ( j^ ) ⋅10|j^|−1+1 ( j^=0 ) ∨ ( j^=3 ) ⋅ ( 100−Vk ) In parallel with this a vector of fusions is updated , such that at every time point , the next element in the fusion vector is set to 1 if a fusion took place , and 0 else . Many repetitions of time consuming stochastic simulations had to be performed , and many sets of ODEs were solved for each choice of parameters . Therefore , simulations were carried out on the computer grid on The Bioinformatics Center , University of Copenhagen . This allowed running repetitions in parallel with MATLAB’s Parallel Computing toolbox using between 5 and 100 cores depending on the simulation job . In order to calculate the eEJC , we needed a vector of the SV fusions at different time points . Both deterministic and stochastic simulations yielded the vectors time_outcome and fuse_outcome , which is a pair of vectors of the same length but with changing time steps . For the sampling we generated a time vector , time_sample , with a fixed time step of 1 µs . From here , the determining of the SV fusion times differ between deterministic and stochastic simulations . In the deterministic simulations , we simulated a sample of distances , bins , as described earlier . Each bin gave rise to a set of ODEs , which could be simulated independently , and the fuse_outcome is continuously changing based on the rates . In MATLAB the interpolation for bin k was done as follows:fuse_interpk=interp1 ( time_outcome , fuse_outcome , time_sample ) fuse_interpk contained the cumulative fused SVs over time in a single bin sampled at the time points of the vector time_sample . These were summed to find the total number of fused SVs:Vk=Vk+1 ( j^≠3 ) ⋅signj^⋅10j^-1+1j^=0∨ ( j^=3 ) ⋅100-Vk Therefore the SVs fused per time step were be the difference between neighboring values in the fuse_interp vector:fusion_vec=[0 , diff ( fuse_interp ) ] This vector was the basis for the computation of the eEJC . In the stochastic simulations , the fuse_outcome vector contains discrete SV fusions at certain time points . We therefore sample the SV fusions by assigning them to the nearest time points on the time_sample vector . That is , each fusion time was rounded to the nearest microsecond , thereby giving rise to the fusion_vec , which in the stochastic case contained whole numbers of SV fusions at different time points . In both deterministic and stochastic simulations the mEJC was generated as a vector , mEJC_vec , with the same time step as the time_sample and fuse_vec . This allows us to calculate the eEJC with MATLAB’s convolve function , conv , such thateEJC=conv ( fuse_vec , mEJC_vec ) where fusion_vec is a vector with the same time step , each element being the number of SV fusions at each time point . The eEJC1 amplitude was determined as the minimum current of the eEJC within the time interval ( 0 , 10 ) ms . Similar to the analysis of experimental eEJC data , we fitted an exponential function to the decay for estimation of the base value for the second response ( see Figure 2—figure supplement 1A ) . The eEJC2 amplitude was the difference between the second local minimum and the fitted exponential function extrapolated to the time point of the second local minimum ( as described for the analysis of electrophysiology experiments ) . Because deterministic simulations cannot predict PPR values ( due to Jensen’s inequality , see above ) , but stochastic simulations cannot be fitted to data , we first ran deterministic simulations comparing the simulated first and second absolute eEJC amplitudes to the experimental amplitudes ( not the PPR , see Materials and methods ) . Afterwards we ran stochastic simulations with the optimised parameters in order to compare PPRs and variances to experimental results . To determine the optimal parameters for the deterministic simulations at the five experimental extracellular Ca2+ concentrations , the models were fitted to the two peak amplitudes , eEJC1 and eEJC2 , by minimizing the following cost value:fuse_interp = ∑k=1nbinsfuse_interpkwhere we sum over the five different experimental Ca2+ concentrations . Note that in deterministic simulations , eEJC1 and eEJC2 amplitudes are precise estimates of average amplitudes in stochastic simulations allowing us to do deterministic optimizations . When fitting the models , we used the inbuilt MATLAB function fminsearch , which uses the Nelder-Mead Simplex Search , to minimize the above cost function . The cost calculation in each iteration was a two-step process taking advantage of the fact that the total number of SVs scales the eEJC1 and eEJC2 values in the deterministic simulations . For each choice of parameters the simulation was run with 180 sites ( the initial number of sites is arbitrary , but matched the number of distance bins ) , and the optimal number of sites were determined afterwards . Thus , a given set of parameters gave rise to amplitudes eEJC1 , init and eEJC2 , init from simulations with 180 sites . After that we determined eEJC=conv ( fuse_vec , mEPSC_vec ) such that cost ( eEJC1 , sim , eEJC2 , sim ) =∑k=15eEJC1 , sim , k-eEJC1 , exp , k2eEJC1 , exp , k+eEJC2 , sim , k-eEJC2 , exp , k2eEJC2 , exp , k was minimized . The number of sites in the given iteration was therefore 180⋅csites and the cost of that particular iteration wascsites∈R+ In this way the optimization algorithm did not have to include nsites in the parameter search algorithm , which reduced the number of iterations significantly . In the stochastic simulations , the number of SVs was set to 180⋅csites rounded to nearest integer . | Cells in the nervous system of all animals communicate by releasing and sensing chemicals at contact points named synapses . The ‘talking’ ( or pre-synaptic ) cell stores the chemicals close to the synapse , in small spheres called vesicles . When the cell is activated , calcium ions flow in and interact with the release-ready vesicles , which then spill the chemicals into the synapse . In turn , the ‘listening’ ( or post-synaptic ) cell can detect the chemicals and react accordingly . When the pre-synaptic cell is activated many times in a short period , it can release a greater quantity of chemicals , allowing a bigger reaction in the post-synaptic cell . This phenomenon is known as facilitation , but it is still unclear how exactly it can take place . This is especially the case when many of the vesicles are not ready to respond , for example when they are too far from where calcium flows into the cell . Computer simulations have been created to model facilitation but they have assumed that all vesicles are placed at the same distance to the calcium entry point: Kobbersmed et al . now provide evidence that this assumption is incorrect . Two high-resolution imaging techniques were used to measure the actual distances between the vesicles and the calcium source in the pre-synaptic cells of fruit flies: this showed that these distances are quite variable – some vesicles sit much closer to the source than others . This information was then used to create a new computer model to simulate facilitation . The results from this computing work led Kobbersmed et al . to suggest that facilitation may take place because a calcium-based mechanism in the cell increases the number of vesicles ready to release their chemicals . This new model may help researchers to better understand how the cells in the nervous system work . Ultimately , this can guide experiments to investigate what happens when information processing at synapses breaks down , for example in diseases such as epilepsy . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Rapid regulation of vesicle priming explains synaptic facilitation despite heterogeneous vesicle:Ca2+ channel distances |
The recognition of pathogens and subsequent activation of defense responses are critical for the survival of organisms . The nematode Caenorhabditis elegans recognizes pathogenic bacteria and elicits defense responses by activating immune pathways and pathogen avoidance . Here we show that chemosensation of phenazines produced by pathogenic Pseudomonas aeruginosa , which leads to rapid activation of DAF-7/TGF-β in ASJ neurons , is insufficient for the elicitation of pathogen avoidance behavior . Instead , intestinal infection and bloating of the lumen , which depend on the virulence of P . aeruginosa , regulates both pathogen avoidance and aversive learning by modulating not only the DAF-7/TGF-β pathway but also the G-protein coupled receptor NPR-1 pathway , which also controls aerotaxis behavior . Modulation of these neuroendocrine pathways by intestinal infection serves as a systemic feedback that enables animals to avoid virulent bacteria . These results reveal how feedback from the intestine during infection can modulate the behavior , learning , and microbial perception of the host .
In nature , all organisms are continuously exposed to a complex environment and are under constant threat of attack by pathogenic microbes . To ensure their survival , organisms must recognize pathogens and mount a robust defense in response to their attack . Physical avoidance of pathogens is also a critical defense strategy to reduce pathogen infections ( Kavaliers et al . , 2019; Medzhitov et al . , 2012 ) . Chemosensation of bacterial metabolites and toxins appears to play an important role in eliciting pathogen avoidance behaviors . In mammals , olfactory chemosensory neurons and nociceptor sensory neurons detect bacterial toxins , quorum-sensing molecules , formyl peptides , and lipopolysaccharides through distinct molecular mechanisms that lead to rapid avoidance behaviors ( Boillat et al . , 2015; Chiu et al . , 2013; Meseguer et al . , 2014; Rivière et al . , 2009; Tizzano et al . , 2010; Yang and Chiu , 2017 ) . Similarly , in the fruit fly Drosophila melanogaster , olfactory and gustatory neurons have been reported to detect geosmin ( the smell of mold ) , phenol , and lipopolysaccharides via distinct molecular mechanisms , allowing the organism to avoid food contaminated with bacteria ( Mansourian et al . , 2016; Soldano et al . , 2016; Stensmyr et al . , 2012 ) . A deeper understanding of the various mechanisms of pathogen avoidance has the potential to uncover conserved host defense responses that are important against pathogen infections . The free-living nematode Caenorhabditis elegans feeds on bacteria . Pathogenic bacteria such as Pseudomonas aeruginosa infect and kill C . elegans , and upon exposure to P . aeruginosa , C . elegans elicits an innate immune response that results in the activation of microbial-killing pathways and a pathogen-avoidance behavior that improve its survival ( Martin et al . , 2017; Meisel and Kim , 2014; Styer et al . , 2008; Troemel et al . , 2006 ) . Chemosensation of the P . aeruginosa metabolites phenazine-1-carboxamide and pyochelin has been proposed to be responsible for eliciting the avoidance behavior ( Meisel et al . , 2014 ) . These metabolites activate a G-protein signaling pathway in the ASJ chemosensory neuron pair that induces expression of the neuromodulator DAF-7/TGF-β within minutes of exposure to the pathogen . However , the animals are initially attracted towards lawns of P . aeruginosa and only begin to avoid the pathogen after exposure for hours ( Meisel et al . , 2014; Singh and Aballay , 2019a; Sun et al . , 2011 ) . The initial attraction of C . elegans towards P . aeruginosa is mediated by sensing of quorum-sensing molecules acylated homoserine lactones ( Beale et al . , 2006; Ha et al . , 2010 ) . In addition , in a two-choice preference assay between Escherichia coli and P . aeruginosa , the naïve animals show a higher preference for P . aeruginosa ( Ha et al . , 2010; Zhang et al . , 2005 ) . The animals only change their choice and move towards E . coli lawns after an exposure time to the two bacteria exceeding 4 hr ( Zhang et al . , 2005 ) . Thus , the activation of DAF-7/TGF-β signaling in ASJ neurons , which takes place after minutes of exposure to P . aeruginosa , does not fully explain the later aversive learning towards P . aeruginosa . Here we show that chemosensation of P . aeruginosa metabolites , which leads to the induction of DAF-7/TGF-β in ASJ neurons , does not correlate with the avoidance behavior . Instead , bloating of the intestinal lumen caused by the pathogen infection underlies the avoidance behavior via modulation of both DAF-7/TGF-β and the G-protein coupled receptor NPR-1 neuroendocrine pathways , which regulate aerotaxis . We further show that the modulation of these neuroendocrine pathways by intestinal infection or genetic modulation of aerotaxis by loss of ocr-2 and osm-9 drives the change in preference of animals from P . aeruginosa to E . coli lawns . Our findings demonstrate that signaling from the intestine via neuroendocrine pathways modulates microbial perception during infection .
Because P . aeruginosa-produced phenazines lead to induction of the neuromodulator DAF-7/TGF-β in the ASJ neuron pair ( Meisel et al . , 2014 ) , we studied the role of phenazines in the elicitation of the pathogen avoidance behavior . P . aeruginosa uses a well-characterized biosynthetic pathway to generate four different phenazines ( Dietrich et al . , 2006 ) ( Figure 1A ) . Phenazine-1-carboxylic acid , the precursor of all other phenazines produced by P . aeruginosa , is synthesized from chorismate by a full set of functional phenazine-1-carboxylic acid biosynthetic enzymes encoded by the phzA1-G1 and phzA2-G2 operons ( Dietrich et al . , 2006 ) . Phenazine-1-carboxylic acid can be modified by other enzymes to make 1-hydroxyphenazine , phenazine-1-carboxamide , or pyocyanin ( Figure 1A ) . Phenazine-1-carboxamide , but not other phenazines , activates DAF-7/TGF-β expression in ASJ neurons ( Meisel et al . , 2014 ) . We tested whether any of the four purified phenazines produced by P . aeruginosa elicited avoidance when added to lawns of E . coli . We found that animals exposed to E . coli lawns containing phenazine-1-carboxamide , which is required for the activation of DAF-7/TGF-β expression in ASJ neurons ( Meisel et al . , 2014 ) , did not exhibit the avoidance behavior ( Figure 1B ) . Additionally , the animals did not avoid E . coli lawns with added phenazine-1-carboxylic acid and pyocyanin ( Figure 1B ) . In contrast , the animals did avoid lawns of E . coli containing 1-hydroxyphenazine ( Figure 1C ) . We observed that animals feeding on E . coli lawns that were supplemented with 1-hydroxyphenazine , but not other phenazines , showed bloating of the intestinal lumen ( Figure 1D , E ) . Bloating of the intestine is known to elicit microbial avoidance behavior , including the avoidance of E . coli ( Kumar et al . , 2019; Singh and Aballay , 2019a ) . Therefore , the avoidance of E . coli lawns containing 1-hydroxyphenazine was likely caused by intestinal bloating of the animals induced by the toxin ( Figure 1D , E ) . We also observed that animals feeding on E . coli lawns supplemented with 1-hydroxyphenazine , but not other phenazines , showed drastically reduced survival ( Figure 1F ) . More than 50% of the animals feeding on E . coli lawns containing 1-hydroxyphenazine died within 24 hr , while animals feeding on control E . coli lawns or E . coli lawns containing other phenazines remained alive at the same time point ( Figure 1F ) . However , P . aeruginosa mutants deficient in phenazine production were not compromised in their ability to kill C . elegans compared with wild-type P . aeruginosa ( Figure 1G ) , suggesting that the addition of 1-hydroxyphenazine may have non-physiological effects . Moreover , all the phenazine mutants induced an avoidance behavior that was indistinguishable from that caused by the wild-type P . aeruginosa ( Figure 1H ) . These results suggest that while high amounts of externally added 1-hydroxyphenazine elicit microbial avoidance behavior , the normal amount of phenazines produced by P . aeruginosa during infection is insufficient for induction of the avoidance behavior . C . elegans is initially attracted towards lawns of P . aeruginosa , and after an initial phase of interaction , the animals begin to avoid the bacterial lawns . However , the initial phase of interaction before elicitation of the avoidance behavior is variable , leading to different avoidance kinetics in various studies ( Hao et al . , 2018; Hilbert and Kim , 2017; Ma et al . , 2017; Martin et al . , 2017; Meisel et al . , 2014; Singh and Aballay , 2019a; Sun et al . , 2011 ) . It is likely that the variations in avoidance kinetics are due to differences in the production of P . aeruginosa factors governing the avoidance behavior . We reasoned that different growing conditions for the P . aeruginosa lawns might be responsible for the differences in avoidance behaviors . We observed that one of the major differences in bacterial lawn preparations is the variation in incubation periods of P . aeruginosa on agar plates before animal exposure ( Hao et al . , 2018; Hilbert and Kim , 2017; Ma et al . , 2017; Meisel et al . , 2014; Singh and Aballay , 2019a; Sun et al . , 2011 ) . To examine whether the differences in culture conditions of P . aeruginosa on agar plates before transferring C . elegans could be the underlying reason for the different avoidance kinetics , we varied the P . aeruginosa incubation times ( Figure 2A ) . The avoidance behavior of the animals was enhanced with the incubation period of the P . aeruginosa lawns ( Figure 2B ) . Animals exposed to P . aeruginosa lawns that were incubated at 37°C for 24 hr followed by 25°C for 48 hr ( referred to as 72 hr lawn ) showed a significantly enhanced avoidance rate in comparison to animals exposed to P . aeruginosa lawns that were incubated at 37°C for 12 hr ( referred to as 12 hr lawn ) ( Figure 2C ) . We tested whether the differences in avoidance kinetics exhibited by the animals on 12 and 72 hr lawns could be due to differences in the induction of daf-7 expression in ASJ neurons . We found that while animals exposed to 12 hr lawns showed robust induction of daf-7 expression in ASJ neurons that did not differ from animals exposed to 72 hr lawns ( Figure 2D , E ) , animals exposed to 12 hr lawns exhibited a delayed avoidance compared with animals exposed to 72 hr lawns ( Figure 2C ) . In addition , the induction of daf-7 expression in ASI neurons was not significantly different in animals exposed to 12 or 72 hr lawns ( Figure 2—figure supplement 1 ) . Because the induction of daf-7 expression in ASJ neurons was indistinguishable in animals on 12 and 72 hr lawns , the results suggest that the induction of daf-7 expression in ASJ may not be the only cause of the elicitation of avoidance behavior . Because 1-hydroxyphenazine induces both avoidance behavior and intestinal bloating ( Figure 1C , D ) , we tested whether bloating could account for the faster avoidance exhibited by animals exposed to the 72 hr lawn compared with those exposed to the 12 hr lawn . We found that animals exposed to 72 hr lawns showed bloated intestines as early as 8 hr , while the lumens of animals exposed to 12 hr lawns for 8 hr were comparable to those of animals fed E . coli ( Figure 3A–C ) . Because bloating of the intestine leads to the induction of genes that are part of the NPR-1 neuroendocrine pathway ( Singh and Aballay , 2019a ) , we examined the expression levels of the npr-1 , flp-18 , and flp-21 genes . As shown in Figure 3D , animals exposed to 72 hr lawns , but not to 12 hr lawns , showed higher expression levels of the npr-1 , flp-18 , and flp-21 genes compared with the control animals on E . coli . To further confirm the relationship between avoidance behavior and intestinal bloating of animals exposed to 72 hr lawns , we studied avoidance in animals deficient in the nol-6 gene . Previous studies have shown that RNA interference ( RNAi ) -mediated knockdown of nol-6 , a nucleolar RNA-associated protein , reduces bloating of the intestinal lumen caused by bacterial infection ( Fuhrman et al . , 2009 ) . We found that nol-6 RNAi delayed pathogen avoidance ( Figure 3E , Figure 3—figure supplement 1A ) . Animals deficient in nol-6 , which failed to avoid P . aeruginosa at 8 hr ( Figure 3E ) , did not exhibit intestinal bloating at the same time point when exposed to 72 hr lawns ( Figure 3F , G ) . Consistent with the idea that bloating induces pathogen avoidance , nol-6 RNAi animals exposed to 72 hr lawns avoided P . aeruginosa at 24 hr ( Figure 3E ) , at which time they also exhibited intestinal bloating ( Figure 3—figure supplement 1B , C ) . These results indicate that nol-6 RNAi animals are not generally defective in avoidance behavior , and the delayed avoidance behavior is due to delayed intestinal bloating . Despite diminishing the avoidance behavior , knockdown of nol-6 did not affect the induction of daf-7 expression in either ASJ ( Figure 3H ) or ASI neurons ( Figure 3—figure supplement 1D ) . Taken together , these results suggest that bloating of the intestine , but not induction of daf-7 in ASJ neurons , underlies the avoidance behavior . We next tested whether the enhanced intestinal bloating of animals exposed for 8 hr to 72 hr lawns was due to increased bacterial colonization compared with animals exposed to 12 hr lawns for the same time . We found that while animals exposed to 12 hr lawns showed a consistent increase in intestinal colonization , animals exposed to 72 hr lawns did not show any increase in colonization during the first 8 hr of exposure ( Figure 4A ) . These results indicate that intestinal bloating on 72 hr lawns is independent of bacterial colonization . Thus , we investigated whether the survival of animals on the two types of lawns was different . Animals exposed to 72 hr lawns died significantly faster than animals exposed to 12 hr lawns ( Figure 4B ) , indicating that the virulence of P . aeruginosa was higher on 72 hr than on 12 hr lawns . It has been shown that a non-virulent P . aeruginosa strain deficient in the gene gacA , a global activator of gene expression and virulence , fails to elicit C . elegans avoidance ( Hao et al . , 2018; Singh and Aballay , 2019a ) . P . aeruginosa gacA mutants are also hampered in the induction of daf-7 in the ASJ chemosensory neuron ( Meisel et al . , 2014 ) . Thus , it is not clear whether the inability of P . aeruginosa gacA mutants to elicit an avoidance behavior is due to reduced daf-7 induction in the ASJ neuron or to their reduced virulence . To distinguish between these two possibilities and to test the role of P . aeruginosa virulence in C . elegans avoidance behavior , we tested the induction of daf-7 in ASJ neurons and avoidance behavior elicited by several strains of P . aeruginosa with reduced virulence . We selected a diverse set of P . aeruginosa mutants that have attenuated virulence in C . elegans ( Feinbaum et al . , 2012 ) . We confirmed that all the studied P . aeruginosa mutants exhibited reduced virulence compared with wild-type P . aeruginosa ( Figure 4—figure supplement 1A ) . All of these mutants were also deficient in the elicitation of pathogen avoidance behavior ( Figure 4C ) . We next tested the induction of daf-7::GFP in the ASJ neurons upon exposure to the aforementioned P . aeruginosa mutants . We found that while kinB and rhlR mutants of P . aeruginosa were deficient in the induction of daf-7 in ASJ neurons , the induction of daf-7 by lasR , lysC , and ptsP mutants was comparable to the induction by wild-type P . aeruginosa ( Figure 4D , E ) . The induction of daf-7 expression in ASI neurons by all of these P . aeruginosa mutants was indistinguishable from the induction by wild-type P . aeruginosa ( Figure 4—figure supplement 1B ) . The levels of daf-7 induction in the ASJ neurons induced by different P . aeruginosa mutants did not show any correlation with the mean occupancy of the animals on P . aeruginosa mutant lawns ( Figure 4F ) . In contrast , the mean survival of the animals on different P . aeruginosa mutants showed a strong correlation with the mean occupancy of the animals on P . aeruginosa mutant lawns ( Figure 4G ) . Taken together , these results show that the virulence of P . aeruginosa , and not its ability to induce daf-7 in ASJ neurons , correlates with C . elegans avoidance behavior . Based on the finding that intestinal bloating caused by infection , and not chemosensation of P . aeruginosa phenazines , elicits the avoidance behavior , we hypothesized that intestinal infection may be responsible for the associative learning of pathogens . It is known that prior exposure to P . aeruginosa for several hours enables C . elegans to preferentially choose nonpathogenic E . coli over P . aeruginosa ( Zhang et al . , 2005 ) . Similarly , in a two-choice assay , when naïve animals were given a choice between E . coli and P . aeruginosa , the animals changed their preference from P . aeruginosa to E . coli after 8 hr of exposure ( Figure 5A , B ) . The P . aeruginosa choice index ( CI ) , described in Figure 5A , measures the preference of animals for P . aeruginosa with values ranging from −1 to 1 . The values 1 , –1 , and 0 indicate that all animals are on P . aeruginosa , all animals are away from P . aeruginosa , and an equal number of animals is on P . aeruginosa and E . coli , respectively . Because aerotaxis plays a role in pathogen avoidance ( Meisel et al . , 2014; Reddy et al . , 2009; Singh and Aballay , 2019a; Styer et al . , 2008 ) , we examined whether aerotaxis is also important for changes in preference from pathogenic to nonpathogenic bacteria . We reasoned that if animals experience different levels of oxygen on lawns of different bacteria , aerotaxis-regulating pathways might affect the microbial preference . It is known that P . aeruginosa lawns have lower oxygen levels than E . coli lawns ( Gray et al . , 2004; Reddy et al . , 2011 ) . To determine whether the animals on E . coli and P . aeruginosa respond to the different levels of oxygen of the two types of lawns , we used the cysl-2p::GFP reporter strain . Expression of the gene cysl-2 is regulated by hypoxia-inducible factor 1 ( HIF-1 ) , a transcription factor that is induced by low levels of oxygen ( Ma et al . , 2012 ) . Since HIF-1 is degraded at higher and accumulates at lower oxygen levels ( Jiang et al . , 2001 ) , the expression levels of cysl-2 correlate inversely with the oxygen levels experienced by the animals . We found that animals exposed for 24 hr to P . aeruginosa lawns had higher GFP levels compared with those exposed to E . coli lawns ( Figure 5C , D ) . We observed that the loss of function mutants daf-7 and npr-1 , which are deficient in DAF-7/TGF-β , and NPR-1 signaling , respectively , inhibited both pathogen avoidance ( Figure 5—figure supplement 1A , B ) and the change in preference from P . aeruginosa to E . coli ( Figure 5E , F ) . These two neuroendocrine pathways are known to act in parallel ( Chang et al . , 2006; de Bono et al . , 2002 ) ; therefore , we investigated the behaviors of daf-7 ( ok3125 ) ;npr-1 ( ad609 ) animals . We found that the aforementioned phenotypes were enhanced in daf-7 ( ok3125 ) ;npr-1 ( ad609 ) animals ( Figure 5G , Figure 5—figure supplement 1C ) . Taken together , these results suggest that aerotaxis-regulating pathways are required for the change in microbial preference upon pathogen infection . The genetic interactions for oxygen preference have been well characterized in C . elegans ( Chang et al . , 2006; Chang and Bargmann , 2008 ) . The increased preference for low oxygen in loss of function npr-1 and daf-7 mutants requires the function of the transient receptor potential channel vanilloid ( TRPV ) genes osm-9 and ocr-2 ( Chang et al . , 2006 ) . Because loss of function ocr-2 and osm-9 mutants have an increased preference for high oxygen levels ( Chang et al . , 2006 ) , and because E . coli lawns have relatively higher oxygen levels ( Gray et al . , 2004; Reddy et al . , 2011 ) , we reasoned that these mutants should rapidly change their preference to E . coli if given the choice between E . coli and P . aeruginosa . First , we studied whether the loss of function mutants ocr-2 ( ak47 ) and osm-9 ( yz6 ) showed enhanced avoidance of P . aeruginosa lawns . As expected , ocr-2 ( ak47 ) and osm-9 ( yz6 ) animals exhibited a strong enhancement of avoidance behavior ( Figure 6A , Figure 6—figure supplement 1A ) . These animals also showed a rapid change in preference to E . coli lawns in the two-choice assay ( Figure 6B , Figure 6—figure supplement 1B ) . The preference for high oxygen in ocr-2 ( ak47 ) and osm-9 ( yz6 ) mutants is suppressed by the loss of function mutation in egl-9 , a negative regulator of HIF-1 ( Chang and Bargmann , 2008 ) . Consistent with the function of EGL-9 , we observed that the egl-9 mutation suppressed both the enhanced pathogen avoidance and E . coli preference of osm-9 ( yz6 ) animals ( Figure 6C , D ) . These results show that aerotaxis regulates both P . aeruginosa avoidance and changes in microbial preference . While the above results showed that aversive learning and changes in microbial preference require aerotaxis pathways , they do not provide insights into the signaling during microbial infection that leads to changes in microbial preference . Bloating of the intestinal lumen upon pathogen infection activates the NPR-1/GPCR and DAF-7/TGF-β pathways , which results in a preference towards high oxygen ( Singh and Aballay , 2019a ) . Thus , we reasoned that animals with bloated intestines should show much more rapid learning and change in preference to E . coli in the two-choice assay . We examined the change in preference from P . aeruginosa to E . coli in the two-choice assay of aex-5 and egl-8 knockdown animals . Knockdown of these genes caused bloating of the intestinal lumen and led to enhanced avoidance of P . aeruginosa ( Singh and Aballay , 2019a ) . As shown in Figure 6E , F , inhibition of these genes by RNAi and mutations also elicited a rapid change in preference to E . coli lawns in the two-choice assay . We were also able to elicit a rapid preference towards E . coli by exposing the animals to 5% oxygen ( Figure 6—figure supplement 2 ) . Exposure to low oxygen levels alone does not affect either intestinal bacterial colonization or bloating ( Figure 6—figure supplement 3 ) . Because bloating accelerates the preference of the animals towards E . coli , we predicted that animals resistant to infection and bloating would not be capable of changing bacterial preferences . We studied the change in microbial preference in nol-6 RNAi animals that are resistant to P . aeruginosa infection and bloating . As shown in Figure 6G , nol-6 RNAi animals were defective in aversive learning toward P . aeruginosa . Together , these results suggest that intestinal bloating caused by pathogen infection , which modulates aerotaxis-regulating neuroendocrine pathways , is important for the learning process that leads to change in microbial preference from bacterial lawns containing relatively lower to those containing relatively higher oxygen levels .
Our study establishes that intestinal infection and bloating of the lumen , which depend on the virulence of P . aeruginosa , regulate both pathogen avoidance and aversive learning by modulating the neuroendocrine pathways NPR-1/GPCR and DAF-7/TGF-β that control aerotaxis behavior ( Figure 7 ) . Enhanced activities of these pathways , as a consequence of intestinal infection , lead to the avoidance of low oxygen , resulting in the avoidance of bacterial lawns with low oxygen due to microbial metabolism . Intestinal infection-mediated avoidance of low oxygen also drives the movement of animals from P . aeruginosa to E . coli lawns , which have been reported to have relatively higher oxygen levels . Thus , modulation of aerotaxis-regulating neuroendocrine pathways by intestinal infection plays a role in the learning process , resulting in changes in the preference of animals from P . aeruginosa to E . coli ( Figure 7 ) . Our study also indicates that microbial perception is controlled by inputs from the intestine during infection in C . elegans . P . aeruginosa secondary metabolites , including phenazine-1-carboxamide , have been shown to activate the DAF-7/TGF-β pathway in the chemosensory neuron pair ASJ ( Meisel et al . , 2014 ) . However , we found that phenazine-1-carboxamide does not play a role in elicitation of the avoidance behavior ( Figure 1B , H ) . The induction of DAF-7/TGF-β in ASJ chemosensory neurons was observed within 6 min of exposure to P . aeruginosa , and no further changes were observed up to 24 hr ( Meisel et al . , 2014 ) . However , the avoidance behavior was observed only after hours of interaction with P . aeruginosa . Therefore , while the activity of the DAF-7/TGF-β pathway is required to elicit the avoidance behavior ( Figure 5—figure supplement 1A ) , its rapid activation in ASJ neurons by chemosensation appears to be insufficient for induction of this behavior . Because the DAF-7/TGF-β and NPR-1/GPCR pathways are induced by intestinal bloating ( Singh and Aballay , 2019a ) and act synergistically to elicit pathogen avoidance ( Figure 5G , Figure 5—figure supplement 1C ) , it is likely that animals integrate multiple inputs , including intestinal infection and chemosensation , to induce defense responses . Indeed , in addition to O2 , animals sense CO2 and NO from bacterial lawns via multiple chemosensory neurons to control the pathogen avoidance behavior ( Brandt and Ringstad , 2015; Hao et al . , 2018 ) . C . elegans possesses an innate attraction to the smell of P . aeruginosa , and a brief exposure to P . aeruginosa does not lead to changes in preference to E . coli ( Ha et al . , 2010; Zhang et al . , 2005 ) . Because the olfactory aversive learning of P . aeruginosa and changes in preference to E . coli require several hours of exposure to P . aeruginosa , the mechanism by which the animals learn to avoid pathogens over time remain unclear . Here we show that bloating of the intestine caused by P . aeruginosa infection is required for the learning process . We demonstrate that the learning involves modulation of aerotaxis-regulating neuroendocrine pathways by intestinal bloating . By genetic modulation of the aerotaxis behavior of the animals , we are able to induce a rapid change in preference from P . aeruginosa to E . coli . Feeding on several pathogenic , but not nonpathogenic , bacteria leads to colonization and bloating of the intestine ( Fuhrman et al . , 2009; Irazoqui et al . , 2010; Kurz et al . , 2003; Yuen and Ausubel , 2018 ) . Thus , the bloating-induced defense response might be a generalized response of C . elegans to infection by different pathogens and enables the animals to distinguish between pathogenic and innocuous microbes . Indeed , the induction of intestinal bloating by genetic manipulation activates avoidance behavior towards even nonpathogenic bacteria ( Kumar et al . , 2019; Singh and Aballay , 2019a ) . Interestingly , the neuropeptide Y ( NPY ) -related signaling neuroendocrine pathway NPR-1 , which is activated by intestinal bloating in C . elegans , appears to have a conserved mode of activation and action across a variety of disparate species ( Duvall et al . , 2019; Singh and Aballay , 2019b ) . For instance , gut distension caused by a complete blood meal may activate the NPY-related signaling in mosquitoes and induce host aversion ( Singh and Aballay , 2019b ) . It will be interesting to study whether intestinal bloating is a conserved danger signal that activates immune responses in different species . The precise mechanism by which bacterial pathogens cause bloating remains unclear . C . elegans possesses a rhythmic defecation cycle that is timed by an oscillation in intestinal pH ( Pfeiffer et al . , 2008 ) . Defects in the pH wave lead to defects in the defecation cycle causing intestinal bloating ( Pfeiffer et al . , 2008 ) . Pathogenic bacteria such as P . aeruginosa and Enterococcus faecalis , but not nonpathogenic E . coli , disturb the pH wave in C . elegans intestine ( Benomar et al . , 2018 ) . It is possible that it is this disturbance of pH wave by pathogens that leads to intestinal bloating in C . elegans . However , it remains to be determined how pathogenic bacteria disturb the pH wave . Replication of bacteria in the intestinal lumen alone and other possible mechanisms could also lead to intestinal bloating . Increasing evidence shows that C . elegans recognizes potential pathogen attack by sensing homeostasis perturbations , including perturbations of core cellular processes , DNA damage , and intestinal bloating ( Dunbar et al . , 2012; Ermolaeva et al . , 2013; Kumar et al . , 2019; McEwan et al . , 2012; Melo and Ruvkun , 2012; Pukkila-Worley , 2016; Singh and Aballay , 2019a ) . Thus , these studies show that the physiological changes induced by pathogen infection or toxins generate modifications in C . elegans homeostasis that elicit an innate immune response and microbial aversion behavior via neuroendocrine signaling . It remains to be determined whether C . elegans is also capable of recognizing infecting microbes through more traditional mechanisms like those capable of recognizing microbe-associated molecular patterns ( MAMPs ) . C . elegans mounts immune responses toward heat-killed pathogenic bacteria ( Pukkila-Worley and Ausubel , 2012; Yuen and Ausubel , 2018 ) , suggesting the possible existence of MAMP-like mechanisms . However , the interpretation of these results is not straightforward because heat-killed bacteria could lead to intestinal bloating and activation of defense responses ( Singh and Aballay , 2019a ) . Our studies indicate that chemosensation of P . aeruginosa phenazines , which leads to the rapid induction of DAF-7/TGF-β in ASJ chemosensory neurons in C . elegans , is insufficient for the elicitation of pathogen avoidance behavior . We cannot rule out the possibility that other metabolites or virulence factors play a role in the elicitation of pathogen avoidance . Our results suggest that intestinal bloating caused by microbial infection plays a crucial role in aversive learning and microbial preference from pathogenic P . aeruginosa to nonpathogenic E . coli . Recent studies in Drosophila melanogaster also highlight the role of intestinal infection in modulating the immune response and social behavior ( Chen et al . , 2019 ) . Therefore , the modulation of behavior by intestinal infection appears to be conserved across species . Future research further deciphering these conserved pathways will aid in better understanding intestinal infection and gut dysbiosis-mediated behavioral and physiological changes .
The following bacterial strains were used: Escherichia coli OP50 , E . coli HT115 ( DE3 ) , E . coli DH5α-GFP , Pseudomonas aeruginosa PA14 , P . aeruginosa PA14-GFP , P . aeruginosa PA14 ΔkinB , P . aeruginosa PA14 ΔlasR , P . aeruginosa PA14 ΔlysC , P . aeruginosa PA14 ΔrhlR , P . aeruginosa PA14 Δphz ( lacking both the phzA1-G1 and phzA2-G2 operons ) , P . aeruginosa PA14 ΔphzH , P . aeruginosa PA14 ΔphzM , P . aeruginosa PA14 ΔphzS , and P . aeruginosa PA14 ΔptsP . The cultures of these bacteria were grown in Luria-Bertani ( LB ) broth at 37°C . C . elegans hermaphrodites were maintained on E . coli OP50 at 20°C unless otherwise indicated . Bristol N2 was used as the wild-type control unless otherwise indicated . Strains FK181 ksIs2 [daf-7p::GFP + rol-6 ( su1006 ) ] , DMS640 nIs470 [cysl-2p::GFP + myo-2p::mCherry] , CX4544 ocr-2 ( ak47 ) , JY190 osm-9 ( yz6 ) , JT23 aex-5 ( sa23 ) , MT1083 egl-8 ( n488 ) , JT307 egl-9 ( sa307 ) , DA609 npr-1 ( ad609 ) , and RB2302 daf-7 ( ok3125 ) were obtained from the Caenorhabditis Genetics Center ( University of Minnesota , Minneapolis , MN ) . The osm-9 ( yz6 ) ;egl-9 ( sa307 ) and daf-7 ( ok3125 ) ;npr-1 ( ad609 ) double mutants were obtained by standard genetic crosses . The daf-7 ( ok3125 ) and daf-7 ( ok3125 ) ;npr-1 ( ad609 ) hermaphrodites were maintained on E . coli OP50 at 15°C . RNAi was used to generate loss-of-function RNAi phenotypes by feeding nematodes E . coli strain HT115 ( DE3 ) expressing double-stranded RNA ( dsRNA ) homologous to a target gene ( Fraser et al . , 2000; Timmons and Fire , 1998 ) . RNAi was carried out as described previously ( Singh and Aballay , 2017 ) . Briefly , E . coli with the appropriate vectors were grown in LB broth containing ampicillin ( 100 μg/mL ) and tetracycline ( 12 . 5 μg/mL ) at 37°C overnight and plated onto NGM plates containing 100 μg/mL ampicillin and 3 mM isopropyl β-D-thiogalactoside ( IPTG ) ( RNAi plates ) . RNAi-expressing bacteria were allowed to grow overnight at 37°C . Gravid adults were transferred to RNAi-expressing bacterial lawns and allowed to lay eggs for 2 hr . The gravid adults were removed , and the eggs were allowed to develop at 20°C to young adults for subsequent assays . The RNAi clones were from the Ahringer RNAi library . The bacterial cultures were grown by inoculating individual bacterial colonies into 2 mL of LB broth and growing them for 10–12 hr on a shaker at 37°C . Then , 20 µL of the culture was plated onto the center of 3 . 5-cm-diameter standard slow-killing ( SK ) plates ( modified NGM agar plates ( 0 . 35% instead of 0 . 25% peptone ) ) . The plates were then incubated under the following conditions: 37°C for 12 hr; 37°C for 24 hr; 37°C for 24 hr followed by 25°C for 24 hr; and 37°C for 24 hr followed by 25°C for 48 hr . The P . aeruginosa lawns obtained upon incubation at 37°C for 12 hr were used for avoidance assays unless otherwise indicated . Thirty synchronized young gravid adult hermaphroditic animals grown on E . coli HT115 ( DE3 ) containing control vector or an RNAi clone targeting a gene were transferred outside the indicated bacterial lawns , and the numbers of animals on and off the lawns were counted at the specified times for each experiment . Three 3 . 5-cm-diameter plates were used per trial in every experiment . The experiments were performed at 25°C . The percent occupancy was calculated as ( Non lawn/Ntotal ) ×100 . At least three independent experiments were performed . E . coli OP50 cultures were grown by inoculating individual bacterial colonies into 10 mL of LB broth and growing them for 10–12 hr on a shaker at 37°C . The cultures were concentrated 10–20-fold , and 20 µL was plated onto the center of 3 . 5-cm-diameter modified NGM agar plates and incubated at 37°C for 12 hr . The stock solutions of different phenazines , which were prepared in ethanol , were diluted to either 10 or 20 µg in M9 salt solution to a final volume of 20 µL and added onto the E . coli lawns . For control plates , the equivalent amount of ethanol was mixed with M9 salt solution and added onto the E . coli lawns . These plates were then incubated at room temperature for 30 min before seeding with 20 synchronized young gravid adult hermaphroditic animals . The experiments were performed at 25°C . The percent occupancy was calculated as ( Non lawn/Ntotal ) ×100 . At least three independent experiments were performed . P . aeruginosa and E . coli HT115 cultures were grown by inoculating individual bacterial colonies into 2 mL and 10 mL of LB broth , respectively , and growing them for 10–12 hr on a shaker at 37°C . E . coli HT115 cultures were concentrated 10 to 20-fold before seeding on plates . Then , 20 µL of each inoculum was plated diagonally opposite onto 3 . 5-cm-diameter SK plates and incubated at 37°C for 12 hr . The plates were cooled to room temperature for at least 30 min before seeding with animals . Thirty synchronized young gravid adult hermaphroditic animals grown on E . coli HT115 ( DE3 ) containing control vector or an RNAi clone targeting a gene were transferred to the center of plates equidistant from both the lawns . For the experiments at 5% oxygen , the hypoxia chamber containing the plates was purged with 5% oxygen ( balanced with nitrogen ) for 5 min at a flow rate of 25 L/min . The chamber was then sealed and incubated at 25°C . For control , the two-choice assay plates were incubated at ambient oxygen . The numbers of animals on both lawns were counted at the specified times for each experiment . Three 3 . 5-cm-diameter plates were used per trial in every experiment . Experiments were performed at 25°C . The P . aeruginosa choice index ( P . aeruginosa CI ) was calculated as follows:P . aeruginosa CI=[No . of worms on P . aeruginosa-No . of worms on E . coli][No . of worms on P . aeruginosa+No . of worms on E . coli] At least three independent experiments were performed . E . coli DH5α-GFP cultures were grown by inoculating individual bacterial colonies into 25 mL of LB broth containing ampicillin ( 100 µg/mL ) and growing them for 10–12 hr on a shaker at 37°C . The cultures were concentrated 10–20-fold , and 300 µL were plated onto the center of 6-cm-diameter NGM agar plates containing ampicillin ( 100 µg/mL ) and incubated at 25°C for 48 hr . Fifty synchronized young gravid adult hermaphroditic N2 animals were transferred to plates containing E . coli DH5α-GFP lawns . The plates were then placed in a hypoxia chamber and the lids of the plates were removed . The hypoxia chamber was then purged with 8% oxygen ( balanced with nitrogen ) for 5 min at a flow rate of 25 L/min . The chamber was then sealed and incubated at 25°C for 24 hr . The control plates were incubated at ambient oxygen . After 24 hr of incubation , the animals were imaged for bacterial colonization and intestinal lumen diameter quantification . The C . elegans killing assays were carried out on wild-type P . aeruginosa PA14 lawns that were incubated at 37°C for 12 hr , or 37°C for 24 hr followed by 25°C for 48 hr . The bacterial lawns were prepared as described above . For full-lawn killing assays , 20 µL of an overnight culture of P . aeruginosa PA14 variants grown at 37°C was spread on the complete surface of 3 . 5-cm-diameter SK plates . The plates were incubated at 37°C for 12 hr and then cooled to room temperature for at least 30 min before seeding with synchronized young gravid adult hermaphroditic animals . The killing assays were performed at 25°C , and live animals were transferred daily to fresh plates . Animals were scored at the times indicated and were considered dead when they failed to respond to touch . The killing assays with different P . aeruginosa PA14 mutants were carried out on full lawns . Intestinal bacterial loads were quantified as described previously ( Singh and Aballay , 2019a ) . Briefly , P . aeruginosa-GFP lawns were prepared as described above . The plates were cooled to room temperature for at least 30 min before seeding with young gravid adult hermaphroditic animals . The assays were performed at 25°C . At the indicated times of exposure , the animals were transferred from P . aeruginosa-GFP plates to the center of fresh E . coli plates for 10 min to eliminate P . aeruginosa-GFP stuck to their body . Animals were transferred to the center of a new E . coli plate for 10 min to further eliminate external P . aeruginosa-GFP . Animals were transferred to fresh E . coli plates a third time for 10 min . Subsequently , ten animals/condition were transferred into 50 µL of PBS plus 0 . 01% Triton X-100 and ground . Serial dilutions of the lysates ( 101 , 102 , 103 , 104 ) were seeded onto LB plates containing 50 µg/mL of kanamycin to select for P . aeruginosa-GFP cells and grown overnight at 37°C . Single colonies were counted the next day and represented as the number of bacterial cells or CFU per animal . Three independent experiments were performed for each condition . To measure the induction of daf-7p::GFP in the ASI and ASJ neuron pairs , 12 hr ( incubated at 37°C for 12 hr ) or 72 hr ( incubated at 37°C for 24 hr followed by 25°C for 48 hr ) lawns of wild-type P . aeruginosa PA14 were used . The bacterial lawns were prepared as described above . The plates were cooled to room temperature for at least 30 min before seeding with daf-7p::GFP reporter young gravid adult hermaphroditic animals . These plates were incubated at 25°C for the indicated times , and then the animals were prepared for fluorescence imaging . To measure the induction of daf-7p::GFP upon exposure to different P . aeruginosa PA14 mutants , the bacterial lawns were prepared by incubation at 37°C for 12 hr . Young gravid adult hermaphroditic daf-7p::GFP animals were transferred to these plates and incubated at 25°C for 4 hr before preparing the animals for fluorescence imaging . The fluorescence intensity in the ASI and ASJ neurons was quantified using Image J software . Fluorescence imaging was carried out as described previously ( Singh and Aballay , 2017 ) with slight modifications . Briefly , the animals were anesthetized using an M9 salt solution containing 50 mM sodium azide and mounted onto 2% agar pads . The animals were then visualized using a Leica M165 FC fluorescence stereomicroscope . After the indicated treatment , the animals were anesthetized using an M9 salt solution containing 50 mM sodium azide , mounted onto 2% agar pads , and imaged . The diameter of the intestinal lumen was measured using the Leica LAS v4 . 6 software . At least 10 animals were used for each condition . Synchronized young gravid adult hermaphroditic cysl-2p::GFP animals grown on E . coli HT115 were transferred onto E . coli HT115 and P . aeruginosa lawns . P . aeruginosa and E . coli HT115 cultures were grown by inoculating individual bacterial colonies into 2 mL and 10 mL of LB broth , respectively , and growing them for 8–10 hr on a shaker at 37°C . E . coli HT115 cultures were concentrated 10 to 20-fold before seeding on plates . Then , 20 µL of each inoculum was plated onto the center of 3 . 5-cm-diameter SK plates . For full lawns of P . aeruginosa , 20 µL of inoculum was spread to completely cover the surface of 3 . 5-cm-diameter SK plates . The plates were incubated at 37°C for 12 hr and then cooled to room temperature for at least 30 min before seeding with synchronized young gravid adult hermaphroditic cysl-2p::GFP animals . The COPAS Biosort machine ( Union Biometrica ) was used to measure the time of flight ( length ) and fluorescence of individual worms . At least 100 worms were measured for each condition . Animals were synchronized by egg laying . Approximately 35 N2 gravid adult animals were transferred to 10 cm RNAi plates seeded with E . coli HT115 and allowed to lay eggs for 4 hr . The gravid adults were then removed , and the eggs were allowed to develop at 20°C . The animals were grown on the E . coli HT115 plates at 20°C until the young adult stage . Subsequently , the animals were transferred to 3 . 5-cm-diameter SK plates seeded with 20 µL of P . aeruginosa and pre-incubated at either 37°C for 12 hr or 37°C for 24 hr followed by 25°C for 48 hr . The control animals were maintained on E . coli HT115 . After transfer of the animals , the plates were incubated at 25°C for 8 hr . Subsequently , the animals were collected , washed with M9 buffer , and frozen in TRIzol reagent ( Life Technologies , Carlsbad , CA ) . Total RNA was extracted using the RNeasy Plus Universal Kit ( Qiagen , Netherlands ) . Residual genomic DNA was removed using TURBO DNase ( Life Technologies , Carlsbad , CA ) . A total of 6 μg of total RNA was reverse-transcribed with random primers using the High-Capacity cDNA Reverse Transcription Kit ( Applied Biosystems , Foster City , CA ) . qRT-PCR was conducted using the Applied Biosystems One-Step Real-time PCR protocol using SYBR Green fluorescence ( Applied Biosystems ) on an Applied Biosystems 7900HT real-time PCR machine in 96-well-plate format . Twenty-five-microliter reactions were analyzed as outlined by the manufacturer ( Applied Biosystems ) . The relative fold-changes of the transcripts were calculated using the comparative CT ( 2-ΔΔCT ) method and normalized to pan-actin ( act-1 , –3 , −4 ) . The cycle thresholds of the amplification were determined using StepOnePlus software ( Applied Biosystems ) . All samples were run in triplicate . The primer sequences have been described earlier ( Singh and Aballay , 2019a ) . The statistical analysis was performed with Prism 8 ( GraphPad ) . All error bars represent the standard deviation ( SD ) . The two-sample t test was used when needed , and the data were judged to be statistically significant when p<0 . 05 . In the figures , asterisks ( * ) denote statistical significance as follows: * , p<0 . 05 , ** , p<0 . 001 , *** , p<0 . 0001 , as compared with the appropriate controls . The Kaplan-Meier method was used to calculate the survival fractions , and statistical significance between survival curves was determined using the log-rank test . All experiments were performed in triplicate . | The bacteria that cause disease may be microscopic , but animals can use senses other than sight to protect themselves from infection . Some bacteria produce harmful toxins , which animals can instinctively recognize as being dangerous using their sense of smell or taste . This is called chemosensation , an innate ability that allows animals to react to chemical stimuli . But animals can also ‘learn’ to avoid harmful bacteria , though it remains unclear how they do so . Now , Singh and Aballay have used roundworms as a model organism to study this phenomenon . Roundworms feed on bacteria , so they need to be able to distinguish between disease-causing strains and harmless ones . However , they only start avoiding harmful bacteria after several hours of exposure , which would not necessarily be expected if they were using chemosensation . This prompted Singh and Aballay to investigate whether another mechanism could be teaching the roundworms to avoid disease-causing bacteria , by comparing roundworms that had been exposed to harmful or benign strains . As observed previously , the roundworms learned to avoid the harmful bacterium Pseudomonas aeruginosa . However , exposing the worms to certain chemicals produced by P . aeruginosa was not enough to teach them to avoid the bacterium . Instead , the experiments showed that when roundworms ingested disease-causing bacteria , the infection caused intestinal bloating . The more toxic the bacteria , the more the intestine swelled , triggering a neural pathway associated with a preference for oxygen . In a few hours , the worms learned to avoid the low oxygen environment associated with P . aeruginosa and developed a preference for high oxygen conditions surrounding harmless bacteria such as Escherichia coli . These results show how an intestinal infection can send signals to the nervous system to modulate animal behavior . Moreover , Singh and Aballay have identified a neural pathway that stimulates a behavioral host response to defend against infection . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"immunology",
"and",
"inflammation"
] | 2019 | Intestinal infection regulates behavior and learning via neuroendocrine signaling |
We computationally study how an icosahedral shell assembles around hundreds of molecules . Such a process occurs during the formation of the carboxysome , a bacterial microcompartment that assembles around many copies of the enzymes ribulose 1 , 5-bisphosphate carboxylase/ oxygenase and carbonic anhydrase to facilitate carbon fixation in cyanobacteria . Our simulations identify two classes of assembly pathways leading to encapsulation of many-molecule cargoes . In one , shell assembly proceeds concomitantly with cargo condensation . In the other , the cargo first forms a dense globule; then , shell proteins assemble around and bud from the condensed cargo complex . Although the model is simplified , the simulations predict intermediates and closure mechanisms not accessible in experiments , and show how assembly can be tuned between these two pathways by modulating protein interactions . In addition to elucidating assembly pathways and critical control parameters for microcompartment assembly , our results may guide the reengineering of viruses as nanoreactors that self-assemble around their reactants .
Encapsulation is a hallmark of biology . A cell must co-localize high concentrations of enzymes and reactants to perform the reactions that sustain life , and it must safely store genetic material to ensure long-term viability . While lipid-based organelles primarily fulfill these functions in eukaryotes , self-assembling protein shells take the lead in simpler organisms . For example , viruses surround their genomes with a protein capsid , while bacteria use large icosahedral shells known as bacterial microcompartments ( BMCs ) to sequester the enzymes and reactions responsible for particular metabolic pathways ( Kerfeld et al . , 2010; Axen et al . , 2014; Shively et al . , 1998; Bobik et al . , 1999; Erbilgin et al . , 2014; Petit et al . , 2013; Price and Badger , 1991; Shively et al . , 1973; Shively et al . , 1973; Kerfeld and Erbilgin , 2015 ) . Within diverse bacteria , BMC functions have been linked to bacterial growth , carbon fixation , symbiosis , or pathogenesis ( Kerfeld and Erbilgin , 2015 ) . Other protein-based compartments are found in bacteria and archea ( e . g . encapsulins ( Sutter et al . , 2008 ) and gas vesicles ( Pfeifer , 2012; Sutter et al . , 2008 ) ) and even eukaryotes ( e . g . vault particles ( Kickhoefer et al . , 1998 ) ) , while some viruses may assemble around lipidic globules ( Lindenbach and Rice , 2013; Faustino et al . , 2014 ) . Thus , understanding the factors that control microcompartment assembly and encapsulation is a central question in modern cell biology . From the perspectives of synthetic biology and nanoscience , there is great interest in reengineering BMCs or viruses as nanoreactors that spontaneously encapsulate enzymes and reagents in vitro ( e . g . Luque et al . , 2014; Douglas and Young , 1998; Rurup et al . , 2014; Patterson et al . , 2014; Patterson et al . , 2012; Zhu et al . , 2014; Rhee et al . , 2011; Rurup et al . , 2014; Wörsdörfer et al . , 2012; Comas-Garcia et al . , 2014 ) , or as customizable organelles that assemble around a programmable set of core enzymes in vivo , introducing capabilities such as carbon fixation or biofuel production into bacteria or other organisms ( e . g . Kerfeld and Erbilgin , 2015; Bonacci et al . , 2012; Parsons et al . , 2010; Choudhary et al . , 2012; Lassila et al . , 2014 ) . However , the principles controlling such co-assembly processes have yet to be established , and it is not clear how to design systems to maximize encapsulation . In this article we take a step toward this goal , by developing theoretical and computational models that describe the dynamical encapsulation of hundreds of cargo molecules by self-assembling icosahedral shells . Although our models are general , we are motivated by recent experiments on a type of BMC known as the carboxysome ( Kerfeld et al . , 2010; Schmid et al . , 2006; Iancu et al . , 2007; Tanaka et al . , 2008 ) . Carboxysomes are large ( 40–400 nm ) , roughly icosahedral shells that encapsulate a dense complex of the enzyme ribulose-1 , 5-bisphosphate carboxylase/oxygenase ( RuBisCO ) and other proteins to facilitate the Calvin-Bensen-Bassham cycle in autotrophic bacteria ( Price and Badger , 1991; Shively et al . , 1973; Shively et al . , 1973; Iancu et al . , 2007; 2010; Kerfeld et al . , 2010; Tanaka et al . , 2008 ) . Recently , striking microscopy experiments visualized β-carboxysome shells assembling on and budding from procarboxysomes ( the condensed complex of RuBisCO and other proteins found in the interior of carboxysomes ) ( Cameron et al . , 2013; Chen et al . , 2013 ) . Genomic analysis suggests that many BMCs with diverse functions assemble via similar pathways ( Cameron et al . , 2013; Kerfeld and Erbilgin , 2015 ) . However , the mechanisms of budding and pinch-off to close the shell remain incompletely understood because of the small size and transient nature of assembly intermediates . Moreover , experiments suggest that α-carboxysomes ( another form of carboxysome ) assemble by a different mechanism , in which shell assembly encapsulates an initially diffuse pool of RuBisCO ( Iancu et al . , 2010; Cai et al . , 2015 ) . The factors determining which of these assembly pathways occurs are unknown . BMC assembly is driven by a complex interplay of interactions among the proteins forming the external shell and the interior cargo . It is difficult , with experiments alone , to parse these interactions for those mechanisms and factors that critically influence assembly pathways , especially due to the lack of an in vitro assembly system . Models which can correlate individual factors to their effect on assembly are therefore an important complement to experiments . Previous experimental and theoretical studies of encapsulation by icosahedral shells , e . g . the assembly of viral capsids around their nucleic acid genomes ( e . g . Hu and Shklovskii , 2007; Kivenson and Hagan , 2010; Elrad and Hagan , 2010; Perlmutter et al . , 2013; 2014; Mahalik and Muthukumar , 2012; Zhang et al . , 2013; Zhang and Linse , 2013; Hagan , 2008; Devkota et al . , 2009; Dixit et al . , 2006; Borodavka et al . , 2012; Dykeman et al . , 2013; 2014; Zlotnick et al . , 2013; Johnson et al . , 2004; Patel et al . , 2015; Cadena-Nava et al . , 2012; Comas-Garcia et al . , 2012; 2014; Garmann et al . , 2014a; 2014b; Malyutin and Dragnea , 2013 ) , have demonstrated that the structure of the cargo can strongly influence assembly pathways and products . However , BMCs assemble around a cargo which is topologically different from a nucleic acid — a fluid complex comprising many , noncovalently linked molecules . We demonstrate here that changing the cargo topology leads to new assembly pathways and different critical control parameters . We present phase diagrams and analysis of dynamical simulation trajectories showing how the thermodynamics , assembly pathways , and emergent structures depend on the interactions among shell proteins and cargo molecules . Within distinct parameter ranges , we observe two classes of assembly pathways , which resemble those suggested for respectively α- or β-carboxysomes . We find that tunability of cargo loading is a key functional difference between the two classes of pathways . Shells assembled around a diffuse cargo can be varied from empty ( containing almost no cargo ) to completely full , whereas assembly around a condensed , procarboxysome-like complex invariably produces full shells . While we find that the encapsulated cargo becomes ordered due to confinement , complete crystalline order in the globule before encapsulation inhibits budding . We discuss these results in the context of recent observations on carboxysome assembly , and their implications for engineering BMCs , viruses or drug delivery vehicles that assemble around a fluid cargo ( e . g . Refs . [Kerfeld and Erbilgin , 2015; Parsons et al . , 2010; Choudhary et al . , 2012; Lassila et al . , 2014; Luque et al . , 2014; Douglas and Young , 1998; Rurup et al . , 2014; Patterson et al . , 2014; Patterson et al . , 2012; Zhu et al . , 2014; Rhee et al . , 2011; Rurup et al . , 2014; Wörsdörfer et al . , 2012] ) .
We begin by discussing assembly behavior when the cargo-cargo interactions are strong enough to drive equilibrium phase coexistence ( εCC≥1 . 6 ) . Near the phase boundary ( εCC=1 . 6 ) a system of pure cargo particles is metastable on the timescales we simulate . However , for εSC>4 , adding shell subunits drives nucleation of a cargo globule with shell subunits adsorbed on the surface . The subsequent fate of the globule depends on parameter values; typical simulation end-states are shown as a function of parameter values in Figure 3 . For moderate interaction strengths ( 2 . 5≤εSS≤3 . 5 ) the globule grows to a large size , typically containing at least twice the cargo molecules that can be packaged within a complete shell . Adsorbed shell subunits then reversibly associate to form ordered clusters . Once a cluster acquires enough inter-subunit interactions to be a stable nucleus , it grows by coagulation of additional subunits or other adsorbed clusters . For the parameter set corresponding to Figure 2A , nucleation is fast in comparison to cluster growth , and thus two nuclei grow simultaneously . The last three images show the system immediately preceding and following detachment of the lower shell . Missing only one of its 32 subunits , the shell is connected to the remainder of the droplet only by a narrow neck of cargo . Insertion of the final subunit breaks the neck and completes shell detachment . The complete shell contains 120–130 cargo particles , which is slighty above random close packing ( ≈120 particles ) but below fcc density ( ≈150 particles , see appendix 1 . 2 ) . 10 . 7554/eLife . 14078 . 009Figure 3 . Results of assembly around a cargo globule . ( A ) The most frequently observed assembly outcome is overlaid on a color map of the theoretical free energy density difference Δfassem ( Equation ( 3 ) ) between assembled shells and the unassembled globule . Results are plotted against the shell-cargo adsorption strength εSC and the shell-shell interaction strength εSS for indicated values of the cargo-cargo interaction strength εCC . ( B ) Representative snapshots of the predominant assembly outcomes shown in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 00910 . 7554/eLife . 14078 . 010Figure 3—source data 1 . List of all simulation outcomes for Figures 3A , 5A . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01010 . 7554/eLife . 14078 . 011Figure 3—source data 2 . Criteria used to categorize assembly outcomes . The sizes of each cargo globule and shell assemblage , and associations between shell assemblages and cargo globules , were determined by clustering . The outcome was then categorized according to the criteria listed in this table . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01110 . 7554/eLife . 14078 . 012Figure 3—figure supplement 1 . The distribution of assembly outcomes in Figure 3A is shown as a function of εSC for indicated values of εCC and εSS . Ten simulations were performed at each set of parameter values . Representative snapshots corresponding to each outcome are shown in Figure 3B . Simulations were performed for 3×108 timesteps . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01210 . 7554/eLife . 14078 . 013Figure 3—figure supplement 2 . Results of assembly around a pre-equilibrated cargo globule . The most frequently observed assembly outcome is overlaid on a color map of the theoretical free energy density difference Δfassem ( Equation ( 3 ) ) between assembled shells and the unassembled globule . Results are plotted against the shell-cargo adsorption strength εSC and the shell-shell interaction strength εSS for indicated values of the cargo-cargo interaction strength εCC . Outcomes are defined as in Figure 3B of the main text . The outcome of each simulation for this figure is listed in Figure 3—figure supplement 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01310 . 7554/eLife . 14078 . 014Figure 3—figure supplement 2—source data 1 . List of all simulation outcomes for Figure 3—figure supplement 1—2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01410 . 7554/eLife . 14078 . 015Figure 3—figure supplement 3 . The number of cargo particles packaged as a function of parameters . ( A ) The mean number of cargo molecules encapsulated by shells assembled in dynamics simulations for εCC=1 . 6 . The results are averaged over all complete shells ( for any εSS ) assembled at each value of εSC , the error bars indicate 95% confidence intervals . ( B ) The equilibrium number of cargo particles packaged in shells as a function of the shell-cargo and cargo-cargo interaction strengths . The equilibrium cargo loading was calculated by performing simulations initialized with a pre-assembled shell , for which the excluders on one subunit were made permeable to cargo particles . We then performed two simulations at each parameter set , each of length 5×105 timesteps . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 015 Increasing the shell-shell interaction strength drives faster shell assembly and closure , thus limiting the size of the globule before budding . For the largest interaction strength we simulated ( εSS=4 . 5 ) the globule typically does not exceed the size of a single shell , and multiple globules nucleate within the simulation box ( Figure 2—figure supplement 1 ) . This observation could place an upper bound on shell-shell interaction strengths , since multiple nucleation events were rare in the carboxysome assembly experiments ( Cameron et al . , 2013 ) ( however , we discuss potential complicating factors within the cellular environment below ) . To quantify the relationship between assembly mechanism and parameter values , we calculate an assembly order parameter , defined as the maximum number of unassembled subunits adsorbed onto a globule during an assembly trajectory . The order parameter is shown as a function of the interaction strengths in Figure 4 . For εCC≥1 . 6 and εSS≤3 we observe large values of the order parameter ( e . g . >32 , the red and yellow regions in Figure 4 ) , which indicate formation of a large amorphous globule consisent with the procarboxysome precursor to carboxysome shell assembly ( Cameron et al . , 2013 ) . 10 . 7554/eLife . 14078 . 016Figure 4 . Dependence of assembly pathway on shell-cargo and shell-shell interaction strength . The assembly order parameter , defined as the maximum number of unassembled shell subunits adsorbed on a globule at any point during a trajectory , is shown as a function of εSC and εSS for indicated values of the cargo-cargo interaction εCC . Large numbers of adsorbed unassembled subunits ( >32 ) indicate the two step assembly mechanism ( Figure 2A ) , whereas smaller values correspond to simultaneous assembly and cargo condensation ( Figure 2B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01610 . 7554/eLife . 14078 . 017Figure 4—figure supplement 1 . Assembly order parameter values for εCC=1 . 8 and εCC=2 . 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 017 For εCC=1 . 3 the cargo forms an equilibrium vapor phase in the absence of shell subunits . However , above threshold values of εSS and εSC , the diffuse cargo molecules drive nucleation of shell assembly . The subsequent assembly pathway depends sensitively on the shell-cargo interaction strength . For low εSC ( Figure 2C ) , assembly captures only a few cargo molecules , leading to complete , but nearly empty shells . For larger εSC ( Figure 2B , and Simulation Video 2 ) , the shell-cargo interactions drive local condensation of cargo molecules . Shell assembly and cargo complexation then proceed in concert , resembling the mechanism proposed for assembly of α-carboxysomes ( Iancu et al . , 2010 ) . Thus , tuning the shell-cargo interaction dramatically affects cargo loading , with a sharp transition from empty to filled shells around εSC=2 . This transition closely tracks the equilibrium filling fraction ( Figure 5C ) , measured by simulating a complete shell made permeable to cargo molecules . This effect is comparable to the condensation of water vapor below its dew point inside of hydrophilic cavities . In contrast , assembly around a globule only generates full shells . 10 . 7554/eLife . 14078 . 018Figure 5 . Results of assembly around a cargo with weak interactions ( εCC=1 . 3kBT ) . ( A ) The most frequently observed assembly outcome as a function of εSS and εSC . The distribution of outcomes for εSS=4 is shown in Figure 3—figure supplement 2 , and a data file containing the outcome for each trial at each parameter set is included ( Figure 3—source data 1 ) . ( B ) Representative snapshots for the outcomes shown in ( A ) . The complete shell outcomes are shown with the excluders rendered opaque ( left ) and transparent ( right ) to enable visualizing the encapsulated cargo . ( C ) The number of cargo molecules encapsulated by shells assembled in dynamics simulations ( red symbols ) is compared to the results of equilibrium simulations ( black line ) . The dynamics results are averaged over all complete shells ( for any εSS ) assembled at each value of εSC , the error bars indicate 95% confidence intervals . Most simulations were performed for 3×108 timesteps; simulations with εSS=4 . 5 , εSC≤4 , and εCC=1 . 3 exhibited partially assembled shells at 3×108 timesteps , and were continued up to 7 . 2×109 timesteps . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01810 . 7554/eLife . 14078 . 019Figure 5—figure supplement 1 . Assembly yields calculated by simulation and theory . ( A ) The color map shows the fraction of subunits in assembled shells ( fc ) obtained from numerically solving Equation ( 1–2 ) , using the parameter values described in appendix 2 , as a function of εSC and εSS for assembly around gas phase cargo εCC=1 . 3 ( left ) and assembly and budding from a pre-equilibrated globule εCC=1 . 6 ( right ) . The white circles overlaid on the plots quantify the fraction of dynamical simulations that led to at least one well-formed capsid ( defined as shells containing 12 pentamers and 20 hexamers , each interacting with respectively 5 or 6 neighbors ) . The size of each white circle is proportional to the yield obtained from dynamical simulations at that parameter set , with the largest circle corresponding to 100% . ( B ) The color map shows the simulation result for the fraction of subunits in of any type of assemblage ( defined as any assemblage comprising 10 or more subunits ) as a function of εSC and εSS for εCC=1 . 3 ( left ) and εCC=1 . 6 ( right ) . We see that the theoretical prediction of the onset of assembly roughly corresponds to the boundary between assembly and no assembly in the simulations , except that the simulation boundary is seen at slightly higher parameter values in all cases , and for εCC=1 . 3 the simulation boundary slopes upward with εSC more sharply than the theoretical prediction . Both of these differences can be attributed to prohibitive nucleation barriers which arise for parameter values near the threshold equilibrium values . As discussed in the main text , decreasing εSC reduces the ability for a partially assembled shells to condense cargo molecules , leading to longer nucleation timescales and hence a wider range of εSS between the equilibrium threshold for assembly and the threshold for observing nucleation within our simulation timescale . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 01910 . 7554/eLife . 14078 . 020Figure 5—figure supplement 2 . The effect of varying cargo diameter on assembly . ( A ) The fraction of trajectories resulting in assembly of a complete shell is shown as a function of εSC for indicated cargo diameters ( σC ) and εCC=1 . 3 . Each data point corresponds to 10 independent simulations . ( B ) The maximum number of cargo particles encapsulated into a complete shell for each diameter in the simulations shown in ( A ) . ( C ) Cutaway view of assembled shells corresponding to each data point in ( B ) . Further information . For most results shown in this article , we set the size of the cargo to be commensurate with the size of the shell subunits . This is qualitatively consistent with BMCs; e . g . , for carboxysomes the diameter of a RuBisCO holoenzyme is about twice the circum-diameter of a hexamer or pentamer . To investigate how sensitive cargo encapsulation is to the ratio of cargo and shell subunits sizes , we performed additional simulations with cargo diameters in the range σC∈[0 . 6 , 2] , where σC is the cargo diameter in Equation ( A5 ) . In these simulations we maintained a constant cargo volume fraction and box size , so the number of simulated cargo particles varies inversely with the cargo volume . As shown in ( A ) , assembly can accommodate such variations in the cargo diameter , but the yields and robustness to variations in εSC diminish as σC varies from 1 . This may suggest that commensurate shell subunits and cargo sizes are optimal for encapsulation; however , further exploration is required to determine whether varying other parameters such as the cargo volume fraction or the length scale of the subunit-cargo interaction would change this result . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 020 Assembly of full shells ( by either pathway , Figure 2A or Figure 2B ) is typically about two orders of magnitude faster than assembly of empty shells ( Figure 2C ) . This disparity demonstrates the key role that the cargo plays in promoting shell association , during all stages of assembly . Cargo molecules initially promote shell nucleation by stabilizing interactions among small , sub-nucleated clusters . Then , the presence of a condensed globule provides a large cross-section for adsorption of additional subunits , significantly enhancing the flux of subunits to the partial capsid , thus increasing its growth rate . The condensed cargo particularly facilitates insertion of the last few subunits , which are significantly hindered by steric interactions , as noted previously for simulations of empty virus capsids ( Nguyen et al . , 2007 ) . Figure 5A shows how the products of assembly around cargo with weak interactions depends on parameters . While moderate parameter values lead to complete assembly , overly weak εSC and εSS ( lower left region of Figure 5A ) prevent shell nucleation , leading to the ‘Unnucleated’ outcome . In the limit of large εSC but weak εSS the shell-cargo interaction stabilizes small disordered globules ( ∼50 cargo particles , lower right region of Figure 5A ) , while under strong subunit and weak cargo interactions ( εSS=4 . 5 , εSC<5 ) shells nucleate but cannot condense the cargo , leading to the complete but slow assembly just discussed . As for assembly around a globule , overly strong interactions lead to overnucleation and malformed shells . However , the predominant mode of malformation is now shell collapse . Because the cargo is below its dew point , the locally condensed globule leads to a negative pressure on the shell subunits , which can flatten the shell and thus prevent closure of a symmetric shell . The simple free energy model ( Equations ( 1–2 ) ) reproduces the threshold parameter values required for shell assembly with no adjustable parameters ( color map in Figure 3 ) . Since it is an equilibrium model and only considers the free energy difference between complete and unassembled configurations , it cannot distinguish between parameter values that lead to complete assembly or kinetic traps at the long but finite simulation times . However , the thermodynamic calculation does suggest that the simulations resulting in ‘Attached’ shells would eventually reach completion on a longer timescale . We do not show Δfassem in Figure 5A because the globule is always less favorable than assembled shells for εCC=1 . 3 , but the yield of well-formed shells in our simulations roughly follows the prediction of the equilibrium theory ( Figure 5—figure supplement 1 ) . To investigate whether the results described above depend on assumptions within our model , we performed several sets of additional simulations . Firstly , we performed simulations in which the ratio between cargo diameter in shell subunit size was varied . As shown in Figure 5—figure supplement 2 , assembly is most robust for our default cargo diameter ( for which the model was parameterized ) , but productive assembly occurs for cargo diameters varied over a factor of four . Secondly , we performed assembly simulations with anisotropic cargo molecules with a shape motivated by the octomer structure of the RuBisCO holoenzyme ( Figure 2—figure supplement 2 ) . Thirdly , we performed a set of simulations in which we pre-equilibrated the cargo globule before introducing shell subunits into the system ( Figure 3—figure supplement 2 , Simulation Video 3 ) . Investigating this alternative initial condition was motivated by the fact that RuBisCO is present in the cell before induction of the carboxysome gene in the experiments of Ref . ( Cameron et al . , 2013 ) , and by the observation that multiple carboxysomes bud sequentially in time from a single procarboxysome . For εCC=1 . 6 the results are very similar to those obtained without pre-equilibrating the cargo . However , for εCC>1 . 6 , successful assembly and detachment is limited to more narrow ranges of shell-shell and shell-cargo interaction strengths than in Figure 3 , due to an increased prevalence of ‘Attached’ and ‘Stalled’ configurations . The latter are particularly common for εCC=3 , when the cargo forms a hexagonally close packed crystal which strongly resists deformation by shell protein assembly . 10 . 7554/eLife . 14078 . 021Video 3 . Animation of a simulation with a pre-equilibrated cargo globule . Parameters are εCC=1 . 6 , εSC=6 , and εSS=3 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 021 Taken together , the results from both assembly protocols ( Figure 3 and Figure 3—figure supplement 2 ) suggest that moderate effective cargo-cargo interactions are most consistent with the observations of shell assembly and budding in Refs . ( Cameron et al . , 2013; Chen et al . , 2013 ) . Such interactions are strong enough to drive cargo globule formation , but malleable enough to allow shell assembly to deform and eventually sever intra-globule interactions . Studies of assembled carboxysomes report varying degrees of order for the encapsulated cargo , ranging from none to paracrystalline order ( Iancu et al . , 2007; 2010; Kaneko et al . , 2006; Schmid et al . , 2006 ) . We therefore studied the relationship between cargo order and interaction parameters using equilibrium simulations ( see Figure 6 and Figure 6—figure supplement 1 ) . Below εCC<3kBT , we do not observe true fcc order of the encapsulated cargo . However , for all parameters leading to significant filling , even those well below the cargo liquid-vapor transition , the cargo becomes organized in concentric layers ( Figure 6 ) . We observe similar cargo organizations within shells which have budded from cargo globules in dynamical simulations . These results demonstrate that ordering of the cargo does not require crystallinity of the initial globule . Moreover , the magnitude of ordering increases with cargo loading , but , for fixed loading , is essentially independent of the cargo-shell interaction strength εSC . We observe ordering within filled shells due to confinement , even if even if εSC is set to 0 ( Figure 6—figure supplement 1 ) , as previously noted by Iancu et al . ( Iancu et al . , 2007 ) . 10 . 7554/eLife . 14078 . 022Figure 6 . Order of the encapsulated cargo . The spherically averaged density of cargo molecules inside a shell is shown as a function of radius for ( A ) εCC=1 . 6 and ( B ) εCC=1 . 3 for indicated values of the cargo-shell adhesion strength εSC , measured in equilibrium simulations . The density of the encapsulated cargo ranges from below random close packing to near hexagonal close packing density as εCC and εSC are increased ( see Figure 3—figure supplement 3 ) . A snapshot of cargo inside the shell is shown in Figure 5—figure supplement 2 . The raw data for this figure is provided in Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 02210 . 7554/eLife . 14078 . 023Figure 6—source data 1 . Raw data for Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 02310 . 7554/eLife . 14078 . 024Figure 6—figure supplement 1 . Ordering of the encapsulated cargo is primarily driven by confinement , not adhesion to the inner surface of the shell . The spherically averaged density distribution is shown as a function of distance from the shell center , for simulations in which a preset number n of cargo molecules are trapped within a complete shell , with the cargo-shell attraction turned off ( εSC=0 ) . The value of n corresponding to each curve is given in the legend , and the value of the subunit-shell energy εSC is shown above each plot . These simulations were each run for 5×105 timesteps . The simulations shown in Figure 6 were also run with a complete shell; however , one excluder was rendered permeable to cargo molecules allowing the number of encapsulated cargo molecules to equilibrate . Those simulations were also each run for 5×105 timesteps . Raw data for this figure is provided in Figure 6—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 02410 . 7554/eLife . 14078 . 025Figure 6—figure supplement 1—source data 1 . Raw data for Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 02510 . 7554/eLife . 14078 . 026Table 1 . Description of the assembly outcomes presented in Figures 3 , 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14078 . 026SymbolNameDescription▪Complete shell ( full ) Complete shell , full of cargo molecules◆Complete shell ( empty ) rComplete shell , almost empty of cargo molecules⚫AttachedNearly complete shells attached to a globule by a neck of cargo✳Over-nucleated/MalformedMultiple globules , with incomplete or malformed shells on their surfaces×StalledLarge globule with multiple incomplete or malformed shells on its surface□GlobuleCargo globule with unassembled shell subunits on its surface⊙UnnucleatedDiffuse subunits and cargo molecules
We have described an equilibrium theory and a dynamical computational model for the assembly of shells around a fluid cargo . Our simulations show that assembly can proceed by two classes of pathways: ( i ) a multi-step process in which the cargo forms a dense globule , followed by adsorption , assembly , and budding of shell proteins , or ( ii ) single-step assembly , with simultaneous aggregation of cargo molecules and shell assembly . This result demonstrates that the minimal interactions included in our model are sufficient to drive both classes of assembly pathways , suggesting that they are a generic feature of assembly around a fluid cargo . Moreover , while we cannot rule out the existence of active mechanisms in biological examples such as carboxysomes , our model demonstrates that the same interactions which drive assembly of shells can also drive budding from and closure around an amorphous globule of cargo . Our results suggest bounds on the relative strengths of interactions that drive BMC assembly in cells . The decisive control parameter determining the assembly pathway is the cohesive energy between cargo molecules , which could arise through direct cargo-cargo interactions or be mediated by auxiliary proteins ( Cameron et al . , 2013 ) . Relatively weak cargo interactions lead to single-step assembly pathways , while stronger interactions favor formation of the cargo-shell globule . However , the strength of cargo-shell and shell-shell interactions also play a role . Strong shell-shell interactions cause assembly to proceed rapidly during globule formation , limiting the size of the globule . Moreover , if a large globule is already present ( e . g . due to time-dependent protein concentrations within a cell ) , strong interactions tend to drive malformed assemblies . We find that an important functional difference between the two classes of assembly pathways is control over the amount of packaged cargo . While the multi-step assembly pathways always generate a shell filled with cargo molecules , shells assembling around a diffuse cargo can be tuned from nearly empty to completely full by controlling the strength of cargo-shell interactions . These results have implications for reengineering BMCs to encapsulate new core enzymes . Recent works demonstrated that protein cargos can be targeted to BMCs via encapsulation peptides that mediate cargo-shell interactions . However , packaged amounts were much lower than for native core enzymes ( Parsons et al . , 2010; Choudhary et al . , 2012; Lassila et al . , 2014 ) . Our simulations show that both cargo-shell and cargo-cargo interactions ( direct or mediated ) must be controlled to assemble full shells . We also find that a general equilibrium theory describes the ranges of parameter values for which assembly occurs . However , the dynamical simulations demonstrate that , at finite timescales , there is a rich variety of assembly morphologies . Formation of ordered , full shells requires a delicate balance of cargo-cargo , cargo-shell , and shell-shell interactions , all of which must be on the order 5-10kBT . This constraint is consistent with previous studies on viruses and other assembly systems , which found that formation of ordered states requires multiple , cooperative weak interactions between subunits ( Hagan , 2014; Whitelam and Jack , 2015 ) . Outside of optimal parameter regimes , the simulations predict alternative outcomes , ranging from no assembly to various alternative trapped intermediates , with the morphology depending on which interaction is strongest . We find that assembly is least robust to parameter variations when the cargo crystallizes before shell assembly . The assembling shell is unable to deform or penetrate the cargo complex , leading to defect-riddled , non-budded complexes . Within the limits of our simplified model , this observation suggests that procarboxysome complexes are at least partially fluid prior to successful shell assembly . Moreover , we find that observations of ordered cargo within assembled shells may be explained by packing constraints . An important limitation of the present study is that the model interactions are specific to the shell geometry shown in Figure 1 ( containing 20 hexamers ) because alternating edges on hexagonal subunits have attractive interactions only with pentagonal subunits . In reality BMCs contain many more hexamers ( formed from multiple protein sequences ) and thus must include a greater range of hexamer-hexamer interactions . Extension of the model to allow for this possibility would allow consideration of two important questions: ( 1 ) The mechanism controlling insertion of the 12 pentagons required for a closed shell topology . ( 2 ) The relationship between assembly pathway and BMC size polydispersity . In particular , experiments suggest that β-carboxysomes are more polydisperse than α-carboxysomes ( Price and Badger , 1991; Shively et al . , 1973; Shively et al . , 1973; Iancu et al . , 2007; 2010; Kerfeld et al . , 2010; Tanaka et al . , 2008 ) . We speculate that in the case of assembly around vapor-phase cargo , the size of the assembling shell will be primarily dictated by the preferred shell protein curvature and thus relatively uniform . However , during assembly around a condensed globule , the shell protein interactions could be strained to accommodate a globule which is larger or smaller than the preferred curvature , causing the shell size to depend on a complex balance of intermolecular interaction strengths and variables such as the local RuBisCO concentration . Our model is minimal , intended to elucidate general principles of assembly around a fluid cargo , and thus may apply to diverse systems including prokaryotic microcompartments , viruses , and engineered delivery vehicles . The predicted trends for how assembly mechanisms and morphologies vary with control parameters can be experimentally tested by microscopy experiments . Such testing will be most straightforward in vitro ( e . g . Luque et al . , 2014; Douglas and Young , 1998; Rurup et al . , 2014; Patterson et al . , 2014; Patterson et al . , 2012; Zhu et al . , 2014; Rhee et al . , 2011; Rurup et al . , 2014; Wörsdörfer et al . , 2012 ) , where subunit-subunit interactions can be tuned by varying solution conditions and the stoichiometries of shell and cargo species can be readily varied . While there is currently no BMC assembly system starting from purified components , our findings can be tested in vivo by mutations which alter known protein binding interfaces , or by altering expression levels of RuBisCO or carboxysome proteins . We anticipate that our model can serve as a qualitative guide for understanding how such multicomponent complexes assemble in natural systems , or to reengineer them for new applications . More broadly , our results demonstrate that the properties of encapsulated cargo , such as its topology , geometry and interaction strengths , strongly influence assembly pathways and morphologies .
To complement the finite-time simulations , we have developed a general thermodynamic description of assembly around a fluid cargo . We consider shells composed of species α=1 , 2 , …M , with nαshell subunits of species α in a complete shell , which encapsulates n0 cargo molecules ( the index 0 refers to cargo molecules henceforth ) . Assembly occurs from a dilute solution of cargo molecules with density ρ0 , shell subunits with density ρα for each species , and the density of assembled , full shells as ρshell . These are in equilibrium with a globule containing n0glob cargo molecules and nαglob subunits for each species α . We assume that , due to the asymmetric nature of the shell-cargo interaction , the shell subunits reside at the exterior of the globule ( as we observe in our simulations ) . The globule containing unassembled shell subunits thus resembles a spherical microemulsion droplet ( Safran , 1994 ) . Minimizing the total free energy ( see appendix 2 ) gives: ( 1 ) v0ρshell=exp[− ( Gshell−∑αnαshellμα ) /kBT] where Gshell is the interaction free energy of the assembled shell and μα are the chemical potentials of free cargo molecules and shell subunits , given by μα=kBTln ( ραv0 ) , with v0 a standard state volume and the globule composition given by ( 2 ) ∂Gglob ( {nαglob} ) ∂nαglob=μαfor α=0…M , with Gglob ( nsglob , n0glob ) as the globule free energy . ( 1 ) – ( 2 ) are the general equilibrium description for a system of assembling shells with a disordered-phase intermediate; application to a specific system requires specifying the forms of Gshell and Gglob . In appendix 2 we specify these equations for our computational model , allowing us to compare the equilibrium calculation with simulation results , using no free parameters . To compare the relative stabilities of the globule and assembled shells , we also calculate the free energy difference ( 3 ) Δfassem=ftot ( {nαglob=0} ) -ftot ( ρshell=0 ) , where the first term on the right-hand side is the minimized free energy for a system containing shells and free subunits but no globule , while the second term corresponds to the minimized free energy for a system containing subunits and the globule , but no assembly . | Bacterial microcompartments are protein shells that are found inside bacteria and enclose enzymes and other chemicals required for certain biological reactions . For example , the carboxysome is a type of microcompartment that enables the bacteria to convert the products of photosynthesis into sugars . During the formation of a microcompartment , the outer protein shell assembles around hundreds of enzymes and chemicals . This formation process is tightly controlled and involves multiple interactions between the shell proteins and the cargo – the enzymes and other reaction ingredients – they will enclose . Understanding how to control which enzymes are encapsulated within microcompartments could help researchers to re-engineer the microcompartments so that they contain drugs or other useful products . Recent studies have used microscopy to visualize how microcompartments are assembled . However , most of the intermediate structures that form during assembly are too small and short-lived to be seen . It has therefore not been possible to explore in detail how shell proteins collect the necessary cargo and then assemble into an ordered shell with the cargo on the inside . Experiments alone are probably not enough to understand the process , especially since microcompartment assembly can currently only be studied within live cells or cellular extract . Within these complex environments it is difficult to determine the effect of any individual factor on the overall assembly process . Perlmutter , Mohajerani and Hagan have now taken a different approach by developing computational and theoretical models to explore how microcompartments assemble . Computer simulations showed that microcompartments could assemble by two pathways . In one pathway , the protein shell and cargo coalesce at the same time . In the other pathway , the cargo molecules first assemble into a large disordered complex , with the shell proteins attached on the outside . The shell proteins then assemble , carving out a piece of the cargo complex . The simulations showed that many factors affect how the shell assembles , such as the strengths of the interactions between the shell proteins and the cargo . They also identified a factor that controls how much cargo ends up inside the assembled shell . Perlmutter , Mohajerani and Hagan found that , in addition to revealing how microcompartments may assemble within their natural setting , the simulations provided guidance on how to re-engineer microcompartments to assemble around other components . This would enable researchers to create customizable compartments that self-assemble within bacteria or other host organisms , for example to carry out carbon fixation or make biofuels . A future challenge will be to investigate other aspects of microcompartment assembly , such as the factors that control the size of these compartments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"physics",
"of",
"living",
"systems",
"computational",
"and",
"systems",
"biology"
] | 2016 | Many-molecule encapsulation by an icosahedral shell |
In mammals , testicular differentiation is initiated by transcription factors SRY and SOX9 in XY gonads , and ovarian differentiation involves R-spondin1 ( RSPO1 ) mediated activation of WNT/β-catenin signaling in XX gonads . Accordingly , the absence of RSPO1/Rspo1 in XX humans and mice leads to testicular differentiation and female-to-male sex reversal in a manner that does not requireSry or Sox9 in mice . Here we show that an alternate testis-differentiating factor exists and that this factor is Sox8 . Specifically , genetic ablation of Sox8 and Sox9 prevents ovarian-to-testicular reprogramming observed in XX Rspo1 loss-of-function mice . Consequently , Rspo1 Sox8 Sox9 triple mutant gonads developed as atrophied ovaries . Thus , SOX8 alone can compensate for the loss of SOX9 for Sertoli cell differentiation during female-to-male sex reversal .
During primary sex determination in mammals , a common precursor organ , the bipotential gonad , develops as a testis or ovary . In humans and mice , testicular development begins when SRY and SOX9 are expressed in the bipotential XY gonad . These transcription factors promote supporting cell progenitors to differentiate as Sertoli cells and form sex cords ( Gonen et al . , 2018; Chaboissier et al . , 2004; Barrionuevo et al . , 2006 ) , and this triggers a cascade of signaling events that are required for the differentiation of other cell populations in the testis ( Koopman et al . , 1991; Vidal et al . , 2001 ) . In XX embryos , the bipotential gonad differentiates as an ovary through a process that requires RSPO1-mediated activation of canonical WNT/β-catenin ( CTNNB1 ) signaling in somatic cells ( Parma et al . , 2006; Chassot et al . , 2008 ) . Ovarian fate also involves activation of FOXL2 , a transcription factor that is required in post-natal granulosa cells ( Schmidt et al . , 2004; Ottolenghi et al . , 2005; Uhlenhaut et al . , 2009 ) , which organize as follicles during embryogenesis in humans and after birth in mice ( McGee and Hsueh , 2000; Mork et al . , 2012 ) . For complete differentiation of testes or ovaries , an active repression of the opposite fate is necessary ( Kim et al . , 2006 ) . Inappropriate regulation within the molecular pathways governing sex determination can lead to partial or complete sex reversal phenotypes and infertility ( Wilhelm et al . , 2009 ) . Studies in humans and mice have shown that the pathway initiated by SRY/SOX9 or RSPO1/WNT/β-catenin signaling are indispensable for sex specific differentiation of the gonads . For example , in XY humans , SRY or SOX9 loss-of-function mutations prevent testis development ( Berta et al . , 1990; Houston et al . , 1983 ) . In mice , XY gonads developing without SRY or SOX9 lack Sertoli cells and seminiferous tubules and differentiate as ovaries that contain follicles ( Lovell-Badge and Robertson , 1990; Chaboissier et al . , 2004; Barrionuevo et al . , 2006; Lavery et al . , 2011; Kato et al . , 2013 ) , indicating Sry/Sox9 requirement . In XX humans and mice , SRY/Sry or SOX9/Sox9 gain-of-function mutations promote Sertoli cell differentiation and testicular development ( Sinclair et al . , 1990; Koopman et al . , 1991; Bishop et al . , 2000; Vidal et al . , 2001; Huang et al . , 1999 ) , indicating that SRY/SOX9 function is also sufficient for male gonad differentiation . With respect to the ovarian pathway , homozygous loss-of-function mutations for RSPO1/Rspo1 trigger partial female-to-male sex reversal in XX humans and mice ( Parma et al . , 2006; Chassot et al . , 2008 ) . In XX Rspo1 or Wnt4 mutant mice , Sertoli cells arise from a population of embryonic granulosa cells ( pre-granulosa cells ) that precociously exit their quiescent state , differentiate as mature granulosa cells , and reprogram as Sertoli cells ( Chassot et al . , 2008; Maatouk et al . , 2013 ) . The resulting gonad is an ovotestis containing seminiferous tubule-like structures with Sertoli cells and ovarian follicles with granulosa cells , indicating that SRY is dispensable for testicular differentiation . In addition , stabilization of WNT/CTNNB1 signaling in XY gonads leads to male-to-female sex reversal ( Maatouk et al . , 2008; Harris et al . , 2018 ) . Thus , RSPO1/WNT/CTNNB1 signaling is required for ovarian differentiation and female development in humans and mice . Given the prominent role of SOX9 for testicular development ( Chaboissier et al . , 2004; Barrionuevo et al . , 2009 ) , it was hypothesized that SOX9 is responsible for Sertoli cell differentiation in XX gonads developing without RSPO1/Rspo1 . This hypothesis was tested by co-inactivation of Rspo1 or Ctnnb1 and Sox9 in Rspo1-/-; Sox9fl/fl; Sf1:creTg/+ ( Lavery et al . , 2012 ) and in Ctnnb1fl/fl; Sox9fl/fl; Sf1:creTg/+double mutant mice ( Nicol and Yao , 2015 ) . Unexpectedly , XY and XX Rspo1 or Ctnnb1 mutant gonads lacking Sox9 exhibited Sertoli cells organized as testis cords ( Nicol and Yao , 2015; Lavery et al . , 2012 ) . Specifically , gonads in XX Rspo1-/-; Sox9fl/fl; Sf1:creTg/+ double mutant mice developed as ovotestes as in XX Rspo1-/- single mutants , and XY Rspo1-/-; Sox9fl/fl; Sf1:creTg/+ mutant mice developed hypoplastic testes capable of supporting the initial stages of spermatogenesis . These outcomes indicate that at least one alternate factor can promote testicular differentiation in Rspo1 mutant mice also lacking Sox9 in XY mice , and lacking both Sry and Sox9 in XX animals . This or these factors remained to be identified . Among the candidate genes that could promote testicular differentiation in the absence of Sry and Sox9 are the other members of the SoxE group of transcription factors that includes Sox9 , Sox8 and Sox10 ( Lavery et al . , 2012; Nicol and Yao , 2015 ) . However , Sox10 expression in testes depends on Sox8 and Sox9 ( Georg et al . , 2012 ) , and Sox10 loss-of-function mice are fertile ( Britsch et al . , 2001; Peirano and Wegner , 2000 ) , suggesting that Sox10 would not be the best candidate gene . For Sox8 , loss-of-function analyses in XY gonads show testicular development , indicating that Sox8 is not required for Sertoli cell differentiation during embryonic development ( Sock et al . , 2001 ) . However , a Sox8-null background enhanced the penetrance of the testis-to-ovary sex reversal phenotype in mice with reduced Sox9 expression ( Chaboissier et al . , 2004 ) , suggesting that Sox8 supports the function of Sox9 . Furthermore , in XY Sox9fl/fl; Sf1:creTg/+ single mutant mice and in XY Sox8-/-; Sox9fl/fl; Amh:creTg/+ and Sox8-/-; Sox9fl/fl; Wt1-CreERT2/+ double mutant mice where Sox9 is inactivated after sex determination , the single and double mutant mice initially form testis cords containing Sertoli cells . However , these cells then lose their identity and begin to express granulosa cell markers like FOXL2 ( Barrionuevo et al . , 2009; Barrionuevo et al . , 2016; Georg et al . , 2012 ) . In addition , following tamoxifen induction of Cre recombinase and subsequent deletion of Sox9 , Sertoli cells in Sox8-/-; Sox9fl/fl; Wt1-CreERT2/+ testes become apoptotic leading to a complete degeneration of the seminiferous tubules . This indicated that a concerted effort by Sox8 and Sox9 is required in XY gonads for the maintenance of Sertoli cells after sex determination . Beyond mice , in humans , SOX8 contributes to testis differentiation or homeostasis , given the 46 , XY gonadal dysgenesis phenotype associated with mutations/rearrangements at the SOX8 locus ( Portnoi et al . , 2018 ) . Although Sox8 expression is dispensable for Sertoli cell differentiation in XY gonads , it may have a key role for testicular differentiation in XX sex reversal gonads or in cases of Sox9-independent testicular differentiation in XY gonads . This led us to hypothesize that Sox8 can compensate for loss of Sox9 and is the alternate factor capable of: ( i ) triggering sex reversal in XX Rspo1 knockout gonads lacking Sry and Sox9 , and ( ii ) promoting testicular development in XY Rspo1 knockout gonads lacking Sox9 . To test this hypothesis , we have generated triple Rspo1 , Sox8 , and Sox9 loss-of-function mutant mice models . We show here that Sox8 and Sox9 are individually dispensable for testicular development in XY and XX mice lacking Rspo1 , indicating the presence of redundant testicular pathways . In the absence of both Sox factors , Sertoli cell differentiation is precluded and XY and XX Rspo1-/-; Sox8-/-; Sox9fl/fl; Sf1:creTg/+ triple mutants develop atrophied ovaries . Together , our data show that Sox8 or Sox9 is required to induce testicular development in XY and XX mice lacking Rspo1 .
We first performed expression analyses for Rspo1 ( Figure 1A a-h ) , Sox8 ( Figure 1 Ba-l ) , and Sox9 ( Figure 1 Ca-l ) , in control and mutant gonads . We chose to study embryonic day 17 . 5 ( E17 . 5 ) fetal gonads , when testis cords form in Rspo1 sex reversal mice , and juvenile postnatal day 10 ( P10 ) gonads , when gonadal fate is likely to be set ( Lavery et al . , 2012 ) . In XY gonads , Rspo1 is mostly localized to the coelomic epithelium at E17 . 5 and to the tunica albuginea at P10 ( Figure 1 Aa , c ) . In fetal ovaries , Rspo1 is expressed in somatic cells at E17 . 5 and down-regulated after birth , as shown in post-natal P10 ovaries ( Figure 1 Af , h ) . In XY and XX mice lacking Sox8 and Sox9 ( i . e . , Sox8-/-; Sox9fl/fl; Sf1:creTg/+ , referred to as Sox8KO Sox9cKO double mutants ) , high Rspo1 expression levels were observed in embryonic gonads and down-regulated after birth , indicating ovarian differentiation ( Figure 1 Ab , d , e , g ) , as previously described ( Chaboissier et al . , 2004 ) . Together , these data confirmed that although Rspo1 is expressed in both XY and XX gonads , robust Rspo1 expression in cells throughout the gonad is a feature of ovarian development in fetuses . In XY control , XY Rspo1-/- ( referred to as Rspo1KO ) , and XY Rspo1-/-; Sox9fl/fl; Sf1:creTg/+ ( referred to as Rspo1KO Sox9cKO ) mice , immunostaining revealed SOX8 expression in Sertoli cells organized as testis cords at E17 . 5 and seminiferous tubules at P10 , in agreement with previous reports ( Figure 1 Ba , d , b , e , g , j; Schepers et al . , 2003; Lavery et al . , 2012 ) . In XX mice , though SOX8 is not expressed in control ovaries ( Figure 1 Bi , l ) , expression was observed in XX Rspo1KO and XX Rspo1KO Sox9cKO sex reversal gonads ( Figure 1 Bh , k , g , j ) . Co-immunolabeling with AMH confirmed the identity of Sertoli cells ( Figure 1 Ba-f , g-h , j-l ) and LAMA1 staining at P10 demarcated both testis cords ( Figure 1 Bd-f , j-k ) and follicles ( Figure 1l ) , which do not express SOX8 . In summary , these data corroborated that Sox8 is expressed in gonads lacking Rspo1 , and that its expression can be independent of Sox9 ( Lavery et al . , 2012 ) . Next , immunostaining revealed SOX9-positive testis cords in XY Rspo1KO testes ( Figure 1 Cb , e ) , XX Rspo1KO ovotestes ( Figure 1 Ch , k ) , as in control testes ( Figure 1 Ca , d ) , as previously described ( Chassot et al . , 2008 ) . Co-immunolabeling with AMH confirmed the identity of Sertoli cells , since AMH-positive granulosa cells do not express Sox9 , and given that ovarian steroidogenic theca cells expressing Sox9 are AMH-negative ( Figure 1Cl ) . In addition , deletion of Sox8 did not alter the expression of Sox9 in XY or XX Rspo1KO gonads ( i . e . , in Rspo1KO Sox8KO gonads ) ( Figure 1Cc , f , g , j ) . Altogether , our results show that Sox8 and Sox9 are expressed in the absence of each other in Rspo1 mutant gonads when testis cords are present or when partial sex reversal occurs . Next , we asked how inactivation of both Rspo1 and Sox8 would impact gonad development in XY and XX Rspo1KO Sox8KO double mutants by comparison with controls ( Figure 2a–y and Figure 2—figure supplement 1a–h ) . In XY Rspo1KO Sox8KO mice , the anogenital distance in adult P40 animals was comparable to XY control males ( Figure 2a , b ) . In contrast , XX control females exhibited a short anogenital distance ( Figure 2m ) . Internally , XY Rspo1KO Sox8KO mice developed epididymides , vasa deferensia , seminal vesicles and prostate , as in control males ( Figure 2—figure supplement 1a , b ) . Histological analyses by PAS staining revealed seminiferous tubules with no obvious defects in P10 and P40 XY Rspo1KO Sox8KO animals ( Figure 2—figure supplement 1c , d and Figure 2e , f ) , and these mice were fertile . Testicular development in XY Rspo1KO Sox8KO mice was confirmed by immunostaining experiments on embryonic ( E17 . 5 ) and post-natal ( P10 , and P40 ) gonads that contained SOX9 and DMRT1 positive Sertoli cells forming testicular sex cords and seminiferous tubules ( Figure 1Cc , f , Figure 2g–j , and Figure 2—figure supplement 1e–h ) . DMRT1 expression was also observed in germ cells , which are TRA98-positive ( Figure 2i , j and Figure 2—figure supplement 1g , h; Matson et al . , 2010 ) . Thus , loss of both Rspo1 and Sox8 does not impair testis differentiation . For XX Rspo1KO Sox8KO mice , the question is whether the double mutant gonads developed as ovaries or as ovotestes , as in XX Rspo1KO single mutant ( Figure 2k–y and Figure 2—figure supplement 1i–t ) and as in XX Rspo1KO Sox9cKO double mutant mice ( Lavery et al . , 2012 ) . Externally , as in XX control mice , both XX Rspo1KO and XX Rspo1KO Sox8KO mice developed a short anogenital distance , as shown in adult P40 animals ( Figure 2k–m ) . Internally , although XX Rspo1KO Sox8KO mice exhibited rare testis cords during embryonic development ( Figure 1Cg ) , seminiferous tubules devoid of germ cells were apparent at P10 , suggesting a delay in ovo-testicular development in double mutant gonads ( Figure 2—figure supplement 1l , m ) . Indeed , by P40 , both XX Rspo1KO and XX Rspo1KO Sox8KO mice were essentially indistinguishable with respect to gonad morphology ( Figure 2n , o ) , reproductive tract development ( Figure 2—figure supplement 1i , j ) , ovo-testicular organization ( Figure 2q , r ) , and the presence of SOX9 and DMRT1 positive Sertoli cells in the testicular area ( Figure 2t , u , w , x ) . Altogether , studies performed in Rspo1KO Sox8KO mice demonstrate that like Sox9 ( Lavery et al . , 2012 ) , Sox8 is dispensable for testicular development in XY and XX Rspo1KO gonads . Moreover , our data suggests that SOX9 likely compensates for the loss of Sox8 in Rspo1KO Sox8KO double mutants . Our genetic mouse models allowed us to investigate gonadal fate in XY and XX Rspo1KO mice lacking both Sox8 and Sox9 ( i . e . , in XY and XX Rspo1KO Sox8KO Sox9cKO triple mutant mice ) . We first studied gonads in E17 . 5 fetuses ( Figure 3a–j'' and Figure 3—figure supplement 1a–j' ) , which is when differentiated granulosa cells reprogram as Sertoli cells in XX Rspo1KO gonads ( Maatouk et al . , 2013 ) . As shown , XX control gonads contained granulosa cells expressing FOXL2 , but not Sertoli cells expressing SOX9 or DMRT1 ( Figure 3f and Figure 3—figure supplement 1a ) , indicating ovarian development . The granulosa cells remained quiescent , as evidenced by expression of the mitotic arrest marker CDKN1B ( also known as P27 ) throughout the E17 . 5 gonad , and the absence of AMH expression indicated that these cells were fetal or pre-granulosa cells ( Figure 3a; Maatouk et al . , 2013 ) . In contrast , CDKN1B is down-regulated in the anterior area of XX Rspo1KO Sox8KO Sox9cKO triple mutant gonads ( n = 4 triple mutant fetuses , Figure 3e ) , as in XX Rspo1KO single , as well as in XX Rspo1KO Sox8KO and XX Rspo1KO Sox9cKO double mutants ( Figure 3b , c , d; Maatouk et al . , 2013 ) . In addition , these mutants contained cells expressing AMH ( Figure 3b'–e' asterisks ) , indicating precocious granulosa cell differentiation , as previously described ( Maatouk et al . , 2013 ) . However , while SOX9 and DMRT1 positive , TRA98-negative Sertoli cells were readily detectable in the anterior area of the XX Rspo1KO gonads ( Figure 3—figure supplement 1b' and Figure 3g' , white arrowheads ) , these cells were noticeably absent or rare in XX Rspo1KO Sox8KO Sox9cKO triple mutant fetuses ( 1 out of 8 XX triple mutant gonads studied from n = 4 fetuses ) ( Figure 3—figure supplement 1e and Figure 3j ) . This was also the case in XX Rspo1KO Sox8KO and XX Rspo1KO Sox9cKO double mutants ( Figure 3—figure supplement 1c , d and Figure 3h , i ) . Together with these observations , quantification of immunostained cells expressing DMRT1 , FOXL2 , and CDKN1B per gonadal section area demarcated by DAPI ( Figure 3—figure supplement 2a–f ) confirmed the lack of Sertoli cells and presence of granulosa cells in XX double and triple mutant gonads at E17 . 5 ( Figure 3—figure supplement 2a , c , e ) . In addition to the presence of mature granulosa cells , gonads in the XX single , double , and triple mutant fetuses also exhibited NR5A1- and HSD3β-positive cells ( Figure 3—figure supplement 1g–j ) , which were absent in XX control ovaries ( Figure 3—figure supplement 1f; Chassot et al . , 2008; Lavery et al . , 2012 ) . Thus , these data indicated that ablation of Sox8 and/or Sox9 in XX fetuses lacking Rspo1 does not prevent the appearance of steroidogenic cells and precocious granulosa differentiation , two characteristics of XX Rspo1KO gonads ( Maatouk et al . , 2013; Chassot et al . , 2008 ) . We then examined the phenotype of Rspo1KO Sox8KO Sox9cKO gonads in E17 . 5 XY fetuses ( Figure 4a–h''' and Figure 4—figure supplement 1a–h' ) . As shown , XY Rspo1KO Sox8KO double mutant gonads contained SOX9 and DMRT1 positive Sertoli cells forming testis cords , as in control fetal testes ( Figure 4e , f and Figure 4—figure supplement 1a , b ) . Also , XY Rspo1KO Sox9cKO gonads exhibited DMRT1-positive testis cords ( Figure 4g , g' ) , which were more pronounced than testis cords in XX Rspo1KO and XX Rspo1KO Sox9cKO gonads at this stage ( Figure 3g , g' , i , i'; Lavery et al . , 2012 ) . Thus , in XY fetuses lacking Rspo1 , inactivation of one Sox gene is dispensable for Sertoli cells . However , in fetuses lacking both Sox8 and Sox9 in XY triple mutant gonads , Sertoli cells expressing DMRT1 were not readily obvious ( 6 of 6 XY triple mutant gonads studied from n = 3 fetuses ) ( Figure 4h and Figure 4—figure supplement 1d ) . Instead , as in XY Rspo1KO Sox9cKO gonads at this stage , XY triple mutant gonads exhibited FOXL2-positive pre-granulosa cells ( Figure 4h , h' , h''' , yellow arrowheads ) , and AMH expression suggested that some mature granulosa cells were present ( Figure 4d' , asterisk ) . Quantification of cells expressing DMRT1 , FOXL2 , and CDKN1B confirmed these observations ( Figure 3—figure supplement 2b , d , f ) . Like XX triple mutants , XY triple mutants also contained steroidogenic cells expressing NR5A1 and HSD3β ( Figure 4—figure supplement 1h ) . Altogether , fetal XY and XX Rspo1KO Sox8KO Sox9cKO gonads resembled gonads from XX Rspo1KO Sox8KO , as well as XY and XX Rspo1KO Sox9cKO fetuses , with respect to the presence of steroidogenic cells and mature granulosa cells . However , fetal triple mutant gonads lacked Sertoli cells that were present in fetal ( Figure 4f , g ) or post-natal ( Figure 1Bf , j , Cf , j ) double mutant mice . Thus , while pre-granulosa cells in triple mutants differentiated precociously , their reprogramming as Sertoli cells forming testis cords at E17 . 5 appears to be blocked , or delayed . In order to further address the development of triple mutant gonads , we extended our analyses to juvenile ( P10 ) and adult ( P40 ) mice ( Figure 5a–d'; Figure 5—figure supplement 1a–x; Figure 5—figure supplement 2a–j and Figure 5—figure supplement 3a–o''' ) . Both XY and XX Rspo1KO Sox8KO Sox9cKO triple mutant mice developed externally as female with a short anogenital distance , as in XX control mice ( Figure 5c , p , r ) . Internally , both XY and XX triple mutants displayed hermaphroditism of the reproductive tracts , as shown by concomitant presence of vasa deferensia and uteri ( Figure 5—figure supplement 1c , m ) . This was also observed in XY and XX Rspo1KO Sox9cKO mice ( Figure 5—figure supplement 1b , n ) , as well as in XX Rspo1KO and in XX Rspo1KO Sox8KO mice ( Figure 2—figure supplement 1i , j ) . Histological analyses revealed that XY and XX triple mutant gonads developed as ovaries containing primary follicles at P10 ( Figure 5—figure supplement 1f , p ) , which matured up to the antral follicle stage at P40 , though some exhibited irregular granulosa cell organization ( Figure 5i , v , blue arrowheads ) . The triple mutant gonads occasionally contained immature or atrophied follicles ( Figure 5—figure supplement 2a , b ) . Both XY and XX Rspo1KO Sox8KO Sox9cKO gonads lacked testicular sex cords ( Figure 5i , v , Figure 5—figure supplement 1f , p , and Figure 5—figure supplement 2a , b ) , which were found in XY and XX Rspo1KO mice lacking Sox8 ( Figure 2f , q and Figure 2—figure supplement 1d , l ) or Sox9 ( Figure 5h , w and Figure 5—figure supplement 1e , q ) . Immunostaining experiments on P10 and P40 triple mutant gonads confirmed the presence of follicles with granulosa cells expressing FOXL2 and the absence of testis cords with Sertoli cells expressing DMRT1 ( Figure 5l , o , y , b’ and Figure 5—figure supplement 1i , l , s , v ) . In 3 of 10 XY and 6 of 16 XX post-natal gonads studied , a cluster of cells expressing DMRT1 were found , but further analyses revealed that these cells did not express the mature Sertoli cell marker GATA1 ( Beau et al . , 2000; Figure 5—figure supplement 3f , f' , f'' , l , l' , l'' , o , o' , o'' ) . Instead , these cells expressed the embryonic supporting cell marker GATA4 ( Tevosian et al . , 2002 ) , which suggests rudimentary testis cord formation ( Figure 5—figure supplement 3c , c' , c'' , i , i' , i'' , asterisks ) . We also noticed some cells expressing DMRT1 and FOXL2 , though these cells were rare ( Figure 5—figure supplement 3l'' , l''' , arrowheads ) . In fact , immunostaining for FOXL2 confirmed that the vast majority of the supporting cells in triple mutants were granulosa cells , which did not undergone reprogramming into Sertoli cells ( Figure 5l , o , y , b' and Figure 5—figure supplement 1i , l , s , v ) . While observing atrophied follicles in adult XY and XX Rspo1KO Sox8KO Sox9cKO triple mutant mice , a distinct interstitial compartment was also apparent ( Figure 5i , v , asterisks and Figure 5—figure supplement 2a , b ) . The identity of this compartment was confirmed by immunostaining for NR5A1 and HSD3β ( Figure 5—figure supplement 2g , h ) . In triple mutant gonads , the interstitial cells were arranged individually or in small clusters when compared with XX control ovaries and XX Rspo1KO ovotestes . In addition , XY and XX triple mutant interstitial cells mildly atrophied , appeared collapsed/dysplastic , and lacked interstitial sinusoids ( Figure 5—figure supplement 2e , f ) . No evidence of neoplasia was present in XY and XX triple mutant and in XX Rspo1KO gonads . In summary , gonads in XY and XX triple mutants developed as atrophied ovaries . Altogether , our data clearly demonstrate that Sox8 or Sox9 is required and sufficient for testicular differentiation in XY and XX Rspo1KO Sox9cKO or Rspo1KO Sox8KO double mutants , respectively .
Our results emphasize the essential role of SOX genes in testis differentiation as we show that Sox genes are required for Sertoli cell differentiation in XX ovotestis . The critical domain of SOX proteins is the DNA binding domain , the HMG ( High-Mobility Group ) -domain that binds in a sequence-specific manner ( Mertin et al . , 1999 ) . Remarkably , an HMG-box gene is associated with male sex-specific region in the brown algae Ectocarpus ( Ahmed et al . , 2014 ) . The sexual cycle of this species consists of an alternation between a diploid sporophyte ( with both the U and the V chromosomes ) , which after meiosis produces either a female haploid gametophyte ( with the U chromosome ) or male gametophyte ( with the V chromosome ) . The sex-specific region of the Ectocarpus V-chromosome contains an HMG-domain gene , suggesting a conserved function of the HMG-domain containing genes in maleness throughout evolution . In mice , when the HMG box of SRY is replaced with that of SOX3 or SOX9 , these composite Sox transgenes induce Sox9 expression and Sertoli cell differentiation ( Bergstrom et al . , 2000 ) . Also , transgenic expression of Sox3 or Sox10 in XX gonads results in Sox9 expression and testicular differentiation ( Sutton et al . , 2011; Polanco et al . , 2010 ) . These examples demonstrated functional conservation among Sox genes or HMG-box domains and also suggests that male fate centers on transactivation of Sox9 . However , testicular differentiation was reported in XY and XX Rspo1/Ctnnb1 Sox9 double mutant mice ( Nicol and Yao , 2015; Lavery et al . , 2012 ) , suggesting that another Sox gene can substitute for the absence of Sox9 in this context . Given that Sox8 is up-regulated in the double mutant gonads ( Nicol and Yao , 2015; Lavery et al . , 2012 ) , we hypothesized that Sox8 and Sox9 can act redundantly for testicular development in mice lacking Rspo1 . Here , we demonstrated this by showing that in XY and XX Rspo1KO mice: ( i ) Sox8 and Sox9 are expressed independently; ( ii ) Sox8 or Sox9 is sufficient for Sertoli cell differentiation in Rspo1KO Sox9cKO and Rspo1KO Sox8KO mice , respectively; and ( iii ) Sox8 or Sox9 are required for testicular differentiation , as evidenced by the development of atrophied ovaries in Rspo1KO Sox8KO Sox9cKO triple mutant mice . Together our data show that Sox8 is able to substitute for Sox9 to induce Sertoli cell differentiation in XX sex reversal . The gonad fate in wildtype , Sox and Rspo1 mutant mice is summarized in Figure 6 . In wildtype mice , SOX9 promotes testicular differentiation in XY gonads and RSPO1 promotes ovarian differentiation in XX gonads ( Figure 6a ) . This is also the case in mice lacking Sox8 , since it is dispensable for testis and ovarian development ( Figure 6a; Sock et al . , 2001 ) . As shown , there is an antagonistic relationship between the testis and ovarian pathways , such that the activation of one pathway also leads to the repression of the other to ensure one gonadal fate ( Figure 6a ) . In XY Sox9cKO mice , the testis pathway is not activated , and the ovarian pathway is not repressed , leading to ovarian differentiation ( Figure 6b ) . In XX Sox9cKO mice , loss of SOX9 does not impair ovarian development ( Figure 6b ) . In XY Rspo1KO Sox8KO or XY Rspo1KO Sox9cKO mice , gonads develop as testes or hypo-plastic testes , since one SOX factor is sufficient for Sertoli cell differentiation and seminiferous tubule formation ( Figure 6c , d ) . This is also exemplified by ovo-testicular development in XX Rspo1KO Sox8KO and XX Rspo1KO Sox9cKO mice ( Figure 6c , d ) , where Sertoli cells arise from reprogramming of pre-granulosa cells that have precociously differentiated ( Maatouk et al . , 2013 ) . We found that inactivation of both SOX factors in mice lacking RSPO1 prevents testicular development in XY and XX animals . In XY and XX Rspo1KO Sox8KO Sox9cKO triple mutant embryos , though pre-granulosa cells differentiate precociously , the absence of both SOX factors impedes granulosa-to-Sertoli reprogramming in embryos and gonads develop as atrophied ovaries ( Figure 6e ) . This atrophied ovary outcome suggests that FOXL2 and other ovarian factors cannot fully compensate for the loss of RSPO1 ( Figure 6e ) . Interestingly , in XX Rspo1KO single mutant gonads and in XX Rspo1KO gonads lacking Sox8 and/or Sox9 in double and triple mutants , pre-granulosa cells differentiate as mature granulosa cells expressing AMH in an anterior-to-posterior wave or gradient ( [Maatouk et al . , 2013] and present data ) . Such a gradient was also found in XX gonads with an Sry transgene – supporting cells transiently express SOX9 , after which this ability is lost in an anterior-to-posterior wave ( Harikae et al . , 2013 ) . This suggests that somatic cell differentiation in ovaries proceeds in a spatiotemporal , anterior-to-posterior , manner . As shown here , apparently , this wave of somatic cell differentiation is conserved in XX sex reversal associated with Rspo1 mutations . How Sox8 operates in pathophysiological cases of testicular differentiation is not yet known . In wildtype mice , Sox8 expression in XY gonads has been described as coinciding with Sox9 at E11 . 5 ( Jameson et al . , 2012; Stévant et al . , 2018; Schepers et al . , 2003 ) or occurring after robust expression of Sox9 at E12 . 5 ( Schepers et al . , 2003 ) . Together , these observations suggest that SRY might activate Sox8 , as predicted ( Li et al . , 2014 ) , and that Sox8 expression is reinforced by Sox9 . However , the activation of Sox8 by SOX9 is likely indirect given that SOX9 does not bind the Sox8 locus in mice . Interestingly , SOX9 binding to Sox8 has been shown in cattle ( Rahmoun et al . , 2017 ) . Also , the expression of Sox8 and Sox9 are independent in sex cords in XY mouse gonads ( ( Barrionuevo et al . , 2009 ) , our present results ) . Thus , it is more plausible that SRY activates Sox8 expression in XY Rspo1/Ctnnb1 Sox9 double mutant mice . In fact , in these mice , Sry expression is extended beyond E12 . 5 ( Lavery et al . , 2012; Nicol and Yao , 2015 ) , a time when Sry is normally down-regulated in mice ( Hacker et al . , 1995 ) . Whereas Sox8 expression can result from SRY activation in XY embryos , it is not obvious how Sox8 is upregulated in the absence of Sry in XX Rspo1/Ctnnb1 Sox9 double mutants . The pro-ovarian factors Rspo1 or Ctnnb1 are required on one hand , to prevent precocious maturation of granulosa cells , which are capable of transdifferentiation into Sertoli cells . On the other hand , these factors repress ectopic steroidogenesis in XX gonads as evidenced by the presence of steroidogenic cells in XX Rspo1 or Rspo1KO Sox9cKO embryonic gonads ( Chassot et al . , 2008; Lavery et al . , 2012 ) . When E13 . 5 ovaries are transplanted to kidneys of XY mice , circulating androgens promote partial trans-differentiation towards a testis fate through a mechanism involving up-regulation of Sox8 before Sox9 ( Miura et al . , 2019 ) . Ablation of Sox8 in the transplanted ovaries did not prevent sex reversal and this outcome is likely attributed to the presence of Sox9 . Furthermore , before up-regulation of Sox8 , supporting cells in the fetal ovary transplant express Amh , a phenotype that is strikingly similar to sex reversal in XX Rspo1KO gonads ( Maatouk et al . , 2013 ) . However , inactivation of Amh in the transplanted ovaries was also dispensable for sex reversal , suggesting that TGF-β signaling driven by other TGF-β factors like Activin or unknown factors may promote Sox gene expression in sex reversal conditions . It is noteworthy that WNT/CTNNB1 signaling regulates the level of ActivinB as evidenced by its up-regulation in XX Wnt4KO or Ctnnb1cKO gonads ( Yao et al . , 2006; Liu et al . , 2010 ) . Moreover ablation of InhibinA/B , two antagonist members of Activin , promotes sex cord development in XX gonads ( Matzuk et al . , 1992 ) . Thus , TGF-β signaling is likely involved in XX sex reversal when the WNT/CTNNB1 pathway is compromised . The factors including Activin/Inhibin to control Sox gene expression in XX transplanted ovaries and in XX mice lacking Rspo1/Ctnnb1 remain to be identified . The identification of Sox8 as a key factor in pathophysiological testicular development is somewhat of a paradox , given evidence indicating that aside from Sry and Sox9 , no other Sox gene tested so far play key roles in Sertoli cell differentiation in XY wildtype gonads ( She and Yang , 2017 ) . In mice , Sox8 is dispensable for Sertoli cell differentiation ( Sock et al . , 2001 ) , but an inefficient or late deletion of Sox9 leads to XY sex reversal only if there is additional deletion of Sox8 ( Lavery et al . , 2011; Chaboissier et al . , 2004; Barrionuevo et al . , 2009 ) . This suggests that Sox8 reinforces Sox9 during testis differentiation . In addition , Sox8 is required for Sertoli cell maintenance along with Sox9 , since Sertoli cells in XY Sox8 Sox9 double loss-of-function gonads undergo apoptosis ( Barrionuevo et al . , 2016 ) . Thus , although Sox8 is an important factor for testis differentiation and maintenance , the rapid and high level of expression of Sox9 induced by SRY , minimizes the role of Sox8 in XY differentiating testes . During XX sex reversal , early and high induction of Sox8/9 does not occur , given the absence of Sry . Hence , ovarian differentiation is initiated , but absence of pro-ovarian gene such as Rspo1 leads to a succession of events from ectopic steroidogenesis to accelerated maturation of granulosa cells that ultimately promote the expression of Sox8 and Sox9 . In XX Rspo1KO gonads , both factors are similarly important and can compensate for the absence of the other to induce transdifferentiation of granulosa cells to Sertoli cells during late embryogenesis . Thus , although Sox8 is dispensable for testicular differentiation in wildtype mice , our current study demonstrates that Sox8 is essential for testicular differentiation in sex reversal conditions . Functional redundancy between SOX8 and SOX9 does not seem to operate in humans . For example , XY sex reversal can result from inactivating mutations of one SOX9 allele , indicating haploinsufficiency ( Wagner et al . , 1994; Foster et al . , 1994 ) . Also , SOX8 mutations were associated with a range of phenotypes including complete gonadal dysgenesis ( streak gonads with immature female genitalia ) and hypoplastic testes in three 46 , XY patients ( Portnoi et al . , 2018 ) . Thus , it appears that the impact of a single gene mutation can vary , according to the nature of the mutation and genetic background of the individual . Nevertheless , the human cases of XY sex reversal show that SOX8 is emerging to be an important regulator of testicular gonadal development and by extension , overall male development .
The experiments described here were carried out in compliance with the relevant institutional and French animal welfare laws , guidelines , and policies . These procedures were approved by the French ethics committee ( Comité Institutionnel d’Ethique Pour l’Animal de Laboratoire; number NCE/2011–12 ) . All mouse lines were kept on a mixed 129Sv/C57BL6/J background . Rspo1-/- ( Chassot et al . , 2008 ) , Sox8-/- ( Sock et al . , 2001 ) , Sox9fl/fl ( Akiyama et al . , 2002 ) , and Sf1:creTg/+ ( Bingham et al . , 2006 ) mice were obtained previously , and the generation of Sox9fl/fl; Sf1:creTg/+ ( Lavery et al . , 2011 ) and Rspo1-/-; Sox9fl/fl; Sf1:creTg/+ ( Lavery et al . , 2012 ) mice was described previously . For Rspo1KO Sox8KO mice: Rspo1-/- males were mated with Sox8-/- females to obtain Rspo1+/-; Sox8+/- males and females . Matings between these littermates allowed us to obtain Rspo1-/-; Sox8-/- double mutant mice , referred to as Rspo1KO Sox8KO mice , and control animals . For Rspo1KO Sox8KO Sox9cKO mice: first , Rspo1-/-; Sox8-/- males were mated with Sox8-/-; Sox9fl/fl; Sf1:creTg/+ females to generate Rspo1+/-; Sox8-/-; Sox9fl/+ males and Rspo1+/-; Sox8-/-; Sox9fl/+; Sf1:creTg/+ females . Matings between these littermates then produced Rspo1-/-; Sox8-/-; Sox9fl/fl males and Rspo1+/-; Sox8-/-; Sox9fl/fl; Sf1:creTg/+ females . Finally , matings between these littermates then allowed us to obtain Rspo1-/-; Sox8-/-; Sox9fl/fl; Sf1:creTg/+ triple mutant mice , referred to as Rspo1KO Sox8KO Sox9cKO mice , and control animals . Embryos were collected from timed evening matings that was confirmed by the presence of a vaginal plug the following morning . This marked embryonic day 0 . 5 ( E0 . 5 ) . The day of delivery was defined as post-natal day 0 ( P0 ) . Genotyping was performed as described in Chaboissier et al . ( 2004 ) ; Chassot et al . ( 2008 ) ; Bingham et al . ( 2006 ) by using DNA extracted from tail tip or ear biopsies of mice . The presence of the Y chromosome was determined , as described previously ( Hogan et al . , 1994 ) . Gonad samples were fixed with 4% paraformaldehyde overnight , processed for paraffin embedding , and then sectioned at 5–7 μm thick . The in situ hybridizations for Figure 1e–h were carried out essentially as described by Lavery et al . ( 2012 ) . For analyses in Figure 1a–d , RNAscope technology was used ( Wang et al . , 2012 ) . The Rspo1 probe was purchased from the manufacturer ( Advanced Cell Diagnostics ) and the protocol was performed according to the manufacturer’s instructions using the Fast Red dye , which can be visualized using light or fluorescence microscopy . The in situ hybridization experiments were performed on gonads from at least three mice for each genotype . Gonad samples were fixed with 4% paraformaldehyde overnight , processed for paraffin embedding , and sectioned at 5 μm thick . The following dilutions of primary antibodies were used: AMH/MIS ( c-20 , sc-6886 , Santa Cruz ) , 1:200; DMRT1 ( HPA027850 , Sigma ) , 1:100; FOXL2 ( NB100-1277 , Novus ) , 1:200; GATA1 ( N6 , sc-265 , Santa Cruz ) , 1:200; GATA4 ( C20 , sc-1237 , Santa Cruz ) , 1:200; 3βHSD ( P18 , sc-30820 , Santa Cruz ) , 1:200; P27 ( Kip1 , sc-528 , Santa Cruz ) , 1:200; LAMA1 ( L9393 , Sigma ) , 1:150; SF1 ( kindly provided by Ken Morohashi ) , 1:1000; SOX8 ( kindly provided by Elisabeth Sock [Stolt et al . , 2005] ) , 1:1000; SOX9 ( HPA001758 , Sigma ) , 1:200; and TRA98 ( ab82527 , Abcam ) , 1:200 . Counterstain with DAPI was used to detect nuclei . Immunofluorescence of secondary antibodies were detected with an Axio ImagerZ1 microscope ( Zeiss ) coupled to an Axiocam mrm camera ( Zeiss ) or a LSM 780 NLO inverted Axio Observer . Z1 confocal microscope ( Carl Zeiss Microscopy GmbH , Jena , Germany ) using a Plan Apo 10X dry NA 0 . 45 objective . Images were processed with Axiovision LE and Serif Affinity Photo software . Immunostaining experiments were performed on gonads from at least three mice for each genotype . Immunostaining analyses were performed , as described above . For analyses at E17 . 5 , immunostaining were performed on 2 to 17 sections spaced 20–30 μm apart in each gonad . Then , for each section , the ratio of cells positive for DMRT1 , FOXL2 , or CDKN1B to total gonad area , as visualized by DAPI staining , were manually tabulated . Next , the individual ratios and mean for each genotype were plotted in a histogram using Graphpad software . Finally , the data was analysed by one-way ANOVA and Tukey-Kramer post tests . For p-values<0 . 05 , <0 . 01 , <0 . 001 , and <0 . 0001 , asterisks ( * , ** , *** , **** ) represent significant differences compared with XY control cell numbers , respectively and ampersands ( & , && , &&& , &&&& ) represent significant differences compared with XX control cell numbers , respectively . Gonad samples were fixed with Bouin’s solution overnight , processed for paraffin embedding , sectioned at 5 μm thick , and then stained according to standard procedures for periodic acid Schiff ( PAS ) or hematoxylin and eosin ( H and E ) staining . Images were taken with an Axiocam mrm camera ( Zeiss ) and processed with Serif Affinity Photo software . Histology staining was performed on gonads from at least three mice for each genotype . | In humans , mice and other mammals , genetic sex is determined by the combination of sex chromosomes that each individual inherits . Individuals with two X chromosomes ( XX ) are said to be chromosomally female , while individuals with one X and one Y chromosome ( XY ) are chromosomally males . One of the major differences between XX and XY individuals is that they have different types of gonads ( the organs that make egg cells or sperm ) . In mice , for example , before males are born , a gene called Sox9 triggers a cascade of events that result in the gonads developing into testes . In females , on the other hand , another gene called Rspo1 stimulates the gonads to develop into ovaries . Loss of Sox9 in XY embryos , or Rspo1 in XX embryos , leads to mice developing physical characteristics that do not match their genetic sex , a phenomenon known as sex reversal . For example , in XX female mice lacking Rspo1 , cells in the gonads reprogram into testis cells known as Sertoli cells just before birth and form male structures known as testis cords . The gonads of female mice missing both Sox9 and Rspo1 ( referred to as “double mutants” ) also develop Sertoli cells and testis cords , suggesting another gene may compensate for the loss of Sox9 . Previous studies suggest that a gene known as Sox8 , which is closely related to Sox9 , may be able to drive sex reversal in female mice . However , it was not clear whether Sox8 is able to stimulate testis to form in female mice in the absence of Sox9 . To address this question , Richardson et al . studied mutant female mice lacking Rspo1 , Sox8 and Sox9 , known as “triple mutants” . Just before birth , the gonads in the triple mutant mice showed some characteristics of sex reversal but lacked the Sertoli cells found in the double mutant mice . After the mice were born , the gonads of the triple mutant mice developed as rudimentary ovaries without testis cords , unlike the more testis-like gonads found in the double mutant mice . The findings of Richardson et al . show that Sox8 is able to trigger sex reversal in female mice in the absence of Rspo1 and Sox9 . Differences in sexual development in humans affect the appearance of individuals and often cause infertility . Identifying Sox8 and other similar genes in mice may one day help to diagnose people with such conditions and lead to the development of new therapies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2020 | Sox8 and Sox9 act redundantly for ovarian-to-testicular fate reprogramming in the absence of R-spondin1 in mouse sex reversals |
A hallmark of adult hematopoiesis is the continuous replacement of blood cells with limited lifespans . While active hematopoietic stem cell ( HSC ) contribution to multilineage hematopoiesis is the foundation of clinical HSC transplantation , recent reports have questioned the physiological contribution of HSCs to normal/steady-state adult hematopoiesis . Here , we use inducible lineage tracing from genetically marked adult HSCs and reveal robust HSC-derived multilineage hematopoiesis . This commences via defined progenitor cells , but varies substantially in between different hematopoietic lineages . By contrast , adult HSC contribution to hematopoietic cells with proposed fetal origins is neglible . Finally , we establish that the HSC contribution to multilineage hematopoiesis declines with increasing age . Therefore , while HSCs are active contributors to native adult hematopoiesis , it appears that the numerical increase of HSCs is a physiologically relevant compensatory mechanism to account for their reduced differentiation capacity with age .
HSC-derived hematopoiesis has usually been studied in the setting of transplantation ( Benz et al . , 2012; Biasco et al . , 2016; Dykstra et al . , 2007; Lu et al . , 2011; Rundberg Nilsson et al . , 2015; Wu et al . , 2014 ) , an experimental paradigm that has been the foundation of hematopoietic research for decades ( Siminovitch et al . , 1963; Till and McCulloch , 1961 ) and which has established hallmark properties of HSCs such as multi-potency and self-renewal . However , while the transplantation assay has provided key insights , not the least with relevance for the clinical use in bone marrow ( BM ) transplantation , it might not accurately reflect the contribution of HSCs to ongoing and unperturbed steady state hematopoiesis . This is because transplantation is conducted under highly non-physiological conditions wherein HSCs are forced to proliferate to rebuild an entire hematopoietic hierarchy in a myeloablated bone marrow micro-environment . Therefore , there is a need to approach HSC biology also in more unperturbed settings . While the overall structure of hematopoiesis is rather well established ( Bryder et al . , 2006 ) , the degree by which HSCs contribute to adult hematopoiesis in the steady state is more unclear . This includes whether the proposed differentiation routes for the hematopoietic lineages are obligatory , or whether alternative/complementary pathways exist . Furthermore , cells of the different hematopoietic lineages have not only distinct homeostatic functions and maintenance mechanisms ( Bando and Colonna , 2016; Dzierzak and Philipsen , 2013; Rodvien and Mielke , 1976 ) but also display dramatically different lifespans ( Galli et al . , 2011; Harker et al . , 2000; Van Putten , 1958; Westera et al . , 2013 ) . As a consequence , the rates by which separate adult-derived blood cell lineages must be replenished differ substantially . At the extreme end , certain hematopoietic cell types generated during the fetal period appear devoid of replenishment from adult progenitors , and rather rely on homeostatic proliferation for their maintenance ( Ginhoux and Guilliams , 2016; Kantor et al . , 1995 ) . Recent developments of transgenic mouse models that allow for identification ( Acar et al . , 2015; Chen et al . , 2016; Gazit et al . , 2014 ) and evaluation of HSCs biology have facilitated studies of native in vivo hematopoiesis ( Busch et al . , 2015; Sawai et al . , 2016; Sun et al . , 2014; Wilson et al . , 2008 ) . Using one such model , we recently revealed that most adult HSCs are highly quiescent , which is strikingly different in the transplantation scenario ( Säwén et al . , 2016 ) . Other models have been used for lineage tracing from HSCs ( Busch et al . , 2015; Sawai et al . , 2016; Sun et al . , 2014 ) . In one of these , lineage tracing was conducted via random genetic integration of an inducible transposable genetic element , leading to the proposition that native hematopoiesis involves a large number of actively contributing progenitor cell clones , which are only rarely shared among hematopoietic lineages ( Sun et al . , 2014 ) . More common approaches for lineage tracing involve the use of cell type specific recombinases , that function to irreversibly mark a cell of interest and with time its descendants . While elegant and extensively used among developmental biologists , such approaches have only sparsely been applied to adult HSCs , and with seemingly contradictory results . Using a Tie2-driven model , Busch et al . concluded a substantial hematopoietic contribution/maintenance from progenitors rather than HSCs ( Busch et al . , 2015 ) , which at least to some extent would appear compatible with the results from Sun et al . ( 2014 ) . By contrast , Sawai et al . utilized a Pdzk1ip1-based CreERT2 system and suggested robust HSC labeling and hematopoiesis from adult HSCs ( Sawai et al . , 2016 ) . To try to assess these potential ambiguities , we here investigated the degree to which HSCs contribute to steady state adult hematopoiesis by using an inducible Fgd5-based HSC lineage tracing model ( Gazit et al . , 2014 ) . We observed dramatic differences with regards to HSC contribution to adaptive immunity ( slow ) and the myeloerythroid lineages ( fast ) , with HSCs contributing to the platelet lineage with the most rapid kinetics . The regeneration of terminal cell fates was closely mirrored at the level of each intermediate myeloerythroid precursor . These findings are consistent with adult HSCs as highly active contributors to multilineage hematopoiesis not only following transplantation , but also during the steady state . However , when approached in the situation of chronological aging , we noted diminished mature blood cell output from aged HSCs that could be traced to the first differentiation events from HSCs . These results suggest that the previously proposed fetal to adult switch ( Bowie et al . , 2007 ) , in which HSCs alter their properties from more excessive proliferation/differatiation to a more dormant state in the adult , extends gradually throughout adulthood . As a consequence , the well-known numerical increase of HSCs with age ( Morrison et al . , 1996; Rossi et al . , 2005; Sudo et al . , 2000 ) appears to represent a physiologically relevant mechanism to account for reduced HSC differentiation with age .
Using a transcriptome based screen of more than 40 different hematopoietic cell types , Fgd5 ( FYVE , RhoGEF and PH domain containing 5 ) was identified as a HSC-expressed gene that is rapidly downregulated upon differentiation . That Fgd5 expression marks all HSCs was confirmed through functional studies using an Fgd5 knock-in reporter strain ( Gazit et al . , 2014 ) . To further detail the HSC specificity of Fgd5 , we first acquired transcriptome data from 11 , 581 individual lineage-marker negative , c-kit positive and CD45 positive bone marrow cells ( Lin-kit+ ) . The Lin-kit+ population contains a range of different immature hematopoietic progenitor cells ( Pronk et al . , 2007 ) . Therefore , Lin-kit+ cells provided a benchmark to which other more defined/specific hematopoietic progenitor subsets could be compared . Next , we took advantage of an Fgd5 reporter strain in which a ZsGreen-2A-CreERT2 allele was knocked into the endogenous Fgd5 locus ( hereafter Fgd5CreERT2/+ mice ) ( Figure 1B ) ( Gazit et al . , 2014 ) . We sorted either Lin-kit+Fgd5+ cells ( Figure 1A middle; 793 cells , Fgd5+ ) , or Fgd5+ cells with a stringent Lin-kit+Sca-1+CD48-CD150+ HSC phenotype ( Figure 1A right , 519 cells , HSC-Fgd5+ ) . All Fgd5+ and HSC-Fgd5+ data were aggregated with the Lin-kit+ transcriptome data , which was followed by identification of the most significant gene vectors using principal component analysis ( PCA ) . Data was then visualized using t-distributed stochastic neighbor embedding ( tSNE ) dimensionality reduction ( Figure 1A ) . Lin-kit+ cells were extensively scattered across the two dimensions ( Figure 1A , left ) , in agreement with the heterogeneity of these cells . By contrast , Fgd5+ cells , regardless if sorted based on additional HSC markers , formed a distinct and highly overlapping cluster ( Figure 1A , middle and right ) . This cluster localized to a region with very few cells when evaluating Lin-kit+ cells ( Figure 1A , left , dotted area ) , emphasizing the HSC-specificity of the Fgd5 reporter and the low HSC frequency within the larger Lin-kit+ fraction . We next generated a lineage tracing model by crossing Fgd5CreERT2/+ mice to Rosa26-Lox-Stop-Lox-Tomato mice ( hereafter Rosa26lsl-Tomato/+ ) ( Figure 1B ) . In this model , HSCs can be identified based on ZsGreen expression , while Tamoxifen administration leads to irreversible and heritable Tomato labeling of HSCs and , over time , their offspring ( Figure 1C ) . To confirm the model , we evaluated Tomato label in HSC and BM progenitor cells 48 hr after a single injection ( 1x ) of Tamoxifen . This revealed labeling of a fraction of candidate HSCs , with virtually no labeling in other c-kit+ progenitor fractions ( Figure 1D and Figure 1—figure supplement 1 ) . This established HSC specific labeling and a relatively low differentiation rate of HSCs in steady state ( Säwén et al . , 2016; Wilson et al . , 2008 ) . To illustrate our ability to detect Tomato label in peripheral blood ( PB ) cells , we assessed Tomato expression in defined cell types from mice that had received Tamoxifen 8–48 weeks previously ( Figure 1D , lower right ) . Complementary to immunophenotypic identification of initially labeled BM cells as HSCs ( Figure 1D and data not shown ) , we evaluated the proliferation history of Tomato labeled HSPCs 5 days after a pulse of Tomato labeling by evaluation of transgenic H2B-mCherry label retention ( Figure 1E ) ( Säwén et al . , 2016 ) . Among HSCs , this revealed a strong correlation between a restricted proliferative history and Tomato labeling . Of note , a single dose of Tamoxifen was insufficient to label all candidate Fgd5-expressing HSCs ( Figure 1E and data not shown ) . Finally , to corroborate that Tomato labeled phenotypic HSCs are bona fide HSCs , we injected mice with Tamoxifen and isolated candidate Tomato positive and negative HSCs 48 hr later . Sorted cells were transplanted at limiting dilution ( 5 cells/mouse ) . This revealed long-term multilineage reconstitution in 5/8 recipients transplanted with Tomato+ HSCs ( Figure 1F ) . Encouraged by the highly specific HSC label observed after Tamoxifen administration to Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice ( Figure 1 ) , we next set out to perform label tracing studies of hematopoietic generation from HSCs . For this , we labeled cohorts of Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice with one injection of Tamoxifen and chased groups of mice for different periods of time up to 83 weeks after labeling . At end point analyses , the fraction of Tomato+ cells was determined in various hematopoietic compartments to assess the HSC contribution to progenitor pools and mature blood cell subsets ( Figure 1—figure supplement 1 ) . The frequencies of Tomato+ cells for each investigated subset were next related to the fraction of Tomato labeled HSCs ( Figure 2A , mean 13% ± 9% ) in individual mice ( Figure 2B , and Figure 2—figure supplement 1 ) . The fraction of labeled HSCs was generally higher in mice analyzed beyond 4 days of chase compared to mice analyzed after shorter chase periods . However , no further increase in HSC labeling was noted after longer periods of chase ( Figure 2A ) . First , we investigated Tomato label progression into the immature lineage negative , Sca-1 positive and c-kit positive ( LSK ) compartment , fractionated further using the Slam markers CD48 and CD150 ( Kiel et al . , 2005 ) ( Figure 2B and Figure 1—figure supplement 1 ) . We used this approach to identify HSCs ( LSKCD150+CD48- ) and different multipotent progenitor fractions ( MPPs: LSKCD150-CD48- , MPP2: LSKCD150+CD48+ , MPP3/4: LSKCD150-CD48+ ) . LSKCD150-CD48- MPPs are immature multipotent progenitors distinguished from HSCs by their limited self-renewal potential ( Kiel et al . , 2005; Kiel et al . , 2008; Ugale et al . , 2014 ) . Of the evaluated progenitor subsets in our work , this subset was generated from HSCs with the fastest kinetics , with near equilibrium to HSC label reached already by 4 weeks ( Figure 2B ) . MPP2 cells represent a rare subset of cells with more undefined lineage/developmental affiliations . This prompted us to first elucidate their developmental potential . First , we aimed to place these cells within a transcriptional framework established by other , more established , hematopoietic progenitors . For this , we obtained gene expression data from a panel of defined stem and progenitor cells and MPP2 cells using a multiplexed qRT-PCR approach for 48 genes , selected to include cell surface markers , cell cycle regulators and transcription factors associated with hematopoiesis ( Supplementary file 2 ) . Visualization of this data using PCA revealed that MPP2 cells clustered closely to Meg/E progenitors ( Figure 2C ) . Consistent with a close association to the Meg/E lineages , short-term ( 6 days ) culture experiments revealed a more robust generation of both megakaryocyte and erythroid containing colonies from MPP2s compared to other LSK subsets ( Figure 2—figure supplement 2 ) . When investigating Tomato label progression , MPP2 cells reached label equilibrium with HSCs after 32 weeks in 1x injected mice ( Figure 2B ) . MPP3/4 cells lack , for the most part , Meg/E lineage potential ( Adolfsson et al . , 2005; Arinobu et al . , 2007; Pietras et al . , 2015; Pronk et al . , 2007 ) . MPP3/4 cells acquired Tomato label with much slower kinetics compared to other LSK fractions ( Figure 2B ) . Of the distinct progenitor fractions within the Lin-kit+ fraction ( Figure 1—figure supplement 1A ) , megakaryocyte progenitors ( MkP ) acquired label with the fastest kinetics , reaching label equilibrium with HSCs after 32 weeks . Other myeloerythroid progenitors , including pre-megakaryocytic/erythroid ( preMeg/E ) , pre-colony forming unit-erythroid ( pre CFU-E ) and pre-granulocyte-macrophage ( preGM ) progenitors acquired Tomato label with very similar kinetics despite their distinct lineage affiliations , although they never quite reached an equilibrium with HSCs throughout the course of the experiments ( Figure 2B and Figure 2—figure supplement 1 ) . Mature effector cells represent the terminal progeny of HSCs . We observed distinct generation kinetics for different lineages ( Figure 2B ) . First , we made the general observation that myeloerythroid cells acquired label more rapidly than lymphoid cells . Among the myeloid subsets , platelets acquired Tomato label with the fastest kinetics , followed by granulocytes and erythrocytes . Among lymphoid cell types , NK cells displayed faster labeling kinetics followed by B cells . T cells showed the slowest labeling kinetics among lymphoid cells and CD4+ T cells acquired label faster than CD8+ T cells ( Figure 2B ) . Because the frequency of Tomato+ cells increased over time in all evaluated lineages , this data demonstrate a continuous contribution of HSCs to all hematopoietic lineages . While multiple studies have defined populations of hematopoietic progenitors that associate with distinct developmental and/or stages of differentiation ( Bryder et al . , 2006 ) , it is unknown whether such described progenitors are obligatory intermediates and/or their quantitative association relative to their anticipated mature offspring . Therefore , we interrogated the relationships between the rates of ( re ) generation of candidate committed myeloerythroid progenitors to those of their proposed mature cell lineage . At the earliest time points evaluated , we observed for all evaluated fractions a higher label in their corresponding progenitors ( Figure 2D ) . However , this was resolved during the course of the experiments and reached similar equilibrium ratios for all evaluated lineages , although the erythroid lineage displayed somewhat slower kinetics ( Figure 2D ) . Collectively , these experiments are in line with the view that progenitor generation precedes the generation of mature cells and that previously proposed progenitors appears to be , at least for the most part , obligatory intermediates . Hematopoiesis after transplantation of HSCs is fundamentally different from unperturbed hematopoiesis ( Busch et al . , 2015; Sun et al . , 2014 ) . However , to what extent the pre-conditioning regimen and co-transplantation of mature cells and progenitors influence on hematopoiesis from HSCs is less established . Therefore , we next transplanted wild type recipient mice on continuous Tamoxifen diet with purified Fgd5CreERT2/+; Rosa26lsl-Tomato/+ HSCs or WBM cells . Here , recipient mice were pre-conditioned by either lethal irradiation or antibody mediated CD45-depletion ( Palchaudhuri et al . , 2016 ) . Due to the HSC specificity of the model , this approach allowed us to monitor the kinetics of the HSC contribution to all lineages after transplantation and compare it to the HSC contribution in steady state ( Figure 2E ) . Compared to steady state , label progression in transplanted mice were faster ( Figure 2E ) . When label progression kinetics was compared between HSC and WBM transplanted animals , HSC transplantation resulted in faster label progression , especially into the B cell lineage ( Figure 2E ) . This likely reflects a significant contribution to the regeneration of the B cell lineage by co-transplanted long-lived B-lineage progenitors and mature cells after WBM transplantation . Comparison of label progression after WBM transplantation into irradiated or non-irradiated/antibody-mediated conditioned recipient mice revealed similar label progression kinetics into most mature lineages , with the exception of platelets that displayed a faster label progression in irradiated mice . This suggests that progenitors for platelets are more effectively ablated by irradiation than antibody-mediated pre-conditioning . While a labeling regimen of one Tamoxifen injection allows for accurate kinetic evaluations ( Figure 2A–B , D ) , this experimental strategy labels only a fraction of HSCs ( Figure 1E and Figure 2A ) and thus necessitates correlation of label in HSCs to other evaluated cell subsets ( Busch et al . , 2015 ) ( Figure 2B ) . If the original HSC label is low , this might as a consequence not allow for evaluation of the activity of the entire pool of HSCs . To explore whether we could label the HSC pool more extensively , Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice were fed Tamoxifen containing food pellets for 16 weeks . To rule out adverse effects of prolonged Tamoxifen treatment on HSC proliferation , these experiments were preceded by a control label retaining experiment using Col1a1tetO-H2B-mCherry/tetO-H2B-mCherry; ROSA26rtTA/rtT mice ( Säwén et al . , 2016 ) . Following H2B-mCherry induction with Doxycycline , mice were chased for 5 weeks in the presence or absence of Tamoxifen . Prolonged Tamoxifen treatment did not induce any additional proliferation within the HSC compartment , while more differentiated progenitors had readily proliferated in both settings ( Figure 3A ) . The 16 weeks labeling period was followed by an extensive ( up to 41 weeks ) chase period , during which mice received normal chow ( Figure 3B ) . This labeling strategy resulted in labeling of virtually all candidate HSCs ( Figure 3C ) . The blood of labeled mice was analyzed regularly to determine the fraction of Tomato+ cells in PB cell subsets ( Figure 1—figure supplement 1B ) . Similar to after 1x Tamoxifen labeling , we observed robust label progression into all PB cell subsets , with similar kinetics in between different lineages ( Figure 3B ) . However , a more complete HSC labeling resulted in a somewhat faster and more robust label progression into all PB cell subsets compared to 1x Tamoxifen labeling ( Figure 2E ) . This was most evident for the lymphoid lineages , where the majority of PB cells had been generated from HSCs at the experiment end point upon prolonged Tamoxifen administration , whereas the ratio of labeled lymphocytes vs . labeled HSCs was low ( >0 , 5 ) even after 83 weeks of chase in 1x Tamoxifen labeled mice ( Figure 2B , Figure 2E and Figure 3B ) . From endpoint mice in which the pool of HSCs was almost completely labeled ( Figure 3C ) , we next interrogated the skin epidermis for Tomato+ contribution to granulocytes and Langerhans cells . Granulocytes were almost completely Tomato positive , while Langerhans cells were devoid of label ( Figure 3D , lower left ) , in line with the fetal origin and self-maintenance of the latter cells ( Collin and Milne , 2016 ) . Next , we interrogated Tomato expression in Vγ3δ+ T cells , an established fetal derived T cell subset ( Havran and Allison , 1990 ) . To our surprise , this revealed robust Tomato labeling of Vγ3δ+ T cells ( Figure 3D , lower left ) . However , closer examination revealed high expression of ZsGreen in these cells ( Figure 3B , lower middle ) . Therefore , rather than establishing adult contribution to this lineage , these experiments established Fgd5-CreERT2 transgene expression in Vγ3δ+ T cells . B1a B cells represent an invariant subtype of B cells with a fetal origin that is primarily located in the peritoneum ( Hayakawa et al . , 1985 ) , where they co-exists with more traditional B1b and B2 B cells in adult mice . While less than 10% of B1a B cells displayed Tomato label , around 50% of B1b B cells and the vast majority of B2 B cells ( Figure 3D right ) were Tomato+ ( comparable to levels in PB ) . This is in line with a more strict fetal/postnatal origin of B1a B cells , the ontogenically mixed origin of B1b B cells ( Kantor et al . , 1995 ) and an adult HSC origin of most B2 B cells . Finally , we investigated adult HSC contribution to microglial cells of the brain , a subset of central nervous system myeloid cells that has been proposed to arise entirely from embryonic precursor cells ( Alliot et al . , 1999 ) . Evaluations by confocal microscopy of the brain parenchyma revealed no detectable Tomato expression in IBA-1+ microglia ( Figure 3E ) , while Fgd5 expressing endothelial cells ( Cheng et al . , 2012; Gazit et al . , 2014 ) displayed abundant Tomato expression ( Figure 3E middle ) . We next set out to investigate how chronological aging influence on HSC contribution to hematopoiesis . To achieve rapid and robust labeling of HSCs , we labeled juvenile and aged Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice by injecting Tamoxifen for five consecutive days ( 5x ) . Labeling was followed by evaluation of the fraction of Tomato positive cells in HSC and MPP fractions of the BM LSK compartment one day after the last Tamoxifen injection . In aged mice , the initial labeling was highly specific to HSCs , with only low levels of labeling in MPP2 cells . In sharp contrast , a larger fraction of LSKCD150-CD48- MPPs were labeled in juvenile mice ( Figure 4A ) . Next , we correlated how increasing age influences on the HSC generation of other LSK/MPP subsets . Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice between 6 and 96 weeks of age were labeled using a 5x Tamoxifen injection scheme , before evaluation of Tomato label in HSCs/MPP2-421 days later . Ratios of the fraction of labeled MPPs vs . labeled HSCs in corresponding mice was calculated and plotted against mouse age at labeling ( Figure 4B ) . This established that label progression into all MPP subsets in aged mice was substantially lower when compared to young adult mice and further revealed that the HSC contribution to MPPs and MPP3/4s gradually declines with age towards very little replenishment of in particular MPP3/4 in very old age ( Figure 4B ) . To evaluate the functional potential of initially Tomato-labeled MPPs , we FACS sorted HSCs and different MPP subsets isolated from non-labeled Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice and transplanted cells into lethally irradiated wild type mice . Recipient mice were provided Tamoxifen containing food pellets throughout the experiment . We analyzed donor chimerism and Tomato labeled PB cells after repetitive PB blood sampling . As expected , we did not observe any multilineage long-term reconstitution or Tomato+ cells in MPP2/MPP3/4 transplanted mice ( Figure 4C ) . 10 out of 10 HSC transplanted mice were multi lineage reconstituted at 16 weeks post transplantation with high levels of Tomato+ donor cells in all evaluated lineages ( Figure 4C ) . More surprisingly 2 out of 9 MPP transplanted mice displayed donor reconstitution levels > 1% in all lineages 16 weeks after transplantation ( Figure 4C ) . This long-term multilineage reconstitution potential from MPPs was accompanied with robust Tomato labeling among donor cells and revealed that Tomato labeled phenotypic MPPs perform as bona fide HSCs after transplantation ( Figure 4C ) . By contrast , mice that received MPP cells and displayed only transient myeloid reconstitution never displayed any Tomato+ cells . This demonstrates , in young mice , the presence of a minor CD150- HSC activity that appears exclusively coupled to Fgd5 expression . Finally , we were interested in determining whether the age-related decrease in HSC derived MPPs might influence on the generation kinetics of lineage-restricted progenitors and mature blood cells . For this , we labeled juvenile ( 29 days ) and aged ( 87–89 weeks ) Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice with 1x Tamoxifen , which was followed by evaluation of label progression after 18 weeks ( Figure 4D and Figure 4—figure supplement 1 ) . While the frequency of Tomato labeled HSCs was similar among the two groups ( Figure 4—figure supplement 1 ) , all progeny exhibited reduced frequencies of Tomato labeled cells in aged mice , demonstrating a strikingly reduced multilineage differentiation capacity of HSCs as a consequence of age .
In this work , we explored the cellular contribution from HSCs using a HSC-specific lineage tracing approach . Our work revealed robust HSC contribution to adult multilineage hematopoiesis and , while not focused at studying fetal hematopoiesis , support the previously reported fetal/juvenile origins of several other specific hematopoietic subsets . We observed the fastest label progression into the platelet lineage , which might be related to the recent observations that at least a subset of HSCs appears platelet-biased ( Sanjuan-Pla et al . , 2013; Shin et al . , 2014 ) . Thereafter , erythrocytes and granulocytes acquired label with similar kinetics , although Tomato+ erythrocytes emerged in the peripheral blood somewhat later than granulocytes . This likely reflects the slower turnover of mature erythrocytes compared to other myeloid cells . On the other hand , among granulocytes , we did not find any noticeable differences when evaluating HSC contribution to the neutrophil or eosinophil lineages ( data not shown ) , despite the seemingly distinct transcriptional underpinnings of these lineages ( Drissen et al . , 2016 ) . Among lymphocytes , the HSC contribution was most rapid and robust to the NK cell lineage . At present , we can say little on whether this reflects an early and distinct progenitor intermediate for NK cells ( Wu et al . , 2014 ) or whether NK cells are , at least partially , regenerated through myeloid progenitors ( Chen et al . , 2015; Grzywacz et al . , 2011 ) . While the extremely slow HSC generation of adaptive immune components ( and T cells in particular ) have implications for our understanding of HSCs , by proposing that adult HSCs to a large extent present as myeloid-biased in an unperturbed scenario , our results also suggest the necessity of very harsh conditioning to achieve an ‘immunological reboot’ ( the generation of naïve lymphocytes ) in certain autoimmune situations ( Atkins et al . , 2016 ) . According to current models of hematopoiesis , lineage committed progenitors reside developmentally in between HSCs and their mature progeny , and much work has been aimed at detailing these stages ( Adolfsson et al . , 2005; Akashi et al . , 2000; Arinobu et al . , 2007; Oguro et al . , 2013; Pietras et al . , 2015; Pronk et al . , 2007 ) . What has remained more unknown is the relationships of these defined progenitors not only to HSCs and their mature offspring , but also whether they are obligatory . In our work , we could demonstrate that the generation rates of each evaluated myeloid progenitor subset correlated highly to their corresponding mature offspring , although the generation kinetics varied depending on lineage . Thus , although some challenges have recently been raised on how lineage commitment occur from HSCs based on inferences from large-scale single-cell RNA sequencing experiments ( Paul et al . , 2015; Velten et al . , 2017 ) , our data support the more conventional view that the generation of mature myeloid cells is preceded by the generation of obligatory lineage-committed intermediates . Investigations of immature progenitors revealed that the most rapid label progression associated with LSKCD150-CD48- MPPs . MPPs share many defining properties of HSCs , including multilineage differentiation potential and very low proliferation rates in steady state ( Säwén et al . , 2016 ) . The distinction of MPPs from HSCs is mainly thought to result from pronounced differences in self-renewal; a property so far entirely evaluated by transplantation . Intriguingly , our work revealed that LSKCD150-CD48- MPPs displayed more rapid label kinetics in young mice , which was gradually declining with advancing age . At the same time , we could in agreement with previous studies ( Weksberg et al . , 2008 ) also demonstrate a minor multilineage HSC activity in this compartment , which correlated exclusively to Fgd5 expression/Tomato labeling . Together with an overall decline in multilineage HSC contribution of aged mice , these results strongly propose a model in which aging associates with reduced/compromised HSC differentiation , which in combination with the well-established expansion of HSCs with age ( Morrison et al . , 1996; Rossi et al . , 2005; Sudo et al . , 2000 ) appears to represent a physiologically relevant compensatory mechanisms to sustain multilineage hematopoiesis from HSCs ( Figure 4E ) . Compared to MPPs , MPP3/4 are perhaps easier to approach given their restrictions in lineage potential ( lack of Meg/E potential ) ( Adolfsson et al . , 2005; Arinobu et al . , 2007; Pietras et al . , 2015; Pronk et al . , 2007 ) . We found MPP3/4 to be regenerated from HSCs with slow kinetics compared to other downstream myeloid progenitor cells , but also to MPP2 cells , that we in agreement with other studies ( Pietras et al . , 2015 ) find ‘primed’ towards Meg/E development . Intriguingly , our data proposes a significant self-renewal activity of at least a subset of MPP3/4 , with the demonstration that this fraction never reached label equilibrium with HSCs in any evaluated experimental setting . This might be particularly relevant in the setting of age , a situation in which HSC was found to generate MPP3/4 very inefficiently ( Figure 4E ) . While limited , a few groups have recently approached HSC contribution to native hematopoiesis . Evaluations of hematopoiesis using transposon mobilization led to the conclusion that HSCs are not major contributors to adult hematopoiesis ( Sun et al . , 2014 ) . To some degree , this conclusion was later corroborated by CreER-mediated labeling of a minor fraction of the adult HSC pool using a Tie2-based CreER driver ( Busch et al . , 2015 ) . Limited HSC contribution to adult hematopoiesis is in sharp contrast to the results we present here and to results from another recent study ( Sawai et al . , 2016 ) . Our studies would propose that absence of a HSC specific driver , as in the work from Sun et al . , makes interpretations of HSC contribution very complicated , not the least for the lymphoid lineages , while the labeling of only a minor fraction of HSCs , as in the work from Busch et al . , might select for a subset of HSCs with a rather distinct functional behavior . In summary , we conclude that although the study of native hematopoiesis highlights fundamental differences , with in particular slower regeneration times from HSCs to those seen after transplantation , they regardless corroborate decades of research derived from transplantation experiments in which HSCs has been proposed to continuously contribute to hematopoiesis .
For inducible marking of HSCs in vivo , we crossed Fgd5-2A-ZsGreen-CreERT2 mice ( Gazit et al . , 2014 ) ( JAX 027789 ) to Rosa26-LoxP-Stop-LoxP-Tomato ( Madisen et al . , 2010 ) ( JAX 007905 ) mice , resulting in Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice . For simultaneous in vivo tracking of proliferation history and marking of HSCs , Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice were crossed with Col1a1tetO-H2B-mCherry/tetO-H2B-mCherry; ROSA26rtTA/rtTA mice ( JAX 014602 ) to generate Fgd5CreERT2/+; Rosa26lsl-Tomato/rtTA; Col1a1tetO-H2B-mCherry/+ . Such mice were administered doxycycline in food pellets ( 2 g/kg; Ssniff Spezialdiäten ) for 2 weeks followed by 5 weeks of chase before HSC marking by a single i . p . Tamoxifen injection ( 50 mg/kg ) and analysis 5 days later . Tamoxifen was purchased from Sigma-Aldrich and suspended at 100 mg/ml in ethanol and mixed with sunflower oil to a concentration of 10 mg/ml . Tamoxifen was administered by intraperitoneal injections at 50 mg/kg body weight once ( 1x ) or for 5 ( 5x ) consecutive days . To acquire full/maximal HSC labeling , cohorts of mice were continuously fed Tamoxifen containing food pellets for 16 weeks . Mice on Tamoxifen food were regularly bled ( sparsely; 1–2 drops ) during the labeling period and during the chase period . Transplanted recipient mice were subjected to lethal irradiation ( 950 rad ) except CD45 . 1/2 mice that were CD45-depleted by intravenous injection of an immunotoxin ( 3 mg/kg ) consisting of CD45 . 2-biotin ( clone 104 ) and streptavidin-Saporin ( Advanced Targeting Systems ) in a 1:1 molar ratio 3 days prior to transplantation , as described ( Palchaudhuri et al . , 2016 ) . All transplanted cells were isolated from CD45 . 2 Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice . CD45-depleted recipients were transplanted with 107whole bone marrow ( WBM ) cells and irradiated mice were transplanted with 3 × 106 WBM cells ( n = 7 ) , 200 ( n = 10 ) or 100 ( n = 2 ) HSCs , 635 MPPs ( n = 10 ) , 1200 MPP2s ( n = 5 ) or 7300 MPP3/4s ( n = 4 ) . Tomato+ and Tomato- HSCs were isolated from Fgd5CreERT2/+; Rosa26lsl-Tomato/+ mice injected with Tamoxifen 2 days before isolation and transplantation into congenic C57BL/6 ( CD45 . 1+ ) mice on normal chow . Before transplantations of HSCs or MPPs FACS sorted cells were mixed with 3 × 105 WBM competitor cells in 200 µl PBS supplemented with 2 mM EDTA and 2% FBS before injection . Where indicated , recipient mice were given Tamoxifen containing food pellets ( 400 mg/kg Tamoxifen Citrate , Ssniff Spezialdiäten ) throughout the experiments . At the indicated time point , PB was collected from the tail vein for reconstitution analysis . H2B-mCherry labeling in Col1a1tetO-H2B-mCherry/tetO-H2B-mCherry; ROSA26rtTA/rtTA mice was induced by administration of doxycycline ( Säwén et al . , 2016 ) . Thereafter , mice were chased for 5 weeks while eating Tamoxifen containing food pellets or normal chow ( No TAM ) , followed by FACS analysis to assess H2B-mCherry dilution in HSCs/progenitors . All mice were maintained at the animal facilities at BMC at Lund University and all experiments were performed with consent from the Malmö/Lund animal ethics board , reference number M186-15 . Immunophenotyping by FACS was done as described ( Säwén et al . , 2016 ) ( Supplementary file 1 ) . For platelet and erythrocyte analysis , 1 μl of whole blood was taken to 300 μl PBS before FACS analysis . Cells were sorted and/or analyzed on a FACS Aria III cell sorter ( Becton Dickinson ) or on a LSRFortessa ( Becton Dickinson ) . For isolation of peritoneal cells , peritoneal lavage was performed using 10 mL PBS . For isolation of skin epidermal cells , the flank of the mouse was shaved before excision of skin . The skin was incubated for 25 min at 37°C in a dissociation buffer ( PBS containing 2 . 4 mg/ml of dispase ( Roche ) and 3% FCS ) before separation of dermis from the epidermis . Pieces of epidermis were incubated for 30 min at 37°C in digestion buffer ( PBS supplemented with 1 mg/ml collagenase IV [Sigma-Aldrich] , 100 U/ml DNase I [Sigma] , 2 . 4 mg/ml dispase [Roche] and 3% FBS ) and thereafter filtered and stained against indicated markers . Before analysis , cells were incubated with Propidium Iodide ( Invitrogen ) to exclude dead cells . Mice were deeply anaesthetized with an overdose of pentobarbital and transcardially perfused with cold saline . Brains were post-fixed for 48 hr in 4% paraformaldehyde ( PFA ) and incubated in 20% sucrose for 24 hr before being cut in 30 μm thick coronal sections on a microtome . Sections were incubated in blocking solution ( 5% normal serum and 0 . 25% Triton X-100 in 0 . 1 M potassium-phosphate buffered solution ) for one hour and subsequently overnight at 4°C with the primary antibody ( Iba1 1:1000 Wako ) . Fluorophore-conjugated secondary antibody ( Molecular Probes or Jackson Laboratories ) was diluted in blocking solution and applied for 2 hr at room temperature . Tomato label could be detected without any staining . Nuclei were stained with Hoechst ( Molecular Probes ) for 10 min and sections were mounted with Dabco mounting medium . Images were obtained using confocal microscopy ( Zeiss , Germany ) . Single LSKCD150+CD48+ cells , MPP3/4s and HSCs were sorted into Terasaki wells containing 20 μl of media ( OptiMEM supplemented with 10% FCS , 1:1000 Gentamicin ( Invitrogen ) , 1:100 GlutaMAX ( Invitrogen ) and 1:500 β-mercaptoethanol ( Invitrogen ) supplemented with cytokines ( mSCF ( Peprotech ) 100 ng/ml , TPO ( Peprotech ) 10 ng/ml , IL-3 ( Peprotech ) 5 ng/ml , EPO ( Janssen ) 5units/ml , human G-CSF ( Amgen ) 10 ng/ml ) . After 6 days of culture at 37°C , wells were scored and evaluated for the presence of megakaryocytes and erythroid cells by visual inspection in microscope . qRT-PCR analyses using the Fluidigm Biomark HD Platform was done as described ( Säwén et al . , 2016 ) ( Supplementary file 2 ) . PCA on gene expression data from all reference populations was performed using Clustvis , ( http://biit . cs . ut . ee/clustvis/ ) . Single-cell RNA seq libraries were generated using a Chromium system ( 10x Genomics ) according to the manufacturer’s instructions . Two consecutive sequencing runs were performed to achieve enough sequencing depth and data was combined and further analyzed using the Cell RangerTM pipeline ( 10x Genomics ) . The accession number for the single-cell RNA sequencing data reported in this paper is GSE122473 . Data were analyzed using Microsoft Excel ( Microsoft ) and Graphpad Prism ( GraphPad Software ) . All FACS analyses were performed using Flowjo software ( TreeStar ) . | As far as we know , all adult blood cells derive from blood stem cells that are located in the bone marrow . These stem cells can produce red blood cells , white blood cells and platelets – the cells fragments that form blood clots to stop bleeding . They can also regenerate , producing more stem cells to support future blood cell production . But , our understanding of the system may be incomplete . The easiest way to study blood cell production is to watch what happens after a bone marrow transplant . Before a transplant , powerful chemotherapy kills the existing stem cells . This forces the transplanted stem cells to restore the whole system from scratch , allowing scientists to study blood cell production in fine detail . But completely replacing the bone marrow puts major stress on the body , and this may alter the way that the stem cells behave . To understand how adult stem cells keep the blood ticking over on a day-to-day basis , experiments also need to look at healthy animals . Säwén et al . now describe a method to follow bone marrow stem cells as they produce blood cells in adult mice . The technique , known as lineage tracing , leaves an indelible mark , a red glow , on the stem cells . The cells pass this mark on every time they divide , leaving a lasting trace in every blood cell that they produce . Tracking the red-glowing cells over time reveals which types of blood cells the stem cells make as well as provides estimates on the timing and extent of these processes . It has previously been suggested that a few types of specialist blood cells , like brain-specific immune cells , originate from cells other than adult blood stem cells . As expected , the adult stem cells did not produce such cells . But , just as seen in transplant experiments , the stem cells were able to produce all the other major blood cell types . They made platelets at the fastest rate , followed by certain types of white blood cells and red blood cells . As the mice got older , the stem cells started to slow down , producing fewer blood cells each . To compensate , the number of stem cells increased , helping to keep blood cell numbers up . This alternative approach to studying blood stem cells shows how the system behaves in a more natural environment . Away from the stresses of transplant , the technique revealed that blood stem cells are not immune to aging . In the future , understanding more about the system in its natural state could lead to ways to boost blood stem cells as we get older . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2018 | Murine HSCs contribute actively to native hematopoiesis but with reduced differentiation capacity upon aging |
Mitochondrial complex III ( CIII2 ) and complex IV ( CIV ) , which can associate into a higher-order supercomplex ( SC III2+IV ) , play key roles in respiration . However , structures of these plant complexes remain unknown . We present atomic models of CIII2 , CIV , and SC III2+IV from Vigna radiata determined by single-particle cryoEM . The structures reveal plant-specific differences in the MPP domain of CIII2 and define the subunit composition of CIV . Conformational heterogeneity analysis of CIII2 revealed long-range , coordinated movements across the complex , as well as the motion of CIII2’s iron-sulfur head domain . The CIV structure suggests that , in plants , proton translocation does not occur via the H channel . The supercomplex interface differs significantly from that in yeast and bacteria in its interacting subunits , angle of approach and limited interactions in the mitochondrial matrix . These structures challenge long-standing assumptions about the plant complexes and generate new mechanistic hypotheses .
The canonical mitochondrial electron transport chain ( mETC ) , composed of four integral membrane protein complexes ( complexes I–IV; CI–CIV ) located in the inner mitochondrial membrane ( IMM ) , transfers electrons from NADH and succinate to molecular oxygen . The concomitant pumping of protons ( H+ ) across the IMM establishes an electrochemical proton gradient that is used by ATP synthase to produce ATP ( Nicholls , 2013 ) . Whereas the atomic details of the respiratory complexes and several supercomplexes ( higher-order complex assemblies ) are known for yeast , mammals and bacteria ( Vinothkumar et al . , 2014; Fiedorczuk et al . , 2016; Gu et al . , 2016; Letts et al . , 2016; Wu et al . , 2016; Zhu et al . , 2016; Zickermann et al . , 2015; Blaza et al . , 2018; Guo et al . , 2017; Letts et al . , 2019; Agip et al . , 2018; Parey et al . , 2018 ) , the high-resolution structural details of the respiratory complexes and supercomplexes of plants have remained mostly unknown . Complex III ( CIII2 ) , also called the cytochrome bc1 complex or ubiquinol-cytochrome c oxidoreductase , is an obligate dimer that transfers electrons from ubiquinol in the IMM ( reduced by CI , CII , or alternative NADH dehydrogenases ) to soluble cytochrome c in the intermembrane space ( IMS ) ( Nicholls , 2013 ) . This redox reaction is coupled to the pumping four H+ to the IMS . CIII2 is composed of three conserved subunits present in all organisms ( cytochrome b , COB; cytochrome c1 , CYC1; and the iron-sulfur ‘Rieske’ subunit , UCR1 ) , as well as a varying number of accessory subunits present in eukaryotes ( Iwata et al . , 1998; Xia et al . , 2013; Xia et al . , 1997 ) . Each CIII monomer contains one low-potential heme b ( bL ) and one high-potential heme b ( bH ) in COB , a heme c in CYC1 , a 2Fe-2S iron-sulfur cluster in UCR1 , as well as two quinone-binding sites ( QP and QN close to the positive/IMS and negative/matrix sides respectively ) in COB . Given that CIII2 is a dimer , these sites in each CIII monomer are symmetrical within the dimer in isolation . However , the symmetry may be broken when CIII assembles into asymmetrical supercomplexes ( Letts et al . , 2016; Letts et al . , 2019; Letts and Sazanov , 2017 ) . CIII2’s redox and proton pumping occur via the ‘Q-cycle’ mechanism ( Cramer et al . , 2011 ) , which allows for efficient electron transfer between ubiquinol ( a two-electron donor ) and cyt c ( a one-electron acceptor ) . To this end , one electron is transferred from ubiquinol in the Qp site to the 2Fe-2S in the UCR1 head domain . The head domain then undergoes a large conformational swinging motion from its ‘proximal’ position close to the QP site to a ‘distal’ CYC1 binding site adjacent to heme c1 ( Zhang et al . , 1998 ) . The electron is then transferred via heme c1 to cyt c bound to CIII2 in the IMS . Of note , the UCR1 head domain belongs to the opposite CIII protomer relative to COB and CYC1 . The second electron donated by ubiquinol is transferred via hemes bL and bH to a quinone in the QN site , reducing it to ubisemiquinone . The cycle is repeated to regenerate ubiquinol in the QN site , ultimately reducing two molecules of cyt c and pumping four protons . In eukaryotes , the large CIII2 accessory subunits exposed to the mitochondrial matrix have homology to mitochondrial processing peptidases ( MPP ) of the pitrilysin family ( Gakh et al . , 2002 ) . These metalloendopeptidases—composed of an active β subunit and an essential but catalytically inactive α subunit—cleave mitochondrial signal sequences of proteins that are imported into the mitochondria ( Gakh et al . , 2002 ) . Whereas in yeast the CIII2 accessory subunits with MPP homology ( ScCor1/Cor2 ) have completely lost MPP enzymatic activity , the mammalian CIII2 homolog ( UQCR1/UQCR2 heterodimer ) retains basal activity to only one known substrate ( the Rieske subunit ) ( Gakh et al . , 2002; Taylor et al . , 2001 ) . Hence , in yeast and mammals this enzymatic activity is carried out by soluble MPP heterodimers in the mitochondrial matrix . In contrast , in vascular plants , there is no additional soluble MPP enzyme , and all MPP activity is provided by the CIII2 MPP heterodimer ( MPP-α/β ) . Thus , in plants CIII2 serves a dual role as a respiratory enzyme and a peptidase ( Braun et al . , 1992; Emmermann et al . , 1993; Eriksson et al . , 1994; Braun et al . , 1995; Eriksson et al . , 1996; Braun and Schmitz , 1995a; Glaser and Dessi , 1999 ) . This integration of respiratory and peptidase activities may have occurred early in eukaryogenesis ( Braun and Schmitz , 1995b ) . The bioenergetic implications of this dual function of plant CIII2 remain unknown . Complex IV ( CIV ) , also called cytochrome c oxidase , transfers electrons from cyt c to molecular oxygen , reducing it to water ( Nicholls , 2013 ) . The redox reaction is coupled to the pumping of four protons into the IMS . Like CIII2 , CIV is composed of three conserved subunits ( COX1 , COX2 , COX3 ) and a variable number of accessory subunits , depending on the organism . Electrons are transferred from cyt c to oxygen via COX1’s dinuclear copper ( CuA ) , heme a and copper-associated heme a3 ( CuB , binuclear center ) . The passage of protons from the matrix to the IMS occurs through distinct ‘channels’ formed by protonatable amino-acid residues ( Rich , 2017; Rich and Maréchal , 2013; Yoshikawa and Shimada , 2015; Wikström et al . , 2018 ) . It is currently believed that , whereas yeast CIV pumps protons through the K and D transfer pathways ( named after key amino-acid residues in each pathway ) , mammalian CIV uses an H channel in addition to the K and D channels ( Rich , 2017; Rich and Maréchal , 2013; Yoshikawa and Shimada , 2015; Wikström et al . , 2018; Maréchal et al . , 2020 ) . The contribution of the K , D , H pathways in plant CIV has not been characterized . In the IMM , respiratory complexes can be found as separate entities or as higher-order assemblies known as supercomplexes ( Schägger and Pfeiffer , 2000 ) . Although it was initially hypothesized that supercomplexes would allow for direct substrate channeling between complexes , evidence has mounted against this view ( Gu et al . , 2016; Letts et al . , 2016; Letts et al . , 2019; Yu et al . , 2018; Sousa et al . , 2016; Blaza et al . , 2014; Fedor and Hirst , 2018 ) . Instead , supercomplexes may have roles in improving the stability of the complexes , providing kinetic advantages to the electron transfer or reducing the production of reactive oxygen species or of aggregates in the IMM ( Letts and Sazanov , 2017; Milenkovic et al . , 2017 ) . Supercomplexes of various stoichiometries between CIII2 and CIV ( e . g . SC CIII2+CIV , SC CIII2+CIV2 ) have been seen ( Schägger and Pfeiffer , 2000; Eubel et al . , 2004; Eubel et al . , 2003 ) . In plants , the CIII2-CIV supercomplex of highest abundance is a single CIII dimer with a single CIV monomer ( SC III2+IV ) ( Eubel et al . , 2004 ) . High-resolution structures of the model yeast Saccharomyces cerevisiae and Mycobacterium smegmatis CIII2-CIV supercomplex ( SC III2+IV2 ) have recently been determined ( Hartley et al . , 2019; Rathore et al . , 2019; Gong et al . , 2018; Wiseman et al . , 2018 ) . Although there is currently no high-resolution structure for a mammalian CIII2-CIV supercomplex , the supercomplex between CI , CIII2 and CIV ( SC I+III2+IV , the respirasome ) shows a distinct interaction interface between CIII2 and CIV relative to the SC III2+IV2 from yeast and bacteria ( Gu et al . , 2016; Letts et al . , 2016 ) . Similar to that seen in comparative tomographic studies of plant SC I+III2 and bovine and yeast SC I+III2+IV ( Davies et al . , 2018 ) , the above SC III2+IV2 studies revealed that , while the general configuration of the individual CIII2 and CIV are conserved , the location of the binding interface between CIII2 and CIV in the supercomplex is divergent , with different subunits involved in the different organisms . For plant CIII2 and CIV , the only currently available structural information is from low-resolution , 2D-averages of negative-stain EM samples from A . thaliana ( Dudkina et al . , 2005 ) and potato ( Bultema et al . , 2009 ) . High-resolution structures or atomic models for CIII2 , CIV or their supercomplexes are not currently available for the plant kingdom . Here we present the cryoEM structures of CIII2 and SC III2+IV from the vascular plant Vigna radiata ( mung bean ) at nominal resolutions of 3 . 2 Å and 3 . 8 Å , respectively . Using focused refinements around CIV , we achieved a nominal resolution for CIV of 3 . 8 Å . The structures reveal plant CIII2 and CIV’s active sites , as well as the plant-specific configuration of several CIII2 and CIV subunits . The structures also show the SC III2+IV binding interface and orientation , which is unique to plants . Additionally , using cryoEM 3D-conformation variability analysis ( Punjani and Fleet , 2020 ) , we were able to visualize the swinging motion of CIII2’s UCR1 head domain at 5 Å resolution in the absence of substrate or inhibitors . We also observed complex-wide coupled conformational changes in the rest of CIII2 . These results question long-standing assumptions , generate new mechanistic hypotheses and provide the initial structural basis for the development of more selective agricultural inhibitors of plant CIII2 and CIV .
By docking the individually refined CIII2 and CIV models into the SC III2+IV composite map ( Figure 1 , Figure 1—figure supplement 2 , Video 3 ) , we were able to define the binding interface of the plant supercomplex . Direct contacts between the complexes occur in one site in the matrix side and one site in the IMS ( Figure 8 ) . Site 1 ( matrix side ) , shows a single hydrophobic contact between one residue of VrQCR8 ( Pro31 ) and one residue of VrCOX2 ( Trp59 ) ( Figure 8B ) . Our cryoEM reconstruction also contains a short stretch of weak , unassigned density near the first modelled residue of VrCOX5c ( BtCOX6c , ScCox9 ) , which could maximally represent an additional four amino acids . If so , the N-terminus of VrCOX5c could potentially provide additional contacts with CIII2 . However , this density may also correspond to a bridging lipid bound between the two complexes . The limited matrix-side contacts in V . radiata are in stark contrast to the supercomplex interface in S . cerevisiae , where binding is dominated by interactions between ScCor1 ( VrMPP-β ) and ScCOX5a ( VrCox4 ) on the matrix side . Instead , V . radiata’s site 2 ( IMS side ) provides the bulk of the protein-protein interactions of the supercomplex , with a hydrophobic interaction between VrQCR6 and VrCOX5c ( Leu26 and Leu51 respectively ) as well as an interface between VrQCR6 ( Pro19 and Lys20 ) and VrCOX4 ( Arg114-Phe117 ) ( Figure 8C ) . Despite the potential for lipid bridges at the matrix leaflet of the IMM , there are no direct contacts in the membrane and no protein contact at the IMS leaflet of the IMM . The overall limited binding interactions between V . radiata’s CIII2 and CIV , and , consequently , the lower expected stability of the plant supercomplex compared to yeast's , are consistent with the fact that ( CIII2+CIV ) n supercomplexes have only been experimentally identified in a few of the plant species studied ( Dudkina et al . , 2006; Braun , 2020 ) . Unsurprisingly given the large differences in contacts , there is a significant difference in the angle between CIII2 and CIV in V . radiata versus yeast , resulting in a more ‘open’ orientation in V . radiata ( Figure 8—figure supplement 1 ) . This orientation leads to a difference of 18° in the angle between the CIII2 and CIV as measured by the relative positions of the bh-hemes in CIII2 and the a-hemes in CIV . The difference in orientation also results in a larger estimated distance between the CIII2- and CIV-bound cyt c in V . radiata ( ~70 Å ) than in yeast ( ~61 Å ) ( Figure 8—figure supplement 1E ) .
Plant CIV subunit composition has been previously analyzed by mass spectrometry of proteins from 2-dimensional blue-native gels ( BN-PAGE ) ( Millar et al . , 2004; Klodmann et al . , 2011; Senkler et al . , 2017 ) . Although these studies were mostly in agreement , several putative CIV subunits—including several putative plant-specific subunits—differed between datasets . Given the considerable technical challenges in obtaining plant CIV samples , experimental evidence for the stoichiometric presence of these putative subunits in complex IV remained limited , with strongest evidence for COX-X1 , COX-X2 and COX-X4 ( Senkler et al . , 2017; Braun , 2020 ) . The structure of V . radiata CIV presented here offers a complementary approach to determine the complex’s subunit composition . Our structure shows that CIV obtained from etiolated V . radiata sprouts is composed of 10 subunits ( three mitochondrially encoded subunits and seven accessory subunits ) . Only three of the previous plant-specific candidates are seen in our structure ( COX-X2 , COX-X3 and COX-X4 ) . Moreover , structural analysis shows that COX-X2 , COX-X3 and COX-X4 are homologs of mammalian and yeast CIV subunits ( BtCOX4/ScCOX5 , BtCOX7c/ScCox8 and BtCOX7a/ScCox7 , respectively ) rather than being plant-specific subunits . Although mass spectrometry analysis of our mixed sample also shows some evidence for the occurrence of COX-X1 , this protein is not present in our structure , and its function remains unknown . Our structure provides a new definition for plant CIV’s composition; however , we note that the arrangement may differ between free CIV and that in supercomplexes . Moreover , its composition may be dynamically regulated in different metabolic states ( e . g . different light or oxygen levels ) , as is known to occur for certain subunit isoforms in other organisms ( Burke et al . , 1997 ) . The 3DVA allowed us to observe the full swing of the UCR1 ( Rieske subunit ) head domain ( Figure 4 , Video 10 ) without the need for any CIII2 inhibitor to capture the proximal , distal or intermediate positions in the mobile state ( Esser et al . , 2019 ) . As such , it provides direct confirmation for a multitude of previous crystallographic , mutational , kinetic and molecular dynamics studies , mostly done in the presence of inhibitors , that showed the flexibility and mobility of this domain ( Xia et al . , 2013; Cooley , 2013; Berry and Huang , 2011; Izrailev et al . , 1999; Huang and Berry , 2016 ) . Together , the findings definitively show that the swinging motion is an inherent property of the UCR1 in the absence of substrates . Moreover , the conformational heterogeneity analysis suggests that the movement of the UCR1 head domains is coordinated between the CIII2 protomers to a large degree ( Figure 4C , Video 10 ) . Determining the nature of and mechanism for this inter-protomer coordination will have implications on electron transfer in CIII2 and its supercomplexes . Unfortunately , however , this conformational flexibility precluded us from building an atomic model for the head domain . Thus , we were not able to evaluate the H-bonding pattern of the UCR1 head domain with COB , or the implications of such pattern to the Q-cycle electron bifurcation mechanism ( Xia et al . , 2013; Ho et al . , 2020; Belt et al . , 2017 ) . Similarly , the resolution of the 3DVA was not sufficient to evaluate changes in the positions of COB’s cd1 and ef helices . These helices are critical components of the UCR1’s COB binding ‘crater’ ( Xia et al . , 2013 ) , and their position changes in response to binding of different CIII2 inhibitors ( Esser et al . , 2006 ) , with important implications for the Q-cycle mechanism . It is important to note that 3DVA only reveals conformational changes , with no information on kinetics or occupancy rates . Nonetheless , we demonstrate here that cryoEM conformational heterogeneity tools such as 3DVA ( Punjani and Fleet , 2020 ) and others ( Zhong et al . , 2020 ) are a valuable complementary approach to study the conformational changes of CIII2’s Q-cycle in its native state , as well as in the presence of inhibitors and substrates . Moreover , the 3DVA revealed that CIII2 can undergo different types of complex-wide motions that are coordinated across sides of the membrane and between protomers ( Videos 6–9 ) . Changes at the ‘top’ of the complex on one side of the membrane co-vary ( i . e . are coupled ) with movement at the ‘bottom’ of the complex on the other side of the membrane . This long-range conformational coupling across the entire CIII2 could be the basis for symmetry-breaking and coordination of the UCR1 head domain motion between the CIII2 protomers . The long-range conformational coupling is particularly relevant in the context of plant CIII2’s dual roles in signal-peptide processing and respiration . The potential interdependence between these two functions was investigated using CIII2 inhibitors ( Eriksson et al . , 1994; Eriksson et al . , 1996 ) , ultimately leading to the interpretation that these functions are independent ( Glaser and Dessi , 1999 ) . In the presence of the CIII2 respiratory inhibitors antimycin A ( QN-site inhibitor ) and myxothiazol ( QP-site inhibitor ) at concentrations that inhibit ~90% of spinach CIII2’s respiratory activity , the complex’s peptidase activity is inhibited 30–40% ( Eriksson et al . , 1994; Eriksson et al . , 1996 ) . Given that higher concentrations of inhibitors are needed to elicit MPP inhibition than respiratory inhibition , the authors initially speculated that the effects on the peptidase activity could be due to the inhibitors preventing necessary conformational changes ( Eriksson et al . , 1994 ) . However , when crystal structures of metazoan CIII2 in complex with these inhibitors became available and revealed the locations of the inhibitor binding sites ( Iwata et al . , 1998; Xia et al . , 1997; Zhang et al . , 1998 ) , the large distances between these sites and the MPP domain were interpreted to reinforce the notion that the dual roles of plant CIII2are independent , as long-range coupled conformational changes were deemed unlikely ( Glaser and Dessi , 1999 ) . In contrast , our 3DVA results showed that long-range coupled motions are intrinsic to V . radiata CIII2 . Moreover , our atomic model of plant CIII2 revealed additional contacts and secondary-structure elements not previously seen in other organisms that enhance the interaction between the MPP domain and the rest of the complex . For example , the extended N-termini of MPP-α and -β bridge across the dimer and provide plant-specific contacts with CIII2’s membrane subunits ( Figure 3 , Figure 3—figure supplements 1–2 ) . Moreover , UCR1’s longer N-terminus in plants also provides plant-specific contacts with the MPP domain ( Figure 2—figure supplements 1 and 3; Figure 3—figure supplement 1C ) . Given its span across the membrane and its domain-swapping across protomers , UCR1 may have roles as a ‘conformational coupler’ beyond its essential function in the Q-cycle . Together , our 3DVA results challenge long-standing assumptions on plant CIII2’s suitability for conformational coupling and call for a re-evaluation of the relationship between the respiratory and the processing activities of the plant complex . The orientation and binding interfaces of SC III2+IV vary significantly among organisms ( Hartley et al . , 2019; Rathore et al . , 2019; Gong et al . , 2018; Wiseman et al . , 2018; Sousa and Vonck , 2019 ) . For instance , the CIII surface used by yeast to bind to CIV is instead used by mammals to bind to CI ( Hartley et al . , 2019 ) . Given these disparities , it is not surprising that the differences seen in the VrCIV subunits are concentrated in the subunits that form the supercomplex interfaces in the different organisms ( Figure 6 ) . Moreover , while some of the supercomplex interactions in V . radiata are reminiscent of the supercomplex interface in yeast , there are significant differences in the protein:protein sites , interacting subunits and angle of orientation within the SC ( Figure 8 and Figure 8—figure supplement 1 ) . In yeast , the main interface is on the matrix side , with the N-terminal helical domain of ScCox5a ( VrCOX4 ) interacting with ScCor1 ( homolog of VrMPP-β ) . In V . radiata , this interface is lacking , as plant COX4 does not possess the ~100 amino-acid helical N-terminal domain present in yeast and mammals ( Figure 6 and Figure 6—figure supplement 1 ) . In contrast , the main supercomplex interface in mung bean is in the IMS , driven by contacts between VrQCR6 and VrCOX4 . In the yeast supercomplex , the homologs of VrQCR6 and VrCOX4 ( ScQCR6 and ScCOX5a/b ) also interact in the IMS side , but in a much more limited fashion . In light of VrCIII2’s conformational heterogeneity and the potential interdependence between respiratory and peptidase functions of plant CIII2 , an intriguing possibility is that matrix-side interactions between CIII2 and CIV are minimized in the plant supercomplex to prevent steric constraints on the MPP domain , which is catalytically active and likely requires flexibility for its peptidase activity . Thus , the plant-specific supercomplex interface may have evolved to accommodate the particularities of plant CIII2’s dual respiratory and processing functions . Nevertheless , whereas the details differ , the overall location of the CIII2:CIV interface in V . radiata and yeast is similar . A related observation has been made for the supercomplexes between CIII2 and CI ( SCI+III2 ) of plants , yeast and mammals as seen by sub-tomogram averaging ( Davies et al . , 2018 ) . In this case , although the interfaces between CI and CIII2 in the different organisms were also similar , there was a ~10° difference in the angle between CI and CIII2 . Additional functional/structural studies of supercomplexes from organisms of diverse phylogenetic origins could determine whether the location of the supercomplex interface has been achieved by convergent or divergent evolution . In turn , this would shed light on the evolution and potential functional significance of the interface sites . What can already be concluded is that—as seen in yeast ( Hartley et al . , 2019; Rathore et al . , 2019 ) —the benefit of the SC III2+IV arrangement in plants does not involve direct electron transfer from CIII2 to CIV by simultaneously bound cyt c on each complex , as the calculated distance between the bound cyt c is too large ( ~70 Å , Figure 8—figure supplement 1 ) . Recent quantitative-proteomics estimations of the stoichiometry of plant respiratory-chain components indicate that the average plant mitochondrion contains ~6500 copies of CIII monomers ( i . e . ~3250 CIII2 ) , ~2000 copies of CIV and ~2250 copies of cytochrome c ( Fuchs et al . , 2020 ) . This implies a maximum of ~2000 copies of SC III2+IV and , thus , a roughly 1:1 ratio between cyt c and SC III2+IV . ( The ratio has been estimated to be 2–3 in S . cerevisiae [Stuchebrukhov et al . , 2020] ) . Based on recent theoretical analyses of electron flow between CIII2 and CIV ( Stuchebrukhov et al . , 2020 ) , at this low 1:1 ratio , electron flow would be limited by the time constant of cyt c equilibrating with the bulk IMS phase . Under these conditions , the formation of SC III2+IV in the plant mitochondrion would provide a kinetic advantage to electron flow between CIII2 and CIV by reducing the distance between them relative to CIII2 and CIV freely diffusing in the plane of the membrane . It is important to note that this possible kinetic advantage does not imply substrate trapping or channeling between the complexes , and is thus consistent with a single cyt c pool ( Stuchebrukhov et al . , 2020 ) . Our work provides the first high-resolution structure of SC III2+IV in plants , revealing plant-specific features of the complexes and supercomplex . Detailed comparisons of plant CIII2 and CIV sites with existing structures of inhibitor-bound complexes in other species will allow for the development of more selective inhibitors for plant CIII2 and CIV , frequently used as agricultural herbicides and pesticides ( Esser et al . , 2014 ) . The structures also allow for the generation of new mechanistic hypotheses—for example , related to proton translocation in CIV—and a re-evaluation of long-standing assumptions in the field—for instance , related to CIII’s capacity for long-range coordinated motion and the relationship between its respiratory and processing functions . Together with biochemical , cellular and genetic studies , further comparative analyses of these atomic structures with the growing number of respiratory complexes and supercomplexes across the tree of life will allow for the derivation of the fundamental principles of the respiratory electron transport chain .
The cryoEM dataset used in this paper is the same sample , grid and micrographs as those used in Maldonado et al . , 2020 . CryoEM data processing for the structures reported in this paper and those reported in Maldonado et al . , 2020 diverged after 2D classification ( see Figure 1—figure supplement 1 ) . Further data processing for the structures shown here is described in detail below . V . radiata seeds were purchased from Todd’s Tactical Group ( Las Vegas , Nevada , USA ) . Seeds were incubated in 1% ( v:v ) bleach for 20 min and rinsed until the water achieved neutral pH . Seeds were subsequently imbibed in a 6 mM CaCl2 solution for 20 hr in the dark . The following day , the imbibed seeds were sown in plastic trays on damp cheesecloth layers , at a density of 0 . 1 g/cm2 and incubated in the dark at 20 °C for 6 days . The resulting etiolated mung beans were manually picked , and the hypocotyls were separated from the roots and cotyledons . The hypocotyls were further processed for mitochondria purification based on established protocols ( Millar et al . , 2007 ) . Briefly , hypocotyls were homogenized in a Waring blender with homogenization buffer ( 0 . 4 M sucrose , 1 mM EDTA , 25 mM MOPS-KOH , 10 mM tricine , 1% w:v PVP-40 , freshly added 8 mM cysteine and 0 . 1% w:v BSA , pH 7 . 8 ) before a centrifugation of 10 min at 1000 x g ( 4 °C ) . The supernatant was collected and centrifuged for 30 min at 12 , 000 x g ( 4 °C ) . The resulting pellet was resuspended with wash buffer ( 0 . 4 M sucrose , 1 mM EDTA , 25 mM MOPS-KOH , freshly added 0 . 1% w:v BSA , pH 7 . 2 ) and gently centrifuged at 1000 x g for 5 min ( 4 °C ) . This supernatant was then centrifuged for 45 min at 12 , 000 x g . The resulting pellet was resuspended in wash buffer , loaded on to sucrose step gradients ( 35% w:v , 55% w:v , 75% w:v ) and centrifuged for 60 min at 52 , 900 x g . The sucrose gradients were fractionated with a BioComp Piston Gradient Fractionator connected to a Gilson F203B fraction collector , following absorbance at 280 nm . The fractions containing mitochondria were pooled , diluted 1:5 in 10 mM MOPS-KOH , 1 mM EDTA , pH 7 . 2 and centrifuged for 20 min at 12 , 000 x g ( 4 °C ) . The pellet was resuspended in final resuspension buffer ( 20 mM HEPES , 50 mM NaCl , 1 mM EDTA , 10% glycerol , pH 7 . 5 ) and centrifuged for 20 min at 16 , 000 x g ( 4 °C ) . The supernatant was removed , and the pellets were frozen and stored in a −80 °C freezer . The yield of these mitochondrial pellets was 0 . 8–1 mg per gram of hypocotyl . Frozen V . radiata mitochondrial pellets were thawed at 4°C , resuspended in 10 ml of chilled ( 4 °C ) double-distilled water per gram of pellet and homogenized with a cold Dounce glass homogenizer on ice . Chilled KCl was added to the homogenate to a final concentration of 0 . 15 M and further homogenized . The homogenate was centrifuged for 45 min at 32 , 000 x g ( 4 °C ) . The pellets were resuspended in cold Buffer M ( 20 mM Tris , 50 mM NaCl , 1 mM EDTA , 2 mM DTT , 0 . 002% PMSF , 10% glycerol , pH 7 . 4 ) and further homogenized before centrifugation at 32 , 000 x g for 45 min ( 4 °C ) . The pellets were resuspended in 3 ml of Buffer M per gram of starting material and further homogenized . The protein concentration of the homogenate was determined using a Pierce BCA assay kit ( Thermo Fisher , Waltham , Massachusetts , USA ) , and the concentration was adjusted to a final concentration of 10 mg/ml and 30% glycerol . Washed membranes were thawed at 4°C . Digitonin ( EMD Millipore , Burlington , Massachusetts , USA ) was added to the membranes at a final concentration of 1% ( w:v ) and a digitonin:protein ratio of 4:1 ( w:w ) . Membrane complexes were extracted by tumbling this mixture for 60 min at 4 °C . The extract was centrifuged at 16 , 000 x g for 45 min ( 4 °C ) . Amphipol A8-35 ( Anatrace , Maumee , Ohio , USA ) was added to the supernatant at a final concentration of 0 . 2% w:v and tumbled for 30 min at 4°C , after which gamma-cyclodextrin ( EMD Millipore , Burlington , Massachusetts , USA ) was added stepwise to a final amount of 1 . 2x gamma-cyclodextrain:digitonin ( mole:mole ) . The mixture was centrifuged at 137 , 000 x g for 60 min ( 4 °C ) . The supernatant was concentrated with centrifugal protein concentrators ( Pall Corporation , NY , NY , USA ) of 100 , 000 MW cut-off , loaded onto 10–45% ( w:v ) or 15–45% ( w:v ) linear sucrose gradients in 15 mM HEPES , 20 mM KCl , pH 7 . 8 produced using factory settings of a BioComp Instruments ( Fredericton , Canada ) gradient maker and centrifuged for 16 hr at 37 , 000 x g ( 4 °C ) . The gradients were subsequently fractionated with a BioComp Piston Gradient Fractionator connected to a Gilson F203B fraction collector , following absorbance at 280 nm . For grid preparation , the relevant fractions were buffer-exchanged into 20 mM HEPES , 150 mM NaCl , 1 mM EDTA , pH 7 . 8 ( no sucrose ) and concentrated to a final protein concentration of 6 mg/ml and mixed one-to-one with the same buffer containing 0 . 2% digitonin ( w:v ) , for a final concentration of 0 . 1% digitonin ( w:v ) . Mitochondrial membrane extractions were diluted in 2X BN-loading buffer ( 250 mM aminocaproic acid , 100 mM Tris-HCl , pH 7 . 4 , 50% glycerol , 2 . 5% ( w:v ) Coomassie G-250 ) , loaded on pre-cast 3–12% NativePAGE Bis-Tris gels ( Invitrogen , Carlsbad , CA ) and run at 4 °C . The cathode buffer was 50 mM Tricine , 50 mM BisTris-HCl , pH 6 . 8 plus 1X NativePAGE Cathode Buffer Additive ( 0 . 02% Coomassie G-250 ) ( Invitrogen , Carlsbad , CA ) and the anode buffer was 50 mM Tricine , 50 mM BisTris-HCl , pH 6 . 8 . Gels were run at 150 V constant voltage for ∼30 min , after which the cathode buffer was switched for a ‘light blue’ cathode buffer containing 50 mM Tricine , 50 mM BisTris-HCl , pH 6 . 8 plus 0 . 1X NativePAGE Cathode Buffer Additive ( 0 . 002% Coomassie G-250 ) ( Invitrogen , Carlsbad , CA ) . The settings were changed to 200 V constant voltage and run for another ∼90 min . The CI in-gel NADH-dehydrogenase activity assay was performed based on Schertl and Braun , 2015 . The BN-PAGE gel was incubated in 10 ml of freshly prepared reaction buffer ( 1 . 5 mg/ml nitrotetrazoleum blue in 10 mM Tris-HCl pH 7 . 4 ) . Freshly thawed NADH stock ( 20 mM ) was added to the container with the gel , to a final concentration of 150 μM . The gel with the complete reaction buffer was rocked at room temperature for ∼10 min . Once purple bands indicating NADH-dehydrogenase activity appeared , the reaction was quenched with a solution of 50% methanol ( v:v ) and 10% acetic acid ( v:v ) . Spectroscopic activity assays were performed based on Letts et al . , 2019; Huang et al . , 2015; Barrientos et al . , 2009 , with modifications . Reduced-decylubiquinone ( DQ ) : cyt c activity was measured by spectroscopic observation of cyt c reduction at 550 nm wavelength at room temperature using a Molecular Devices ( San Jose , CA ) Spectramax M2 spectrophotometer . Reactions were carried out in 96-well plates . DQ ( Santa Cruz Biotechnology , Dallas , TX ) was freshly reduced . The required amount of ethanol-diluted 100 mM DQ was aliquoted and further diluted to ~300 μl with 100% ethanol . A couple of lithium borohydride crystals were added to reduce the DQ , turning the solution transparent . Excess lithium borohydride was quenched by the dropwise addition of 1 N HCl until no further bubbles were observed . The ethanol was then evaporated with a stream of argon gas until a volume of ~50 μl was obtained . This reduced-DQ was added to a master mix reaction buffer ( 100 mM HEPES , pH 7 . 8 , 50 mM NaCl , 10% glycerol , 0 . 1% CHAPS , 1 mg/ml BSA , 0 . 25 mg/ml 4:1 asolectin:cardiolipin in 0 . 1% CHAPS , 25 U/ml SOD , 4 μM KCN , 15 μM piericidin ) at a final concentration of 100 μM and mixed by vortexing . The pH of the reaction buffer master mix was checked to be 7–8 with pH strips . The reaction master mix was aliquoted , and CIII2 inhibitors antimycin A and myxothiazol were added at 1 μM final concentration where relevant and mixed by vortexing . Protein samples ( 5 μg ) were added to the respective aliquots of reaction buffer to a total volume of 200 μl and mixed by vortexing . The reaction was initiated by addition of equine cyt c ( Sigma Aldrich , St Louis , MO ) to a final concentration of 100 μM , briefly mixed by pipetting and plate stirring for 10 s before recording for 3 min every 4 s . Measurements were done in 3–5 replicates , averaged and background-corrected . An extinction co-efficient of 28 mM−1 cm−1 ( Huang et al . , 2015 ) was used in the calculations . Statistical significance was determined using two-tailed t-tests . Spectroscopic activity assays were performed based on Letts et al . , 2019; Huang et al . , 2015; Barrientos et al . , 2009 , with modifications . Cytochrome c oxidase activity was measured by spectroscopic observation of the oxidation of reduced cyt c at 550 nm wavelength at room temperature using a Molecular Devices ( San Jose , CA ) Spectramax M2 spectrophotometer . Reactions were carried out in 96-well plates . Equine cyt c ( Sigma Aldrich , St Louis , MO ) , diluted in 20 mM HEPES , pH 7 . 4 , 50 mM NaCl , 10% glycerol buffer , was freshly reduced based on manufacturer’s instructions with modifications . Dithiothreitol ( DTT ) was added at a 10 mM final concentration . After ~20 min and a visible change in color , cyt c reduction was confirmed spectroscopically . Given the spectrophotometer’s specifications , the simultaneous measurement of A550 and A565 was suboptimal ( A550:A565 ratio of ~9 ) . Therefore , the A550:A575 was measured instead , as per manufacturer’s recommendations . Cyt c reduction was confirmed at A550:A575 ratio ~22–24 . For the spectroscopic activity assay , the reaction master mix consisted of 20 mM HEPES , pH 7 . 4 , 50 mM NaCl , 10% glycerol , 0 . 1% CHAPS ( w:v ) with additional 4 μM KCN wherever appropriate . Protein samples ( 5 μg ) were added to the respective aliquots of reaction buffer to a total volume of 200 μl and mixed by vortexing . The reaction was initiated by addition of reduced cyt c to a final concentration of 100 μM , briefly mixed by pipetting and plate stirring for 10 s before recording for 3 min every 4 s . Measurements were done in 3–4 replicates , averaged and background-corrected . An extinction co-efficient of 28 mM−1 cm−1 ( Huang et al . , 2015 ) was used in the calculations . Statistical significance was determined using two-tailed t-tests . The sample used for mass spectrometry was the sample used to blot the cryoEM grid that was used for here and in Maldonado et al . , 2020 . This sample corresponds to concentrated , pooled peak two fractions from the sucrose gradient shown in Maldonado et al . , 2020 ( fractions 10–11 , Figure 1—figure supplement 2H ) . This sample is roughly equivalent to fractions 11–13 from Figure 1—figure supplement 4 here . Thus , the mass spectrometry results of this mixed sample include complex I subunits in addition to CIII2 and CIV subunits . See below for full dataset availability and accession codes . Samples were digested with the S-Trap micro ( PROTIFI ) digestion . Digestion followed the S-trap protocol . The proteins were reduced and alkylated , the buffer concentrations were adjusted to a final concentration of 5% SDS , 50 mM TEAB , 12% phosphoric acid was added at a 1:10 ( v:v ) ratio with a final concentration of 1 . 2% and S-trap buffer ( 100 mM TEAB in 90% MEOH ) was added at a 1:7 ratio ( v:v ) ratio . The protein lysate S-trap buffer mixture was then spun through the S-trap column and washed three times with S-Trap buffer . Finally , 50 mM TEAB and 1 µg of trypsin ( 1:25 ratio ) was added and the sample was incubated overnight with one addition of 50 mM TEAB and trypsin after two hours . The following day the digested peptides were released from the S-trap solid support by spinning at 1 min for 3000 x g with a series of solutions starting with 50 mM TEAB which is placed on top of the digestion solution then 5% formic acid followed by 50% acetonitrile , 0 . 1% formic acid . The solution was then vacuum centrifuged to almost dryness and resuspended in 2% acetonitrile , 0 . 1% triflouroacetic acid ( TFA ) and subjected to Fluorescent Peptide Quantification ( Pierce ) . Digested peptides were analyzed by LC-MS/MS on a Thermo Scientific Q Exactive plus Orbitrap Mass spectrometer in conjunction Proxeon Easy-nLC II HPLC ( Thermo Scientific ) and Proxeon nanospray source . The digested peptides were loaded on a 100 micron x 25 mm Dr . Masic reverse phase trap where they were desalted online before being separated using a 75 micron x 150 mm Magic C18 200 Å 3U reverse phase column . Peptides were eluted using a 70 min gradient with a flow rate of 300 nL/min . An MS survey scan was obtained for the m/z range 300–1600 , MS/MS spectra were acquired using a top 15 method , where the top 15 ions in the MS spectra were subjected to HCD ( High Energy Collisional Dissociation ) . An isolation mass window of 2 . 0 m/z was used for the precursor ion selection , and normalized collision energy of 27% was used for fragmentation . A twenty second duration was used for the dynamic exclusion . Tandem mass spectra were extracted and charge state deconvoluted by Proteome Discoverer ( Thermo Scientific ) . All MS/MS samples were analyzed using X ! Tandem ( The GPM , thegpm . org; version X ! Tandem Alanine ( 2017 . 2 . 1 . 4 ) ) . X ! Tandem was set up to search the Uniprot Vigna radiata database ( October 2019 , 35065 entries ) the cRAP database of common laboratory contaminants ( http://www . thegpm . org/crap; 117 entries ) plus an equal number of reverse protein sequences assuming the digestion enzyme trypsin . X ! Tandem was searched with a fragment ion mass tolerance of 20 PPM and a parent ion tolerance of 20 PPM . Carbamidomethyl of cysteine and selenocysteine was specified in X ! Tandem as a fixed modification . Glu->pyro Glu of the n-terminus , ammonia-loss of the n-terminus , gln->pyro Glu of the n-terminus , deamidated of asparagine and glutamine , oxidation of methionine and tryptophan and dioxidation of methionine and tryptophan were specified in X ! Tandem as variable modifications . Scaffold ( version Scaffold_4 . 8 . 4 , Proteome Software Inc , Portland , OR ) was used to validate MS/MS based peptide and protein identifications . Peptide identifications were accepted if they could be established at greater than 98 . 0% probability by the Scaffold Local FDR algorithm . X ! Tandem identifications required score of at least 2 . Protein identifications were accepted if they could be established at greater than 6 . 0% probability to achieve an FDR less than 5 . 0% and contained at least two identified peptides . Protein probabilities were assigned by the Protein Prophet algorithm ( Nesvizhskii et al . , 2003 ) . Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony . Proteins sharing significant peptide evidence were grouped into clusters . The sample ( 6 mg/ml protein in 20 mM HEPES , 150 mM NaCl , 1 mM EDTA , 0 . 1% digitonin , pH 7 . 8 ) was applied onto glow-discharged holey carbon grids ( Quantifoil , 1 . 2/1 . 3 300 mesh ) followed by a 60 s incubation and blotting for 9 s at 15°C with 100% humidity and flash-freezing in liquid ethane using a FEI Vitrobot Mach III . CryoEM data acquisition was performed on a 300 kV Titan Krios electron microscope equipped with an energy filter and a K3 detector at the UCSF W . M . Keck Foundation Advanced Microscopy Laboratory , accessed through the Bay Area CryoEM Consortium . Automated data collection was performed with the SerialEM package ( Schorb et al . , 2019 ) . Micrographs were recorded in super-resolution mode at a nominal magnification of 60 , 010 X , resulting in a pixel size of 0 . 8332 Å2 . Defocus values varied from 1 . 5 to 3 . 0 µm . The dose rate was 20 electrons per pixel per second . Exposures of 3 s were dose-fractionated into 118 frames , leading to a dose of 0 . 72 electrons per Å2 per frame and a total accumulated dose of 51 electrons per Å2 . A total of 9816 micrographs were collected . Software used in the project was installed and configured by SBGrid ( Morin et al . , 2013 ) . All processing steps were done using cryoSPARC and RELION 3 . 0 ( Zivanov et al . , 2018; Punjani et al . , 2017 ) unless otherwise stated . Motioncor2 ( Zheng et al . , 2017 ) was used for whole-image drift correction of each micrograph . Contrast transfer function ( CTF ) parameters of the corrected micrographs were estimated using Ctffind4 ( Rohou and Grigorieff , 2015 ) . After motion correction and CTF correction , a set of 8541 micrographs was selected for further processing . Automated particle picking using crYOLO ( Wagner et al . , 2019; Wagner and Raunser , 2020 ) resulted in ~1 . 5 million particles . The particles were extracted using 4002 pixel box binned two-fold and sorted by reference-free 2D classification in Relion using ( --max_sig 5 ) , followed by re-extraction at 5122 pixel box . Reference-free 2D classification in Relion resulted in the identification of 502 , 224 particles that were then imported into cryoSPARC for further reference-free 2D classification . A set of 121 , 702 particles were identified by 2D classification in cryoSPARC to contain CIII2 alone or SC III2+IV ( Figure 1—figure supplement 1 ) . These particles were subjected to ab initio model generation with four targets to remove contaminant particles resulting in a set of 99 , 937 particles across three classes . Each individual class was subjected to an additional round of ab initio model generation with three targets . This separated CIII2 alone particles from the SC III2+IV class and allowed the recovery of CIII2 alone particles from the poor particle class . CIII2 alone particles from across the ab initio model generation jobs were pooled , defining a final class of 48 , 111 particles . The multiple rounds of ab initio model generation resulted in only one good class of 28 , 020 SC III2+IV particles . Poses for these two particle sets ( CIII2 alone and SC III2+IV ) were refined using cryoSPARC’s Homogeneous Refinement ( New ) algorithm including Defocus Refinement and Global CTF refinement . This resulted in reconstructions at 3 . 2 Å and 3 . 8 Å resolution for CIII2 alone and SC III2+IV respectively , according to the gold standard FSC criteria ( Figure 1—figure supplement 1; Scheres and Chen , 2012 ) . In parallel , a set of 69 , 876 particles were identified by further 2D and 3D classification in Relion ( Figure 1—figure supplement 2 ) . These particles , which contained a mixture of SC III2+IV and CIII2 alone particles , were aligned using a SC III2+IV model and mask . They were then subjected to five rounds of masked classification using a CIV model and mask aligned with the position of CIV in the SC . Three parallel masked classifications ( all using T = 8 ) varied in the degree of rotational searches , with either no searches , 0 . 1° sampling interval over +/- 0 . 2° search range and 3 . 7° sampling interval over +/- 7 . 5° search range . The masked 3D classification without searches was repeated successively for three rounds , inputting the best particles from the previous round into the subsequent round . The best CIV class from each 3D classification were selected and combined while removing overlaps ( any particle within 200 pixels of another was considered as an overlap and discarded ) . This 3D classification strategy resulted in a set of 38 , 410 particles . The coordinates of these particles were used to extract two sets of re-centered SC particles , one centered on CIII2 and one centered on CIV . These two sets of particles were independently 3D-refined , CTF-refined and Bayesian-polished using a model and mask centered around CIII2 or CIV respectively . The CIV-centered shiny particles were subjected to a final round of 3D classification , defining a final set of 29 , 348 CIV particles . Although this final round of 3D classification did not improve the nominal resolution of the map , it increased map quality at the periphery of the complex . These final CIII2 and CIV classes resulted in reconstructions at 3 . 7 Å and 3 . 8 Å resolution for CIII2 and IV from the SC respectively , according to the gold standard FSC criteria ( Figure 1—figure supplement 1; Scheres and Chen , 2012 ) . These two maps were aligned with the full SC map and combined to make a composite map using Phenix . 3D variability analysis ( 3DVA ) was performed on CIII2 alone in cryoSPARC using their built-in algorithm ( Punjani and Fleet , 2020 ) . Two separate instances of 3DVA were performed , each solving for the four largest principal components . The first instance used a mask around the entire CIII2 and data was low-pass filtered to 6 Å resolution to remove the influence of high-resolution noise from the amphipol detergent belt . The second instance used a mask focused around the IMS domain of CIII2 and was low-pass filtered to 5 Å resolution . All other parameters were kept as default . Starting template models for V . radiata CIII2 was ovine CIII2 ( PDB: 6Q9E ) . Starting template models for V . radiata CIV were from S . cerevisiae ( PDB: 6HU9 ) and bovine ( PDB: 5B1A ) CIV . Additionally , starting models for the V . radiata subunits were generated using the Phyre2 web portal ( Kelley et al . , 2015 ) . Real-space refinement of the model was done in PHENIX ( Liebschner et al . , 2019; Goddard et al . , 2018; Pettersen et al . , 2004 ) and group atomic displacement parameters ( ADPs ) were refined in reciprocal space . The single cycle of group ADP refinement was followed by three cycles of global minimization , followed by an additional cycle of group ADP refinement and finally three cycles of global minimization ( Letts et al . , 2019 ) . The refined CIII2 and CIV models were docked into the SC III2+CIV map without subsequent refinement . Molecular graphics and analyses were performed with UCSF Chimera ( Pettersen et al . , 2004 ) and ChimeraX ( Goddard et al . , 2018 ) developed by the Resource for Biocomputing , Visualization , and Informatics at the University of California , San Francisco , with support from NIH P41-GM103311 and R01-GM129325 and the Office of Cyber Infrastructure and Computational Biology , National Institute of Allergy and Infectious Diseases . PyMOL Molecular Graphics System , Version 2 . 0 Schrödinger , LLC was also used . | Most living things including plants and animals use respiration to release energy from food . Respiration requires the activity of five large protein complexes typically called complex I , II , III , IV and V . Sometimes these complexes combine to form supercomplexes . The complexes are similar across plants , animals and other living things , but there are also many differences . Detailed structures of the respiratory complexes have been determined for many species of animals , fungi and bacteria , highlighting similarities and differences between organisms , and providing clues as to how respiration works . Yet , there is still a lot to learn about these complexes in plants . To bridge this gap , Maldonado et al . used a technique called cryo electron microscopy to study the structure of complexes III and IV and the supercomplex they form in the mung bean . This is the first study of the detailed structure of these two complexes in plants . The results showed many similarities to other species , as well as several features that are specific to plants . The way the two complexes interact to form a supercomplex is different than in other species , as are several other , smaller , structural features . Further examination of complex III revealed that it is flexible and that movements are coordinated across the length of the complex . Maldonado et al . speculate that this may allow it to coordinate its role in respiration with its other cellular roles . Understanding how plant respiratory complexes work could lead to improvements in crop yields or , since respiration is required for survival , result in the development of herbicides that block respiration in plants more effectively and specifically . Further researching the structure of the plant respiratory complexes and supercomplexes could also shed light on how plants adapt to different environments , including how they change to survive global warming . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2021 | Atomic structures of respiratory complex III2, complex IV, and supercomplex III2-IV from vascular plants |
The large number of individuals placed into quarantine because of possible severe acute respiratory syndrome coronavirus 2 ( SARS CoV-2 ) exposure has high societal and economic costs . There is ongoing debate about the appropriate duration of quarantine , particularly since the fraction of individuals who eventually test positive is perceived as being low . We use empirically determined distributions of incubation period , infectivity , and generation time to quantify how the duration of quarantine affects onward transmission from traced contacts of confirmed SARS-CoV-2 cases and from returning travellers . We also consider the roles of testing followed by release if negative ( test-and-release ) , reinforced hygiene , adherence , and symptoms in calculating quarantine efficacy . We show that there are quarantine strategies based on a test-and-release protocol that , from an epidemiological viewpoint , perform almost as well as a 10-day quarantine , but with fewer person-days spent in quarantine . The findings apply to both travellers and contacts , but the specifics depend on the context .
Quarantining individuals with high risk of recent infection is one of the pillars of the non-pharmaceutical interventions to control the ongoing severe acute respiratory syndrome coronavirus 2 ( SARS-CoV-2 ) pandemic ( Kucharski et al . , 2020 ) . Owing to the large fraction of transmission of SARS-CoV-2 that is pre-symptomatic or asymptomatic ( Ashcroft et al . , 2020; Buitrago-Garcia et al . , 2020; Ferretti et al . , 2020b; He et al . , 2020 ) , quarantine can prevent a substantial fraction of onward transmission that would not be detected otherwise . In mathematical modelling studies , it was estimated that thermal screening at airports would allow more than 50% of infected travellers to enter the general population ( Quilty et al . , 2020; Gostic et al . , 2020 ) , which could have been prevented by mandatory quarantine . Quarantine is also a fundamental part of the test–trace–isolate–quarantine ( TTIQ ) intervention strategy to break chains of transmission within a community ( Salathé et al . , 2020 ) . With the high or increasing case numbers that are observed in many places around the globe , however , more and more people are being placed into quarantine . There is ongoing public debate about the appropriateness of quarantine and its duration . Quarantine lowers onward transmission in two ways: first , preventing transmission prior to symptom onset ( with the assumption that symptomatic individuals will isolate ) and decreasing overall transmission from persistently asymptomatic individuals . The appropriate length of quarantine thus depends on both incubation period and the temporal profile of infectiousness . In theory , quarantine periods could be avoided altogether through widespread and regular testing programmes , but the low sensitivity of reverse transcriptase PCR ( RT-PCR ) tests , particularly in early infection ( Kucirka et al . , 2020 ) , as well as limitations on testing capacity in most countries preclude this approach . Quarantine has high economic , societal , and psychological costs ( Nicola et al . , 2020; Brooks et al . , 2020 ) . It restricts individual freedoms ( Parmet and Sinha , 2020 ) , although the level of restriction imposed is generally judged to be proportionate , given the severity of coronavirus disease 2019 ( COVID-19 ) . The low number of individuals placed in quarantine that turn out to be infected is another argument that is given against quarantine . Individuals are generally placed into quarantine for one of two reasons: either they have been identified as a recent close contact of a confirmed SARS-CoV-2 case by contact tracing , or they have returned from recent travel to an area with community transmission that has been assessed to pose significant epidemiological risk ( WHO , 2020 ) . These groups of quarantined individuals differ in two important ways: compared with traced contacts , travel returners may have lower probability of being infected and have less precise information about the likely time of exposure . This raises the question whether the duration of quarantine should be the same for these two groups of individuals . To our knowledge , there are no published analyses of surveillance data that directly assess the impact of duration of quarantine on SARS-CoV-2 transmission ( Nussbaumer-Streit et al . , 2020 ) . In this study , we present a mathematical model that allows quantification of the effects of changing quarantine duration . We use the distributions of incubation time ( time from infection to onset of symptoms ) , infectivity ( infectiousness as a function of days since symptom onset ) , and generation time ( difference of timepoints of infection between infector and infectee ) . These distributions have been estimated by Ferretti et al . , 2020b , combining multiple empirical studies of documented SARS-CoV-2 transmission pairs ( Ferretti et al . , 2020a; Xia et al . , 2020; Cheng et al . , 2020; He et al . , 2020 ) . Using the model , we explore the effect of duration of quarantine for both traced contacts of confirmed SARS-CoV-2 cases and for returning travellers on the fraction of prevented onward transmission . We assess the effects of test-and-release strategies and the time delay between test and result . These considerations are particularly important given that multiple testing has been shown to be of little benefit ( Clifford , 2020 ) . We also address the role of pre-symptomatic patients becoming symptomatic and therefore being isolated independent of quarantine . Furthermore , as one of the arguments for shortening the duration of quarantine is to increase the number of people complying with the recommendation , we investigate by how much adherence needs to increase to offset the increased transmission due to earlier release from quarantine . Finally , we assess the role of reinforced individual-level prevention measures , such as mask wearing , for those released early from quarantine . Making policy decisions about the duration of quarantine fundamentally requires specifying how the effectiveness of quarantine relates to its costs . The effectiveness can be measured in terms of the overall reduction of transmission , while economic , societal , and individual costs are likely a function of the number of days spent in quarantine . In addition to the epidemiological outcome , which considers only the reduction in transmission , we also present results based on the ratio of transmission prevented to the average number of days spent in quarantine .
In the absence of quarantine , individuals that are infected with SARS-CoV-2 can infect further individuals in the population . In the model , the timing of onward transmission from an infected individual is determined by the generation time distribution , which describes the time interval between the infection of an infector and infectee ( see Figure 1—figure supplement 1 ) . To quantify how much transmission is prevented by quarantining individuals who have been infected with SARS-CoV-2 , we need to know the time at which the individual was exposed ( t_E ) , as well as when they enter ( t_Q ) and are released from ( t_R ) quarantine . The fraction of transmission that is prevented by quarantine is then the total transmission probability ( i . e . the area under the curve ) that lies between t_Q and t_R ( Figure 1 ) . We refer to this fraction of prevented transmission as quarantine efficacy and is defined in Equation ( 1 ) in 'Materials and methods' . Unless otherwise stated , we assume that adherence to quarantine is 100% . Under the standard quarantine strategy , all potentially exposed individuals are quarantined for the same duration . An alternative approach is the test-and-release strategy , which uses virological testing during quarantine to release individuals with a negative test result earlier . Individuals with a positive test result are isolated until they are no longer infectious . The timing of the test ( t_T ) is important due to the substantial false-negative rate of the RT-PCR test in the early stages of infection ( Kucirka et al . , 2020 ) . A false-negative test result would release an infected individual into the community prematurely , leading to further transmission ( Figure 1A ) . In this case , quarantine efficacy is defined as the expected fraction of transmission that is prevented by quarantine across false-negative and positive testing individuals , as defined in Equation ( 2 ) in 'Materials and methods' . As well as the epidemiological benefit of quarantine ( i . e . the fraction of transmission prevented by quarantining an infected individual ) , we can also quantify the economic and societal costs in terms of the expected number of person-days spent in quarantine . We can then define the utility of a quarantine strategy as the ratio between the quarantine efficacy and the average time spent in quarantine , that is , the transmission prevented per day spent in quarantine , as defined in Equation ( 4 ) in 'Materials and methods' . This utility measure is dependent on the fraction of individuals in quarantine that are infected . This definition of utility should be considered as an example of such a utility function , but this may not be the best way to quantify quarantine utility . Details of the calculations used can be found in 'Materials and methods' . Further extensions to the model , including the role of reinforced hygiene measures , asymptomatic infections , and adherence to quarantine , are described in Appendix 1 . Traced contacts have a known ( last ) time of exposure to a confirmed case . There is usually a delay between this exposure time and the start of quarantine . Under the standard quarantine protocol , traced contacts are released from quarantine once a number of days have passed after the last exposure time . In Switzerland , for example , quarantine lasts until 10 days after the last exposure . Any shortening of a traced contact’s quarantine duration will lead to an increase in transmission from that individual if they are infected , but the degree of increase depends on the extent of the shortening . The expected onward transmission that is prevented by quarantine shows the diminishing return of increasing the quarantine duration ( black line in Figure 2A ) . Increasing quarantine duration beyond 10 days shows almost no additional benefit ( Figure 2—figure supplement 1A ) : the standard quarantine protocol ( here with a 3-day delay between exposure and the start of quarantine ) can maximally prevent 90 . 8% [95% CI: 79 . 6% , 97 . 6%] of onward transmission from an infected traced contact , while release on day 10 prevents 90 . 1% [CI: 76 . 0% , 97 . 5%] . The maximum attainable prevention also applies to the test-and-release strategy: the onward transmission prevented under a test-and-release strategy will always be below this level ( coloured lines in Figure 2A ) . This is because of the chance of prematurely releasing an infectious individual who received a false-negative test result . On the other hand , it is always better to test a person prior to release from quarantine so that individuals with asymptomatic and pre-symptomatic infections can be detected and prevented from being released . Hence , these scenarios provide upper and lower bounds for the efficacy of the test-and-release strategy . The fraction of transmission that is prevented increases if we test later in quarantine because we not only increase the duration of quarantine but also reduce the false-negative probability . The delay between testing and release from quarantine can have a substantial effect on the efficacy . Current laboratory-based RT-PCR tests have a typical turnaround of 24–48 hr ( Quilty et al . , 2021 ) . This delay is primarily operational , and so could be reduced by increasing test throughput . There are also rapid antigen-detection tests , which can provide same-day results , but with lower sensitivity and specificity than RT-PCR tests ( Guglielmi , 2020 ) . Here we assume that tests have the same sensitivity and specificity regardless of the delay to result . Compared to a test with 2-day delay until result , we observe that using a rapid test with same-day release can reduce the quarantine duration of individuals with a negative test result by 1 day while maintaining the same efficacy ( Figure 2A ) : the extra accuracy gained by waiting one extra day until testing balances the increased transmission caused by reducing the duration . For example , a rapid test on day 6 has roughly the same efficacy ( 80 . 5% [CI: 67 . 9% , 88 . 7%] ) as testing on day 5 and releasing on day 7 ( 82 . 3% [CI: 68 . 2% , 93 . 4%] ) while shortening the quarantine duration of individuals with a negative test result from 7 to 6 days . In Figure 2 we have assumed a fixed delay of 3 days between exposure and the start of quarantine . Shortening this delay increases the maximum efficacy of quarantine because pre-quarantine transmission is reduced ( Figure 2—figure supplement 1A ) . If the duration of quarantine is longer than 10 days , then little can be gained in terms of prevention by quarantining for longer , but reducing the delay between exposure and quarantine does lead to increased efficacy . Note that we have assumed that the contact was infected at the last time of exposure . If there have been multiple contacts between them and the index case , then transmission may have occurred earlier and we would overestimate the efficacy of quarantine . For the standard quarantine strategy , the duration of quarantine is independent of whether individuals in quarantine are infected . Therefore , the utility of the standard quarantine strategy ( i . e . the ratio of efficacy to duration ) is directly proportional to the fraction of individuals in quarantine that are infected . By comparing two different standard quarantine strategies through their relative utility ( i . e . the ratio of the utilities ) , we can eliminate the dependence on the fraction of infecteds in quarantine ( see 'Materials and methods' ) . Therefore , the argument that we should shorten quarantine because of the low probability of quarantined individuals being infected is misguided in this situation . By calculating the relative utility for the standard quarantine strategy compared to the baseline 10-day quarantine , we observe that there is a quarantine strategy ( release after 7 days ) which maximises the ratio between the fraction of transmission prevented and the number of days spent in quarantine ( black line in Figure 2B ) . The optimal strategy lies between 6 and 8 days if we vary the delay between exposure and the start of quarantine ( Figure 2—figure supplement 1B ) . Under the test-and-release quarantine protocol , the average time spent in quarantine is dependent on the fraction of infecteds in quarantine; only the infected individuals can test positive and face a longer period of isolation ( i . e . we assume there are no false-positive test results ) . Hence the utility of the test-and-release strategy , as well as the relative utility of test-and-release compared to the standard quarantine protocol , is dependent on the fraction of individuals in quarantine that are infected . In Figure 2B , we fix the fraction of infecteds in quarantine to 10% . By calculating the relative utility for the test-and-release quarantine strategies shown in Figure 2A compared to the baseline 10-day quarantine , we see that testing-and-releasing before day 10 always increases the utility ( Figure 2B ) . Testing on day 5 and releasing test-negative individuals on day 7 has a relative utility of 1 . 53 [CI: 1 . 45 , 1 . 62] compared to a standard 10-day quarantine . Reducing the delay between test and result leads to a corresponding increase in utility: a rapid test ( zero delay between test and result ) on day 6 has a relative utility of 1 . 90 [CI: 1 . 83 , 1 . 98] for an almost equivalent efficacy . In Figure 2 , we have made the following assumptions: ( i ) individuals released from quarantine have – in the post-quarantine phase – the same transmission probability as individuals who were not quarantined; ( ii ) adherence to quarantine is 100%; and ( iii ) the transmission prevented by quarantine for cases who develop symptoms is attributed to quarantine . We now relax these assumptions to assess their impact on quarantine efficacy . Reinforced prevention measures post-quarantine , where individuals who are released from quarantine must adhere to strict hygiene and social distancing protocols until 10 days after exposure have passed , can reduce post-quarantine transmission . Considering a 50% reduction of post-quarantine transmission leads to large increases in both efficacy and utility for early testing strategies , but with diminishing returns as the release date is increased towards day 10 ( Figure 2—figure supplement 2; see 'Appendix 1: Reinforced prevention measures after early release' ) . Note that we assume no contribution to the number of days spent in quarantine in the utility function due to mask wearing and social distancing in the post-release phase . Adherence to quarantine is unlikely to be 100% and could depend on the proposed duration of quarantine . For simplicity we treat adherence to quarantine as a binary variable: a fraction of individuals adhere to quarantine completely for the proposed duration , while the remaining fraction do not undergo any quarantine . We now ask: by how much would the fraction of those who adhere to quarantine have to increase to maintain the efficacy of quarantine if the duration is shortened ? In the absence of testing during quarantine , shortening from 10 to 5 days would require almost three times as many individuals to adhere to the quarantine guidelines in order to maintain the same overall efficacy ( relative adherence 2 . 90 [CI: 2 . 15 , 4 . 36]; black line in Figure 3A ) . This threefold increase is not possible if adherence to the 10-day strategy is already above 33% as the maximum adherence cannot exceed 100%; the required increase in adherence grows rapidly as quarantine is shortened and soon becomes infeasible . Hence the argument of shortening quarantine to increase adherence is of limited use . Shortening to 7 days ( without testing ) may be effective provided that adherence can increase by 30% ( relative adherence 1 . 30 [CI: 1 . 08 , 1 . 55] ) . Under the test-and-release strategy , however , the efficacy of the standard 10-day quarantine can be matched with release on day 5 or 6 if adherence is also increased by 30% . Releasing earlier than day 5 would seemingly be infeasible given the sharp increase in adherence required . As a final consideration , we note that our quantification of the fraction of transmission prevented by quarantine is more relevant to individuals with persistently asymptomatic SARS-CoV-2 infection than to those who develop symptoms during quarantine and are subsequently isolated . If symptomatic cases go into isolation once symptoms appear , then quarantine has no further impact on transmission after symptom onset as these cases would anyway be isolated . To account for this , we can modify the model such that cases are removed from the infectious pool upon symptom onset ( see Appendix 1 ) . For example , in a fully asymptomatic population a 10-day quarantine can prevent 90 . 1% [CI: 76 . 0% , 97 . 5%] of transmission . However , if 25% of cases are asymptomatic , then only 50 . 8% [CI: 42 . 8% , 56 . 5%] of transmission is prevented by quarantine , while 39 . 3% is prevented by the self-isolation of symptomatic cases ( Figure 3B ) . We assume that self-isolation occurs immediately after symptom onset , but any delay between symptom onset and self-isolation would mean that more transmission is prevented by quarantine ( Figure 3—figure supplement 1 ) . The fraction of transmission prevented by quarantine is an increasing function of the fraction of asymptomatic cases ( Figure 3B ) . This means that we likely overestimate the efficacy of quarantine as we are also counting transmission that could be prevented by isolation following symptom onset . Furthermore , we have assumed that the false-negative rate is the same between symptomatic and asymptomatic cases . If the test is less sensitive ( higher false-negative probability ) for asymptomatic cases , then quarantine efficacy would be further reduced . The rules for whether travellers returning from abroad are quarantined are frequently changed according to the epidemiological scenario in the travel destination and/or in the home country . A high risk of infection while abroad due to high prevalence , or the possibility of returning with a new virological variant , can lead to the imposition or reinstatement of quarantine measures ( Russell et al . , 2021 ) . Countries that have already eliminated the infection may be even stricter in their quarantine approach to prevent new community-transmission clusters from being seeded . Here we do not discuss these scenarios or the concept of relative risk , we simply quantify how effective quarantine strategies would be at preventing transmission if the returning traveller was infected while abroad . Should quarantine rules be instated or modified , these results can help determine the optimal quarantine duration and/or testing strategy . The timing of infection of a traveller during a trip abroad is generally unknown . We assume that infection could have happened on each day of the trip with equal probability . Quarantine begins immediately upon return , which we refer to as day 0 , and lasts for a number of days ( e . g . currently 10 days in Switzerland ) from this timepoint ( Figure 1B ) . We consider the fraction of local transmission that is prevented by quarantine . That is , the fraction of the transmission that could occur in the local country that is prevented by quarantine [Equation ( 8 ) ] . For a 7-day trip , as in Figure 4 , the maximum transmission that could occur in the local country is 73 . 3% [CI: 65 . 7% , 80 . 3%] . The remaining infectivity potential was already used up before arrival . A standard ( no test ) 10-day quarantine will prevent 99 . 9% [CI: 98 . 0% , 100 . 0%] of local transmission if the individual was infected during a 7-day trip ( Figure 4A ) . There is little benefit to gain by increasing the duration of quarantine beyond 10 days . On the other hand , standard quarantine efficacy decreases quickly as the duration is shortened . The test-and-release strategy can improve the efficacy of shorter-duration quarantines . Testing on day 5 and releasing on day 7 ( to account for test processing delays ) performs similarly to 10-day quarantine , preventing 98 . 5% [CI: 95 . 5% , 99 . 6%] of local transmission ( Figure 4A ) . Testing and releasing on day 6 ( i . e . no delay between test and result ) still prevents 97 . 8% [CI: 94 . 4% , 99 . 0%] of local transmission . Hence , if the rapid test has the same sensitivity and specificity as the laboratory-based RT-PCR test , then the duration of quarantine of individuals with a negative test result can be shortened by 1 day with minimal loss in efficacy compared to a test with a 48 hr turnaround . The timing of the test can have a significant impact on prevented transmission; late testing reduces the false-negative probability but increases the stay in quarantine . An important consequence of this is that testing on arrival is a poor strategy for limiting transmission: testing and releasing on day 0 would prevent only 35 . 2% [CI: 35 . 1% , 35 . 3%] of local transmission , while testing on arrival and releasing after 2 days prevents only 54 . 1% [CI: 49 . 5% , 59 . 4%] . As was the case for the traced contacts , the fraction of local transmission prevented by standard quarantine bounds the efficacy of the test-and-release quarantine strategy from below ( Figure 4A ) . We again measure the utility of quarantine by calculating the efficacy ( local transmission prevented across all individuals in quarantine , assuming 100% adherence ) per day spent in quarantine , and then comparing these utilities for different quarantine strategies to the utility of the standard 10-day quarantine through the relative utility ( Figure 4B ) . In the absence of testing , the duration of quarantine , and hence the relative utility , is independent of the fraction of individuals in quarantine that are infected . For travellers returning from a 7-day trip , the relative utility is a decreasing function of quarantine duration ( black line in Figure 4B ) . The maximum utility strategy would then be to shorten quarantine as much as possible . As was the case for traced contacts , under the test-and-release quarantine protocol the average time spent in quarantine , the utility , and the relative utility compared to the standard 10-day quarantine will depend on the fraction of individuals in quarantine that are infected . This fraction may change depending on disease prevalence at the travel destination and the duration of travel . For example , the infected fraction of travellers returning from a long stay in a high-risk country is likely to be higher than the infected fraction of travellers returning from a short stay to a low-risk country . In Figure 4B , we keep this fraction fixed at 10% . Early testing greatly reduces the average duration of quarantine and hence leads to increased utility despite the low fraction of transmission that is prevented ( coloured lines in Figure 4B ) . The average quarantine duration increases linearly with the fraction of infecteds in quarantine [Equation ( 3 ) in 'Materials and methods'] . The ratio of quarantine efficacy to the average quarantine duration will also increase , such that quarantine is of higher utility if the fraction of infecteds is higher . However , the relative utility of test-and-release quarantine compared to the standard 10-day protocol will decrease and approach 1 as the fraction of infecteds increases . Hence , if the disease prevalence among those returning from travel abroad is high , then test-and-release may not bring substantial benefits over the standard 10-day protocol . Our assumption that infection occurs with uniform probability across each day of a trip leads to some interesting results . Returning travellers that have been infected on a short journey will have , on average , used up less of their infectivity potential by the time they return than a traveller who was infected on a long journey . Hence , the total transmission that can be prevented by a long quarantine period ( e . g . 10 days ) upon arrival is greater for short trips ( Figure 4—figure supplement 1A ) . When considering the fraction of local transmission that can be prevented by quarantine , then shorter quarantine durations have a greater impact on long than short trips ( Figure 4—figure supplement 1C ) . Again , this is because , on average , the traveller on a long trip would have been exposed earlier and they will be infectious for a shorter time period after arrival . If an individual traveller is to be quarantined , then the optimum duration of quarantine , based on our metric of utility , would depend on the duration of their travel , with shorter journeys requiring longer quarantine ( Figure 4—figure supplement 1B , D ) . This might be counterintuitive because individuals who have been on longer journeys to high-risk countries have a higher probability of being infected . The absolute utility ( transmission prevented by quarantine across all individuals in quarantine divided by the average quarantine duration ) of quarantining such individuals could indeed be higher than for individuals returning from shorter journeys . However , here , we are not considering the question of whether to quarantine or not , but we are assuming that the individual is quarantined and are trying to optimise the duration of quarantine in response to the expected infection dynamics . We observe an almost-linear response between quarantine duration and the relative utility of the standard ( no test ) quarantine: for every day that quarantine is shortened , we see the same additive increase in relative utility ( black line in Figure 4B ) . This almost-linear response is coincidental to the 7-day trip duration: longer or shorter trips show non-linear responses ( Figure 4—figure supplement 1D ) . Trips shorter than 7 days have a maximum relative utility of between 4 and 7 days , while trips longer than 7 days have maximum utility for maximally shortened quarantine durations . Enforcing additional hygiene and social distancing guidelines post-quarantine can increase both efficacy and utility , but with diminishing returns as the release date is increased ( Figure 4—figure supplement 2 ) . As discussed for traced contacts , the loss of efficacy due to shortening quarantine could be offset by increasing quarantine adherence . Shortening from 10 to 5 days would require adherence to increase by 20% ( relative adherence 1 . 20 [CI: 1 . 12 , 1 . 35] ) in order to maintain the same overall efficacy ( Figure 4—figure supplement 3A ) . With test-and-release this required increase in adherence is even smaller . We note that the change in adherence required to balance a change in efficacy for shortened quarantine durations is dependent on the travel duration , with short travel durations requiring a greater increase in adherence compared with longer travel durations .
Quarantine is one of the most important measures in controlling the ongoing SARS-CoV-2 epidemic due to the large fraction of pre-symptomatic and asymptomatic transmission . A quarantine period of 10 days from exposure , as currently implemented in Switzerland , is long enough to prevent almost all onward transmission from infected contacts of confirmed cases from the point of entering quarantine: increasing the duration of quarantine beyond 10 days has no extra benefit . Reducing the delay to quarantining individuals increases the fraction of total transmission that is preventable . The same 10-day quarantine duration will prevent almost all local onward transmission from infected travel returners from the time of arrival , independent of their travel duration . Any decrease in the duration of quarantine of an infected individual will result in increased onward transmission . Furthermore , our analyses suggest that this increase in transmission cannot realistically be compensated by increased adherence for significantly shortened quarantine ( fewer than 5 days ) . However , there are diminishing returns for each day that we add to quarantine: increasing the duration from 10 days has a negligible effect in terms of reduced transmission . One therefore has to assess how much human cost , measured in terms of days spent in quarantine , we are willing to spend to prevent disease transmission . By comparing the ratio of prevented transmission to quarantine duration , we have shown that maximal utility strategies can exist . This ratio is maximised for quarantine durations of 6–8 days after exposure for traced contacts , and potentially less for returning travellers depending on their duration of travel . Importantly , under this metric the fraction of individuals in quarantine that are infected does not affect the optimal duration of quarantine . Therefore , the argument that we should shorten quarantine because of the low probability of being infected is misguided under our definition of utility and in the absence of testing during quarantine . A test-and-release strategy will lead to a lower average quarantine duration across infected and non-infected individuals . However , due to the considerable false-negative probability of the RT-PCR test ( Kucirka et al . , 2020 ) , this strategy also leads to increased transmission as infectious individuals are prematurely released . Nevertheless , test-and-release strategies prevent more transmission than releasing without testing , and hence test-and-release increases the utility of quarantine . Reducing the delay between test and result leads to further reduced transmission and increased utility , and reinforcing individual prevention measures after release is also effective for short quarantine periods . The ratio of transmission prevented versus days spent in quarantine is only one possible definition of utility . Defining the appropriate function is ultimately a policy question: the economic , societal , and individual costs are likely a function of the number of days spent in quarantine , but we cannot determine the shape of this function . Furthermore , the local epidemiological situation will dictate which metric of quarantine efficacy is to be optimised . In situations where the goal is to prevent the ( re ) introduction of SARS-CoV-2 , one should focus on maximising the reduction of transmission ( and hence minimising the transmission risk ) . If the virus is already endemic , then considering the trade-off between transmission reduction and quarantine duration could determine the optimum strategy . Another perspective is that the utility of preventing transmission is crucially dependent on whether it brings the effective reproductive number under 1 . Ultimately , bringing the reproductive number below 1 requires a combination of effective measures including isolation , physical distancing , hygiene , contact tracing , and quarantine ( Kucharski et al . , 2020 ) . Effective quarantine is only possible in the presence of efficient contact tracing to find the potentially exposed individuals in a short time , as well as surveillance of disease prevalence to identify high-risk travel . Further reducing the time taken to quarantine a contact after exposure and reducing the delay between test and result will allow average quarantine durations to be shorter , which increases the benefit-to-cost ratio of quarantine . The scenarios of returning travellers and traced contacts of confirmed SARS-CoV-2 cases differ in the probability of having been exposed and infected and on the information available about the likely window of exposure . The impact of quarantining returning travellers depends on the duration of travel and whether we consider the local prevention of transmission or the total transmission prevented by quarantine . However , a single test done immediately after return can only prevent a small fraction of the transmission from a returning traveller because of the false-negative rate of the RT-PCR test early in infection . Therefore testing should be postponed until as late as possible , and utilising rapid tests could be crucial if their performance characteristics are acceptable . This same principle also applies to traced contacts . Our findings are aligned with those of two recent simulation studies which estimate the role that quarantine plays in limiting transmission from returning travellers ( Clifford , 2020 ) and from traced contacts ( Quilty et al . , 2021 ) . Our results are based on the latest estimates of the generation time distribution of COVID-19 ( Ferretti et al . , 2020b ) . Potential limitations to our approach could be that these distributions may change throughout the epidemic , particularly depending on how people respond to symptoms ( Ali et al . , 2020 ) . Furthermore , these distributions , and also the test sensitivity profile , could be different between persistently asymptomatic and symptomatic individuals ( Buitrago-Garcia et al . , 2020 ) , which ultimately lead to an overestimation of how much transmission is prevented by quarantine . In addition , we have primarily assumed that symptom onset during quarantine has no impact on quarantine efficacy . However , this symptomatic transmission should not be counted towards the efficacy of quarantine as the infected individual should already self-isolate after symptom onset . We have quantified this effect and have shown that this assumption leads us to overestimate quarantine efficacy . For travellers , another consideration is that lengthy quarantine is seen as a deterrent to travel to high-risk areas ( IATA , 2020 ) . Any shortening of quarantine may lead to an increase in travel volume , potentially leading to a compounded increase in disease transmission . In the absence of empirical data about the effectiveness of different durations of quarantine , mathematical modelling can be used objectively to explore the fraction of onward transmission by infected contacts or returning travellers that can be prevented . However , determining the optimal quarantine strategy to implement depends on the impact that shortening the duration of quarantine has on individuals , society , and the economy versus how much weight is assigned to a consequential increase in transmission . Both the individual , societal , and economic impact , as well as the weight of transmission increase , will have to be considered based on the current epidemiological situation . We have shown that there are quarantine strategies based on a test-and-release protocol that , from an epidemiological viewpoint , perform almost as well as the standard 10-day quarantine , but with a lower cost in terms of person-days spent in quarantine . This applies to both travellers and contacts , but the specifics depend on the context .
For an infected individual who was exposed at time t_E , the fraction of transmission that is prevented by the standard quarantine strategy is given by the area under the generation time distribution , q ( t ) ( Figure 1—figure supplement 1B ) , between the times at which the individual enters ( t_Q ) and leaves ( t_R ) quarantine ( Grantz et al . , 2020 ) , that is , ( 1 ) F_qs ( t_E , t_Q , t_R ) =∫_t_Qt_Rdtq ( t-t_E ) . The duration of time that the individual spends in quarantine is then D_qs=t_R-t_Q . The test-and-release strategy uses virological testing during quarantine to release individuals with a negative test result and to place those with a positive test result into isolation . As illustrated in Figure 1A , test is issued at time t_T≥t_Q . If the test is negative , the individual is released when the test result arrives at time t_R . Otherwise , the individual is isolated until they are no longer infectious . One challenge with this strategy is the high probability of a false-negative RT-PCR test result ( i . e . an infectious individual is prematurely released into the community ) . As reported by Kucirka et al . , 2020 , the false-negative rate is 100% on days 0 and 1 post-infection , falling to 67% ( day 4 ) , 38% ( day 5 ) , 25% ( day 6 ) , 21% ( day 7 ) , 20% ( day 8 ) , and 21% ( day 9 ) , before rising to 66% on day 21 . We use linear interpolation and label this function f ( t ) , the false-negative probability on day t after infection . The fraction of transmission prevented by quarantining an infected individual under the test-and-release strategy is ( 2 ) F_qtr ( t_E , t_Q , t_T , t_R ) =f ( t_T-t_E ) ∫_t_Qt_Rdtq ( t-t_E ) +[1-f ( t_T-t_E ) ]∫_t_Qt_enddtq ( t-t_E ) , where the first term captures the fraction of individuals who receive a false-negative test result and are released at time t_R , and the second term captures individuals who return a positive test and are subsequently isolated until they are no longer infectious at time t_end . A further challenge with this false-negative rate is that it was calculated by Kucirka et al . , 2020 from symptomatic cases only . Here we assume that this test sensitivity profile also applies to asymptomatic cases . Quarantine is applied pre-emptively , such that we do not know the infection status of individuals when they enter quarantine . If only a fraction s of the individuals that are quarantined are infected , then the average reduction in transmission across all individuals in quarantine is sF , where F is the fraction of transmission prevented when an infected individual is quarantined [i . e . Equation ( 1 ) or ( 2 ) ] . For the standard quarantine protocol , the average number of days spent in quarantine is independent of s: all individuals are quarantined for the same duration . However , under the test-and-release protocol , only the individuals who are actually infected can test positive and remain isolated after t_R . All non-infected individuals ( 1-s ) will receive a negative test result and are released at time t_R . Among the infected individuals in quarantine ( s ) , a fraction f ( t_T-t_E ) will receive a false-negative test result and will be released at time t_R , while the remaining fraction [1-f ( t_T-t_E ) ] will receive a positive test result and are isolated until they are no longer infectious . Hence the average number of days spent in quarantine for test-and-release is ( 3 ) Dqtr= ( 1−s ) ( tR−tQ ) +s[f ( tT−tE ) ( tR−tQ ) +[1−f ( tT−tE ) ] ( tend−tQ ) ]= ( tR−tQ ) +s[1−f ( tT−tE ) ] ( tend−tR ) , where s[1-f ( t_T-t_E ) ] is the fraction of quarantined individuals who return a positive test result . We see that the average test-and-release quarantine duration increases linearly with the fraction of individuals in quarantine that are infected ( s ) . Model parameters and timepoints are summarised in Table 1 . One possible metric to relate the effectiveness of quarantine to its negative impact on society is to consider the ratio between the amount of overall transmission prevented and the number of person-days spent in quarantine . We refer to this ratio as the utility of quarantine . Concretely , for an efficacy F [F_qs or F_qtr as defined by Equation ( 1 ) or ( 2 ) , respectively] , fraction of individuals in quarantine that are infected s , and average time spent in quarantine D ( D_qs or D_qtr ) , we define the utility as ( 4 ) U ( s , F , D ) =sFD . We can then compare the utility of two quarantine strategies by calculating the relative utility , that is , the ratio between the two utilities: ( 5 ) RU ( s , F , D , F∗ , D∗ ) =sF/DsF∗/D∗=F/DF∗/D∗ , where F and D are the efficacy and duration of quarantine of a new strategy , and F* and D* are the efficacy and duration of the baseline quarantine strategy to which we compare . The efficacies F and F* in Equation ( 5 ) are independent of fraction of individuals in quarantine that are infected s . For the standard quarantine strategy , the durations D=D_qs and D*=D_qs* are also independent of s , and hence the relative utility of the standard quarantine strategy is independent of s . For the test-and-release strategy , however , the duration is a linearly increasing function of s [D=D_qtr ( s ) ; Equation ( 3 ) ] . Hence the relative utility of the test-and-release strategy is dependent on s: ( 6 ) RU[s , F_qtr , D_qtr ( s ) , F_qs* , D_qs*]=F_qtr/D_qtr ( s ) F_qs*/D_qs* . In Appendix 1 we show that the relative utility of the test-and-release quarantine strategy is a decreasing function of s . We consider the scenarios of a traced contact and a returning traveller differently because the values of t_E , t_Q , and t_R are implemented differently in each case . To complement the results in this paper , and to allow readers to investigate different quarantine scenarios , we have developed an online interactive application . This can be found at https://ibz-shiny . ethz . ch/covidDashboard/quarantine . | The COVID-19 pandemic has led many countries to impose quarantines , ensuring that people who may have been exposed to the SARS-CoV-2 virus or who return from abroad are isolated for a specific period to prevent the spread of the disease . These measures have crippled travel , taken a large economic toll , and affected the wellbeing of those needing to self-isolate . However , there is no consensus on how long COVID-19 quarantines should be . Reducing the duration of quarantines could significantly decrease the costs of COVID-19 to the overall economy and to individuals , so Ashcroft et al . decided to examine how shorter isolation periods and test-and-release schemes affected transmission . Existing data on how SARS-CoV-2 behaves in a population were used to generate a model that would predict how changing quarantine length impacts transmission for both travellers and people who may have been exposed to the virus . The analysis predicted that shortening quarantines from ten to seven days would result in almost no increased risk of transmission , if paired with PCR testing on day five of isolation ( with people testing positive being confined for longer ) . The quarantine could be cut further to six days if rapid antigen tests were used . Ashcroft et al . ’s findings suggest that it may be possible to shorten COVID-19 quarantines if good testing approaches are implemented , leading to better economic , social and individual outcomes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"medicine"
] | 2021 | Quantifying the impact of quarantine duration on COVID-19 transmission |
Cortical blood flow can be modulated by local activity across a range of species; from barrel-specific blood flow in the rodent somatosensory cortex to the human cortex , where BOLD-fMRI reveals numerous functional borders . However , it appears that the distribution of blood capillaries largely ignores these functional boundaries . Here we report that , by contrast , astrocytes , a major player in blood-flow control , show a striking morphological sensitivity to functional borders . Specifically , we show that astrocyte processes are structurally confined by barrel boundaries in the mouse , by the border of primary auditory cortex in the rat and by layers IIIa/b and Cytochrome Oxidase ( CO ) -blobs boundaries in the human primary visual cortex . Thus , astrocytes which are critical elements in neuro-hemodynamic coupling show a significant anatomical segregation along functional boundaries across different mammalian species . These results may open a new anatomical marker for delineating functional borders across species , including post-mortem human brains .
The cerebral cortex has a remarkable capacity to regulate its own blood supply according to local neuronal activity demands ( Buxton and Frank , 1997 ) . This mechanism is of great interest , since it is the basis of optical imaging of intrinsic signals ( Vanzetta and Grinvald , 1999 ) as well as BOLD-fMRI imaging ( Logothetis et al . , 2001 ) . In particular , blood flow modulations have been used to demarcate functional columnar boundaries ( Blasdel and Salama , 1986; Grinvald et al . , 1986; Lieke et al . , 1989 ) . A particularly striking example of such modulation has been found across the boundaries of the rodent's vibrissae related 'barrels' ( Cox et al . , 1993; Derdikman et al . , 2003; Petersen and Sakmann , 2001; Woolsey et al . , 1996; Woolsey and Van der Loos , 1970; Yang et al . , 1997 ) . Given the tight coupling between columnar neuronal activation and blood flow , one could envision that the cortical vascular architecture should show anatomical shaping according to these boundaries . Such structural confinements by columnar boundaries were previously demonstrated in the dendritic and axonal arbors of barrel field neurons . These dendrites and axons appear to be 'repulsed' as they approach the barrel's boundary ( Brecht and Sakmann , 2002; Lendvai et al . , 2000; Petersen and Sakmann , 2000; Shepherd et al . , 2005 ) . However , detailed analysis of the micro-vascular organization in rodent barrel fields has failed to reveal such structural barrel-related modifications in the vascular capillary bed ( Blinder et al . , 2013 ) . Similar failure of anatomical confinement was additionally reported in the primate visual cortex ( Adams et al . , 2015 ) , including human cytochrome oxidase blobs evident in layers 2–3 of striate cortex , which constitute a robust example of functional 'mosaics' in the primate striate ( Livingstone and Hubel , 1984 ) . While the capillaries are a critical component in controlling blood circulation , another major constituent that is hypothesized to play a crucial role in the coupling between neurons and blood flow are glial cells , particularly astrocytes . These cells provide a crucial bridge between neuronal activity and vascular blood circulation . The basic concept is that astrocyte branches are capable of sensing the level of neuronal synaptic communications , integrate this information and then transmit it to the blood vessels , with which they are in contact . It has previously been proposed that signals related to neuronal activity levels can control the capillary’s diameter through astrocytes in this manner ( Attwell et al . , 2010; Haydon and Carmignoto , 2006; Petzold and Murthy , 2011; Takano et al . , 2006 ) . Importantly the role of astrocytes in blood flow control has recently been demonstrated using Ca2+ imaging ( Otsu et al . , 2015 ) . Interestingly , this research emphasized the importance of astrocytes' processes in blood flow control . Indeed , such control can be exquisitely precise , reaching sub-millimeter resolution ( Fukuda et al . , 2005 ) . Another interesting aspect of astrocyte morphology is the tight coupling between the neuronal and astrocyte processes . In particular , recent evidence has demonstrated the existence of tight anatomical co-localization of astrocyte processes and neuronal synapses in what has been termed a 'tri-partite synapse' formation ( Araque et al . , 1999; Halassa et al . , 2007a; Perea et al . , 2009; Santello et al . , 2012 ) . Such links further suggest that astrocyte morphology may follow boundaries defined by neuronal functional compartments . Finally , previous research has uncovered that in parallel to the neuro-vascular coupling- astrocytes play a significant role in activity dependent astrocyte-neuron lactate shuttle for the supply of metabolic substrates to neurons ( Pellerin et al . , 1998 ) . Given the crucial functional role of astrocytes in neuro-vascular coupling and other aspects of neuronal modulations , we hypothesized that the astrocytes , rather than the blood vessels , may show anatomical remodeling over time according to persistent functional boundaries . Indeed , in the case of olfactory bulb glomeruli were neuronal connectivity is highly ordered , dye coupling experiments highlight a preferential communication between astrocytes within glomeruli but not between astrocytes in adjacent glomeruli ( Roux et al . , 2011 ) . To examine this hypothesis , we stained the entire astrocyte population in three species ( mouse , rat and human ) and in four functional boundaries ( barrel field , auditory cortex , layer IIIa/b and Blob-interblob of human striate cortex ) , using an immunohistochemical procedure on thin paraffin sections . Superimposing the astrocytes onto the functional boundaries revealed a significant structural relationship . Thus , astrocyte processes showed a significant anatomical confinement at the functional boundaries . This finding raises the possibility of using such astrocyte formations as an anatomical marker for the identification of functional boundaries across different cortical areas and species , including post-mortem human tissue .
Our examination of barrel field tissue was based on 154/146 ( barrel/septa ) measurements conducted in 4 sections in five left hemispheres of five mice ( sectioned at 8 µM ) and on 56/54 ( barrel/septa ) measurements calculatedin two additional left hemispheres ( sectioned at 4 µM ) . For inspection of the details of astrocyte morphology , tangential ( parallel to the cortical surface ) sections were collected from all cortical layers . Prior to astrocyte staining , the sections were photographed using dark field illumination to highlight the barrel field borders . Figure 1A provides a low magnification ( 3 . 2x ) example of the barrel boundaries obtained in one animal ( dark field illumination , inverted contrast ) . The barrel boundaries are clearly evident as dark septa surrounding a lighter core . 10 . 7554/eLife . 15915 . 003Figure 1 . Relationship of astrocyte branches and capillaries to the barrels borders in cortical layer IV at medium and high magnifications . ( A ) A dark field illumination image ( inverted contrast ) of the barrel field ( BF ) area of a flattened left hemisphere . The barrel borders are evident as dark septa surrounding a lighter core . Barrel rows A–E and arcs 1–5 as well as the medial–lateral and rostral–caudal axes are indicated . ( B ) Reconstruction of the capillaries ( Cp ) visualized using fluorescein dextran . Barrel borders ( derived from A ) marked by yellow lines , were superimposed on the drawing of capillaries . The two images were aligned using penetrating blood vessels as landmarks . ( C ) Fluorescent micrograph of GFAP immunostaining ( CY3 conjugated ) of astrocyte processes ( As ) . The image consists of 90 tiles that were assembled together revealing the lateral distribution of the processes . ( D ) A reconstruction of the astrocyte processes of image C performed by manual drawing of their branches at high magnification . The borders of the barrel field ( marked in turquoise ) and astrocyte processes were aligned as in B . Green arrowheads indicate three barrel borders shown in enlarged view in E and F . All the images were taken from the same tissue section . ( E ) Enlarged florescent image of stained astrocyte processes ( same as in C ) showing further details of the processes . Arrow heads point to a clearly visible gap in the astrocyte distribution at three barrel boundaries . Green corners indicate the enlarged image shown in G . ( F ) A reconstruction of the astrocyte processes ( drawn manually at high magnification ) shown in E . ( G ) An enlarged fluorescent image view of the area bounded by the green box in E , showing the astrocyte distribution relative to the barrel borders ( arrows ) . ( H ) A high magnification confocal image demonstrates the fine processes of two astrocytes at the border between barrels and septa ( green arrowhead ) indicated by a yellow line . The astrocyte cell bodies are indicated by yellow arrows and examples of their processes by red ones . DOI: http://dx . doi . org/10 . 7554/eLife . 15915 . 003 Following the barrel boundary demarcation , sections were immunostained by GFAP antibody to detect the astrocyte morphology . This staining revealed the branching patterns of individual astrocyte processes ( Figure 1C ) . An example of the branching of astrocyte processes is shown at intermediate ( 40x ) and high ( 60x and 100x ) magnifications in Figure 1E , G and H respectively . The spread of astrocyte processes was not uniform , but tended to be highly modulated , forming patches of dense and sparse regions . To reveal the precise details of such modulations , and their possible relationship to barrel borders , we manually drew the distribution of astrocyte processes at high magnification ( see Materials and methods ) . The astrocyte process drawings for the sample cases are shown at low , ( 20x , Figure 1D ) and intermediate , ( 40x , Figure 1F ) magnifications . As can be clearly seen in Figure 1C , E , G and H , reductions in the density of astrocyte processes were not randomly distributed , but overlapped the barrel boundaries ( e . g . see arrows-heads Figure 1C–H and dashed line in 1H ) . A high magnification confocal image demonstrates the fine processes of astrocytes reaching the barrel border and abruptly stopping ( red arrows Figure 1H ) . To compare the distributions of the astrocyte processes to that of the blood vessel capillary bed , the blood vessel distribution ( traced by FITC dextran ) was examined ( see Materials and methods ) . Figure 1B depicts the distribution of blood vessels for the same field shown in Figure 1C . Unlike the clear boundary-related gaps that were evident in the astrocyte process distribution , blood capillaries appeared to ignore barrel boundaries and their tangential distribution was largely uniform across both the barrel septa and core . The astrocyte confinement within barrel borders was a ubiquitous and robust phenomenon , and was observed in all animals studied . In all cases , there was a clear reduction in septa crossing by astrocyte branches . To obtain a quantitative measure of this phenomenon , the number of astrocyte process crossings was separately counted in the barrel core and septa ( Figure 2 illustrates the procedure ) . Barrels , with clearly demarcated borders , having a length of between 350–450 µm obtained from dark field imaging , were selected for analysis . Based on these photographs , straight lines ( 375 µm length ) were superimposed on the center of the barrel border , aligned with the border's orientation . For comparison , lines of identical length and orientation were then placed at the center of the barrel cores ( Figure 2C ) . These septa and core-related lines were then superimposed on the corresponding drawings of the astrocyte processes ( see illustration in Figure 2A ) and the number of astrocyte branch crossings of these lines was calculated . The same procedure was repeated for blood vessel crossings ( Figure 2B ) . 10 . 7554/eLife . 15915 . 004Figure 2 . Quantitative comparison of astrocyte and capillary crossings in the core and septa of the barrels in layer IV as well as of the corresponding regions in the supra-granular and infra-granular layers . ( A ) Astrocyte process ( As ) reconstructions from the barrel field ( BF ) of layer IV . ( B ) Capillary ( Cp ) reconstruction of the same field . ( C ) A matrix of lines ( 375 µm length ) was drawn at the center of the septa aligned with the border's orientation . For comparison , lines of identical length and orientation were superimposed at the center of the barrel cores ( see Materials and methods for details ) . Septa and core related lines superimposed on the astrocyte and the capillary drawings in A and B respectively were aligned using penetrating arterioles . ( D and E ) Drawings of astrocyte distribution in layers III ( supra-granular ) ( left ) and layer V ( infra-granular ) ( right ) of the cortex , 80 µm above and below the barrel field borders . The boundaries of the barrels were transferred from the dark field image of layer IV and superimposed upon the astrocyte processes of layer III and V . All the images were aligned according to penetrating arterioles . The matrix of lines was also transferred from layer IV ( not shown ) . ( F ) Histograms depicting the number of astrocyte processes and capillaries respectively , crossing the septal and barrel cores at layers III-V . ( G ) Histograms displaying septa/barrel crossing ratio ( see Materials and methods ) for the astrocyte ( left ) and capillary ( middle ) crossings , across the different layers , as well as a histogram comparing the ratio of astrocytes with capillaries in layer IV ( right ) . Means ± s . e . m . , *p<0 . 0001; n . s . , non-significant . DOI: http://dx . doi . org/10 . 7554/eLife . 15915 . 004 The results of this analysis for 5 animals were performed on 4 neighboring sections with interval of about 20 µm between them . The number of astrocyte process crossings in 154 ± 7 barrel cores and 146 ± 6 septa , are shown in Figure 2F . There was a three-fold ( 11 . 02 ± 0 . 58 vs 3 . 78 ± 0 . 35 , means ± s . e . m , t298 = 9 . 7 , p<0 . 001 . 95% confidence interval , 10 . 26 to 11 . 82 and 3 . 3 to 4 . 2 respectively; Cohen's d = 1 . 72; Figure 2F , left ) difference in the number of astrocyte crossings of barrel borders compared to the crossing numbers within the barrel’s core . In contrast , there was only a slight trend of reduced crossings of blood vessels at barrel's border ( 2 . 64 ± 0 . 18 vs 2 . 97 ± 0 . 18 . 95% confidence intervals , 2 . 47 to 2 . 81 and 2 . 77 to 3 . 17 ) which did not reach significance ( t156 = 1 . 28 , p=0 . 2; Figure 2F , right ) . It should be noted that the barrel boundaries are less anatomically defined in the supra and infra-granular layers ( Durham and Woolsey , 1977; Woolsey and Van der Loos , 1970 ) . Consequently , our barrel boundary for the layers above and below layer IV was based on a vertical extrapolation of layer IV boundaries ( relying on penetrating arterioles to align the borders . Figure 2D and E are representative drawings taken from layer III ( D ) and layer V ( E ) . Although the distribution of astrocyte processes was suggestive of barrel confinement , the relationship to the barrel boundaries was much less evident . Quantitative comparison of the astrocyte crossings within and across barrels of three animals ( 85 barrel cores and 76 septa ) , performed in a similar manner to that conducted for the layer IV analysis , showed a lower , albeit significant reduction in septa-barrel crossings , relative to crossings of the barrel core in supra-granular layer III ( 16 . 2 ± 0 . 8 vs 11 . 2 ± 0 . 7 , t159 = 3 . 6 , p<0 . 001 . 95% confidence interval , 15 . 6 to 16 . 8 and 10 . 7 to 11 . 7 respectively , Cohen's d = 1 . 84; Figure 2F , left ) . A similarly significant confinement was found in the infra-granular layer V; 9 . 4 ± 0 . 8 vs 14 . 3 ± 0 . 7 , t154 = 3 . 9 p<0 . 001 . 95% confidence interval , 9 . 0 to 9 . 8 and 13 . 7 to 14 . 9 respectively; Cohen's d = 1 . 92; Figure 2F , left ) . In contrast , there were no significant differences in crossings of capillaries at the barrel's border when compared to the barrel core in either layer III ( 1 . 66 ± 0 . 25 vs 1 . 69 ± 0 . 18; t155 = 1 . 31 , p = 0 . 19 . 95% confidence interval , 1 . 49 to1 . 86 and 1 . 57 to 1 . 81 respectively , Figure 2F , right ) or in layer V ( 2 . 46 ± 0 . 29 vs 1 . 73 ± 0 . 28; t153 = 0 . 29 , p = 0 . 77 . 95% confidence interval , 2 . 28 to 2 . 64 and 1 . 64 to 1 . 82 respectively; Figure 2F , right ) . Comparison of the ratio between crossings within the septa and barrel core between layers revealed that astrocytes in layer IV showed a significantly higher level of astrocyte boundary confinement when compared to the supra and infra granular layers ( 0 . 79 ± 0 . 07 in layer III , 0 . 36 ± 0 . 03 in layer IV , and 0 . 68 ± 0 . 05 , in V , using a one-way ANOVA on log-transformed ratios , followed by a Tukey post-hoc test ( F2 , 105 = 20 . 6 , p<0 . 001; Figure 2G , left ) . Conversely , comparison of the septa/barrel ratio of capillary crossings between layers , revealed nonsignificant changes ( 0 . 91 ± 0 . 2 in layer III , 1 . 05 ± 0 . 09 in Layer IV , and 1 . 68 ± 0 . 39 and in Layer V ) using the same statistical analysis as in 2G , left ( Figure 2G , middle ) . Comparing the ratio between crossings of the septa and the barrel core of the astrocyte processes and capillaries in layer IV , the former was significantly lower ( 0 . 36 ± 0 . 03 vs 1 . 05 ± 0 . 09 , t282 = 6 . 6 , p<0 . 001; Figure 2G , right ) . To examine possible effects related to section's thickness , we compared 4 and 8 µm thick sections in layer IV ) . The analysis revealed the expected reduction of about 35% in the number of astrocyte process crossing in the 4 compared to the 8 µm thick sections , but the ratio between the septa/barrels for the astrocyte branches was similar in both thick and thin sections ( 0 . 37 ± 0 . 05 and 0 . 36 ± 0 . 03 for 4 and 8 µm sections , respectively ) . While our results demonstrate a clear confinement of astrocyte processes to the barrel borders in the mouse cortex , it could be argued that such a phenomenon is unique to the barrel cortex or to the mouse . To examine how general this phenomenon is , we examined the density of astrocyte processes in the auditory cortex of the rat . The analysis was based on 82 ± 9 measurements on 4 neighbored sections ( the interval between sections was about 20 µm ) conducted in the left hemispheres of 4 rats ( sectioned at Layer IV ) . For convenience , we will refer to all the secondary fields around A1 ( Polley et al . , 2007 ) as A2 . The border of A1/A2 was defined by dark field illumination , the results of which are depicted in Figure 3 . To examine the density of astrocyte processes in the A1 vs . A2 cortex , we obtained ten bands 25 µm wide and 100 µm long , their centers located at the borders between A1/A2 ( Figure 3C , middle ) . Quantitative measurement of their density was taken along each pixel of the bands ( Figure 3C right ) . In addition , we plotted straight lines ( 50 µm length ) , which were superimposed at a distance of 50 and 200 µm from the border of A1/A2 on both sides , aligned with the border's orientation . These A1 and A2 lines were then superimposed on the corresponding drawings of the astrocyte processes ( Figure 3E and G ) and the number of astrocyte branch crossings were calculated . 10 . 7554/eLife . 15915 . 005Figure 3 . The relationship between astrocyte processes and the A1/A2 border in layer IV of rat auditory cortex . ( A ) Low magnification of a dark field illuminated image of a tangential section through layer IV . ( B ) Dark field illuminated image of areas A1 ( primary auditory cortex , dark round area ) and A2 ( secondary auditory fields , bright area surrounding A1 ) , taken from the green bounded region in A . ( C ) Left panel: graphical representation of the number of astrocyte processes ( AS ) crossing a matrix of lines ( 50 µm length ) which were drawn at A1 and A2 areas , 50 and 200 µm from the A1/A2 border aligned with the border's orientation ( not shown ) . Right panel: density plot measurements of ten 25 µm wide bands located perpendicular to the border’s orientation ( middle panel ) in areas A1/A2 ( individual traces-red , average-blue ) . ( D and F ) An enlarged fluorescent images of the astrocyte processes in the area bounded by the yellow and green boxes in B left ( L ) and right ( R ) border of A1/A2 , respectively . ( E and G ) A manual reconstruction of the astrocyte processes in images D and F , respectively . The border between A1 and A2 is demarcated by a green line . SS , somatosensory cortex . V1 , primary visual cortexDOI: http://dx . doi . org/10 . 7554/eLife . 15915 . 005 As can be seen from the histogram in Figure 3C right , there was a sharp and significant drop ( 37 ± 3% ) in the density of astrocyte processes between A1 and A2 . The histogram in Figure 3C left , reveals reduction in the number of astrocyte crossings immediately outside the A1 border ( 50 µm ) ; 14 . 1 ± 0 . 93 vs 7 . 4 ± 0 . 69 ( t52 = 6 . 46 , p<0 . 001 ) . Similar results were obtained when we compared between the two areas at a distance of 200 µm from the border; 13 . 96 ± 0 . 54 vs 7 . 18 ± 96 ( t50 = 5 . 86 , p<0 . 001 ) . Importantly , no significant changes were observed at different distances from the border on either side , ( 14 . 1 ± 0 . 93 vs 13 . 9 ± 0 . 54 , t52 = 0 . 25 , p=0 . 79 , at a distance of 50 and 200 µm respectively in A1 and 7 . 48 ± 0 . 69 vs 7 . 18 ± 0 . 45 , t50 = 0 . 45 , p=0 . 64 , 50 and 200 µm respectively in A2 ) , indicating that the change in the density of astrocyte processes was abrupt at the A1/A2 border and did not gradually change approaching this border . If the confinement of astrocyte processes at functional borders is a general phenomenon , such a finding could be particularly informative for studying the human cortex . Thus , examining the density of astrocyte processes may offer a novel 'window' into functional boundaries that may be revealed even in post-mortem tissue . To explore the possible extension of our hypothesis to the human cortex , we examined the boundaries of sub-laminae IIIa/IIIb of human primary visual cortex V1 . Previous research has uncovered functional distinctions between these two sub-layers in the primate cortex ( Sceniak et al . , 2001 ) . We analyzed coronal sections from the right hemisphere of one human post-mortem brain ( age 35 ) . Figure 4—figure supplement 1 depicts the orientation of the sections relative to the occipital pole . To reveal the IIIa/IIIb border we used non-phosphorylated neurofilament ( NPNF ) immunostaining ( Campbell and Morrison , 1989; Preuss et al . , 1999 ) and related this border to inter-laminar astrocyte processes ( see Figure 4D ) . Astrocyte processes were not uniformly spread in layer III , but appeared to be constrained by the border of IIIa/IIIb . To quantitatively examine this effect , we manually drew the distribution of astrocyte processes at high magnification ( Figure 4C ) . Measuring the density of astrocyte processes between the two sublayers revealed a dramatic and sharp reduction in the density ( 66 ± 4% ) at the border of layer IIIb ( Figure 4E , left ) . The number of astrocyte branches crossing straight lines ( 375 µm length , see Materials and methods ) at both sides of the border was also calculated in 4 neighboring sections ( interval of about 20 µm between sections , Figure 4E , right ) . There were no differences between the crossings of astrocyte processes in the same sub laminar layer at different distances ( 50 and 200 µm ) from the border; 27 . 2 ± 1 . 15 vs 28 . 5 ± 1 . 07 for IIIa ( t68 = 1 . 66 , p=0 . 10 ) and 13 . 3 ± 0 . 54 vs 12 . 4 ± 1 . 37 for IIIb ( t62 = 1 . 18 , p=0 . 24 ) . However , a highly significant change was found in the number of crossings when comparing the two sub-laminar areas at similar distances from the borders; 27 . 2 ± 1 . 15 vs 13 . 3 ± 0 . 54 , t66 = 22 . 39 , p<0 . 001 , 50 µm , and 28 . 5 ± 1 . 07 vs 12 . 4 ± 1 . 37 , t68 = 18 . 44 , p<0 . 001 , 200 µm from the border , indicating an abrupt change in astrocyte crossing at the sublaminar border . 10 . 7554/eLife . 15915 . 006Figure 4 . Relationship between interlaminar astrocyte processes and the sublaminar border of layers IIIa/IIIb in the post-mortem human striate cortex . Immunohistochemical staining of coronal sections of human area V1 , illustrating the relationship between interlaminar astrocyte processes and the sublaminar border of layers IIIa/IIIb . ( A ) Bright field photograph of coronal section through V1 stained for NPNF ( non-phosphorylated neurofilament ) , a higher magnification , image of the area is shown in Figure 4—figure supplement 1D . The border between IIIa/IIIb is demarcated by the heavily ( II & IIIa ) and lightly ( IIIb ) stained bands . ( B and D ) Immunofluorescent double staining for NPNF ( B ) and GFAP ( D ) , respectively , on a section adjacent to that of A . The border between IIIa/IIIb is demarcated by a green line . ( C ) A manual reconstruction of the astrocyte processes ( As ) in image D . ( E ) Left panel: superimposed density plot measurements of ten 30 µm wide bands perpendicular to the border ( plotted in C ) in layers IIIa/IIIb ( individual traces-red , average-blue ) . Right panel: a graphic representation of the number of astrocyte processes crossing a matrix of lines ( 375 µm length ) , which were drawn at layers IIIa and IIIb , 50 and 200 µm from the IIIa/IIIb border aligned with the border's orientation ( not shown ) . ( F–H ) Enlarged image of stained pyramidal cell bodies and apical dendrites , as well as astrocyte processes taken from the green bounded region in B–D , respectively , showing further details of the astrocyte processes . Arrow heads point to a clearly visible IIIa/IIIb borderDOI: http://dx . doi . org/10 . 7554/eLife . 15915 . 00610 . 7554/eLife . 15915 . 007Figure 4—figure supplement 1 . Low magnification view of the human occipital lobe . ( A ) A semi-schematic illustration of the right hemisphere of the brain . The vertical dotted line in the occipital lobe corresponds to the plane of sections B–D . ( B ) A Coronal section through the Calcarine sulcus of area V1 ( marked by red line in A ) , stained for myelin . The dark , densely stained stripe demarcates the Stria of Gennary ( arrow heads ) . ( C ) Higher magnification micrograph of the section in B , rotated 90º CW , demonstrating the Stria of Gennary in layer IVb . ( D ) A sequential section of C , immunostained for non-phosphorylated neurofilament ( NPNF ) , showing dark staining bands in layers IIIa , IVb , V and VI . The green corners indicate the area of the image in Figure 4A ( see main text ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15915 . 007 Finally , we examined the potential segregation of astrocyte processes along the compartmental division of CO-blobs vs inter-blobs in human V1 . It should be noted that , in contrast with the previous examples , the blob boundaries are not sharply defined and , in fact , dendritic trees , unlike barrel-field borders , are not 'repelled' by these fuzzy borders ( Malach , 1992; 1994 ) . However , an interesting issue is whether the density of astrocytes significantly changes as one moves from the CO-blob into the inter-blob regions . To examine this question , we mapped neighboring sections ( about 100 µm apart ) of human post-mortem tissue with both the CO-blob locations , defined through CO histochemistry ( 15 µm thick sections ) and the astrocyte distribution through immunohistochemistry for GFAP . Images were aligned by matching blood vessels as landmarks . Figure 5 depicts the results of this analysis . We counted the number of astrocyte crossings across each of the vertical 500 µm lines that were crossing blobs or inter-blobs area at layers II and IIIa ( Figure 5D ) . As can be seen from the histogram ( Figure 5E ) , there were no significant differences between the crossings of astrocyte processes in the inter-blobs at different distance from the blobs border; 11 . 2 ± 1 . 08 vs 12 . 1 ± 1 . 19 ( p=0 . 58 ) . However , we found a significant increase in astrocyte processes crossing at inter-blobs compared to the CO blobs proper ( 11 . 25 ± 1 . 08 vs 5 . 25 ± 0 . 72; t ( 48 ) = 9 . 1 p<0 . 001 . 95% confidence interval , 10 . 33 to 12 . 17 and 4 . 57 to 5 . 93 respectively and 12 . 1 ± 1 . 19 vs 5 . 25 ± 0 . 72; t ( 48 ) = 11 . 2 p<0 . 001 ) . 95% confidence interval , 11 . 28 to 12 . 92 and 4 . 57 to 5 . 93 ( Cohen's d = 1 . 2 and 0 . 9 respectively . Figure 5E ) . Thus , we found a significant modulation in astrocyte density that follows the changes of CO-organization . 10 . 7554/eLife . 15915 . 008Figure 5 . Relationship between CO-blobs and astrocyte density in the post-mortem human striate cortex . The relationship between the intensity of cytochrome oxidase ( B , blue ) and astrocytic GFAP staining ( C , red ) along cortical layer II and IIIa is depicted in a coronal section from human striate cortex . ( A ) 1–3; transmission plots of bright-field CO histochemistry ( high-CO indicated by low levels of transmission , blue lines ) and astrocyte process density ( red ) . R-values ( Pearson coefficients of correlation ) indicate a significant CO-modulation of astrocyte density and blob-interblob organization . ( B ) A section stained for cytochrome oxidase shows a dense band in layer IV and periodic vertical regions of enhanced enzyme activity in the upper layers ( blobs ) , interspersed with inter-blob regions at about 1 mm period . Horizontal lines ( numbered 1–3 ) depict the density measurement lines shown in A1-3 respectively . ( C ) Astrocyte process reconstruction in layer II and IIIa of an adjacent section stained for GFAP . Horizontal blue lines , similar to B . ( D ) Same sections as in B and C , indicating the location of 500 micron vertical lines that were placed at the center of blobs ( yellow ) , the center of inter-blobs ( green ) and at the left margin of inter-blobs ( blue ) . ( E ) Histograms depict the number of astrocyte crossings across each of the vertical lines shown in D . Note the significant increase in astrocyte processes crossing in inter-blobs compared to the CO blobs proper ( *p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15915 . 008
First , we demonstrated a consistent anatomical confinement of astrocyte processes to somato-sensory barrel boundaries . In contrast to astrocyte confinement , and in agreement with previous studies ( Blinder et al . , 2013; Woolsey et al . , 1996; Wu et al . , 2014 ) , the distribution of blood capillaries did not show a significant relation to barrel boundaries . In that sense , the astrocyte processes appear to be more similar to dendritic arbors of barrel neurons , which were also reported to show a clear anatomical restriction in relation to barrel borders ( Brecht and Sakmann , 2002; Feldmeyer et al . , 2002; Lefort et al . , 2009; Lendvai et al . , 2000; Lubke et al . , 2000; Petersen and Sakmann , 2000; Shepherd et al . , 2005 ) . In contrast to layer IV , where barrel boundaries were clearly demarcated , the columnar segregation in supra- and infra-granular layers was less distinct . This is reflected both in structural segregation and in neuronal signaling ( Feldmeyer et al . , 2002; Lubke et al . , 2000; Petersen and Sakmann , 2001 ) . It is interesting to note that such boundary 'blurring' was also reflected in the astrocyte geometry , since our quantitative analysis revealed a significant reduction in the anatomical confinement by barrel boundary , both in supra- as well as infra-granular layers ( Figure 2 ) . Importantly , a significant , albeit smaller decrease in border crossing was observed in these layers as well . This result suggests that the level of anatomical border confinement could be a gradual process rather than an all or nothing categorical 'toggle' of a columnar confinement process . However , a word of caution should be noted here regarding the methodological accuracy of barrel border definition . In layer IV , such demarcation was straightforward , since barrel boundaries were easily demarcated in this layer ( Welker and Woolsey , 1974; Woolsey and Van der Loos , 1970 ) . Thus , determining the septa and core boundaries could be achieved with great accuracy . By contrast , defining barrel borders outside layer IV was less accurate , relying on vertical extrapolations from layer IV borders by alignment of some of the penetrating arterioles . Such extrapolation may have introduced some inaccuracies in our border delineations . Therefore , the estimates of border crossings in the upper and lower layers should be considered less accurate compared to our layer IV results . Our study reveals a significant dissociation between capillary distribution , which appears to ignore columnar boundaries , and the anatomical shaping of astrocyte processes . However , it could be argued that this observation was unique to the mouse barrel fields which indeed show a particularly striking functional segregation . To examine whether the astrocyte structural confinement may reflect a more general phenomenon , we extended our analysis to additional species ( rat and human ) and to additional functional boundaries ( A1/A2 sub-layer IIIa/IIIb and blob/inter-blob organization respectively ) . In all cases ( with the exception of CO-blob boundaries , which are not precisely defined to begin with , see ( Malach , 1992 ) , we found a significant and abrupt reduction in astrocyte processes near the regional boundaries ( see Figures 3–5 ) . It could be argued that the staining of astrocyte processes was not complete although our procedure ( thin sections , antigen retrieval , long time incubations ) enhanced the staining sensitivity . While we cannot rule out this possibility , such limited staining could not account for the selective drop in the density of astrocyte processes that was observed near functional boundaries , and not away from them as seen in all three cases studied ( see histograms in Figures 2F , 3C and 4E ) . Thus , we can safely conclude that astrocyte process morphology can be confined by regional and columnar boundaries across species and systems . Of course , our limited sample does not prove that all cortical boundaries should have a similar impact on astrocyte structure , but the observed phenomenon opens a potentially informative direction of future research , in which such astrocyte boundaries could be mapped in great detail across species and brain areas . It should be noted that a number of previous studies have demonstrated that astrocyte activity may be modulated according to the functional properties of the modules , within which they reside . For example , ( Schipke et al . , 2008 ) have demonstrated barrel specific Ca2+ response in astrocytes . However , it is important to emphasize that such functional distinctions , while offering a possible source for the morphological restructuring reported here , do not on their own constitute a demonstration of such anatomical reshaping . For example , in the Schipke et . al . experiment , the astrocyte selectivity was rapidly abolished by GABAA receptor antagonist application , while the astrocyte process' reshaping clearly operates ( if at all ) on much slower time scales . Similarly , previous studies have demonstrated synaptic and molecular selectivity of astrocytes- for example Houades et al . ( Houades et al . , 2008 ) have shown that gap junction density may be modulated along barrel boundaries , while Voutsinos-Porche et al . ( Voutsinos-Porche et al . , 2003 ) have shown a transient selective appearance of glutamate transporter in barrels which disappeared in the adult cortex . These prior studies further confirm the potential functional selectivity of astrocytes . However they are not relevant to the issue of the astrocyte morphology . Thus , selective aggregation of gap junctions may occur with or without anatomical confinement of astrocyte processes reported here . Astrocytes have been suggested to play a critical role in neurovascular coupling , controlling the blood flow upon an increase in neuronal activity ( Attwell et al . , 2010; Haydon and Carmignoto , 2006; Petzold and Murthy , 2011; Takano et al . , 2006 ) . Furthermore , Astrocyte processes have been demonstrated to manifest tight anatomical formations with neuronal dendrites and the activity of the brain actually arises from the coordinated activity of a network comprises of both neurons and astrocytes ( Araque et al . , 1999; Halassa et al . , 2007a; Perea et al . , 2009; Santello et al . , 2012 ) . Interestingly , a number of recent studies have demonstrated a principle of overlap-avoidance among individual astrocytes . Thus , astrocyte processes appear to form no-overlapping domains which may endow each astrocyte with a unique functional coupling to the neighboring neuronal population ( Halassa et al . , 2007b; Nedergaard et al . , 2003; Oberheim et al . , 2006 ) . An attractive hypothesis , suggested by the present results , could be that long term and habitual confinement of neuronal activation to specific functional domains may have gradually adapted the individual astrocyte arbor domains to the neuronal functional boundaries in parallel with the dendritic and axonal confinement . It is important to emphasize that the astrocyte confinement does not necessarily preclude blood flow across functional boundaries- since the capillaries themselves freely cross these boundaries . However , the confinement may allow a more efficient , precise and localized direction of blood flow to functional domains ( Petersen and Sakmann , 2001 ) . Compatible with this suggestion , optical imaging of blood flow during vibrissae activation reflects a clear confinement of blood supply within the barrel's borders ( Derdikman et al . , 2003; Woolsey et al . , 1996 ) . Indeed , as has been thoroughly examined in previous studies ( Blinder et al . , 2013 ) , it appears that the highly localized functional selectivity reflected in intrinsic optical imaging is likely due to direct and localized modulation at the level of the microvessels , while more pathological blockage of penetrating arterioles and venules can lead to a more wide-spread vascular effects . The mechanism that brings about astrocyte structural restriction is currently unknown . If , indeed such confinement serves an adaptive role , an interesting possibility is that this anatomical confinement reflects a dynamic plasticity process . In such a process , the distribution of astrocyte processes may reflect the level of sustained co-activation across neighboring neuronal assemblies , analogous to a Hebbian learning process . Relevant to this hypothesis , previous research has demonstrated activity-dependent dynamic changes in the coupling between astrocyte processes in the case of olfactory glomeruli ( Roux et al . , 2011 ) . It is noteworthy that this hypothesis reverses the causal chain . We do not propose that the morphological shaping of astrocytes somehow endows the underlying tissue with its functional boundaries . Rather , we hypothesize that the astrocyte processes are shaped by the functional selectivity of the underlying neuronal population . Finally , one cannot rule out a more interactive shaping , in which both systems remodel each other . More generally the finding that astrocyte processes respect functional boundaries could open the way to a variety of plasticity experiments . For example , examining astrocyte remodeling during development ( Diamond et al . , 1993; Olavarria et al . , 1987 ) , or studying the impact of functional deprivation , such as removing a single or a row of vibrissae on astrocyte morphology . Furthermore , global parameters such as age and pharmacological factors can be examined . Finally , the generality of the astrocyte confinement requires further examination . It should be noted that previous studies ( Colombo and Reisin , 2004; Oberheim et al . , 2009 ) showed that human astrocytes that populate the superficial cortical layer are larger in diameter ( 2 . 6 fold ) and show a more complex organization compared to rodents . However , despite these striking differences , our results revealed such confinement in all three systems , including human cortex . It will be interesting to examine whether similar anatomical astrocyte confinements can be observed in other well defined functional boundaries , and the parameters that determine the level of astrocyte confinement in different systems . Our demonstration that astrocyte density modulation can be revealed in micro-structures in human V1 opens the exciting possibility that the non-homogeneity in astrocyte spread could be employed as a new marker for additional areal , columnar and functional discontinuities in the human brain . This is particularly important since the currently available information about such functional boundaries in post-mortem brain tissue is confined to more basic anatomical and histological aspects ( Lorenz et al . , 2015; Zilles and Amunts , 2010 ) .
C57BL/6 mice and Wistar rats were used in this study . The mice and the rats were purchased from Harlan ( Jerusalem , Israel ) . Mice and rats ( females ) , 8–12 weeks of age , were used and kept in a specific pathogen free ( SPF ) environment . All experiments were approved by the Institutional Animal Care and Use Committee of the Weizmann Institute . To investigate the striate cortex of the human brain , we examined the right hemisphere of 35-year-old male patient , who died from non-neurological causes . The brain was photographed for purposes of orientation and scale . The occipital lobe through the calcarine sulcus up to the parieto-occipital sulcus , was removed ( see Figure 4—figure supplement 1A ) , and was cut into 4 coronal pieces . Two parts were embedded in paraffin blocks while the others were taken for free floating sectioning . Fluorescein isothiocyanate ( FITC ) -dextran , ( molecular weight 500 , 000 Daltons , Sigma , St . Louis , MO ) , 10 mg/ml , was dissolved in PBS and filtered through 0 . 45 µm mesh . 200 µl of the solution was injected into the tail vein of each animal . Ten min after FITC-dextran injection , mice were deeply anesthetized by a peritoneal injection of ketamin and xylazine ( 1:1 , Kepro , Holland ) , their brains were removed and post fixed in 2 . 5% paraformaldehyde in PBS , pH 7 . 4 , for 24 hr . The left and right cerebral hemispheres of the mice and rats were separated along the longitudinal fissure and the cortex of each side was divided from the underlying white matter . The isolated cortex was then gently flattened between two glass slides , separated by 1 . 4 mm spacer and soaked for 4–6 days in 1% paraformaldehyde in PBS . The flattened cortices were embedded in paraffin and cut tangentially parallel to the pial surface in thicknesses of 4 or 8 μm by a microtome ( Leica , Wetzlar , Germany ) . Sections were collected and mounted on SuperFrost+ slides ( Thermo Scientific , Waltham , MA ) . Dark field and fluorescent images were taken immediately after the cutting ( before the sections were dried ) by a Leica M165 FC stereo microscope connected to a digital monochrome camera ( Leica , DFC345 FX ) at magnifications of X1 and X3 . 5 . The human primary visual cortex was taken along the banks of the calcarine fissure of the occipital lobe ( right hemisphere , age 35 ) . The tissue was embedded in paraffin and cut coronally ( from pia to white matter , see Figure 4—figure supplement 1B ) . CO staining followed procedures described previously ( Horton and Hedley-Whyte , 1984; Murphy et al . , 1995; Wong-Riley , 1979 ) , with some modifications; sections were incubated in a solution containing 100 mg diaminobmuenzidine , 60 mg cytochrome C ( type III , Sigma ) and 40 mg catalase per 100 ml of 0 . 1 M phosphate buffer . The reaction usually took 5d at 39°C . The incubation solution was refreshed daily . The staining was done on four neighboring sections per sample , with an interval of about 18–20 µm . Paraffin sections were deparaffinized and rehydrated . Antigen retrieval was performed in 10 mM citric acid pH 6 for 10 min using a low boiling program in the microwave to break protein cross-links and unmask the antigens and epitopes . After pre-incubation with 20% normal horse serum and 0 . 2% Triton X-100 , slides were incubated with rabbit anti-GFAP ( 1:100 , Dako , Glostrup , Denmark ) and/or mouse anti non-phosphorylated neurofilament proteins ( NPNF , SMI-32; Covance , Dedham , MA , USA ) . ; diluted 1: 100 ) , at 4ºC for 7d . To enhance the signal , secondary antibodies , biotinylated anti rabbit or anti mouse ( 1:100 , Jackson ImmunoResearch , West Grove , PA ) were added for 90 min , followed by Cy2 or Cy3-conjugated Streptavidin ( 1:200 , Jackson ImmunoResearch West Grove , PA ) for 2 hr . Sections were counterstained with Hoechst 33 , 258 ( Molecular Probes , Eugene , OR ) for nuclear labeling . For single immunohistochemistry we the eliteABC kit ( Vector Lab Burlingame , CA , USA ) followed by DAB ( Sigma ) reaction . Stained sections were examined and photographed by a fluorescence or bright field microscope ( Eclipse Ni-U; Nikon , Tokyo , Japan ) equipped with Plan Fluor objectives ( 20x; 40x; 60x ) connected to a monochrome camera ( DS-Qi1 , Nikon ) . To form one large image of the total barrel field area , auditory cortex or V1 cortex , digital images at intermediate magnifications ( 20x ) were collected and stitched together automatically , using NIS element software ( Leica ) . Each large image of the barrel field consists of 60–120 tiles . In addition , to demonstrate the fine processes of astrocytes reaching the barrel border , some sections were imaged with a confocal scanning microscope ( LSM 700 , Zeiss ) . We used an oil immersion objective ( 100x , NA 1 . 40; Zeiss ) . A z-series of the entire 8 µm section was imaged at 0 . 42 µm spacing followed by a maximum intensity projection . The staining sensitivity to GFAP was substantially enhanced in the cortex ( up to five times ) by employing three procedures . First , staining was conducted on thin sections ( 4–8 µm ) . Second , an antigen retrieval technique was used , finally we used a long incubation sections with the first antibody ( 7d at 40ºC ) . Barrel ( mice ) as well as A1/A2 ( rat ) boundary demarcation was based on the dark field ( contrast inversed ) photographs of the sections in layer IV . The boundaries between cortical sublayers IIIa/IIIb were defined based on NPNF staining ( Figure 4A , B ) . The borders between blobs/inter-blobs were based on CO staining ( Figure 5D ) . These boundaries were then superimposed on the fluorescent images of the astrocytes processes or the capillaries . The superposition was based on careful alignment of the patterns of penetrating arterioles that were identified separately in the two images . To reveal the precise details of the distribution of astrocyte processes ( GFAP immuno-positive staining ) and the capillaries ( traced by FITC Dextran ) in the barrel field area , manual plotting was performed from the assembled images using Adobe Photoshop software ( Adobe Systems , San Jose , CA ) at magnifications of 20x . The septal borders were also drawn manually according to the dark field ( inversed ) images and superimposed on the astrocyte ( Figure 1D , Figure 2D and E ) or capillary drawings ( Figure 1B ) . The image alignment was done by adjusting the penetrating vessels of the images . A set of lines ( 375 µm length ) was drawn at the centers of the septa , aligned with the border's orientation . For comparison , lines of identical length and orientation were superimposed at the center of the barrel cores . These septa and core related lines were copied and superimposed on the astrocyte and capillary drawings ( Figure 2A–C ) . The alignment was done according to the penetrating blood vessels . Only the large barrels were analyzed and only the astrocyte branches/capillaries that crossed the lines were calculated . To reveal the precise details of the distribution of astrocyte processes ( GFAP immuno-positive staining ) in A1 versus A2 areas , straight lines ( 50 µm length ) were superimposed at a distance of 50 and 200 µm from the border of A1/A2 on both sides aligned with the border's orientation . These A1 and A2 lines were then superimposed on the corresponding drawings of the astrocyte processes ( Figure 3E and G ) and the number of astrocyte branches crossings were calculated . To quantitatively examine the distribution of astrocytes processes in layer III of the human primary cortex , around the border of IIIa/IIIb , we manually drew the distribution of astrocyte processes at high magnification ( see Materials and methods ) in both areas . Straight lines ( 375 µm length ) were superimposed on the NPNF image , 50 and 200 µm from the border of the two sublayers , aligned with the border's orientation and then superimposed on the corresponding drawings of astrocyte processes . To quantitatively examine the relationship between the intensity of cytochrome oxidase and astrocytic GFAP staining along cortical layer II and IIIa we superimposed density plots of both types of staining at three parallel lines and calculated the R-values using Pearson's coefficients of correlation . In addition we examined the distribution of astrocyte processes in blobs/inter-blobs zones . A set of vertical lines ( 500 µm length ) were drawn at the center of blobs , the center of inter-blobs and at the left margin of the inter-blobs . These lines were then copied and superimposed on the astrocyte drawings ( Figure 5D ) . The alignment was done according to the pattern of blood vessels . The astrocyte branches that crossed the lines were counted ( Figure 5E ) . All the analysis was performed by a person unaware of the experimental question . For the mice barrel cortex , 8 µm thick , sections were taken one from each of 5 left cortical hemispheres , and on two , 4 µm thick , single sections were taken from two additional hemispheres . The thin sections were analyzed separately to examine possible effects of section thickness on the septa/core crossing ratios . For the rat auditory cortex analysis , we used of one section of each of 4 left hemispheres . For the analysis of the human primary cortex , we used 8 sections from the same brain . The number of line crossings was compared between septa and barrels ( separately for astrocyte processes and capillaries ) in the mice barrel cortex , as well as between A1 and A2 in the rat auditory cortex and layer IIIa and IIIb in human primary cortex , using independent-sample Student’s t-test . The septa/barrels ratio was compared between astrocyte processes and capillaries using an independent-sample t-test . The septa/barrels ratio was compared between layers ( separately for astrocyte processes and capillaries ) using a one-way ANOVA on log-transformed ratios , followed by a Tukey post-hoc test where appropriate . All tests were performed by Statsoft's Statistica , version 12 . | The brain is subdivided into many specialized regions that each has distinct roles . A key aim of brain research is to define the boundaries of these areas . Researchers have attempted to map the transitions between brain regions by identifying changes in the properties and activity of neurons ( the cells that transmit information around the brain ) . However , these approaches cannot be used in some circumstances , such as when studying the living human brain , where only non-invasive experimental techniques can be used . Cells other than neurons are also present in the brain . Astrocytes ( a sub-type of glia cells ) are support cells that have an extensive array of branches that project from each astrocyte’s cell body , often giving it a characteristic star shape . Now , using high-magnification light microscopy , Eliam et al . show that the branches of individual astrocytes tend to avoid crossing the borders of brain regions with different roles . These changes in crossing densities define measurable boundaries between such subdivisions . These density-change boundaries formed by the astrocytes are present in multiple species – mouse , rat and human – and in multiple systems: touch , auditory and visual . This discovery could provide a new window into the functional organization of the brain . It may also offer insights into how the brain optimizes its blood-flow control across different subregions . The results of this study raise an additional question: is the confinement of astrocytes to single regions of the brain shaped by experience or is it present from birth ? Exposing animals to different sensory experiences at different developmental stages will hopefully shed further light on this phenomenon . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Astrocyte morphology is confined by cortical functional boundaries in mammals ranging from mice to human |
Oriented cell intercalation is an essential developmental process that shapes tissue morphologies through the directional insertion of cells between their neighbors . Previous research has focused on properties of cell–cell interfaces , while the function of tricellular vertices has remained unaddressed . Here , we identify a highly novel mechanism in which vertices demonstrate independent sliding behaviors along cell peripheries to produce the topological deformations responsible for intercalation . Through systematic analysis , we find that the motion of vertices connected by contracting interfaces is not physically coupled , but instead possess strong radial coupling . E-cadherin and Myosin II exist in previously unstudied populations at cell vertices and undergo oscillatory cycles of accumulation and dispersion that are coordinated with changes in cell area . Additionally , peak enrichment of vertex E-cadherin/Myosin II coincides with interface length stabilization . Our results suggest a model in which asymmetric radial force balance directs the progressive , ratcheted motion of individual vertices to drive intercalation .
A common characteristic of many tissues and organisms is an elongation along a primary dimensional axis . The oriented intercalation of cells is one of the fundamental mechanisms utilized to direct tissue elongation ( Keller et al . , 2000 ) . Tissue elongation is essential to the shaping of an elongated body axis ( Keller et al . , 2000; Irvine and Wieschaus , 1994 ) , as well as the development of many internal organs , such as the palate , cochlea , gut , and kidney ( Chalmers and Slack , 2000; Tudela et al . , 2002; Wang et al . , 2005; Lienkamp et al . , 2012 ) . Epithelial cell intercalation drives elongation of the Drosophila body axis during gastrulation , in a process known as germband extension ( GBE; Irvine and Wieschaus , 1994 ) . The intercalary behaviors driving GBE occur through a remodeling of cell topologies , with cells contracting shared anterior-posterior ( AP , vertical or T1 ) interfaces to a single point , followed by newly juxtaposed dorsal-ventral ( DV ) cells constructing horizontally-oriented interfaces between them ( Irvine and Wieschaus , 1994; Bertet et al . , 2004; Blankenship et al . , 2006; Collinet et al . , 2015; Yu and Fernandez-Gonzalez , 2016 ) . This is referred to as a topological T1 process , and results in a cumulative contraction of the embryonic epithelium along the DV axis , which helps to drive a perpendicular elongation along the AP axis . Previous research into the genetic factors associated with GBE has shown that global polarizing cues from maternal AP patterning are translated into asymmetric protein distributions at the cellular level ( Irvine and Wieschaus , 1994; Blankenship et al . , 2006 ) . At AP interfaces , Myosin II forms both supracellular cables and smaller , transient networks . Protein populations associated with adhesion ( E-cadherin , ß-catenin , Bazooka/Par-3 ) are found enriched at non-contracting interfaces ( Blankenship et al . , 2006 ) . This body of work led to a model in which actomyosin networks mediate higher line tensions along AP interfaces to direct contraction ( Fernandez-Gonzalez et al . , 2009; Rauzi et al . , 2008 ) . However , these studies have been limited to the molecular and mechanical characteristics of interfaces between two cells . The discrete regions where these interfaces overlap , tricellular vertices , have never been comprehensively examined . As a result of the focus on cell-cell interfaces , many studies on force-generation during intercalation have addressed forces oriented along the cell cortex ( Bertet et al . , 2004; Fernandez-Gonzalez et al . , 2009; Rauzi et al . , 2008; Rauzi et al . , 2010; Kasza et al . , 2014; Simões et al . , 2014; Collinet et al . , 2015; Yu and Fernandez-Gonzalez , 2016; Sun et al . , 2017 ) . However , Myosin II populations are highly active and are transiently present in multiple locations in epithelial cells , including in medial and apical cell regions ( Rauzi et al . , 2010; Fernandez-Gonzalez and Zallen , 2011; Sawyer et al . , 2011; Sun et al . , 2017 ) . These medial actomyosin networks drive a number of morphogenetic processes by mediating oscillations in cell area ( Martin et al . , 2009; Roh-Johnson et al . , 2012; Simões et al . , 2017; An et al . , 2017 ) . Indeed , during GBE medial Myosin II flows direct apical area oscillations that contribute to AP/DV anisotropy within a cell ( Rauzi et al . , 2010; Fernandez-Gonzalez and Zallen , 2011; Sawyer et al . , 2011 ) . However , the mechanisms by which they could be linked to cell-neighbor exchange have been unclear . Here , we demonstrate that vertices move independently of one another during T1 contraction , and exhibit distinct molecular dynamics that are required for effective intercalation . We show that intercalation proceeds through a sliding vertex mechanism that physically couples vertex motion to radially-oriented forces . E-cadherin and Myosin II are strikingly enriched at vertices , and this vertex enrichment coincides with length stabilization post-sliding . E-cadherin recruitment at vertices is coordinated with apical cell area oscillations , and is favored at vertices associated with AP interfaces . Finally , perturbing Myosin II function reduces E-cadherin enrichment and dynamics at vertices , and leads to a loss of productive intercalation . Together , these observations provide a mechanism by which area oscillations are coupled to cyclic molecular dynamics , and further introduce a link between the molecular properties of tricellular vertices and the emergent biophysical properties of the tissue .
To study in more detail how cell topologies are remodeled , we first examined the localization of endogenously tagged E-cadherin ( E-cad:GFP ) at the onset of intercalation . Strikingly , and in contrast to the previously reported homogenous distribution of E-cadherin along DV interfaces at mid-GBE ( Blankenship et al . , 2006 ) , we found that E-cadherin was highly enriched at vertices ( Figure 1A , B and C , and Figure 1—figure supplement 1A–C ) . This strong vertex-association of E-cadherin began at the onset of GBE , with E-cadherin diffusely present at apical cell interfaces prior to intercalation ( Figure 1D and D’ ) . Systematic quantification of vertex-associated E-cadherin averaged across the entire time of GBE ( see Materials and methods; Figure 1—figure supplement 1D ) also revealed that E-cadherin maintains a higher vertex:interface ratio than does a control , plasma membrane-associated marker ( Gap43:mCh; Figure 1B and C ) . Interestingly , although E-cadherin is enriched at vertices , it possesses a much wider distribution of intensities than the control marker , which suggested that vertex associated E-cadherin might undergo cycles of enrichment and dispersion ( Figure 1C ) . Analysis of vertex E-cadherin confirmed that the intensity of enrichment fluctuates in time ( Figure 1E and F; Video 1 ) . These results show that the vertex-specific localization of E-cadherin temporally coincides with intercalary movements , and suggest that vertex E-cadherin may function in directing GBE . As previous literature has focused on interface behaviors during GBE , and since we observed predominantly vertex-localized E-cadherin , we next developed a computational method for following individual vertex trajectories during interface contraction . A central physical expectation from previously described line-tension models ( Fernandez-Gonzalez et al . , 2009; Rauzi et al . , 2010 ) is that the inward movement of vertices connected by a contracting interface should show evidence of mechanical coupling ( Figure 2A ) . Surprisingly , however , we observed no evidence of this hypothesized coupling ( Figure 2B and C ) . Indeed , in systematic pair-wise analyses of cell vertices , physical coupling could only be observed in radial directions ( e . g . between vertices 3 and 6; Figure 2B’ and C ) . In other words , an inward correlation of vertex motion was only found between vertex pairs on opposite sides of the cell , with the largest correlations between those diametrically opposed ( Figure 2C ) . These results indicate that during the contraction of an AP interface , the motion of the two vertices toward the middle of the interface ( referred to as ‘productive’ motion ) occurs independently of each other , while all vertices undergo coupled motion into the radial direction . These results argue against a line tension-driven model of interface contraction , and suggest that intercalary movements should be reconsidered in terms of cell vertices and radially exerted forces . As we continued our analysis of vertex steps in vivo , a novel behavior began to emerge: cell vertices associated with contracting interfaces often underwent periods of productive sliding along the plasma membrane ( Figure 2D; Video 2 ) . This suggested that the uncoupled motion of T1 vertices is due to vertex sliding , a previously uncharacterized form of cell-shape deformation . Measurement of interface lengths showed that as a vertical interface contracts ( Figure 2D and E; blue ) the interface adjacent to it elongates ( Figure 2D and E; red ) , and consequently the total length stays constant ( Figure 2D and E; black ) . This compensatory increase in adjacent interface length is contrary to what would be expected through canonical models of interface contraction , in which the contracting interface shortens while adjacent interfaces maintain a constant length . Additionally , analysis of the lengths of all contracting and adjacent interfaces throughout GBE demonstrated that this behavior is a systematic component of T1-associated vertex movements ( Figure 2F ) . We also observe instances where vertices take turns sliding over the course of a T1 contraction ( Figure 2G and H ) . These results are consistent with the observed , uncoupled movement of vertices , and provide a mechanism by which a single vertex can independently produce the changes in cell shape that drive cell intercalation . Since our data demonstrate that the motion of vertices is coupled radially , we sought to understand the forces that could drive this movement . Previous research has shown that an apical actomyosin network drives area oscillations during GBE ( Rauzi et al . , 2010; Fernandez-Gonzalez and Zallen , 2011 ) ; Figure 3A and A’; Video 3 ) . This appears to be a common feature of many cell shaping processes , as similar apical area oscillations occur in the invagination of the Drosophila ventral furrow ( Martin et al . , 2009; Roh-Johnson et al . , 2012 ) , neuroblast ingression ( Simões et al . , 2017; An et al . , 2017 ) , and the internalization of the C . elegans endodermal precursor cells ( Roh-Johnson et al . , 2012 ) . To study the dynamics of cell vertices and interface lengths in terms of these radially oriented oscillations , we developed a computational assay to identify the instantaneous phase of cell area oscillations . We could then interpolate vertex motion and interface length changes in this area phase space for large numbers of cell oscillations ( see Materials and methods; Figure 3—figure supplement 1A–A’ ) . These instantaneous phase data follow a coordinate system where area contraction corresponds to angles from −180 to 0 degrees ( Figure 3B and B’; gray shading ) , while area expansion occurs in the period from 0 to +180 degrees ( Figure 3B and B’; blue shading ) . Using this method , we measured the systematic changes in interface lengths during apical area oscillations . Under isotropic conditions , the expected theoretical behavior would be that interface lengths would oscillate along with oscillations in cell area ( Figure 3C and C’; black ) . However , an intriguing behavior is observed when empirical interface lengths are plotted against cell phase . We observed a larger than isotropic decrease in vertical interface length during area contraction ( Figure 3C and C’; blue; an individual example is shown in Figure 3—figure supplement 1B ) . Notably , this decrease is preserved even as area contractions are reversed . Conversely , transverse interfaces undergo a smaller than isotropic decrease in interface length , and appear to undergo a compensatory increase in interface length during area expansion ( Figure 3C and C’; red ) . To account for the fact that area contraction and expansion generally are associated with decreases and increases in cell perimeter , we also plotted the effective length changes , or fractional length ( defined as interface length divided by cell perimeter ) , of vertical and transverse interfaces ( Figure 3D ) . The fractional length metric has the advantage that isotropic area changes do not result in changes of total fractional length . These results show that about 64% of the effective length contraction of vertical interfaces occurs during area contraction , while only 36% occurs during area expansion when absolute length is stabilized ( Figure 3D ) . Thus , vertex sliding occurs opportunistically during area contraction phases in a ratchet-like fashion , so that shortened vertical interface lengths are stabilized during area expansion . We also used this phasic analysis to examine cell contours during periods when cells are actively contracting in cell area as compared to periods when cells are expanding ( Figure 3E , F ) . We generated a convexity/concavity cell shape metric by comparing the area ratio between a theoretical straight-line Euclidean geometry that connects cell vertices to the experimentally defined cell contours . Interestingly , cells possess concave cell contours during contraction , and convex boundaries during apical area expansion ( Figure 3E–H ) . These results are consistent with cell vertices , but not cell interfaces , leading overall changes in cell shape and again suggest that cell vertices are key structures that govern cell topologies . We then examined the behaviors of E-cadherin in cells during observed oscillations in apical area . As cell areas contracted , we found specific vertices where E-cadherin became distinctly more enriched ( Figure 4A , yellow arrowhead ) . In these instances , the associated T1 interface length decrease observed during area contraction was stabilized during subsequent area expansion . In contrast , other vertices had very little E-cadherin enrichment , and the interface length change of these vertices scaled with area oscillations ( Figure 4A , white arrowhead ) . A cross-correlation function relating E-cadherin intensity and cell area showed that as areas decreased , vertex-associated E-cadherin intensity systematically increased ( Figure 4B; Figure 4—figure supplement 1A–A’’’ ) . Interestingly , this cross-correlation also showed that vertex E-cadherin intensity peaks just before cell area is in its most contracted state and before the onset of area expansion ( Figure 4B ) . Together , this phase relationship suggests that E-cadherin intensity is coordinated with cell area oscillations , and that enrichment of E-cadherin acts to stabilize sliding vertices . These results further indicate a model in which area oscillations provide the force for intercalation , with E-cadherin dynamics functioning as a molecular ratchet to harness vertex-sliding into productive movement . We then hypothesized that if oscillations in E-cadherin enrichments are essential to vertex movements and sliding , then artificially stabilizing and/or increasing E-cadherin at the plasma membrane should disrupt vertex displacement . We therefore inhibited endocytosis by injecting embryos with a small-molecule inhibitor , chlorpromazine ( Levayer et al . , 2011 ) . This increased the total amount of E-cadherin at cell vertices and interfaces compared to control embryos; additionally , cells maintained a vertex-associated enrichment of E-cadherin throughout GBE ( Figure 4C and D; Figure 4—figure supplement 2A ) . Importantly , endocytic inhibition prevented E-cadherin from oscillating to a lower enrichment state ( Figure 4E and F; Figure 4—figure supplement 2A; Video 4 ) . Although E-cadherin dynamics were disrupted , oscillations in cell areas still occurred ( Figure 4—figure supplement 2B; Video 4 ) . However , under these conditions we found that there was virtually no difference between the theoretical isotropic length change and the observed length changes for either vertical or transverse interfaces ( Figure 4G ) . To account for the possibility that vertex sliding is reduced due to other effects of endocytic inhibition we also measured length changes in E-cadherin overexpressing embryos and found that length ratcheting is also reduced ( Figure 4—figure supplement 2C–C’’ ) . Indeed , vertices in Ubiquitin-E-cadherin:GFP overexpressing embryos had displacements that were reduced by 41% as compared to control embryos ( Figure 4—figure supplement 2C’’ ) . Thus , stabilizing E-cadherin prevents vertex sliding and uncouples the ratcheted motion of individual vertices from the motive oscillations in cell area . This further suggests that increased E-cadherin levels restrict vertex sliding , and supports a model in which E-cadherin enrichment at the end of an apical area oscillation serves as a ratchet that stabilizes vertices post-sliding . To this point , our phase-based area analysis had focused only on individual cell behaviors ( Figure 5A ) . However , since vertices receive molecular and mechanical inputs from three different cells ( Figure 5B ) , we developed new computational approaches to study vertex motion with respect to the phases of all three involved cells . To do so , we first decomposed the displacement of a given vertex into two independent components: one tangential and one radial to the cell center ( Figure 5A’ , cyan and blue lines , respectively ) . Since radial displacement would be expected from isotropic area contraction , productive sliding displacements were defined from the analysis of tangential motions ( Figure 5A’ , cyan ) . This was then plotted against the phases of each cell that share a common vertex , permitting the measurement of the relative contribution of each cell to productive vertex motion . When tangential displacement rates were plotted with the phases of cells A and B ( the two AP neighboring cells at a T1 interface , Figure 5C’ ) , productive displacements occurred when both cells were contracting ( Figure 5C , C’ , E and E’ ) . However , when the phases of cells A and C were examined , productive displacements only occurred during the coordinated expansion of cell C with the contraction of cell A ( Figure 5D , D’ , E and E’ ) . Strikingly , the magnitude of displacement was much greater in this latter condition than in the first condition ( Figure 5E and E’ ) . This demonstrates that the phase of cell C is more predictive of productive vertex sliding events than cells A or B , and provides a mechanism for how anisotropic , radial force balance between three cells sharing a common vertex could couple individual T1-associated vertex steps to cell area oscillations . In light of this phase-correlated mechanical anisotropy , we then asked whether the molecular properties of vertices also displayed symmetry-breaking differences between AP and DV interface-associated vertices . We therefore examined E-cadherin intensities with respect to both area phase and vertex position ( Figure 5F ) . As before , we observed a strong increase in E-cadherin intensity as cell areas contracted ( Figure 5G ) . Interestingly , this correlation was observed specifically in vertices with polar angles near 90˚ , corresponding to vertices associated with T1 interfaces ( Figure 5G and H ) . Indeed , only contractile , T1-associated vertices show the maximal intensity increases that peak at the end of area contractions , while DV-associated vertices experienced reduced E-cadherin recruitment ( Figure 5G and H ) . Stabilizing E-cadherin by disrupting endocytosis reduced the correlation between area phase and vertex movements ( Figure 5—figure supplement 1 ) , again consistent with E-cadherin dynamics directing productive vertex sliding events . These results demonstrate that E-cadherin is specifically recruited to contracting vertices during apical cell area oscillations and suggest a mechanism to achieve anisotropic stabilization of vertex movements . Previous work has shown that medial Myosin II networks flow towards cell interfaces , and locally cluster E-cadherin for endocytic uptake ( Rauzi et al . , 2010; Levayer et al . , 2011 ) . This led us to examine whether medial Myosin II networks could similarly influence the localization and dynamics of E-cadherin at vertices . Simultaneous imaging of Myosin II:mCherry and E-cad:GFP revealed that , in addition to its previously described junctional and medial populations , Myosin II is highly , and dynamically , enriched at vertices ( Figure 6A–A’’; Figure 6—figure supplement 1A , B–B’’ with Zipper:GFP ) . This unexpected population of Myosin II exhibited a strong colocalization with vertex-associated E-cadherin ( Figure 6A and A’ ) . Furthermore , systematic quantitation of vertex Myosin II and vertex E-cadherin demonstrated a strong cross-correlation , showing that the dynamics of enrichment of both proteins are temporally coupled ( Figure 6B , Figure 6—figure supplement 1C–C’’’ , D ) . Interestingly , this cross-correlation function also revealed a slight temporal offset , with Myosin II becoming enriched at vertices about 3 . 3 s before E-cadherin ( Figure 6B ) . Similar to E-cadherin ( Figure 5G ) , phase-based analysis of Myosin II indicated an enrichment during periods of cell area contraction that is polarized to T1 associated vertices ( Figure 6C and C’; Figure 6—figure supplement 1E with Zipper:GFP ) . However , these dynamics again subtly preceded those of E-cadherin ( compare Figure 6C and C’ and Figure 5G and H ) . As was previously described for interfaces ( Rauzi et al . , 2010 ) , live imaging of individual cell behaviors showed that approximately 50% of medial Myosin II flows moved toward vertices , resulting in vertex enrichment ( Figure 6D and E , and Video 5 ) . Taken together , these observations show that Myosin II dynamics at vertices are similar to , but temporally precede , those of E-cadherin , and suggest a role for Myosin II in clustering E-cadherin at vertices . This additionally provides a potential mechanistic link between apical cell area oscillations and fluctuations in E-cadherin intensities at cell vertices . To investigate if Myosin II is required for vertex-associated E-cadherin behaviors , we functionally disrupted Myosin II by injection of the Rok inhibitor Y-27632 . Injection of a high concentration of Y-27632 ( 100 mM ) disrupted E-cadherin enrichment to cell vertices , as well as E-cadherin dynamics ( Figures 7A , C and E ) . Additionally , analysis of sqhAX mutant embryos produced a similar defect in vertex-associated E-cadherin ( Figure 7—figure supplement 1A–B ) . The phenotype of Y-27632 injected embryos was dose dependent , as injection of a lower concentration ( 25 mM Y-27632 ) preserved an enrichment of E-cadherin at cell vertices ( Figure 7B and D ) . In this background , however , E-cadherin was not significantly polarized to T1 associated vertices , and it lacked cycles of vertex enrichment ( Figures 7D , F and G ) . As a consequence , vertex displacement was severely disrupted in Y-27632 injected embryos at both concentrations ( Figure 7H and I ) . Interestingly , although productive vertex displacements are observed ( Figure 7H and I , red areas ) , these displacements are offset by backwards displacements ( Figure 7H and I , blue areas ) . These results are consistent with Myosin II functioning upstream of the ratchet-like behavior of E-cadherin at vertices . It is additionally important to note that cell area oscillations are highly disrupted in Y-27632 injected embryos ( Figure 7—figure supplement 1C ) , and that the observed lack of E-cadherin dynamics could represent a disruption of coupling between vertex populations of E-cadherin and medial actomyosin-dependent area oscillations . Either way , these results are consistent with a functional role for Myosin II in clustering E-cadherin at vertices and directing the stabilization of productive sliding movements .
In summary , we have shown that radial force coupling drives ratchet-like contractions of AP interfaces . These results also introduce a new functional unit capable of regulating cell topologies – tricellular vertices . We identify a new mechanism driving cell shape remodeling , in which tricellular vertices slide laterally in response to medial force generation ( Figure 7J ) . Much of the previous focus in studying intercalary behaviors has been on changes in cell adhesion and force generation at cell interfaces ( Bertet et al . , 2004; Blankenship et al . , 2006; Fernandez-Gonzalez et al . , 2009; Rauzi et al . , 2008; 2010; Kasza et al . , 2014; Simões et al . , 2014; Collinet et al . , 2015; Munjal et al . , 2015 ) . While higher line tensions at AP interfaces clearly exist and direct distinct aspects of intercalary cell behaviors ( such as interface alignment along the DV axis , recoil velocities upon laser ablation , and boundary element behaviors ) , it will be interesting to further explore the mechanisms regulating tricellular vertex function ( Rauzi et al . , 2008; Fernandez-Gonzalez et al . , 2009; Tetley et al . , 2016 ) . As vertices are connected to three ( or more ) interfaces as well as the radial coupling reported on here , their displacement will rely on the summed total of these local force contributions . Some of the mechanisms may involve the endocytic uptake of plasma membrane and adhesion proteins at interfaces which have been recently described ( Levayer et al . , 2011; Jewett et al . , 2017 ) . These molecular models of regulated adhesion are not necessarily dependent on line tensions , and could contribute to the biases in lateral vertex displacements in particular tangential directions . Although our data argue against a predominant function of interface-spanning line tensions in directing interface contraction , local regions of either medial- or interface-associated Myosin II are likely to impact vertex displacements as well . Indeed , although E-cadherin and Myosin II are primarily located at cell vertices early in GBE , by mid-GBE interface localization of both significantly strengthen and is consistent with interface as well as radial contributions to vertex displacements . Thus , cell vertices are well positioned to integrate the many different force-generating networks that will ultimately determine changes in cell shape and topology . Our results also show that intercellular adhesion dynamics are required for vertex movement . Under conditions in which E-cadherin exhibited greater enrichment and/or stability than is present in wild-type embryos , AP and transverse interface lengths oscillated identically , suggesting that cell vertices were unable to slide productively due to increased adhesive stability . Stabilizing E-cadherin was achieved by inhibiting endocytosis ( Levayer et al . , 2011 ) , and suggests that endocytic events may underlie the dispersion phase of E-cadherin dynamics at cell vertices . Given that endocytic pathways centered on asymmetric planar behaviors of Clathrin , Dynamin , and Rab35-dependent functions have been previously identified ( Levayer et al . , 2011; Jewett et al . , 2017 ) , it will be interesting to further explore if these same pathways function at or near cell vertices to direct E-cadherin dynamics , and whether endocytosis at vertices or at interfaces is more responsible for vertex E-cadherin redistribution . However , one interesting implication of our work is that these endocytic pathways should alter the balance of adhesion on either side of a vertex , potentially through the uptake of E-cadherin adhesion molecules , but the combined length of T1 and transverse interfaces would remain largely unchanged . This again suggests that the positioning of cell vertices will reflect the combined activities of contractile and adhesion elements that are located directly at the vertex as well as in local regions near the vertex . It also underlines the importance of cell vertices , and suggests a primary importance of vertices in determining cell topologies , which is further indicated by the strong enrichments of E-cadherin and Myosin II at cell vertices during early GBE . Regardless , it will require additional work to tease apart the contributions of E-cadherin stabilization and turnover pathways at cell vertices versus cell interfaces . We have also shown that as cells contract their apical area , E-cadherin and Myosin II are preferentially recruited to AP vertices to provide the adhesive force necessary to stabilize vertex position and interface length during area relaxation . This , as well as anisotropy in cell area oscillations and local imbalances in interface-associated forces ( Rauzi et al . , 2010; Fernandez-Gonzalez and Zallen , 2011; Sawyer et al . , 2011 ) , are likely responsible for enforcing the directionality of intercalation . Our results also suggest that , while radial forces in cells sharing a contracting AP interface are important for vertex displacement , vertex displacement has the strongest correlation with expansive motion in the adjacent , DV-oriented cells . This is intriguing , and suggests a homology to recent results during interface extension in which the adjacent cells provide motive force for extension rather than the cells that share the newly growing interface ( Collinet et al . , 2015; Yu and Fernandez-Gonzalez , 2016 ) . At the molecular level , AP vs DV anisotropy at the level of vertices could be a result of stress anisotropy and a mechanosensory feedback loop . Alternatively , vertices may also experience differentially positioned signaling networks , allowing for Myosin II and E-cadherin vertex enrichment . While the planar sliding behaviors described here are a novel mechanism underlying intercalary behaviors , it is also interesting to note that a similar adherens junction sliding behavior has been observed during three dimensional epithelial folding events as well as during cell ordering in the Drosophila notum ( Wang et al . , 2012; Wang et al . , 2013; Curran et al . , 2017 ) . In the formation of the epithelial folds that occur on the dorsal surface of the embryo in response to GBE , there is a basal shift in adherens junction position in response to Rap1 signaling events ( Wang et al . , 2012; Wang et al . , 2013; Takeda et al . , 2018 ) . It will be intriguing to explore if a similar Rap1-dependent pathway operates on cell vertices during cell intercalation . It is also highly interesting that , in the Drosophila pupal notum , a similar conservation of total junctional lengths , referred to as ‘continuous neighbor exchange’ has been observed ( Curran et al . , 2017 ) . It may well be that the repositioning of junctional/vertex elements will represent a new and conserved paradigm in how cell topologies are re-shaped during development . Finally , there is a growing body of work on tricellular vertices as unique epithelial structures . Previous studies have shown that cell vertices possess important molecular characteristics capable of coordinating complex cell morphologies ( Staehelin , 1973; Graf et al . , 1982; Schulte et al . , 2003; Ikenouchi et al . , 2005; Blankenship et al . , 2006; Byri et al . , 2015 ) . Indeed , recent work has shown that tricellular junctions act as key sensors of cell shape that serve as landmarks to orient epithelial cell divisions ( Bosveld et al . , 2016 ) , and intestinal stem cells require tricellular function to maintain appropriate homeostatic levels ( Resnik-Docampo et al . , 2017 ) . This suggests that vertices represent unique domains of the cell surface and that tricellular vertices have a special role as centers of functional signaling and physical networks within epithelial sheets .
Image and data analysis were performed in MATLAB . Cells were segmented using a seeded watershed algorithm ( see Figure 1—figure supplement 1A ) and tracked in time . The 'skeletonized' representation of the tissue directly yields vertex positions , interface contours , lengths and orientation angles , apical cell areas and perimeters , which we store together with cell-cell and vertex-vertex connectivity matrices . Cell areas were measured as the sum of the pixels within the contour of the watershed segmentation lines ( multiplied by the pixel area ) . Interface lengths were calculated as the Euclidian distances between the corresponding vertices . Line plots were generated as single pixel intensities along an interface line drawn manually in ImageJ , and the minimum value was subtracted from all data points and plotted using MATLAB . Images were cropped and leveled in Adobe Photoshop , and figures were prepared in Adobe Illustrator . Intensity measurements of vertices , local interfaces , and local background were automated using the watershed segmentation data to generate ROIs ( using distance transforms ) that were then used to measure average pixel intensities ( see Figure 1—figure supplement 1B for image of ROIs ) . Vertex ROIs were generated by making a binary matrix of the vertex pixel and using the distance transform to identify all pixels within 3 pixels of the vertex ( forming a pixelated disk seven pixels in diameter ) . The local interface and background intensities were measured within a 41-by-41 pixel square centered at the vertex , and local interface and background intensities were used to account for non-uniform illumination and varying junctional protein enrichment . Within the square neighborhood , the interface ROI was acquired using the distance transform to identify all pixels within 3 pixels of the interfaces and removing the vertex ROI . The local background ROI was acquired using a distance transform to identify all pixels at least seven pixels in distance from the segmentation lines: the seven pixel value was chosen to create a three pixel buffer zone between the interface/vertex ROIs and the background . The intensity ratio for each vertex was calculated as ( V-B ) / ( J-B ) where V is the vertex , B is the background , and J is the junction ( interface ) intensity measurements . Quantification of the intensity ratio was performed by averaging over all vertex time-points during GBE and all embryos ( Figures 1B , 6A’’ and 7E ) or resolved in time ( Figure 1C ) . To measure the dynamics of protein enrichment at vertices , we calculated the standard deviation of the intensity ratio trajectory of each vertex . To quantify this we averaged over all vertices . With respect to the embryo or the imaging field of view , vertices undergo primarily two different components of motion: drift ( or translational motion ) as the germband elongates , and cell-centric motion such as cell shape changes and intercalation movements . Thus , to analyze the motions associated with cell shape changes and intercalation , we removed the drift component by measuring vertex positions with respect to the centroid of the cell , c⇀ ( t ) =[xc ( t ) , yc ( t ) ] , where xc ( t ) and yc ( t ) are the x- and y-coordinates of the cell centroid at time t . The centroids of each cell were obtained over time using MATLAB’s regionprops function . In this cell-centric reference frame the position of a vertex is given by v⇀ ( t ) =[xv ( t ) −xc ( t ) , yv ( t ) −yc ( t ) ] , where [xv ( t ) , yv ( t ) ] is the position of the vertex in the image frame of reference . For measuring vertex tangential motion , vertex positions were converted to polar coordinates , r⇀ ( t ) =[r ( t ) , θ ( t ) ] , where r ( t ) is the radial position of the vertex from the centroid of the cell and θ ( t ) is the angle . Tangential displacements of the vertex are given by Δs=r×Δθ , where , by convention , displacements towards the opposing vertex of the interface were assigned a negative sign . Thus , interface-contracting displacements of an individual vertex are negative and interface-elongating displacements are positive . We quantified the motion coupling of each vertex with each neighbor vertex within the same cell via the cross-correlation at zero-lag of the vertices’ rate of displacement towards each other . The motions correlated are the components of displacement in the direction parallel to the two vertices , that is , along the line connecting the pair of vertices . The parallel component is calculated by taking the dot product of the displacement with the average vertex-to-vertex vector before and after displacement , the vertex position trajectories are then computed by taking the cumulative sum of the parallel displacement components . Rates of vertex displacement were calculated over 25 s to avoid correlating localization error ( Weber et al . , 2012 ) . Like pairs of vertices were combined in the quantification ( Figure 2C ) : 1 vs . 5 includes 2 vs . 4 representing horizontally coupled vertices , 1 vs . 4 includes 2 vs . 5 representing diagonally coupled , etc . The highlighted plot regions ( Figure 2B and B’ ) were manually selected to show periods in which the vertex is highly active in inward or outward motion . Vertex movements towards or away from each other result in a positive correlation , while a negative correlation value indicates that they move in the same direction . One prediction of vertex sliding is length ( i . e . plasma membrane ) compensation of the adjacent interfaces , that is as a vertex slides one interface gets longer by the same amount that the other gets shorter . To quantify this on time scales of interface contraction , we took the sum of the contracting interface with both of its transverse interfaces ( note: this accounts for the sliding motion of both vertices of an interface ) to see if total length was conserved . This sum was performed on the last 5 min of all available fully contracting interfaces that had a lifetime of at least 5 min . Of all the fully contracting interfaces that were present at the beginning of the movies this represented 91 . 8% . Individual cases were aligned such that the last time point was T = 0 min . Each contracting interface has two pairs of two transverse interfaces , those for cell A and for cell B , and both pairs were counted in the average . To get the instantaneous phase ( Figure 3B’ ) of an oscillating signal such as the area of a cell , we used a method known as the Osculating Circle Method ( Hsu et al . , 2011 ) , which is based on the Hilbert Transform but better suited for signals with non-zero mean . To get meaningful results using this method , the signal must be filtered to remove the noise that will result in artificial high frequency oscillations . A Savitzky-Golay filter of polynomial order three and frame size 45 s ( the exact number of frames varied based on the frame rate of the movies ) was used because it performed best at removing noise and preserving true cell area oscillations ( Figure 3—figure supplement 1A–A’ ) . Before applying the Savitzky-Golay filter , we de-trended the signal by subtracting a long time scale Gaussian filtered ( sigma = 180 s ) version of the signal . Instantaneous phase was shifted such that −180 to 0 degrees represents peak-to-trough and 0 to 180 degrees trough-to-peak . The Osculating Circle Method also generates an instantaneous amplitude , which was averaged over time to calculate each cell’s average oscillation amplitude ( Figure 4—figure supplement 2B and Figure 7—figure supplement 1C ) . Once the instantaneous phase trajectory of a cell’s area oscillations is obtained , other time-dependent readouts of the cell dynamics , such as the length of an interface , motion of a vertex , or intensities of a vertex , can be remapped from the time domain to the cell phase domain to average how those parameters change over the course of a cell’s oscillation cycle; this averaging is difficult or impossible in the time domain due to variations in oscillation cycle length . Mapping into the phase domain results in a 2D scatter plot of parameter values versus phase values for multiple cycles of phase; we interpolated these data by Gaussian weighted averaging of the parameter to a grid of phase values . To determine how interface lengths change over the course of an area oscillation cycle the data were mapped into the phase space of the adjacent cells , separately for the phases of cell A and cell B . For each phase cycle the interface length trajectory was shifted ( by subtracting the initial length ) so that it starts from L = 0 with each cycle . This ‘reset’ to zero with the start of each cycle was done so that positive lengths represent interface growth and negative lengths represent interface contraction with each phase cycle . The 2D set of data points ( phases and shifted lengths ) for all phase cycles ( from both cell A and cell B , for all AP interfaces , and all movies ) were combined and interpolated as described above . For the quantification ( Figure 3C’ ) we measured the average shifted length within two 10 degree phase bins: one bin centered at 0 degrees to capture the length change after the area decreasing phase and one centered at 175 degrees ( 170–180 degrees ) to capture the average length change after a full cycle . From the values in the bin centered at 175 degrees we subtract the average of the 0 degree bin to get the length change over just the area expanding phase . The methods for fractional length are the same as above , except that interface lengths are replaced with the ratio of the interface length divided by cell perimeter . Contour cell area is defined as the area within the watershed segmentation lines plus half the area of the pixels that make up the watershed boundary lines ( otherwise the contour areas would be artifactually small ) . The Euclidean area is defined as the area of the polygon whose vertices are the cell’s detected vertices . The contour over Euclidean area ratio ( or cell shape metric ) is a simple ratio of these two values . When interpolating the cell shape matric to phase space we used the phase of the Euclidean cell area . The quantification in Figure 3F was done on 10 degree phase bins centered at ±180 , ±90 , and 0° . To see whether individual vertex motion was correlated to apical cell area oscillation cycles , vertex displacements were tracked with respect to the phases of the three cells that make up the tri-cellular vertex . The particular type of vertex displacement measured was the cell-tangential component , which is perpendicular to the vector from the cell centroid to the vertex , and therefore independent , to the radial motion associated with cell area oscillations . Cell-tangential displacements were assigned a negative sign if the displacement contracted the vertical interface . For each vertex there are two tangential motion trajectories , one with respect to cell A , and one with respect to cell B , thus the methods described here were performed twice for each vertex , once with respect to each cell . For one vertex motion trajectory the instantaneous phases of cells A , B , and the other cell in the triad ( C or D ) are collected for each time point in the trajectory . Vertex displacements were interpolated to a 3D grid with the three cell phases on the three axes . To account for edge effects of interpolation , data near the edges ( ±180° ) were wrapped around by adding or subtracting 360° to the phases . Vertex intensities of Ecad:GFP and mCherry:Sqh were also interpolated into area phase space in a spatial angle resolved manner . Each vertex has an angle with respect to the centroid of each cell it borders , which also varies over time . These angles were measured with respect to a global reference angle , the angle along the dorsal-ventral axis of the embryo towards the ventral ( Figure 5F ) . The intensity trajectory of each vertex was normalized ( zero mean and s . d . of 1 ) . Normalized intensities were interpolated to a 2D grid of vertex angle and cell phase angle . To account for edge effects of interpolation , data near the edges ( ±180° ) were wrapped around by adding or subtracting 360° to the phases . To quantify the temporal correlation between E-cad and Myosin enrichment at vertices , we performed a cross-correlation of the rates of change of the vertex intensity ratios ( Figure 6B ) . Rates of change were calculated over 25 s to avoid correlating localization error . Signals were normalized such that mean and standard deviation were 0 and 1 , respectively , and an unbiased cross-correlation was computed via MATLAB’s xcorr function . The temporal lag was estimated by calculating the centroid of the positive peak ( Figure 6B ) . For medial myosin destination measurements , over 160 medial myosin flow events in 10 contracting cells were counted manually in the first 12 min of GBE in each of 3 E-Cad:GFP; mCh:Sqh movies . The flow of myosin to the area of 3 pixels radius around the tricellular vertex was considered as vertex destination , otherwise , it was considered as interface destination . Stocks were kept at 25°C and maintained by standard procedures . Fly stocks used in this study were endogenous E-cad:GFP ( gift of Y . Hong , University of Pittsburg , BL-60584 ) , Zipper:GFP ( Kyoto 115–082 ) , Gap43:mCh/TM3 ( gift of A . Martin , MIT ) , Ubi-E-cad:GFP ( BL-58471 ) , sqhAX3 ( BL-25712 ) , and mCherry:Sqh ( gift of A . Martin , MIT ) . E-cad:GFP and Zipper:GFP were expressed from the endogenous loci on the second chromosome and are homozygous viable . Embryos were collected on apple juice agar and dechorionated in 50% bleach for 2 min , then rinsed with water and either staged on apple juice agar or transferred to a gas permeable microscope slide and covered with Halocarbon 27 oil . All imaging was performed on a CSU10b Yokogawa spinning disk confocal from Zeiss and Solamere Technologies Group with a 63x/1 . 4 NA objective , with the exception of Myosin II movies , which were obtained on a CSUX1FW Yokogawa spinning disk confocal from Nikon and Solamere Technologies Group with a 60x/1 . 4 NA objective . Ecad:GFP; Gap43:mCh images are a summed projection of 5 z-slices taken at 0 . 5 μm steps starting sub-apically , at ~3 s/frame . Ecad:GFP; mCh:Sqh images are a maximum intensity projection of 8–10 z-slices taken at 0 . 75 μm steps , at ~6 s/frame . Following dechorionation as previously described , embryos were staged and aligned on apple juice agar , glued to a coverslip with heptane glue , and desiccated . Embryos were covered with Halocarbon 700 oil then injected with 10 mM chlorpromazine , or 25 or 100 mM Y-27632 . Embryos at the beginning of GBE were injected in the perivitelline space at 50% egg length . All measurements were quantified from a minimum of 3 embryos , and represented at least two individual trials . | Cells need to come together to form tissues of different shapes and sizes . Cells can move about in different ways to shape the tissues . For example , a process called cell intercalation is vital for creating elongated structures like the spinal cord and inner ear . In intercalation , a cell slots itself between neighboring cells to lengthen tissues in one direction . Most of the work to understand cell intercalation has examined the interfaces that form between two neighboring cells . But there are points called vertices where three cells make contact with each other . Vanderleest , Smits et al . have now used microscopy and computational analysis to examine these contact points , known as vertices , in fruit flies . It was thought that vertices that are connected by a single interface coordinate how they move . However , Vanderleest , Smits et al . now show that these connected vertices move independently of each other . Instead , the movements of unconnected vertices on opposite sides of the cell show coordination . Vanderleest , Smits et al . also found that two proteins build up at the vertices in the early stages of intercalation . One of these , called E-cadherin , enables cells to stick to each other . The other protein , called Myosin II , helps E-cadherin to localize to the vertices and also enables cells to contract . These results suggest that the vertices help to guide intercalation and changes in cell shape . Tracking the vertices over time revealed that they slide around the surface of the cells . During this sliding the total length of the interfaces that meet at the vertex remains the same – so as one becomes shorter , neighboring interfaces will become longer . This creates a zipper-like movement of the vertices that tugs the cells into line and suggests a new mechanism by which interconnected cells can change shape . Future work will focus on identifying the molecules that specify these unique vertex behaviors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2018 | Vertex sliding drives intercalation by radial coupling of adhesion and actomyosin networks during Drosophila germband extension |
The precise recognition of appropriate synaptic partner neurons is a critical step during neural circuit assembly . However , little is known about the developmental context in which recognition specificity is important to establish synaptic contacts . We show that in the Drosophila visual system , sequential segregation of photoreceptor afferents , reflecting their birth order , lead to differential positioning of their growth cones in the early target region . By combining loss- and gain-of-function analyses we demonstrate that relative differences in the expression of the transcription factor Sequoia regulate R cell growth cone segregation . This initial growth cone positioning is consolidated via cell-adhesion molecule Capricious in R8 axons . Further , we show that the initial growth cone positioning determines synaptic layer selection through proximity-based axon-target interactions . Taken together , we demonstrate that birth order dependent pre-patterning of afferent growth cones is an essential pre-requisite for the identification of synaptic partner neurons during visual map formation in Drosophila .
The identification of mechanisms that regulate the precise formation of neural circuits has been one of the major goals in developmental neurobiology . The Chemoaffinity hypothesis , formalized by Roger Sperry ( Sperry , 1963 ) , suggests that growing neurons must carry individual identification tags that allow the recognition between synaptic partners . Although many cell-type specific recognition molecules essential for neural circuit assembly have been identified in recent years ( reviewed in Missaire and Hindges , 2015; Yogev and Shen , 2014 ) , the precise developmental context in which these molecular tags control cell recognition and specify synaptic identity remain largely elusive . The temporal pattern in which different types of neurons are generated and specified has been shown to influence their connectivity during further development ( Kohwi and Doe , 2013; Osterhout et al . , 2014; Pujol-Marti et al . , 2012 ) . In addition to cell type specification , all neurons undergo similar steps of cellular differentiation including the growth of specific processes with different molecular and functional properties ( Rolls , 2011; Tahirovic and Bradke , 2009 ) accompanied by the expression of general neuronal molecules like N-Cadherin ( Gärtner et al . , 2015 ) . These molecules common to most neurons also influence axon targeting ( Brusés , 2011; Sakai et al . , 2012 ) and synaptogenesis ( Basu et al . , 2015; Bekirov et al . , 2008; Seong et al . , 2015 ) but how their ubiquitous expression can support neuronal recognition is not well understood . The Drosophila visual system , due to its highly stereotypic arrangement and genetic tractability , provides an excellent system to understand the mechanisms involved in neural circuit assembly ( Clandinin and Zipursky , 2002 ) . Each of the compound eyes is composed of approximately 800 units called ommatidia ( Campos-Ortega , 1980 ) and each ommatidium contains eight photoreceptor or retinula cells ( R1-R8 ) . Axons of R1-R6 photoreceptors terminate in the outermost lamina neuropile ( Fischbach and Dittrich , 1989 ) . In contrast , R8 and R7 axons project topographically through the lamina and terminate in the medulla ( Figure 1 A ) . This topographic projection leads to the formation of medulla columns that receive input from R7/R8 cells of the same ommatidium . Within the medulla column R8/R7 axons terminate in two different layers , M3 and M6 respectively ( Fischbach and Dittrich , 1989 ) , in which they contact their post-synaptic partner neurons ( Fischbach and Dittrich , 1989; Gao et al . , 2008; Karuppudurai et al . , 2014; Melnattur and Lee , 2011; Ting et al . , 2014 ) . 10 . 7554/eLife . 13715 . 003Figure 1 . Initial positioning of R cell growth cones in the developing medulla target field . ( A ) Overview of the developing Drosophila visual system at 24 hr APF . Arrows indicate the developmental gradient of photoreceptor differentiation in the retina and corresponding axonal targeting in the medulla neuropile . ( B ) Model of initial innervation of R cell axons and growth cone segregation in the medulla . C–E’’ . R7/R8 axon innervation in the medulla target field at different developmental stages . ( C , D , E ) R8 growth cones labelled with UAS-mCD8-GFP expressed under sens-Gal4 . C’ , D’ , E’ . R7 growth cones labelled with UAS-mCD8-GFP expressed under sev-Gal4 . C , C’ . At 6 hr APF , 18 R8 ( C arrowhead ) and 15 R7 ( C’ arrowhead ) axons innervate the medulla . D , D’ . At 12 hr APF , 24 R8 and 21 R7 axons innervate the medulla . In addition to 21 R7 axons that are already present in the medulla field , 22nd R7 axon can be seen entering at the anterior medulla ( D’ arrowhead ) . E , E’ . At 24 hr APF , 32 R8 and 29 R7 axons have innervated the medulla field . Scale bar shown in all images is 20μm . All photoreceptor axons are visualized using 24B10 antibody ( Fujita et al . , 1982 in red ) , R8 ( in A , C , D , E ) and R7 growth cones in ( C’ , D’ , E’ ) are stained with anti-GFP antibody ( in green ) and medulla neuropile is stained using anti-N-Cadherin antibody ( in Blue ) . C’’ , D’’ , E’’ . Quantification of the sequential innervation of R7/R8 axons in the medulla field . Error bars indicate Standard Deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 00310 . 7554/eLife . 13715 . 004Figure 1—source data 1 . R cell innervation quantification . The data shows quantification of total number of R8 and R7 axons innervating medulla neuropile at three different developmental stages ( 6/12/24 hr APF ) and axons labelled by 24B10 staining at each stage . The analysis shows the number of R7 and R8 axons for each brain as well as the average numbers for each stage at the bottom . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 00410 . 7554/eLife . 13715 . 005Figure 1—source data 2 . Sequoia expression quantification . The file depicts the average Normalised Fluorescence Intensity ( NFI ) of Sequoia expression analysed using anti-Sequoia antibody . The average NFI for each row of R8 cells is shown . Average NFI for Sequoia expression in each row was calculated for 60 different 3rd instar eye discs and was normalised against background to avoid individual differences in staining of each disc . The number represents the ratio of NFI for each R8 cell in a row against the background staining in each disc . To minimize the effect of variations in the staining of individual discs following modifications were made to imaging and analysis protocol- The discs were imaged as a single slice so that complete nuclei of all R8 cells are captured simultaneously to avoid fluorescence decay between slices . The fluorescence of each R8 nucleus was normalised against the background signal of exactly the same area . Further , to get rid of variation between samples , ratio of average R8 fluorescence against average fluorescence of background of each disc was considered while quantifying the NFI of Sequoia expression for each row . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 00510 . 7554/eLife . 13715 . 006Figure 1—figure supplement 1 . Onset of cell-type specific marker expression in the developing R cells . ( A ) Quantification of Sequoia expression levels in the R8 cells . The graph shown normalized fluorescence intensity plotted against rows in the 3rd instar eye disc . The Sequoia expression timing in R8 and R7 cells in the 3rd instar eye disc is shown in comparison to the photoreceptor cell type marker expression . ( B–D ) Expression of R cell specific drivers in the 3rd instar eye disc . B . R7 labelling with GFP expression under sevenless-Gal4 compared to R8 differentiation determined by anti-Senseless antibody . GFP expression can be seen from 3rd ommatidial row onwards , which is one row prior to the R7 specification . This is likely due to the expression of sev-Gal4 in the R7 precursor cells . Specific labelling of cell types is indicated in the figure panel . ( C ) R8 labelling by senseless-Gal4 compared to Chaoptin expression in the 3rd instar eye disc . Chaoptin expression ( visualized using 24B10 antibody ) can be detected from 9th ommatidial row onwards . ( D ) GFP expression under senseless-Gal4 compared to R8 differentiation determined by anti-Senseless antibody . GFP expression can be seen from the 4th ommatidial row onwards suggesting a delay in the Gal4 driven expression of GFP as compared to Senseless protein expression , visible from 1st row onwards . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 006 The layer specific targeting of R7/R8 axons can be divided into two main developmental phases: First , targeting of R cell axons to distinct temporary layers in the early medulla and second , the selection of correct synaptic target layer within the mature medulla neuropile ( Hadjieconomou et al . , 2011; Ting and Lee , 2007; Ting et al . , 2005 ) . Multiple cell type specific molecules involved in layer specific targeting of R8 axons ( Hakeda-Suzuki and Suzuki , 2014; Lee et al . , 2003; Ohler et al . , 2011; Pappu et al . , 2011; Senti et al . , 2003 Shinza-Kameda et al . , 2006; Timofeev et al . , 2012; Tomasi et al . , 2008 ) and R7 axons ( Astigarraga et al . , 2010a; 2010b; Choe et al . , 2006; Lee et al . , 2001; Morey et al . , 2008; Nern et al . , 2005; Prakash et al . , 2010; Ting et al . , 2005 ) have been identified . Interestingly , most of these molecules function in the second phase of axon targeting and the molecular mechanisms governing the initial innervation of R8 and R7 axons as well as the importance of this temporary positioning for subsequent synaptic layer targeting remain elusive . The Zn finger transcription factor Sequoia and the cell adhesion molecule N-Cadherin are both expressed in R7 as well as R8 cells but are primarily required for the temporary layer positioning of R7 axons ( Lee et al . , 2001; Petrovic and Hummel , 2008; Ting et al . , 2005 ) . Additionally , the LRR molecule Capricious , expressed only in R8 cells , has been described to control R8 axon targeting during the second step ( Shinza-Kameda et al . , 2006 ) but its role in the initial targeting of R8 axons to temporary medulla position has not been addressed despite an early onset of expression . Here we show that early self-patterning of R7/R8 afferents , mediated by relative difference in Sequoia levels , organizes an initial topographic map . This is achieved by a birth-order defined sequence of R7-R8 growth cone segregation leading to their differential positioning in the target field . Shortly afterwards , cell adhesion molecules , like Capricious in R8 , consolidate these growth cone positions , which is critical for subsequent steps of synaptic partner recognition .
In the developing Drosophila visual system , photoreceptor axons project from the eye disc into the optic lobe and target to the lamina and medulla neuropiles ( Figure 1A ) . Photoreceptor differentiation begins in the 3rd instar eye disc in a defined sequential fashion with R8 specified first followed by the outer R1-R6 cells and finally R7 in every ommatidium and can be visualized in developing ommatidial rows ( Figure 1B , Tomlinson and Ready , 1987 ) . We examined how R8/R7 sequential specification is represented in the arrival of R cell axons at the medulla target region using cell type specific reporter lines ( sens-Gal4 for R8 , sev-Gal4 for R7 , Figure 1C–E” , Figure 1—figure supplement 1 ) . By quantifying the number of R8/R7 growth cones at three consecutive stages of early pupal development ( 6/12/24 hr After Puparium Formation , APF ) , we could show that R cell axon innervation in the medulla mirrors the temporal pattern of R cell specification and follows a consistent sequence of growth cone segregation for each stage ( Figure 1C–E’ , summarized in B ) . At the anterior edge of the medulla the youngest R8 axons arrive sequentially and position at the superficial layer of the medulla neuropile ( 'R8 positioning' , Figure 1C arrowhead ) , followed by the arrival of the first R7 axon at the respective R8 position about 6 hr later and three columns posterior to the youngest R8 axon ( Figure 1C’ arrowhead ) . For the next 6-8 hr , represented by 3–4 columns , R7/R8 growth cones are in close contact followed by their segregation into adjacent positions , with R7 growth cones locating proximal with respect to the superficial R8 growth cones in the corresponding columns ( 'R7-R8 segregation' , Figure 1C’-E’ ) . Afterwards , R7 growth cones move deeper into the medulla neuropile whilst maintaining their columnar topography and a separate R7 temporary layer becomes visible ( 'R7 positioning' , illustrated in Figure 1B ) . As reported previously , Sequoia is critical for R cell target layer selection ( Petrovic and Hummel , 2008 ) . Next we tested if Sequoia is involved in the regulation of these initial steps of growth cone positioning . Using MARCM ( Lee and Luo , 2001 ) we generated sequoia mutant R8 cells by activating Flippase under heat shock promoter and visualized their growth cones with an R8 specific reporter line ( sens-Gal4 , Figure 2A–B’ ) . Axons of sequoia mutant R8 cells project to the anterior medulla in the wild type sequence but convergence of neighbouring R8 growth cones can be observed , disrupting the topographic organization ( Figure 2A–B' ) . Similar phenotype in R8 growth cone positioning can be observed in eye specific sequoia mosaics ( ey3 . 5-FLP ) , excluding an effect of unlabelled sequoia mutant cells generated in the brain in the hs-FLP background ( data not shown ) . 10 . 7554/eLife . 13715 . 007Figure 2 . Sequoia mediates growth cone segregation of R cell axons in the medulla . ( A , A’ ) Wild type position of R8 growth cones as they arrive at the anterior region of the medulla ( A’ ) . ( B , B’ ) sequoia mutant R8 growth cones converge upon arrival at the anterior medulla ( B’ arrowhead ) leaving gaps in their normal position ( B’ arrow ) . ( C , C’ ) Growth cones of wild type R7 cells segregate from R8 growth cones at 6 hr APF and are positioned immediately proximal to the R8 growth cones in respective columns . ( D , D’ ) Growth cones of sequoia mutant R7 cells fail to segregate from R8 growth cones and are positioned with R8 growth cones at the superficial medulla position at 6 hr APF ( D’ arrowhead ) . ( E , E’ ) Wild type R7 growth cones reach their temporary target layer in the deeper medulla position at 24 hr APF . F , F’ . sequoia mutant R7 growth cones fail to reach their temporary target layer and remain at the superficial medulla position with R8 growth cones at 24 hr APF ( G’ arrowheads ) . ( G , G’ ) Two neighbouring sequoia mutant R7 growth cones ( G’ arrows ) converge into a single column ( G’ arrowhead ) in the superficial medulla position . ( H , H’ ) Wild type R7 axons target to medulla layer M6 and R8 axons target to layer M3 in the adults ( H’ arrowhead ) . ( I , I’ ) sequoia mutant R7 axons mis-target to M3 , the target layer for R8 axons ( I’ arrowhead ) leaving layer M6 empty ( I’ asterisk ) . ( J , J’ ) Wild type R7 axons target to medulla layer M6 in the adult ( J’ arrowhead ) even when R8 axons are retained in layer M1 due to expression of Golden Goal ( GMR-gogo ) . ( K , K’ ) sequoia mutant R7 axons in presence of GMR-gogo mis-target to medulla layer M1 along with R8 axons ( K’ arrowheads ) leaving both layers M3 and M6 empty ( K’ asterisks ) . Schematics in all panels illustrate growth cone positioning ( A–G’ ) or axon targeting ( H–K’ ) phenotypes and numbers indicate quantification of respective phenotype . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 00710 . 7554/eLife . 13715 . 008Figure 2—figure supplement 1 . Cell-autonomous effects of Sequoia loss-of-function in R8 cells ( A , A’ ) Wild type single R8 cell clones . A' Shows a single R8 clone in the medulla . A’ . Wild type R8 cell growth cone is positioned correctly in the superficial medulla position and is segregated from its wild type neighbour growth cones ( A’ arrowhead ) . B–B’ Sequoia loss-of-function leads to cell-autonomous defects in single R8 cell clones . B . Shows a single sequoia mutant R8 cell generated in the medulla . B’ . Shows defects in the segregation of sequoia mutant R8 cell which leaves its normal position , leaving a gap ( B’ arrow ) and converges onto the neighbouring wild type cell leading to non-segregation ( B’ arrowhead ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 008 Similar to R8 , no defects can be detected in the extension and arrival of sequoia mutant R7 axons but they fail to segregate from R8 growth cones within the same column as early as 6 hr APF ( Figure 2C–D' ) . Subsequently , these sequoia mutant R7 growth cones remain at the superficial medulla position together with R8 growth cones even at 24 hr APF ( Figure 2E–F' ) . Interestingly , sequoia mutant R7 growth cones from two neighbouring columns also converge in the superficial medulla position ( Figure 2G–G' ) . Single cell analysis revealed a cell-autonomous function of Sequoia in growth cone segregation ( Figure 2—figure supplement 1A–B' , arrowhead ) . Therefore , loss of Sequoia function disrupts the sequential segregation of R8-R8 , R8-R7 and R7-R7 growth cones within as well as between layers . The targeting phenotype of sequoia mutant R7 cells varies depending on the genetic background of R8 in the same medulla column: In case of wild type R8 axons , sequoia mutant R7 from the same ommatidium will terminate together in M3 ( Figure 2I , I' ) , whereas a sequoia mutant R8 and R7 co-terminate in M1 ( Petrovic and Hummel , 2008 ) . To test if mis-targeting of sequoia mutant R7 cell axons is caused by a change in their target layer recognition or the consequence of an afferent segregation defect , we analysed the phenotype of wild type and sequoia mutant R7 axons in a background with a modified R8 axon position . R8 axons were retained in the superficial medulla layer M1 via Golden-goal ( gogo ) over-expression using GMR-gogo , which has no effect on the targeting of R7 axons ( Tomasi et al . , 2008 ) . In this background , we labelled wild type and sequoia mutant R7 cells using MARCM and analysed axon targeting in the adult medulla . Wild type R7 growth cones segregate from ectopic R8 growth cones and reach the layer M6 ( Figure 2J–J' ) , indicating that the ectopic position of R8 axons does not influence wild type R7 axon targeting . In contrast , sequoia mutant R7 axons mis-target to layer M1 along with ectopic R8 axons ( Figure 2K–K' ) . This indicates that mis-targeting of sequoia mutant R7 axons is the consequence of growth cone segregation defects rather than a change in target layer recognition . In summary , Sequoia regulates two main steps of growth cone segregation of R7/R8 axons . For axons of the same R cell type but from different ommatidia , Sequoia supports point-to-point spacing within the temporary layer and thereby controls the subsequent columnar restriction . In addition , for the pair of R7/R8 cells from a single ommatidium , which innervate the same medulla column , Sequoia mediates the segregation of growth cones between layers . Sequoia shows a highly restricted expression during early differentiation of all R cells ( Petrovic and Hummel , 2008 ) . A short peak of high Sequoia expression at the onset of R cell differentiation is followed by a rapid decline in protein levels ( Figure 1—figure supplement 1A graph ) . Due to the sequential development of ommatidia in the eye field as well as R cell types within each ommatidium ( Tomlinson and Ready , 1987 ) , the short peak of Sequoia expression in each cell leads to small differences of Sequoia levels in sequentially projecting R8 cells from adjacent ommatidial rows ( 'inter-ommatidial differences' , Figure 1—figure supplement 1A , Figure 3K1a ) . In addition , each R7 develops approximately 8 hr after R8 in every ommatidium corresponding to four ommatidial rows in the eye disc . Therefore , the highest difference in the expression levels of Sequoia can be found between these two cell types with almost no detectable Sequoia in R8 at the time of maximal Sequoia expression in the R7 of the same ommatidium ( 'intra-ommatidial differences' , Figure 3—figure supplement 1C , inset 3 , Figure 3K1b ) . 10 . 7554/eLife . 13715 . 009Figure 3 . Relative levels of Sequoia mediate growth cone segregation . ( A ) Wild type R7 growth cones segregate within the deeper medulla position thereby maintaining the topographic columnar arrangement . ( B ) Mis-expression of Sequoia in R7 cells alone under sev-Gal4 ( all R8 Seqlow , all R7 Seqhigh ) does not affect segregation of R7 growth cones from R8 growth cones but disrupts segregation among R7 growth cones within deeper medulla position . This leads to the loss of topographic arrangement illustrated by gaps in the deeper medulla position ( B’ arrowheads ) . R8 and R7 growth cone positions in the medulla are indicated in A’–B’ . C–E” . Wild type and Sequoia mis-expressing clones of R7 cells are generated using GMR-FLP induced MARCM . ( C’ ) shows a single cell and C’’ shows a two cell R7 clone ( C’ , C’’ arrowheads ) . ( D , E ) Sequoia mis-expressing clones of R7 and R8 respectively . D’ shows a single Sequoia mis-expressing R7 cell clone ( R7 Seqhigh ) that segregates from growth cones of neighbour wild type R7 cells ( R7 Seqlow ) and extends beyond the normal R7 position into the medulla; D’’ shows a two cell clone of neighbouring R7 cells that mis-express Sequoia ( R7 Seqhigh-R7 Seqhigh ) but do not extend growth cones beyond the normal R7 position in the medulla ( D’’ arrowhead ) . ( E’ ) Single Sequoia mis-expressing R8 cell clone ( R8 Seqhigh ) extends growth cone to the medulla regions beyond the R7 position ( E’ arrow ) thus leaving a gap in the superficial R8 position in the medulla ( E’ arrowhead ) . ( E” ) Multiple R8 cell clones exhibit shift of their growth cones to deeper medulla position ( E” arrowheads showing empty R8 position ) but are retained in this position similar to the two cell R7 clones ( R8 Seqhigh-R8 Seqhigh E” arrow ) . The brain regions are visualized using anti-N-cadherin antibody ( in blue ) and labelling of photoreceptors axons is indicated in the figure panels . ( F ) Quantification of the overshooting phenotype at the Seqlow -Seqhigh clone boundary exhibited by single vs . multiple cell clones of R7 and R8 cells . Sequoia expression in- all R cells- R7 n=96 , R8 n=128 , multiple cell clones- R7 n=37 , R8 n=43 , two cell clones- R7 n=97 , R8=6 and single cell clones- R7 n=121 , R8 n=7 . ( G–J” ) Lamina plexus assembly is disrupted by loss and gain of Sequoia function . ( G , G” ) - Wild type clones of lamina targeting R1-R6 cells are generated using ey3 . 5-FLP and are labelled with ro-τ-LacZ to visualize growth cones of R2/R5 cells . ( H , H” ) Disruption of lamina plexus assembly in sequoia mutant R1-R6 clones ( H’ arrowhead ) . R2/R5 cell growth cones visualized with ro-τ-LacZ . ( I , I” ) Wild type clones of R1-R6 , generated using ey3 . 5-FLP and labelled with LGMR-Gal4 , target normally to lamina plexus . ( J , J” ) Sequoia mis-expressing clones of R1-R6 cells labelled with LGMR-Gal4 . Growth cones of Sequoia mis-expressing R1-R6 cells ( labelled with GFP ) segregate from the growth cones of wild type cells and assemble into an additional layer between marginal ( MaG ) and medulla glia ( MeG ) cells ( J’ arrowheads ) . G”–J” show schematics of wild type , sequoia mutant and Sequoia mis-expressing R1-R6 growth cones in the lamina plexus . ( K ) Schematics depicting ( 1 ) the relative differences in Sequoia expression levels among R8-R8 and R8-R7 cells in wild type development , ( 2 ) R cell growth cone phenotypes in Sequoia loss and gain of function scenarios . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 00910 . 7554/eLife . 13715 . 010Figure 3—source data 1 . R cell axon overshooting quantification . The overshooting phenotype of R cell axons in Sequoia gain of function clones was quantified and calculated as percentage of axons overshooting . Each brain was manually analysed and total number of clone axons ( GFP positive ) were individually counted against number of overshooting axons . The row 1 depicts type of clone , row 2 depicts cell type , row 3 depicts the percentage of overshooting axons and row 4 shows the raw number of axons counted . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01010 . 7554/eLife . 13715 . 011Figure 3—figure supplement 1 . Loss and gain of Sequoia function in single lamina targeting R cell-type ( R4 ) is sufficient to induce changes in growth cone segregation . ( A ) Wild type R7 growth cones segregate from R8 growth cones present in the superficial medulla position and are positioned in the deeper medulla position . ( B ) Prolonged expression of Sequoia in both R7 and R8 cells under LGMR-Gal4 ( all R8Seqhigh , all R7 Seqhigh ) leads to convergence of their growth cones in the deeper medulla position . ( C ) Expression of Sequoia in 3rd instar eye disc . The coloured blocks depict the expression period of Sequoia in R8 ( red ) , R1-R6 ( green ) and R7 ( blue ) . 1 , 2 and 3 insets show Sequoia expression in R8 , R1-R6 and R7 respectively . ( 1 ) Shows high expression of Sequoia in R8 cells ( asterisks ) . ( 2 ) Shows the loss of Sequoia expression in R8 cells ( arrows ) whereas high Sequoia expression in R1-R6 cells ( asterisks ) . ( 3 ) Shows high Sequoia expression in R7 cells and no Sequoia expression R8 as well as R1-R6 cells . The large ( intra-ommatidial ) difference in Sequoia expression levels between R8 and R7 cells is depicted by outlining the R7/R8 cells in same ommatidium . No detectable expression of Sequoia in R8 ( 3 , arrows ) and high Sequoia expression in R7 ( 3 , arrowheads ) . ( D–F’ ) Sequoia loss and gain of function affects segregation of even single cell type ( R4 ) in the lamina . ( D ) Clones of wild type R4 cells in the lamina plexus . R4 growth cones are labelled using mδ0 . 5 Gal4 and show normal patterning within lamina plexus . ( E , E’ ) sequoia mutant R4 cells labelled with mδ0 . 5 Gal4 show abnormal patterning in the lamina plexus and disruption of lamina assembly is demonstrated by occurrence of gaps in the plexus ( B’ arrowhead ) . ( F , F’ ) Sequoia mis-expression under mδ0 . 5 Gal4 leads to segregation between growth cones of Sequoia mis-expressing R4 cells ( labelled with GFP ) and wild type R1-R6 cells , and shift of Sequoia mis-expressing R4 cell growth cones to a distinct position within the lamina neuropile . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 011 To determine the role of Sequoia expression dynamics in R7/R8 growth cone positioning we next analysed how inter- and intra-ommatidial differences in Sequoia levels influence growth cone segregation . Abolishing the inter-ommatidial differences via prolonged Sequoia expression using LGMR-Gal4 leads to a convergence of R7/R8 growth cones in the deeper medulla position ( Figure 3—figure supplement 1A–B' ) . Similarly , equalizing Sequoia levels in only R7 cells using sev-Gal4 , which shows transient , low-level expression as compared to LGMR-Gal4 , resulted in frequent R7 growth cone convergence and corresponding innervation gaps in the deeper medulla layer ( Figure 3A–B' , arrowheads in B’ ) . This result shows that the loss and gain of Sequoia activity leads to similar growth cone convergence phenotypes indicating that the presence of Sequoia expression is not sufficient to mediate growth cone segregation . To determine if the observed differences in Sequoia levels between R cells are critical for their growth cone segregation into different layers , we sought to model the endogenous Sequoia expression difference between R8 and R7 cells among neighbouring R7 cells targeting within a layer . For this , we analysed the axon projections in Sequoia gain-of-function R7 cell mosaics ( Seqhigh ) at the clonal boundary with adjacent wild type R7 cells ( Seqlow; Figure 3C–F ) . Here we observed that the majority of single Seqhigh R7 axons segregate from the surrounding wild type Seqlow axons and project deeper into the medulla ( Figure 3D' ) . In contrast to the single cell clones , most of the mosaics with two adjacent Seqhigh R7 cells show growth cone convergence and termination at the appropriate medulla position ( Figure 3D” ) . Interestingly , single Seqhigh R7 cells from distant columns , which overshoot the temporary layer , often converge in the inner medulla neuropile ( Figure 3D' arrow ) or even within the lobula complex ( data not shown ) . Similar to R7 , mosaics of R8 cells with different Sequoia expression levels segregate during initial axon positioning: single Seqhigh R8 axons segregate from the surrounding wild type Seqlow R8 axons ( Figure 3E' arrowhead ) and extend further into the medulla neuropile ( Figure 3E' arrow ) , whereas two or more neighbouring Seqhigh R8 growth cones converge and terminate at the R7 position ( Figure 3E" arrow ) . Thus , together with the loss-of-function , this gain-of-function analyses show that ‘equalized’ Sequoia levels prevent growth cone segregation thereby highlighting the role of relative Sequoia levels in this process ( summarized in Figure 3K2 ) . To test if relative differences in Sequoia levels also control growth cone patterning of other photoreceptors , we analysed the outer R1-R6 cells . These neurons develop in a temporal window between R8 and R7 specification ( Tomlinson and Ready , 1987 ) and their growth cones are positioned in a single temporary layer , the lamina plexus , which corresponds to their small relative differences in Sequoia expression ( Figure 3—figure supplement 1C , Figure 3G , G' ) , We generated sequoia mutant clones of R1-R6 cells using ey3 . 5-FLP and visualized the growth cones of R2/R5 cells ( ro-τ-LacZ ) or R4 cells ( m∂0 . 5-Gal4 ) ( see Materials and methods ) . The sequoia mutant R1-6 growth cones fail to segregate normally , illustrated by gaps within the lamina plexus ( Figure 3H–H” , Figure 3—figure supplement 1E-E’ ) , which is similar to the defects in inter-ommatidial R8 as well as R7 growth cone segregation . Strikingly , clones of Seqhigh R1-6 cells segregate their growth cones from the wild type ( Seqlow ) ones thereby organizing the formation of a novel layer within the lamina neuropile ( Figure 3I–J” , Figure 3—figure supplement 1F , F’ ) . This demonstrates that large differences in Sequoia levels among R cells , either inherently present as in case of developing R7/R8 cells or ectopically generated as in case of R1-R6 cells , segregate their growth cones into distinct layers . In all cases , cells with higher Sequoia levels position their growth cones deeper in the neuropile . From these data we propose a 'Growth Cone Segregation' model , in which small differences in Sequoia levels lead to an evenly–spaced growth cone positioning within a layer , whereas large Sequoia differences result in segregation of growth cones into separate layers . In contrast , equal levels of Sequoia in projecting R cells , either in loss- or gain-of-function context , cause the convergence of growth cones and termination of axonal extension ( Figure 3K ) . We next determined the temporal dynamics of initial R8/R7 growth cone segregation into superficial vs . deeper medulla layers . To temporally restrict the Gal4-driven expression of Sequoia , we utilized the Gal80ts ( TARGET system , McGuire , 2003 ) . At the permissive temperature ( 18°C ) functional Gal80 prevents Sequoia expression whereas an inactive Gal80 at the restrictive temperature ( 29°C ) allows the induction of Sequoia expression . Using this method , an R8 growth cone shift to deeper medulla position can be observed following Sequoia expression until 24 hr APF ( Figure 4—figure supplement 1C ) . In contrast , an onset of Sequoia expression after 24 hr APF , at which time all photoreceptor axons have arrived in the target field , has no effect on R7/ R8 growth cone segregation ( Figure 4—figure supplement 1D ) . This indicates that R7/R8 cells are sensitive to Sequoia-induced growth cone segregation only during a narrow developmental window . Interestingly , although R8 growth cones do not change the layer upon Sequoia expression after 24 hr APF , they still respond to these elevated Sequoia levels by leaving their topographic position and converging onto neighbouring growth cones ( Figure 4—figure supplement 1D ) . This indicates that the restriction of R cell growth cones to distinct temporary layers occurs immediately following their segregation whereas the process of columnar restriction continues into the later steps of visual map formation ( Ferguson et al . , 2009; Ting et al . , 2007 ) . To further characterize the developmental sequence of the transition from initial positioning to the consolidation of R8 growth cones , we induced short pulses of Sequoia expression in early pupal stages using the Gal80ts method described above and analysed R8 growth cones at 24 hr APF ( Figure 4A–F ) . Upon the induction of Sequoia expression at 6 hr APF with about 18 R8 growth cones in the medulla target region , we observed a defined shift of the youngest 10 of these R8 growth cones to the deeper R7 position whereas the more posterior , and therefore older , 8 R8 growth cones remain at their superficial medulla position . This indicates that posterior R8 growth cones had consolidated their position prior to the effects of induced Sequoia expression ( Figure 4B ) . Two hours later ( at 8 hr APF ) , four more R8 growth cones ( 12 out of 20 ) are no longer responsive to elevated Sequoia levels as they do not leave their superficial medulla positions ( Figure 4C ) . Upon the onset of Sequoia expression at 12 hr APF , 16 out of 24 R8 growth cones are retained in the superficial medulla position ( Figure 4D ) . Additionally , continuous Sequoia expression at later developmental stages does not disrupt the superficial layer positioning of R8 growth cones that are consolidated prior to Sequoia induction ( Figure 4E ) . Together , the consolidation of each R8 growth cone occurs approximately 18 hr after its arrival at the superficial medulla layer ( Figure 4F ) . The fixed correlation between the number of consolidated R8 growth cones and developmental time indicates that for each R8 growth cone there is a defined transition from the initial positioning to the consolidation shortly after R7-R8 segregation . 10 . 7554/eLife . 13715 . 012Figure 4 . Initial position of R cell axons in the medulla is developmentally consolidated . ( A ) Early mis-expression of Sequoia from 3rd instar stage leads to shift of all R8 growth cones to deeper medulla position and results in convergence with R7 growth cones . ( B ) Induction of Sequoia mis-expression from 6 hr APF shows consolidation of 8 posterior R8 growth cones in their superficial medulla position at 24 hr APF . ( C ) Sequoia mis-expression from 8 hr APF shows 12 posterior R8 growth cones to be consolidated . ( D ) Sequoia mis-expression from 12 hr APF onwards shows consolidation of 16 posterior R8 growth cones in superficial medulla position at 24 hr APF . ( E ) Continued Sequoia mis-expression until 40 hr APF following induction at 12 hr APF does not disrupt the consolidation of 16 posterior R8 growth cones . ( F ) Quantification of R8 growth cone consolidation . Error bars indicate Standard Deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01210 . 7554/eLife . 13715 . 013Figure 4—source data 1 . R8 axon consolidation quantification . The file depicts quantification of number of R8 axons consolidated in the superficial medulla position following induction of Sequoia expression at different developmental stages . Rows depict different developmental stages at which Sequoia expression was induced . Total number of R8 axons was counted at the stage of Sequoia expression induction and number of R8 axons consolidated in the superficial medulla position was counted at 24 hr APF . For each stage , 20 brains were analysed and the average number is shown in the source file . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01310 . 7554/eLife . 13715 . 014Figure 4—figure supplement 1 . Temporally restricted induction of Sequoia expression has differential effects on R7/R8 growth cone segregation . ( A–D ) The induction of Sequoia expression in temporally restricted manner during R cell growth cone positioning . Schematics on the left illustrate the temporal windows of Sequoia expression and schematics on the right illustrate the resulting growth cone phenotypes . ( A ) Wild type targeting of R8 and R7 growth cones at 24 hr APF in the medulla . ( B ) Continued mis-expression of Sequoia causes R8 and R7 growth cones to remain converged forming a single layer in deeper medulla position at 40 hr APF . ( C ) Early mis-expression of Sequoia until 24 hr APF is sufficient for continued convergence of R8 and R7 growth cones at 40 hr APF . ( D ) Late mis-expression of Sequoia after 24 hr APF cannot induce convergence of R8 and R7 growth cones in deeper medulla position at 40 hr APF but results in loss of topographic arrangement ( D arrowheads ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 014 To gain further insights into the molecular mechanism underlying the consolidation of R8 cell growth cones we tested candidate molecules expressed during initial axon targeting . Capricious ( Caps ) is expressed in projecting R8 cells and has been proposed to mediate R8 axon targeting ( Shinza-Kameda et al . , 2006 ) . We determined the relative levels of Capricious expression in the developing medulla at the R8 and R7 positions by measuring the ratio of normalized fluorescence intensities ( NFI ) of Capricious staining at the R8 and R7 growth cones against the surrounding medulla region ( Figure 5 Table ) . At the anterior medulla , Capricious shows a homogeneous expression with NFI ratios for R cell growth cones and corresponding medulla region being close to 1 ( 0 . 94 in R8 and 0 . 87 in R7 growth cone position , Figure 5A–D ) . Following the phase of growth cone segregation , Capricious enriches during the process of consolidation at the region of R8 growth cones with an NFI ratio increasing to 1 . 39 and a four fold decline at the R7 position ( with the ratio of 0 . 24 , Figure 5A’–D’ ) . This indicates that during the process of consolidation , Capricious levels in R8 growth cones increase . At the same time , Capricious-negative R7 growth cones move to a deeper medulla position devoid of any Capricious expression . 10 . 7554/eLife . 13715 . 015Figure 5 . Capricious mediates initial position consolidation of R8 axons . ( A–D’ ) Expression pattern of Capricious protein in the developing medulla at 24 hr APF . A , B , C and D show Capricious expression at the anterior medulla corresponding to the initial positioning of R8/R7 growth cones as they innervate the medulla and A’ , B’ , C’ and D’ show the Capricious expression at the posterior side of medulla corresponding to the region where R8 growth cones are consolidated in their superficial medulla position . ( D–D’ ) The heat map of Capricious expression ( measured in terms of normalized relative fluorescence intensity ) in the developing medulla region . The Arbitrary Fluorescence Units used for plotting the heat map are shown in D . The table shows the quantification of Capricious expression intensity measured as ratio between the Normalized Fluorescence Intensities ( NFI ) at R8 or R7 growth cone position vs . the surrounding medulla region . ( E–H” ) Loss of Capricious function disrupts R8 growth cone consolidation . ( E , E” ) Wild type R8 clones at 24 hr APF . Wild type R8 growth cones are positioned at the superficial medulla position at both anterior ( E’ ) and posterior ( E” ) side of medulla ( 98% , n=58 ) . ( F , F” ) Wild type R8 axons at 75 hr APF target to medulla layer M3 ( F’ , F” , 97% , n=49 ) . ( G–H” ) capricious mutant R8 clones . ( G , G” ) capricious mutant R8 clones at 24 hr APF . capricious mutant R8 growth cones are positioned correctly at the anterior side of medulla ( G’ , 94% , n=37 ) as they innervate medulla whereas at the posterior side ( G” ) capricious mutant R8 growth cones prematurely extend towards the deeper medulla position ( G” arrowheads , 62% , n=24 ) . ( H–H” ) capricious mutant R8 axons at 75 hr APF mis-target to medulla layer M6 instead of M3 ( H” arrowheads , 71% , n=38 ) . ( I–L ) Reduction of Capricious levels leads to the disruption of R8 growth cone consolidation . I . UAS-CapriciousRNAi expression under LGMR-Gal4 does not affect R8 growth cone consolidation at 24 hr APF . ( J ) Sequoia mis-expression from 12 hr APF leads to consolidation of 16 posterior R8 growth cones at 24 hr APF . ( K ) Sequoia mis-expression from 12 hr APF in sensitized background of UAS-CapriciousRNAi severely disrupts R8 growth cone consolidation at 24 hr APF . ( I , J , K ) Dotted lines depict region in the medulla with R8 growth cone consolidation and solid lines depict region in the medulla with R8 growth cones extending to the deeper R7 position . ( L ) Quantification of the UAS-CapriciousRNAi mediated disruption of R8 growth cone consolidation . Error bars indicate Standard Deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01510 . 7554/eLife . 13715 . 016Figure 5—source data 1 . UAS-Seq/ UAS-Seq; UAS-Caps RNAi consolidation quantification . The file depicts the quantification of R8 axon consolidation following Sequoia expression induction at 12 hr APF with and without CapriciousRNAi in the background . For Sequoia expression induction without Capricious RNAi , R8 axon consolidation was quantified for 13 brains whereas 19 brains were analysed for Sequoia expression induction in Capricious RNAi background . Average number of axons and Standard Deviation were accordingly calculated for both conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01610 . 7554/eLife . 13715 . 017Figure 5—figure supplement 1 . Antagonistic interaction between Sequoia and Capricious mediates proper positioning of R-cell growth cones in the developing medulla . ( A–A’ ) R-cell growth cone positioning is disrupted by Sequoia mis-expression . A . Sequoia mis-expression under LGMR-Gal4 leads to mis-positioning of R8 growth cones in the deeper medulla position at 24 hr APF ( Arrowhead shows mis-positioned R8 growth cones in deeper medulla position , arrow shows empty superficial medulla position ) . Co-expression of Sequoia and Capricious under LGMR-Gal4 rescues Sequoia mediated mis-positioning of R8 growth cones , thereby R8 growth cones are positioned at the superficial medulla position ( arrow ) whereas R7 growth cones are positioned in their normal deeper medulla position ( arrowhead ) at 24 hr APF . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 017 To determine the role of Capricious in the process of growth cone consolidation we followed the projection of capricious mutant R8 cells using MARCM . Interestingly , no defect could be detected in the initial position of R8 growth cones as well as subsequent R7/R8 segregation ( Figure 5E–E” , G , G’ ) . R8 growth cone positioning defects first appear during the phase of consolidation where capricious mutant growth cones extend towards the deeper R7 position ( Figure 5G” ) . Later in development , these capricious mutant R8 axons mis-target to layer M6 as previously reported ( Figure 5F–H" , Shinza-Kameda et al . , 2006 ) . This shows that Capricious is not required for the initial positioning of R8 growth cones and R7/R8 segregation but is critical for the subsequent step of R8 growth cone consolidation in the superficial medulla . To further elucidate the role of Capricious in the consolidation of R8 growth cones we modified Capricious levels in R8 cells with prolonged Sequoia expression . The partial reduction of Capricious levels via targeted RNAi does not interfere with the consolidation of R8 growth cones in the superficial medulla position ( Figure 5I ) . As described above , approximately 16 R8 growth cones in the posterior medulla region are consolidated in their initial position at 12 hr APF , making them insensitive to the induction of Sequoia expression ( Figure 5J ) . In contrast , the RNAi-mediated Capricious reduction in the background of elevated Sequoia expression severely affected the R8 growth cone consolidation resulting in most of the posterior R8 growth cones shifting to the deeper R7 position ( Figure 5K , L ) . Furthermore , co-expression of Sequoia and Capricious prevents Sequoia-induced shift of R8 growth cones and leads to segregation of R7 and R8 growth cones in a wild type pattern ( Figure 5—figure supplement 1A–B’ ) . Therefore , Sequoia mediated growth cone segregation and Capricious mediated adhesion are antagonistic forces that are capable of balancing each other to achieve proper positioning of R-cell growth cones in the developing medulla . Following the sequential positioning of R7/R8 growth cones during the first half of pupal development , all R7 and R8 axons simultaneously extend during the second half of medulla circuit assembly to the M6 and M3 layer respectively ( Ting and Lee , 2007; Özel et al . , 2015 ) . To determine how the initial growth cone positioning influences subsequent steps of synaptic layer selection we followed the development of R8 growth cones which have been displaced to deeper R7 position . Wild type R8 growth cones are positioned at the superficial medulla at 24 hr APF ( Figure 6A , A’ ) and later target to layer M3 in the adult ( Figure 6B , B’ ) . A pulse of Sequoia expression until 24 hr APF and subsequent repression using Gal80ts leads to an initial shift of all R8 growth cones to the deeper R7 position ( Figure 6C , C’ ) . Interestingly , in the subsequent steps of synaptic layer targeting , these R8 axons , together with R7 axons , terminate in the layer M6 even without further Sequoia expression ( Figure 6D , D’ ) . Additionally , an early Sequoia pulse until 6 hr APF leads to a shift of 8–10 posterior R8 growth cones towards the deeper R7 medulla position ( Figure 6E , E’ ) . Following this pattern during the subsequent steps of medulla development reveals that R8 axons recognize their final target layer exactly according to their initial growth cone position ( Figure 6F , F’ ) , with 10 most posterior R8 axons targeting to the layer M6 ( Figure 6F’ arrowhead ) and the remaining R8 axons terminating in the layer M3 ( Figure 6F’ arrow ) . We analysed if this R8 mis-targeting to layer M6 results from Sequoia-induced changes in the expression of Capricious and the known R8 guidance receptor Frazzled ( Pecot et al . , 2014; Shinza-Kameda et al . , 2006; Timofeev et al . , 2012 ) . However , no difference in Capricious and Frazzled expression can be detected in R8 cells with elevated Sequoia levels ( Figure 6G-J' ) , which is in line with earlier data showing that Sequoia does not influence cell-type specific differentiation programs of R7 and R8 ( Petrovic and Hummel , 2008 ) . 10 . 7554/eLife . 13715 . 018Figure 6 . Initial position determines final medulla layer targeting and synaptogenesis . ( A–F’ ) Initial growth cone position correlates with final target layer . ( A ) Wild type R7/R8 growth cone position in the medulla at 24 hr APF . ( B ) Wild type innervation of R8 axons in the adult to layer M3 and R7 axons to layer M6 . ( C ) Induced Sequoia expression mediates mis-positioning of R8 growth cones in the deeper medulla position at 24 hr APF . ( D ) R8 axons mistarget to layer M6 in the adult even when Sequoia mis-expression is stopped from 24 hr APF onwards . ( E ) Sequoia mis-expression until 6 hr APF leads to mis-positioning of 6–8 posterior R8 growth cones to deeper medulla position . ( F ) Initially mis-positioned R8 growth cones mis-target to layer M6 whereas normally positioned R8 growth cones later target to layer M3 . A’-F’ shows magnifications of A-F . ( G-J’ ) Changes in Sequoia expression do not affect expression of known R8 targeting molecules . ( G , G’ ) Homozygous sequoia mutant cells are visualized using loss of Ubi-GFPnls expression as a clonal marker . Arrowheads show individual R8 cells labelled with Senseless and Frazzled and cells without GFP are sequoia mutant R8 cells . ( H , H’ ) Visualization of Frazzled expression following Sequoia mis-expression using LGMR-Gal4 . All cells mis-express Sequoia and individual R8 cells are labelled with Senseless ( Blue ) . Frazzled expression is visualized using anti-Frazzled antibody ( Red ) . All R8 cells express Frazzled suggesting mis-expression of Sequoia does not repress Frazzled expression . Additional Frazzled staining at the ommatidial boundaries is from a different imaging plane . ( I–J’ ) Elevated Sequoia levels do not affect Capricious expression . I , I’- Wild type pattern of Capricious expression reported by caps-LacZnlsenhancer trap ( Shishido et al . , 1998 ) and visualized using anti-LacZ antibody ( Green ) in the R8 cells labelled with Senseless ( Blue ) . ( J , J’ ) Sequoia mis-expression does not transcriptionally repress the expression of Capricious ( J , J’- Arrowheads show R8 nuclei with Sequoia mis-expression , arrows show R8 nuclei without Sequoia mis-expression ) . ( K–M” ) Synaptogenesis of R8 axons in the ectopic layer M6 with Dm8 neurites as shown by GFP Reconstitution Across Synaptic Partners ( syb-GRASP ) . ( K–K” ) Control GRASP between R7 and Dm8 at layer M6 . R8 and R7 axons target to layers M3 and M6 ( K’ arrows ) and GRASP signal is observed in layer M6 ( K” arrows ) but not in layer M3 ( K” arrowheads ) . ( L–L” ) Mistargeting of R7 axons to layer M3 upon UAS-CapriciousID expression under sev-Gal4 leads to loss of GRASP between R7 and Dm8 . Escaper R7 axons that target to M6 show GRASP with Dm8 ( L’ arrows ) whereas R7 axons that mistarget to M3 do not show GRASP signal ( L” arrowheads ) . ( M–M” ) R8 axons ectopically targeting to layer M6 upon Sequoia mis-expression show GRASP signal with Dm8 ( M” arrows ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01810 . 7554/eLife . 13715 . 019Figure 6—figure supplement 1 . Ectopic R8 growth cones are present in close proximity of Dm8 neurites early in the development . ( A–B’ ) . Initially mis-positioned R8 growth cones can contact Dm8 neurites at 24 hr APF . ( A , A’ ) At 24 hr APF , Dm8 neurites are present at the R7 initial position in the medulla . This R7 and Dm8 neurite position is proximal to the Capricious expression domain in the medulla . Dotted lines show R8 and R7 axon positions . ( B , B’ ) Sequoia mis-expression causes R8 axons to mis-position along with R7 axons in the deeper position in the medulla just proximal to the Capricious expression domain . This is the position where Dm8 neurites are present and are therefore in close proximity to the mis-positioned R8 growth cones . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 01910 . 7554/eLife . 13715 . 020Figure 6—figure supplement 2 . Role of N-Cadherin in stabilizing R7 growth cones in deeper medulla position and subsequent M6 layer targeting . ( A–B’ ) Loss of N-Cadherin leads to mis-targeting of R7 axons to M3 layer . ( A , A’ ) CadN mutant R7 growth cones shift to superficial R8 position in the medulla ( A’ , arrowhead ) leaving gaps in their normal medulla position ( A’ asterisk ) . The CadN mutant R7 axons are generated using GMR-FLP and labelled with elav-Gal4 UAS-mCD8GFP . ( B , B’ ) CadN mutant R7 axons that are initially mis-positioned subsequently mistarget to layer M3 along with R8 axons ( B’ , arrowhead ) leaving a gap in the M6 layer ( B’ , asterisk ) . ( C–D’ ) In the absence of N-Cadherin , initial positioning to the temporary deeper medulla layer is not sufficient for M6 layer targeting . ( C , C’ ) Induction of Sequoia expression under weak elav-Gal4 in cadN mutant R7 cells can rescue the initial mis-positioning of their growth cones . The CadN mutant R7 growth cones with induced Sequoia expression reach their normal temporary target layer in the deeper medulla position ( C’ , arrowhead ) . ( D , D’ ) Early induction of Sequoia expression is not sufficient to rescue the later R7 axon mis-targeting to layer M3 . The CadN mutant R7 axons mis-target to layer M3 in the adult ( D’ , arrowhead ) leaving gaps in layer M6 ( D’ , asterisk ) even though these R7 growth cones are initially positioned in normal medulla layer ( C , C’ ) . ( E-F’ ) N-Cadherin is required for mistargeting of R8 axons to M6 layer . ( E , E’ ) Early Sequoia mis-expression in large clones of R8 cells under sens-Gal4 is sufficient for their mis-targeting to M6 layer in the adults ( E’ arrowheads ) leaving layer M3 empty ( E’ asterisk ) . ( F , F’ ) CadN mutant R8 axons cannot mis-target to layer M6 even after early mis-expression of Sequoia under sens-gal4 . Thus , due to absence of N-Cadherin both R7 and R8 axons fail to reach layer M6 leaving it empty ( F’ asterisk ) and are present in layer M3 ( F’ arrowheads ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 020 Interestingly , the main R7 synaptic target cell , Dm8 , is already in close proximity of R7 growth cones immediately following the R7/R8 segregation and positioning of R7 growth cones in the deeper medulla ( Figure 6—figure supplement 1A–A' ) , Ting et al . , 2014 . Therefore the Dm8 neurites and mis-positioned R8 growth cones are also in close proximity raising the possibility that these two cell types can directly interact with each other ( Figure 6—figure supplement 1B–B' ) . To test if the ectopic R8 axons form synapses with Dm8 neurites in layer M6 , we made use of the recently modified GFP reconstitution method ( syb-GRASP; Karuppudurai et al . , 2014 ) . In this method , the GFP1-10 fragment is fused to the C-terminus of Drosophila n-synaptobrevin ( n-syb ) producing n-syb::spGFP1-10 chimera , resulting in reconstitution with GFP11 only after vesicle fusion ( Macpherson et al . , 2015 ) and a preferential labelling of ‘active-synapses’ rather than neuronal contacts made at any time during development . In wild type ( Figure 6K–K” ) , the expression of pre-synaptic syb::spGFP1-10 in R7/R8 ( LGMR-Gal4 ) and post-synaptic LexAop-spGFP11::CD4 , in Dm8 cells ( OrtC1-3 LexA DBD , OrtC2B dVP16AD; Karuppudurai et al . , 2014; See Materials and methods for details ) , results in GRASP-positive R7->Dm8 connections in layer M6 ( Figure 6K” arrows ) and no GRASP-positive R8->Dm8 connections can be detected in layer M3 ( Figure 6K” arrowheads ) . When R7 growth cones are retained at the R8 temporary layer via UAS-CapriciousID expression using sev-Gal4 ( Shinza-Kameda et al . , 2006 ) , GRASP signal , demonstrating R7->Dm8 contacts , can be observed only in columns where 'escaper' R7 axons target to layer M6 ( Figure 6L” arrows ) , but not in columns with mis-targeted R7 axons ( Figure 6L” arrowhead ) . Surprisingly , GRASP-positive R8->Dm8 , similar to R7->Dm8 , connections can be detected in medulla layer M6 following the early shift of R8 growth cones to the deeper medulla position ( Figure 6M–M” ) , indicating that both R7 and R8 can recognize Dm8 processes and form synaptic contacts .
Here we demonstrate that the early , birth order dependent , segregation of R cell growth cones determines later synaptic layer identity in the Drosophila visual system . Small inter-ommatidial differences in Sequoia levels organize R cell growth cones within a layer whereas large intra-ommatidial differences segregate growth cones between layers . Changes in the positioning of growth cones directly correlate with changes in synaptic layer selection without affecting the expression of known cell-type specific targeting molecules . These results highlight the importance of initial afferent growth cone positioning for visual map formation prior to synaptic partner recognition ( Figure 7 ) . 10 . 7554/eLife . 13715 . 021Figure 7 . Role of early growth cone patterning in synaptic layer selection . ( A ) R cell axons arrive asynchronously in the developing medulla field in a pattern that reflects their specification in the eye disc . The growth cones of the same R cell type topographically segregate within the temporary layer . This segregation within a layer is mediated by low inter-ommatidial differences in the Sequoia levels ( illustrated for R8 cells ) . The segregation of growth cones of different cell type ( R7/R8 ) between distinct temporary layers occurs as a result high intra-ommatidial differences in Sequoia levels . The differences in Sequoia levels are an outcome of temporal sequence of photoreceptor specification in the eye disc ( Tomlinson and Ready , 1987 ) . Therefore , birth-order dependent differential positioning generates this early patterning of afferent growth cones in the medulla . ( B ) Once the initial patterning is achieved , the cell adhesion molecules consolidate the growth cones in their distinct temporary layers . In case of R8 growth cones , Capricious mediates afferent-target interactions and thereby stabilizes R8 growth cones in the superficial medulla position . The R7 growth cones , segregated into deeper medulla position most likely interact with neurites of Dm8 cells , the primary post-synaptic target of R7 cells via N-Cadherin ( Özel et al . , 2015 ) . This sequential process continues until all R cell axons arrive in the developing medulla field , are segregated into and consolidated in distinct temporary layers . ( C ) Following the Capricious mediated consolidation of R8 growth cones in the superficial medulla and possible R7->Dm8 interaction in the deeper medulla position , axons of both cell types synchronously extend towards their final target layers . The guidance of R8 axons to M3 layer is mediated by localised NetB signal from L3 neurites ( Timofeev et al . , 2012; Pecot et al . , 2014 ) . On the contrary , the R7 axons are suggested to passively dislocate and reach layer M6 via their interactions with Dm8 neurites that are gradually pushed deeper in the medulla as a result of growth of the medulla field ( Özel et al . , 2015 ) . Furthermore , the extension of R7 axons to reach layer M6 , following their initial positioning , requires N-cadherin function ( Özel et al . , 2015; Ting et al . , 2005 ) . Therefore , M6 targeting of R7 axons seems to be a result of pre- and post-synaptic neuron interactions mediated by general synaptogenetic molecules rather than cell type specific factors . Thus , cellular proximity determines M6 layer targeting and establishment of synaptic contacts . DOI: http://dx . doi . org/10 . 7554/eLife . 13715 . 021 An early shift in the R8 growth cones to R7 position , induced by a short pulse of Sequoia expression , allows them to recognize the R7 target cell Dm8 as synaptic partner later during the development . Similarly , if R7 growth cones fail to segregate from R8 growth cones , they terminate together with R8 axons in the M3 synaptic layer independent of their intrinsic differentiation and targeting program . This extension of Frazzled-negative R7 axons ( Timofeev et al . , 2012 ) towards layer M3 could be explained by the default setting of R7 axons to tightly fasciculate and follow the R8 pioneer axons towards the target region ( Fischbach and Heisinger , 2008; Meinertzhagen and Hanson , 1993 ) . R cell growth cone segregation can be controlled by axon-target interactions or axon-axon interactions or both . Although there is no experimental demonstration for direct R7-R8 afferent interactions , the following set of data indicate that such interactions occur during development . Support for direct interaction between R7-R8 afferents comes from Maurel-Zaffran et al . , 2001 , where the expression of LAR as membrane tethered ligand in R8 cells alone ( using an R8-specific Gal4 driver line ) could induce a response from R7 axons , indicating a direct signalling between R8-R7 axons . Further , induced Capricious expression in R8 and R7 cells , in Capricious null background ( therefore resulting in a target region without any Capricious expression ) is sufficient to mis-target R7 axons to layer M3 , again indicating a direct R7-R8 afferent interaction ( Berger-Müller et al . , 2013 ) . Here we show that final target layer of sequoia mutant R7 axons depends on the targeting of R8 axons , further suggesting that mis-targeting of sequoia mutant R7 axons to ectopic synaptic layer is the consequence of segregation defect rather than a change in target layer recognition . Interactions among afferent axons have been implicated in the assembly of visual and olfactory circuits in vertebrates as well as invertebrates ( Brown et al . , 2000; Clandinin and Zipursky , 2002; Ebrahimi and Chess , 2000; Feinstein and Mombaerts , 2004; Komiyama et al . , 2007; Petrovic and Schmucker , 2015; Sweeney et al . , 2007 ) . It has recently been shown that Eph-Ephrin signalling mediates local sorting of RGC axons in mammalian visual system ( Brown et al . , 2000; Suetterlin and Drescher , 2014 ) . Notch signalling was demonstrated to play a role in spacing of DCN cluster neuron axons via neighbour axon interactions ( Langen et al . , 2013 ) . But , whether these afferent interactions influence synaptic partner recognition is not known . Here we show that in the Drosophila visual system , relative levels of Sequoia determine the segregation of afferent R7/R8 growth cones within or between layers . By creating Seqhigh-Seqlow R cell combinations using Sequoia gain-of-function R7 mosaics we observed that difference in Sequoia levels among neighbouring cells could induce growth cone segregation . The endogenous differences in Sequoia levels most likely arise as a result of the temporal sequence of R cell specification , suggesting a self-patterning mechanism in early visual circuit assembly ( Hassan and Hiesinger , 2015; Roignant and Treisman , 2009 , Tomlinson and Ready , 1987 ) . How the relative differences in Sequoia levels in the nuclei of R cells translate into growth cone segregation remains elusive . We have tested candidate signalling pathways including Semaphorin/Plexin ( Cafferty et al . , 2006; Hsieh et al . , 2014; Pecot et al . , 2013; Yu et al . , 2010 ) , TGF-beta ligand Activin and its receptor Baboon ( Ting et al . , 2007 ) and Notch ( Langen et al , 2013 ) but did not find evidence for a critical role in initial growth cone segregation ( data not shown ) . This suggests a so far unknown molecular mechanism in which the growth state of an axon is directly coupled to differential growth cone adhesion . As we could demonstrate a cell-autonomous function of Sequoia in R8 for columnar segregation as well as in R7 for layer segregation , we envision a mechanistic model related to the concept of cell competition , in which strong cell-cell interactions induce cell-autonomous responses ( Rhiner et al . , 2010 ) . The initial segregation of afferent growth cones into distinct positions is then consolidated by expression of Capricious in R8 axons in the same posterior-to-anterior pattern in which they arrive in the medulla . We speculate that Capricious mediated growth cone consolidation serves two purposes: 1 . It removes the temporal difference in the arrival of R8 axons and 2 . It maintains R8 axons in the position where they are responsive to subsequent NetrinB signal provided by L3 neurites ( Pecot et al . , 2014; Timofeev et al . , 2012 ) . This is supported by two different sets of results: First , as we show in this study , the displacement of R8 growth cones to deeper medulla position leads to their mis-targeting to layer M6 in spite of normal Frazzled expression in these R8 cells . Second , the ectopic expression of Frazzled in R7 cells cannot re-direct them to M3 layer in response to localized NetrinB signal present in the superficial position but R8 axons can be re-directed to a different layer ( M1/M2 ) by ectopic expression of localized NetrinB in a position deeper to the superficial R8 medulla position ( Nern et al . , 2008; Pecot et al . , 2013; Timofeev et al . , 2012 ) . Taken together , these observations suggest that M3 layer targeting via L3-mediated NetrinB signalling requires R8 axons to be positioned superficially in the medulla further underscoring the importance of R8 growth cone consolidation in this position ( Figure 7 ) . Previous studies have identified several molecules necessary for M6 targeting of R7 axons ( Clandinin et al . , 2001; Hofmeyer and Treisman , 2009; Hofmeyer et al . , 2006; Maurel-Zaffran et al . , 2001; Ting et al . , 2005; Tong et al . , 2011 ) , including Liprin-alpha/beta/gamma , PTP69D and D-Lar . The loss of these molecules specifically affects the stabilization of R7 growth cones during the second step of targeting . Additionally , these molecules along with N-Cadherin have been shown to be critical for establishment of synaptic contacts between pre and post-synaptic neurons ( Arikkath and Reichardt , 2008; Astigarraga et al . , 2010b; Garrity et al . , 1999; Hummel and Zipursky , 2004; Lee and Godenschwege , 2015; Nagaoka et al . , 2014; Prakash et al . , 2010; Reines et al . , 2012 ) . We confirm previous observations , that R7 growth cones are in close proximity with their primary post-synaptic target neurons , Dm8 , at the end of growth cone segregation ( Figure 6—figure supplement 1A–B' , Ting et al . , 2014 ) . This raises the possibility that targeting of R7 axons to M6 layer , later in the development , could be the direct result of R7->Dm8 contacts mediated by N-Cadherin . Recently it was shown that N-Cadherin function is necessary for stabilizing R7 growth cones in the deeper medulla position but not for targeting and subsequent extension of R7 axons to M6 layer seems to be a result of passive dislocation ( Özel et al . , 2015 ) . Additional support for the role of N-Cadherin in the formation and maintenance of R7->Dm8 contacts , following their initial segregation from R8 growth cones , comes from our observation that early expression of Sequoia in CadN mutant R7 cells under weak elav-Gal4 driver can rescue the mis-targeting of R7 growth cones in the superficial medulla position along with R8 growth cones at 24 hr APF ( Figure 6—figure supplement 2A , A’ , C , C’ ) , but fails to rescue the later mis-targeting to layer M3 eventually resulting in a mis-targeting phenotype identical to CadN mutant R7 axons ( Figure 6—figure supplement 2B , B’ , D , D’ ) . Interestingly , the R8 growth cones initially mis-positioned in the deeper medulla eventually mis-target to layer M6 and form synaptic contacts with Dm8 . In addition , these R8 cells , with axons mis-targeted to layer M6 , do not show changes in any of their known cell-type specific molecules including early specifier of cell identity ( Senseless ) , guidance receptors ( Frazzled , Capricious ) and sensory receptors ( Rh6 ) . In addition , no expression of R7 specific molecules ( Prospero , R3 , Rh4 ) can be detected . Thus , the R8 cells interact with Dm8 neurons most likely via ubiquitously expressed molecules such as N-Cadherin expressed in both , R7 as well as R8 , cells . This is supported by our observation that N-Cadherin is required for stabilization of R8 axons at the layer M6 ( Figure 6—figure supplement 2E–F’ ) . We observed that R8 axons form functional synapses with Dm8 , a known R7 target neuron , in the layer M6 . This raises the fundamental question of how synaptic layer selection influences synaptic partner recognition . The cellular complexity of potential post-synaptic target layer encountered by ingrowing R cell axons has not been fully determined , leaving room for selective recognition for synaptogenesis within a layer . In fact , it has been shown that within M6 , R7 axons form synapses with Dm8 but not with Tm5c ( Karuppudurai et al . , 2014 ) which also arborize the M6 layer . In addition , we have identified various medulla columnar neurons within M6 that are not contacted by R7 axons ( unpublished results ) . Similarly in layer M3 R8 and L3 select distinct post-synaptic partners ( Takemura et al . , 2013 ) . The types of neurons present in the medulla at the time of R8 and R7 axon innervation have not been fully identified . Based on the data from Hasegawa et al . , 2011; Li et al . , 2013; Suzuki et al . , 2013 , the medulla neurons are generated in temporal fashion ( reviewed in Sato et al . , 2013; Suzuki and Sato , 2014 ) and therefore they likely innervate the medulla at different time points . Experiments presented here support a developmental scenario in which the medulla context for arriving R cell axons reduces the complexity of synaptic partner selection . For example R8 and L3 have different arrival times at M3 ( Nern et al . , 2005; Pecot et al . , 2013; 2014 ) , thereby would encounter a different local environment of potential post-synaptic partners competent for synaptogenesis . It is plausible that some form of temporal co-ordination of afferent axons and their post-synaptic partner cell neurites would actually simplify the synaptic partner matching . The concept of temporal identity would argue that R7 and R8 axons arriving at the same medulla position approximately the same time , as shown in the Sequoia gain-of-function background , will pick the same synaptic partners exemplified by Dm8 . Support for such proximity-based axon-target interaction for synaptogenesis comes from earlier analysis of ectopic axons in Drosophila as well as Zebrafish ( Edwards and Meinertzhagen , 2009; Berger-Müller et al . , 2013; Pujol-Martí et al . , 2014 ) . From an evolutionary perspective , such proximity-induced synapse formation has several advantages over mechanisms that require regulation and expression of distinct sets of cell recognition molecules . Considering R7 as the most recently added cell to the precursor ommatidium ( Mavromatakis and Tomlinson , 2012 ) : During development , R7 is recruited using mechanisms similar to R8 and therefore possesses default R8 specification program . However , this default R8 program is suppressed to facilitate R7 specification ( Cook et al . , 2003; Morey et al . , 2008 ) . Thus , a temporally separated , novel R7 cell is generated with basic neuronal differentiation similar to that of an R8 cell ( Brennan and Moses , 2000; Friedrich et al . , 2011; Roignant and Treisman , 2009 ) . Interestingly the temporal difference in the R8/R7 differentiation is then translated into Sequoia mediated layer segregation of their growth cones ( Petrovic and Hummel , 2008 ) , with Sequoia expression being part of common differentiation program . Thus , the evolutionary recent R7 cell seems to recognize its synaptic targets via pan neuronal molecules like N-Cadherin as part of the default neuronal differentiation program , instead of the invention of an additional recognition code .
The flies were raised at 25°C unless otherwise mentioned . The following flies were used in this study: CantonS , FRT42B , FRT2A , Gal80/II , Gal80/III , FRT80 , GMR-GFP , UAS-mCD8GFP , hs-FLP , ey3 . 5-FLP , GMR-FLP , tub-Gal80ts , elav-Gal4 , sev-Gal4 , LGMR-Gal4 , m∂0 . 5-Gal4 , ro-tau-LacZ , Rh6-EGFP , PanR7-Gal4 were obtained from Bloomington Drosophila Stock Center . sens-Gal4/CyO and sens-Gal4/TM6 were gift from Bassem Hassan . sequoia5 is a previously generated Sequoia loss-of-function allele ( Petrovic and Hummel , 2008 ) . UAS-Sequoia was obtained from Jay Brenman . UAS-Capricious , UAS-CapriciousID ( intracellular deletion ) , CapriciousC18fs FRT2A and caps-LacZnls flies were kindly provided by Akinao Nose . GMR-gogo was a generous gift from Takashi Suzuki . PanR8-Gal4 was a gift from Claude Desplan . UAS-CapriciousRNAi was obtained from VDRC ( VDRC Transformant ID No . 27097 ) . PM181-Gal4 and CadN405FRT40 ( Lee et al . , 2001 ) were used in Sequoia analysis , Early Dm8 labelling OK371-VP16AD/CyO; ortC2-Gal4DBD/TM2 , adult Dm8 specific OrtC1-3 LexA DBD , OrtC2B dVP16AD/CyO ( Ting et al . , 2014 ) and syn-GRASP constructs UAS-Syb::spGFP1-10 and LexAop spGFP11::CD4/TM2 ( Karuppudurai et al . , 2014 ) were used in syb-GRASP experiments . The primary antibodies used were: Rabbit anti-GFP ( 1:1000 , Invitrogen , Carlsbad , California , USA ) , Mouse anti-GFP ( 1:100 , Invitrogen ) , Chicken anti-GFP ( 1:1000 Abcam ) , Mouse 24B10 anti-Chaoptin ( 1:50 , DSHB ) , Rat anti-CadN ( 1:20 , DSHB ) , Rabbit anti-Sequoia ( 1:1000 , Brenman et al . , 2001 ) , Mouse Anti-Elav ( 1:20 DSHB ) , Rabbit anti-LacZ ( 1:200 , Invitrogen ) , Guinea pig anti-Senseless ( 1:1000 ) was kindly provided by Hugo Bellen , Guinea pig anti-Repo ( 1:200 ) and Rabbit anti-Frazzled ( 1:10 ) were kindly provided by Benjamin Altenhein . Rabbit anti-Caps ( 1:50 ) antibody was generated for this study ( See below ) . The Secondary antibodies used were: Goat Anti-Rabbit Alexa-488 ( 1:500 ) , Goat anti-Rabbit Alexa-568 ( 1:300 ) , Goat anti-Mouse Alexa-488 ( 1:300 ) , Goat anti-Mouse Alexa-560 ( 1:500 ) , Goat anti-Mouse Alexa-647 ( 1:500 ) , Goat anti-Rat Alexa-647 ( 1:300 ) , Goat anti-Guinea pig Alexa-568 ( 1:500 ) , Goat anti-Guinea pig Alexa-647 ( 1:300 ) . All secondary antibodies were obtained from Invitrogen ( Carlsbad , CA ) . Pupal and adult brains were dissected in PBS and fixed with 3 . 7% formaldehyde in PBS for 20 min . Fixed brains were blocked with 10% Goat Serum for one hour and then incubated with primary antibody in 10% Goat Serum ( in 0 . 3% PBS-T ) over night at 4°C . Following three times washing ( 15 min each ) , brains were incubated with secondary antibody diluted in 0 . 3% PBS-T overnight at 4°C . After three times washing ( 20 min each ) brains were mounted in Vectashield ( Vector Laboratory , Burlingame , CA ) anti-fade mounting medium for confocal microscopy . Images were obtained using a Leica TCS SP5II confocal microscope and processed with ImageJ and Adobe Photoshop CS5 . 1 . MARCM clones were generated as previously described ( Lee and Luo , 1999 ) . Briefly , ey3 . 5- and GMR-FLP were used to generate eye and R7 specific clones , respectively . To generate R8 specific clones FLP was expressed under heat shock responsive promoter ( hs-FLP ) . 2nd instar larvae were collected and subjected to heat shock at 37°C for 25 min ( for large clones ) or 5 min ( single cell clones ) . Post heat shock , the larvae were incubated at 25°C , following a brief incubation at 18°C for half an hour , and brains were dissected at appropriate pupal stages . To generate an antibody specific against Capricious , we used the C-terminal peptide sequence AAGGYPYIAGNSRMIPVTEL as the antigen ( Shishido et al . , 1998 ) . Using the standard services from GenScript ( www . genscript . com ) , we generated a rabbit polyclonal antibody against the specified peptide . The specificity of the antibody was tested using immunostaining in the wing disc . The antibody was then used in the brain tissue after standardization for concentration . The normalized fluorescence intensity ( NFI ) was measured as described in Komiyama et al . ( 2007 ) using ImageJ . Briefly , the images were de-convoluted and fluorescence intensity of region of interest was normalized against the background of the image . The fluorescence intensity for each R8 cell in the same row was measured and normalized against the background . The ratio of average intensity for all R8 cells in the row and average intensity of same amount of area in the background was taken as the average Sequoia expression in R8 cells in that row . The average intensity of different R8 cell rows from multiple images was measured and plotted . For visualization of synaptic contacts , syb-GRASP method was utilized as previously described ( Karuppudurai et al . , 2014 ) . Briefly , the larvae and pupae were kept on a 12 hr light/dark cycle before eclosion . Following eclosion , the flies were raised in constant light ( 50lx ( 14W cm-2 ) condition for 3–5 days . The brains were dissected in 4% PFA to avoid diffusion of the reconstituted GFP signal . Following 20 min incubation in 4% PFA , the brains were processed and imaged as previously described . | A nervous system requires a precise network of connections between cells called neurons to work properly . Within the brain , the fiber-like connections between pairs of neurons are not running crisscross like a pile of spaghetti . Instead , connected partner neurons are organized into distinct layers and columns . Many questions remain about how these partner neurons find each other and how the layers of fiber-like connections form . To answer these questions , scientists often study the part of the fruit fly nervous system that controls the insect’s vision . This brain-like structure is simple and can be easily manipulated with genetic engineering . Fruit fly studies have helped identify some molecules that play a role in helping partner cells find one another and connect . These studies have also shown that the timing of brain cell development appears to play a role . But the role that layer formation plays in the process is still a mystery . Now , Kulkarni et al . show that the birthdate of neurons in the fruit fly visual system helps organize them into layers . These neurons are generated early in the development of the fly . Shortly after birth , a molecular clock under the control of a protein called Sequoia starts within each newly generated neuron . The Sequoia protein is a transcription factor and controls the activity of many genes , and the molecular clock provides the growing neuron fibers with information about where and when to look for its partner neurons . By manipulating the amount and time that Sequoia is produced in the fly visual system , Kulkarni et al . show that this clock helps arrange the growing cells into layers . Cells with similar birthdates connect and are arranged into layers . How much and when Sequoia is produced dictates where each new layer begins . The next steps for the research will be to learn more about how the clock works and identify any intermediaries between the clock and cell growth patterns . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Birth order dependent growth cone segregation determines synaptic layer identity in the Drosophila visual system |
Acquisition of genes through horizontal gene transfer ( HGT ) allows microbes to rapidly gain new capabilities and adapt to new or changing environments . Identifying widespread HGT regions within multispecies microbiomes can pinpoint the molecular mechanisms that play key roles in microbiome assembly . We sought to identify horizontally transferred genes within a model microbiome , the cheese rind . Comparing 31 newly sequenced and 134 previously sequenced bacterial isolates from cheese rinds , we identified over 200 putative horizontally transferred genomic regions containing 4733 protein coding genes . The largest of these regions are enriched for genes involved in siderophore acquisition , and are widely distributed in cheese rinds in both Europe and the US . These results suggest that HGT is prevalent in cheese rind microbiomes , and that identification of genes that are frequently transferred in a particular environment may provide insight into the selective forces shaping microbial communities .
Great strides have been made in characterizing the composition of microbiomes , and in understanding their importance in the ecology of natural systems , in agriculture and in human health . However , despite these advances , the forces that shape the diversity , structure , and function of microbiomes remain poorly understood ( Widder et al . , 2016 ) . Investigating these underlying mechanisms in situ is difficult , as observational and sequenced-based analysis rarely enables causal conclusions ( Nemergut et al . , 2013 ) . Replicating microbial communities in vitro is also an enormous challenge , because of high levels of diversity and the difficulties in establishing pure cultures of most bacterial species . These obstacles significantly hamper our ability to move from observations of microbial diversity to the molecular mechanisms shaping key processes such as species interactions and microbial evolution . Horizontal gene transfer ( HGT ) is a major force in microbial evolution and can lead to the wholesale acquisition of novel functions . In some cases , these novel functions can have significant adaptive consequences , such as in the transfer of antibiotic resistance genes ( Ochman et al . , 2000 ) . HGT also allows rapid adaptation to new niches ( Wiedenbeck and Cohan , 2011 ) , as ecologically relevant genes may be acquired by species not previously adapted to a particular niche ( Tasse et al . , 2010; Hehemann et al . , 2010 ) . The movement of microbes to new environments has been shown to increase both the rate and impact of HGT , and HGT is most frequent for genes under positive selection ( Niehus et al . , 2015 ) . In moving to a new environment , microbes can face novel abiotic conditions ( temperature , moisture , salinity , pH , and nutrients ) and novel biotic challenges and opportunities resulting from the presence of microbial neighbors . Evaluating HGT within the context of microbial communities could provide new insights concerning the extent , mechanisms , and ecological impact of this important process . Advances in genome sequencing have begun to provide a glimpse into HGT within environmentally , medically and economically important microbiomes ( McDaniel et al . , 2010; Andam et al . , 2015 ) . For example , extensive gene sharing has been observed throughout the commensal human microbiome , including sharing of genes that enable nutrient acquisition from novel food sources ( Hehemann et al . , 2010; Smillie et al . , 2011 ) , and pathogenicity islands and antibiotic resistance genes in pathogenic microbes ( McCarthy et al . , 2014; Hiramatsu et al . , 2001; Forsberg et al . , 2012 ) . There is evidence of extensive HGT in other natural habitats , such as soil ( Coombs and Barkay , 2004; Heuer and Smalla , 2012 ) and aquatic environments ( McDaniel et al . , 2010; Frischer et al . , 1994 ) . Although these studies offer valuable insights into the frequency and potential impact of genes that can be transferred in microbial communities , the complexity of these systems makes difficult any further examination of the effects of these HGT events on their evolution and ecology . The microbial communities of fermented foods experience strong selection as a result of growing in novel , human-made environments . Previous work has demonstrated that HGT can be a major driver of adaptation in food systems and other human-managed environments ( Andam et al . , 2015; Rossi et al . , 2014 ) . Prior analysis of microbial species from cheese has revealed several instances of HGT in this environment . Lactic acid bacteria ( LAB ) such as Lactobacillus and Lactococcus , which are used in the initial fermentation of milk , are known to harbor antibiotic resistance genes and may be reservoirs for transfer to pathogenic enterococci ( Wang et al . , 2006; Mathur and Singh , 2005 ) and other pathogenic microbes . Other food-associated bacteria may also contribute to antibiotic resistance gene transfer ( Cocconcelli et al . , 2003; Delorme , 2008; Flórez et al . , 2005 ) . In yogurt , another dairy ferment using LAB , HGT of metabolic genes has been observed between protocooperative species L . bulgaricus and S . thermophilus ( Li et al . , 2013; Liu et al . , 2009 ) . Sequencing of Penicillium species isolated from the cheesemaking environment identified HGT of large genomic islands between these key fungal inhabitants of cheese ( Cheeseman et al . , 2014; Ropars et al . , 2015; Gibbons and Rinker , 2015 ) . During the aging of traditional styles of cheese in caves or aging rooms , bacteria and fungi form a multi-species biofilm called the rind ( Button and Dutton , 2012 ) . We have previously shown that these communities can be used to examine community-based processes , such as succession and interspecies interactions , within an experimentally tractable system ( Wolfe et al . , 2014; Kastman et al . , 2016 ) . Given that biofilms such as these are densely populated , and microbes in cheese rinds are under strong selection to obtain limited nutrients ( e . g . free amino acids , iron ) as well as tolerate cell stress ( Monnet et al . , 2015 ) , we predicted that HGT might be widespread in cheese rind microbiomes and therefore might provide a useful experimental model for HGT within microbial communities . We sought to determine the diversity , distribution , and functional content of HGT in bacterial species isolated from cheese rinds . Specifically , we predicted that ( 1 ) HGT would be widespread , ( 2 ) HGT genes would be enriched for functions related to survival in cheese environment , and ( 3 ) there would be uneven distribution of HGT events across taxa . We analyzed the genomes of newly isolated and sequenced cheese-associated bacterial species ( 31 genomes ) and those available in public databases ( 134 additional genomes ) . We present data which suggest that there is extensive HGT in cheese-associated bacteria . The regions of DNA identified appear to encode a number of functions which would be expected to provide adaptive advantages within the cheese environment . In particular , we identified three large multi-gene islands that are shared within multiple Actinobacteria , Proteobacteria , and Firmicutes species . These genomic regions are not related , but appear to have analogous functions involving iron acquisition , and are widely distributed in geographically distant cheeses . This work provides foundational knowledge in an experimentally tractable system in which future work could help to provide insight into the role of HGT within microbiomes .
To establish a diverse database of cheese-associated bacterial genomes , we isolated species from cheese samples collected as part of previous work ( Wolfe et al . , 2014 ) . A total of 31 isolates , representing four bacterial phyla and 11 genera , were selected for genome sequencing using Illumina and PacBio ( Supplementary file 1a ) . Recently , a large collection of cheese-associated bacterial genomes were sequenced ( Almeida et al . , 2014 ) , allowing inclusion of additional genomes in our analysis . Our isolates were from cheeses produced in the United States , Spain , Italy , and France , while the Almeida et al . collection was almost exclusively from France . We also included genomes from the NCBI reference sequence ( RefSeq ) database that are associated with cheese , for a total of 165 bacterial genomes . We next developed a computational pipeline for identification of putative horizontally transferred genes , adapted from work on the human microbiome ( Smillie et al . , 2011 ) . We built a central BLAST database containing all ORFs from all cheese-associated genomes . For each gene in each genome , we performed BLAST against this database , and compiled a list of hits ( Figure 1—figure supplement 1 , Materials and methods ) . For each hit , we examined the length and percent identity of aligned regions . Closely related species will have many nearly identical genes as a result of vertical inheritance . To avoid capturing these genes in our analysis , we determined the pairwise average nucleotide identity ( ANI ) between species within the same genus ( Chan et al . , 2012; Rodriguez-R and Konstantinidis , 2016 ) . ANI provides a measure of the overall similarity of two genomes . We tested varying thresholds for length and ANI to examine the effects of these parameters on the results ( Supplementary file 1c ) . Higher maximum ANI cutoffs and shorter lengths are more likely to yield false positives , as closely related species are more likely to share short stretches of high nucleotide identity . At the same time , a lower maximum ANI cutoff may exclude legitimate HGT events , especially considering that closely related species are also more likely to engage in HGT . Based on our most conservative gene identity parameters ( minimum 99% identity over 500 nucleotides ) , we identified at least one putative horizontally transferred gene in 130/165 ( 78 . 8% ) cheese-associated species in the analysis , for a total of 4733 genes ( Figure 1A , Supplementary file 1d ) . Because this analysis depends on the species included for comparison , this list of HGT is almost certainly an underestimate . 10 . 7554/eLife . 22144 . 003Figure 1 . Extensive horizontal gene transfer in the cheese microbiome . ( A ) All HGT events in analyzed cheese-associated bacteria . Connection thickness is scaled to number of shared protein coding sequences . Maximum likelihood tree based on 16S RNA alignment using Ribosomal Database Project ( RDP ) . ( B ) HGT events clustered into 264 ‘groups’ based on genomic proximity . Groups are plotted based on total nucleotide content ( x-axis , from low to high ) , and the mean number of genes per species ( y-axis ) . Diameter of each circle is proportional to the total number of species in the group . Groups highlighted in red are described further in the text . ( C ) Quantification of KEGG modules and submodules for protein coding genes ( CDS ) identified as horizontally transferred . Annotations were generated by BLAST-Koala . Genes without function prediction are not depicted . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 00310 . 7554/eLife . 22144 . 004Figure 1—figure supplement 1 . Schematic of software pipeline to identify HGT . ( 1 ) Sequenced genomes are annotated with IMG/ER and downloaded in Genbank format . ( 2 ) All annotated genes in all genomes are used to assemble a BLAST database using BLAST+ command-line tools . ( 3 ) All protein coding genes ( CDS ) from all species are queried against the BLAST database . Hits from the same species are discarded; hits from species with an ANI > 89% are discarded; other hits are saved . ( 4 ) For each species , coding sequences that have at least one BLAST hit are grouped into islands based on proximity . Genes that are within 5 kb of each other on the same contig are considered to be part of the same island . ( 5 ) Islands in each species are compared with islands in other species to form groups . Islands that share at least one gene in common according to BLAST parameters in step 3 are placed in the same group . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 00410 . 7554/eLife . 22144 . 005Figure 1—figure supplement 2 . Same as Figure 1A with branch labels . All HGT events in analyzed cheese-associated bacteria . Connection thickness is scaled to # of shared protein coding sequences . Phylogenetic tree based on 16S RNA alignment using Ribosomal Database Project ( RDP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 005 As multiple genes can be transferred in a single HGT event , we next assembled the putative HGT genes into groups based on genomic proximity . Individual coding sequences ( CDS ) for each species were grouped into islands if they were found within 5000 nucleotides of one another on the same contig . These islands were then clustered with islands in other species that shared at least one CDS . The 4733 genes were clustered into 264 individual groups ( Figure 1B , Supplementary file 1d ) . Mobile elements such as transposons complicate our method of group clustering , as non-contiguous islands may be grouped together if they share a common transposon . Indeed , this appears to have occurred with Group 1 , which contains genes from several disparate genomic regions . In other cases , a single species may have genes in a single group spread across multiple contigs ( Figure 2—figure supplement 1 ) , but this may accurately represent a single HGT event . Most HGT groups identified ( 231 , 87 . 5% ) contain only members of the same phylum , or even a single genus ( 183 or 69% , Supplementary file 1d ) . This supports previous studies which suggest that HGT is most prevalent among closely related species ( Ravenhall et al . , 2015 ) . However , we uncovered several notable exceptions . For example , Alkaibacterium kapii FAM208 . 38 , a Firmicute , has a substantial ( ~8 kb ) fraction of Group 1 , which is predominantly found in Actinobacteria species . Groups 2 and 3 each have hits found in three different phyla ( although both are predominantly found in Actinobacteria ) . HGT can enable rapid evolution of microbes entering a new environment , with genes that are under positive selection observed more frequently ( Wiedenbeck and Cohan , 2011 ) . Identifying the functions of genes that are frequently transferred could help to identify the selective forces that are most important for adapting to the cheese rind environment . Across all of the genes identified in our analysis , the most abundant gene functions are transposases , conjugal transfer genes , phage-related proteins , and other mobile elements ( 631/4733 or 13% of all protein coding sequences ) . A third ( 86/264 or 32 . 6% ) of all HGT groups contain mobile elements . These genetic elements are likely involved either directly or indirectly in mobilization and transfer of DNA . To determine whether gene functions other than mobile elements are enriched in identified HGT regions , we used BlastKOALA ( Kanehisa et al . , 2016 ) to assign KEGG functional annotations ( Figure 1C , Figure 2—figure supplement 1c ) . In approximately half ( 53% ) of the genes , KEGG annotations could not be assigned . Of the KEGG-annotated genes , the most frequent module ( 281/2264 or 11% ) was ‘metal ion , iron siderophore and vitamin B12 transport systems’ . Five of the 10 largest HGT groups as measured by total number of genes ( Groups 1 , 2 , 3 , 7 and 8 ) contained siderophore transport systems ( K02013-K02016 ) . Low availability of iron in cheese is known to limit the growth of several bacterial species ( Monnet et al . , 2012 , 2010 ) . Previous work has also shown that genes involved in iron acquisition are present in higher numbers in cheese-associated species compared with closely related species from other environments ( Monnet et al . , 2010; Walter et al . , 2014 ) . Many other horizontally transferred genes ( 267/2264 or 12% of KEGG-annotated genes ) are also involved in transport of nutrients relevant for growth in the cheese environment . Lactate is an abundant carbon source in freshly made cheese , as the initial stages of cheesemaking involve fermentation of lactose to lactate by lactic acid bacteria ( Button and Dutton , 2012 ) . We observed a large number of genes ( 63/2264 or 2 . 8% of KEGG-annotated genes ) involved in lactate import ( lactate permease - K03303 ) or lactate metabolism . Lactate dehydrogenase ( K00101 ) , which reduces lactate to pyruvate , represented nearly 1% of all horizontally transferred protein coding sequences . Apart from lactate , the primary source of energy for microbial growth in cheese is metabolism of the abundant lipids and proteins , particularly casein ( Monnet et al . , 2015 ) . We also identified glutamate importers ( 43/2264 or 1 . 9% , eg . K12942 , K10005-K10008 ) and short peptide/nickel transporters ( 88/2264 or 3 . 9% eg . K03305 ) in our analysis , suggesting pathways for utilization of casein degradation products . Transporters for micronutrients , including phosphonate ( K05781 , K06163-K06165 ) , molybdate ( K02017 , K02019 , K02020 , K03750 , K03750 , K03639 ) , and metal ions like zinc and manganese were also identified . HGT of drug resistance genes is of particular concern , as mobile resistance genes from food-associated microbes may be transferred to animal- and human-associated microbes ( Rossi et al . , 2014 ) . Cheese rind communities often contain filamentous fungi including Penicillium species and other microbes that could potentially produce antimicrobial compounds and thus select for antibiotic resistance in co-occurring species . However , less than 1% of KEGG-annotated genes in this dataset are related to drug resistance . A tetracycline resistance gene was identified in eight Brevibacterium species ( group 10 ) and a tripartite multidrug resistance system ( K03446 , K03543 ) in three Pseudomonas species ( group 37 ) . We also noted a small number of genes from the core genome that were not expected to be horizontally transferred . For example , group 27 is found in all 10 strains of B . linens in this dataset , as well as the closely related B . antiquum CNRZ918 , and contains the SSU ribosomal protein S1p , as well as DNA Polymerase 1 . It is possible that these results are false positives as B . linens and B . antiquum have an ANI ~88% , and these genes are typically more highly conserved than average . At the same time , other ribosomal genes that should also be highly conserved protein coding genes have substantially lower homology between these species than S1p ( Supplementary file 1e ) . Further , another gene within this HGT group ( SAM-dependent methyltransferase ) is not typically highly conserved , but nevertheless is >99% identical between these Brevibacterium species . We cannot exclude the possibility that this is a false positive , but this may be an example of homologous recombination facilitated by the high sequence identity of the ribosomal protein gene . Several other groups also contain ribosomal proteins ( 42 L5p and S3p , 180 L4p , 219 S3p ) , but these groups do not contain any other protein coding genes , and they are clustered with other ribosomal protein coding genes , which is a more typical arrangement . The abundance of iron acquisition genes identified as HGT suggests that iron is a driving force in adaptation to growth on cheese . The largest HGT region we identified in cheese-associated bacteria , Group 1 , includes an island of ~47 kbp ( ~1% of the genome of B . linens JB5 ) and 34 genes . This island is found in whole or in part in 15 different species in five different Actinobacterial genera ( Brachybacterium , Brevibacterium , Corynebacterium , Microbacterium , and Glutamicibacter , formerly Arthrobacter ) , and one Firmicute ( Alkalibacterium ) . The core of this region , flanked by AraC-like transcriptional regulators ( e . g . Ga0099663_102740 and Ga0099663_102753 from JB182 ) , contains several genes predicted to form a siderophore import complex , including two cell-surface associated substrate binding protein genes ( Ga0099663_102743–44 ) , two membrane permease genes ( Ga0099663_102745–46 ) , and an ATPase subunit ( Ga0099663_102747 ) . A siderophore reductase ( Ga0099663_102741 ) is present immediately downstream of the AraC regulator , but has less than 99% identity between the species we analyzed ( Figure 3A , Supplementary file 1c ) . We named this region iRon Uptake/Siderophore Transport Island ( RUSTI ) . 10 . 7554/eLife . 22144 . 006Figure 2 . HGT Groups in Actinobacteria , Firmicutes , and γ-Proteobacteria groups . ( A ) The 31 largest HGT groups that contain predominantly Actinobacteria . The areas of circles are scaled to log2 ( n ) , where n is the total number of nucleotides in that group for each species . The largest circle size represents the largest HGT group in that phylum . Phylogenies ( left ) are based on small subunit ribosomal RNA alignment . ( B ) The 25 largest HGT groups that contain predominantly Firmicutes . ( C ) The 28 largest groups that contain predominantly γ-Proteobacteria . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 00610 . 7554/eLife . 22144 . 007Figure 2—figure supplement 1 . Group A: Expected clustering: contiguous genes in multiple species are in a single group . Although island 6 ( i6 ) lacks one gene present in i1 and i4 , ( possibly because of a transposon insertion ) , it is still considered related . Group B: Ambiguous grouping: islands 2 and 3 from species 1 are found on different contigs , but are grouped together . They may be found in close proximity in the genome , but on different sides of a gap in the assembly , or they may be quite distant from each other . The grouping of related genes in species 2 into a single island suggests that they may have been transferred in a single event , but the possibility of two unrelated HGT events landing in the same spot cannot be excluded . Group C: Possible mis-grouping of two HGT events in a single group: although species 4 does not share any genes with species 1 and 2 , these islands are nevertheless clustered because of the proximity of coding sequences in species 3 . This may correctly represent a single gene cluster that subsequently diverged in each species , or unrelated HGT that happened to insert in close proximity . Group D: Mis-grouping because of mobile element: Mobile elements ( outlined in red ) found in multiple locations in multiple genomes may insert next to unrelated HGT islands , causing spurious grouping by the algorithm . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 00710 . 7554/eLife . 22144 . 008Figure 3 . Structure of RUSTI islands . ( A ) At-scale schematics for genomic context of HGT Group 1 ( top ) for B . linens JB5 and G . arilaitensis JB182 and alignment of RUSTI operon ( bottom ) . Nucleotide position values ( top ) refer to contigs Ga0099665_11 and Ga0099663_102 respectively . Dotted line for JB5 depicts regions of the contig that are not shown . Nucleotide position values ( bottom ) refer to operon starting from stop codon of leading AraC coding sequence . ( B ) At-scale schematics for genomic context of HGT Group 7 for Halomonas sp . JB37 and V . casei JB196 ( top ) and alignment of iron and phosphonate metabolism genes ( bottom ) . Nucleotide position values ( top ) refer to contigs Ga0099667_11 and Ga0099672_104 respectively . Grey lines for JB196 depict gaps in the alignment resulting from insertions in JB37 . Nucleotide position values ( bottom ) refer to operon starting from stop codon of leading protein coding sequence . ( C ) At-scale schematics for genomic context of HGT Group 31 for S . fleuretti . CIP106114 and S . vitulinus Ma1 ( top ) . For both species , the group is split across two different contigs and nucleotide position values ( top ) refer to the relative position for that contig . Alignment of iron and phosphonate metabolism genes from Group 31 ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 008 Horizontally transferred genes are not always expressed in the recipient genome , because of possible incompatibilities in promoter sequence ( Ochman et al . , 2000 ) . As iron is a limiting resource in cheese ( Kastman et al . , 2016; Monnet et al . , 2012 ) , we reasoned that if RUSTI is a functional operon , it would likely have increased expression in the presence of additional competition for iron . To assess whether RUSTI genes are regulated in the presence of competition , we grew G . arilaitensis JB182 alone or in the presence of Penicillium and performed RNA sequencing ( RNA-seq ) to monitor gene expression . The genes in RUSTI were significantly upregulated in the presence of a competing microbe relative to growth alone ( Figure 3B , Supplementary file 1f ) , suggesting that this horizontally transferred region is transcriptionally active and may be responding to competition for limited iron in cheese . Hundreds of different siderophores have been identified belonging to three major classes: hydroxamate , catecholate , and α-hydroxycarboxylate ( Hider and Kong , 2010 ) . To predict the function of the RUSTI transporters , we compared their protein sequences to the Transporter Classification Database ( TCDB ) ( Saier et al . , 2016 ) using BLAST ( Supplementary file 1g ) . The two genes annotated as permease subunits and one gene annotated as an ATP binding subunit each share substantial homology with the catechol ferric enterobactin transport system ( FepD , FepG , and FepC , respectively ) in E . coli ( Elkins and Earhart , 1989; Shea and McIntosh , 1991; Chenault and Earhart , 1991 ) . Two genes annotated as substrate binding proteins have weak homology to vibriobactin and iron ( III ) dicitrate binding proteins from Vibrio cholerae and E . coli , respectively . Siderophore-related genes are also well-represented in γ-Proteobacterial HGT groups . Like Group 32 in Actinobacteria , Group 39 contains both siderophore acquisition and siderophore biosynthesis genes and is found in three species of Psychrobacter . The HGT group with the most protein coding genes that we identified in γ-Proteobacteria ( Group 7 ) is found in several Vibrio and Halomonas species , and like ActinoRUSTI contains an ABC siderophore transport system with individual substrate-binding , permease , and ATP-binding domains ( Figure 3C ) . Although this group appears to have analogous function in the acquisition of iron with RUSTI from Actinobacteria , this ProteoRUSTI does not appear to be related . TCDB analysis suggests homology to hemin transporters in Yersinia pestis and Bordetella pertussis ( Supplementary file 1g ) . The same gene island also contains genes related to the Phn family involved in phosphonate import and metabolism ( Jochimsen et al . , 2011 ) . Phosphonate metabolism genes have previously been associated with iron siderophore acquisition in acidic environments ( Osorio et al . , 2008 ) , and cheese is typically close to pH5 during the initial periods of rind community growth . Interestingly , BLAST of this region against the NCBI RefSeq database reveals that several uropathogenic E . coli strains share identical DNA sequences ( Supplementary file 1h ) . Highly similar sequences are found in Oligella urethralis , another gram-negative pathogen of the urogenital tract , and Vibrio harveyi , a bioluminescent ocean-dwelling microbe . Iron sequestration by animals is a common defense against pathogens ( Parrow et al . , 2013 ) , and enhanced iron acquisition is commonly associated with virulence . Mammals produce lactoferrin in milk for the same reason ( Ellison , 1994 ) , and these data suggest that the same genes would be adaptive in both pathogenesis and growth on cheese . The convergence of strategies for both pathogenesis and growth on cheese was also demonstrated in the Firmicutes ( Figure 3D ) . Two species of cheese-associated Staphylococcus ( S . fleuretti CIP106114 and S . vitulinus Ma1 ) share a large ( ~20 kb ) cluster of genes ( Group 31 ) that includes hemolysin and fibronectin binding protein ( FnBP ) , which are involved in virulence in S . aureus ( Sinha et al . , 1999; Cheung and Ying , 1994 ) . Hemolysin ( also known as alpha toxin ) forms pores in cell membranes , and is so-called because of its ability to lyse red blood cells . FnBP enables binding to and invasion of cells , and has been implicated in formation of biofilms in methicillin resistant S . aureus ( McCourt et al . , 2014 ) . It is unlikely that these genes provide a selective advantage to cheese-associated Staphylococci , but Group 31 also contains genes for iron acquisition . These genes are homologous to the EfeUOB systems in E . coli and Bacillus subtilis and FepABC system in S . aureus , which are active in low-pH conditions ( Cao et al . , 2007; Miethke et al . , 2013; Turlin et al . , 2013 ) . These iron acquisition genes are also found in association with hemolysin and FnBP in S . aureus and the animal-associated S . sciuri , although at only ~80% nucleotide identity ( Supplementary file 1h ) ( Kloos and Schleifer , 1976 ) . As iron acquisition genes may be adaptive both on cheese and in pathogenesis , it is possible that this region was acquired from an animal pathogen , and the virulence genes have been preserved because of their association with these genes . The genomes sequenced are from a limited number of cheeses from Europe and the United States . To determine the distribution of RUSTI across a more expansive sample of cheese microbiomes , we used BLAST searches against assembled metagenomic data from 38 different cheeses using representative Proteo- , Actino- , and Staph-RUSTI sequences ( Figure 4 ) . Gene islands at least 97% identical to ActinoRUSTI were readily identified in 23 ( 61% ) of these metagenomes , in both natural and washed rind cheeses from the United States and Europe . Although less common ( 26% of metagenomes ) , ProteoRUSTI was also identified in diverse cheeses in both the US and Europe . StaphRUSTI could not be found in any of the metagenomes we analyzed . These data demonstrate that siderophore-associated HGT islands are widespread in cheese rind microbiomes . Whether independent HGT events occur within each cheese production and aging facility , or whether they occurred before the widespread distribution of these microbes across cheese production regions is unknown . 10 . 7554/eLife . 22144 . 009Figure 4 . Presence of RUSTI in cheese metagenomes . Genes in ActinoRUSTI ( G . arilaitensis JB182 ) and ProteoRUSTI ( V . casei JB196 ) regions were compared with 32 assembled metagenomes from the US and Europe . Filled CDS represents a positive ( >97% identical nucleotides ) hit in that metagenome . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 009 To begin to understand potential mechanisms which could mediate HGT in cheese-associated bacteria , we analyzed the sequences surrounding the RUSTI region of Glutamicibacter JB182 . Conjugative elements are commonly involved in HGT ( Wozniak and Waldor , 2010 ) . Integrative and conjugative elements ( ICEs ) can in part be identified by the presence of signature proteins associated with core functions of integration into and excision from the host genome ( recombinase ) , replication as an extrachromosomal element ( polymerase ) , and conjugation from the host to recipient cell ( conjugation ) ( Ghinet et al . , 2011 ) . Analysis of the Glutamicibacter JB182 RUSTI region revealed homologs of each of these protein classes ( Figure 3A ) : a recombinase of the site-specific tyrosine recombinase XerD family ( Ga0099663_102762 ) ( Subramanya et al . , 1997 ) , a hexameric ATPase conjugation protein of the VirD4/TraG/TraD family ( Ga0099663_102784 ) ( Hamilton et al . , 2000 ) , and a homolog of the bi-functional primase-polymerase DNA replication protein family ( Ga0099663_102766 ) . Interestingly , Actinobacterial ICE systems typically use conjugation apparatus belonging to the SpoIIE/FtsK family , which allows transfer of double-stranded DNA ( te Poele et al . , 2008; Bordeleau et al . , 2012 ) . However , the conjugation machinery here is more reminiscent of gram-negative and Firmicute systems of single-stranded transfer ( Burrus and Waldor , 2004 ) . ICE integration is site-specific , and frequently occurs at the 3’ end of tRNA genes ( Ghinet et al . , 2011 ) . Immediately downstream of the RUSTI region in Glutamicibacter is a leucine tRNA . The 3’ end of the tRNA forms an imperfect repeat with the region immediately upstream of the RUSTI region , which strongly suggests that the tRNA-Leu is used at the integration site ( att site ) for this ICE . To determine whether this ICE was still active , we performed PCR using primers within and flanking the putative integration site ( Figure 5A ) . We were able to detect PCR products which suggest that at least a portion of the cells within the population have lost the RUSTI ICE from their chromosome , and it is present as an extrachromosomal circular form ( Figure 5B ) . Sequencing of the PCR product ( primers 1 + 6 ) that spans the predicted excision site matched the predicted remaining sequence , containing Repeat element B ( Figure 5D ) . Sequencing of the PCR product ( primers 2 + 5 ) that spans the predicted circularization site matched the predicted sequence , containing Repeat element A ( Figure 5E ) . 10 . 7554/eLife . 22144 . 010Figure 5 . Mobility of RUSTI . ( A ) Schematic for PCR primer design - see Materials and methods for details . ( B ) PCR testing for the presence of RUSTI and for the excision of the ICE in an overnight culture of G . arilaitensis JB182 . ( C ) DNA was extracted from five commercially available starter cultures and tested for the presence of RUSTI using PCR with primers specific for the HGT region ( Materials and methods ) . Starter culture 3 was plated on PCAMS media , and four isolates selected based on colony morphology were also tested . The expected size for the amplicon is ~1 . 4 kb . Sequencing of the 16S ribosomal RNA genes for these isolates suggested that two isolates are Glutamicibacter arilaitensis and two are Brevibacterium linens . G . arilaitensis . JB182 was used as a positive control . ( D ) The ~2000 bp band from the PCR amplification using primers 1 and 6 and ( E ) the ~500 bp band from amplification using primers 2 and 5 were extracted and sequenced . Alignment with the JB182 genome reveals 100% alignment with expected and the spliced chromosomal region containing the 3’ repeat and the excision circle containing the 3’ repeat respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 22144 . 010 There are several possible explanations for the widespread distribution of nearly identical ActinoRUSTI . Initial transfer events may have occurred in a single location , on the surface of a cheese or in livestock , and subsequently been dispersed to many separate cheesemaking facilities . The continued mobility of the ICE in JB182 raises the alternate possibility that it may be continually introduced to many cheeses from a common source . Cheese producers often use commercially available ‘starter’ cultures that contain desirable species , including many Actinobacteria ( Robinson , 2005 ) . We tested five common starter cultures for the presence of ActinoRUSTI by PCR , and positively identified it in one of these ( Figure 5C ) . This culture is known to contain two species of Actinobacteria , Brevibacterium linens and Glutamicibacter nicotinae . To identify the RUSTI donor species , we plated the starter culture and isolated four distinct strains based on colony morphology . SSU sequencing revealed that two isolates were Glutamicibacter and two were Brevibacterium . One of the Brevibacterium isolates tested positive for ActinoRUSTI by PCR ( Figure 5C ) . Although we have no evidence for direct transmission of ActinoRUSTI from this starter culture , and thus cannot definitively conclude that this was the source , these data are consistent with the hypothesis that HGT from a starter culture could explain some of the dissemination of ActinoRUSTI .
In this paper , we provide evidence of extensive horizontal gene transfer in cheese-associated bacteria . Many of the transferred regions are large multi-gene islands , and are shared by numerous species . Genes involved in nutrient acquisition , especially iron and lactate , are particularly abundant , suggesting that HGT may provide a selective advantage in the iron- and sugar-limited environment of cheese . The largest HGT we identified appears to be an active ICE and is found in a starter culture , raising the possibility that we are observing contemporary processes that may have ongoing importance . These data support previous studies that show HGT is an important factor in the evolution of microbial communities ( Wiedenbeck and Cohan , 2011 ) and suggest that cheese rind communities may provide a useful model for studying this process in greater detail . For this study , we focused on bacterial members of the cheese microbiome , but many cheeses also contain numerous fungal species . Indeed , HGT has been previously documented in cheese-associated Penicillium species ( Cheeseman et al . , 2014 ) . HGT between bacteria and fungi has also been documented in other environments , although it is thought to be rare ( Keeling and Palmer , 2008 ) . Evaluating bacterial-fungal gene transfer in cheese could provide additional insights into the extent and importance of gene exchange in microbial communities . Further sequencing of bacterial genomes from cheese could also continue to reveal HGT . We were able to show that ActinoRUSTI in G . arilaitensis JB182 is likely contained within an integrative and conjugative element , although many other mechanisms for transfer are likely at play . Indeed , the appearance of phage-related genes , transposases , and other mobile elements in many HGT groups suggests that we are observing the results of multiple methods for mobilizing DNA . Our method for identifying HGT does not permit determination of the direction of gene flow , and indeed it seems likely that the original sources of many of these genetic elements are not present in our dataset . In some cases , we can infer a possible origin , such as the Brevibacterium strain in a starter culture that may be the source of RUSTI in multiple cheese species around the world . However , this does not permit identification of where this species acquired them . Further characterization of cheese-associated microbes , as well as those found in dairy farms or in cheese caves may provide a more complete picture , but the evidence that at least some of these genetic elements are found in human pathogens and ocean-dwelling bacteria suggests that genes are shared across diverse environments . Although previous studies demonstrated that iron is limiting for Glutamicibacter , Brevibacterium , and Corynebacterium species growing on cheese ( Monnet et al . , 2012 , 2010 ) , the preponderance of siderophore and other iron acquisition genes we observed being horizontally transferred suggests that the same is true across bacterial phyla . Limiting iron is a deliberate strategy on the part of mammalian hosts to block the growth of infectious microbes , and this strategy influences the composition of milk because of the presence of lactoferrin ( Ellison , 1994 ) . Interestingly , convergent strategies for acquiring iron are used by pathogens and by cheese-associated microbes and we observe that in some cases these disparate species appear to have shared genes through horizontal transfer . The presence of these same genes in a microbe found in ocean habitats suggests that these genes have broad utility for the common challenge of iron limitation . We have yet to demonstrate the functional consequences of these genes on individual species or on the community as a whole . Given that iron is limiting in cheese , and that ActinoRUSTI genes are upregulated in response to other species ( Figure 4C ) , it is likely that these genes are functional and may play a role in competition . The prevalence of siderophore import , but not siderophore synthesis pathways suggests that species may cheat by scavenging the biosynthetic products of others ( Cordero et al . , 2012 ) . The identification of widespread sharing of genes in cheese microbial communities could have important implications . In particular , the possibility that a starter culture is the source of mobile gene elements suggests that the genomic content , rather than just specific species , must be considered when designing microbial supplements . In addition to starter cultures used for fermented foods , living microbial supplements ( ‘probiotics’ ) are increasingly being adopted in agriculture ( Verschuere et al . , 2000; Chaucheyras-Durand and Durand , 2010 ) and for a wide range of human health conditions ( Cuello-Garcia et al . , 2015; Onubi et al . , 2015; ( IBS Dietetic Guideline Review Group on behalf of Gastroenterology Specialist Group of the British Dietetic Association ) et al . , 2016 ) , and even as cosmetics ( Whitlock et al . , 2016 ) . The need to screen for clinically relevant elements such as antibiotic resistance genes is widely recognized ( Sanders et al . , 2010 ) , but other mobile gene elements from these organisms may also enter native microbial populations with unknown consequences . Although we and others have observes a large number of HGT events in microbial species across a diverse range of environments ( McDaniel et al . , 2010; Smillie et al . , 2011; Ravenhall et al . , 2015 ) , the biotic and abiotic conditions that affect this frequency , and what effects HGT may have on the community remain unclear . A model system to study HGT in a community context is particularly important , as sequence-based characterization of complex communities has particular limitations when it comes to HGT . Further , even if complete characterization in situ were possible , many microbial communities are difficult to experimentally manipulate in vitro . By contrast , cheese rind-associated bacteria are readily isolated and cultured , and model communities may enable identification of features of microbial communities and their environments that alter frequency and extent of HGT . The cheese rind model system provides an opportunity to observe HGT as it happens and to investigate how community composition affects the frequency of transfer and the persistence of genes . The in vitro cheese system enables experimental manipulation to investigate the role of community composition in driving HGT . Further , as many gene products may only have survival benefits in the context of community competition and cooperation , investigating the role of RUSTI and other horizontally transferred genes on microbial growth in the context of their natural community is critical . Having an experimentally tractable microbial community will allow us to test these ideas under controlled conditions in the laboratory and generate predictions about how these processes work in more complex natural systems . Many species of cheese-associated bacteria have close relatives in the soil , on skin , in the ocean , making insight from this system potentially applicable to diverse environments . In addition , horizontal acquisition of iron-uptake genes has been documented in numerous environments including the oceans and in human pathogens ( Gyles and Boerlin , 2014; Richards et al . , 2009 ) , suggesting that the specific processes occurring in cheese may also be generally informative across systems . Understanding the extent of HGT in the cheese microbiome is the first step towards addressing how the movement of genes shapes , and is shaped by , a microbial community . Using cheese rinds as a model system could help elucidate the factors that influence the frequency of HGT , how it impacts competition and cooperation , and helps shape a microbiome .
Bacterial strains JB4 , 5 , 7 , 37 , 110 , 182 , 196 , and 197 were isolated from cheeses in a single geographic region and sequenced using a combination of Illumina short-read ( 100 bp , paired end ) and Pacbio long-read sequencing . DNA was extracted using Genomic Tip 100/G ( Qiagen , USA ) or Power Soil ( MoBio , USA ) . Illumina library preparation and sequencing were performed at Harvard University by the Bauer Core facility . Pacbio library preparation and sequencing were performed by the University of Massachusetts Medical School Deep Sequencing Core . De novo hybrid assembly was performed using SPAdes ( v3 . 5 . 0 ) ( Bankevich et al . , 2012 ) . Genomes were annotated using the Integrated Microbial Genomes Expert Review ( IMG/ER ) annotation pipeline ( Markowitz et al . , 2012 ) . In addition , we also sequenced eight additional rind isolates of Brachybacterium ( strains 341 . 9 , 738 . 10 , 862 . 8 , 876 . 9 , 900 . 8 , 908 . 11 , 947 . 1 , 962 . 10 ) and Brevibacterium ( strains 341 . 13 , 738 . 8 , 862 . 7 , 876 . 7 , 900 . 6 , 908 . 7 , 947 . 7 , 962 . 8 ) and three additional isolates of Glutamicibacter ( strains BW77 , 78 , 80 ) from different cheeses in a broad geographic distribution . For these isolates , we prepared draft genomes using Illumina short-read sequencing and assembled with CLC genomics workbench . The annotated genomes used can be found on Zenodo ( Bonham et al . , 2016 ) 16S sequences were retrieved from the sequenced genomes and aligned using the structure-based aligner , Infernal v1 . 1rc4 ( Nawrocki and Eddy , 2013 ) , as implemented in the Ribosomal Database Project release 11 ( Cole et al . , 2009 ) . The alignment was imported into Geneious v9 ( Biomatters , LTD ) , and a tree was calculated using the maximum likelihood method PHYML ( GTR model ) ( Guindon et al . , 2003 ) . The tree was rooted using Thermus thermophilus . The tree was then uploaded to Interactive Tree of Life ( iTOLv3 ) ( Letunic and Bork , 2016 ) to enable mapping of HGT data ( connections and group abundance profiles ) . Four replicate transcriptomes from two treatments were sequenced: ( 1 ) G . arilaitensis alone and ( 2 ) G . arilaitensis + Penicillium . We used a strain of Penicillium solitum that was isolated from a natural rind cheese and was used for experiments in Wolfe et al . ( 2014 ) . For each experimental unit , approximately 80 , 000 CFUs of Glutimicibacter arilaitensis were spread across the surface of a 100 mm Petri dish containing 20 mL of cheese curd agar ( 10% freeze-dried fresh cheese , 3% NaCl , 1 . 7% agar , 0 . 5% xanthan gum ) ( Wolfe et al . , 2014 ) . For the + Penicillium treatment , approximately 100 , 000 CFUs were co-inoculated onto the plates with the Glutimicibacter . Plates were incubated in a bin with moist paper towel ( >90% relative humidity ) at 24°C for 5 days . Rind biofilms were then harvested by scraping the cheese curd surface and stored in RNAProtect Reagent ( Qiagen ) to stabilize mRNA frozen at −80°C . RNA was extracted using a standard phenol-chloroform protocol used for many different fungal and bacterial species , which has been adopted from transcriptomics work in gut microbiomes ( see [David et al . , 2014] ) . This protocol uses a standard bead-beating step in a lysis buffer to release cell contents from biofilms stored in RNAProtect . DNA was removed from the samples using a TURBO DNA-free kit ( Life Technologies ) , and 5S tRNA and large rRNA was depleted using MEGAClear ( Life Technologies ) and RiboZero ( Illumina ) kits , respectively . To remove both fungal and bacterial large rRNA , we used an equal mixture of Ribo-Zero Yeast and Bacteria rRNA Removal Solution . To confirm that the samples were free of DNA contaminants , a PCR of the 16S rRNA gene was performed with standard primers ( 27 f and 1492 r ) . Overall quantity and quality of the RNA preps were confirmed by Nanodrop and Agilent 2100 Bioanalyzer using the RNA 6000 Nano kit . RNA-seq libraries were constructed from purified mRNA using the NEBNext Ultra RNA Library Prep Kit for Illumina ( New England Biolabs ) where each library received a unique six base pair barcode for demultiplexing after the sequencing run . Each library was run on an Agilent 2100 Bioanalyzer High Sensitivity DNA chip to confirm that primer dimers and adapter dimers were not present in the sample and to determine the size of the library . Final libraries were standardized to 10 nM each after quantification with a Qubit dsDNA HS Assay Kit ( Life Technologies ) and then pooled in equal amounts to get similar sequencing coverage across all libraries . The pooled library samples were sequenced using paired-end 100 bp reads on an Illumina HiSeq Rapid Run by the Harvard Bauer Core Sequencing Core Facility . To quantify gene expression and determine whether genes within RUSTI were differentially expressed when grown with the competitor Penicillium , we used Rockhopper ( McClure et al . , 2013 ) . Only forward reads were used for this analysis . The assembled and annotated Glutamicibacter arilaitensis strain JB182 ( described above ) genome were used as a reference genome for mapping . We considered genes that had a greater than fourfold difference in expression when grown with Penicillium or Staphylococcus and were significantly different ( based on Rockhopper’s q-values , which control for false discovery rate using the Benjamini-Hochberg procedure ) to be differentially expressed genes . PCR reactions were performed using Q5 Hot Start Mastermix ( New England Biolabs ) . Where JB182 RUSTI is integrated in the chromosome , PCR using primer 1 ( CAACTGTGCCACGCAATTCA ) and primer 2 ( CGGCTACTTCTCGGATGGTC ) are expected to produce a 1037 bp product that includes the 5’ ICE repeat . Primer 3 ( CGCAATCGTGGTGTATCTGC ) and primer 4 ( GACGGGATCAGGAACGACG ) should produce a 1410 bp product , while primer 5 ( GCCGCATCTACCTCGATGAA ) and primer 6 ( CCAAATCGCGACGCATTGAT ) are expected to form a 1467 bp product . Primers 1 and 6 are separated by approximately 59 kb when RUSTI is present and are not expected to form a PCR product , but should form a 1937 bp product if RUSTI is excised . Primers 2 and 5 should not form a PCR product when RUSTI is integrated , but would form a 500 bp product if the excision circle were present . Annotated genomes were compared using blastn from BLAST+ ( v2 . 3 . 0 ) ( Camacho et al . , 2009 ) . Protein coding genes were considered to be potential HGT if their sequence was at least 99% identical for at least 500 nucleotides . Neighboring candidate HGT were identified as part of the same island if they were separated by no more than 5000 nucleotides . Scripts to import and store genome information and blast results and to analyze results are available on github ( Bonham , 2016; https://github . com/kescobo/kvasir ) . A copy is archived at https://github . com/elifesciences-publications/kvasir . Genomic average nucleotide identity ( ANI ) was calculated using the ‘ani . rb’ script from the enveomics collection ( commit ‘e8faed01ff848222afb5820595cccc4e50c89992’ ) with default settings ( Rodriguez-R and Konstantinidis , 2016 ) . Shotgun metagenomic data from ( Wolfe et al . , 2014 ) and ( Kastman et al . , 2016 ) were assembled with CLC Genomic Workbench 8 . 0 . Representative sequences for ActinoRUSTI or ProteoRUSTI were compared with assembled metagenomes by BLAST . Hits with >97% similarity were considered to be positive hits for target regions . Newly sequenced genomes were registered with NCBI with the bioproject ID PRJNA387187 . Biosample accession numbers for individual genomes are shown in Supplementary file 1a , and are as follows: Brevibacterium linens 341_13: SAMN07141149 , Brevibacterium linens 738_8: SAMN07141150 , Brevibacterium linens 862_7: SAMN07141151 , Brevibacterium linens 876_7: SAMN07141152 , Brevibacterium linens 900_6: SAMN07141153 , Brevibacterium linens 908_7: SAMN07141154 , Brevibacterium linens 947_7: SAMN07141155 , Brevibacterium linens 962_8: SAMN07141156 , Brachybacterium alimentarium 341_9: SAMN07141157 , Brachybacterium alimentarium 738_10: SAMN07141158 , Brachybacterium alimentarium 862_8: SAMN07141159 , Brachybacterium alimentarium 876_9: SAMN07141160 , Brachybacterium alimentarium 900_8: SAMN07141161 , Brachybacterium alimentarium 908_11: SAMN07141162 , Brachybacterium alimentarium 947_11: SAMN07141163 , Brachybacterium alimentarium 962_10: SAMN07141164 , Glutamicibacter sp . BW77: SAMN07141165 , Glutamicibacter sp . BW78: SAMN07141166 , Glutamicibacter sp . BW80: SAMN07141167 , Microbacterium sp . JB110: SAMN07141168 , Halomonas sp JB37: SAMN07141169 , Brevibacterium linens JB5: SAMN07141170 , Psychrobacter sp . JB193: SAMN07141171 , Brachybacterium sp . JB7: SAMN07141172 , Sphingobacterium sp . JB170: SAMN07141173 , Vibrio casei JB196: SAMN07141174 , Arthrobacter sp . JB182: SAMN07141175 , Corynebacterium sp . JB4: SAMN07141176 , Pseudoalteromonas sp . JB197: SAMN07141177 . | From the depths of the ocean to the lining of the human gut , almost every environment on Earth is home to a unique community of microorganisms referred to as a microbiome . Within these communities , unrelated microorganisms can exchange genetic information through a process known as horizontal gene transfer . For example , genes linked to antibiotic resistance are often transferred between different microorganisms , which can create increasingly drug resistant microbes and has important implications for human health . Horizontal gene transfer has been studied for almost 100 years , but examining it directly is challenging because , almost by definition , it requires studying a community of microbes rather than one microbe in isolation . As such , researchers are looking for simple models of microbial communities that can be easily manipulated in experiments . Bonham et al . have now turned to the outer surface of cheese , also known as cheese rind , to better understand horizontal gene transfer . As a model system , the cheese rind microbiome is relatively simple to work with because cheese rind is easy to replicate in the laboratory , and the microbes growing on cheese can be grown on their own or in combinations with other microbes . By comparing the genetic material of 165 cheese-associated bacteria to one another , Bonham et al . identified over 4 , 000 genes that were shared between the bacteria , including several large clusters of genes that were shared by many species . Many of the identified genes ( about 23% to be precise ) appear to help the microorganisms acquire nutrients that are known to be in short supply on the surface of cheese surface , including iron . Bacteria typically use specialized molecules called siderophores to scavenge for iron and uptake systems to carry the iron-bound siderophore back into the cell . Notably , only the genes associated with the uptake systems were found in some of the shared gene clusters . This finding suggests that horizontal gene transfer has allowed some microbes to “cheat” and take up iron-bound siderophores without expending energy to produce the siderophores themselves . Using the cheese rind microbiome as a model system , it becomes possible to explore how horizontal gene transfer works in more detail than before . A better understanding of this process can then be applied to other important microbiomes , including those where genes conferring antibiotic resistance are commonly exchanged . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"genetics",
"and",
"genomics"
] | 2017 | Extensive horizontal gene transfer in cheese-associated bacteria |
The fission yeast actin cytoskeleton is an ideal , simplified system to investigate fundamental mechanisms behind cellular self-organization . By focusing on the stabilizing protein tropomyosin Cdc8 , bundling protein fimbrin Fim1 , and severing protein coffin Adf1 , we examined how their pairwise and collective interactions with actin filaments regulate their activity and segregation to functionally diverse F-actin networks . Utilizing multi-color TIRF microscopy of in vitro reconstituted F-actin networks , we observed and characterized two distinct Cdc8 cables loading and spreading cooperatively on individual actin filaments . Furthermore , Cdc8 , Fim1 , and Adf1 all compete for association with F-actin by different mechanisms , and their cooperative association with actin filaments affects their ability to compete . Finally , competition between Fim1 and Adf1 for F-actin synergizes their activities , promoting rapid displacement of Cdc8 from a dense F-actin network . Our findings reveal that competitive and cooperative interactions between actin binding proteins help define their associations with different F-actin networks .
The self-organization of complex structures from interactions between basic components is a general phenomenon of chemistry and material sciences , as well as more complicated biological systems ( Karsenti , 2008 ) . Examples include the generation of self-segregating PAR domains in the developing C . elegans embryo ( Hoege and Hyman , 2013; Goldstein and Macara , 2007 ) and the organization of a mitotic spindle around DNA-coated microspheres ( Heald et al . , 1996 ) . How individual interactions within the cell coalesce to generate complex patterns or structures remains a fundamental biological question . The actin cytoskeleton is an ideal system to study complex cellular self-organization . Multiple functionally diverse F-actin networks , each with a distinct architecture and dynamics , assemble at the correct time and place within a single crowded cytoplasm . Distinct sets of actin binding proteins ( ABPs ) help to define the characteristics of each F-actin network by performing tasks such as actin filament ( F-actin ) nucleation , bundling , severing , and capping ( Pollard , 2016 ) . Therefore , proper localization of ABPs to the correct network is crucial to generate F-actin networks defined for specific processes . The biochemical activity and cellular functions of many individual ABPs have been well-studied . However , we are only beginning to understand how ABPs function in concert , how they compete with each other for association with individual actin filaments , and how these interactions contribute to the proper sorting of ABPs to diverse F-actin networks on a whole-cell scale ( Michelot and Drubin , 2011; Skau and Kovar , 2010; Jégou and Romet-Lemonne , 2016 ) . Fission yeast is an ideal simplified system in which to study the underlying molecular mechanisms behind F-actin network self-organization ( Kovar et al . , 2011 ) . Fission yeast has three primary actin cytoskeleton networks , in which all of the actin filaments are assembled by a distinct actin assembly factor: endocytic actin patches ( Arp2/3 complex ) , polarizing actin cables ( formin For3 ) , and the cytokinetic contractile ring ( formin Cdc12 ) . Moreover , each of these F-actin networks contains a distinct set of ABPs . We hypothesize that ABP competition for association with actin filaments is critical for their proper sorting to distinct F-actin networks . We previously discovered that competition between ABPs tropomyosin Cdc8 , fimbrin Fim1 , and ADF/cofilin Adf1 results in the exclusion of tropomyosin from actin patches ( Skau and Kovar , 2010 ) . Here , we use a combination of multi-color TIRF microscopy ( TIRFM ) of reconstituted F-actin networks and mathematical modeling to elucidate the underlying molecular mechanisms behind this series of competitive ABP interactions in fission yeast . By understanding ABP competition at a mechanistic level , we can gain insight into how an ABP’s intrinsic physical properties help dictate its interactions with other ABPs and its association with specific F-actin networks . In this study , we have found that different modes of active and passive competition exist between different ABPs . Furthermore , we have determined that cooperativity affects ABP competition by defining the ability of ABPs to both associate and be dissociated from an F-actin network . Finally , we have found that the combination of cooperative and competitive interactions between a set of ABPs defines the ABP composition and F-actin organization of the associated F-actin network .
Tropomyosin has been implicated in ABP sorting and F-actin network organization in many organisms including fission yeast ( Gunning et al . , 2015; Skau and Kovar , 2010; Clayton et al . , 2010 ) . Tropomyosin is a two-chained , parallel coiled-coil composed of two polypeptide chains with a characteristic heptad-repeat of hydrophobic residues ( Figure 1—figure supplement 1 ) . Individual tropomyosin coiled coils associate end-to-end to form continuous tropomyosin cables that extend along both sides of the helical actin filament ( Hanson and Lowy , 1963; Skoumpla et al . , 2007 ) . Vertebrate tropomyosins span six or seven actin subunits in the filament . Yeast tropomyosins are shorter: the two S . cerevisiae tropomyosin isoforms span four or five actins , while S . pombe tropomyosin Cdc8 ( hereafter called Cdc8 ) spans four actin subunits . Although Cdc8 , like other tropomyosins , has been purified and characterized in steady state bulk assays ( Cranz-Mileva et al . , 2015; Skau et al . , 2009; Skoumpla et al . , 2007 ) , our general mechanistic understanding of how tropomyosins load onto actin filaments and are affected by other ABPs is unclear . Multi-color TIRFM has been successfully utilized as a sensitive probe to study the detailed interactions between multiple ABPs and F-actin ( Bombardier et al . , 2015; Jansen et al . , 2015; Winkelman et al . , 2016 ) . However , fluorescently labeling tropomyosins is generally problematic as mutations or insertions within the protein can potentially disrupt its coiled-coil structure ( Greenfield and Hitchcock-DeGregori , 1995 ) . Additionally , tropomyosins labeled on the N- or C-terminus are not fully functional , as the presence of a label blocks end-to-end associations between tropomyosin molecules ( Brooker et al . , 2016 ) . Therefore , we created three distinct Cdc8 mutants , each containing an engineered cysteine mutation that could be labeled for visualization by TIRFM ( Figure 1—figure supplement 1 , Methods ) . We examined the functionality of each Cdc8 mutant in vitro and in vivo to identify the mutant most similar to wild-type Cdc8 . Two mutants , Cdc8 ( I76C ) and Cdc8 ( D142C ) , bound F-actin similarly to wild type Cdc8 ( Figure 1—figure supplement 1B–C ) , were able to be labeled and visualized by TIRFM ( 100% of actin filaments associated with tropomyosin , Figure 1—figure supplement 1D ) , and caused only very mild cytokinesis defects as the sole copy of Cdc8 in fission yeast ( Figure 1—figure supplement 2A–D ) . Therefore , we chose one of these mutants , Cdc8 ( I76C ) , for further study in TIRFM experiments . We utilized two-color TIRFM to examine the loading characteristics of Cy5-labeled Cdc8 on assembling Alexa 488-labeled actin filaments over a range of concentrations ( Figure 1A–B , Video 1 ) . At concentrations below 1 μM , Cdc8 was not observed to bind F-actin . However , at 1 μM Cdc8 , short stretches of Cdc8 were observed to load on actin filaments . These dynamic stretches varied in size over time , but could be as small as the observable limit of ~100 nm and generally remained within the constraints of ~0 . 5–2 μm ( ~2% of total actin filament coverage ) over time . At 1 . 25 μM Cdc8 , spreading of Cdc8 cables from initial ‘seeds’ was observed , with ~40% of actin filament sites coated with Cdc8 . At concentrations of 2 μM Cdc8 and higher , ~98% of F-actin was coated with Cdc8 . This rapid shift in actin filament occupancy over a small Cdc8 concentration range indicates a high degree of cooperativity . One possibility is that Cdc8’s high cooperativity is a result of the previously demonstrated end-to-end associations between tropomyosin molecules ( ‘end-to-end cooperativity’ ) ( Figure 1C ) ( Caspar et al . , 1969; Greenfield et al . , 2006 ) . However , the cooperativity of muscle tropomyosin on F-actin has been found to not directly correlate with its end-to-end binding ability ( Willadsen et al . , 1992 ) , suggesting that other interactions may also influence tropomyosin’s cooperativity on F-actin ( Tobacman , 2008 ) . We therefore considered whether indirect interactions between Cdc8 cables on opposing sides of an actin filament or via long-range interactions along the length of the F-actin lattice ( ‘indirect cooperativity’ ) also contribute to Cdc8’s high cooperativity ( Figure 1C ) . We focused on investigating a potential role for indirect cooperativity in Cdc8 loading onto an actin filament by observing ( Figure 2 ) and quantifying ( Figure 3 ) Cdc8 loading events in detail under conditions near the inflection point of Cdc8’s cooperativity ( 1 . 25 μM Cdc8 ) . 10 . 7554/eLife . 23152 . 003Figure 1 . Tropomyosin Cdc8 loads cooperatively onto actin filaments . ( A ) Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 labeled ) with a range of concentrations of tropomyosin Cdc8 ( Cy5-labeled ) . Scale bar , 5 µm . ( B ) Plot of the fraction of actin filament bound by Cdc8 ( ‘Cdc8 occupancy’ ) over free Cdc8 dimer concentration . Data were fit to a Hill function , revealing a Hill coefficient >1 ( Hill=14 . 6 ) , that indicates cooperativity . Error bars represent standard error of the mean; n = 2 reactions . ( C ) Schematic of Cdc8 loading onto an actin filament . Observed cooperativity of Cdc8 could be the result of end-to-end binding of tropomyosin molecules ( ‘End-to-end cooperativity’ ) and/or indirect interactions between tropomyosin molecules via changes in the actin filament ( ‘Indirect cooperativity’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 00310 . 7554/eLife . 23152 . 004Figure 1—figure supplement 1 . Characterization of Tropomyosin Cdc8 mutants L38C , I76C , and D142C in vitro . ( A ) Characteristic tropomyosin coiled-coil heptad repeat organization . Residues with low sequence conservation localized at b- , c- , or f-sites on the outside of the coiled-coil were chosen for mutation to cysteine . ( B–C ) High-speed sedimentation assay of 1 μM of wild-type ( WT ) or mutant ( L38C , I76C , or D142C ) tropomyosin Cdc8 dimer binding to increasing ( 0–10 μM ) concentrations of actin ( B ) and quantification of Cdc8 in pellet normalized to actin in pellet ( C ) . ( D ) Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488-labeled ) and 2 . 5 μM Cdc8 mutants L38C ( left ) , I76C ( middle ) , or D142C ( right ) ( Cy5-labeled ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 00410 . 7554/eLife . 23152 . 005Figure 1—figure supplement 2 . Characterization of Tropomyosin Cdc8 mutants L38C , I76C , and D142C in vivo . ( A ) Morphology ( DIC and DAPI/calcofluor ) and actin organization ( BODIPY-phallicidin ) of cells with Cdc8 mutants replacing the endogenous cdc8 gene . Scale bars , 5 μm . ( B–D ) Quantification of the number of nuclei ( B ) , abnormal septa ( C ) , and the time of ring assembly ( D ) in cells expressing tropomyosin cdc8 mutants . Two-tailed t-tests for data sets with equal variance yielded p-values: WT vs . L38C: *p-value=1 . 80×10−10; WT vs . I76C: p-value=0 . 27 , WT vs . D142C: p-value=0 . 57 . n ≥ 10 cells for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 00510 . 7554/eLife . 23152 . 006Video 1 . Tropomyosin Cdc8 loads cooperatively onto F-actin , related to Figure 1 . Two-color TIRF microscopy of 1 . 5 μM actin ( Alexa-488 labeled ) with a range of concentrations of tropomyosin Cdc8 ( Cy5-labeled ) . Scale bar , 5 μm . Time in sec . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 00610 . 7554/eLife . 23152 . 007Figure 2 . Two distinct Tropomyosin Cdc8 cables load cooperatively onto a single actin filament . ( A–B ) Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 ) with 1 . 25 μM tropomyosin Cdc8 dimer ( Cy5-labeled ) . ( A ) Timelapse of an elongating actin filament ( left ) and Cdc8 loading and spreading events ( right ) . Scale bar , 5 μm . ( B ) Kymograph of the elongating actin filament and associated Cdc8 events . The first Cdc8 cable loading event ( i ) , second Cdc8 cable loading event ( ii ) , and Cdc8 cable spreading event ( iii ) are boxed . Scale bar , 5 μm . Time bar , 30 s . ( C ) Fluorescent images and corresponding fluorescence intensity line scans of actin ( green ) and Cdc8 ( magenta ) from the boxed regions in ( B ) . Scale bar , 1 μm . ( Ci ) A single Cdc8 cable on an actin filament segment . The dotted line ( 1 ) marks one Cdc8 cable on the actin filament segment . ( Cii ) A second Cdc8 cable loading event on an actin filament . Dotted ( 1 ) and solid ( 2 ) lines mark the first and second tropomyosin cables on the actin filament segment . ( Ciii ) A spreading Cdc8 cable . Arrows denote spreading direction of first ( dotted line , 1 ) and second ( solid line ( 2 ) Cdc8 cables . Scale bars , 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 00710 . 7554/eLife . 23152 . 008Figure 3 . Quantification of loading and spreading of first and second Tropomyosin Cdc8 cables . Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 ) with 1 . 25 μM tropomyosin Cdc8 dimer ( Cy5-labeled ) . ( A ) Depiction of potential sites for the first Cdc8 loading event . The actin filament and Cdc8 molecule are depicted by green and purple lines , respectively . ( B ) Plot of the first Cdc8 association event ( purple circles ) on actin filaments ( black lines ) , with F-actin pointed ends ( P ) aligned at the left . n = 82 events . ( C ) Depiction of Cdc8 cable spreading toward the barbed ( B ) and pointed ( P ) ends of actin filaments . ( D ) Spreading rates of Cdc8 cables toward the barbed or pointed end . Purple line denotes mean . Two-tailed t-test for data sets with equal variance yielded p-value *p=0 . 037 . n > 12 elongation events . ( E ) Depiction of site of second Cdc8 loading event , which can occur at a naked actin site ( top cartoon ) or across from the first-bound Cdc8 cable ( bottom cartoon ) . The larger percentage of the F-actin surface coated by the initial Cdc8 cable ( 1 ) increases the probability that the second Cdc8 cable ( 2 ) will associate across from a site already bound by Cdc8 . ( F ) Plot of the fraction of second Cdc8 events that associate with a F-actin site already coated by Cdc8 , binned by percentage of F-actin already coated by Cdc8 ( light purple bars ) . n = 38 events . Modeling of predicted degree of association given no indirect cooperativity ( c = 1X , purple circles ) , positive indirect cooperativity ( c = 2X , dark purple circles ) , and negative indirect cooperativity ( c = 0 . 5X , light purple circles ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 00810 . 7554/eLife . 23152 . 009Figure 3—figure supplement 1 . Tropomyosin Cdc8 first-binding events are consistent with random binding . ( A–B ) Modeling of random binding compared to experimental data of first tropomyosin Cdc8 binding events . Experimental TIRFM data ( purple filled circles ) , theoretical fit of random binding ( purple line ) and random binding model with generated noise ( purple open square ) . ( A ) ( Top ) Depiction of actin filament age at time of initial Cdc8 binding . ( Bottom ) Analysis and modeling of age of the actin filament at time of initial Cdc8 binding . ( B ) ( Top ) Depiction of age of site first bound by Cdc8 at time of Cdc8 binding . ( Bottom ) Analysis and modeling of age of Cdc8-bound site at the time of binding . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 009 Examination of timelapse images ( Figure 2A ) and kymographs ( Figure 2B ) of an elongating actin filament and associated tropomyosin Cdc8 molecules revealed the loading and spreading behavior of Cdc8 on F-actin . Cdc8 loading is complex and characterized by several initial ‘seed’ events , followed by Cdc8 cable extension along the actin filament toward both the barbed and pointed ends , interrupted by frequent stops and starts ( Figure 2A–B ) . Sites of initial Cdc8 seed association were identified by an increase in Cdc8 fluorescence at the site of binding ( Figure 2Ci ) . Unlike Drosophila tropomyosin Tm1A ( Hsiao et al . , 2015 ) , which favors association near the actin filament pointed end , initial Cdc8 loading events had no preference for the barbed or pointed end of the actin filament , with binding patterns consistent with stochastic association ( Figure 3A–B , Figure 3—figure supplement 1 ) . Once an initial Cdc8 site was initiated , spreading of Cdc8 occurred toward both the barbed and pointed ends at similar rates ( 3 . 2 and 2 . 4 Cdc8 molecules sec−1 μM−1 , respectively , p-value=0 . 037 ) . However , there was considerable variation in spreading rates in both directions , between ~1–6 Cdc8 molecules sec−1 μM−1 ( Figure 3C–D ) . As the majority of Cdc8 molecules associated adjacent to a previously-bound Cdc8 , these findings suggest that end-to-end binding is one key feature contributing to the high cooperativity of tropomyosin . A cryo-electron microscopy structure of mammalian tropomyosin , as well as negative-stain electron microscopy of S . pombe Cdc8 has shown that a single actin filament accommodates two tropomyosin cables , one on each side of the helical actin filament ( Hanson and Lowy , 1963; Moore et al . , 1970; Skoumpla et al . , 2007 ) . In our TIRFM assays , we observed the loading of these two distinct Cdc8 cables on the same actin filament ( Figure 2Cii–iii , Video 2 ) . A second Cdc8 cable loading event was evident by a doubling in fluorescence intensity at sites of previously loaded Cdc8 cable ( Figure 2C , right panels ) . The single pixel resolution in our TIRFM experiments is 100 nm , or a continuous stretch of ~5 Cdc8 molecules bound end-to-end . We assume that Cdc8 spreading events are the result of end-to-end binding between Cdc8 molecules . However , this resolution does not allow us to rule out the possibility of unconnected Cdc8 molecules binding near , but not immediately next to , previously loaded Cdc8 molecules . The doubling in Cdc8 fluorescence intensity upon association of the second Cdc8 cable allows us to explore the possibility of actin filament-mediated cooperative Cdc8 interactions . 10 . 7554/eLife . 23152 . 010Video 2 . Two distinct cables of Tropomyosin Cdc8 load onto a single actin filament , related to Figure 2 . Two-color TIRF microscopy of 1 . 5 μM actin ( Alexa-488 labeled ) with 1 . 25 μM tropomyosin Cdc8 ( Cy5-labeled ) . Green arrowhead indicates elongating F-actin barbed end . Purple arrowhead indicates the loading and spreading of Cdc8 cables . Scale bar , 5 μm . Time in sec . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 010 We investigated whether the two distinct Cdc8 cables load independently , or whether there is a loading bias of the second Cdc8 cable to regions occupied by the first Cdc8 cable . In an initial Cdc8 loading event , a Cdc8 ‘seed’ forms and elongates along one face of the actin filament . Second Cdc8 loading events can occur on the actin face opposite the first Cdc8 cable or on either face at a stretch of the actin filament free of Cdc8 ( Figure 3E ) . As expected , if the first Cdc8 cable covered a higher fraction of the actin filament face , the second Cdc8 binding event was more likely to occur across from it ( Figure 3F , purple bars ) . However , comparisons to a model of stochastic binding suggested that the observed binding opposite a first Cdc8 cable was higher than what was predicted ( Figure 3F , c = 1X circles ) . The addition of an indirect cooperativity factor of 2 ( doubling the likelihood of binding opposite a first Cdc8 cable ) more closely replicated the data ( Figure 3F , c = 2X circles ) . These findings suggest that although initial Cdc8 molecules bind stochastically on the actin filament , there is a bias for subsequent Cdc8 molecules to bind opposite previously-established Cdc8 stretches , indicating a potential role for indirect cooperativity ( Figure 1C ) . The preference of second Cdc8 loading events for already Cdc8-occupied regions described above is based on a statistical calculation from a limited number of second Cdc8 binding events ( n = 37 ) . To further address whether indirect cooperativity plays a role in tropomyosin Cdc8 loading , we created two variations of a lattice simulation describing Cdc8 loading onto a growing actin filament ( Figure 4 ) . The first model described Cdc8 loading with only end-to-end cooperativity ( w ) ( Figure 4Bi ) , while the second described Cdc8 loading with both end-to-end ( w ) and indirect cooperativity ( c ) across the actin filament ( Figure 4Ci ) . We sought to determine whether either model was sufficient to reproduce the complex behavior observed experimentally ( Figure 4Ai–iii ) . To select values of w , c and koff , we first fixed c , and varied koff and w until generating ( 1 ) the best match to the experimental data in Figure 1B as kon is varied , ( 2 ) a similar koff to single molecule experimental measurements ( Figure 4—figure supplement 1 ) , and ( 3 ) a similar initial Cdc8 binding time to experimental measurements ( Figure 3—figure supplement 1 ) . In the first model , cooperativity occurred only via end-to-end binding ( w ) ( Figure 4Bi ) . A kymograph generated using those parameters replicated many of the observed Cdc8 loading characteristics , specifically the frequent starts and stops and variation in spreading rate ( Figure 4Bii ) . However , a much lower fraction of F-actin was doubly coated with Cdc8 in the model ( Figure 4Biii ) compared to the experimental data ( Figure 4Aiii ) . Therefore , end-to-end cooperative binding alone does not fully account for the experimentally observed Cdc8 loading behavior . We therefore added a factor of positive indirect cooperativity ( c ) to the model ( Figure 4Ci ) . An indirect cooperativity ( c ) value of 1 . 25 most closely replicated our experimental data ( Figure 4C ) . This model replicated the observed dynamics of Cdc8 loading ( Figure 4Cii ) , and also replicated the observed bias towards double-coating of actin filaments by Cdc8 ( Figure 4Ciii , compare to Figure 4Aiii ) . These findings , along with the second Cdc8 binding event data ( Figure 3F ) , are consistent with a role for indirect cooperativity in the high overall cooperativity of Cdc8 . 10 . 7554/eLife . 23152 . 011Figure 4 . Modeling of Tropomyosin Cdc8 dynamics on growing actin filaments . ( Ai-iii ) Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 ) with 1 . 25 μM tropomyosin Cdc8 dimer ( Cy5-labeled ) . ( Ai-ii ) Timelapse and corresponding kymograph of Cdc8 loading and spreading . The green line indicates the actin filament barbed end . Scale bar , 5 μm . ( Aiii ) Quantification of the fraction of F-actin coated by one ( Single , checkered purple ) or two ( Double , solid purple ) Cdc8 cables . Total coverage ( purple line ) is from initial quantification in Figure 1B . Hill=14 . 6 . n = 2 reactions . ( Bi-iii ) Modeling of Cdc8 association with an actin filament with exclusively end-to-end interactions . ( Bi ) Lattice model schematic with parameters for actin elongation ( vgrow ) , rates of association ( kon , ( a ) ) or dissociation ( koff , ( b ) ) of single Cdc8 molecules with the actin filament , and rates of association ( kon*w ( c ) , kon*w2 ) and dissociation ( koff/w ( d ) , koff/w2 ( e ) ) at sites within a Cdc8 cable . ( Bii ) Kymograph of simulated loading and spreading of modeled Cdc8 under parameters in ( Bi ) . The green line indicates the actin filament barbed end . ( Biii ) Quantification of simulated data from end-to-end cooperativity model . Hill=14 . 9 . ( Ci-iii ) Modeling of Cdc8 association with an actin filament that includes both end-to-end interactions and indirect cooperativity . ( Ci ) Schematic of lattice model , which includes all parameters from ( B ) as well as additional parameters added for ( C ) : rates of association ( kon*c , ( a ) ) and dissociation ( koff/c , ( b ) ) of Cdc8 molecules across from a site already bound by a Cdc8 molecule , and rates of association ( koff*cw ) and dissociation ( koff/cw ( c ) ) of Cdc8 molecules within a cable and across from an already-bound Cdc8 . ( Cii ) Kymograph of simulated loading and spreading of modeled Cdc8 under parameters in ( Ci ) . ( Ciii ) Quantification of simulated data from end-to-end with indirect cooperativity model . Hill=13 . 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 01110 . 7554/eLife . 23152 . 012Figure 4—figure supplement 1 . Residence time of tropomyosin Cdc8 on actin filaments . Three-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488-labeled ) and 2 . 5 μM tropomyosin Cdc8 ( 0 . 5% TMR-labeled and 30% Cy5-labeled ) . Residence time of single Cdc8-TMR molecules on F-actin was quantified in the presence of Cdc8-Cy5 in order to determine that full Cdc8 coverage of F-actin occurred . n = 102 events . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 012 Tropomyosin Cdc8’s high cooperativity allows it to rapidly coat actin filaments . This property could be physiologically important as it allows tropomyosins to quickly define the functional composition of an F-actin network by associating with F-actin and influencing the subsequent association of other ABPs ( Gunning et al . , 2015; Johnson et al . , 2014; Tojkander et al . , 2011 ) . In fission yeast , Cdc8 inhibits several actin patch ABPs from binding F-actin , including ADF/cofilin Adf1 and myosin-I Myo1 ( Clayton et al . , 2010; Skau and Kovar , 2010 ) . Though Cdc8's presence at the contractile ring and actin cables likely prevents the unwanted association of several types of ABPs at those F-actin networks , opposing mechanisms must be in place to prevent Cdc8 from associating with actin patches . Cdc8’s association with actin patches is regulated by competition with the F-actin crosslinking protein fimbrin Fim1 ( Skau and Kovar , 2010 ) , which localizes predominantly to actin patches in fission yeast ( Nakano et al . , 2001; Wu et al . , 2001 ) . Though Cdc8 does not associate with actin patches normally , Cdc8 localizes to actin patches in the absence of fimbrin ( fim1-1Δ cells ) ( Skau and Kovar , 2010 ) . In addition , Fim1 prevents Cdc8 from binding F-actin in steady state in vitro bulk sedimentation assays ( Skau and Kovar , 2010 ) . However , the mechanism by which Fim1 prevents the association of Cdc8 with F-actin is unclear . We suspected that Fim1 could inhibit the initial ability of Cdc8 to load onto F-actin . Alternatively , Fim1 could actively facilitate the dissociation of Cdc8 from actin filaments . To differentiate between these and other potential mechanisms , we utilized three-color TIRFM to follow the assembly of labeled actin in the presence of labeled Fim1 and Cdc8 to simultaneously observe their associations with F-actin ( Figure 5 ) . Cdc8 initially coated individual actin filaments ( Figure 5B , first panel ) , with loading and spreading patterns similar to those observed in the absence of Fim1 . However , as actin filaments became bundled by Fim1 ( Figure 5A , C ) , the amount of Fim1 associated with the actin filaments increased and Cdc8 was actively removed from the bundled region ( Figure 5B–D , Video 3 ) . We observed Cdc8 displacement to some extent from every Fim1-mediated bundle formed . Kymographs of elongating actin filaments bundled by Fim1 showed that Cdc8 displacement occurred in a cooperative manner at the bundled site , where Cdc8 cables appeared to be ‘stripped’ away from initial sites of dissociation ( Figure 5E ) . Fim1 fluorescence intensity increased at F-actin bundles compared to single actin filaments ( Figure 5F ) , while Cdc8 intensity decreased at F-actin bundles ( Figure 5G ) , indicating that Cdc8 displacement by Fim1 occurred preferentially at sites of F-actin bundling ( Figure 5G ) . These findings demonstrate that Fim1-mediated bundling displaces Cdc8 from the actin network . Additionally , as Cdc8 appears to be rapidly ‘stripped’ from a few initial dissociation points , it is possible that the cooperativity of Cdc8 assists not only its assembly onto filaments but also its rapid displacement specifically from F-actin networks bundled by Fim1 . 10 . 7554/eLife . 23152 . 013Figure 5 . Fimbrin Fim1-mediated bundling induces cooperative removal of Tropomyosin Cdc8 . ( A–E ) Timelapse of three-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488-labeled ) with 2 . 5 μM tropomyosin Cdc8 ( Cy5-labeled ) and 250 nM fimbrin Fim1 ( TMR-labeled ) . Arrowheads and dotted line mark actin filament barbed ends and bundled region , respectively . ( E ) Kymographs of actin , Fim1 , and Cdc8 during bundle formation . Dotted lines indicate the bundled region . Scale bars , 5 μm . Time bar , 30 s . ( F–G ) Box plots of the amount of Fim1-TMR ( F ) or Cdc8-Cy5 ( G ) fluorescence on single actin filaments , two-filament bundles , or bundles containing more than two filaments . Open circles indicate outliers . Two-tailed t-tests for data sets with unequal variance yielded p-values: *p-value=4 . 67 × 10−10 , **p-value=4 . 74 × 10−11 , ***p-value=1 . 38 × 10−5 , †p-value=4 . 52 × 10−7 , ††p-value=1 . 16 × 10−12 , †††p-value=0 . 01 . n > 42 measurements . ( H ) Cartoon model of how Fim1 and Cdc8 affect each other’s association with single and bundled actin filaments . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 01310 . 7554/eLife . 23152 . 014Video 3 . Fimbrin Fim1 actively displaces Tropomyosin Cdc8 from F-actin bundles , related to Figure 5 . Three-color TIRF microscopy of 1 . 5 μM actin ( Alexa-488 labeled ) with 250 nM fimbrin Fim1 ( TMR-labeled ) and 2 . 5 μM tropomyosin Cdc8 ( Cy5-labeled ) . Arrowheads indicating two distinct F-actin elongating barbed ends . Scale bar , 5 μm . Time in sec . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 014 ADF/cofilin Adf1 is an F-actin severing protein that localizes predominantly to actin patches and is required for proper actin patch dynamics ( Nakano et al . , 2001; Nakano and Mabuchi , 2006 ) , but also localizes to the contractile ring and is critical for its assembly ( Chen and Pollard , 2011 ) . Adf1 is an important F-actin network disassembly factor that severs actin filaments and allows recycling of the actin monomers ( Andrianantoandro and Pollard , 2006; Chen and Pollard , 2013 ) . Competition between tropomyosins and ADF/cofilins in many systems has been well-established ( Bernstein and Bamburg , 1982; DesMarais et al . , 2002; Kuhn and Bamburg , 2008; Ono and Ono , 2002 ) . In fission yeast , Adf1-mediated severing is decreased in the presence of Cdc8 ( Skau and Kovar , 2010 ) . We used multi-color TIRFM to investigate whether Cdc8 inhibits Adf1-mediated severing by decreasing the initial association of Adf1 with F-actin , or by another mechanism . Wild-type Adf1 labeled with a cysteine-reactive dye was not observed to associate with F-actin ( data not shown ) . We therefore engineered an Adf1 labeling mutant ( C12S , C62A , D34C ) for use in TIRFM , based on a similar strategy used for budding yeast ADF/cofilin Cof1 ( Figure 6—figure supplement 1 , Methods ) ( Suarez et al . , 2011 ) . The TMR-labeled Adf1 mutant is less active than wild-type Adf1 ( Figure 6—figure supplement 1A ) , but was observed to bind F-actin in a cooperative manner and sever actin filaments at boundaries of Adf1 bound/unbound regions ( Figure 6—figure supplement 1B and C ) , characteristics previously observed for ADF/cofilins from a variety of organisms including fission yeast ( Andrianantoandro and Pollard , 2006; Hayakawa et al . , 2014; De La Cruz , 2009; Michelot et al . , 2007 ) . At high Adf1 concentrations ( 5 μM ) , the majority of F-actin was rapidly coated with Adf1 ( Figure 6A , Figure 6—figure supplement 1B ) . However , in the presence of Cdc8 , little Adf1 was observed to initially associate with actin filaments ( Figure 6A–C , Video 4 ) . Over time , Adf1 began to load near the pointed end of Cdc8-associated actin filaments in small ‘clusters’ that then spread cooperatively along the filament ( Figure 6B ) . Importantly , though Cdc8 affected the initial association ( kon ) of Adf1 with actin filaments , the dissociation rate ( koff ) of Adf1 from F-actin was unaffected by Cdc8 ( Figure 6D ) . These findings are supported by three-color TIRFM imaging of the assembly of labeled actin in the presence of labeled Adf1 and Cdc8 . Labeled Cdc8 and Adf1 show mutually exclusive localization on actin filaments ( Figure 6E ) . Initially , the majority of actin filaments are coated with Cdc8 . Over time , Adf1 puncta arise and displace Cdc8 from those regions . Adf1 domains then spread in a cooperative manner , further displacing Cdc8 ( Figure 6F ) . Cdc8 continues to associate with the elongating actin filament barbed end , where Adf1 is not yet associated , but is ultimately displaced by Adf1 over time . As Adf1 preferentially associates with ADP-F-actin ( Andrianantoandro and Pollard , 2006 ) , we suspect that the transition from ADP-Pi- to ADP-actin increases the kon of Adf1 for ADP-bound F-actin , allowing a few Adf1 molecules to associate with the actin filament . Adf1 then cooperatively spreads from these regions of association , and displaces Cdc8 . Collectively , these data suggest that while Cdc8 inhibits the initial binding of Adf1 to actin filaments , Adf1 is capable of associating and ultimately displacing segments of Cdc8 ( Figure 6G ) . 10 . 7554/eLife . 23152 . 015Figure 6 . Tropomyosin Cdc8 inhibits initial binding of ADF/cofilin Adf1 to actin filaments . ( A–C ) Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 labeled ) with 5 μM ADF/cofilin Adf1 ( TMR-labeled ) in the presence or absence of 2 . 5 μM tropomyosin Cdc8 ( unlabeled ) . ( A ) Micrograph of Adf1 association with actin filaments in the absence or presence of Cdc8 . ( B ) Kymograph of an elongating actin filament ( top ) and associated Adf1 ( bottom ) in the absence or presence of unlabeled Cdc8 . Scale bar , 5 μm . Time bar , 30 s . ( C ) Box plot of average Adf1 fluorescence intensity on actin filaments in the absence or presence of unlabeled Cdc8 . Open circles indicate outliers . Two-tailed t-test for data sets with unequal variance yielded *p-value: 6 . 77 × 10−82 . n ≥ 145 measurements . ( D ) Residence time of single Adf1 ( TMR-labeled ) molecules in the absence or presence of unlabeled Cdc8 . n ≥ 81 events . ( E–F ) Three-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 labeled ) with 5 μM Adf1 ( TMR-labeled ) and 2 . 5 μM Cdc8 ( Cy5-labeled ) . ( E ) Micrograph of actin filaments associated with Adf1 and Cdc8 . Scale bar , 5 μm . ( F ) Timelapse of Adf1 association with an actin filament , and subsequent dissociation of Cdc8 . Scale bar , 5 μm . Time in sec . ( G ) Cartoon model of how Cdc8 and Adf1 affect each other’s association with F-actin . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 01510 . 7554/eLife . 23152 . 016Figure 6—figure supplement 1 . Characterization of Cofilin Adf1 labeling mutant . ( A ) Spontaneous assembly assay of 1 . 5 μM actin ( 10% pyrene-labeled ) and 50 or 500 nM of WT Adf1 ( green lines ) or Adf1 labeling mutant ( blue lines ) . ( C ) Actin filament severing occurs at Adf1 boundaries . ( B ) Two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 ) with a range of concentrations of cofilin Adf1 mutant ( C12S , C62A , D34C ) ( TMR-labeled ) . Scale bar , 5 μm . ( C ) ( Top ) Two-color TIRF microscopy of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 ) 500 nM Adf1 mutant ( C12S , C62A , D34C ) ( TMR-labeled ) . Yellow arrowhead and asterisks indicate future sites of severing and severing events , respectively . Scale bar , 5 μm . ( Bottom ) Linescan of Adf1 fluorescence ( purple ) along actin filament before severing . Yellow dotted line indicates sites of severing . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 01610 . 7554/eLife . 23152 . 017Video 4 . Tropomyosin Cdc8 prevents initial association of Cofilin Adf1 with F-actin . Two-color TIRF microscopy of 1 . 5 μM actin ( Alexa-488 labeled ) with 5 μM cofilin Adf1 ( TMR-labeled ) . Scale bar , 5 μm . Time in sec . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 017 Within an actin patch , a dense array of Arp2/3 complex-mediated branched F-actin provides the network architecture that propels newly-generated endocytic vesicles inward ( Collins et al . , 2011; Young et al . , 2004 ) . We previously postulated that fimbrin Fim1 may be required for proper actin patch motility indirectly , by inhibiting tropomyosin Cdc8 association and therefore allowing ADF/cofilin Adf1-mediated severing to occur ( Skau and Kovar , 2010 ) . However , a separation of function Fim1 mutant that can bind but not bundle F-actin , and retains the ability to displace Cdc8 from F-actin , still shows abnormal actin patch dynamics ( Skau et al . , 2011 ) . Therefore Fim1 has additional roles aside from protecting actin patches from Cdc8 . Therefore , we examined the relationship between Fim1 and Adf1 utilizing multi-color TIRFM with labeled proteins ( Figure 7 ) . Unlike the mutually-exclusive localization of Cdc8 and Adf1 ( Figure 6E ) , Fim1 and Adf1 co-localized at both single actin filaments and multi-filament bundles throughout the TIRF chamber ( Figure 7B ) . However , distinct Fim1 or Adf1 domains could still be observed , and severing often occurred at these interfaces ( Figure 7C , yellow line ) , likely because of sharp changes in Adf1 density and/or actin filament flexibility between those two regions ( Elam et al . , 2013; McCullough et al . , 2008; Suarez et al . , 2011 ) . Consistent with this finding , the Adf1 severing rate was increased in the presence of Fim1 ( Figure 7F ) . As the severing rate could only be directly observed on single actin filaments and two-filament bundles , our measured severing rate in the presence of Fim1 is likely underreported ( Figure 7F , indicated by # ) . Severed filaments were quickly incorporated into nearby bundles , resulting in an increase in Fim1-mediated bundling in the presence of Adf1 ( Figure 7D , Video 5 ) . Fim1-mediated bundles became extremely large and dense in the presence of Adf1 ( compare Figure 7B to 7A ) , as each severing event resulted in the formation of a new elongating barbed end . This rapid generation of barbed ends yielded a ~11 fold change in actin fluorescence after 300 s ( Figure 7E , solid lines ) , compared to a ~2 fold change in experiments lacking Adf1 ( Figure 7E , dashed lines ) . Together , these findings support a potential role for Adf1 as both a disassembly factor ( via severing ) and an assembly factor ( via generation of new barbed ends ) . We speculate that the presence of Fim1 on actin patches may be important not only for exclusion of Cdc8 from the network but also for enhancement of Adf1-mediated severing via generation of single filament/bundle interfaces ( Figure 7G ) . 10 . 7554/eLife . 23152 . 018Figure 7 . Fimbrin Fim1 and ADF/cofilin Adf1 competition generates a dense F-actin network . ( A ) Timelapse of two-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488-labeled ) with 500 nM fimbrin Fim1 ( Cy5-labeled ) . ( B ) Timelapse of three-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488-labeled ) with 500 nM Fim1 ( Cy5-labeled ) and 5 μM ADF/cofilin Adf1 . ( C ) Timelapse showing severing at the boundary between a Fim1-mediated bundle and a single actin filament . White arrow indicates the elongating actin filament barbed end . A yellow line and asterisk indicate the severing site and severing event , respectively . ( D ) Quantification of percent of total actin filaments bundled with Fim1 alone or Fim1 and Adf1 . Error bars represent standard deviation of the mean; n = 2 reactions . ( E ) Fold-change over time in total fluorescence intensity for either actin ( dotted green line ) and Fim1 ( dotted red line ) in the absence of Adf1 , or for actin ( solid green line ) , Fim1 ( solid red line ) , and Adf1 ( solid blue line ) in the presence of Adf1 . ( F ) Severing rate of high concentrations of Adf1 alone or in the presence of Fim1 . # indicates under-reporting , as severing events could not be measured on dense bundles . Error bars represent standard deviation of the mean; n = 2 reactions . ( G ) Cartoon model of how Adf1 and Fim1 influence each other’s interactions with actin and affect F-actin network formation . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 01810 . 7554/eLife . 23152 . 019Video 5 . Fimbrin Fim1 and Cofilin Adf1 synergize to generate a dense F-actin network , related to Figure 7 . Three-color TIRF microscopy of 1 . 5 μM actin ( Alexa-488 labeled ) with 500 nM fimbrin Fim1 ( Cy5-labeled ) and 5 μM cofilin Adf1 ( TMR-labeled ) . Arrowheads indicate sites of severing . Scale bar , 5 μm . Time in sec . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 019 Our findings and previous work have suggested that the synergistic activities of fimbrin Fim1 and ADF/cofilin Adf1 may serve to inhibit the association of tropomyosin Cdc8 with F-actin ( Skau and Kovar , 2010 ) . We tested this possibility using four-color TIRFM to examine how Fim1 , Adf1 , and Cdc8 collectively affect each other’s interactions with F-actin ( Figure 8 ) . As in the previous experiments with only Fim1 and Adf1 ( Figure 7 ) , a dense , bundled F-actin network was formed , containing filaments coated with both Adf1 and Fim1 ( Figure 8A ) . Cdc8 , on the other hand , was rapidly dissociated from nearly every actin filament in the chamber ( Figure 8A–C , Video 6 ) . Closer observation revealed individual growing actin filaments initially coated with Cdc8 ( Figure 8D , right panel ) . These filaments are often severed at single/bundled actin filament boundaries ( Figure 8Di , yellow line ) , and the severing events created new actin filament barbed ends that were encompassed into F-actin bundles , where Cdc8 was displaced by Fim1 ( Figure 8Dii ) . We did not observe Adf1 directly displacing Cdc8 , and therefore we suspect that Adf1’s primary role in displacing Cdc8 is by rapidly generating new actin filament barbed ends that are encompassed into F-actin bundles mediated by Fim1 ( Figure 8E ) . 10 . 7554/eLife . 23152 . 020Figure 8 . Competitive interactions between Cofilin Adf1 and Fimbrin Fim1 result in rapid displacement of Tropomyosin Cdc8 from F-actin networks . ( A–D ) Four-color TIRFM of 1 . 5 μM Mg-ATP actin ( 15% Alexa 488 labeled ) with 2 . 5 μM tropomyosin Cdc8 ( Cy5-labeled ) , 500 nM fimbrin Fim1 ( TMR-labeled ) , and 5 μM cofilin Adf1 ( Alexa 405-labeled ) . ( A ) Timelapse of F-actin network generation in the presence of actin ( green ) , Fim1 ( red ) , Adf1 ( cyan ) , and Cdc8 ( magenta ) . ( B ) Fold-change over time in fluorescence intensity of actin ( green line ) , Fim1 ( red line ) , Adf1 ( blue line ) , and Cdc8 ( purple line ) . ( C ) Fold change in fluorescence intensity of each ABP after 300 s . Error bars represent standard deviation of the mean . n = 2 independent experiments . ( D ) Enlargement of the area within the dotted box in ( A ) . Fim1 and Adf1 synergize to create a dense F-actin bundle ( white dotted line ) , while Cdc8 associates with a single actin filament . Severing occurs at the boundary of the single actin filament and F-actin bundle ( i , yellow asterisk ) , creating a new elongating F-actin barbed end . The bundle extends to incorporate the single actin filament and Cdc8 is displaced ( ii ) . Scale bar , 5 μm . ( E ) Cartoon model of events occurring in ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 02010 . 7554/eLife . 23152 . 021Video 6 . Fimbrin Fim1 and Cofilin Adf1 work together to displace Tropomyosin Cdc8 from the F-actin network . Four-color TIRF microscopy of 1 . 5 μM actin ( Alexa 488-labeled ) with 500 nM fimbrin Fim1 ( TMR-labeled ) , 2 . 5 μM tropomyosin Cdc8 ( Cy5-labeled ) and 5 μM cofilin Adf1 ( Alexa 405-labeled ) . Scale bar , 5 μm . Time in sec . DOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 021
Our work demonstrates that cooperativity is an important characteristic for multiple aspects of F-actin network formation , organization , and ABP sorting . We show that fission yeast tropomyosin Cdc8 binds extremely cooperatively to F-actin ( Figure 1 ) . Though tropomyosins from other organisms have been labeled for use in TIRFM ( Hsiao et al . , 2015; Schmidt et al . , 2015 ) , ours is the first study to observe two distinct tropomyosin cables associating with a single actin filament ( Figure 2 ) . This high degree of resolution has provided us the unique opportunity to differentiate between the contributions of ( 1 ) end-to-end association and ( 2 ) indirect interactions via the actin filament in tropomyosin's cooperative binding to F-actin ( Figure 4 ) . End-to-end binding has been considered to be the primary source of tropomyosin’s cooperativity . Mutations in the N- or C-terminal domains of Cdc8 affect its ability to bind cooperatively and polymerize on an actin filament ( East et al . , 2011 ) . However , it has been suggested that other interactions between tropomyosin and the actin filament may also be involved in tropomyosin cooperativity as end-to-end attachments between tropomyosin molecules tend to be rather weak ( Sousa and Farah , 2002 ) and muscle tropomyosin with impaired ability to associate end-to-end is still cooperative on actin filaments ( Willadsen et al . , 1992; Tobacman , 2008 ) . Our findings suggest that , though end-to-end binding is a key factor in tropomyosin cooperativity , indirect interactions via the actin filament may also be important for enhancing tropomyosin coating of F-actin . Our results suggest that a small increase in indirect cooperativity ( increase in binding affinity of Cdc8 by a factor of ~2 ) is sufficient to account for Cdc8's high cooperativity and observed loading characteristics ( Figure 4C ) . The precise interactions within the actin filament that allow for this cooperativity remain to be determined . However , due to the logarithmic relationship between free energy differences and equilibrium constants , a factor of 2 increase in binding affinity corresponds to a very small increase in free-energy stabilization of -kBTln ( 2 ) =~0 . 4 kcal/mol . Hence , if this indirect cooperativity is due to a structural change in the actin filament , we expect that change to be quite subtle . Tropomyosin is known to increase the persistence length of F-actin ( Fujime and Ishiwata , 1971; Isambert et al . , 1995 ) . As a result , the stabilizing effect of an initial Cdc8 binding to one side of the actin filament may favor binding of a second Cdc8 to the opposing side . Additionally , if the rigidity provided by initial Cdc8 binding is propagated slightly further up or down the actin filament , more tropomyosin molecules could associate with the actin filament as a result of an increase in F-actin rigidity . More complex mammalian cells express >40 tropomyosin isoforms that vary in affinity for F-actin and degree of cooperativity ( Pittenger et al . , 1994; Schevzov et al . , 2011 ) , suggesting that these characteristics may differentially affect their distinct cellular roles and ability to sort to different F-actin networks ( Gunning et al . , 2015 ) . How does this high degree of cooperativity affect tropomyosin’s influence on the activity of other ABPs ? The nature of tropomyosin as a strand-like , end-to-end associated protein makes it an ideal F-actin ‘gatekeeper’ ( Gunning et al . , 2008 , 2015 ) . Individual tropomyosin Cdc8 molecules associate poorly with actin filaments . However , once a Cdc8 ‘seed’ has been initiated , end-to-end binding and potential indirect interactions promote Cdc8’s rapid coating of F-actin . The avidity generated from multiple F-actin-Cdc8 and Cdc8-Cdc8 interactions regulates the binding of other ABPs such as ADF/cofilin ( Figure 6A–B ) . In addition , this large number of interactions between a Cdc8 cable and the actin filament could allow a Cdc8 cable to remain associated despite perturbations along the actin filament that may result from the association of other ABPs with the actin filament . However , these same cooperative characteristics also make Cdc8 easily removable from actin filaments once a threshold of perturbations has been bypassed . For example , at regions of high fimbrin Fim1 association ( F-actin bundles ) , Cdc8's end-to-end associations allow it to rapidly peel away from those actin filaments ( Figure 5B ) . In addition , the poor ability of individual Cdc8 molecules to bind to an actin filament makes it unlikely to be able to displace other ABPs once they’re bound to an actin filament , resulting in Cdc8's complete exclusion from certain F-actin networks , such as actin patches . We suspect that Fim1’s long residence time on F-actin bundles enhance its ability to create regions of high Fim1 occupancy that displace Cdc8 , suggesting that ABP dynamics ( on-/off-rate ) may partially dictate the ability of an ABP to compete . An ABP’s dynamics could be driven by its intrinsic biochemical properties ( Winkelman et al . , 2016 ) , post-translational modifications ( Miao et al . , 2016 ) , or mechanical stresses within the F-actin network ( Srivastava and Robinson , 2015 ) , resulting in the fine-tuning of ABP sorting by many factors . Another impact of Cdc8’s high cooperativity may be that only a slight bias toward certain actin filaments could generate an ‘all-or-nothing’ sorting toward those networks . Several studies have suggested that the actin assembly factor may be the key to generating this initial bias ( Kovar et al . , 2011 ) . In fission yeast , altering formin localization in the cell results in the relocalization of both acetylated and unacetylated forms of Cdc8 ( Johnson et al . , 2014 ) . This bias may be the result of formin-induced conformational changes that could be propagated down the actin filament ( Bugyi et al . , 2006; Papp et al . , 2006 ) . Future work will involve deciphering potential ‘initiating’ signals for ABP sorting and their effects on ABP cooperativity and competition . Our work also addresses the importance of ABP competition in network organization and ABP sorting . An outstanding question is how ADF/cofilin , a protein that severs at low concentrations , can be present at high concentrations in the cell and yet still rapidly disassemble F-actin networks . Other ABPs , such as Aip1 ( Gressin et al . , 2015; Nadkarni and Brieher , 2014 ) , Coronin ( Jansen et al . , 2015 ) , and Twinfilin ( Johnston et al . , 2015 ) have been found to enhance ADF/cofilin-mediated severing via multiple mechanisms , including potential side-binding by Aip1 ( Chen et al . , 2015 ) . Our study additionally implicates side-binding ABPs not involved in severing ( fimbrin Fim1 ) as important for the enhancement of ADF/cofilin Adf1 severing , likely by generating ADF/cofilin boundaries or changes in flexibility that result from Fim1-mediated bundle formation ( Figure 7C ) . The idea of competition between ADF/cofilin and other factors as important for enhancing ADF/cofilin-mediated severing is an exciting area of investigation ( Elam et al . , 2013 ) , and future work remains to determine how different ABPs regulate ADF/cofilin severing to different extents on different F-actin networks . In addition , our study suggests an important role for ADF/cofilin not only as a disassembly factor , but also as a potent F-actin network assembly factor through the creation of new barbed ends as in the model of dendritic nucleation within the lamellipodia of migrating cells ( DesMarais et al . , 2004; Ghosh et al . , 2004; Ichetovkin et al . , 2002 ) . The combination of Fim1 and Adf1 generates dense F-actin bundles composed of many more actin filaments than Fim1 alone ( Figure 7A–B ) ( Skau et al . , 2011 ) It was previously shown that expression of an ADF/cofilin mutant deficient in severing results in delayed actin patch assembly in fission yeast , as a lack of severing prevents the creation of new mother filaments for actin patch initiation ( Chen and Pollard , 2013 ) . Our findings also potentially implicate Adf1 involvement in increasing the number of barbed ends within the dense actin patch network . The creation of barbed ends via severing takes advantage of the inherent polarity of an actin filament , as the newly created barbed end at the severed site is automatically oriented in the same direction as the original barbed end . This type of mechanism could be ideal for force-generating networks such as those at endocytic actin patches or the lamellipodia of migrating cells ( Ghosh et al . , 2004 ) . However , the proper balance of assembly vs . disassembly must be achieved for proper actin patch dynamics , and likely involves the concerted effort of many ABPs . Finally , our work highlights that the collective efforts of multiple ABPs can enhance or modulate the effects of individual ABPs on F-actin . There are many examples of multiple ABPs working together to create F-actin networks of defined organization and dynamics , including those within Listeria comet tails ( Loisel et al . , 1999 ) , the leading edge of motile cells ( Blanchoin et al . , 2000 ) , contractile F-actin networks ( Ennomani et al . , 2016; Reymann et al . , 2012 ) , and disassembling actin patches ( Jansen et al . , 2015 ) . We show here that a similar idea holds true for competition-mediated ABP segregation . Tropomyosin Cdc8 is rapidly displaced from F-actin networks due to the combined activities of fimbrin Fim1 and ADF/cofilin Adf1 . Though Cdc8 alone is capable of preventing association of Adf1 with F-actin ( Figure 6 ) , Fim1 rapidly displaces Cdc8 from actin filaments ( Figure 5 ) , allowing Adf1-mediated severing to occur ( Figure 8 ) . Adf1-mediated severing feeds back on Fim1's F-actin bundling activity by creating small , severed filaments that are easily incorporated into bundles ( Figure 7 ) , facilitating further bundling by Fim1 and increased Cdc8 displacement ( Figure 8 ) . Our work has implications not only for ABP sorting , but for the formation of any organized F-actin network that contains multiple components competing for the same binding substrate . Our lab has previously shown that competition for F-actin monomer is important for regulating the density and size of F-actin networks ( Burke et al . , 2014; Suarez et al . , 2015 ) . In addition , competitive interactions between DNA methylation and surrounding transcription factors mediate transcription factor association with certain regions of the genome ( Domcke et al . , 2015 ) , and competition for membrane or receptor binding has been suggested to be involved in cargo sorting in a number of contexts ( Soza et al . , 2004 ) . Overall , our work and the work of others suggest that competitive interactions between individual components can have large-scale effects on cellular organization in many contexts .
All plasmid design and construction were performed using SnapGene software ( from GSL Biotech; available at snapgene . com ) . Three amino acids in S . pombe tropomyosin Cdc8—Leucine 38 , Isoleucine 76 , and Aspartate 142—were chosen as potential labeling sites for mutation to cysteine based on four criteria: ( 1 ) localization on the outside of the coiled coil ( at b , c , or f locations ) , ( 2 ) low sequence conservation amongst fungal tropomyosins ( Cranz-Mileva et al . , 2013 ) , ( 3 ) present outside the first half of each period and therefore unlikely to affect Cdc8's association with F-actin ( Barua et al . , 2013 ) , and ( 4 ) present away from C-terminal end so as to not affect end-to-end associations between Cdc8 molecules . To create Cdc8 mutants L38C , I76C , and D142C for protein expression , QuikChange Site-Directed Mutagenesis ( Agilent Technologies , Inc . , Santa Clara , CA ) was used to engineer distinct base pair substitutions within acetylation-mimic Cdc8 expression vector pET3a-AS-Cdc8 ( Monteiro et al . , 1994 ) , ( Clayton et al . , 2014 ) and modifications were confirmed by sequencing . In vitro high-speed sedimentation assays and preliminary TIRFM assays at high Cdc8 concentrations determined that the I76C and D142C Cdc8 mutants behaved closest to wild-type Cdc8 in ability to bind to F-actin ( Figure 1—figure supplement 1B–D ) . In addition , replacing the endogenous fission yeast cdc8 gene with each of these mutants revealed that the I76C and D142C mutant strains behaved more similarly to wild-type than the L38C mutant ( Figure 1—figure supplement 2 ) . Therefore , a labeled I76C mutant was used in TIRFM assays for the rest of the study . All three mutations were introduced within exon 2 of the cdc8 gene . For insertion of cdc8 mutants into the S . pombe genome , the portion of the cdc8 mutant corresponding to exon 2 of cdc8 was amplified from pET3a-AS-Cdc8mut protein expression vectors ( using primers AAGGCGCGCCAGATCTAAAATTAATGCCGCTCGTGCTGAG and CAAGCTAAACAGATCTCTACAAATCCTCAAGAGCTTGGTGAAC ) , and the product was cloned into gene targeting vector pFA6-kanMX6 at BglII using In-Fusion HD Cloning ( Clontech Laboratories , Mountain View , CA ) . cdc8mut-kan was then amplified ( using primers AGGTATGAGATGATAGCTTTTCATTGGAAAATCAAGTTGCTAATATTTGCTTTTTATTTAGAAAATTAATGCCGCTCGTGC and AGAAGATATAAAAAAGGTGGTATGTTTCTTCTATGTTCGTCAAGCTTTTCGCTATGAATTCGAGCTCGTTTAAAC ) and transformed into a temperature sensitive mutant cdc8-27 strain . Colonies were screened for absence of temperature sensitivity and resistance to kanamycin . Candidate colonies were then screened for proper insertion by PCR and sequenced to confirm insertion at the cdc8 locus . Strains created are listed in Table 1 . 10 . 7554/eLife . 23152 . 022Table 1 . Fission yeast strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 23152 . 022Strain nameGenotypeReferenceFY527h- , leu1-32 , his3-D1 , ura4-D18 , ade6-M216Forsburg labMBY6663h+ , pAct1 Lifeact-GFP::Leu+; ade6-m216; leu1-32; ura4-D18Huang et al . ( 2012 ) KV920h ? , cdc8-D142C::KanMX6 , ade6-m216; ura4-D18This studyKV921h ? , cdc8-I76C::KanMX6 , ade6-m216; ura4-D18This studyKV922h ? , cdc8-L38C::KanMX6 , ade6-m216; ura4-D18This studyKV969h ? cdc8-I76C::KanMX6 , pAct1 Lifeact-GFP::Leu+ , ade6-m216; ura4-D18This studyKV970h ? cdc8-D142C::KanMX6 , pAct1 Lifeact-GFP::Leu+ , ade6-m216; ura4-D18This studyKV971h ? cdc8-L38C::KanMX6 , pAct1 Lifeact-GFP::Leu+ , ade6-m216; ura4-D18This study Three mutations were made in S . pombe ADF/cofilin Adf1 for labeling in TIRFM experiments . Endogenous cysteine residues were converted to alanine ( C12S and C62A ) , and aspartate 34 was converted to cysteine ( D34C ) as previously described for ADF/cofilin from S . cerevisiae ( Suarez et al . , 2011 ) . These mutations were made by QuikChange Site-Directed Mutagenesis in expression vector pMW-SpCofilin ( Skau and Kovar , 2010 ) . Modifications were confirmed by sequencing . Fimbrin Fim1 , ADF/cofilin Adf1 ( WT and mutant D34C , C12S , C62A ) , and tropomyosin AlaSer-Cdc8 ( WT and L38C , I76C and D142C mutants ) were expressed in BL21-Codon Plus ( DE3 ) -RP ( Agilent Technologies , Santa Clara , CA ) and purified as described previously ( Skau and Kovar , 2010 ) . Briefly , Fim1 was purified with Talon Metal Affinity Resin ( Clontech , Mountain View , CA ) . Adf1 was purified by an ammonium sulfate precipitation , size exclusion chromatography , and anion exchange chromatography . Cdc8 was purified by boiling the cell lysate , ammonium sulfate precipitation , and anion exchange chromatography . Actin was purified from chicken skeletal muscle or rabbit skeletal muscle acetone powder ( Pel-Freez , Rogers , AR ) as described previously ( Spudich and Watt , 1971 ) . A280 of purified proteins was taken using a Nanodrop 2000c Spectrophotometer ( Thermo-Scientific , Waltham , MA ) . Protein concentration was calculated using extinction coefficients Fim1: 55 , 140 M−1 cm−1 , Cdc8 ( WT and mutants ) : 2980 M−1 cm−1 , Adf1 ( and mutant ) : 13 , 075 M−1 cm−1 . Proteins were labeled with CFTM405M ( Sigma-Aldrich , St . Louis , MO ) , TMR-6-maleimide ( Life Technologies , Grand Island , NY ) or Cy5-monomaleimide ( GE Healthcare , Little Chalfont , UK ) dyes as per manufacturer’s protocols immediately following purification , and were flash-frozen in liquid nitrogen and kept at −80°C . Cdc8 was reduced with DTT prior to labeling . For proteins labeled on one cysteine residue ( Cdc8 and Adf1 mutants ) , labeling efficiency was determined by taking the absorbance at the emission max of the dye and calculating the coupling efficiency ( Kim et al . , 2008 ) . All reported Cdc8 concentrations are of the two-chain ( dimer ) molecule . Coverslips and microscope slides ( #1 . 5; Fisher Scientific ) for TIRFM were prepared by washes in acetone , isopropanol , and water followed by sonication for 30 min in isopropanol . Washed glass was then cleaned by plasma cleaning for 3 min using a Harrick PDC-32G plasma cleaner ( Harrick Plasma , Ithaca , NY ) . Cleaned coverslips and microscope slides were immediately passivated by incubation in 1 mg/mL PEG-Si ( 5000 MW ) in 95% ethanol for 18 hr ( Winkelman et al . , 2014 ) . Coverslips and slides were then rinsed in ethanol and water , and flow chambers were assembled as described previously ( Zimmermann et al . , 2016 ) . Time-lapse TIRFM movies were obtained using a cellTIRF 4Line system ( Olympus ) fitted to an Olympus IX-71 microscope with through-the-objective TIRF illumination and a iXon EMCCD camera ( Andor Technology , Belfast , UK ) . Mg-ATP-actin ( 15% Alexa 488-labeled ) was mixed with labeled or unlabeled ABPs and a polymerization mix ( 10 mM imidazole ( pH 7 . 0 ) , 50 mM KCl , 1 mM MgCl2 , 1 mM EGTA , 50 mM DTT , 0 . 2 mM ATP , 50 μM CaCl2 , 15 mM glucose , 20 μg/mL catalase , 100 μg/mL glucose oxidase , and 0 . 5% ( 400 centipoise ) methylcellulose ) to induce F-actin assembly ( Winkelman et al . , 2014 ) . The mixture was then added to a flow chamber and imaged at 2 . 5 or 5 s intervals at room temperature . Line scans were performed by drawing a three pixel width line along the actin filament and recording the fluorescence intensity along the line using ‘Plot Profile’ in FIJI ( Schindelin et al . , 2012; Schneider et al . , 2012 ) . The same region of interest was then applied to the ABP channel of interest . If necessary , the region of interest was adjusted to account for filament movement during channel switching . The occupancy of Cdc8 on actin filaments was determined from one frame from each TIRFM movie . For each movie , the frame of interest was chosen based on F-actin density ( between 7 and 11 μm of filament per square μm ) rather than time point . Segmented line ROIs were used to measure total actin filament length ( 488 channel ) and total Cdc8 cable length ( 647 channel ) at that frame . History of the TIRFM movie as well as Cdc8 fluorescence intensity was used to determine sites at which two Cdc8 cables were present on a single actin filament stretch , and in these cases each Cdc8 stretch was counted as a separate measurement . As there are two Cdc8 cables on a single actin filament , Total Cdc8 Occupancy=Total Cdc8 Length/ ( Actin Filament Length*2 ) . Single vs double Cdc8 occupancy was determined by measuring the length of single- vs double-Cdc8-coated stretches and calculating the total length of single- or double-coated stretches divided by total actin filament length . Free Cdc8 was calculated by ( [Cdc8 added]−250* ( Total Cdc8 Occupancy ) /4 ) *0 . 001 ( adapted from [Hsiao et al . , 2015] ) . 250 ( nM ) refers to the F-actin concentration at the average time point that total Cdc8 occupancy was measured . The F-actin concentration was determined by a spontaneous pyrene actin assembly assay . Four refers to the ratio of bound Cdc8 to F-actin ( 1:4 ) . The data were fit to a Hill equation θ=[L]n/ ( Kd+[L]n where θ is the fraction of actin sites that are bound by Cdc8 , [L] is the free Cdc8 concentration , Kd is the apparent dissociation constant , and n is the Hill coefficient . The fluorescence intensity of fimbrin Fim1 , tropomyosin Cdc8 , or ADF/cofilin Adf1 was used to determine amount of ABP associated with actin filaments under different conditions . Analysis was performed on movie frames with a similar filament density for each compared condition . Segmented line ROIs ( line width five pixels ) were used to define each actin filament in the actin channel ( 488 channel ) . ROIs were then transferred to the ABP channel of interest and mean fluorescence intensity was measured for each actin filament . For comparing single actin filaments to F-actin bundles , the history of the actin channel and the actin fluorescence intensity was used to determine single filament versus bundled regions and separate ROI sets were generated and used to measure fluorescence intensity . To calculate residence times for individual ADF/cofilin Adf1 or tropomyosin Cdc8 molecules , a high total concentration of Adf1 or Cdc8 was included in the TIRFM assay to ensure total coating of the actin filaments ( 5 μM Adf1 or 2 . 5 μM Cdc8 ) . ~20% of Adf1 or Cdc8 was labeled with Cy5 dye in order to visualize the extent of coating of the protein on F-actin . The actin and Cy5-labeled Adf1 or Cdc8 was visualized every 10 s . A low ( 0 . 5–1% ) percentage of Adf1 or Cdc8 was labeled with TMR in order to visualize single molecules , and fast imaging ( five frames/sec ) was performed in this channel . To measure residence time of Adf1 in the presence of Cdc8 , 2 . 5 μM of unlabeled Cdc8 was included in the reaction . Single molecules were tracked using MTrackJ ( Meijering et al . , 2012 ) . Only single molecules that moved were tracked , as static molecules were assumed to be adsorbed to the coverglass . Both censored and uncensored events were obtained and a Kaplan-Meier analysis was performed ( Kaplan and Meier , 1958 ) . Events were fit to a single exponential f ( x ) =f0e−xT1 that was used to determine residence time and koff . Site of the first tropomyosin Cdc8 binding event on an actin filament was determined by observing TIRFM movies performed at Cdc8 concentrations at the inflection point of the Hill plot ( 1 . 25 μM Cdc8 ) . As our resolution limit is 100 nm , and individual Cdc8 molecules may only briefly associate with the actin filament before dissociating or forming a ‘seed’ , we cannot determine explicitly the number of Cdc8 molecules in each of these initial association events . At the point of first observation of Cdc8 binding , the length of the actin filament was measured as well as the distance from the pointed end to the site of Cdc8 binding . The barbed and pointed ends of the actin filament were identified by observing photobleaching of the older , pointed end of the actin filament that occurred over time . To compute the first binding times of a molecule to a substrate , we modeled the reaction as a master equation with two states , bound and unbound , with unbound transforming to bound at rate k12 and the reverse process occurring with rate k21 = 0 . Once a binding event is observed for a filament , that filament was unable to go back to having been never bound . We described P ( t ) as a vector representing the populations of the two states . ( 1 ) dP→dt=WP→ where W is a rate matrix whose columns must sum to zero . Hence: ( 2 ) W=[k22k12k21k11]=[0k120−k12] In the case of binding to an extending F-actin substrate , if we assumed an average constant growth rate vgrow , the length of the substrate l ( t ) = vgrowt . With the assumption of uniform binding affinity , the rate of going from unbound to singly bound k12 depended on time , therefore konl ( t ) = konvgrowt . kon is the rate per unit length of observing Cdc8 stably residing on a filament , which arises from the complex cooperative binding process discussed above and below . Equation 1 is solved formally by the equation ( 3 ) P→ ( t ) =e∫0tW ( t′ ) dt′P→ ( 0 ) in which case: ( 4 ) P→ ( t ) =exp ( [0konvgrowt220−konvgrowt22] ) P→ ( 0 ) =[11−e−konvgrowt220e−konvgrowt22]P→ ( 0 ) Hence , since P→ ( 0 ) = ( P2 ( 0 ) P1 ( 0 ) ) = ( 01 ) , P1 ( t ) =e−konvgrowt22 , the probability of a first binding event happening at time t = τ is the probability of still being unbound at time τ , P1 ( τ ) , times the rate of binding at time τ , k12 ( τ ) , hence: ( 5 ) P ( τ ) =k12 ( τ ) P1 ( τ ) =konvgrowτ e−konvgrowτ22 Given this equation for the binding time , we also determined the probability of binding to a particular site on the actin filament that has been in the filament for an amount of time τage , P ( τage ) . Under our assumptions , at time t , the actin filament has length l ( t ) = vgrowt and the probability of Cdc8 binding anywhere along the filament is uniform . The probability of binding at a distance x0 from the end of the actin filament of length l is ( 6 ) Pbind ( x0|l ) ={1/ll≥x00l<x0 Consequently , given our assumption of a constant average growth rate vgrow , the probability of binding to a spot of a given age τage is ( 7 ) Pbind ( τage|l=vgrowt ) ={1vgrowtτage≤t0τage>t To determine the probability that a molecule binds to a part of the substrate of age τage , we integrated Equation 7 against the probability that the binding event happened at time t , as given by Equation 5 . Therefore , ( 8 ) Pbind ( τage ) =∫0∞dt Pbind ( τage|l=vgrowt ) P ( t ) ( 9 ) =∫τage∞dt kone−konvgrowt22 ( 10 ) =konvgrow2∫τagekonvgrow2∞dx e−x2 ( 11 ) =πkonvgrow2erfc ( τagekonvgrow2 ) where erfc ( x ) is the so-called complementary error function and was evaluated numerically . Using these equations , we chose kon to fit the data in Figure 3—figure supplement 1 and we found that a value of 4 × 10−6 sec−1 nm−1 gave good agreement with the data . The validity of the aforementioned expressions was tested by comparison with a simple simulation . The simulation methodology moreover provided a way to mimic the noise level that arose due to sample size limitations . A Monte Carlo scheme was performed by taking discrete time steps of size dt and at each time assuming the probability of binding in that time dt is ( 12 ) Pbind ( dt ) =1−e−konvgrowt dt An example implementation:from math import exp from random import random def simulate_binding ( k_on , v_grow , dt ) : t = 0 while True: fil_length = v_grow*t p_bind = 1 - exp ( -k_on * v_grow * t * dt ) # pick a random number # from zero to one and decide # if binding will occur if random ( ) <p_bind: # choose a random position # to bind from barbed end # to pointed end binding_position = random ( ) *fil_length binding_age = binding_position/v_grow return t , binding_age t = t + dt def simulate_experiment ( n_trials , k_on , v_grow , dt ) : binding_time_list = [] age_list = [] for i in range ( n_trials ) : t , binding_age = simulate_binding ( k_on , v_grow , dt ) binding_time_list . append ( t ) age_list . append ( binding_age ) # histogram/bootstrap age_list # and binding_time_list . . . Elongation rates of individual tropomyosin Cdc8 cables were determined by creating a region of interest ( ROI ) of an individual actin filament in the actin 488 channel over many time points . The actin ROIs were then applied to the Cdc8 ( 647 ) channel and adjusted slightly if necessary to account for movement of the actin filament during channel switching . A kymograph of the Cdc8 647 channel was created from the ROIs using an adapted version of Kymograph - Time Space Plot ImageJ plugin ( http://www . embl . de/eamnet/html/kymograph . html ) . Spreading rates were determined from kymographs by identifying examples of constant growth over at least 15 s ( three frames ) . The distance of growth over time was calculated to determine a rate of spreading . To determine whether the two tropomyosin Cdc8 cables on each side of the actin filament are influenced by each other , we first quantified whether a second Cdc8 binding event on an actin filament was more or less likely to bind at a site that is already bound by Cdc8 on one side . First Cdc8 binding events were determined as stated above , and those actin filaments were observed until a second Cdc8 binding event occurred , and whether or not the binding occurred at a site already occupied by Cdc8 or not was determined . At that frame , the length of the actin filament as well as the total Cdc8 cable length was calculated to determine the current percent occupancy of Cdc8 on the actin filament . These ‘current occupancies’ were then binned into 0–12 . 5% , 12 . 5–25% or 25–50% occupancy . Initial occupancy cannot exceed 50% as one Cdc8 binding event could at most cover one side of the actin filament , or half of the available binding sites . To determine whether the proportion of observed second binding events that bind opposite a tropomyosin Cdc8-bound segment ( Figure 3E ) would be expected in the absence of indirect cooperativity , or whether they are more likely given inclusion of some positive or negative indirect cooperativity factor , we performed a bootstrapping-type analysis on the n = 37 observed events . For this analysis we asked , given experimental data , what is the probability that the second binding events are binding randomly vs binding in a biased fashion towards or against being across from a Cdc8-coated segment ? For an actin filament of length La and a first Cdc8 cable of length L1 starting at position x1 ( in the range 0 to La-L1 ) along the actin filament , we determined the probability of a second Cdc8 cable of length L2 binding across from the first Cdc8 cable . We then compared the probability generated from different models with our experimental results . To account for experimental resolution , we divided the actin filament into a grid of 100 nm segments . In order to compute the probability of the second binding event overlapping with the first Cdc8 cable , we counted the number of places the second cable could bind that overlaps with the first cable of length L1 , and divided by the total number of potential places a second Cdc8 cable of length L2 could bind . As a helical actin filament has two grooves , there are twice as many potential binding sites on F-actin unoccupied by Cdc8 as those occupied by a single Cdc8 cable . However , if there is any overlap of second Cdc8 cable binding with the first Cdc8 cable , we assumed that there was only one potential binding face . For random binding , Cdc8 binding to each actin site was given an equivalent likelihood . To account for potential indirect cooperativity , we performed a weighted sum , where the second Cdc8 cable binding is proportionally more or less likely by a factor of c at all sites overlapping with the first cable of Cdc8 . The probability of an overlapping binding was therefore given by the following three expressions:N1=2∑x=0La−L2 ( ( x+L2 ) <x1 ) | ( x> ( x1+L1 ) ) N2=∑x=0La−L2{ ( x1−x ) +c ( L2− ( x1−x ) ) x<x1 & ( x+L2 ) >x1 ( L2− ( ( x1+L1 ) −x ) ) +c ( ( x1+L1 ) −x ) x≥x1 & ( x<x1+L1 ) ( with the terms in outer parentheses being restricted to minimum value 0 and maximum L2 ) p2=N2N1+N2 We then compared the expected outcomes for random ( c = 1X ) vs positive or negative indirect cooperativity values ( c = 2X or c = 0 . 5 respectively ) ( Figure 3F ) . Finally , we performed a bootstrapping analysis . For each experimentally observed set of events , where a given ( La , L1 , and x1 ) was measured , we computed the probability that a cable of length L2 binding would overlap with the first Cdc8 cable ( p2 ) . We then re-performed the 37 experiments 5000 times with a chosen indirect cooperativity factor c , choosing randomly with probability p2 whether or not overlapping binding occurred . Binning these events similar to the experimental data , we computed the values shown in Figure 3F . To probe the microscopic origins of the high cooperativity observed for tropomyosin Cdc8 ( Figure 1B ) , we distinguished between two potential cooperative mechanisms: ( 1 ) end-to-end cooperativity and ( 2 ) indirect cooperativity ( Figure 1C ) . We distinguished between these two models by modeling the binding kinetics of Cdc8 to a growing actin filament using a lattice model ( Figure 4Bi , 4 Ci ) . As two distinct Cdc8 cables can bind to an actin filament ( one on the surface of each groove of the helical actin filament ) , we represented the actin filament as a lattice with two rows representing the two actin surfaces potentially bound by Cdc8 . Each Cdc8 molecule interacts across four actin monomers , meaning that a bound Cdc8 extends over the length of 8 actin monomers total . Therefore , as each Cdc8 cable is represented separately , we ascribed to each lattice site a length l of approximately 21 . 6 nm ( assuming an approximate actin spacing of 2 . 7 nm ) . The dynamics of Cdc8 molecules associating with this lattice were then simulated using a kinetic Monte Carlo procedure ( Newman and Barkema , 1999 ) . First , we considered the case for a non-elongating actin filament , represented by a fixed lattice of length N . Each potential Cdc8 binding site can either be occupied ( 1 ) or unoccupied ( 0 ) by Cdc8 . We described binding site site i in row j as xij . Every unoccupied site had a base on-rate kon ( 0→1 ) and a base off-rate koff ( 1→0 ) . To include the effect of end-to-end cooperativity , the on-rate for unoccupied site ij was multiplied by a factor of w and the off-rate for occupied site ij was divided by a factor of w for each occupied immediate neighbor in the same row ( Figure 4Bi ) . Using these parameters , we ran a model describing Cdc8 binding with end-to-end cooperativity as the sole form of cooperativity ( Figure 4B ) . In order to additionally probe the potential role of indirect cooperativity , we adjusted the same model to include additional parameters for indirect cooperativity . In this model , on-rates were multiplied by a factor c and off-rates were divided by a factor c for each occupied site in the opposite row ( row 1-j ) ( Figure 4Ci ) . For a lattice size 2xN sites , there were precisely 2xN possible events . The dynamics were then solved using kinetic Monte Carlo simulation ( Newman and Barkema , 1999 ) up to a chosen time max_time or until the lattice was completely filled . In order to factor in an elongating actin filament , an additional type of event representing actin filament growth was included , with a rate equal to the input growth rate of actin , vgrow . In this case , whenever that event was selected by the algorithm , time was advanced and N was set equal to N = floor ( new_time*site_extension_rate ) To make kymographs and to compute the occupancy of Cdc8 as accurately as possible , we processed the output raw simulation data in three ways: The coverage data in Figure 4Biii and 4Ciii as computed by averaging this fluorescence level over the filament at the time when the filament had length 6 μm over 24 such simulations . 6 μm was the average length of the actin filaments at the frame used for quantification of Cdc8 coverage in TIRFM experiments ( 4Aiii ) . In order to choose the values of w , c and koff , we first fixed c , and tried combinations of koff and w that gave ( a ) the best match to the data shown in Figure 1B as kon is scanned , and ( b ) gave kymographs whose first time for observed binding was similar to what was observed experimentally . Due to cooperative binding effects , modeling is required to find a value for kon that is commensurate with experimental data . In Figure 4—figure supplement 1 , koff was measured from single molecule Cdc8 events on an actin filament fully coated by Cdc8 . In our model , the microscopic koff corresponding to this situation was koffmeasured = koffmodel/w2/c . Using koffmodel values far from those measured in Figure 4—figure supplement 1 did not give a good agreement with the data . Constraining koff to the value measured experimentally results in a single value of w that best matched the experimental data ( Figure 1B , parameters koff = 116 . 8 sec−1 , w = 40 , c = 1 , Kd = 0 . 08 sec-1 ) , with the results from these simulations shown in Figure 4B . However , these simulations did not match the experimental single/double coverage data ( Figure 4Aiii ) . Relaxing the restraint on koff did not alleviate this problem . For a fixed value of c > 1 , there is a single w that best matched the experimental data . For c = 1 . 25 , while these simulations matched the experimental data ( Figure 1B ) , they over-stabilized double cable association . Hence , we slightly relaxed the restriction on koff and the simulations with parameters ( koff = 300 sec−1 , w = 125 , c = 1 . 25 , Kd = 0 . 02 sec-1 ) matched the experimental data ( Figure 4A ) very closely as well as qualitatively reproducing many of the aspects we observed for Cdc8 loading ( Figure 4C ) . We note that a limitation of this model is that it did not allow for any ‘frame shifts’ in the Cdc8 binding . Hence there cannot be defects in Cdc8 occupancy on a cable that are smaller than four actin monomers in these representations , and we did not allow the Cdc8 to ‘slide’ except by binding and unbinding . The first part of this could be addressed by using a lattice of 4x higher resolution . However , since we did not know how to account for the end-to-end cooperativity in this case , or how to represent the sliding dynamics , we chose the simpler representation for the purposes of this study . The rates of a lattice with these w and c parameters fixed can be written in terms of energies for a two-row Ising model in a field . Since this system is still one dimensional in nature , for the infinite-length case any equilibrium properties of the system can be derived using a transfer matrix approach ( Tsuchiya and Szabo , 1982 ) in the same way as for a standard Ising model ( except that the transfer matrix is 4 × 4 ) . However , given that the growth of the actin filament and the way that the data is analyzed play significant roles in the coverage fractions computed , we have found the simulations useful since they are able to replicate the full dynamics as shown in Figure 4Bii and 4Cii . The percentage of actin filaments bundled was quantified at similar F-actin densities ( between 800 and 1100 μm per field ) for each experiment . The total actin filament length in the chamber was measured manually by creating ROIs for every actin filament . ROIs for every segment of actin filament present in a bundle were then created , and the ratio of actin filament present in a bundle vs . total actin filament length was quantified . Fold-change in fluorescence intensity over time was taken by taking a total fluorescence measurement of the entire TIRFM field for each frame of the movie . The fold-change in fluorescence of each frame compared to the first frame of the movie was quantified and plotted over time . ROIs were created for all single filaments and two-filament bundles at that time point and total filament length was measured . Severing events that occurred within those filaments prior to the selected frame were quantified , and severing events per micron per second was calculated . We were only able to directly observe severing of single filaments and two-filament bundles , and were unable to determine severing rate within the dense bundles generated by a combination of fimbrin Fim1 and ADF/cofilin Adf1 . The spontaneous assembly of 1 . 5 µM Mg-ATP-actin monomers ( 10% pyrene-labeled ) was carried out in a 96 well plate as described ( Zimmermann et al . , 2016 ) . WT ADF/cofilin Adf1 or TMR labeled Adf1 were mixed with 10X KMEI ( 500 mM KCl , 10 mM MgCl2 , 10 mM ethylene glycol tetraacetic acid [EGTA] , and 100 mM imidazole , pH 7 . 0 ) and placed in the lower row of the plate . Pyrene fluorescence readout over time was detected using an Infinite M200 Pro ( Tecan Systems , Inc . , San Jose , CA ) fluorescence plate reader . Fission yeast cells were grown in YE5S media at 25°C for 24–36 hr . Cells were fixed in 100% cold methanol and nuclei ( DAPI ) and septa ( calcofluor ) were visualized . For staining , cells were incubated in 300 μL of 50 mM sodium citrate and 4 μL Calcofluor White Stain ( Fluka Analytical , Sigma-Aldrich , St . Louis , MO ) at 37°C for 5 min . Cells were then washed with 1 mL of 50 mM sodium citrate , and resuspended in 15 μL sodium citrate and 4 μL DAPI stock ( 1 mg/mL in H20 , Life Technologies , Carlsbad , CA ) . Stained cells were kept on ice until imaging . Cells were imaged using differential interference contrast ( DIC ) and epifluorescence with an Orca-ER camera ( Hamamatsu , Bridgewater , NJ ) fitted to an IX-81 microscope ( Olympus , Tokyo , Japan ) , with a 60 × 1 . 4 N . A . Plan Apo objective . Time required for fission yeast strains to assemble a contractile ring was quantified using strains expressing Lifeact-GFP as a general actin marker ( Table 1 ) . Single z-plane movies were taken ( one frame per minute ) using an epifluorescence microscope as in DAPI/calcofluor imaging . Time was quantified from the first appearance of fluorescence at the midzone to the formation of a compact ring . BODIPY-phallacidin staining of fission yeast to visualize their actin cytoskeleton was performed as described previously ( Sawin and Nurse , 1998 ) . BODIPY-phallacidin powder ( Thermo Fisher Scientific , Waltham , MA ) was suspended to 0 . 2 units/µL in methanol , aliquoted , and lyophilized for storage at −20°C . Fission yeast grown in YE5S were fixed in 16% paraformaldehyde at room temperature for 5 min . Cells were washed with PEM buffer three times at room temperature and permeabilized in PEM with 1% triton X-100 for exactly 1 min . Cells were then spun at 7000 RPM for 30 s and the supernatant was removed quickly . Cells were washed in PEM buffer three times and resuspended in 10 µL PEM buffer following the final wash . Lyophilized phallacidin was resuspended to one unit/µL and 1 µL of resuspended phallacidin was added to 10 µL of cells and incubated at room temperature for 30 min in the dark . Following phallacidin incubation , cells were washed with 1 mL PEM and spun at 7000 RPM for 30 s or until nearly all cells were spun down . Supernatant was removed leaving the resuspended cells in a small volume . Resuspended cells were imaged on glass slides using a Zeiss Axiovert 200M fitted with a 100x , 1 . 4 NA objective and Yokogawa CSU-10 spinning disk unit ( McBain , Simi Valley , CA ) equipped with a 50-milliwatt 473 nm DPS laser and Cascade 512B EM-CCD camera ( Photometrics , Tucson , AZ ) . Sedimentation assays were performed as previously described ( Skau and Kovar , 2010 ) . 15 μM Mg-ATP actin monomers were spontaneously assembled in 10 mM imidazole , pH 7 . 0 , 50 mM KCl , 5 mM MgCl2 , 1 mM EGTA , 0 . 5 mM DTT , 0 . 2 mM ATP and 90 µM CaCl2 for 1 hr to generate F-actin . F-actin was then incubated with tropomyosin Cdc8 mutants ( 2 µM ) for 20 min at 25°C and spun at 100 , 000g at 25°C . Supernatant and pellets were separated by 15% SDS-PAGE gel electrophoresis and stained with Coomassie Blue for 30 min , destained for 16 hr and analyzed by densitometry with ImageJ . | Cells use a protein called actin to provide shape , to generate the forces needed for cells to divide , and for many other essential processes . Inside a cell , individual actin proteins join up to form long filaments . These actin filaments are organized in different ways to make networks that have distinct properties , each tailored for a specific process . For instance , bundles of straight actin filaments help a cell to divide , whereas a network of branched actin filaments allows cells to move . The different proteins that bind to actin filaments influence how quickly actin filaments are assembled and organized into networks . Therefore , many of the properties of an actin filament network are due to the actin binding proteins that are associated with it . Two actin binding proteins called fimbrin and cofilin associate with a type of actin filament network known as the actin patch . A third actin binding protein called tropomyosin associates with a different network that forms a ring . It is not known how particular actin binding proteins choose to associate with one actin network instead of another . Christensen et al . used a fluorescence microscopy technique to study how fimbrin , cofilin and tropomyosin associate with different actin networks in a single-celled organism called fission yeast . This technique involved incubating actin and actin binding proteins together in a microscope chamber . The experiments show that some actin binding proteins , like tropomyosin , cooperate to bind to actin . Individual tropomyosin molecules find it difficult to bind actin filaments on their own , but once one tropomyosin molecule is attached to the filament , others rapidly join to coat the filament . On the other hand , some actin-binding proteins compete for binding to filaments . For example , the binding of fimbrin to actin filaments causes tropomyosin to be removed from the actin network . Further experiments revealed that fimbrin and cofilin work with each other to rapidly generate a dense actin network and displace tropomyosin . Together , the findings of Christensen et al . suggest that competitions between actin binding proteins determine which actin binding proteins are associated with an actin network . The next challenge is to understand how the most competitive actin-binding proteins are kept off actin networks where they do not belong . Further studies will shed light on how these interactions cause large changes in how the cell is organized . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | Competition between Tropomyosin, Fimbrin, and ADF/Cofilin drives their sorting to distinct actin filament networks |
Highly distributed neural circuits are thought to support adaptive decision-making in volatile and complex environments . Notably , the functional interactions between prefrontal and reciprocally connected thalamic nuclei areas may be important when choices are guided by current goal value or action-outcome contingency . We examined the functional involvement of selected thalamocortical and corticothalamic pathways connecting the dorsomedial prefrontal cortex ( dmPFC ) and the mediodorsal thalamus ( MD ) in the behaving rat . Using a chemogenetic approach to inhibit projection-defined dmPFC and MD neurons during an instrumental learning task , we show that thalamocortical and corticothalamic pathways differentially support goal attributes . Both pathways participate in adaptation to the current goal value , but only thalamocortical neurons are required to integrate current causal relationships . These data indicate that antiparallel flow of information within thalamocortical circuits may convey qualitatively distinct aspects of adaptive decision-making and highlight the importance of the direction of information flow within neural circuits .
To reach specific goals in volatile environments , living organisms must integrate current internal motivational states with an up-to-date causal understanding of the relationships between external events ( Rangel et al . , 2008; Dickinson , 2012 ) . Such complex cognitive abilities are supported by highly evolved brain structures . Past research points to a central role for the dorsomedial prefrontal cortex ( dmPFC ) in adaptive decision-making ( Corbit and Balleine , 2003; Killcross and Coutureau , 2003 ) . A circuit-level analysis of the functional role of the PFC also indicates a major role for areas innervating the PFC ( O'Doherty , 2011 ) . In this respect , the mediodorsal thalamus ( MD ) appears of special interest due to the extensive reciprocal projections connecting these two areas ( Groenewegen , 1988; Gabbott et al . , 2005; Alcaraz et al . , 2016a ) . These anatomical considerations helped to shape a new functional view of the thalamus , wherein its role is not limited to that of a relay ( Sherman , 2005; Mitchell , 2015; Wolff et al . , 2015a; Sherman , 2016 ) . Indeed experimental interventions aimed at the MD produce a vast array of specific cognitive deficits on both rodents and primates ( Corbit et al . , 2003; Baxter , 2013; Parnaudeau et al . , 2013; Parnaudeau et al . , 2015; Alcaraz et al . , 2016b; Chakraborty et al . , 2016 ) , supporting the view that this thalamic area plays an integrative role within thalamocortical circuits ( Schmitt et al . , 2017 ) . Surprisingly , the functional significance of reciprocal projections , which are the hallmark of thalamocortical organization , has not been directly examined in the context of adaptive decision-making . A recent report provided initial evidence that thalamocortical and corticothalamic pathways recruited by the same behavioral task may support qualitatively distinct aspects of working memory ( Bolkan et al . , 2017 ) . This further underscores the importance of the functional interactions between cortical and thalamic areas in high-order cognition . Gaining clearer insight into the functioning of thalamocortical circuits therefore requires manipulating thalamocortical and corticothalamic pathways separately . In the present study , we applied a chemogenetic strategy in rats to specifically inhibit projection-defined dmPFC or MD neurons during a classic instrumental task requiring adaptive actions . We found that distinct goal attributes , namely current goal value and current action-outcome contingency , are differentially supported by thalamocortical and corticothalamic pathways .
To express an inhibitory DREADD receptor ( Armbruster et al . , 2007 ) only in dmPFC-projecting MD cells , an adeno-associated virus carrying a floxed hM4Di receptor expression cassette was injected in the MD , while a retrograde CAV-2 vector ( Junyent and Kremer , 2015 ) carrying the Cre recombinase was injected in the dmPFC ( Figure 2A ) . As a result , only thalamic cells projecting to the dmPFC were infected by both vectors and therefore expressed mCherry and hM4Di . In general mCherry expression was more evident in the lateral portion of the MD , in agreement with our current knowledge of these thalamocortical projections ( Alcaraz et al . , 2016a ) . mCherry expression was also visible to some degree in adjacent dmPFC-projecting thalamic areas such as the intralaminar group ( PC and CL mostly ) and , to some extent , the CM and the PV . In some cases , fluorescence was also observed in the habenula . Eight rats showed only minimal ( or unilateral ) levels of DREADD expression and were therefore excluded from the analyses ( saline: n = 7 , CNO: n = 7 ) . Figure 2B and C illustrate the extent of mCherry expression at the thalamic level . Instrumental learning took place 1 month postsurgery . Instrumental performance progressively increased over training for both CNO- and saline-treated groups , as shown by the significant effect of Session ( F ( 5 , 60 ) = 11 . 7 , p<0 . 0001 ) , but CNO-treated rats tended to perform fewer lever presses overall ( F ( 1 , 12 ) = 11 . 4 , p=0 . 0055 ) ( Figure 3A ) . There was no significant Drug X Session interaction however ( F < 1 ) , confirming efficient instrumental learning even for the CNO-treated group . In addition , the asymptotical performance did not differ between saline-treated and CNO-treated groups during the final session of training ( F ( 1 , 12 ) = 2 . 58 , p=0 . 1338 ) . During devaluation by specific satiety , both groups of rats consumed the same amount of food ( Saline group: 10 . 8 ± 0 . 5 g , CNO group: 12 . 8 ± 0 . 8 g; F ( 1 , 12 ) = 1 . 7 , p=0 . 2214 ) , indicating that basic motivational processes were not altered by CNO treatment . The ability to use current goal value to guide behavior was assessed during a choice test conducted immediately after devaluation , under extinction conditions . While the group of rats that received saline exhibited the expected adaptive behavior during that test , expressing a clear bias toward the action associated with the still valued outcome , rats that received CNO showed only little differential response toward either actions ( Figure 3B ) . Consistent with these observations , the critical Devaluation X Drug interaction approached significance ( F ( 1 , 12 ) = 4 . 0 , p=0 . 0679 ) , while the main effect of Devaluation ( F ( 1 , 12 ) = 10 . 0 , p=0 . 0081 ) but not of Drug ( F < 1 ) reached significance . When considering each group separately , a significant effect of Devaluation was evident for the saline ( F ( 1 , 6 ) = 7 . 7 , p=0 . 0324 ) but not the CNO group ( F ( 1 , 6 ) = 2 . 7 , p=0 . 1543 ) . To determine if this was due to a performance deficit during this test , we verified the dynamics of responding over time by analyzing the data as 2 min blocks . This analysis confirmed the existence of extinction with a significant effect of Block ( F ( 9 , 108 ) = 8 . 3 , p<0 . 0001 ) and this factor did not interact with Drug ( Block X Drug ( F ( 9 , 108 ) = 1 . 2 , p=0 . 2908 ) ; Block X Devaluation X Drug ( F ( 9 , 108 ) = 1 . 6 , p=0 . 1335 ) . Moreover , analyses conducted on each group separately confirmed that responding gradually decreased over time during this test ( Saline: Block , F ( 4 , 24 ) = 6 . 6 , p=0 . 0010; CNO: Block , F ( 4 , 24 ) = 7 . 4 , p=0 . 0005 ) . Thus , responding was initially higher and then declined to comparable rates for both saline- and CNO-treated groups suggesting that the impairment in the CNO group was not the result of a performance deficit . Collectively , these data therefore suggest that the CNO treatment produced a mild deficit in the ability to update goal value representation and/or its use to guide behavior . Importantly , a consumption test performed immediately after the devaluation test confirmed the effectiveness of the sensory-specific satiety , which was left unaltered by CNO administration . That is , both saline- and CNO-treated rats preferably consumed the still valued outcome when they could freely select from the two outcomes ( Devaluation ( F ( 1 , 12 ) = 25 . 6 , p=0 . 0003; Drug and Drug X Devaluation , Fs < 1; Figure 3C ) . After two sessions of retraining under standard conditions , rats were subjected to a new phase of instrumental training , during which the contingency between one of the actions and its associated outcome was selectively degraded . On this occasion , rats continued to receive the same treatment ( CNO or saline ) as that during initial instrumental learning ( see Figure 1 ) . During the degradation phase , differential responding was evident for both saline- and CNO-treated groups with lower responding when action-outcome contingency was degraded ( Figure 3D ) , as shown by the significant Degradation effect ( F ( 1 , 12 ) = 52 . 8 , p<0 . 0001 ) as well as the significant Session X Degradation interaction ( F ( 5 , 60 ) = 3 . 1 , p=0 . 0152 ) . Drug treatment did not produce any visible effect on this occasion ( Drug , F ( 1 , 12 ) = 1 . 6 , p=0 . 2305; Session x Drug , F ( 5 , 60 ) = 1 . 6 , p=0 . 1820; all remaining Fs < 1 ) . Thus , when the sensory feedback provided by the outcome was available , all rats were capable of exhibiting adaptive decision-making , showing that the consequences of their actions strongly affected their behavior . Finally , a critical choice test was conducted again under extinction conditions ( Figure 3E ) Interestingly , the behavior exhibited by the two groups of rats was now markedly different . While saline-treated rats continued to express differential responding for both actions , CNO-treated rats were unable to do so . In line with these observations , the critical Degradation X Drug interaction was significant ( F ( 1 , 12 ) = 5 . 6 , p=0 . 0354 ) , as were the main effects of Drug ( F ( 1 , 12 ) = 6 . 8 , p=0 . 0225 ) and of Degradation ( P ( 1 , 12 ) = 6 . 0 , p=0 . 0310 ) . Further analyses confirmed that the main effect of Degradation was significant for the saline-treated ( F ( 1 , 6 ) = 12 . 2 , p=0 . 0129 ) but not the CNO-treated group ( F < 1 ) . We provide as an Appendix supplemental data showing that these effect did not result from CNO alone because neither CNO nor DMSO treatment altered behavior throughout testing ( Appendix 1—figures 1–3 ) . Thus , inhibiting dmPFC-projecting MD neurons produced selective impairments when rats were forced to rely on representations to guide behavior . The impairment appeared to be mild when rats were required to use current goal value , but more pronounced when the contingency between an action and its consequence was altered . Overall , these data indicate a central role for thalamocortical pathways in the ability to guide choice based on current knowledge of the causal link between actions and their outcomes . Next , we used the same strategy in a distinct set of rats to examine the behavioral outcome of inhibiting dmPFC neurons projecting to the MD . For this purpose , injections sites for either viral construct were reversed ( Figure 2D ) . The resulting mCherry expression at the cortical level is shown in Figure 2E and F . A marked expression of mCherry was evident in deep cortical layers , consistent with the existence of abundant corticothalamic projections targeting the MD from cortical layers 5/6 ( Gabbott et al . , 2005 ) . Although we did not quantify the number of labelled cells , comparing experiments 1 and 2 shows that greater labelling was evident for CT cells , consistent with the view that CT cells outnumber TC cells ( eg . , Haber and Calzavara , 2009 ) . Six animals showed little or unilateral DREADD expression and were not considered for analyses ( saline: n = 8 , CNO: n = 6 ) . Instrumental learning was comparable between saline- and CNO-treated rats ( Figure 4A ) , with improved instrumental learning over training ( F ( 5 , 12 ) = 99 . 3 , p<0 . 0001 ) . Instrumental learning was not affected by CNO ( Drug , F ( 1 , 12 ) = 3 . 1 , p=0 . 1050; Session X Drug interaction , F < 1 ) . During devaluation , all rats again consumed an equal amount of food , irrespective of whether they were treated with saline or CNO ( Saline group: 9 . 1 ± 0 . 6 g , CNO group: 9 . 1 ± 0 . 7 g; F < 1 ) . Immediately after the devaluation procedure however , the choice test conducted in extinction revealed a markedly distinct pattern of response in the two groups of rats ( Figure 4B ) . While the saline group expressed a clear bias for the still valued option , the CNO group responded similarly for both actions , consistent with the view that they failed to use current goal value to guide behavior . Importantly , the critical Drug X Devaluation interaction reached significance ( F ( 1 , 12 ) = 9 . 2 , p=0 . 0103 ) , providing compelling support for these observations . In addition , the main effect of Devaluation ( F ( 1 , 12 ) = 8 . 1 , p=0 . 0149 ) and Drug ( F ( 1 , 12 ) = 6 . 6 , p=0 . 0244 ) also reached significance . Separate analyses confirmed the existence of a selective deficit in CNO-treated ( Devaluation , F < 1 ) but not saline-treated ( Devaluation , F ( 1 , 7 ) = 23 . 1 , p=0 . 0020 ) rats . Again , the presence of extinction was confirmed by analyzing the data as blocks of 2 min ( Block , ( F ( 19 , 108 ) = 3 . 9 , p=0 . 0002 ) . In addition , drug treatment did not interact with the general dynamics of responding during this test ( Block X Drug ( F ( 9 , 108 ) = 1 . 3 , p=0 . 2721 ) ; Block X Devaluation X Drug ( F ( 9 , 108 ) = 1 . 4 , p=0 . 2062 ) . Further analyses confirmed a significant effect of Block for both saline- ( F ( 4 , 28 ) = 5 . 0 , p=0 . 0035 ) and CNO-treated ( F ( 4 , 28 ) = 5 . 6 , p=0 . 0034 ) groups suggesting that responding decreased over time in a similar fashion for both groups . Thus , floor effect alone cannot account for the specific impairment evident during this test . Consumption tests conducted immediately after yielded essentially the same results as in experiment 1: all rats expressed a clear bias for the still valued outcome ( the sensory-specific satiety procedure was efficient ) and CNO treatment did not alter behavior at this stage ( Devaluation , ( F ( 1 , 12 ) = 44 . 0 , p<0 . 0001; Drug , ( F ( 1 , 12 ) = 1 . 9 , p=0 . 1934; Drug X Devaluation , F < 1 , Figure 4C ) . After two days of retraining , the degradation phase began . Differential responding was evident for both saline- and CNO-treated groups with lower responding when action-outcome contingency was degraded ( CNO treatment produced no visible effect , Figure 4D ) . As a consequence , the main effect of Degradation ( F1 , 12 ) =15 . 4 , p=0 . 0020 ) and Session ( F ( 5 , 60 ) = 4 . 6 , p=0 . 0014 ) as well as the interaction between these factors ( F ( 5 , 60 ) = 9 . 8 , p<0 . 0001 ) reached significance . The main effect of Drug did not reach significance ( F < 1 ) and no interaction was observed with this factor ( Degradation X Drug , F ( 1 , 12 ) = 1 . 6 , p=0 . 2252; Session X Drug , F ( 5 , 60 ) = 1 . 6 , p=0 . 1719; Degradation X Session X Drug , F < 1 ) . During the final choice test conducted in extinction ( Figure 4E ) , all rats continued to select the action with reliable consequences as attested by the significant effect of Degradation ( F ( 1 , 12 ) = 5 . 8 , p=0 . 0331 ) . Importantly , inhibiting corticothalamic pathways produced no noticeable effect and did not prevent rats to display adaptive decision-making even when they could only rely on represented information ( Drug and Drug X Degradation , Fs <1 ) . Thus , dissociable patterns of performance were obtained when inhibiting thalamocortical and corticothalamic pathways during choice tests conducted under extinction conditions . A clear deficit in the ability to use recently updated goal value was observed when inhibiting the corticothalamic pathway , but the same treatment did not prevent animals to adapt to a selective change in the contingency between an action and its outcome . The latter ability was however abolished by the inhibition of the thalamocortical pathway , which also produced a mild deficit in the ability to use goal value to guide behavior .
In this study we sought to disentangle the functional contribution of thalamocortical and corticothalamic pathways connecting the dmPFC with the MD in the context of goal-directed behaviors . The present data indicate an important contribution for both cortical and thalamic neurons in the ability to perform adaptive actions . However , while both neuronal populations were found to be important to guide behavior based on current goal value , only thalamic neurons critically supported choice based on current causal relationships . Importantly , inhibiting corticothalamic and thalamocortical pathways produced very specific patterns of behavioral alterations , apparent only when no rewards were available . Thus , deficits only appeared during tests conducted under extinction conditions , suggesting that functional interactions between cortical and thalamic areas are important to guide behavior based on the current content of mental representations . Even then , inhibition of the corticothalamic pathway left the ability to guide choice based on current action-outcome contingency unaltered . Importantly , control experiments showed that CNO injections alone did not alter instrumental behavior at any stage of the task ( Appendix 1 ) . Together with the absence of effects of CNO on consumption ( Figures 3C and 4C ) , this appears to be sufficient to rule out any non-specific impact of CNO , e . g . due to clozapine conversion ( Gomez et al . , 2017 ) . But more importantly , it points to a specific role for projections-defined thalamic and cortical neurons in cognitive processes that are necessary when the task includes unobservable information ( Bradfield et al . , 2015 ) . Recent studies have emphasized a consistent role for thalamic nuclei to sustain cortical activity when holding online information is important for subsequent choice ( Bolkan et al . , 2017; Schmitt et al . , 2017 ) . The present data support this view as only dmPFC-projecting MD neurons were found to be important to support choice based on the current mental representation of action-outcome contingency . Thus , one role of the MD could be to provide online information to support choice when no observable element can help to retrieve action-outcome contingency . By itself , this result is also consistent with the effects of global chemogenetic inhibition of the MD ( Parnaudeau et al . , 2015 ) . Similarly , recent findings obtained in primate suggested that MD-lesioned monkeys are unable to persist in successful strategies ( Chakraborty et al . , 2016 ) , hinting at a similar problem of maintaining an accurate representation of the associative structure of the current task over time . The role of the dmPFC in goal-directed actions is now well established ( Corbit and Balleine , 2003; Killcross and Coutureau , 2003; Tran-Tu-Yen et al . , 2009; Hart and Balleine , 2016 ) , and is crucial for the acquisition of instrumental action-outcome associations ( Tran-Tu-Yen et al . , 2009 ) . However , following acquisition , the role of the dmPFC in adapting instrumental responses to contingency changes appears complex , depending on the accessibility of the reward ( Corbit and Balleine , 2003 ) and the nature of contingency changes ( Coutureau et al . , 2012 ) . As a result , the available data therefore suggest that the dmPFC is differentially implicated in guiding choice based on current goal value or current action-outcome contingency ( Naneix et al . , 2009 ) . Our data show that MD-projecting dmPFC neurons were necessary for representing/using current goal value , but not current action-outcome contingency . It is therefore possible that other pathways originating from the dmPFC and preserved in the present study may be relevant to track contingency changes . Overall levels of responding appeared to be somewhat low during the initial devaluation test , especially when inhibiting dmPFC-projecting MD neurons ( Experiment 1 ) . Interestingly , this last feature is reminiscent from classic studies showing lower levels of instrumental performance in rats sustaining MD lesions ( Corbit et al . , 2003 ) . However , in our study the same rats exhibited high level of responding during the degradation phase , suggesting that this disturbance was at best transitory . Low levels of performance are sometimes reported even in controls during devaluation tests ( Corbit et al . , 2003; Bradfield and Balleine , 2017 ) . It seems unlikely that low levels of performance alone could account for the specific impairments during devaluation tests because CNO-treated rats exhibited normal extinction at this occasion . Since responding diminished over time during this test for all rats in a comparable fashion , performance was initially above floor level . We cannot exclude however that the impairments resulted from generalization on the two actions available during the choice test . We were actually concerned beforehand about this possibility , which prompted us to use two clearly distinct manipulanda ( a lever and a tilt ) , unlike the two levers most commonly adopted in the literature . This should limits the possibility that rats generalize current goal value for both actions . In experiment 1 in particular , CNO-treated rats behave as if they were generalizing but this could result from an inability to select the correct option in the absence of the sensory feedback provided by the reward . The identification of a specific role for corticothalamic projections not only strongly argues against the view of a thalamus acting only as a relay , but also suggests a specific role for these pathways in cognition ( Crandall et al . , 2015; Guo et al . , 2017 ) . Understanding the functional relevance of these corticothalamic pathways is an important issue as conceptual views posit that they may contribute to cortical functioning by enabling transthalamic communication between cortical areas , thus offering supplemental integrative opportunities ( Sherman and Guillery , 2011; Sherman , 2016 ) . The functional contribution of thalamocortical pathways appears to be consistent with that of a general role of the thalamus to direct attention toward task’s elements relevant for successful performance ( Wolff et al . , 2015a; Wolff et al . , 2015b ) , not only in the presence of cues , but also when using the current content of mental representation is required for successful performance . In conclusion , we provide causal evidence that thalamocortical and corticothalamic pathways connecting the dmPFC and the MD support at least partially dissociated goal attributes . These results highlight the directionality of the functional exchanges within neural circuits as one of their fundamental features ( see also Bolkan et al . , 2017; Lichtenberg et al . , 2017 ) , which calls for a more systematic functional assessment of reciprocally connected pathways . Past research has indicated a time-limited role for both the dmPFC ( Ostlund and Balleine , 2005; Tran-Tu-Yen et al . , 2009 ) and the MD ( Ostlund and Balleine , 2008 ) in the acquisition of goal-directed behaviors . While studies that have directly examined functional interactions between cortical and thalamic areas have generally used permanent interventions ( Bradfield et al . , 2013; Browning et al . , 2015 ) , as was the case in the present study , proceeding to stage-limited interventions appears as a valuable prospect to further refine the functional contribution of projections-defined neurons .
42 male Long Evans rats weighting 275 g to 300 g at surgery were obtained from Centre d’Elevage Janvier ( France ) . Rats were initially housed in pairs and accustomed to the laboratory facility for two weeks before the beginning of the experiments . Environmental enrichment was provided by tinted polycarbonate tubing elements , in accordance with current French ( Council directive 2013–118 , February 1 , 2013 ) and European ( directive 2010–63 , September 22 , 2010 , European Community ) laws and policies regarding animal experiments . The facility was maintained at 21 ± 1°C with lights on from 7 a . m . to 7 p . m . The experimental protocols received approval #5012053-A from the local Ethics Committee on December 7 , 2012 . After histological verification ( see below ) , the final group sizes were: thalamocortical: n = 7 for saline , n = 7 for CNO; corticothalamic: n = 8 for saline , n = 6 for CNO . Rats were anaesthetized with 4% Isoflurane and placed in a stereotaxic frame with atraumatic ear bars ( Kopf , Tujunga , CA ) in a flatskull position . Anaesthesia was maintained with 1 . 5–2% Isoflurane complemented by subcutaneous administration of buprenorphin ( Buprecare , 0 . 05 mg/kg ) . CAV-2 and AAV were pressure injected ( Picospritzer , General Valve Corporation , Fairfield , NJ ) into the brain through a glass micropipette ( outside diameter: around 100 µm ) and polyethylene tubing . For MD-to-dmPFC pathway targeting , 1 µl of 1 × 109 genomic copies/µl of CAV2-Cre ( Biocampus PVM , Montpellier , France ) was injected bilaterally in the PL at the following coordinates: AP +3 . 2 mm from bregma , laterality ±0 . 6 mm , ventrality −3 . 4 mm from skull . In the same surgery session , 1 µl of 1 × 109 genomic copies/µl of AAV-hSyn-DIO-hM4Di-mCherry ( UNC Vector Core , USA ) was injected bilaterally in the MD at the following coordinates: AP −2 . 6 mm , laterality ±0 . 7 mm and ventrality −5 . 6 mm . For dmPFC-to-MD pathway targeting , virus injections were reversed , that is , CAV-2 in the MD and AAV in the dmPFC . All injection parameters were the same , except for the mediolateral coordinates of AAV injection in the dmPFC , set at ±0 . 8 mm , in order to preferentially target the cortical layers V and VI which project to the MD . In all groups , the pipette was left in place 5 min after injection before slow retraction . To allow for optimal viral expression , rats were given one month of recovery before behavioral testing began . Rats were first habituated to the magazine dispenser through two daily sessions of magazine training for 2 days . A session consisted in the delivery of 30 food rewards , grain or sucrose pellets , distributed randomly through a 30 min session . The first session took place in the morning , and the second in the afternoon , with the order of rewards counterbalanced between rats and days . Twelve daily sessions of instrumental training began the day after the last session of magazine training , during which rats had to make specific associations between two responses ( lever press or tilt action ) and the two different outcomes . Daily training consisted in instrumental learning with either the lever or the tilt , each specifically associated with one of the outcome ( i . e . either grain or sucrose pellets , see Figure 1 ) . For clarity , blocks of instrumental performance were considered for analyses on two consecutive sessions ( one with the tilt , one with the lever , then averaged for the analysis ) . Daily training was completed when 30 rewards were earned or 30 min had elapsed . The action-outcome associations and the order of their presentations were counterbalanced between rats and days . For the four first sessions , each action was reinforced . Then , for sessions 5 to 8 , a random ratio schedule of 5 was introduced ( 2 to 10 actions were necessary to obtain the reward , probability of receiving an outcome given a response = 0 . 2 ) . Sessions 9 to 12 were performed with a RR10 schedule ( 4 to 20 actions were necessary to obtain the reward , probability of receiving an outcome given a response = 0 . 1 ) . The last instrumental session with each action ( RR10 , highlighted in Figure 1 ) was used as a measure of baseline performance for the devaluation test while the last retraining session after this devaluation test ( RR10 , highlighted in Figure 1 ) was used as a measure of baseline performance for the degradation phase , including the choice test ( see Figure 1 ) . The day after the last session of training , rats were placed in a plastic feeding cage containing free access of 15 g of one of the two outcomes for one hour of devaluation . Half of the rats in each response-outcome assignment received grain pellets , the remaining receiving sucrose pellets . Immediately after , rats were put in the operant cages for a 10 min extinction test . During the test , both actions were available but unrewarded . This ensured that rats were using representations of the response-outcome contingencies and outcome value to guide their behavior . Animals that received saline during training also received saline during the test and the same logic applied for animals that had received CNO . Performance was quantified relative to prior baseline levels . After the extinction test , rats were put in the plastic feeding cage used for outcome devaluation . They had free access first to 5 g of one outcome for 15 min , and then to 5 g of the other outcome for 15 min . Food consumed was then measured for each outcome . Order of outcome presentation was counterbalanced between rats and groups . One day after completion of the consumption test , rats received two supplemental sessions of RR10 to reinstate regular instrumental training . Immediately after , the degradation procedure began . For one of the action-outcome associations , the contingency between the action and its consequences was maintained identical to that used during instrumental training ( RR10 training ) but for the other , the contingency was degraded by delivering the same overall number of rewards randomly even if no action was performed . For both the degradation phase and the test , performance was quantified relative to prior baseline levels . CNO ( Enzo Life Science ) was diluted in saline with 0 . 5% of DMSO at a final concentration of 1 mg/ml . CNO groups received a daily CNO i . p . injection one hour before each training and testing session ( 1 mg/kg ) while saline groups received a daily saline i . p . injection ( 1 ml/kg of 0 . 9% Saline ) one hour before each training and testing session ( see Figure 1 ) . We recently demonstrated the efficacy of CNO administration in reducing neuronal activity at a dose of 1 mg/kg using the same reagents and suppliers ( Parkes et al . , 2017 ) . All animals were submitted to surgery then allocated to CNO or saline groups on a random basis prior to training . Rats were perfused transcardially with 150 ml of saline followed by 400 ml of 4% paraformaldehyde ( PFA ) . Brains were kept in the same PFA solution overnight , then sections of 40 µm of the prefrontal cortex and the thalamus were made using a vibratome . Immunochemistry was performed on the sections to enhance the mCherry staining . First , sections were rinsed in PBS 0 . 1M ( 5 × 5 min ) , and then incubated in a blocking solution for 1 hr ( 4% goat serum and 0 . 2% Triton X-100 in PBS 0 . 1 M ) . Immediately after , sections were put in a bath containing primary antibodies , rabbit anti-RFP ( Clinisciences , PM005 ) primary antibodies diluted at 1/200 in the blocking solution for incubation at 4°C for 48 hr . Sections were then rinsed in PBS 0 . 1 M ( 4 × 5 min ) and placed for 2 hr in a bath containing a goat anti-rabbit coupled to DyLight 549 ( 1/200 in PBS 0 . 1 M ) ( Jackson ImmunoResearch , 111-025-003 ) for two hours . Following four 5 min rinses in PBS 0 . 1 M , Hoechst solution ( bisBenzimide H 33258 , Sigma , B2883 ) for counterstaining was added for 15 min ( 1/5000 in PBS 0 . 1 M ) . Finally , sections were rinsed in PB 0 . 1M ( 4 × 5 min ) , mounted in PB 0 . 05 M onto gelatin-coated slides and coverslipped with the anti-fading reagent Fluoromount G ( SouthernBiotech , 0100–01 ) . Images were then captured using a Nanozoomer slide scanner ( Hamamatsu Photonics ) and analyzed with the NDP . view 2 . 0 freeware ( Hamamatsu Photonics ) . Histology was performed by FA , VF and MW independently , while being blind to behavioral data . The data were submitted to ANOVAs on StatView software ( SAS Institute Inc . ) . For both experiments , Drug ( saline/CNO ) was the between subject factor , and Devaluation ( Devalued/Non Devalued ) , Degradation ( Degraded/Non degraded ) and Session ( averaged over both actions ) were repeated measures when appropriate . The alpha value for rejection of the null hypothesis was 0 . 05 throughout . | Planning and decision-making rely upon a region of the brain called the prefrontal cortex . But the prefrontal cortex does not act in isolation . Instead , it works together with a number of other brain regions . These include the thalamus , an area long thought to pass information on to the cortex for further processing . But signals also travel in the opposite direction , from the cortex back to the thalamus . Does the cortex-to-thalamus pathway carry the same information as the thalamus-to-cortex pathway ? To find out , Alcaraz et al . blocked each pathway in rats performing a decision-making task . The rats had learned that pressing a lever led to one type of reward , whereas moving a rod led to another . Alcaraz et al . reduced the desirability of one of the rewards by giving the rats free access to it for an hour . Afterwards , the rats opted mainly for the action associated with the reward that had remained desirable . However , blocking either the thalamus-to-cortex or cortex-to-thalamus pathway prevented this preference from emerging . This suggests that an information flow in both directions is necessary to update knowledge about the value of a reward . In a second experiment , Alcaraz et al . removed the link between one of the actions and its reward . The reward instead appeared at random , irrespective of the rat’s own behavior . Control rats responded by focusing their efforts on the action that still delivered a reliable reward , and by performing the other action less often . Blocking the thalamus-to-cortex pathway prevented this response , but blocking the cortex-to-thalamus pathway did not . This suggests that only the former pathway is necessary to re-evaluate the relationship between an action and an outcome . Two key aspects of goal-directed behavior – recognizing the value of a reward and the link between an action and an outcome – thus depend differently on the thalamus-to-cortex and cortex-to-thalamus pathways . This same principle may also be at work in other neural circuits with bidirectional connections . Understanding such principles may lead to better strategies for treating disorders of brain connectivity , such as schizophrenia . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Thalamocortical and corticothalamic pathways differentially contribute to goal-directed behaviors in the rat |
We exploit the reduced space of C . elegans postures to develop a novel tracking algorithm which captures both simple shapes and also self-occluding coils , an important , yet unexplored , component of 2D worm behavior . We apply our algorithm to show that visually complex , coiled sequences are a superposition of two simpler patterns: the body wave dynamics and a head-curvature pulse . We demonstrate the precise Ω-turn dynamics of an escape response and uncover a surprising new dichotomy in spontaneous , large-amplitude coils; deep reorientations occur not only through classical Ω-shaped postures but also through larger postural excitations which we label here as δ-turns . We find that omega and delta turns occur independently , suggesting a distinct triggering mechanism , and are the serpentine analog of a random left-right step . Finally , we show that omega and delta turns occur with approximately equal rates and adapt to food-free conditions on a similar timescale , a simple strategy to avoid navigational bias .
Much of our fascination with the living world , from molecular motors to the dynamics of entire societies , is with emergence — where the whole is surprisingly different than the sum of its parts ( see , e . g . , [Laughlin , 2014] ) . Yet , the existence of such collective organization also suggests that living systems , despite their enormous potential complexity , often inhabit only a much smaller region of their potential ‘phase space’ , and evidence for this lower-dimensional behavior is ubiquitous . For example , the motor control system produces movements that are far less complex than what the musculoskeletal system allows ( d'Avella et al . , 2003 ) and this hints at the presence of an organizational principle . In a typical daily movement like walking , the central nervous system is thought to produce the full walking gait by combining low-level ‘locomotory modules’ , some of which appear to be universal among species ( Dominici et al . , 2011 ) . Similarly , the dynamics in brain networks are organized in low-dimensional activity patterns ( Tkačik et al . , 2014; Gao and Ganguli , 2015 ) and these patterns — not individual neurons — might be the carriers of information and computation ( Hopfield , 1982; Yoon et al . , 2013 ) . The emergent dynamics of behavior , how animals move and interact , is particularly important as the ultimate function of the system ( Tinbergen , 1963 ) and the scale on which evolution naturally applies . Yet , our quantitative understanding of behavior is substantially less advanced than the microscopic processes from which it is produced , even as recent efforts have expanded this frontier ( Mirat et al . , 2013; Berman et al . , 2014; Cavagna and Giardina , 2014 ) . How do we analyze high-resolution behavioral dynamics and what does this reveal about an animal’s movement strategy ? How do we build effective models on the behavioral level where a ‘bottom-up’ approach is daunting ? How do we connect analysis on the organism-scale to the properties of molecules , cells and circuits ? We approach these questions through the postural movements of the nematode C . elegans . In C . elegans , the 2D space of body postures can be captured precisely and is also low-dimensional ( Stephens et al . , 2008 ) so that the worm’s motor behavior is faithfully encoded as a time series of only four ‘eigenworm’ variables . These shape projections are collective coordinates in the space of natural worm postures and provide a notable reduction in complexity . However , an important limitation of previous work is the inability to deduce the geometry of self-occluding body shapes . Such coiled body postures occur during ‘omega turns’ ( a maneuver during which the worm’s body briefly resembles the Greek letter Ω [Croll , 1975] ) and are a general and important feature of the worm’s behavioral repertoire , ranging from foraging ( Stephens et al . , 2010; Salvador et al . , 2014 ) , and chemotaxis ( Pierce-Shimomura et al . , 1999 ) , to escape from noxious stimuli ( Mohammadi et al . , 2013 ) . For example , during escape behaviors worms use coiled shapes to reorient by 180∘ and the benefit seems obvious: it steers the worm back to safety . But how does a ‘blind’ organism achieve this result without any visual reference to the outside world ? While some of the neural and molecular mechanisms driving omega turns have been uncovered ( Gray et al . , 2005; Donnelly et al . , 2013 ) and there has been previous work on crossed shapes ( Huang et al . , 2006; Wang et al . , 2009; Roussel et al . , 2014; Nagy et al . , 2015 ) , a quantitative analysis of such self-occluded posture dynamics is lacking . Here , we exploit low-dimensionality to develop a novel and conceptually simple posture tracking algorithm able to unravel the worm’s self-occluding body shapes . We apply our approach to analyze coiled shapes during two important behavioral conditions: the escape response induced by a brief heat shock to the head , and spontaneous turns while foraging on a featureless agar plate . We find that , in general , complex deep turn sequences can be viewed as a simpler superposition of body wave phase dynamics with a bimodal head swing followed by a unimodal curvature pulse . In the escape response we show that , while turning accounts for much of the ~180° reorientation , the full distribution of reorientation angles is shaped by significant contributions from the reversal , turn and post-turn behaviors , a result consistent with the presence and action of the monoamine tyramine during the entire response ( Donnelly et al . , 2013 ) . In natural crawling , the peak amplitudes of the curvature pulse reveal two distinct coiling behaviors — the classical omega turn accomplishing large ventral-side reorientations , and a previously uncharacterized ‘delta’ turn which produces dorsal reorientations by overturning through the ventral side . The omega and delta turns occur independently in time , suggesting a separate triggering process , but have similar rates , as expected if they contribute little overall bias in the trajectories .
Previously , we analyzed movies of C . elegans freely crawling on an agar plate ( Figure 1A ) ( Stephens et al . , 2008 ) . For each movie frame , we identified the body of the worm , and applied a thinning algorithm to find the centerline . The worm’s 2D body posture was characterized as a 100-dimensional vector of tangent angles along this centerline ( Figure 1B–C ) . Principal Component Analysis revealed that more than 95% of the variance in naturally-occurring body postures was captured by just four eigenvectors of the posture covariance matrix ( Figure 1D ) . As a result , any worm posture can be decomposed as a linear combination of these ‘eigenworms’ ( Figure 1E ) . Worm behavior then becomes a smooth , low-dimensional trajectory through posture space ( Figure 1F ) . As an example , forward and backward crawling appear as approximately circular trajectories in the ( a1 , a2 ) plane , and correspond to limit-cycle attractors . However , for coiled shapes such as shown in Figure 1H , the thinning algorithm does not produce a faithful reconstruction of the worm’s actual posture ( Figure 1G ) . 10 . 7554/eLife . 17227 . 003Figure 1 . Inverting posture analysis to generate worm images . ( A–E ) We previously showed that the space of C . elegans body postures is low-dimensional . ( A ) For a set of images of a freely moving worm , ( B ) we find the centerline of the body using image thinning ( black point indicates the head ) . ( C ) At equidistant points along the centerline , we measure the direction θ ( s ) of the tangent t^ . After subtracting ⟨θ⟩ , this gives a description of the worm’s shape that is intrinsic to the worm itself . ( D ) Principal Component Analysis reveals that only four eigenvectors of the shape covariance matrix are needed to account for ∼95% of the variance in θ ( s ) . ( E ) Hence , any body shape can be decomposed as a linear combination of postural ‘eigenworms’ . ( F ) Alternatively , we can think of any body posture as a point in a low-dimensional ‘posture space’ , spanned by the eigenworms ( gray ) . Forward crawling is then represented by clockwise progression along a circular trajectory in the ( a1 , a2 ) plane ( blue oval , body wave phase angle φ ) . ( G ) For any point in this space , we can easily calculate the shape of the backbone . A series of filled circles with radii representing the worm’s thickness , are used to draw an image of the worm’s body ( H ) , inverting the original postural analysis to generate an image . For self-overlapping shapes , image thinning ( H , magenta ) does not produce an accurate reconstruction of the posture ( G , red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 003 The above procedure can also be implemented in reverse to generate worm images . For any point 𝒑 in posture space ( Figure 1F ) , we can reconstruct the shape of the backbone ( Figure 1G ) . Knowing the thickness of the worm at each point along the body ( which we estimate by averaging over many worm images ) , we are able to draw a reconstructed body image ( Figure 1H; see Materials and methods ) . We then track the posture by finding , for each movie frame , the point in posture space ( and thus the correct centerline ) for which the reconstituted worm image is the most similar to the original image . This approach works for all worm postures — in contrast to image thinning , which fails for self-overlapping shapes ( Figure 1H ) . Our ‘inverse’ tracking algorithm consists of three basic elements . ( i ) An image error function ferr quantifies how well a reconstituted worm image 𝐖~ ( 𝒑 ) matches the movie frame 𝐖 ( Figure 2A ) ; ( ii ) an efficient optimization scheme to search for a global error minimum over all possible postures , and; ( iii ) a method to resolve ambiguity , as different self-occluding body shapes can give rise to the same image . We measure image similarity using two specific shape metrics ( Yang et al . , 2008 ) : outline shape , and coarse-grained pixel density ( Figure 2A ) . By mapping this error function onto posture space: ferr ( 𝒑 ) =ferr ( 𝐖 , 𝐖~ ( 𝒑 ) ) , we create a fitness landscape , in which the position of the global minimum corresponds to the tracking solution . We find this minimum using a pattern search algorithm ( a form of direct search [Kolda et al . , 2003] ) . To resolve ambiguity , we retain multiple minima for each frame , until a final step which minimizes total sequence error . We sketch this process for a single mode in Figure 2C . 10 . 7554/eLife . 17227 . 004Figure 2 . Tracking coiled shapes by searching for image matches in posture space . Top: tracking algorithm sequence . ( A ) For each movie frame 𝐖 and reconstituted worm image 𝐖~ ( 𝒑 ) for posture 𝒑 , we apply two metrics , one based on the shape of the boundary ( left ) , and one based on a coarse-grained pixel density matrix ( right ) . ( B ) An error function ferr based on these two shape metrics generates a fitness landscape ( schematically shown ) . The position of the global minimum of ferr corresponds to the tracking solution; if a frame is ambiguous , multiple minima may be present . ( C ) For non-crossed body postures , a simple image thinning algorithm suffices to obtain time series of the modes ai ( blue line , schematically shown ) . For crossed frames , we use the procedure outlined in A–B . Due to the inherent ambiguity of such images , multiple solutions are generally found for each frame ( light gray points ) . Using the filtering algorithm described in the Materials and methods , we identify the correct solutions ( dark gray points ) . The resulting smooth trajectory ( magenta , dotted line ) forms the full tracking solution . ( D ) Sample tracking results ( bottom , white background ) , contrasted with original images ( top , gray background ) , for a turning sequence . Bottom: the inverse algorithm accurately tracks both simple and coiled worm shapes with small error . ( E ) Histogram of tracking errors for non–self-overlapping worm shapes , quantified as the Euclidean distance δθ between the tangent angle vector 𝜽 from our algorithm , and 𝜽 found by image thinning ( magenta ) . For scale , the error due to dimensionality reduction to five postural eigenmodes is shown in black . We also show the Euclidean distance between 𝜽 in consecutive frames , representing the confidence in 𝜽 due to the finite time resolution of the movie ( gray ) . Even for an extreme value of δθ=3rad ( gray arrow ) , backbones from the ‘classic’ algorithm ( top ) and our algorithm ( bottom ) are nearly indistinguishable by eye ( inset ) . ( F ) Tracking error in eigenmode values for the first four modes . For uncrossed worm shapes ( yellow/light ) , our algorithm shows negligible tracking errors . For a smaller set of crossed frames , we compare to a manually found solution ( blue/dark ) . For scale , we show reconstituted images for worms with a single nonzero mode value of ai=1; these ‘error worms’ are essentially flat . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 00410 . 7554/eLife . 17227 . 005Figure 2—figure supplement 1 . The eigenworms {ek=1…4} derived from the fully-tracked data show only minor changes compared to those computed without crossings . We show the original eigenworms ( blue ) ( Stephens et al . , 2008 ) ; as well as those derived from the current foraging data , but without crossings ( red ) ; and those derived from the fully-tracked foraging data ( yellow ) . While the eigenworm shapes are largely similar , and in all cases four modes capture over 95% of the postural variance , the third ( turning ) eigenmode individually accounts for more variance in the fully-tracked data , as expected given its primary role in deep turns . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 005 Tracking results for a typical movie that includes complex , self-occluding shapes are shown in Figure 2D ( see also Videos 1 and 2 ) . In the gray rows at the top are the original movie frames; the reconstituted images from our inverse algorithm are below . While some minor inaccuracies are visible by eye , the overall result is remarkably similar . To quantify posture tracking accuracy , we first compared the results of our algorithm to image thinning , which allows for verification based on a large dataset . We used image thinning to construct a 100-dimensional vector of tangent angles 𝜽 , defined the tracking 'error' as δθ=∥𝜽inv-𝜽thinning∥ , and we plot the distribution of these errors in Figure 2E ( magenta ) . We also show the discrepancy in 𝜽 that results from dimensionality reduction to the postural eigenmodes ( black ) . Additionally , we show Euclidean distances between tangent angle vectors of consecutive frames in a 16 Hz movie , representing limited time resolution ( gray ) . For this dataset of non-crossed frames , our algorithm provides excellent performance , with tracking errors bounded by time resolution and dimensionality reduction . Even for deviations in the tail of the distribution ( δθ=3rad ) , backbones from the thinning and the ‘inverse’ algorithm are quite similar ( inset , gray backbones ) . 10 . 7554/eLife . 17227 . 006Video 1 . Tracking results for the escape response . Left images are data , while on the right , there are reconstructed images from our tracking algorithm . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 00610 . 7554/eLife . 17227 . 007Video 2 . Tracking results for a complex , spontaneous coil . Left images are data , while on the right , there are reconstructed images from our tracking algorithm . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 007 A more relevant quantity for low-dimensional trajectories is the mode discrepancy δai=∥aiinv−aithinning∥ which is negligible for simple shapes , as shown in Figure 2F ( yellow ) . Finally , we created a dataset of self-overlapping body shapes for which backbones were manually drawn . In Figure 2F ( blue ) , we show that , for the majority of crossed frames , the mode error is less than 10% of the total range of naturally occurring mode values . As a visual reference , the reconstituted worm shapes corresponding to mode errors of δai=1 are shown in gray: these are noticeably flat . We first applied our postural tracking algorithm to quantify the full shape dynamics of the C . elegans ‘escape response’ . This is a stereotyped behavioral sequence , consisting of a pause , a reversal and an Ω-turn , that quickly moves the worm away from a threatening stimulus . Featuring only relatively simple coiled shapes , the escape response provided a useful test of our algorithm . While recent work has connected the escape response with genetic , molecular , and neural mechanisms ( Donnelly et al . , 2013 ) , the behavior itself has been described only qualitatively . Here , we elicited an escape response by using an infrared laser pulse administered to the head of the worm , which raised the temperature by ~0 . 5°C . 10 s of pre-stimulus behavior and 20 s of post-stimulus behavior were recorded at 20Hz . Each worm was only assayed once , to prevent adaptation . In total , N=92 worms were recorded , of which N=91 successful trackings were used in the final analysis . A schematic of the response is shown in Figure 3A , with the associated postural mode dynamics in Figure 3B , C . During normal , forward locomotion ( i in Figure 3A , t<10s in Figure 3C ) , the worm crawls by propagating a sine-like wave through its body . This is reflected as a pair of phase-locked sinusoidal oscillations in a1 and a2 and we define the body wave phase angle φ=-arctan ( a2/a1 ) , where the minus sign ensures that dφ/dt is positive during forward crawling . When the worm is stimulated by the infrared pulse ( ii in Figure 3A , pink line in Figure 3C at t = 10 s ) , it immediately backs up ( iii ) , seen as a decrease in φ . The end of this reversal and the beginning of the Ω-turn is marked by a head-swing , visible as a bimodal pulse in a4 . The Ω-turn itself ( iv ) occurs as a large , unimodal pulse in a3 , and propagates head-to-tail . This implies another switch of the direction of the body wave , and hence a return to increasing φ . Finally , as the turn is finished , the worm resumes forward crawling ( v ) . The mode dynamics outlined above illustrate that the complexity of the escape sequence can be seen as a superposition of two simpler patterns: the body wave phase dynamics in ( a1 , a2 ) , and the head-curvature dynamics of ( a3 , a4 ) . An animation of these mode dynamics is available as Video 3 . 10 . 7554/eLife . 17227 . 008Figure 3 . Tracking coiled postures and reorientation in the escape response . ( A ) Schematic overview ( Donnelly et al . , 2013 ) with worm body shapes extracted from tracking data: i forward locomotion and exploratory head motions; ii infrared laser stimulus; iii reversal phase; iv omega turn; v resumption of forward locomotion in the opposite direction . ( B ) Trajectory through posture space . φ indicates direction of increasing body wave phase angle , and color encodes time , with blue for t = 0 and red at t = 30 s . The worm’s reorienting coiling behavior is evident as a large excursion along the third mode , starting at the red arrow . ( C ) The same trajectory as in B , in terms of the body wave phase angle φ and the postural modes ( a3 , a4 ) . The heat shock occurs at t = 10 s ( pink bar ) . The omega turn is initiated by a head swing , as seen in a4 , followed by a large pulse in a3 , and is linked to a ‘re-reversal’ , a return to forward movement . ( D ) An important feature of the escape response is the change in the worm’s overall orientation , and we apply our algorithm to track this reorientation for each response segment . While turning accounts for much of the reorientation , the full response distribution is shaped by significant contributions from all three segments . In particular , the small but biased reorientations of the reversal and post-turn segments originate in the a3 fluctuations outside the turn ( see the time series in C and also Figure 3—figure supplement 1 ) . This is consistent with the release and presence of the monoamine tyramine during the entire response . ( E ) The precision of the escape response is evident in the trial-mean reorientation where we find ⟨Δθ⟩=-0 . 89π ± 0 . 05πrad for the full response and ⟨Δθ⟩=-0 . 97π ± 0 . 04πrad if we exclude ( four ) worms that only make small dorsal reorientations . Notably , the mean reorientation in the reversal and post-turn segments closely cancel , suggesting a correction mechanism at the level of the average response . In the inset to ( D , Omega turn ) , we also show the distribution of a3 amplitudes , which is peaked near coiled shapes in which the worm barely touches . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 00810 . 7554/eLife . 17227 . 009Figure 3—figure supplement 1 . Bias in the turning mode a3 , and resulting reorientation , occurs during all epochs of the escape response . From previous work on the interpretation of the postural eigenmodes , we know that the third eigenmode ( an overall bending of the worm ) is linked to reorientation of the worm ( Stephens et al . , 2008 , 2010 ) . We therefore tested if any asymmetry in the fluctuations of a3 during the reversal phase could be linked to the observed reorientations . Such asymmetry is also visible in Figure 3C as a baseline shift of the third mode during the reversal . ( Top , left & right ) The mean a3 value , versus the resulting orientation change , during the reversal and post-omega behaviors , respectively . The orientation change is strongly correlated with the mean a3 value . ( Top , middle ) Peak amplitude of the a3 peak corresponding to the omega turn , versus the resulting orientation change . ( Bottom , left & right ) Histogram of mean a3 values , during the reversal phase and post-omega phase , respectively , showing similarly asymmetric distributions . ( Bottom , middle ) Histogram of a3 peak amplitudes during the omega turn . We also show a reconstituted worm image for an a3 value of 15 , for which the worm is barely self-occluded . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 00910 . 7554/eLife . 17227 . 010Video 3 . The dynamics of the escape response in the space of the first three eigenworms . On the right , we show the full body posture , which turns red at the moment of the thermal stimulus . On the left are the dynamics in mode space . The large-amplitude omega turn is visible as a ‘figure-8’ trajectory . Note that , even during the turn , the body wave is progressing . In general , turning behavior is a superposition of the body wave and curvature dynamics . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 010 A notable feature of the escape response is how closely the worm controls its reorientation . Our tracking algorithm also makes it possible to track the overall orientation continuously , across the different phases of the escape response . In Figure 3D–E , we calculate how much each of the three response segments reorients the worm . The distribution of reorientations for the full escape response is largely similar to the distribution during the omega turn , but includes contributions from the reversal and post-turn segments . In the trial-averaged reorientation Figure 3E , we find ⟨Δθ⟩=-0 . 89π ± 0 . 05πrad for the full response . The omega turn itself results in ⟨Δθ⟩=-0 . 90π ± 0 . 04πrad , while pre- and post-omega phases show smaller but significant contributions , ⟨Δθ⟩=0 . 13π ± 0 . 03πrad and ⟨Δθ⟩=-0 . 12π ± 0 . 03πrad , respectively ( errors are calculated using bootstrap across trials and are equivalent to standard errors of the mean ) . In Figure 3D , the interval ( 0 , -π ) corresponds to a final ventral-side reorientation , and ( -π , -2π ) to a final dorsal-side reorientation . The small number of reorientations between ( 0 , π ) are also final dorsal-side reorientations but are achieved using a shallow dorsal bend , not an omega turn , and excluding these worms results in a total mean reorientation angle ⟨Δθ⟩=-0 . 97π ± 0 . 04πrad . 10 . 7554/eLife . 17227 . 011Figure 4 . Unraveling coiled shapes during foraging reveals two distinct ventrally-biased classes of large-amplitude turns . ( A ) ( left ) Probability of the amplitude of all local extrema in the time series of the third postural eigenmode a3 . Colors represent the sign of the a3 amplitude , and hence the dorsal ( gray ) or ventral ( blue ) direction of the resulting turn . ( A ) ( right ) As previously , with all negative a3 amplitudes now plotted as positive . The peaked excess in the distribution for large ventral bends corresponds to ‘classic’ Ω ( omega ) shapes , and previously undescribed , deeper δ ( delta ) turns . Insets in A ( left ) show reconstructed worm shapes for the indicated a3 amplitudes . ( B ) Stills from a movie of a worm making a classical omega turn ( left , yellow ) , and a deep delta turn ( right , blue ) . The head is marked with a red dot; dashed lines indicate postures determined from our inverse tracking algorithm . The dynamics of delta turns are largely similar to omega turns , differing primarily in the amplitude of the bending mode a3 , and the overall time to complete the maneuver ( see Figure 4—figure supplement 1 ) . ( C ) Histogram of orientation change ( Δ⟨θ⟩ ) due to ventral omega turns ( yellow/light ) and ventral delta turns ( blue/dark ) . Ventral reorientations are accomplished through omega turns . To reorient to the dorsal side , however , C . elegans employs delta turns , which ‘over-turn’ through the ventral side . ( D ) Average turning rate during the tracking experiment . Ventral omega and delta turns are temporally independent , suggesting a separate triggering mechanism , but occur with approximately equal rates that adapt similarly with time spent away from food , a simple strategy to avoid any dorsal-ventral navigational bias . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 01110 . 7554/eLife . 17227 . 012Figure 4—figure supplement 1 . Omega and delta turns follow similar kinematics; while visually quite distinct , the primary difference is the amplitude of the curvature pulse a3 . ( A ) Typical time series for the postural eigenmodes a1 . . 4 during a deep delta turn . φ is the body wave phase angle , as defined in Figure 1F . Shaded area indicates the delta turn , as defined in the Materials and methods . ( B ) Average eigenmode time series during a delta turn ( blue , N = 348 ) . Gray lines indicate SD . For comparison , the average escape response omega turn is also shown ( red dotted line ) . Time has been normalized with respect to the total length of the turn: 6s ± 2s ( mean ± SD ) for delta turns , 7s ± 3s for escape-response omega turns . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 01210 . 7554/eLife . 17227 . 013Figure 4—figure supplement 2 . The shifted mutual information between delta turn and omega turn time series . Events are localized by the large amplitude peaks in the curvature mode a3 , and bins from 2 to 20 s are used to convert these timings into a binary event series . The mutual information is very small in all cases ( for scale , the entropy of either event series is ∼1 bit ) indicating that the two different turn types occur independently . Dashes are estimated errors from finite sampling . Red lines denote the mutual information between shuffled time series , which would be zero in the infinite data limit . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 01310 . 7554/eLife . 17227 . 014Figure 4—figure supplement 3 . ( left ) Location of one of the 12 tracked worms over the course of a 35-min tracking experiment ( off-food ) , starting at ( 0 , 0 ) ( black arrow ) . Omega turns ( orange ) and delta turns ( blue ) are highlighted . ( right ) An enlargement of the area marked in gray . While both turns occur more frequently at early times , there is no correlated pattern , as consistent with their independence . DOI: http://dx . doi . org/10 . 7554/eLife . 17227 . 014 Remarkably , the mean reorientation in the reversal and post-turn segments precisely cancel , suggesting a correction mechanism at the level of the average response so that the mean overall reorientation is entirely determined by the omega-turn . No such precision is apparent in the variance , where we find δθ2=0 . 69π ± 0 . 16πrad2 for the full response compared to the smaller δθ2=0 . 45π ± 0 . 16πrad2 for the turn segment . Thus , while the omega turn is an effective maneuver for turning away from the stimulus , the full response orientation change is broadened by the reversal δθ2=0 . 23π ± 0 . 05πrad2 and post-omega δθ2=0 . 19π ± 0 . 04πrad2 behaviors . These observations allow us to hypothesize a subtle link between the behavior of the worm and the escape response at the neurotransmitter level ( Donnelly et al . , 2013 ) . As the worm enters the reversal phase , release of tyramine sets up an asymmetry in the worm’s body , and this appears as a baseline shift in the fluctuations of the third mode ( see also Figure 3—figure supplement 1 ) leading to a positive bias in the reorientation , Figure 3D , E ( reversal ) . After the turn , lingering effects of the tyramine produce a similar baseline shift , but as the worm is moving forward instead of backward , this now leads to an opposite orientation bias , Figure 3D , E ( post-omega ) . To analyze more complex coiled shapes , we applied our posture algorithm to foraging worm behavior on a flat agar plate . Under these conditions , worms navigate using a combination of maneuvers ( Gray et al . , 2005 ) , including short and long reversals , pirouettes and also gradual turns ( Iino and Yoshida , 2009 ) . We are particularly interested in the pirouettes , as they involve deep coils . Such body bends are primarily encoded in the third postural eigenmode ( a3 ) and , as discussed in the previous section , peaks in a3 are a characteristic feature of omega turns , and have a known role in reorientation of the worm ( Stephens et al . , 2010 ) . In Figure 4A , we show the full distribution of postural mode a3 for all local extrema . Note that the modes have been normalized so that negative a3 amplitudes correspond to dorsal turns; ventral turns have strictly positive amplitudes . A clear asymmetry can be observed so that on top of a symmetric background distribution of shallow turns in both directions , we see , on the ventral side , two distinct additional peaks . Drawing reconstituted worm images for the center values of these two peaks , it is clear that the peak at a3∼15 corresponds to a ‘classic’ Ω shape . The second peak , at a3∼23 , shows a body shape with a much higher characteristic curvature . In Figure 4A ( right ) , we have ‘folded’ the dorsal side of the distribution over the ventral side , highlighting the ventral asymmetry at high a3 amplitudes . As noted in the figure , we refer to turns in the lower-amplitude peak as omega turns and distinguish these from the higher-amplitude delta ( δ ) turns in the second peak . As for the omega turn , the name delta turn is chosen to reflect the δ-like shape of the worm during a typical sequence . Returning to the original tracking movies , the presence of these two classes of turns is clearly visible . In Figure 4B , we display movie stills for two example turns: one omega turn , and one delta turn . During the classical omega turn , the worm slides its head along its body , similar to the escape response , ending up with a large , primarily ventral reorientation . A delta turn , on the other hand , is much deeper: the worm completely crosses its head over its body , resulting in a dorsal reorientation by ‘over-turning’ across the ventral side . In postural dynamics , δ- and Ω-turns differ primarily in their a3 pulse amplitude; their turn kinematics are otherwise very similar ( Figure 4—figure supplement 1 ) . However , when turns do occur , they result in a dramatically different change of overall orientation . As in the escape response , we use our algorithm to track the worm’s overall body reorientation , and in Figure 4C , we show how the worm reorients using both omega ( orange ) and delta ( blue ) turns . Simply put , omega turns reorient the worm by large , ventral angles , while delta turns reorient the worm dorsally by ‘over-turning’ through the ventral side . The difference in reorientation angle may provide a hint as to why these two behaviors exist . Earlier , we saw that the neural mechanisms that produce the escape-response omega turn , are fundamentally asymmetric , producing only ventral turns ( through disinhibition of the VD motor neurons ) ( Donnelly et al . , 2013 ) . If the worm uses the same neural infrastructure during free crawling , this would only ever allow it to reorient itself towards its ventral side . Lacking a dorsal ‘copy’ of the same neural infrastructure , the worm could instead hyper-activate the existing infrastructure to produce ventral ‘over-turning’ . These ‘over-turns’ are what we call delta turns , and enable the worm to also reorient towards its dorsal side . We also find that delta and omega turns occur seemingly independently; the mutual information between time-binned , time-shifted series for both turning event time series has a maximum of less than a few percent ( see Materials and methods and Figure 4—figure supplements 2 and 3 ) . On the other hand , evidence that the turns can be jointly controlled is shown in Figure 4D . Here , we plot the frequency of turning events over the course of the experiment . As the worm searches for food in a larger area , the turn frequency decreases significantly — a well-known phenomenon ( Gray et al . , 2005; de Bono and Villu Maricq , 2005; Srivastava et al . , 2009 ) — and both omega and delta turns show similar frequencies and adaptation .
The ability to track self-overlapping shapes of C . elegans together with the eigenworm projection of postures , provides a complete and quantitative accounting of the worm’s locomotory behavior in 2D . Among living systems with a nervous system , such an exact behavioral description is unique , and is likely to be especially important as new techniques emerge for the simultaneous imaging of a substantial fraction of the worm’s neurons during free behavior ( Nguyen et al . , 2016; Venkatachalam et al . , 2016 ) . Our posture tracking algorithm itself is conceptually simple and relies on an optimized image search within the low-dimensional space of worm shapes . Indeed , while the identification of low-dimensionality occupies an important role in quantitative approaches to living systems ( see e . g . Machta et al . , 2013; Daniels and Nemenman , 2015; Ganguli and Sompolinsky , 2012 ) , here we have leveraged low-dimensionality to elucidate important and previously unknown aspects of C . elegans coils . Interestingly , we were able to apply the characterization of body postures developed previously for non–self-overlapping body shapes ( Stephens et al . , 2008 ) , to capture shapes that do self-overlap; even the simpler eigenworm space allows for substantial postural diversity . We applied our tracking algorithm to two important behaviors: an evoked escape response; and the deep , spontaneous turns that occur during foraging . Viewing the coiled turn as a trajectory through the low-dimensional posture space , a simple model emerges: a superposition of the body wave ( a circular trajectory in posture space corresponding to simple forward and backward crawling ) , and coupled pulses along the third and fourth mode ( corresponding to the deep coil and a preceding head oscillation ) . This model is consistent with the molecular mechanisms found to orchestrate the escape response ( Donnelly et al . , 2013 ) . Our results also hint at a possible answer as to how reorientations of 180° are accomplished: the worm could use its own body as a ‘guide’ for reorientation . During the omega turn , the distribution of a3 peak amplitudes ( Figure 3D [Omega turn , inset] ) lies close to a value of 15: the lowest a3 value that generates a self-touching body shape . This suggests that the worm might have evolved to coil until it just intersects its own body , which it then slides along to find its way back . While the omega turn has previously been considered as a single class of C . elegans behavior , our analysis of the amplitudes of the curvature mode a3 pulses associated with deep coils , reveals the presence of distinct subpopulations . In foraging , we show that ‘classic’ omega turns , featuring the signature Ω body shape , primarily reorient the worm to the ventral side , while delta turns reorient the worm dorsally by over-turning through the ventral side . These deep dorsal and ventral reorientations occur independently in time with approximately equal rates , which is important if there is to be no overall bias in the trajectories . On the other hand , in an evoked escape response , we observed only Ω-type turns with reorientations of ∼180° . While distinct in visual appearance , omega and delta turns differ only in the amplitude of the curvature mode , and we have shown that these behaviors are discretely separable during foraging . Interestingly , the neuronal basis for omega bend initiation and execution has been studied in some detail ( Gray et al . , 2005 ) , where in particular the SMD and RIV motor neurons are , respectively , implicated in the amplitude and the ventral bias of the turn . Coiling is also observed in other contexts , including a variety of mutants ( Yemini et al . , 2013; Nagy et al . , 2015 ) , and we expect that our methods will be useful in further analyzing such shapes , and as a guide for uncovering coiling behavior . Deep turns and reorientations form an important component of the taxis strategy of C . elegans ( Croll , 1976; Pierce-Shimomura et al . , 1999; Gray et al . , 2005; Stephens et al . , 2010; Salvador et al . , 2014 ) . Under foraging and chemotaxis conditions , these behaviors are seemingly stochastic ( Srivastava et al . , 2009; Gallagher et al . , 2013 ) , producing a broad distribution of reorientation angles analogous to tumbling in the bacteria E . coli ( Berg and Brown , 1972 ) . However , unlike bacterial tumbling ( which occurs through an instantaneous switch in the rotation direction of a molecular motor and the resulting unbundling of the flagellar tail , see , e . g . , Berg , 2006 ) the worm’s reorientation is driven by a long , controlled sequence of stereotyped postural changes . Thus , an important question is how the worm effectively randomizes its direction of motion . We have shown here that half the variability in C . elegans foraging reorientations is due simply to the initial random choice of delta or omega turns . However , even the level of stochasticity can be modulated , as evidenced by the largely deterministic reorientation in the escape response , differing response variability depending on the strength of a thermal stimulus ( Mohammadi et al . , 2013 ) , and the slow adaptation of the reversal rate ( Gray et al . , 2005; Stephens et al . , 2011 ) . Overall , such a combination of behaviors , flexible and stochastic combined with patterned and deterministic , is likely to be observed even in more complex organisms , including humans . In initiating the detailed analysis of C . elegans turning behavior , we hope that our work offers a first step towards a general understanding of these processes .
We used two datasets encompassing both foraging and escape response behavioral conditions ( Broekmans et al . , 2016a ) . The foraging data were explored previously ( Stephens et al . , 2011 ) ; for more details on data collection , see also ( Stephens et al . , 2008 ) . In short , young L4-stage C . elegans N2-strain worms were imaged with a video tracking microscope at f=32Hz . Worms were grown at 20°C under standard conditions ( Sulston and Brenner , 1974 ) . Before imaging , worms were removed from bacteria-strewn agar plates using a platinum worm pick , and rinsed from E . coli by letting them swim for 1 min in NGM buffer . They were then transferred to an assay plate ( 9-cm Petri dish ) that contained a copper ring ( 5 . 1 cm inner diameter ) pressed into the agar surface , preventing the worm from reaching the side of the plate . Recording started approximately 5 min . after the transfer , and lasted for 2100 s ( 35 min ) . In total , data from N = 12 worms was recorded . The second dataset , the ‘escape response’ condition , was recorded following procedures as described in ref . ( Mohammadi et al . , 2013 ) . In short , worm recordings took place in a temperature-controlled room ( 22 . 5°C ± 1°C ) . A 100 ms , 75-mA infrared laser pulse from a diode laser ( λ = 1440 nm ) was administered to the head of the worm , raising the temperature in a FWHM-radius of 220 m by ∼0 . 5°C . 10 s of pre-stimulus behavior and 20 s of post-stimulus behavior were recorded at a frame rate of 20 Hz . Each worm was only assayed once , to prevent adaptation . In total , N = 92 worms were recorded , of which N = 91 successful trackings were used in the final analysis . All movie frames were converted to binary images and cropped , using standard image processing functions in MATLAB ( R2014b , The Mathworks , Natick , MA ) ( Stephens et al . , 2008 ) . For faster processing , before analysis with the inverse tracking algorithm , the foraging data was down-sampled to 16 Hz by dropping every second frame . To reconstitute an image of a worm with a body posture 𝒑= ( a1 , … , a5 ) , we first calculated the vector of backbone tangent angles from 𝛉=∑ipi𝒆^i , with 𝒆^i the i’th eigenworm . Knowing the total arc length l of the worm , we could calculate the position of each of the 100 points along the backbone . At each backbone point j , we then drew a filled circle with radius rj to capture the worm’s body thickness ( see also Figure 1G , H ) and thus create the worm image . Circle radii rj for a particular worm were computed from movies of uncrossed worm postures for that specific worm . In each such frame , after finding the centerline ( backbone ) and outline of the worm ( Stephens et al . , 2008 ) , we could find rj as the minimum distance between backbone point and outline . This was averaged across all frames . Similarly , the total arc length l of the worm was computed by averaging across frames . For the error function described below , the overall orientation of the worm in the image is important , and we generate images of worms in all possible orientations by adding an overall orientation value ⟨θ⟩∈[0 , 2π ) to the backbone tangent angle vector . This gives us a full backbone vector 𝛉F=⟨θ⟩+∑i=15ai𝒆𝒊^ . For the postural dynamics , the eigenworm shape projections were taken from Stephens et al . ( 2008 ) . Recomputing the eigenworms on the fully-tracked data here showed only minor changes ( see Figure 2—figure supplement 1 ) . The shape error function compares two binary worm images 𝐖1 and 𝐖2 , and is computed as ferr=foutline⋅fpixel . For foutline , we calculate a set of tangent angles ψ to the perimeter of the worm shape ( Figure 2A , bottom left ) . We find the 4-connected outline of the worm in the binary image 𝐖i , fit a spline through these points , and discretize it into 201 segments sampled at equal arc length . The 200 resulting angles between the segments form a vector 𝝍i= ( ψi , 1 , ψi , 2 , … , ψi , 200 ) ; the total length of the segments is ℓi . foutline is now foutline=C0|ψ1−ψ2|2+C1 ( ℓ1−ℓ2 ) 2 , for arbitrary constants C0 and C1 . Note that the value of foutline is sensitive to the choice of starting points for tracing the 4-connected outline in each image; this is resolved by choosing the pair of starting points that minimizes foutline . For fpixel , we first align the images 𝐖1 and 𝐖2 so that their centroids overlap . Each image is then divided into a grid of 10x10-pixel ‘blocks’ ( Figure 2A , bottom right ) . For each block ( j , k ) ( j=1 , … , n; ) in image 𝐖i , the fraction di ( j , k ) of black pixels in the block is calculated . This coarse-graining into blocks allows for , e . g . , minor inaccuracies in the generation of worm images from mode values , without affecting the error function . We then calculate fpixel as fpixel=1nm∑j , k ( d1 ( j , k ) -d2 ( j , k ) ) 2 . In earlier trials , we found that using five postural eigenmodes gave us significantly better tracking results than only using four . Since our error function is sensitive to the overall rotation of the worm , we amended the five-dimensional posture space with an extra dimension for the overall orientation ⟨θ⟩ . This means that the search space for our algorithm is six-dimensional , with 5 postural dimensions , and 1 rotational dimension . To find a tracking solution for a frame , we ran hundreds of pattern searches ( using MATLAB’s ‘patternsearch’ function ) from randomly distributed starting points in search space , with the error function described above as objective function . Only solutions with an error value less than 1 . 0 , a threshold value obtained through trial-and-error , were kept . Solutions within a given hypercube of dimensions [3 . 0 , 3 . 0 , 3 . 0 , 3 . 0 , 2 . 5] were merged , leaving only the solution with the lowest error value . This finally resulted in zero , one , or more potential tracking solutions per movie frame . To speed up the optimization , we applied two additional constraints . Firstly , we bounded the absolute value of the eigenmodes to ( 18 , 18 , 34 , 12 , 6 ) , for each of the five modes respectively . We verified that the distributions of eigenvalues ai found in our tracking data tailed off before reaching these limits . Secondly , we set a limit to the maximum local curvature of the worm’s backbone , so that elements in the resulting theta vector that are 10 indices apart must not be different by more than 1 . 95 rad . This limit rules out body shapes that were unnaturally coiled . Importantly , we note that our inverse problem is fundamentally ill-posed: multiple body postures may produce the same two-dimensional worm image ( e . g . , Figure 2B , bottom ) and for each movie frame j=1 , … , N , we generally find multiple potential solutions which we label {𝒑jk} , with k=1 , … , Mj . Even for simple , non-crossed postures , there can be two solutions ( Mj=2 ) , corresponding to the swapped locations of the head and tail . Across the movie , we label the indices of the correct solutions as a vector 𝒃= ( b1 , … , bN ) . We explicitly allow bj=0 in case the optimization process fails , and use a cubic spline to interpolate across any such gaps . Let us call the point in posture space for movie frame j , resulting from this interpolation step , 𝒑~j ( 𝒃 ) . To find 𝒃* for the full , correct tracking solution of the movie , we seek the solution vector that minimizes the total sequence error E ( b ) =∑j=1Nferr[Wj , W~ ( p~j ( b ) ) ] . We constrain the mode changes between two successive frames to be below 𝒗max , which simply reflects the fact that the worm can only change posture continuously . In a first pass of the data , the ‘classic’ worm tracking algorithm based on image thinning was used on all frames ( Stephens et al . , 2008 ) . This fast algorithm yields high-accuracy tracking results for frames with simple , non–self-overlapping body shapes . It also automatically labels crossed frames . For the foraging dataset , the data were cut into smaller segments to allow for faster parallel processing . Each segment consisted of a series of non-crossed frames , followed by a series of crossed frames , followed by more non-crossed frames . This effectively segmented the data by deep turns ( 936 segments in total for the 12 worm trajectories ) . For the escape response dataset , such segmentation was not necessary , due to the smaller size of the data for each worm . Frames that were labeled by the ‘classic’ algorithm as ‘crossed’ were tracked using the inverse algorithm described above . The result was an interpolated , smooth trajectory through posture space . When using this pipeline as-is , the algorithm would occasionally swap the locations of head and tail between frames . To resolve head/tail orientation correctly throughout a segment , we implemented four steps . ( 1 ) During the filtering and interpolation step , we allowed the algorithm to pick , for each non-crossed frame , not just the solution given by the ‘classic’ algorithm; it could also pick an alternative version in which head and tail were swapped ( this version can be trivially computed ) . ( 2 ) We explicitly included a limit for the maximum change of overall orientation ⟨θ⟩ between frames of ∼π rad per second in the maximum velocity vector 𝒗max . Any head/tail swaps between frames violate such a maximum change of ⟨θ⟩ . ( 3 ) After the filtering and interpolation step had produced a full tracking solution , we computed the error for both that tracking solution , as well as a version in which the head and tail were swapped for all frames in the segment . This fixed the overall head/tail orientation for the full segment . ( 4 ) As a final check , we manually verified and , if necessary , corrected head/tail orientations during post-processing . A minimal working set of our tracking code , plus a sample movie that can be successfully tracked using the code’s default parameters is available on Figshare as detailed in the author response ( code: https://figshare . com/s/3ac08fbfec9ae3d5a531 , movie: https://figshare . com/s/658dd86e3847d5926257 ) . A minimal working set of our tracking code , plus a sample movie that can be successfully tracked using the code’s default parameters is available on Figshare ( Broekmans et al . , 2016b; Broekmans et al . , 2016c ) . In total , 92 escape responses and 936 free-crawling segments ( each containing one self-overlapping turn; see above ) were analyzed . The escape response tracking results were inspected manually , and 91 trackings ( 99% ) were considered successful , as they were visually close to the appearance of the original worm . For the free crawling dataset , instead , after inspection of a representative sample of 236 segments across multiple worms , 96% were estimated to be successful . First , we assessed the quality of our tracking algorithm for non-crossed worm shapes ( Figure 2E ) . We used both the ‘classic’ algorithm and the ‘inverse’ algorithm to track N = 15433 non-crossed frames from the foraging dataset . For each frame , we calculated the Euclidean distance between the two resulting 𝜽 vectors giving the ‘inv . tracking’ distribution . In the same figure , the ‘dim . reduct . ’ distribution was calculated from Euclidean distances between the full 𝜽 vector from the classic algorithm , and θreduct=∑i=15aie^i , where 𝒆^i are the eigenworms from ( Stephens et al . , 2008 ) ( see also Figure 2—figure supplement 1 ) . This represents the information lost in only using the first five postural eigenmodes . The ‘time res . ’ distribution represents the Euclidean distance between 𝜽 vectors from consecutive frames in a movie . In Figure 2F , we additionally collected a dataset of four movies , featuring visually distinct types of omega turns . For the N=348 crossed frames in these four movies , backbones were hand-drawn on the worm images , independently from the tracking results . We compared these backbones to the final results of our inverse tracking / filtering and interpolation algorithms . The resulting mode errors δai are plotted as the blue/dark distributions . We also include the mode errors for the set of 15433 non-crossed frames ( yellow ) . For the escape response data , the largest peak in a3 between t = 10 s ( the time of the stimulus ) and t = 29 s was identified as the apex of the omega turn . To locate the end of the omega turn , the first zero of a4 after the apex was found; any point after that root that had a3<3 was considered to be the end of the omega turn . This ensured that the negative peak in a4 , representing a high-curvature state of the tail at the end of the omega turn , had finished , and that the worm had reached a relatively ‘straight’ shape . For such straight shapes , the overall orientation ⟨θ⟩ has a straightforward , intuitive interpretation . The same criterion was used , in the opposite direction , to find the start of the omega turn . If no starting point and/or end point of the omega turn could be found , the recording was excluded from the analysis . ( In the escape response dataset , this was the case for 15 out of 91 recordings ) . We used the same criterion to find both omega and delta turns in the foraging condition . For detection of local extrema in a3 , a standard peak-finding algorithm was used to detect both minima and maxima ( based on the MATLAB ‘findpeaks’ function , which defines a peak as a data point with a greater value than its immediate neighbors ) . Only extrema with a minimum prominence of 0 . 5 were kept , resulting in 1187 large-amplitude |a3|≥10 peaks throughout the entire foraging dataset . Some a3 peaks featured smaller sub-peaks in their shoulders; such sub-peaks were discarded . Orientation changes were computed by comparing the overall orientation ⟨θ⟩ between two reference points around each omega or delta turn . The apex of each deep turn was the largest a3 peak identified previously . The first reference point was the last frame before the turn’s apex that featured a ‘straight’ body shape — i . e . , a body shape with a low maximum local curvature . Only for such relatively ‘flat’ worm shapes does the overall orientation ⟨θ⟩ correspond directly to the intuitive orientation assigned to the worm . Similarly , the second reference point was the first frame after the turn’s apex with such a straight body shape . Importantly , our postural tracking algorithm allows us to continuously follow the orientation angle through coiled shapes and this is important for identifying the ‘overturning’ reorientation effects of delta turns . For the analysis of the worm’s reorientation during the escape response ( Figure 3D , E ) , N = 91 escape responses were analyzed . Each 30-second recording was segmented by first finding the omega turn . After identification of the omega turn , the reversal phase was simply defined as the first frame after the stimulus with a negative body wave phase velocity dφ/dt , up until the start of the omega turn . The ‘post-omega’ phase was any data after the end of the omega turn until the end of the recording at t = 30 s . For reorientation during foraging , we analyzed the angle change for segments with self-overlapping turns . To calculate the mutual information between the omega and delta turns during foraging , we created a binary event time series by first identifying the time of the a3 peak and then binning these times into bins of width 2 , 4 , 10 , or 20 s . We then calculated the mutual information between these binary time series as in ref . ( Strong et al . , 1998 ) . The mutual information was calculated for different relative shifts , ranging from −60 to +60 s and the results are shown in Figure 4—figure supplement 2 . Mutual information across time shifts never exceeded ∼3% of the maximum entropy of each time series , indicating that these turns occur independently . There is also no apparent spatial correlation ( see Figure 4—figure supplement 3 ) . In Figure 4D , we show how the average turn frequencies for omega and delta turns change over the course of the 35 min foraging experiments . Turns were detected by using the peak detection algorithm outlined above , applying the amplitude boundaries |a3|≥10 . The total of these extrema consists of three populations: the tail of a dorsal/ventral symmetric distribution of shallower turns , and two types of ventral deep turns , the delta and omega turns . To find the number of omega turns , we counted the number of a3 peaks with an amplitude between −20 and −10 in each time window , and subtracted this from the total number of a3 peaks with an amplitude between +10 and +20 . We then computed the average number of omega turns per unit time , across the 12 experiments , in a 10-minute sliding window , shifted across the data in 5-minute steps . The first 200 s of each experiment were discarded . An identical procedure with |a3|>20 gives the number of delta turns . Over the foraging time analyzed in Figure 4D , we find 274 ± 64 omega turns and 305 ± 35 delta turns , where the errors denote bootstrap errors produced by resampling the N = 12 different worm recordings with replacement . The equality of turn counts , within error bars , signals an approximate overall balance in turn events , in agreement with the rate calculations . The total turn rate in Figure 4D is comparable to previous work ( e . g . , Gray et al . , 2005 ) , though there are notable differences in turn definitions and experimental conditions . We also note , however , that there are spatiotemporal fluctuations in the turn counts , with an increased number of both turns , as well as a specific bias towards omega turns near the location of the copper ring , likely reflecting an increased rate of ring-induced escape responses . In addition , we find an early-time bias towards delta turns , during which we believe that the behavior is strongly influenced by the mechanical perturbation of picking . In future work , it will be fruitful to examine these spatiotemporal patterns in a larger experimental arena and with increased turn statistics . | We all instinctively recognize behavior: it’s what organisms do , whether they are single cells searching for food , or birds singing to mark their territory . If we want to understand behavior , however , we have to be able to characterize such actions as precisely and completely as their underlying molecular and cellular mechanisms . For the millimeter-sized roundworm C . elegans , video tracking and analysis has produced a compact characterization of naturally occurring worm postures . Simply put: every body posture of the worm is a different mix of four fundamental postures called ‘eigenworms’ . The worm’s snake-like motion is then a series of combinations of these projections , which can be analyzed to provide an automatic and measureable read-out of the worm’s behavior . There is , however , an important caveat: when the worm makes a ‘loop’ , and crosses over itself , such posture analysis is inapplicable . That is unfortunate: some of the worm’s most interesting behavior involves looping . One example is the “omega turn” , named after the shape of the Greek letter Ω . This sharp turn is used by the worm to steer away from harm , and more generally to abruptly reorient during the search for food and for mates . Broekmans et al . have now created an algorithm , based on eigenworms , which can analyze worm images that encompass both looped and normal shapes . The result is a complete ‘behavioral microscope’ that shows how C . elegans moves in 2D . Focusing this microscope in particular on the omega turn , Broekmans et al . found that such turns are not , as has been previously described , a single behavior . Instead , they are two separate behaviors that represent the worm’s equivalent of a left-right step . Together with previous posture analysis the work presented by Broekmans et al . allows for the full and precise measurement of the body shapes of C . elegans in 2D . This , combined with remarkable recent progress in global brain and gene expression imaging , should help to uncover new mechanisms that ultimately produce and control a worm’s behavior . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2016 | Resolving coiled shapes reveals new reorientation behaviors in C. elegans |
Pain signaling in vertebrates is modulated by neuropeptides like Substance P ( SP ) . To determine whether such modulation is conserved and potentially uncover novel interactions between nociceptive signaling pathways we examined SP/Tachykinin signaling in a Drosophila model of tissue damage-induced nociceptive hypersensitivity . Tissue-specific knockdowns and genetic mutant analyses revealed that both Tachykinin and Tachykinin-like receptor ( DTKR99D ) are required for damage-induced thermal nociceptive sensitization . Electrophysiological recording showed that DTKR99D is required in nociceptive sensory neurons for temperature-dependent increases in firing frequency upon tissue damage . DTKR overexpression caused both behavioral and electrophysiological thermal nociceptive hypersensitivity . Hedgehog , another key regulator of nociceptive sensitization , was produced by nociceptive sensory neurons following tissue damage . Surprisingly , genetic epistasis analysis revealed that DTKR function was upstream of Hedgehog-dependent sensitization in nociceptive sensory neurons . Our results highlight a conserved role for Tachykinin signaling in regulating nociception and the power of Drosophila for genetic dissection of nociception .
Neuropeptides are key regulators of behavior . They can act as local neurotransmitters ( Salio et al . , 2006 ) or as tonic “gain controls” on neuronal activity to modify diverse aspects of organismal physiology including appetite , biological rhythms , aggression , and more ( Marder , 2012; Taghert and Nitabach , 2012 ) . Neuropeptide signaling also modulates nociception , the sensory perception of noxious stimuli . For example , Calcitonin Gene-Related Peptide ( CGRP ) and Substance P ( SP ) both regulate nociception in mammals ( Harrison and Geppetti , 2001; Seybold , 2009 ) . Modulation of nociception occurs following tissue damage , where the threshold that elicits aversive behaviors is reduced . This nociceptive sensitization can appear as allodynia - aversive responsiveness to previously innocuous stimuli , or hyperalgesia - exaggerated responsiveness to noxious stimuli ( Gold and Gebhart , 2010 ) . The exact roles of neuropeptides in regulating nociceptive sensitization are not yet clear . In mammals , SP is highly expressed at the central nerve terminals of nociceptive sensory neurons where it is released as a peptide neurotransmitter ( Ribeiro-da-Silva and Hokfelt , 2000 ) . These neurons innervate the skin , are activated by noxious environmental stimuli , and project to second order neurons in laminae I of the spinal cord dorsal horn ( Allen et al . , 1997; Marvizon et al . , 1999 ) . These spinal neurons express a G-Protein-coupled receptor ( GPCR ) , Neurokinin-1 receptor ( NK-1R ) , which binds SP to transmit pain signals to the brain for further processing ( Brown et al . , 1995; Mantyh et al . , 1997 ) . NK-1R is also expressed in nociceptive sensory neurons ( Andoh et al . , 1996; Li and Zhao , 1998; Segond von Banchet et al . , 1999 ) . Once SP engages NK-1R , Gqα and Gsα signaling are activated leading to increases in intracellular Ca2+ and cAMP ( Douglas and Leeman , 2011 ) . Whether other signal transduction pathways , especially other known mediators of nociceptive sensitization , are activated downstream of NK-1R is not known . Drosophila melanogaster has several neuropeptides that are structurally related to SP . The Drosophila Tachykinin ( dTk ) gene encodes a prepro-Tachykinin that is processed into six mature Tachykinin peptides ( DTKs ) ( Siviter et al . , 2000 ) . Two Drosophila GPCRs , TKR86C and TKR99D , share 32 – 48% identity to mammalian neurokinin receptors ( Li et al . , 1991; Monnier et al . , 1992 ) . All six DTKs and mammalian SP can activate TKR99D , increasing cytoplasmic Ca2+ and cAMP levels ( Birse et al . , 2006 ) . In Drosophila , dTk regulates gut contractions ( Siviter et al . , 2000 ) , enteroendocrine homeostasis ( Amcheslavsky et al . , 2014; Song et al . , 2014 ) , stress resistance ( Kahsai et al . , 2010a; Soderberg et al . , 2011 ) , olfaction ( Ignell et al . , 2009 ) , locomotion ( Kahsai et al . , 2010b ) , aggressive behaviors ( Asahina et al . , 2014 ) , and pheromone detection in gustatory neurons ( Shankar et al . , 2015 ) . Whether dTk and its receptors also regulate nociception and , if so , what downstream molecular mediators are involved have not yet been investigated . Drosophila are useful for studying the genetic basis of nociception and nociceptive sensitization ( Im and Galko , 2011 ) . Noxious thermal and mechanical stimuli provoke an aversive withdrawal behavior in larvae: a 360-degree roll along their anterior-posterior body axis ( Babcock et al . , 2009; Tracey et al . , 2003 ) . This highly quantifiable behavior is distinct from normal locomotion and light touch responses ( Babcock et al . , 2009; Tracey et al . , 2003 ) . When a larva is challenged with a sub-threshold temperature ( 38°C or below ) , only light touch behaviors occur , whereas higher thermal stimuli result in aversive rolling behavior ( Babcock et al . , 2009 ) . Peripheral class IV multi-dendritic neurons ( class IV neurons ) are the nociceptive sensory neurons that innervate the larval barrier epidermis by tiling over it ( Gao et al . , 1999; Grueber et al . , 2003 ) and mediate the perception of noxious stimuli ( Hwang et al . , 2007 ) . For genetic manipulations within class IV neurons , ppk1 . 9-GAL4 has been used widely as the 1 . 9 kb promoter fragment of pickpocket1 driving Gal4 selectively labels class IV nociceptive sensory neurons in the periphery ( Ainsley et al . , 2003 ) . When the barrier epidermis is damaged by 254 nm UV light , larvae display both thermal allodynia and thermal hyperalgesia ( Babcock et al . , 2009 ) . This does not model sunburn because UV-C light does not penetrate the Earth’s atmosphere , however , it has proven useful for dissecting the molecular genetics of nociceptive sensitization ( Im and Galko , 2011 ) . What conserved factors are capable of sensitizing nociceptive sensory neurons in both flies and mammals ? Known molecular mediators include but are not limited to cytokines , like TNF ( Babcock et al . , 2009; Wheeler et al . , 2014 ) , neuropeptides , metabolites , ions , and lipids ( Gold and Gebhart , 2010; Julius and Basbaum , 2001 ) . In addition , Hedgehog ( Hh ) signaling mediates nociceptive sensitization in Drosophila larvae ( Babcock et al . , 2011 ) . Hh signaling regulates developmental proliferation and cancer ( Fietz et al . , 1995; Goodrich et al . , 1997 ) and had not previously been suspected of regulating sensory physiology . The main signal-transducing component of the Hh pathway , smoothened , and its downstream signaling components , such as the transcriptional regulator Cubitus interruptus and a target gene engrailed , are required in class IV neurons for both thermal allodynia and hyperalgesia following UV irradiation ( Babcock et al . , 2011 ) . In mammals , pharmacologically blocking Smoothened reverses the development of morphine analgesic tolerance in inflammatory or neuropathic pain models suggesting that the Smoothened/Hh pathway does regulate analgesia ( Babcock et al . , 2011 ) . Interactions between the Hh and SP pathways in regulating nociception have not been investigated in either vertebrates or Drosophila . Transient receptor potential ( TRP ) channels act as direct molecular sensors of noxious thermal and mechanical stimuli across phyla ( Venkatachalam and Montell , 2007 ) . In particular , the Drosophila TRPA family members , Painless ( Pain ) and TrpA1 , mediate baseline thermal nociception in larvae ( Babcock et al . , 2011; Tracey et al . , 2003; Zhong et al . , 2012 ) , as well as thermal sensation ( Kang et al . , 2012 ) and thermal nociception in adults ( Neely et al . , 2010 ) . When larval class IV neurons are sensitized , it is presumably through modification of the expression , localization , or gating properties of TRP channels such as Painless or TrpA1 . Indeed , direct genetic activation of either the TNF or Hh signaling pathway leads to thermal allodynia that is dependent on Painless . Direct genetic activation of Hh also leads to TrpA1-dependent thermal hyperalgesia ( Babcock et al . , 2011 ) . Whether Drosophila TRP channels are modulated by neuropeptides like Tachykinin has not been addressed in the context of nociception . In this study , we analyzed Drosophila Tachykinin and Tachykinin receptor ( TkR99D or DTKR ) in nociceptive sensitization . Both were required for UV-induced thermal allodynia: DTK from neurons likely within the central brain and DTKR within class IV peripheral neurons . Overexpression of DTKR in class IV neurons led to an ectopic hypersensitivity to subthreshold thermal stimuli that required specific downstream G protein signaling subunits . Electrophysiological analysis of class IV neurons revealed that when sensitized they display a DTKR-dependent increase in firing rates to allodynic temperatures . We also found that Tachykinin signaling acts upstream of smoothened in the regulation of thermal allodynia . Activation of DTKR resulted in a Dispatched-dependent production of Hh within class IV neurons . Further , this ligand was then required to relieve inhibition of Smoothened and lead to downstream engagement of Painless to mediate thermal allodynia . This study thus highlights an evolutionarily conserved modulatory function of Tachykinin signaling in regulating nociceptive sensitization , and uncovers a novel genetic interaction between Tachykinin and Hh pathways .
To assess when and where Tachykinin might regulate nociception , we first examined DTK expression . We immunostained larval brains and peripheral neurons with anti-DTK6 ( Asahina et al . , 2014 ) and anti-Leucopheae madurae tachykinin-related peptide 1 ( anti-LemTRP-1 ) ( Winther et al . , 2003 ) . DTK was not detected in class IV neurons ( Figure 1—figure supplement 1 ) . Previous reports suggested that larval brain neurons express DTK ( Winther et al . , 2003 ) . Indeed , numerous neuronal cell bodies in the larval brain expressed DTK and these extended tracts into the ventral nerve cord ( VNC ) ( Figure 1A ) . Expression of a UAS-dTkRNAi transgene via a pan-neuronal Elav ( c155 ) -GAL4 driver decreased DTK expression , except for a pair of large descending neuronal cell bodies in the protocerebrum ( Figure 1—figure supplement 2 ) and their associated projections in the VNC , suggesting that these neurons express an antigen that cross-reacts with the anti-Tachykinin serum . Labeling of anti-DTK6 in the brain was also greatly decreased ( Figures 1B and C ) in homozygous larvae bearing two different dTk alleles , dTkEY21074 and dTkΔ1C , that decrease Tachykinin function ( Figure 1—figure supplement 3 ) . Therefore , we conclude that dTk expression is effectively knocked down both in mutants and by RNAi transgenes . 10 . 7554/eLife . 10735 . 003Figure 1 . Tachykinin is expressed in the larval brain and required for thermal allodynia . ( A–C ) Dissected larval brain wholemounts of the indicated genotypes immunostained with a guinea pig antiserum to DTK6 . Arrowheads , large immunoreactive descending neurons . Arrows , remaining neurons immunoreactive to anti-DTK6 . ( A ) w1118 ( B ) dTkΔ1C ( C ) dTkEY21074 ( D ) Baseline responses to thermal stimulation in the absence of injury at 45°C and 48°C when Tachykinin is targeted by RNAi in all neurons . Larvae of indicated genotypes were stimulated for up to 20 s with a thermal probe set to the indicated temperatures . The resulting behavior was categorized as “no withdrawal” ( white ) if a 360 º aversive roll did not occur , “slow withdrawal” ( gray ) , if the roll occurred between 6 and 20 s of probe contact , or “fast withdrawal” ( black ) , if the roll occurred within 5 s of probe contact . Percent behavioral responses were plotted as mean ± SEM . This scheme was employed for all behavioral quantitation in this study . ( E ) Baseline responses to thermal stimulation at 45°C and 48°C of dTk mutant alleles and relevant controls . ( F–G ) UV-induced thermal allodynia . ( F ) RNAi targeting dTk and controls . ( 1 ) and ( 2 ) refer to non-overlapping UAS-RNAi transgenes targeting Tachykinin . ( G ) Mutant alleles of dTk and controls . All behavior experiments throughout were performed in triplicate sets of n = 30 unless noted otherwise . Statistical significance was determined by the chi-square test . Same statistical significance markers were used throughout all figures . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 00310 . 7554/eLife . 10735 . 004Figure 1—figure supplement 1 . Tachykinin is not expressed in class IV md nociceptive sensory neurons . Dissected larval epidermal whole mounts ( genotype: ppk-eGFP ) immunostained with anti-DTK and anti-GFP antibodies . GFP ( green in merge ) ; anti-DTK ( magenta in merge ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 00410 . 7554/eLife . 10735 . 005Figure 1—figure supplement 2 . Dissected larval brain whole mounts of Elav/+ and Elav>TKRNAi immunostained with anti-LemTRP . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 00510 . 7554/eLife . 10735 . 006Figure 1—figure supplement 3 . Schematic of the dTk locus . Various genomic features and characterization tools are indicated to the left . FRT-containing transposon insertion alleles d08303 and f03824 were used to make a deletion allele of dTk , dTkΔ1C . Primers a through h were used to molecularly map the deleted and neighboring regions . Diagnostic PCR using genomic DNA of w1118 control and dTkΔ1C . In homozygous dTkΔ1C , PCR products were produced only with primers pairs of a/b and g/h that flank the deleted region . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 00610 . 7554/eLife . 10735 . 007Figure 1—figure supplement 4 . Temperature versus behavior dose response curves . Control larvae were tested with thermal stimuli of temperature ranging from 38 to 48°C and their aversive rolling behaviors were monitored . Two sets of data were generated by two experimenters and plotted in parallel . The same set of data was plotted in two different displays: categorical bar graphs ( left ) and in non-categorical line graphs of the accumulated percent response on the Y-axis versus latency on the X-axis ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 00710 . 7554/eLife . 10735 . 008Figure 1—figure supplement 5 . Alternative data presentation of thermal allodynia ( a subset of Figure 1F and a subset of Figure 1G ) in non-categorical line graphs of accumulated percent response as a function of measured latency . Statistical tests were performed using Log-rank ( Mantel-Cox ) test . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 008 Because we observed a knockdown of DTK staining in the brain with mutants and RNAi , and because mammalian SP regulates pain behavior , we tested if dTk loss of function affects nociceptive behaviors . We first tested baseline nociception in the absence of injury , where larvae were challenged with noxious thermal stimuli at 45°C or 48°C , the middle and upper end of their response range , respectively ( Babcock et al . , 2009 ) . For uninjured larvae , the behavioral dose-response to temperature forms a reproducible graded curve ( Figure 1—figure supplement 4 ) . Pan-neuronal knockdown of dTk did not cause baseline nociception defects compared to relevant GAL4 controls ( Figure 1D ) . Similarly , larvae homozygous or transheterozygous for dTkEY21074 ordTkΔ1C had normal baseline thermal nociceptive responses ( Figure 1E ) . Next , we tested UV-induced nociceptive sensitization . Pan-neuronal knockdown of dTk significantly reduced thermal allodynia ( responsiveness to sub-threshold 38°C ) ( Figure 1F and Figure 1—figure supplement 5 ) . Two non-overlapping RNAi transgenes ( TkJF01818 and TkKK112227 ) targeting Tachykinin reduced the allodynia response from 70% to about 20% compared to relevant GAL4 or UAS alone controls 24 hr after UV irradiation ( Figure 1F ) . Consistent with the absence of DTK staining in class IV neurons ( Figure 1—figure supplement 1 ) , class IV-specific knockdown of dTk did not alter thermal allodynia ( Figure 1F ) . As genetic confirmation of the RNAi phenotype , we tested mutant alleles of dTk for tissue damage-induced thermal allodynia . Heterozygous larvae bearing these dTk alleles , including a deficiency spanning the dTk locus , displayed normal thermal allodynia ( Figure 1G ) . By contrast , all homozygous and transheterozygous combinations of dTk alleles drastically reduced thermal allodynia ( Figure 1G ) . Therefore , we conclude that Tachykinin is necessary for the development of thermal allodynia in response to UV-induced tissue damage . Two GPCRs recognize Tachykinins . DTKR ( TkR99D or CG7887 ) recognizes all six DTKs ( Birse et al . , 2006 ) whereas NKD ( TkR86C or CG6515 ) binds DTK-6 and a tachykinin-related peptide , natalisin ( Jiang et al . , 2013; Monnier et al . , 1992; Poels et al . , 2009 ) . Because DTKR binds more broadly to DTKs , we tested if class IV neuron-specific knockdown of dtkr using the ppk-GAL4 driver ( Ainsley et al . , 2003 ) led to defects in either baseline nociception or thermal allodynia . See Figure 2A for a schematic of the dtkr locus and the genetic tools used to assess this gene’s role in thermal allodynia . Similar to dTk , no baseline nociception defects were observed upon knockdown of dtkr ( Figure 2B ) . Larvae homozygous for TkR99Df02797 and TkR99DMB09356 were also normal for baseline nociceptive behavior ( Figure 2C ) . 10 . 7554/eLife . 10735 . 009Figure 2 . Tachykinin Receptor is required in class IV nociceptive sensory neurons for thermal allodynia . ( A ) Schematic of the dtkr genomic locus . Location of transposon insertion alleles and targeted sequences of UAS-RNAi transgenes are shown . ( B , C ) Baseline thermal nociception at 45°C and 48°C . ( B ) dtkr RNAi in class IV neurons and controls . ( C ) dtkr mutant alleles and controls . ( D , E ) UV-induced thermal allodynia at 38°C . ( D ) dtkr RNAi and rescue in class IV neurons . ( E ) dtkr mutant alleles and controls . ( F ) “Genetic” thermal allodynia in the absence of injury upon overexpression of DTKR in class IV neurons . ( G–I ) Dissected larval epidermal wholemounts ( genotype: ppk>DTKR-GFP ) immunostained with anti-LemTRP-1 ( reacts to DTKs ) and anti-GFP . ( G ) DTKR-GFP expression in class IV neuron soma and dendrites . ( H ) Larval brain wholemount . GFP ( green ) ; anti-DTK ( magenta ) . Yellow Box indicates close-up shown in I . ( I ) Axonal tracts expressing DTKR-GFP in class IV neurons juxtaposed with TK-expressing cells in the VNC . Arrows , regions where GFP-expressing axons are closely aligned with DTK-expressing axons . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 00910 . 7554/eLife . 10735 . 010Figure 2—figure supplement 1 . Alternative data presentation of thermal allodynia ( Figure 2D and a subset of Figure 2E ) in non-categorical line graphs of accumulated percent response as a function of measured latency . Statistical tests were performed using Log-rank ( Mantel-Cox ) test . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 010 Although baseline nociception was unaffected , class IV neuron-specific expression of UAS-dtkrRNAi significantly reduced thermal allodynia compared to GAL4 or UAS alone controls ( Figure 2D and Figure 2—figure supplement 1 ) . This reduction was rescued upon simultaneous overexpression of DTKR using a UAS-DTKR-GFP transgene , suggesting that the RNAi-mediated phenotype was not off-target ( Figure 2D ) . We also tested mutant alleles of dtkr for thermal allodynia defects . While all heterozygotes were normal , larvae bearing any homozygous or transheterozygous combination of alleles , including a deficiency spanning the dtkr locus , displayed greatly reduced thermal allodynia ( Figure 2E ) . Restoration of DTKR expression in class IV neurons in a dtkr mutant background fully rescued their allodynia defect ( Figure 2E and Figure 2—figure supplement 1 ) suggesting that the gene functions in these cells . Lastly , we examined whether overexpression of DTKR within class IV neurons could ectopically sensitize larvae . While GAL4 or UAS alone controls remained non-responsive to sub-threshold 38°C , larvae expressing DTKR-GFP within their class IV neurons showed aversive withdrawal to this temperature even in the absence of tissue damage ( Figure 2F ) . Visualization of the class IV neurons expressing DTKR-GFP showed that the protein localized to both the neuronal soma and dendritic arbors ( Figure 2G ) . Expression of DTKR-GFP was also detected in the VNC , where class IV axonal tracts run immediately adjacent to the axonal projections of the Tachykinin-expressing central neurons ( Figures 2H and I ) . Taken together , we conclude that DTKR functions in class IV nociceptive sensory neurons to mediate thermal allodynia . To determine if the behavioral changes in nociceptive sensitization reflect neurophysiological changes within class IV neurons , we monitored action potential firing rates within class IV neurons in UV- and mock-treated larvae . As in our behavioral assay , we UV-irradiated larvae and 24 hr later monitored changes in response to thermal stimuli . Here we measured firing rates with extracellular recording in a dissected larval fillet preparation ( Figure 3A and methods ) . Mock-treated larvae showed no increase in their firing rates until around 39°C ( Figures 3B and D ) . However , UV-treated larvae showed an increase in firing rate at temperatures from 31°C and higher ( Figures 3C and D ) . The difference in change in firing rates between UV- and mock-treated larvae was significant between 30°C and 39°C . This increase in firing rate demonstrates sensitization in the primary nociceptive sensory neurons and correlates well with behavioral sensitization monitored previously . 10 . 7554/eLife . 10735 . 011Figure 3 . Class IV neurons display temperature-dependent changes in firing rates that are modulated by Tachykinin signaling . ( A ) Schematic diagram of assay setup . ( B , C , F , G ) Sample recording traces of the indicated genotypes in response to temperature ramping . ( B ) ppk1 . 9-Gal4 , ppk-eGFP/+ mock ( C ) ppk1 . 9-Gal4 , ppk-eGFP/+ 24 hr following UV ( F ) ppk-Gal4/+ ( G ) ppk-Gal4>DTKR-GFP . ( D ) Changes in firing rates from larvae in ( B ) and ( C ) in response to temperature ramping . n = 11 ( mock ) , and 21 ( UV ) . ( E ) Changes in firing rates between ppk-Gal4>dtkrRNAi ( mock and UV ) , dtkrMB09356/f02797 ( UV ) , and class IV neuron-specific rescue of dtkrMB09356/f02797 ( UV ) in response to temperature ramping . n = 12 ( RNAi mock ) , 11 ( RNAi UV ) , 17 ( dtkr mutant ) , and 12 ( rescue ) . ( H ) Changes in firing rates between Gal4 only control and class IV specific overexpression of DTKR in response to temperature ramping without tissue damage . n = 9 ( control ) , 12 ( Overexpression ) . Inset , Behavioral response to innocuous temperatures when DTKR is overexpressed in class IV neurons without tissue damage . *= P<0 . 05 , **= P<0 . 01 , ***= P<0 . 001 . Statistical significance was determined by either Two-way ANOVA test with Bonferroni correction or two-tailed unequal variance Student’s t-Test for electrophysiology , or by Chi-square analysis for behavior analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01110 . 7554/eLife . 10735 . 012Figure 3—figure supplement 1 . Control genotypes for electrophysiology recordings of class IV neurons . Changes in firing rates of heterozygote dtkr ( dtkrf02797/+ ) and Gal4 alone control in dtkr mutant background ( ppk1 . 9Gal4 , dtkrf02797/dtkrm09356 ) in response to temperature ramping upon UV irradiation . n = 8 ( heterozygote ) , 7 ( Gal4 control ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 012 Next , we wondered if loss of dtkr could block the UV-induced increase in firing rate . Indeed , class IV neurons of dtkr mutants showed little increase in firing rates even with UV irradiation ( Figure 3E ) . Similarly , knockdown of dtkr within class IV neurons blocked the UV-induced increase in firing rate; UV- and mock-treated UAS-dtkrRNAi-expressing larvae showed no statistically significant difference in firing rate ( Figure 3E ) . When DTKR expression was restored only in the class IV neurons in the dtkr mutant background , the firing rates increased with increasing temperature upon UV irradiation ( Figure 3E and see Figure 3—figure supplement 1 for additional control genotypes ) . Thus , dtkr functions in class IV neurons for the UV-induced increase in firing rate in response to increasing temperature . Next we overexpressed DTKR in class IV neurons and tested the effect of gain of function on the neuronal firing rate . Behaviorally , overexpression induced ectopic sensitization even without UV ( Figure 2F ) . When we assayed lower temperatures ( 32–38°C ) , the ectopic thermal allodynia was obvious above 34°C ( Figure 3H ) . Electrophysiologically , we saw similar results . Class IV neurons expressing DTKR-GFP increased their firing rate to thermal stimuli even without UV irradiation ( Figures 3F–H ) . The magnitude of the increase upon overexpression was comparable to that of UV-treated controls ( Figures 3D and H ) . Taken together , electrophysiological recordings corresponded well with the behavioral changes seen upon loss- or gain-of-function of Tachykinin signaling . The electrophysiology further suggests that DTKR signaling modulates the firing properties of class IV nociceptive sensory neurons in response to tissue damaging stimuli like UV radiation . DTKR activation increases cytoplasmic Ca2+ and cAMP levels in a heterologous cell-based assay ( Birse et al . , 2006 ) , suggesting receptor coupling to Gαs and/or Gαq . To identify the particular trimeric G-protein subunits through which Tachykinin and its receptor modulate thermal allodynia , we screened five of six annotated and two putative Gα- , all three annotated and one putative Gβ- , and all two annotated and one putative Gγ-encoding genes ( Figure 4A ) . Several UAS-RNAi transgenes yielded modest defects in thermal allodynia ( Figure 4B and Figure 4—figure supplement 1 ) . When analyzing our behavioral data categorically , Gβ5 was not quite significant , but when the data was analyzed non-categorically ( accumulated percent response versus latency ) the increased statistical power of this method revealed that Gβ5 was significantly different from the control ( Figure 4—figure supplement 1 ) . Indeed , retesting the strongest hits in greater numbers and analyzing them categorically revealed that knockdown of a putative Gαq ( CG17760 ) , Gβ5 ( CG10763 ) , and Gγ1 ( CG8261 ) all significantly reduced thermal allodynia compared to GAL4 and UAS-alone controls ( Figure 4C and Figure 4—figure supplements 1 and 2 ) . To test if these subunits act downstream of DTKR , we asked whether expression of the relevant UAS-RNAi transgenes could also block the ectopic thermal allodynia induced by DTKR-GFP overexpression ( Figure 2F ) . All of them did ( Figure 4D ) . Therefore , we conclude that CG17760 , Gβ5 , and Gγ1 are the downstream G protein subunits that couple to DTKR to mediate thermal allodynia in class IV neurons . 10 . 7554/eLife . 10735 . 013Figure 4 . Specific Trimeric G proteins act downstream of DTKR in class IV neurons in thermal allodynia . ( A ) Schematic of genetic screening strategy for testing G-protein subunit function by in vivo tissue-specific RNAi in class IV neurons . ( B ) UV-induced thermal allodynia on targeting the indicated G protein subunits by RNAi . n = 30 larvae per genotype . ≈ P = 0 . 082 , * P<0 . 05 . Statistical significance was determined by Fisher’s exact test . ( C ) UV-induced thermal allodynia for the three putative hits from the mini-screen in A . ( 1 ) and ( 2 ) indicate non-overlapping RNAi transgenes . ( D ) Suppression of UAS-DTKR-induced “genetic” allodynia by co-expression of UAS-RNAi transgenes targeting the indicated G protein subunits . Seven sets of n=30 for ppk>DTKR-GFP controls , triplicate sets of n=30 for the rest . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01310 . 7554/eLife . 10735 . 014Figure 4—figure supplement 1 . Alternative data presentation of UV-induced thermal allodynia on targeting G protein subunits by RNAi ( Figure 4B ) in non-categorical line graphs of accumulated percent response as a function of measured latency . Statistical tests were performed using Log-rank ( Mantel-Cox ) test . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01410 . 7554/eLife . 10735 . 015Figure 4—figure supplement 2 . UAS alone controls of RNAi targeting G protein subunits do not exhibit defects in UV-induced thermal allodynia . Flies bearing the indicated UAS-RNAi transgenes were crossed to w1118 and their progeny were tested . Triplicate sets of n = 30 . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 015 The signal transducer of the Hedgehog ( Hh ) pathway , Smoothened ( smo ) , is required within class IV neurons for UV-induced thermal allodynia ( Babcock et al . , 2011 ) . To determine if Tachykinin signaling genetically interacts with the Hh pathway during thermal allodynia , we tested the behavior of a double heterozygous combination of dtkr and smo alleles . Such larvae are defective in UV-induced thermal allodynia compared to relevant controls ( Figure 5A and Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 10735 . 016Figure 5 . Tachykinin signaling is upstream of Smoothened and Painless in thermal allodynia . ( A ) Thermal allodynia in indicated dTk and smo heterozygotes and transheterozygotes . ( B ) Schematic of the expected results for genetic epistasis tests between the dTK and Hh pathways . ( C ) Suppression of Hh pathway-induced “genetic” allodynia by co-expression of UAS-dtkrRNAi . UAS-enRNAi serves as a positive control . ( D–E ) Suppression of DTKR-induced “genetic” allodynia . ( D ) Co-expression of indicated transgenes targeting the Hh signaling pathway and relevant controls . ( E ) Co-expression of indicated RNAi transgenes targeting TRP channel , painless . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01610 . 7554/eLife . 10735 . 017Figure 5—figure supplement 1 . Alternative data presentation of thermal allodynia results ( Figure 5A and Figure 5D ) in non-categorical line graphs of accumulated percent response as a function of measured latency . Statistical tests were performed using Log-rank ( Mantel-Cox ) test . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01710 . 7554/eLife . 10735 . 018Figure 5—figure supplement 2 . Genetic epistasis tests between DTKR and TNF pathway . Co-expression of UAS-dtkrRNAi did not suppress TNF-induced genetic allodynia and co-expression of UAS-wengenRNAi did not suppress DTKR-induced genetic allodynia . Data presented in categorical bar graphs ( left ) and in non-categorical line graphs of accumulated percent response as a function of measured latency . Statistical tests were performed using Chi-Square analysis ( bar graph ) or Log-rank ( Mantel-Cox ) test ( line graphs ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01810 . 7554/eLife . 10735 . 019Figure 5—figure supplement 3 . Schematic of painless genomic locus . painless70 was generated by imprecise excision of painlessEP2451 , deleting 4 . 5 kb of surrounding sequence including the ATG of the A splice variant . Primers P1/P2 were used to confirm the deletion in diagnostic PCR amplification of pain70 versus control w1118 genomic DNA . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 01910 . 7554/eLife . 10735 . 020Figure 5—figure supplement 4 . The pain70 deletion allele and UAS-painRNAi transgenes cause defects in baseline thermal nociception . Quantitation of baseline responses to thermal stimulation at 48°C . painless70 was compared to w1118 controls , while two non-overlapping UAS-painlessRNAi transgenes ( v39477 and 31510 ) expressed in class IV neurons were compared to Gal4 or UAS alone controls . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 020 We next performed genetic epistasis tests to determine whether Tachykinin signaling functions upstream , downstream , or parallel of Hh signaling during development of thermal allodynia . The general principle was to co-express an activating transgene of one pathway ( which induces genetic thermal allodynia ) together with an inactivating transgene of the other pathway . Reduced allodynia would indicate that the second pathway was acting downstream of the ectopically activated one ( see schematic of possible outcomes in Figure 5B ) . For example , to test if Tachykinin signaling is downstream of smo , we combined a dominant negative form of Patched ( UAS-PtcDN ) that constitutively activates Smo and causes ectopic thermal allodynia ( Babcock et al . , 2011 ) with UAS-dtkrRNAi . This did not block the ectopic sensitization ( Figure 5C ) while a positive control gene downstream of smo did ( UAS-engrailedRNAi ) , suggesting that dtkr does not function downstream of smo . In a converse experiment , we combined UAS-DTKR-GFP with a number of transgenes capable of interfering with Smo signal transduction . Inactivation of Smo signaling via expression of Patched ( UAS-Ptc ) , or a dominant negative form of smo ( UAS-smoDN ) , or a dominant negative form of the transcriptional regulator Cubitus interruptus ( UAS-CiDN ) , or an RNAi transgene targeting the downstream transcriptional target engrailed ( UAS-enRNAi ) , all abolished the ectopic sensitization induced by overexpression of DTKR-GFP ( Figure 5D and Figure 5—figure supplement 1 ) . Thus , functional Smo signaling components act downstream of DTKR in class IV neurons . The TNF receptor Wengen ( Kanda et al . , 2002 ) is required in class IV nociceptive sensory neurons to elicit UV-induced thermal allodynia ( Babcock et al . , 2009 ) . We therefore also tested the epistatic relationship between DTKR and the TNFR/Wengen signaling pathways and found that they function independently of/in parallel to each other during thermal allodynia ( Figure 5—figure supplement 2 ) . This is consistent with previous genetic epistasis analysis , which revealed that TNF and Hh signaling also function independently during thermal allodynia ( Babcock et al . , 2011 ) . The TRP channel pain is required for UV-induced thermal allodynia downstream of Smo ( Babcock et al . , 2011 ) . Because Smo acts downstream of Tachykinin this suggests that pain would also function downstream of dtkr . We formally tested this by combining DTKR overexpression with two non-overlapping UAS-painRNAi transgenes . These UAS-painRNAitransgenes reduced baseline nociception responses to 48°C although not as severely as pain70 , a deletion allele of painless ( Figure 5—figure supplement 3 , 4 and . As expected , combining DTKR overexpression and pain knockdown or DTKR and pain70 reduced ectopic thermal allodynia ( Figure 5E ) . In sum , our epistasis analysis indicates that the Smo signaling cassette acts downstream of DTKR in class IV neurons and that these factors then act via Painless to mediate thermal allodynia . Where does Hh itself fit into this scheme ? Although hhts2 mutants show abnormal sensitization ( Babcock et al . , 2011 ) , it remained unclear where Hh is produced during thermal allodynia . To find the source of active Hh , we tried tissue-specific knockdowns . However , none of the UAS-HhRNAi transgenes we tested were effective at inducing wing patterning phenotypes in the wing imaginal disc ( Figure 6—figure supplement 1 ) nor exhibited defects in thermal allodynia ( Figure 6—figure supplement 2 ) . Thus , we asked if tissue-specific overexpression of UAS-Hh in a variety of tissues could induce ectopic thermal allodynia in the absence of UV . Among class IV neurons , epidermis , and gut , overexpression of Hh only in class IV neurons resulted in ectopic sensitization ( Figure 6A ) . This suggests that the class IV neurons themselves are potential Hh-producing cells . 10 . 7554/eLife . 10735 . 021Figure 6 . Tachykinin-induced Hedgehog is autocrine from class IV nociceptive sensory neurons . ( A ) “Genetic” allodynia induced by ectopic Hh overexpression in various tissues . Tissue-specific Gal4 drivers , UAS controls and combinations are indicated . The Gal4 drivers used are ppk-Gal4 ( class IV sensory neuron ) , A58-Gal4 ( epidermis ) , and Myosin1A-Gal4 ( gut ) . ( B ) Schematic of class IV neuron isolation and immunostaining . ( C ) Isolated class IV neurons stained with anti-Hh . mCD8-GFP ( green in merge ) ; anti-Hh ( magenta in merge ) . ( D ) Number of Hh punctae in isolated class IV neurons from genotypes/conditions in ( C ) . Punctae per image are plotted as individual points . Black bar; mean gray bracket; SEM . Statistical significance was determined by One-way ANOVA test followed by multiple comparisons with Tukey correction . ( E ) UV-induced thermal allodynia upon UAS-dispRNAi expression with relevant controls . ( F ) Suppression of “genetic” allodynia by co-expression of UAS-dispRNAi in class IV neurons . Genetic allodynia conditions were induced by Hh overexpression , PtcDN expression , or DTKR-GFP overexpression . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 02110 . 7554/eLife . 10735 . 022Figure 6—figure supplement 1 . RNAi-mediated knockdown of hh was not effective . Wing pattern phenotypes of various UAS-RNAi transgenes targeting hh using nubbin-GAL4 and daughterless-GAL4 . UAS-CiDN was used as a positive control . None of the UAS-HhRNAi transgenes tested effectively interfered with wing patterning . This suggests that the lack of phenotype in thermal allodynia assays on expression of these transgenes is largely due to inefficient knockdown of hh function . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 02210 . 7554/eLife . 10735 . 023Figure 6—figure supplement 2 . RNAi-mediated knockdown of hh was not effective in blocking thermal allodynia . UAS-hhRNAi ( 1 ) refers to 4637R2 whereas UAS-hhRNAi ( 2 ) refers to v1403 . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 02310 . 7554/eLife . 10735 . 024Figure 6—figure supplement 3 . A few more examples of isolated class IV neurons stained with anti-Hh . mCD8-GFP ( green in merge ) ; anti-Hh ( magenta in merge ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 02410 . 7554/eLife . 10735 . 025Figure 6—figure supplement 4 . Genetic allodynia in the absence of tissue injury upon overexpression of TNF in class IV neurons . Knockdown of disp did not interfere with TNF-induced genetic allodynia . DOI: http://dx . doi . org/10 . 7554/eLife . 10735 . 025 These gain-of-function results predict that Hh might be produced in an autocrine fashion from class IV neurons following tissue injury . To monitor Hh production from class IV neurons , we performed immunostaining on isolated cells . Class IV neurons expressing mCD8-GFP were physically dissociated from intact larvae , enriched using magnetic beads conjugated with anti-mCD8 antibody , and immunostained with anti-Hh ( see schematic Figure 6B ) . Mock-treated control neurons did not contain much Hh and UV irradiation increased this basal amount only incrementally ( Figure 6C and Figure 6—figure supplement 3 ) . A possible reason for this incremental increase in response to UV is that Hh is a secreted ligand . To trap Hh within class IV neurons , we asked if blocking dispatched ( disp ) function could trap the ligand within the neurons . Disp is necessary to process and release active cholesterol-modified Hh ( Burke et al . , 1999; Ma et al . , 2002 ) . Knockdown of disp by itself ( no UV ) had no effect; however combining UV irradiation and expression of UAS-dispRNAi resulted in a drastic increase in intracellular Hh punctae ( Figures 6C , D and Figure 6—figure supplement 3 ) . This suggests that class IV neurons express Hh and that blocking Dispatched function following UV irradiation traps Hh within the neuron . Finally , we tested if trapping Hh within the class IV neurons influenced UV-induced thermal allodynia . Indeed , class IV neuron-specific expression of two non-overlapping UAS-dispRNAi transgenes each reduced UV-induced allodynia ( Figure 6E ) . Furthermore , we tested whether expression of UAS-dispRNAi blocked the ectopic sensitization induced by Hh overexpression . It did ( Figure 6F ) , indicating that Disp function is required for production of active Hh in class IV neurons , as in other cell types and that Disp-dependent Hh release is necessary for this genetic allodynia . disp function was specific; expression of UAS-dispRNAi did not block UAS-TNF-induced ectopic sensitization even though TNF is presumably secreted from class IV neurons in this context ( Figure 6—figure supplement 4 ) . Expression of UAS-dispRNAi did not block UAS-PtcDN-induced ectopic sensitization , suggesting that this does not depend on the generation/presence of active Hh ( Figure 6F ) . Finally , we tested if UAS-dispRNAi expression blocked the ectopic sensitization induced by UAS-DTKR-GFP overexpression . It could , further supporting the idea that Disp-dependent Hh release is downstream of the Tachykinin pathway ( Figure 6F ) . Thus , UV-induced tissue damage causes Hh production in class IV neurons . Dispatched function is required downstream of DTKR but not downstream of Ptc , presumably to liberate Hh ligand from the cell and generate a functional thermal allodynia response .
To date we have found three signaling pathways that regulate UV-induced thermal allodynia in Drosophila – TNF ( Babcock et al . , 2009 ) , Hedgehog ( Babcock et al . , 2011 ) , and Tachykinin ( this study ) . All are required for a full thermal allodynia response to UV but genetic epistasis tests reveal that TNF and Tachykinin act in parallel or independently , as do TNF and Hh . This could suggest that in the genetic epistasis contexts , which rely on class IV neuron-specific pathway activation in the absence of tissue damage , hyperactivation of one pathway ( say TNF or Tachykinin ) compensates for the lack of the function normally provided by the other parallel pathway following tissue damage . While TNF is independent of Hh and DTKR , analysis of DTKR versus Hh uncovered an unexpected interdependence . We showed that Hh signaling is downstream of DTKR in the context of thermal allodynia . Two pieces of genetic evidence support this conclusion . First , flies transheterozygous for dTk and smo displayed attenuated UV-induced thermal allodynia . Thus , the pathways interact genetically . Second , and more important for ordering the pathways , loss of canonical downstream Hh signaling components blocked the ectopic sensitization induced by DTKR overexpression . We previously showed that loss of these same components also blocks allodynia induced by either UV or Hh hyperactivation ( Babcock et al . , 2011 ) , suggesting that these downstream Hh components are also downstream of DTKR . The fact that Smo is activated upon overexpression of DTKR within the same cell argues that class IV neurons may need to synthesize their own Hh following a nociceptive stimulus such as UV radiation . The data supporting an autocrine model of Hh production are three fold: ( 1 ) only class IV neuron-mediated overexpression of Hh caused thermal allodynia suggesting this tissue is fully capable of producing active Hh ligand , ( 2 ) expression of UAS-dispRNAi within class IV neurons blocked UV- and DTKR-induced thermal allodynia , implicating a role for Disp-driven Hh secretion in these cells , and ( 3 ) the combination of UAS-dispRNAi and UV irradiation caused accumulation of Hh punctae within class IV neurons . Disp is not canonically viewed as a downstream target of Smo and indeed , blocking disp did not attenuate UAS-PtcDN-induced or UAS-TNF-induced allodynia , indicating that Disp is specifically required for Hh production between DTKR and Smo . Thus , Tachykinin signaling leads to Hh expression , Disp-mediated Hh release , or both ( Figure 7 ) . Autocrine release of Hh has only been demonstrated in a few non-neuronal contexts to date ( Chung and Bunz , 2013; Zhou et al . , 2012 ) . This signaling architecture differs from what has been found in Drosophila development in two main ways . One is that DTKR is not known to play a patterning role upstream of Smo . The second is that Hh-producing cells are generally not thought to be capable of responding to Hh during the formation of developmental compartment boundaries ( Guerrero and Kornberg , 2014; Torroja et al . , 2005 ) . Ultimately , a sensitized neuron needs to exhibit firing properties that are different from those seen in the naïve or resting state . Previously , we have only examined sensitization at the behavioral level . Here we also monitored changes through extracellular electrophysiological recordings . These turned out to correspond remarkably well to behavioral sensitization . In control UV-treated larvae , nearly every temperature in the low “allodynic” range showed an increase in firing frequency in class IV neurons upon temperature ramping . Dtkr knockdown in class IV neurons abolished the UV-induced increase in firing frequency seen with increasing temperature and overexpression of DTKR increased the firing rate comparable to UV treatment . This latter finding provides a tidy explanation for DTKR-induced 'genetic allodynia' . The correspondence between behavior and electrophysiology argues strongly that Tachykinin directly modifies the firing properties of nociceptive sensory neurons in a manner consistent with behavioral thermal allodynia . Genetically , knockdown of painless blocks DTKR- or PtcDN-induced ectopic sensitization suggesting that , ultimately , thermal allodynia is mediated in part via this channel . Indeed , the SP receptor Neurokinin-1 enhances TRPV1 function in primary rat sensory neurons ( Zhang et al . , 2007 ) . Tachykinin/Hh activation could lead to increased Painless expression , altered Painless localization , or to post-translational modification of Painless increasing the probability of channel opening at lower temperatures . Because thermal allodynia evoked by UV and Hh-activation requires Ci and En we favor the possibility that sensitization may involve a simple increase in the expression level of Painless , although the above mechanisms are not mutually exclusive . Altered localization has been observed with a different TRP channel downstream of Hh stimulation; Smo activation leads to PKD2L1 recruitment to the primary cilium in fibroblasts , thus regulating local calcium dynamics of this compartment ( Delling et al . , 2013 ) . The exact molecular mechanisms by which nociceptive sensitization occurs is the largest black box in the field and will take a concerted effort by many groups to precisely pin down . Our results establish that Tachykinin/SP modulation of nociception is conserved across phyla . However , there are substantial differences in the architecture of this signaling axis between flies and mammals . In mammals , activation of TRP channels in the periphery leads to release of SP from the nerve termini of primary afferent C fibers in the dorsal horn ( Abbadie et al . , 1997; Allen et al . , 1997 ) . SP and spinal NK-1R have been reported to be required for moderate to intense baseline nociception and inflammatory hyperalgesia although some discrepancies exist between the pharmacological and genetic knockout data ( Cao et al . , 1998; De Felipe et al . , 1998; Mantyh et al . , 1997; Regoli et al . , 1994; Woolf et al . , 1998; Zimmer et al . , 1998 ) . The most profound difference of Drosophila Tachykinin signaling anatomically is that DTK is not expressed and does not function in primary nociceptive sensory neurons . Rather , DTK is expressed in brain neurons and the larval gut ( Siviter et al . , 2000 ) , and DTKR functions in class IV neurons to mediate thermal pain sensitization . Indeed , this raises an interesting possibility for mammalian SP studies , because nociceptive sensory neurons themselves express NK-1R ( Andoh et al . , 1996; Brown et al . , 1995; Segond von Banchet et al . , 1999 ) and SP could conceivably activate the receptor in an autocrine fashion . A testable hypothesis that emerges from our studies is that NK-1R in vertebrates might play a sensory neuron-autonomous role in regulating nociception . This possibility , while suggested by electrophysiology ( Zhang et al . , 2007 ) and expression studies ( Andoh et al . , 1996; Brown et al . , 1995; Segond von Banchet et al . , 1999 ) has not been adequately tested by genetic analyses in mouse to date . In summary , we discovered a conserved role for systemic Tachykinin signaling in the modulation of nociceptive sensitization in Drosophila . The sophisticated genetic tools available in Drosophila have allowed us to uncover both a novel genetic interaction between Tachykinin and Hh signaling and an autocrine function of Hh in nociceptive sensitization . Our work thus provides a deeper understanding of how neuropeptide signaling fine-tunes an essential behavioral response , aversive withdrawal , in response to tissue damage . | Injured animals from humans to insects become extra sensitive to sensations such as touch and heat . This hypersensitivity is thought to protect areas of injury or inflammation while they heal , but it is not clear how it comes about . Now , Im et al . have addressed this question by assessing pain in fruit flies after tissue damage . The experiments used ultraviolet radiation to essentially cause ‘localized sunburn’ to fruit fly larvae . Electrical impulses were then recorded from the larvae’s pain-detecting neurons and the larvae were analyzed for behaviors that indicate pain responses ( for example , rolling ) . Im et al . found that tissue injury lowers the threshold at which temperature causes pain in fruit fly larvae . Further experiments using mutant flies that lacked genes involved in two signaling pathways showed that a signaling molecule called Tachykinin and its receptor ( called DTKR ) are needed to regulate the observed threshold lowering . When the genes for either of these proteins were deleted , the larvae no longer showed the pain hypersensitivity following an injury . Further experiments then uncovered a genetic interaction between Tachykinin signaling and a second signaling pathway that also regulates pain sensitization ( called Hedgehog signaling ) . Im et al . found that Tachykinin acts upstream of Hedgehog in the pain-detecting neurons . Following on from these findings , the biggest outstanding questions are: how , when and where does tissue damage lead to the release of Tachykinin to sensitize neurons ? Future studies could also ask whether the genetic interactions between Hedgehog and Tachykinin ( or related proteins ) are conserved in other animals such as humans and mice . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"neuroscience"
] | 2015 | Tachykinin acts upstream of autocrine Hedgehog signaling during nociceptive sensitization in Drosophila |
Axons fail to regenerate after central nervous system ( CNS ) injury . Modulation of the PTEN/mTORC1 pathway in retinal ganglion cells ( RGCs ) promotes axon regeneration after optic nerve injury . Here , we report that AKT activation , downstream of Pten deletion , promotes axon regeneration and RGC survival . We further demonstrate that GSK3β plays an indispensable role in mediating AKT-induced axon regeneration . Deletion or inactivation of GSK3β promotes axon regeneration independently of the mTORC1 pathway , whereas constitutive activation of GSK3β reduces AKT-induced axon regeneration . Importantly , we have identified eIF2Bε as a novel downstream effector of GSK3β in regulating axon regeneration . Inactivation of eIF2Bε reduces both GSK3β and AKT-mediated effects on axon regeneration . Constitutive activation of eIF2Bε is sufficient to promote axon regeneration . Our results reveal a key role of the AKT-GSK3β-eIF2Bε signaling module in regulating axon regeneration in the adult mammalian CNS .
Axon injury in the adult mammalian central nervous system ( CNS ) causes irreversible damages and permanent loss of functions due to a diminished intrinsic regenerative capability of the mature CNS neurons as well as an inhibitory extrinsic environment ( Horner and Gage , 2000; Yiu and He , 2006 ) . Reactivation of the intrinsic regenerative capability promotes CNS axon regeneration after injury ( Fischer et al . , 2004; Gaub et al . , 2011; Liu et al . , 2010; Moore et al . , 2009; O'Donovan et al . , 2014; Park et al . , 2008; Smith et al . , 2009; Watkins et al . , 2013 ) . Significantly , deletion of PTEN ( phosphatase and tensin homolog ) in adult retinal ganglion cells ( RGCs ) promotes axon regeneration after optic nerve injury ( Park et al . , 2008 ) . Loss of PTEN leads to the accumulation of PIP3 ( phosphatidylinositol-3 , 4 , 5-trisphosphate ) , resulting in the activation of the serine/threonine kinase AKT via PDK1-mediated phosphorylation ( Carnero et al . , 2008; Luo et al . , 2003 ) . AKT is a critical node in cell signaling downstream of growth factors , cytokines , and other cellular stimuli and regulates a wide spectrum of cellular functions , which include cell survival , growth , proliferation , metabolism , and migration ( Manning and Cantley , 2007 ) . The role of AKT in CNS axon regeneration remains to be revealed . Activation of mTOR complex 1 ( mTORC1 ) plays an important role in mediating Pten deletion-induced axon regeneration ( Park et al . , 2008 ) . However , mTORC1-independent pathway ( s ) may exist to regulate axon regrowth in the Pten-deficient neurons because 1 ) pharmacological inhibition of mTORC1 by rapamycin treatment only partially neutralizes the effects of Pten deficiency on axon regeneration ( Park et al . , 2008 ) ; 2 ) genetic ablation of TSC1 ( tuberous sclerosis complex 1 ) , a negative regulator of mTORC1 , or manipulation of downstream effectors of mTORC1 only partially but not completely recapitulated the axon regeneration effects mediated by Pten deletion ( Park et al . , 2008; Yang et al . , 2014 ) . GSK3 ( glycogen synthase kinase 3 ) is a signal transducer of AKT . In mammals , the GSK3 family consists of two members , GSK3α and GSK3β . AKT inactivates the kinase activity of GSK3 through phosphorylation of GSK3α at Ser21 or GSK3β at Ser9 ( Cross et al . , 1995 ) . Studies in the peripheral nervous system ( PNS ) have generated different results regarding the role of GSK3 in regulating PNS axon regeneration . Using Gsk3a-S21A/Gsk3b-S9A double knock-in mice , in which GSK3α/GSK3β cannot be inactivated by AKT-mediated phosphorylation , one group did not observe obviously phenotype in sensory axon regeneration ( Zhang et al . , 2014 ) , whereas the other group reported that sustained GSK3 activity markedly facilitated sciatic nerve regeneration ( Gobrecht et al . , 2014 ) . In the CNS , pharmacological inactivation of GSK3 by the administration of lithium , a GSK3 inhibitor , stimulates axon formation and elongation in the presence or absence of inhibitory substrates after spinal cord injury ( Dill et al . , 2008 ) . However , the mechanism of lithium action is not completely clear as many other potential targets including a number of vital enzymes are also inhibited by lithium in an uncompetitive manner ( Phiel and Klein , 2001 ) . Thus , genetic evidence is needed to examine the role of GSK3 in CNS axon regeneration . In the present study using optic nerve crush ( ONC ) to model CNS axon injury , we delineate the role of GSK3β in regulating AKT-induced axon regeneration . We further identified eIF2Bε ( eukaryotic translation initiation factor 2B epsilon subunit ) , a substrate of GSK3β , as a novel factor to regulate both GSK3β and AKT-mediated effects on axon regeneration .
AKT activation , indicated by phosphorylation at Ser473 ( Bozulic and Hemmings , 2009 ) , was readily detectable in RGCs labeled by anti-Tuj1 immunoreactivity in the developing retina and was significantly reduced in adult mice ( Figure 1—figure supplement 1A–B ) . PTEN negatively regulates the PI3K/AKT signaling pathway . To examine whether the deletion of Pten leads to AKT activation , we injected AAV-Cre into the vitreous body of adult Ptenf/f mice , resulting in Cre-mediated Pten deletion in RGCs ( Figure 1—figure supplement 1C–D ) , ( Park et al . , 2008 ) . Two weeks after viral injection , we detected increased phosphorylation of AKT at Ser473 in RGCs either with or without ONC , in comparison with Ptenf/fmice injected with AAV-GFP as a control ( Figure 1A–B ) . To investigate whether AKT activation is sufficient to promote axon regeneration , we injected AAV-caAKT ( a constitutively active form of AKT [Kohn et al . , 1996] ) , or AAV-GFP as a control , into the vitreous body of adult wild-type mice ( Figure 1—figure supplement 1E–F ) and performed ONC two weeks after viral injection . Axon regeneration was examined two weeks after ONC using immunohistochemistry for GAP-43 ( Koprivica et al . , 2005; Leon et al . , 2000 ) . While very few axon fibers extended beyond the crush site in the AAV-GFP injected control , robust axon regeneration was stimulated in the AAV-caAKT injected retina ( Figure 1C ) . Upon AKT activation , many regenerating axon fibers were observed at the proximal region of the crush site . The number of regenerating axons gradually declined over a longer distance from the lesion site ( Figure 1D ) . To assess the effect of AKT activation on RGC survival , we quantified the number of Tuj1+ cells using immunohistochemistry on retinal flatmount preparations two weeks after ONC ( Figure 1E ) . In comparison to 25 . 7% of RGCs remaining in the AAV-GFP injected control , significantly more RGCs ( 48 . 4% ) were scored in the AAV-caAKT injected retina ( Figure 1F ) . Our results indicate that Pten deletion activates AKT , and AKT activation promotes both axon regeneration and RGC survival after optic nerve injury . 10 . 7554/eLife . 11903 . 003Figure 1 . AKT activation promotes axon regeneration and RGC survival . ( A ) Detection of AKT Phosphorylation at Ser473 in Pten-deleted RGCs labeled by anti-Tuj1 immunohistochemistry , with or without ( un ) optic nerve injury . Scale bar , 20 µm . ( B ) Quantification of phospho-AKT immunofluorescence intensity following Pten deletion . Data are presented as mean ± s . d . , n=4 per group . *p<0 . 001 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after optic nerve injury from AAV-GFP or AAV-caAkt injected eyes . * , crush site . Scale bar , 200 µm . ( D ) Quantification of regenerating axons at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , +p<0 . 05 , Two-way ANOVA with Bonferroni post hoc test . ( E ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( F ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , Student’s t test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 00310 . 7554/eLife . 11903 . 004Figure 1—figure supplement 1 . Developmental analysis of AKT phosphorylation in RGCs . ( A ) Confocal images of retinal sections showing phospho-AKT immunoreactivity in Tuj1+ RGCs at postnatal day 7 , 21 , and 60 . Scale bar , 20 µm . ( B ) Quantification of phospho-AKT immunofluorescence intensity relative to postnatal day 7 . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 001 , One-way ANOVA with Dunnett's test . ( C ) Confocal images of retinal sections showing PTEN immunoreactivity in Tuj1+ RGCs from AAV-GFP or AAV-Cre injected Ptenf/f eyes , Scale bar , 20 µm . ( D ) Quantification of PTEN immunofluorescence intensity . Data are presented as mean ± s . d . , n=3 per group . *p<0 . 001 , Student’s t test . ( E ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and AKT from AAV-GFP or AAV-caAkt injected eyes , Scale bar , 20 µm . ( F ) Quantification of AKT immunofluorescence intensity . Data are presented as mean ± s . d . , n=3 per group . *p<0 . 01 , Student’s t test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 00410 . 7554/eLife . 11903 . 005Figure 1—figure supplement 2 . Inhibition of mTORC1 partially reduces AKT-induced axon regeneration . ( A ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after injury from AAV-GFP or AAV-caAkt injected eyes treated with vehicle or rapamycin . * , crush site . Scale bar , 200 µm . ( B ) Quantification of regenerating axons at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=6 per group . *p<0 . 01 , +p<0 . 05 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( D ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 001 , One-way ANOVA with Dunnett's test . Data from AAV-GFP was used as comparison control . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 005 The mTORC1 pathway plays an important role in mediating Pten deletion-induced axon regeneration . To examine whether blockade of mTORC1 signaling also affects AKT-induced axon regeneration , we administered rapamycin to inhibit mTORC1 in the AAV-caAKT-injected retinas . In comparison with the vehicle treatment as a control , rapamycin treatment partially reduced the number of regenerating axon fibers ( Figure 1—figure supplement 2A–B ) . The residual regenerative effect could be due to incomplete blockage of mTORC1 signaling by the drug treatment , or additional mTORC1-independent pathway ( s ) may exist to regulate axon regeneration downstream of AKT . We also assessed RGC survival after rapamycin treatment . AKT-induced RGC survival was largely neutralized by rapamycin treatment ( Figure 1—figure supplement 2C–D ) . Given the prominent role of GSK3 in developmental axon extension ( Zhou et al . , 2004 ) and implication of GSK3 in the regenerative axon growth in the PNS ( Gobrecht et al . , 2014; Saijilafu et al . , 2013 ) , we next examined the role of GSK3 in CNS axon regeneration . Phosphorylation of GSK3β at Ser9 results in inactivation of its kinase activity ( Cross et al . , 1995 ) . Phospho-GSK3β level was higher in the developing retina relative to the adult age ( Figure 2—figure supplement 1 ) . Two weeks after AAV-caAKT injection in adult wild-type mice , we detected increased phosphorylation of GSK3β at Ser9 in RGCs either with or without ONC , in comparison with the AAV-GFP injected retina as a control ( Figure 2A–B ) . We also examined whether Pten deletion also led to GSK3β phosphorylation at Ser9 in RGCs . As expected , increased phosphorylation of GSK3β was observed in Pten-deficient RGCs ( Figure 2—figure supplement 2 ) . Our results indicate that GSK3β is a downstream target of PTEN/AKT signaling in RGCs . To investigate whether GSK3β inactivation mediates AKT-induced axon regeneration and RGC survival , we co-injected AAV-caAKT with AAV-GSK3β S9A , a Ser-to-Ala mutant of GSK3β that cannot be phosphorylated by AKT and therefore is constitutively active ( Eldar-Finkelman et al . , 1996 ) . In comparison with the AAV-caAKT and AAV-GFP co-injection , regenerative axon growth was significantly reduced by AAV-GSK3β S9A co-injection ( Figure 2C–D ) . However , AKT-induced RGC survival was not affected by the expression of GSK3β S9A ( Figure 2E–F ) . Our results indicate that AKT phosphorylates and thus inactivates GSK3β , leading to improved axon regeneration , but not RGC survival . 10 . 7554/eLife . 11903 . 006Figure 2 . GSK3β is an essential downstream effector to mediate AKT-induced axon regeneration . ( A ) Immunohistochemical detection of GSK3β phosphorylation at Ser9 in retinal sections from AAV-GFP or AAV-caAkt-injected eyes , either with or without optic nerve injury . Scale bar , 20 µm . ( B ) Quantification of phospho-GSK3β immunofluorescence intensity . Data are presented as mean ± s . d . , n=4 per group . *p<0 . 001 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after optic nerve injury from AAV-caAKT + AAV-GFP or AAV-caAkt + AAV-GSK3β S9A injected eyes . * , crush site . Scale bar , 200 µm . ( D ) Quantification of regenerating axons at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , +p<0 . 05 , Two-way ANOVA with Bonferroni post hoc test . ( E ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( F ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 00610 . 7554/eLife . 11903 . 007Figure 2—figure supplement 1 . Developmental analysis of GSK3β phosphorylation in RGCs . ( A ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and phospho-GSK3β at postnatal day 7 , 21 , and 60 . Scale bar , 20 µm . ( B ) Quantification of phospho-GSK3β immunofluorescence intensity relative to postnatal day 7 . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 001 , One-way ANOVA with Dunnett's test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 00710 . 7554/eLife . 11903 . 008Figure 2—figure supplement 2 . Pten deletion results in GSK3β phosphorylation in RGCs . ( A ) Detection of GSK3β phosphorylation in RGCs by double immunolabeling for phospho-GSK3β ( Ser9 ) and Tuj1 in retinal sections from Ptenf/f mice injected with AAV-GFP or AAV-Cre , with or without optic nerve crush . Scale bar , 20 µm . ( B ) Quantification of phospho-GSK3β immunofluorescence intensity . Data are presented as mean ± s . d . , n=4 per group . *p<0 . 001 , Two-way ANOVA with Bonferroni post hoc test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 008 To investigate whether deletion of Gsk3b promotes RGC axon regeneration , we injected AAV-Cre , or AAV-GFP as a control , in the adult Gsk3bf/f mice ( Patel et al . , 2008 ) ( Figure 3—figure supplement 1 ) , and examined axon regeneration two weeks after ONC . While no obvious regenerating axon fibers were observed in the AAV-GFP injected control , Gsk3b deletion resulted in improved axon regeneration ( Figure 3A ) . We next examined the time course of axon regeneration in Gsk3b-deficient RGCs after injury ( Figure 3—figure supplement 2 ) . At 1 day after ONC , optic nerve fibers terminated at the crush site in Gsk3bf/f mice injected with either AAV-GFP or AAV-Cre . At 3 days after ONC , we observed axon sprouting in the Gsk3b-deleted RGCs . At 7 days after ONC , regenerating axon fibers extended beyond the lesion site in Gsk3b-deficient RGCs , overcoming the inhibitory environment at the lesion site labeled by immunohistochemistry for chondroitin sulfate proteoglycan ( CSPG ) and glial fibrillary acidic protein ( GFAP ) . However , deletion of Gsk3a did not stimulate the regenerative response of injured axons ( Figure 3A ) , although AKT activation also increased its phosphorylation at Ser21 ( Figure 3—figure supplement 3 ) . To further examine whether elimination of the kinase activity of GSK3β is responsible for Gsk3b deletion-induced axon regeneration , we injected AAV-GSK3β K85A , a kinase dead mutant of GSK3β that inhibits endogenous GSK3β in a dominant negative manner ( Dominguez et al . , 1995 ) , into the vitreous body of adult wild-type mice . Two weeks after ONC , we observed many regenerating axon fibers extending beyond the lesion site in the AAV-GSK3β K85A injected mice , a regenerative response slightly weaker in comparison with Gsk3b deletion ( Figure 3A , C ) . We also analyzed neuronal survival two weeks after ONC in the retinas of Gsk3a deletion , Gsk3b deletion , and AAV-GSK3β K85A injection . None of these manipulations changed the rate of RGC survival ( Figure 3B , D ) . Our results indicate that the deletion of Gsk3b , but not Gsk3a , promotes RGC axon regeneration likely through inactivation of the kinase activity of GSK3β . 10 . 7554/eLife . 11903 . 009Figure 3 . Deletion or inactivation of GSK3β promotes axon regeneration . ( A ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after optic nerve crush from Gsk3a−/− mice , Gsk3bf/f mice injected with AAV-GFP or AAV-Cre , or AAV-GSK3β K85A injected wild-type mice . * , crush site . Scale bar , 200 µm . ( B ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( C ) Quantification of regenerating axons at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , Two-way ANOVA with Bonferroni post hoc test . ( D ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 00910 . 7554/eLife . 11903 . 010Figure 3—figure supplement 1 . Cre-mediated Gsk3b deletion in RGCs . ( A ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and GSK3β from AAV-GFP or AAV-Cre-injected Gsk3bf/f eyes , Scale bar , 20 µm . ( B ) Quantification of GSK3β immunofluorescence intensity . Data are presented as mean ± s . d . , n=3 per group . *p<0 . 01 , Student’s t test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 01010 . 7554/eLife . 11903 . 011Figure 3—figure supplement 2 . A time course study of axon regeneration in Gsk3b-deleted RGCs . ( A-C ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 1 , 3 , or 7 days from Gsk3bf/f mice injected with AAV-GFP or AAV-Cre . * , crush site . Scale bar , 100 µm . ( D ) Quantification of regenerating axons at 7 days after injury from Gsk3bf/f mice injected with AAV-GFP or AAV-Cre at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=4 per group . *p<0 . 01 , Two-way ANOVA with Bonferroni post hoc test . ( E ) Confocal images of optic nerve sections showing double immunolabeling for GAP43+ regenerating axons and chondroitin sulfate proteoglycan ( CSPG ) . Scale bar , 20 µm . ( F ) Confocal images of optic nerve sections showing double immunolabeling for GAP43+ regenerating axons and glial fibrillary acidic protein ( GFAP ) . Scale bar , 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 01110 . 7554/eLife . 11903 . 012Figure 3—figure supplement 3 . AKT activation increases GSK3α phosphorylation . ( A ) Immunohistochemical detection of GSK3α phosphorylation at Ser21 in retinal sections from AAV-GFP or AAV-caAkt injected eyes , with or without optic nerve injury . Scale bar , 20 µm . ( B ) Quantification of RGC phospho-GSK3α immunofluorescence intensity . Data are presented as mean ± s . d . , n=4 per group . *p<0 . 001 , Two-way ANOVA with Bonferroni post hoc test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 012 Activation of the mTORC1 pathway is a well-established downstream signaling mechanism in Pten deletion-induced axon regeneration ( Park et al . , 2008; Yang et al . , 2014 ) . And our results demonstrate that GSK3β may represent another important signaling pathway downstream of PTEN/AKT to regulate CNS axon regrowth . To distinguish whether Gsk3b deletion-induced axon regeneration is through activation of the mTORC1 pathway or through an mTORC1-independent mechanism , we first examined whether mTORC1 signaling is altered by Gsk3b deletion using immunohistochemistry for phospho-S6 ribosomal protein ( p-S6 ) to monitor the activity of mTORC1 in RGCs ( Hay and Sonenberg , 2004; Park et al . , 2008 ) . While AKT activation significantly increased the mTORC1 activity , evidenced by a higher fluorescence intensity of p-S6 immunoreactivity in RGCs in adult wild-type mice injected with AAV-caAKT compared to the control injection , Gsk3b deletion did not perturb the activity of mTORC1 as the immunofluorescence intensity of p-S6 remained at the same low level as the control ( Figure 4A–B ) . To further assess whether the mTORC1 pathway contributes to the regenerative effects induced by Gsk3b deletion , we administered rapamycin to inhibit mTORC1 activity in Gsk3b-deleted RGCs . While rapamycin markedly reduced AKT-induced axon regeneration and RGC survival ( Figure 1—figure supplement 2 ) , Gsk3b deficiency-induced axon regeneration was not affected by rapamycin treatment , and neither was RGC survival in Gsk3b-deleted RGCs ( Figure 4—figure supplement 1 ) . However , treatment with the protein synthesis inhibitor anisomycin significantly reduced Gsk3b deletion induced axon regeneration ( Figure 4—figure supplement 1 ) . Our results indicate that GSK3β-mediated effects on axon regeneration require protein synthesis but are achieved through an mTORC1-independent mechanism . 10 . 7554/eLife . 11903 . 013Figure 4 . Gsk3b deletion does not alter the mTORC1 pathway . ( A ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and phospho-S6 from AAV-GFP , AAV-caAkt , or AAV-Cre ( Gsk3bf/f ) -injected eyes , with or without optic nerve injury . Scale bar , 20 µm . ( B ) Quantification of phospho-S6 immunofluorescence intensity . Data are presented as mean ± s . d . , n=4 per group . *p<0 . 001 , Two-way ANOVA with Bonferroni post hoc test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 01310 . 7554/eLife . 11903 . 014Figure 4—figure supplement 1 . Gsk3b deletion-induced axon regeneration is sensitive to protein synthesis inhibition but not to mTORC1 inhibition . ( A ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after injury in Gsk3bf/fmice injected with AAV-Cre with vehicle , rapamycin , or anisomycin treatment . * , crush site . Scale bar , 200 µm . ( B ) Quantification of regenerating axons at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=6 per group . *p<0 . 01 , +p<0 . 05 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( D ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 014 GSK3β regulates many cellular processes through phosphorylation of its diverse substrates that include transcription factors , cytoskeletal proteins , motor proteins , and proteins involved in the regulation of cell growth ( Liu et al . , 2012; Saijilafu et al . , 2013; Sutherland , 2011 ) . eIF2B ( eukaryotic translation initiation factor 2B ) is a heteropentameric guanine nucleotide exchange factor that converts eIF2 ( eukaryotic translation initiation factor 2 ) from the inactive GDP–bound form to the active GTP-bound complex , representing a rate-limiting step in protein translation initiation ( Campbell et al . , 2005; Pavitt , 2005 ) . eIF2Bε ( eIF2B epsilon subunit ) , the largest catalytic subunit of eIF2B , is inhibited by GSK3β phosphorylation at the Ser535 site , linking regulation of global protein synthesis to GSK3β signaling ( Pap and Cooper , 2002; Wang et al . , 2001 ) . Due to the importance of protein synthesis in the regenerative growth of axons , we examined the role of eIF2Bε in optic never regeneration . Phosphorylation of eIF2Bε at Ser535 in RGCs was significantly higher in adult mice in comparison with that in the developing retina ( Figure 5—figure supplement 1 ) , indicating that elF2Bε-controlled protein translation is maintained at a low basal level in adult RGCs . Optic nerve injury alone did not activate eIF2Bε as the phosphorylation of eIF2Bε was detected at similar levels to uninjured retinas after ONC ( Figure 5A ) . By contrast , phospho-eIF2Bε ( Ser535 ) immunoreactivity was markedly reduced in Gsk3b-deleted RGCs in the presence or absence of axon injury ( Figure 5A–B ) , indicating increased eIF2Bε activity for protein translation . Gsk3a deletion did not change eIF2Bε phosphorylation ( Figure 5—figure supplement 2 ) , suggesting that GSK3α may not regulate eIF2Bε activity in RGCs . To determine the role of eIF2Bε in GSK3β-mediated axon regeneration , we inhibited the activity of eIF2Bε in Gsk3b-deleted RGCs by co-injection of AAV-eIF2Bε E572A , a dominant negative mutant of eIF2Bε that lacks catalytic activity but is capable of forming heteropentameric eIF2B complex ( Boesen et al . , 2004; Wang and Proud , 2008 ) . Two weeks after ONC , the number of regenerating axons was significantly reduced by AAV-eIF2Bε E572A co-injection ( Figure 5C–D ) . To examine whether protein synthesis is involved in eIF2Bε-regulated axon regeneration , we used Alexa Fluor-conjugated OPP incorporation assay to label newly translated proteins . While the basal level of protein translation in RGCs was suppressed by treatment with the protein synthesis inhibitor anisomycin , Gsk3b deletion resulted in a significant increase in OPP-labeled new synthesized proteins , which was largely negated by eIF2Bε E572A treatment ( Figure 5E–F ) . These results clearly demonstrate that protein synthesis is tightly controlled by GSK3β/eIF2Bε signaling during axon regeneration . However , inhibition of eIF2Bε had no effect on the survival of Gsk3b-deleted RGCs ( Figure 5G–H ) . 10 . 7554/eLife . 11903 . 015Figure 5 . eIF2Bε is required for Gsk3b deletion-induced axon regeneration . ( A ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and phospho-eIF2Bε from Gsk3bf/fmice injected with AAV-GFP ( Control ) or AAV-Cre ( Gsk3b-/- ) , with or without optic nerve injury . Scale bar , 20 µm . ( B ) Quantification of phospho-eIF2Bε immunofluorescence intensity . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 001 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after optic nerve crush from Gsk3bf/fmice injected with AAV-Cre + AAV-GFP or AAV-Cre + AAV-eIF2Bε E572A . * , crush site . Scale bar , 200 µm . ( D ) Quantification of regenerating axons from Gsk3bf/f mice injected with AAV-Cre + AAV-GFP or AAV-Cre + AAV-eIF2Bε E572A at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=9 per group . *p<0 . 01 , Two-way ANOVA with Bonferroni post hoc test . ( E ) OPP Alexa Fluor 594 protein synthesis assay in retinal whole-mounts from Gsk3bf/f mice injected with AAV-GFP ( Control ) , AAV-Cre ( Gsk3b-/- ) + AAV-GFP , or AAV-Cre + AAV-eIF2Bε E572A or treated with anisomycin . Scale bar , 25 µm . ( F ) Quantification of OPP fluorescence intensity . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , One-way ANOVA with Tukey’s test . ( G ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( H ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 01510 . 7554/eLife . 11903 . 016Figure 5—figure supplement 1 . Developmental analysis of eIF2Bε phosphorylation in RGCs . ( A ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and phospho-eIF2Bε at postnatal day 7 , 21 , and 60 . Scale bar , 20 µm . ( B ) Quantification of phospho-eIF2Bε immunofluorescence intensity relative to postnatal day 7 . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 001 , One-way ANOVA with Dunnett's test . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 01610 . 7554/eLife . 11903 . 017Figure 5—figure supplement 2 . Analysis of eIF2Bε phosphorylation in Gsk3a knockout mice . ( A ) Confocal images of retinal sections showing double immunolabeling for Tuj1+ RGCs and phospho-eIF2Bε in Gsk3a-/- mice or their wild type litter mates as a control , with or without optic nerve injury . Scale bar , 20 µm . ( B ) Quantification of phospho-eIF2Bε immunofluorescence intensity . Data are presented as mean ± s . d . , n= 5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 017 As AKT is further upstream of eIF2Bε , we next examined the effect of eIF2Bε inhibition on AKT-induced axon regeneration . At 2 weeks after ONC , the regenerative response stimulated by AAV-caAKT injection was significantly reduced by co-injection of AAV-eIF2Bε E572A ( Figure 6A–B ) , while AKT-induced RGC survival was not perturbed by eIF2Bε E572A inhibition ( Figure 6C–D ) . Taken together , our results demonstrate that eIF2Bε is required for both AKT and GSK3β-mediated effects on axon regeneration , but plays a minimal role in RGC survival after axon injury . 10 . 7554/eLife . 11903 . 018Figure 6 . eIF2Bε is required for AKT-induced axon regeneration . ( A ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after injury from AAV-caAKT + AAV-GFP or AAV-caAkt + AAV-eIF2Bε E572A injected eyes . * , crush site . Scale bar , 200 µm . ( B ) Quantification of regenerating axons from AAV-caAKT + AAV-GFP or AAV-caAkt + AAV-eIF2Bε E572A injected eyes at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 05 , +p<0 . 05 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( D ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 018 To investigate whether activation of eIF2Bε alone is sufficient to promote axon regeneration , we injected AAV-eIF2Bε S535A , a constitutively active mutant of eIF2Bε that is resistant to GSK3β phosphorylation ( Pap and Cooper , 2002 ) , in the adult wild-type retina . As expected , expression of eIF2Bε S535A significantly increased protein synthesis in RGCs compared with AAV-GFP injection as a control ( Figure 7A–B ) . At 2 weeks after ONC , we observed many regenerating axon fibers extending past the crush site in the eIF2Bε S535A treated retinas ( Figure 7C–D ) . The regenerative effect induced by eIF2Bε activation was reduced by protein synthesis inhibition with anisomycin treatment , but was not affected by mTORC1 inhibition with rapamycin treatment ( Figure 7—figure supplement 1 ) , indicating that mTORC1 signaling plays a minimal role in eIF2Bε-induced axon regeneration . We further assessed whether activation of eIF2Bε affects neuronal survival at 2 weeks after ONC , and found that AAV-eIF2Bε S535A injection did not change the rate of RGC survival in comparison with the control injection ( Figure 7E–F ) . These results demonstrate that activation of eIF2Bε promotes axon regeneration with minimal effects on RGC survival . 10 . 7554/eLife . 11903 . 019Figure 7 . eIF2Bε activation promotes axon regeneration . ( A ) OPP Alexa Fluor 594 protein synthesis assay in retinal whole-mounts from AAV-GFP or AAV-eIF2Bε S535A injected eyes . Scale bar , 25 µm . ( B ) Quantification of OPP fluorescence intensity . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , Student’s t test . ( C ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after injury from wild-type mice injected with AAV-GFP or AAV-eIF2Bε S535A . * , crush site . Scale bar , 200 µm . ( D ) Quantification of regenerating axons from retinas injected with AAV-GFP or AAV-eIF2Bε S535A at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , +p<0 . 05 , Two-way ANOVA with Bonferroni post hoc test . ( E ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( F ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 01910 . 7554/eLife . 11903 . 020Figure 7—figure supplement 1 . eIF2Bε-induced axon regeneration is sensitive to protein synthesis inhibition but not to mTORC1 inhibition . ( A ) Confocal images of optic nerve sections showing regenerating axons labeled by anti-GAP43 immunohistochemistry at 2 weeks after injury from AAV-eIF2Bε S535A injected eyes with vehicle , rapamycin , or anisomycin treatment . * , crush site . Scale bar , 200 µm . ( B ) Quantification of regenerating axons at different distances distal to the lesion site . Data are presented as mean ± s . d . , n=5 per group . *p<0 . 01 , Two-way ANOVA with Bonferroni post hoc test . ( C ) Confocal images of retinal whole-mounts showing surviving Tuj1+ RGCs at 2 weeks after optic nerve injury . Scale bar , 50 µm . ( D ) Quantification of RGC survival at 2 weeks after injury , expressed as a percentage of total Tuj1+ RGCs in the uninjured retina . Data are presented as mean ± s . d . , n=5 per group . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 02010 . 7554/eLife . 11903 . 021Figure 7—figure supplement 2 . A schematic illustration of GSK3β/eIF2Bε and mTORC1 signaling in AKT-induced CNS axon regeneration . DOI: http://dx . doi . org/10 . 7554/eLife . 11903 . 021
Unlike the PNS neurons that are capable of regenerating axons after nerve injury , neurons in the adult mammalian CNS fail to regenerate axons after nerve damages . Both the inhibitory extrinsic environment and the diminished intrinsic regenerative capability in the adult CNS contribute to the poor regeneration of axons after injury . Removal of the inhibitory influences in the extracellular environment is insufficient to stimulate a major regenerative response . On the other hand , activation of intrinsic signaling pathways within CNS neurons has yielded encouraging results in promoting axon regeneration ( de Lima et al . , 2012; Kurimoto et al . , 2010; Liu et al . , 2011; Park et al . , 2008; Smith et al . , 2009; Sun et al . , 2011; Yang et al . , 2014 ) . Deletion of Pten , a negative regulator of the PI3K/AKT pathway , stimulated robust axon regeneration ( Park et al . , 2008 ) . We found that AKT was phosphorylated at Ser473 and thus activated in the Pten-deleted RGCs . Significantly , expression of a constitutively active myristoylated AKT promoted both axon regeneration and RGC survival . AKT regulates a number of diverging downstream signaling pathways through phosphorylation of diverse cellular targets ( Manning and Cantley , 2007 ) . Previous studies have established that the mTORC1 pathway is an important downstream signaling in mediating Pten deletion-induced axon regeneration , as reactivation of this pathway by TSC1 deletion or S6K1 activation effectively promotes axon regeneration ( Park et al . , 2008; Yang et al . , 2014 ) . However , mechanisms other than the mTORC1 pathway downstream of Pten deletion have not been elucidated . Several studies have demonstrated the role of GSK3 signaling in developmental axon growth ( Hur et al . , 2011; Hur and Zhou , 2010; Kim et al . , 2006 ) . In the adult PNS , GSK3 signaling regulates mammalian sensory axon regeneration by inducing the expression of Smad1 ( Saijilafu et al . , 2013 ) . In the spinal cord injury model , neuronal deletion of Gsk3b enhances dorsal column axon regeneration via CRMP2-regulated microtubule dynamics , specifically in the growth cone ( Liz et al . , 2014 ) . Here , using the optic nerve injury model , we have demonstrated that GSK3β plays a pivotal role in regulating CNS axon regeneration . Pten deletion or AKT activation led to phosphorylation of GSK3β at the site of Ser9 , resulting in inactivation of its kinase activity in RGCs . Expression of GSK3β S9A , a kinase active mutant of GSK3β , significantly reduced axon regeneration induced by AKT activation . Deletion of Gsk3b or inactivation of GSK3β with the expression of GSK3β K85A , a kinase dead mutant of GSK3β , was sufficient to promote axon regeneration . However , GSK3α did not have these same effects . These results are in line with studies showing GSK3 isoform-specific effects in developmental axon growth ( Chen et al . , 2012 ) . Indeed GSK3α and GSK3β have been shown to act on different substrates in the brain ( Soutar et al . , 2010 ) . We did not assess the results of complete elimination of GSK3 activity , which has been associated with the suppression of axon growth in developmental studies ( Kim et al . , 2006 ) . Interestingly , unlike AKT activation that promoted both axon regeneration and RGC survival , GSK3 inactivation did not promote RGC survival , suggesting that mTORC1 and/or other unidentified factors downstream of AKT are involved in promoting the survival of RGCs . How might GSK3β signaling regulate axon regeneration ? eIF2Bε , the largest catalytic subunit of eIF2B , plays an important role in initiating protein translation in all eukaryotic cells ( Pavitt , 2005 ) . The activity of eIF2Bε is inhibited by GSK3β phosphorylation . In response to insulin , GSK3β kinase activity is inhibited , eIF2Bε is dephosphorylated and thus more active , leading to an enhancement of protein translation . In the CNS , eIF2Bε-controlled protein translation is required for the development and maintenance of brain white matter , which is composed of bundles of myelinated axons ( Fogli and Boespflug-Tanguy , 2006; Geva et al . , 2010 ) . We found that eIF2Bε is a key downstream target of GSK3β to regulate axon regeneration after optic nerve injury . Inhibition of eIF2Bε with the dominant negative mutant eIF2Bε E572A significantly reduced axon regeneration induced by Gsk3b deletion or AKT activation . More importantly , activation of eIF2Bε with the constitutively active mutant eIF2Bε S535A was sufficiently effective to promote axon regeneration , emphasizing the importance of protein synthesis as a major neuronal intrinsic mechanism in CNS axon regeneration . Protein synthesis is primarily regulated at the initiation phase , which involves binding of the methionyl–transfer RNA ( Met–tRNAiMet ) to the small 40s ribosomal subunit to form the 43 s pre-initiation complex , which then binds to an mRNA to form the 48 s pre-initiation complex that scans the AUG start codon . Each of these steps in translation initiation is facilitated by proteins referred to as eukaryotic initiation factors eIFs ( Klann and Dever , 2004 ) . The tRNAiMet binding to the 40 s ribosomal subunit requires the exchange of GTP-bound for GDP-bound eIF2 , which is catalyzed by the guanine-nucleotide exchange factor eIF2B . Therefore , GSK3β-mediated eIF2Bε activation enhances the recycling of eIF2 for further rounds of translational initiation , constituting a critical regulatory mechanism to control the intrinsic regenerative ability of RGCs after nerve injury . The 5’ cap-binding protein eIF4E is required for the binding of the 43 s pre-initiation complex to most eukaryotic mRNAs . mTORC1 activation enhances protein synthesis through phosphorylation of 4E-BP ( translation initiation factor 4E-binding protein ) , a negative regulator of eIF4E . However , 4E-BP inhibition is not sufficient to promote axon regeneration as co-deletion of 4E-BP1/2 in RGCs does not stimulate optic nerve regeneration . Interestingly , activation of S6K1 , another downstream factor of mTORC1 , promotes RGC axon regeneration ( Yang et al . , 2014 ) . S6K1 activation promotes the translation of 5’ TOP ( terminal oligopyrimidine tract ) mRNAs , which encode exclusively components of the translation machinery including ribosomal proteins , elongation factors , and poly ( A ) -binding protein ( PABP ) ( Hay and Sonenberg , 2004 ) . It appears that mTORC1 and GSK3β signaling , although acting independently to regulate protein translation downstream of AKT , converge on a common translation regulatory mechanism for axon regeneration . In summary , our results provide evidence that the AKT-GSK3β-eIF2Bε signaling module plays a central role in determining the intrinsic axonal growth ability of mature CNS neurons ( Figure 7—figure supplement 2 ) . Identifying various signaling pathways may enable combinatorial treatment to promote axon regeneration after CNS injury .
C57BL/6 mice were purchased from The Jackson Laboratory ( Bar Harbor , Maine ) . Ptenf/f mice were obtained from Dr . William Cafferty’s laboratory ( Yale University , New Haven , USA ) . Gsk3a−/− and Gsk3bf/f mice were kindly provided by Dr . James R Woodgett ( McMaster University , Hamilton , Ontario , Canada ) . All studies adhered to the procedures consistent with animal protocols approved by the IACUC at Yale University . pAAV-Cre and pAAV-GFP plasmids were kindly provided by Dr . Kevin Park ( University of Miami ) . For plasmids construction , protein-coding region in pAAV-GFP was replaced by the coding sequence of caAKT ( myrAkt delta4-129 , Addgene plasmid # 10841 , Cambridge , MA ) , GSK3β S9A ( obtained from Dr . Marc B Hershenson , University of Michigan ) , GSK3β K85A ( Addgene plasmid # 14755 ) , eIF2Bε S535A ( obtained from Dr . Geoffrey M Cooper , Boston University ) and eIF2Bε E572A . pAAV2-RC ( Stratagene , La Jolla , CA ) and the Helper plasmid were used for co-transfection in HEK293T cells . Discontinuous iodixanol gradient ultracentrifugation was used to purify AAV . AAV titers , determined by real-time PCR , were in the range of 1–5 x 1012 genome copies per milliliter . Mice were anesthetized with a mix of ketamine ( 100 mg/kg ) and xylazine ( 10 mg/kg ) by intraperitoneal injection . For intravitreal injection , the micropipette was inserted just behind the ora serrata , and 1 µl of AAV solution was injected in the vitreous body . Two weeks after viral injection , optic nerve was exposed and crushed intraorbitally with jeweler’s forceps for 5 s approximately 1 mm behind the optic disc . Eye ointment was applied to protect the cornea after surgery . Eyes with the attached optic nerve segment , surgically removed from perfused mice , were post-fixed in 4% PFA . Retinas were dissected out for either whole-mount preparations or cryosections . The optic nerve was separated from the eye and cut longitudinally with a Leica cryostat . Retinal whole-mounts or sections were blocked in the staining buffer containing 5% normal donkey serum and 0 . 1% Triton X-100 in PBS for 1 hr before incubation with primary antibodies . Primary antibodies used: Tuj1 ( Covance , 1:500 , Princeton , NJ ) , PTEN ( Cell Signaling Technology , 1:250 , Danvers , MA ) , p-AKT ( Cell Signaling Technology , 1:200 ) , AKT ( Cell Signaling Technology , 1:250 ) , p-GSK3α ( Abcam , 1:250 , UK ) , p-GSK3β ( Cell Signaling Technology , 1:400 ) , GSK3b ( Cell Signaling Technology , 1:250 ) , p-S6 ( Cell Signaling Technology , 1:200 ) , p-eIF2Bε ( EMD Millipore , 1:300 , Billerica , MA ) , GAP43 ( obtained from Dr . Larry Benowitz , 1:500 ) , CSPG ( Sigma , 1:200 , St . Louis , MO ) , and GFAP ( Cell Signaling Technology , 1:200 ) . Secondary antibodies used: DyLight Cy3/594/647-conjugated AffiniPure antibodies ( Jackson ImmunoResearch , 1:500 , West Grove , PA ) . Confocal images were acquired using a Zeiss LSM 510 EXCITER microscope . Fluorescence channel colors were switched for co-localization studies if necessary . Images were analyzed and organized using ImageJ and Photoshop . For RGC counting , retinal whole-mounts were immunostained with Tuj1 antibody , and 8–12 fields ( 321 x 321 µm ) were randomly sampled from the peripheral regions of each retina . Regenerating axons was quantified by counting the number of GAP43 positive axons at different distance from the crush site in four sections per nerve , as described previously ( Park et al . , 2008 ) . The cross-sectional width of the nerve was measured at the point at which the counts were taken and was used to calculate the number of axons per millimeter of the width of the nerve . The number of axons per millimeter was then averaged over all sections . Σad , the total number of axons extending distance d in a nerve having a radius of r , was estimated by summing over all sections having a thickness t: Sad= πr2 x [average axons/mm]/t . Rapamycin was administered as previously described ( Park et al . , 2008 ) . Rapamycin ( LC Laboratories , Woburn , MA ) was dissolved at 20 mg/ml in ethanol for stock . Before each administration , rapamycin was diluted in 5% Tween 80 , 5% polyethylene glycol 400 ( 1 . 0 mg/ml ) in PBS . Rapamycin at 6 mg/kg or the vehicle was injected intraperitoneally once every 2 days after AAV injection . Anisomycin ( Cayman Chemical , Ann Arbor , MI ) was dissolved in DMSO to prepare a stock solution and diluted in PBS before each administration . Anisomycin ( 30 mg/kg body weight ) or vehicle was injected subcutaneously daily after optic nerve crush . To inhibit OPP incorporation , anisomycin ( 50 μg/ml ) was injected intravitreally 1 hr before and together with OPP administration . For new protein synthesis analysis and imaging , we used Click-iT Plus OPP Alexa Fluor 594 Protein Synthesis Assay Kit ( Life Technologies , Carlsbad , CA ) . In this assay , OPP ( O-propargyl-puromycin ) is efficiently incorporated into newly translated proteins that can be detected by fluorescently labeled Alexa Fluor dye . One microliter of OPP reagent ( 200 µM in PBS ) was injected into the vitreous body per eye . One hour after injection , eyes were dissected out after perfusion and were post-fixed in 4% PFA for 4 hr . OPP detection was performed according to the manufacturer’s protocol . Retinas were subsequently immunostained with the Tuj1 antibody to label RGCs . | The central nervous system consists of the neurons that make up the brain , retina , and spinal cord . Neurons transmit electrical signals along a cable-like structure called an axon . However , an axon cannot regenerate itself , and so injuries that crush or sever the axons can lead to permanent damage . This happens for two reasons: neurons don’t have the same regenerative ability as other cells , and the environment in the central nervous system restricts cell growth . The optic nerve transmits visual information from the eye to the brain . Studies in mice with a damaged optic nerve show that it is possible to regenerate the axons of neurons that lack a protein known as PTEN . These studies revealed one molecular pathway by which eliminating PTEN helps to boost the regrowth of axons . Now , Guo et al . identify another independent pathway by which eliminating PTEN helps promote axon regeneration in damaged mouse optic nerves . This pathway starts with a growth-promoting enzyme called AKT , which is turned on in neurons that lack PTEN . Indeed , injecting mice with an active form of this enzyme caused the optic nerve fiber to regrow in mice whose optic nerve had been crushed . Further experiments revealed that AKT activates a pathway in which another enzyme called GSK3β acts on a protein called eIF2Bε . A future challenge is to simultaneously manipulate the different signaling pathways that have been linked to axon regrowth to investigate whether this combined approach could help repair damage to the central nervous system . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | GSK3β regulates AKT-induced central nervous system axon regeneration via an eIF2Bε-dependent, mTORC1-independent pathway |
The macaque orbitofrontal cortex ( OFC ) is essential for selecting goals based on current , updated values of expected reward outcomes . As monkeys consume a given type of reward to satiety , its value diminishes , and OFC damage impairs the ability to shift goal choices away from devalued outcomes . To examine the contributions of OFC’s components to goal selection , we reversibly inactivated either its anterior ( area 11 ) or posterior ( area 13 ) parts . We found that neurons in area 13 must be active during the selective satiation procedure to enable the updating of outcome valuations . After this updating has occurred , however , area 13 is not needed to select goals based on this knowledge . In contrast , neurons in area 11 do not need to be active during the value-updating process . Instead , inactivation of this area during choices causes an impairment . These findings demonstrate selective and complementary specializations within the OFC .
Damage to the prefrontal cortex often disrupts the translation of knowledge into action . Patients with such lesions can accurately explain various intentions , rules and social conventions , but nevertheless violate them ( Milner , 1963; Luria , 1966 ) . As Teuber ( 1972a ) put it: "Patients with frontal lobe disease seem to perceive the mistakes they make , but are unable to use the information to guide their behavior . " As a result , it has been argued that a principal function of the frontal lobe lies in translating knowledge into behavioral goals ( Teuber , 1972b ) . The uncoupling of knowledge from goals has been generalized as goal neglect ( Duncan et al . , 1996; 2008 ) , and this concept extends beyond laboratory tasks , such as card sorting , to the social domain . For example , patients with damage to the orbital and medial prefrontal cortex can accurately convey appropriate moral principles and social conventions , but then violate them in the course of their behavior ( Bechara et al . , 1994 ) . In some cases , this deficit occurs in isolation , with normal behavior in other cognitive domains ( Damasio et al . , 1991; Fellows , 2011 ) . An impairment in translating knowledge into behavioral goals could be fundamental to many addictive behaviors , compulsive disorders , and psychopathologies . A simple account for goal neglect is that the cognitive processes leading to goal selection have , in such cases , poor access to knowledge about expected outcomes and their value . One way to assess an animal’s knowledge of outcome values is through reinforcer devaluation in the devaluation task . Although the subjective value of a food or fluid outcome is influenced by many factors , such as probability , magnitude , and the effort required to obtain it , the devaluation task isolates a different and independent aspect of subjective value: an outcome’s worth at a particular moment based on the individual’s current state . In one version of this task , monkeys first learn that some objects are associated with one kind of food ( food 1 ) , while other objects are associated with a different food ( food 2 ) . Next , a selective satiation procedure temporarily devalues one food type , and monkeys are given a series of choice tests in which food-1 objects are pitted against food-2 objects . If monkeys can update the value of expected outcomes and link this information to their goal choices , they will shift these choices away from objects associated with the devalued food , a phenomenon called the devaluation effect . Other kinds of visual stimuli can be substituted for physical objects in these experiments , as in the present experiment . The key question for the present study is when various brain areas make their contribution to the devaluation effect . By inactivating the amygdala of monkeys either before or after the selective satiation procedure , Wellman et al . ( 2005 ) showed that neurons in the amygdala need to be active during the satiation procedure for normal value updating to occur . Specifically , inactivation of the amygdala before selective satiation disrupted devaluation effects , whereas inactivation after satiation , but before the choice phase of the experiment , had no effect . Thus , the amygdala is essential for value updating but , once that has occurred , it is no longer essential for making goal choices based on those valuations . In addition to the amygdala , the orbitofrontal cortex ( OFC ) makes a necessary contribution to devaluation effects in monkeys ( Izquierdo et al . , 2004; Machado and Bachevalier , 2007; Baxter et al . , 2009; Rudebeck et al . , 2013 ) . Further , functional interaction of the amygdala and OFC is required ( Baxter et al . , 2000 ) . The granular prefrontal areas that compose the primate OFC are also essential for representing expected outcomes more generally , including both their sensory properties and subjective value ( Padoa-Schioppa , 2011; Rudebeck and Murray , 2014 ) , and the agranular OFC areas in rodents have similar properties ( Schoenbaum et al . , 2009 ) . Within the primate OFC , areas 13 and 11 , considered together , are necessary for normal devaluation effects , but area 14 is not ( Rudebeck and Murray , 2011 ) . Thus , the functional contribution of the OFC to this kind of value-based goal selection is mediated by the sensory network proposed by Carmichael and Price ( 1996 ) , which receives inputs from gustatory , olfactory , visceral , and visual sensory areas ( Carmichael and Price , 1995; Saleem et al . , 2008 ) . Each of these inputs presumably contributes to representations of outcome expectancies . Consistent with this idea , when monkeys view images that predict specific amounts or types of outcomes , the activity of neurons in both areas 13 and 11 signals the expected value of those upcoming rewards ( Hikosaka and Watanabe , 2000; Tremblay and Schultz , 2000; Wallis and Miller , 2003; Padoa-Schioppa and Assad , 2006 ) . Area 11 lies anterior to area 13 , and these areas differ in both cytoarchitecture and connections ( Preuss and Goldman-Rakic , 1991; Carmichael and Price , 1994; Saleem et al . , 2008 ) . Little , however , is known about their functional specializations . Anterior–posterior dissociations have been reported in electrophysiological studies in macaques ( Mora et al . , 1980; Kobayashi et al . , 2010 ) , and in functional neuroimaging studies in humans ( Sescousse et al . , 2010; Klein-Flugge et al . , 2013 ) , but these correlational findings leave open the causal contributions of these two OFC components to behavior . Here we show that they make selective and complementary contributions to reward-based goal selection , and in particular to translating abstract , updated knowledge about outcome valuations into advantageous choices .
As expected , when saline was infused into either area 13 or 11 the monkeys showed robust devaluation effects ( Figure 4 ) . That is , after selective satiation , monkeys shifted their choices to images that yielded the higher value outcome . The effect of temporary inactivations of the OFC varied markedly across conditions . A repeated-measures ANOVA on proportion shifted with factors of Treatment ( THIP , saline ) , Region ( area 13 , area 11 ) , and Time ( before satiation , after satiation ) revealed a significant three-way interaction ( F ( 1 , 4 ) = 16 . 82 , p = 0 . 015 , partial η2 = 0 . 81 ) , indicating that behavioral effects of inactivation differed both as a function of area and of timing . Post-hoc tests , detailed below , revealed a double dissociation of function within the OFC . When monkeys selected goals based on updated subjective valuations , inactivation of area 13 but not area 11 during the value-updating phase caused an impairment . In contrast , inactivation of area 11 but not area 13 during the goal-selection phase of the experiment caused an impairment . 10 . 7554/eLife . 11695 . 006Figure 4 . Effect of temporary inactivation of area 13 and area 11 on image choices following reinforcer devaluation . Proportion shifted represents the shift in image choices after selective satiation ( Day 4 ) relative to baseline ( Day 2 ) , combined across probe tests; the higher the score the greater the shift away from choices associated with the devalued food . Asterisks indicate significant differences ( area 13 , saline vs . THIP Pre , p = 0 . 005; area 11 , saline vs . THIP Post , p = 0 . 012 ) . Bars represent group means and symbols show the scores of individual monkeys . White bars , saline infusions; black bars , THIP infusions administered before selective satiation; gray bars , THIP infusions administered after selective satiation . DOI: http://dx . doi . org/10 . 7554/eLife . 11695 . 006 Scores obtained when saline was infused before selective satiation did not differ from those obtained when saline was infused after selective satiation ( repeated-measures ANOVA with factors of Region and Time; for all main effects and interactions: F ( 1 , 4 ) < 1 . 48 , p > 0 . 29 ) ; we therefore pooled saline data within each region for subsequent analysis . Planned follow-up comparisons revealed that THIP infusions into area 13 before satiation significantly lowered proportion shifted scores relative to those obtained with saline infusions ( two-tailed paired t-tests with Bonferroni-corrected alpha = 0 . 025 for all follow-up tests; t4 = 5 . 61 , p = 0 . 005 , d = 2 . 51 ) . Infusions of THIP into area 13 after satiation had no effect ( t4 = 0 . 24 , p = 0 . 826 ) . Because normal functioning of area 13 was necessary during , but not after , satiation , area 13 appears to be essential for value updating but not for selecting visual goals based on that information ( Figure 4 , left ) . The converse pattern of results was obtained for area 11 . THIP infusions into area 11 before satiation had no effect on behavior ( t4 = 0 . 44 , p = 0 . 686 ) , whereas THIP infusions after satiation disrupted devaluation effects ( t4 = 4 . 36 , p = 0 . 012 , d = 1 . 95 ) . Thus , area 11 appears to be essential for advantageous goal selection but not value updating ( Figure 4 , right ) . We also conducted the analysis using proportion choices of images associated with the nonsated food as the dependent measure , and found the same pattern of results ( area 13 , pre-sate vs . saline: t4 = 5 . 95 , p = 0 . 004 , d = 2 . 66 , post-sate vs . saline: t4 = − 0 . 14 , p = 0 . 898; area 11 , pre-sate vs . saline: t4 = 0 . 46 , p = 0 . 669 , post-sate vs . saline: t4 = 4 . 72 , p = 0 . 009 , d = 2 . 11 ) . We conducted three control procedures . First , we considered whether the infusions might disrupt monkeys’ choices in the absence of selective satiation . To examine this possibility , in each of the five subjects we infused THIP into either area 11 or area 13 and administered probe tests without any prior selective satiation . These sessions were inserted randomly within the infusion series carried out for the main study . As shown in Figure 5A , in the absence of satiation , infusions of THIP into either area did not change monkeys’ choices of images paired with each food type ( repeated-measures ANOVA; F ( 2 , 8 ) = 0 . 06 , p = 0 . 945 ) . These data , together with the lack of effect of THIP infusions either after selective satiation in area 13 or before selective satiation in area 11 ( Figure 4 ) , rules out the possibility that infusion of THIP , per se , disrupted monkeys’ choices . 10 . 7554/eLife . 11695 . 007Figure 5 . Control procedures . ( a ) Effect of temporary inactivation of area 13 and area 11 of the OFC on image choices in the absence of selective satiation . Proportion choice indicates the proportion of chosen images associated with the either the preferred food ( black , designated Food 1 ) or nonpreferred food ( white , designated Food 2 ) out of the total number of choices averaged across probe tests . There was no effect of THIP infusions on image choices relative to baseline . Baseline: image choices on baseline days; Area 13: image choices after THIP infusions into area 13; Area 11: image choices after THIP infusions into area 11 . ( b ) Effect of THIP infusions into area 11 on image choices for the 60 object discrimination problems learned in the Training phase ( Figure 1a ) . Proportion correct indicates the accuracy on the familiar discrimination problems . There was no effect of THIP infusions on choice accuracy for the familiar discrimination problems . Baseline , no infusions; Drug , THIP infusions into area 11 administered prior to test session . ( c ) Effect of temporary inactivation of area 13 and area 11 on food choices after selective satiation . Proportion choice nonsated food is scored from forced-choice trials involving selection between the sated and nonsated food . White bars , saline infusions; black bars , THIP infusions administered before selective satiation; gray bars , THIP infusions administered after selective satiation . All monkeys reliably chose the higher-value ( nonsated ) food after selective satiation procedures , even after inactivation of area 13 or area 11 . Thus , satiety mechanisms were intact during all experimental conditions . For all panels , bars represent group means and error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 11695 . 007 A more serious concern was that inactivation of area 11 might disrupt image choices generally , as opposed to disrupting image choices based on current value , as we assumed . Accordingly , we carried out a second control procedure in four of the original five monkeys , in which we tested the effect of THIP infusions into area 11 on performance of the 60 discrimination problems that had been learned in the training phase ( Figure 1a ) . THIP infusions into area 11 before test sessions did not affect accuracy on the familiar 60 image pairs ( t3 < 0 . 001 , p > 0 . 999; Figure 5B ) , response speed on correct trials ( median response latencies , mean ( SEM ) ; no-drug baseline = 1 . 53 s ( 0 . 17 ) ; THIP infusion before = 1 . 35 s ( 0 . 17 ) ; t3 = 1 . 45 , p = 0 . 242 ) , or time to complete test sessions ( total session duration , mean ( SD ) ; no-drug baseline = 449 . 8 s ( 46 . 69 ) ; THIP infusion before = 468 . 1 s ( 54 . 92 ) ; t3 = 1 . 63 , p = 0 . 201 ) . These data , together with the lack of effect of inactivation of area 11 in our first control procedure , described above , led us to conclude that THIP infusions into area 11 after selective satiation in our main task specifically disrupted retrieval or use of the recently updated value , as opposed to having a general effect on choosing rewarded images . Finally , we considered whether infusions affected satiety mechanisms directly , in the absence of the need to select goals in the form of visual images that had been paired with food . This third control procedure was run after the data from the main experiment had been collected . In two subjects from the main experiment and a third monkey used for this procedure only , we assessed monkeys’ food choices directly , in the absence of images , to determine if the selective satiation procedure had been effective . Monkeys received a series of 30 choice trials in which food 1 was directly pitted against food 2 . In separate sessions , saline and THIP were infused before and after selective satiation , and we recorded monkeys choices of food relative to baseline choices of food , just as we had for the probe tests with images . When monkeys could choose between visible foods directly , rather than choosing between images that had previously been paired with food , THIP infusions had no effect on choices , regardless of whether that was measured as a function of proportion shifted , as used for image choices , or as a function of the proportion choice of the nonsated food ( repeated-measures ANOVA with factors of region and treatment; for all main effects of region: F ( 1 , 2 ) < 8 . 68 , p > 0 . 098; for all main effects of treatment and interactions: F ( 2 , 4 ) < 0 . 29 , p > 0 . 764; Figure 5C ) . Thus , consistent with our earlier work based on permanent lesions ( Izquierdo et al . , 2004; Rudebeck et al . , 2013 ) , the reversible inactivations had no discernable influence on satiety mechanisms , and the selective satiation procedure produced the desired effect .
Our results reveal a double dissociation of function between the anterior and posterior components of the OFC , areas 11 and 13 , which Carmichael and Price ( 1996 ) characterize as components of an orbitomedial prefrontal ‘sensory network’ . Their mutually supportive and complementary functions agree with this concept of prefrontal organization . When an outcome’s value is altered , area 13—but not area 11—is essential for a value updating function . After updating has occurred , however , area 11—but not area 13—is essential for translating the knowledge about current valuations into advantageous choices . Thus , these two areas play complementary roles in mediating devaluation effects: the caudal part of the OFC updates valuations and the rostral part translates this knowledge into action . The present experiment provides the first demonstration of a double dissociation of function between the anterior and posterior components of the primate OFC , which complements recent studies that have pointed to functional differences between its medial and lateral parts ( Bouret and Richmond , 2010; Noonan et al . , 2010; Rudebeck and Murray , 2011 ) . The finding of an impairment when area 11 was inactivated after—but not before—the selective satiation procedure is particularly significant . The monkeys had by all indications acquired information about the new , updated value of the food outcome , yet their choices did not reflect this . This observation shows that area 11 is not necessary for updating valuations per se , but rather for the use of this knowledge to choose goals based on visual stimuli associated with predicted food outcomes . In essence , once value updating has occurred , area 11 plays the dominant role and area 13 is no longer needed to guide choices between stimuli based on current valuations . Presumably , storage of the updated value occurred either during or shortly after the selective satiation procedure , a process that depended on area 13 , and this information was then broadcast to other areas , including area 11 . Inactivation of area 11 immediately before the probe tests prevented the optimal use of this knowledge . An additional aspect of our double dissociation deserves comment . Inactivation of area 11 before the selective satiation procedure , which had no effect on behavior , would have persisted into the choice phase , when THIP infusions into area 11 did cause an impairment . But it is reasonable to expect an impairment to follow choice-phase inactivation of area 11 , regardless of how long its neurons had been inactivated . We do not believe that the effect of pre-satiation THIP infusions could have dissipated in time for area 11 to participate in goal selection . Infusion of similar volumes of THIP into the pulvinar produced behavioral effects in monkeys lasting several hours ( Wilke et al . , 2013 ) , and smaller volumes administered in rat superior colliculus affected behavior for at least two hours ( Di Scala et al . , 1983 ) , both well beyond the elapsed time between pre-satiation infusion and choice tests in the present study . Instead , our negative results after pre-satiation infusions probably reflect the cooperative function of other prefrontal areas , which would also explain the relatively mild effect of choice-phase area 11 inactivations . The impairment when area 11 was inactive only during goal selection , but not when it was inactive during both value updating and goal selection , suggests that area 11 plays a dominant role in a partially-redundant goal-selection network . Another possibility is that when area 11 was inactivated before the selective satiation procedure , it was effectively taken ‘offline’ , rendering it unable to receive updated valuation signals and therefore unable to participate in goal selection later . According to this account , when area 11 had received updated valuation signals during the satiation procedure it made a significant contribution to goal selection , but otherwise did not . Presumably , neighboring prefrontal areas could mediate goal selection in this circumstance . Notably , inactivation of area 13 yielded a pattern of results virtually identical to that observed after inactivation of basolateral amygdala in monkeys performing the same task ( Wellman et al . , 2005 ) . Given the evidence from neuropsychological studies that the amygdala and the OFC must functionally interact in mediating devaluation effects ( Baxter et al . , 2000 ) , the present results point to an interaction of the basolateral amygdala with area 13 to support value updating , specifically during the selective satiation procedure . One implication of this result is that , during normal devaluation , values get updated gradually during the selective satiation procedure . This does not imply , however , that the areas involved in goal selection receive real-time information about this gradual devaluation . Instead , from a modeling perspective , the effect of satiation might be computed ( read-in ) at the time of the choice . These possibilities are not mutually exclusive , and competing choice-related areas might operate differently in this regard . Area 13 , specifically area 13m , is the first site of convergence in the OFC of gustatory , olfactory , somatic , visceral , and visual sensory inputs . On these grounds , it has been proposed to be the site of sensory representations of food outcomes , including the sensory properties of the food as well as its value ( Carmichael and Price , 1996; Critchley and Rolls , 1996; Rolls , 2000; Saleem et al . , 2008 ) . Consistent with this idea , neurons in area 13 of macaques recorded during performance of reward-guided tasks signal the value and identity of expected outcomes ( Tremblay and Schultz , 1999; Wallis and Miller , 2003; Padoa-Schioppa and Assad , 2006; 2008; 2009 ) , as do neurons in the amygdala ( Paton et al . , 2006 ) . The pattern of results we observed is consistent with functional imaging studies showing a role for the human OFC in signaling changes in food value specific to the food consumed ( Small et al . , 2001; Gottfried et al . , 2003 ) . The present data shed new light on value-based goal selection . Apparently , either during or soon after the value updating process , the new value is transmitted from area 13 to area 11 and potentially to other prefrontal cortical regions as well . The fact that area 13 does not need to be active during the choice phase reveals that area 13 is not an essential part of the circuitry generating goals for action . Human fMRI studies suggest that the anterior OFC may represent foods at an abstract level; the posterior OFC in humans is more active in response to pictures of food relative to nonfood objects ( van der Laan et al . , 2011 ) . Based on BOLD changes consequent to sensory-specific adaptation , Klein-Flugge et al . ( 2013 ) proposed that the anterior OFC , in the region of area 11 , houses a stimulus–food outcome representation , whereas the more posterior OFC , in the region of area 13 , houses representations of reward identity , independent of associated stimuli . This pattern of results is consistent with our findings , but leaves unanswered how visual sensory input translates this knowledge into goals . Because area 11 can mediate adaptive choices independently of area 13 , we conclude that area 11 and associated circuits link visual inputs with the current , updated value of particular food outcomes . This could occur within area 11 , or it could occur in concert with anatomically related areas such as the ventral lateral prefrontal cortex ( VLPFC ) , area 12 , a region known to be important for visual attentional selection ( Rushworth et al . , 2005 ) and for switching among competing behavior-guiding rules ( Bussey et al . , 2001; Rossi et al . , 2007; Baxter et al . , 2009; Buckley et al . , 2009 ) . West et al . ( 2011 ) have performed an experiment like the present one and found that bilateral inactivation of the OFC disrupted both value updating and goal selection based on those valuations . Because they infused muscimol into a central location within the OFC , their infusions probably affected both areas 11 and 13 simultaneously , which accounts for the difference between their results and ours . Other differences , such as the species tested ( pig-tailed vs . rhesus macaques ) , the type of stimulus material ( images vs . objects ) , and the pharmacological agent used ( muscimol vs . THIP ) seem unlikely accounts , although further work is necessary to rule out these variables . Anatomical studies have suggested a hierarchical organization in the anteroposterior dimension of the OFC . Specifically , by analogy with sensory corticocortical connections , the posteriorly-directed projections within the OFC are said to follow a “feedforward” projection pattern characterized by projections that arise in supragranular layers ( layers 2–3 ) of the cortex ( Carmichael and Price , 1996 ) . For example , the preponderance of cells giving rise to the posteriorly-directed projections from area 11 to area 13 and , likewise , from area 13 to the agranular insular cortex , arise from layers 2 and 3 . This picture is consistent with a model in which visual sensory representations in area 11 subsequently activate representations in more posterior areas , where they become integrated with updated value signals . When area 13 is inactive during the probe tests , however , other regions , such as area 12 of the VLPFC or the anterior insula , are sufficient to provide the information about updated value . To influence behavior , the OFC must eventually affect the motor system . There are at least two models of how value-guided choice might be implemented . One idea posits a serial model ( Padoa-Schioppa , 2011 ) in which offer values are compared , perhaps in a common currency . The offer with the greater value is selected , and then an action is selected to implement that choice . Another model suggests that value signals could influence processing in visuomotor pathways that plan movements through a mechanism akin to top–down attention and biased competition ( Pastor-Bernier and Cisek , 2011 ) . By either model , these influences of current value on choice might be mediated by connections between the OFC and the ventral , medial , and dorsolateral prefrontal cortex ( Barbas and Pandya , 1989; Carmichael and Price , 1996 ) . The OFC and the VLPFC contribute to multisynaptic pathways to dorsal premotor areas , mainly via dorsal and dorsolateral prefrontal cortex ( Takahara et al . , 2012 ) , and cingulate premotor areas , mainly via medial prefrontal cortex ( Morecraft and Van Hoesen , 1998; Morecraft et al . , 2012 ) . The VLPFC in particular receives strong projections from areas 13 and 11 as well as from the basolateral amygdala ( Porrino et al . , 1981; Saleem et al . , 2014 ) . Thus , the OFC and the adjacent VLPFC are well situated to influence the selection of targets for action . Taken together , the results from inactivating areas 11 and 13 selectively demonstrate specialized functions for these two components of the macaque OFC . The posterior component , area 13 , functions in conjunction with the basolateral amygdala to update the valuation of expected reward outcomes , based on an animal’s current satiation state . The anterior component , area 11 , plays a critical role in translating this knowledge into goals that produce an advantageous outcome . Importantly , area 11 is needed for choosing which of two visual stimuli to choose when both stimuli are abstractly associated with a particular reward outcome , but not for choices between two visible foods . The inability to translate abstract valuation knowledge into advantageous choices resembles the goal neglect that occurs after damage to the frontal lobe in humans .
Six male rhesus monkeys ( Macaca mulatta ) , experimentally naïve at the beginning of training , served as subjects . Five of the monkeys participated in the main experiment , and one additional monkey served as a subject in a subset of the control procedures only . All monkeys were housed in rooms kept on a 12-hr light/dark cycle ( lights on at 7:00AM ) and testing occurred during the light period . Four of the six monkeys were housed individually while the remaining two were pair housed with monkeys participating in a different experiment . At the beginning of the study , the monkeys ranged in weight from 6 . 4 to 9 . 5 kg . For the duration of the study , the monkeys were given controlled access to food to ensure sufficient motivation to respond in the test apparatus . Water was available ad libitum in the home cage . All procedures were reviewed and approved by the NIMH Animal Care and Use Committee . Testing was carried out in an automated apparatus consisting of a microprocessor linked to a 15-inch color monitor fitted with a touch-sensitive screen . Rewards consisted of two of the following three foods: M&Ms ( Mars Candies , Hackettstown , NJ ) , peanuts , and Skittles ( Mars Candies , Hackettstown , NJ ) . Rewards were dispensed from automated dispensers ( Med Associates , St Albans , VA ) mounted on top of the test chamber . All tasks were controlled and behavioral data collected by a computer using custom software ( Ryklin Software Inc . , New York , NY; software is available for download: ftp://helix . nih . gov/lsn/beethoven/Operant_Testing_System . exe ) . During each test session , the monkey was seated in a primate chair inside a light- and sound-attenuating test chamber . A fan mounted in the ceiling of the chamber provided ventilation and masked extraneous noise . Visual stimuli consisted of rectangular clipart images , approximately 55 mm x 55 mm . The monkey’s head was approximately 230 mm from the monitor screen . Monkeys were first trained to touch a monitor screen using standard shaping procedures ( Murray et al . , 1993 ) . They then learned a standard single-pair , visual discrimination . One image of the pair was arbitrarily designated as the rewarded image ( S+ ) , and the other was designated as the nonrewarded image ( S− ) . The monkey was required to touch one of the two images on the screen . The selected image was then shown with a red frame and flashing red dot in the center . The nonselected image disappeared . If the selected image was the S+ , touching it again led to delivery of a ½ peanut reward and visual feedback of the image on the screen for an additional 1 s . If the selected image was the S− , then touching it again resulted in the screen going blank . The intertrial interval ( ITI ) was set at 5 s after a correct ( rewarded ) response , and 10 s after an incorrect ( nonrewarded ) response . Criterion was set at a mean of 90 percent correct responses over three consecutive 60-trial sessions . Finally , monkeys were required to discriminate 20 image pairs , repeated 3 times each per session . Stimuli were novel at the beginning of training , and general methods were the same as for the single-pair discrimination . Criterion was set at a mean of 90 percent correct responses over five consecutive sessions . After completing pretraining , each monkey began training on the devaluation task . Training began with the acquisition of a list of 60 fixed image pairs presented for visual discrimination ( Figure 1a ) . All images were novel at the beginning of the experiment . One image of each pair was arbitrarily designated the S+ and the other the S− . On each trial , the two images of a pair were presented simultaneously , 150 mm apart , to the left and right of the screen center . If the monkey touched the S+ , the S− image disappeared , and the selected image remained on the screen for one additional second while a food reward was delivered . If the monkey touched the S− , the screen went blank . There was no correction for errors . Each image pair was displayed once per session , in random order , yielding a total of 60 trials . The location of the S+ ( left or right ) also followed a pseudorandom Gellermann order . Each monkey had a fixed image pair list that was unique to the animal . For 30 of the image pairs , food 1 ( e . g . , ½ peanut ) was delivered when the S+ was selected . For the remaining 30 pairs , food 2 ( e . g . , an M&M ) was delivered when the S+ was selected . The S+ image-food assignments were fixed throughout training . The ITI was 5 s after correct responses and 10 s after incorrect responses . Criterion was set at a mean of 90 percent correct responses over five consecutive sessions ( i . e . , 270 or more correct responses in 300 trials ) . After monkeys had reached criterion on the 60 pairs , but before data collection began in the main task , monkeys were transitioned to a four day per week test schedule ( Figure 3 ) adapted from Wellman et al . ( 2005 ) . Day 1 was a review of the 60 discrimination problems , presented just as in initial learning . Day 2 comprised a probe test to assess monkey’s baseline choices . S+ images assigned to different foods ( food 1 and food 2 ) were pitted against each other , yielding 30 trials per session . Whichever image was chosen , the appropriate food was delivered . The images were paired anew for each probe test , with the constraint that pairs of images presented for choice always comprised one image associated with each of the two foods . Day 3 repeated the review of the 60 pairs , as on Day 1 . The session on Day 4 followed the same procedure as on Day 2 , but was carried out after selective satiation . In addition , this was the day on which we carried out infusions of THIP or saline , either before or after the selective satiation procedure . The measure of interest was the number of food-1 and food-2 associated images selected on Day 4 ( after selection satiation ) relative to Day 2 ( baseline ) , which reflects each monkey’s ability to adaptively shift away from choosing images paired with the devalued food . We calculated proportion shifted according to Eq . 1 , where F1 and F2 represent choices of the images paired with the two food types on weeks in which that food type was devalued , and subscripts D and N respectively represent the Devaluation day ( Day 4 ) and the day of that same week on which that food was Not devaluated ( Day 2 ) . Thus , proportion shifted is the total number of image choices shifted due to devaluation , as a fraction of the total possible shift: Proportion shifted=F1N-F1D+F2N-F2DF1N+F2N The selective satiation procedure has been described in detail elsewhere ( Izquierdo et al . , 2004 ) . Unlike earlier studies , where food was given in the home cage , we delivered the to-be-sated food while the monkeys remained seated in a primate chair . This was done to maintain strict control of the elapsed time between satiety and infusions , and between the infusions and initiation of test sessions . To determine whether satiety mechanisms were intact , monkeys were given additional probe tests . In this series of tests , instead of being evaluated for image choices , monkeys were evaluated for food choices ( Figure 3 , Control manipulations ) . Each monkey sat in the testing room with the experimenter . After undergoing the selective satiation procedure , the monkey was given a series of forced-choice trials between food 1 and food 2 , presented on a two-well test tray . The placement of the foods ( left and right ) was pseudorandomized using a Gellermann schedule . To complete a trial , the monkey selected one of the two foods . The experimenter recorded the choice . Each session comprised 30 trials separated by 10 s . All six monkeys underwent three stages of surgery under general anesthesia . In the first operation , monkeys were fitted with a titanium head post held in place with self-tapping titanium screws . After a minimum of four weeks , each animal received a second operation to implant an infusion chamber ( Section on Instrumentation Core , NIH , Bethesda , MD ) . The chamber—which had interior dimensions of 26 . 5 mm × 46 mm and was fabricated from Ultem plastic—was fixed to the cranium using dental acrylic; ceramic screws positioned around the outside of the chamber served to anchor the implant . In the third and final stage of surgery , two craniotomies were made , one per hemisphere , within the bounds of the chamber . In a few instances , the bone regrew over the target areas and an additional operation was required to re-open the craniotomy . During surgery , aseptic procedures were used . Anesthesia was induced with ketamine hydrochloride ( 10 mg/kg , i . m . ) and maintained with isoflurane ( 1 . 0–3 . 0% , to effect ) . Heart rate , respiration rate , blood pressure , expired CO2 , and body temperature were monitored during surgery , and isotonic fluids were given throughout . Cefazolin antibiotic ( 15 mg/kg , i . m . ) was given to prevent infection . The drug was administered for one day before surgery and for one week after surgery . Monkeys received the analgesic ketoprofen ( 10–15 mg , i . m . ) at the end of surgery and for two additional days postoperatively . This was followed by 100 mg of ibuprofen for the next 5 days . Each monkey received multiple structural MRI scans , as needed , to guide chamber placement , cannula placement for infusions , and to confirm that infusions reached the intended targets . In general , the first scan , which was used to guide chamber placement , was carried out in a horizontal-bore 1 . 5 or 3 T scanner . The monkey was sedated with a mixture of ketamine ( 15 mg/kg , i . m . ) and medetomidine ( 20 μg/kg , i . m . ) , supplemented as needed . Glycopyrrolate was administered to reduce secretions ( 0 . 01 mg/kg , i . m . ) and ketoprofen ( 10–15 mg , i . m . ) was given as an analgesic . Monkeys were placed in a MR-compatible stereotaxic frame for the duration of the scan . To guide cannula placement and to calculate the depth of infusion cannulae , the monkeys received an additional structural scan using a 4 . 7T vertical-bore scanner . For this scan , the chamber was filled with a gadolinium–saline solution ( 1:1200 v:v; gadolinium , Bayer HealthCare , Wayne , NJ ) and the chamber grid was placed in the chamber to allow visualization of the vertical channels through which infusion cannulae would be inserted ( see Infusions ) . Finally , to verify that our infusions reached the intended target , each monkey received a gadolinium–saline infusion ( 1:100 v:v , 2 . 0 μl/site ) targeted just dorsal to the tissue of interest , and then received a scan in the vertical-bore magnet as described above . This final type of scan was conducted once for area 11 and again for area 13 before data collection for that area commenced . Sedation was achieved using a mixture of ketamine ( 5–20 mg/kg , i . m . ) and diazepam ( 0 . 5–1 . 5 mg/kg , i . m . ) , supplemented as needed . Glycopyrrolate was administered to reduce secretions ( 0 . 015 mg/kg , i . m . ) . We carried out infusions in area 13 and area 11 , one area followed by the other . The order in which areas were studied was balanced across subjects . The experimental manipulations were carried out on Day 4 ( Figure 3 ) . Infused compounds were either saline ( vehicle; phosphate buffered saline ) or the drug THIP ( 18 mM; Tocris , Bristol , UK ) which , like muscimol , is a GABAA agonist . In each case , the drug was freshly dissolved in vehicle , and the solution , pH 7 . 0–7 . 5 , was sterile filtered ( Corning , Corning , NY ) before injection . Prior to each infusion , the chamber cap was removed and a sterile grid was inserted into the chamber . The grid fit snugly inside the chamber and contained vertical channels spaced 1 mm apart in the form of two 19 x 19 arrays , one each over the left and right hemisphere ( Talbot et al . , 2011 ) . The two-dimensional coordinate frame of the grid permitted reliable placement of the infusion cannulae . By using the same x , y , and z coordinates from one infusion to the next , the target location could be reproduced . When the grid was in place , sterile guide tubes ( thin walled 24 gauge , Component Supply Company , Fort Meade , FL ) and infusion cannulae ( regular walled 30 gauge , Component Supply Company , Fort Meade , FL ) were inserted into the brain . The infusions were performed while the monkeys were awake and seated in the primate chair , with their heads restrained with the implanted head posts . Infusions were conducted using a Harvard Apparatus infusion pump ( PHD 2000; Harvard Apparatus , Holliston , MA ) and 100 µl Hamilton syringes ( model 1710TLL; Hamilton Syringe Company , Reno , NV ) . Each infusion was administered with a flow rate between 0 . 18–0 . 25 µl/min over the course of 10–12 min , for a total volume between 1 . 98–2 . 34 µl . The vast majority of infusions used a total volume of 2 . 0 µl with a flow rate of 0 . 18 µl/min . Thus , drug or saline was infused over a period of ~11 min . The infusion period was immediately followed by a 10–15 min wait with the pump turned off but the infusion cannulae and guide tubes in place to promote diffusion at the site of infusion and to limit drug traveling up the track upon withdrawal of the needle . When the wait-period was completed , the guide tubes and infusion cannulae were withdrawn , the grid was removed , the chamber was rinsed with sterile saline , and the chamber cap was replaced . Approximately 20 min elapsed between the end of the infusion and the beginning of the test session or satiation procedure . In each region of the OFC investigated , two THIP infusions and two saline infusions were carried out before the selective satiation procedure , one for each food . Similarly , two THIP infusions and two saline infusions were carried out after the selective satiation procedure , again , one for each food . In addition to the saline ( vehicle ) infusions conducted as part of the main experiment , we carried out other infusions as needed to address interpretational issues . Intermixed with the experimental conditions , an additional THIP infusion was given without satiation . For this session , the monkey waited in the chair for the time it would take to complete the selective satiation procedure . This was done to test the possibility that either the delay period or the presence of the drug in the brain tissue alone was enough to significantly alter performance ( Figure 3 , Control manipulations , no satiation ) . In area 11 only , an additional infusion with THIP was given in four of the five monkeys . This infusion was performed prior to testing on Day 3 of the test schedule , on which the animals were presented with the list of 60 pairs , to determine if it disrupted the animals’ performance . Two of the monkeys from the main experiment and a third monkey , used for this procedure only , underwent additional infusions to test for effects of the infusions on satiety mechanisms . In each monkey , we carried out four drug and two saline infusions before and after selective satiation involving either food 1 or food 2 . For each monkey , the structural MR scan images were matched to drawings of coronal sections of a standard rhesus monkey brain at 1 mm intervals through the entire frontal lobe . Using the structural scans acquired with gadolinium-saline solutions in the chamber grid , an imaginary line was extended from the grid hole used to target each area ( area 13 or area 11 ) to the orbital surface . This site on the surface was then plotted onto the standard section and then transferred to the ventral view . At the conclusion of the experiment , the retrograde tracers cholera toxin subunit B ( CTB; List Biological , Campbell , CA; 1–2% ) and Fast Blue ( Sigma-Aldrich , St . Louis , MO; 3% ) were injected into area 11 and into area 13 , respectively , in one monkey . The reverse was done in a second monkey; CTB was injected into area 13 and Fast Blue into area 11 . Both sets of tracer injections were performed unilaterally . Hamilton syringes were inserted into the chamber grid at the same x , y location and to the same depth as for the infusions . After a survival period of 13 days , the monkeys were deeply anesthetized with Euthanasia ( 0 . 1 ml/kg , i . v . ) and perfused transcardially with normal saline , followed by 4% paraformaldehyde in 0 . 1M phosphate buffer . The brain was extracted from the skull , blocked , and cryoprotected through a series of glycerols . After 3–4 days , the brain was frozen in dry ice and isopentane , and cut in the coronal plane at 40 µm-thick sections on a sliding microtome . One to two parallel series of sections were immediately mounted on gelatin-coated slides , air-dried , and coverslipped with DPX ( Sigma-Aldrich ) for the examination of fluorescent tracer Fast Blue . The other series was processed immunohistochemically with the Avidin/Biotin immunoperoxidase method for CTB labeling . To visualize CTB , sections were first rinsed in 0 . 1M phosphate buffered saline ( PBS , pH 7 . 4 ) , and then incubated for 2 hr in blocking serum consisting of 0 . 3% Triton X-100 ( TX-100 ) , 2% bovine serum albumin ( BSA ) , 0 . 3% hydrogen peroxidase ( H2O2 ) , and 15% normal rabbit serum ( NRS ) in PBS . Tissue was then incubated in the primary antibody solution ( 1:5000 anti-CTB , added to the serum solution consisting of 2% BSA , 5% NRS , and 0 . 2% TX-100 in PBS ) for 65 hr at 4°C . After several washes in PBS , sections were then incubated in the secondary antibody solution ( 1:200 anti-goat IgG , added to the same serum solution as described in the previous step ) for 1 . 5 hr at room temperature , followed by another wash in PBS . The sections were then processed with the avidin/biotin staining kit ( Vector ABC Elite ) for 90 min at room temperature , after which sections were washed in PBS and placed in a 0 . 035% DAB solution ( 3 , 3-diaminobenzidine tetrahydrochloride as chromogen; Sigma #D5637 ) . After 10 min , approximately 0 . 0125% hydrogen peroxide was added to initiate the staining reaction . The DAB reaction was stopped when satisfactory contrast was achieved ( usually 2–5 min ) . After a final rinse in phosphate buffer , sections were mounted on gelatin-coated slides , air-dried , and dehydrated through ascending grades of ethanol concentrations before being cleared in xylenes and coverslipped in DPX ( Sigma-Aldrich , St . Louis , MO ) . | Everyone knows that somehow , somewhere , the brain translates knowledge into action . In some people , however , knowledge and action become disconnected . These people behave in a way that either ignores or contradicts the knowledge that they have . They know what to do and can explain it to others , but – when the time comes to act – they do something else , something wrong . Murray et al . have now investigated how a brain region called the orbitofrontal cortex helps to link knowledge and action in macaque monkeys , which , unlike rodents , have all of the main brain areas that make up the orbitofrontal cortex of humans . The monkeys learned to associate images with different types of food , and then performed a task where they chose between two images in order to get the food they wanted . On some days , one of the foods was less ‘valuable’ because the monkeys had already eaten a lot of it . In these circumstances , monkeys chose fewer of the images associated with that food . By temporarily inactivating either the front or back region of the monkey’s orbitofrontal cortex at different times , Murray et al . showed that these regions make different contributions to decision making . Inactivating the back region of the orbitofrontal cortex disrupted the ability of monkeys to update their knowledge about the value of a particular foodstuff . However , inactivating the front part of the orbitofrontal cortex disrupted their ability to use this knowledge to select the images that led to the most valuable food . This contradicts the widely held belief that the orbitofrontal cortex acts as a single entity to update values and translate this knowledge into action . Future work will need to investigate how , having translated knowledge into a chosen action , the orbitofrontal cortex stimulates the motor areas of the brain to generate the movements needed to perform that action . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Specialized areas for value updating and goal selection in the primate orbitofrontal cortex |
Neoangiogenesis is a pivotal therapeutic target in glioblastoma . Tumor monitoring requires imaging methods to assess treatment effects and disease progression . Until now mapping of the tumor vasculature has been difficult . We have developed a combined magnetic resonance and optical toolkit to study neoangiogenesis in glioma models . We use in vivo magnetic resonance imaging ( MRI ) and correlative ultramicroscopy ( UM ) of ex vivo cleared whole brains to track neovascularization . T2* imaging allows the identification of single vessels in glioma development and the quantification of neovessels over time . Pharmacological VEGF inhibition leads to partial vascular normalization with decreased vessel caliber , density , and permeability . To further resolve the tumor microvasculature , we performed correlated UM of fluorescently labeled microvessels in cleared brains . UM resolved typical features of neoangiogenesis and tumor cell invasion with a spatial resolution of ~5 µm . MR-UM can be used as a platform for three-dimensional mapping and high-resolution quantification of tumor angiogenesis .
Gliomas are highly malignant brain tumors with poor prognosis ( Wen and Kesari , 2008 ) . Glioblastoma multiforme ( GBM ) is characterized by high cellular proliferation rates and a rapid induction of angiogenesis ( Wen and Kesari , 2008 ) . Angiogenesis is a hallmark of malignant tumors , and most tumors show an exponential ingrowth of neovessels upon a certain tumor size to accommodate their metabolic needs ( Carmeliet and Jain , 2011; Hanahan and Weinberg , 2011 ) . This phenomenon is called 'angiogenic switch' and is characterized by the upregulation of pro-angiogenic molecules like vascular endothelial growth factor ( VEGF ) or angiopoietin 2 ( Ang-2 ) that induce fast tumor angiogenesis . The angiogenic switch is a central feature of malignant tumors and leads to a transition toward a more invasive and aggressive tumor growth ( Bruno et al . , 2014 ) . Hence , many efforts have been made to develop antiangiogenic therapies to 'starve' tumors off their resources ( Huang et al . , 2003 ) . So far , this approach of using anti-VEGF agents like bevacizumab has shown clinical benefit in certain tumor types ( Mittal et al . , 2014 ) but , despite significant improvement of progression-free survival , does not prolong overall survival of primary GBM in an unselected patient cohort ( Chinot et al . , 2014 ) . Similarly , VEGF receptor 2 ( VEGF-R2 ) inhibition , despite strong preclinical data ( Batchelor et al . , 2007; Winkler et al . , 2004 ) , has not been successful in neuro-oncology yet ( Batchelor et al . , 2013 ) . New methods are currently developed to better identify those patients who might benefit from antiangiogenic therapies: Next to molecular approaches ( Sandmann et al . , 2015 ) , magnetic resonance ( MR ) imaging could be used for patient stratification . In fact , accumulating data suggest that vascular normalization that occurs early after the initiation of therapy , might be such a marker which could aid clinical decision making ( Emblem et al . , 2013; Lu-Emerson et al . , 2015 ) . However , improved imaging techniques that faithfully and non-invasively characterize vessel architecture and antiangiogenic treatment effects are needed to facilitate the understanding of biological actions of these therapies , and the development of clinical trials . We employed a MR imaging approach to monitor tumor vessels at single vessel resolution . The technique is based on T2*-weighted ( T2*-w ) high-resolution Blood Oxygenation Level Dependent ( BOLD ) venography ( Park et al . , 2008 ) with standard gadolinium ( Gd ) contrast agents at ultrahigh-field strength ( 9 . 4 Tesla , isotropic 80 µm resolution ) . This visualizes substantially more vascular detail compared to conventional T2*-w imaging . We performed pre- and post-contrast MR scans to define angiogenesis during glioma development in two different glioma models . Also , we assess treatment effects of anti-VEGF or radiation therapy on the vascular compartment . We further mapped the tumor vascularization by correlated , dual-color ultramicroscopy ( UM ) of cleared , unsectioned brains ( Ertürk et al . , 2012; Schwarz et al . , 2015 ) . Fluorescent labeling of the microvasculature using lectins resulted in complementary 3D MR and UM data sets ( dubbed 'MR-UM' ) of the entire tumor 'macro-' and 'microvasculature' , which can be compared side-by-side . Thus , MR-UM bridges the gap between MR and optical imaging and may serve as a platform to better understand underlying mechanisms of antiangiogenic treatment and to identify novel druggable targets for preclinical therapy development .
The mouse glioma cell line GL261 was purchased from National Cancer Institute ( NCI Tumor Repository , Frederick , MD ) and cultured in Dulbecco’s modified Eagle’s medium ( DMEM ) containing 10% FBS , 100 U/ml penicillin , and 100 µg/ml streptomycin ( all from Sigma-Aldrich Chemie GmbH , Taufkirchen , Germany ) . The 1x105 GL261 cells diluted in 2-µl sterile phosphate buffered saline ( PBS , Sigma-Aldrich Chemie GmbH , Taufkirchen , Germany ) were stereotactically implanted into the right brain hemisphere of 6- to 8-week-old female C57Bl/6J mice ( n=45 mice , Charles River Laboratories , Sulzfeld , Germany ) using a Hamilton syringe , driven by a fine step motor ( coordinates: 2 mm lateral and 2 mm ventral of bregma; injection depth: 3 mm below the dural surface ) . Animals were deeply anesthetized with ketamine/xylazine and unresponsive to stimuli during the intracranial injection; GL261 cells were routinely tested for viral or mycoplasma contamination , and mouse cell identity was confirmed by the multiplex cell contamination test ( Multiplexion GmbH , Heidelberg , Germany , Schmitt and Pawlita , 2009 ) . The human S24 cell line was derived as a primary glioblastoma culture from a resected glioblastoma ( after informed consent ) and GBM typical genetic changes were confirmed by comparative genomic hybridization ( Lemke et al . , 2012; Osswald et al . , 2015 ) . For the S24 glioma model , 5x104 S24:td-tomato cells ( stably transduced by lentivirus ) were transplanted orthotopically in 8- to 10-week-old male NMRI nude mice ( Charles River , Sulzfeld , Germany , n=3 mice ) . The cells were cultivated under serum-free conditions in DMEM-F12 as sphere cultures ( Thermo Fisher Scientific Inc . , Waltham , MA ) supplemented with 2% B-27 ( Thermo Fisher Scientific Inc . ) , 5 µg/ml human insulin ( Sigma-Aldrich Corporation , St . Louis , MO ) , 12 . 8 ng/ml heparin ( Sigma-Aldrich Corporation ) , 0 . 4 ng/ml EGF ( R&D Systems Inc . , Minneapolis , MN ) and 0 . 4 ng/ml FGF ( Thermo Fisher Scientific Inc . ) . Cell lines were regularly checked for mycoplasma infections and authenticity ( species control ) . Mycoplasma testing was done using the LookOut Mycoplasma PCR Detection Kit ( Sigma-Aldrich , Germany ) according to the manufacturer’s instructions . All animal experiments were approved by the regional animal welfare committee ( permit number: G187/10 , G188/12 and G145/10 , Regierungspräsidium Karlsruhe ) . MR imaging was performed on a 9 . 4 Tesla horizontal bore small animal NMR scanner ( BioSpec 94/20 USR , Bruker BioSpin GmbH , Ettlingen , Germany ) with a four-channel phased-array surface receiver coil . MR imaging included a standard RARE T2-w and T1-w post-Gd-contrast sequence to monitor tumor volume ( T2-w parameters: 2D sequence , 78 µm in plane resolution , TE: 33 ms , TR: 2500 ms , flip angle: 90° , acquisition matrix: 200 x 150 , number of averages: 2 , slice thickness: 700 µm duration: 2 min 53 s; T1-w parameters: 2D sequence , 100 µm in plane resolution , TE: 6 ms , 1000 TR: ms , flip angle: 90° , acquisition matrix: 256 x 256 , number of averages: 2 , slice thickness: 500 µm , duration: 5 min ) . To assess the tumor vasculature , we used a T2*-weighted gradient echo sequence ( Park et al . , 2008 ) and acquired pre- and post-contrast scans ( 3D sequence , 80 µm isotropic resolution , TE: 18 ms; TR: 50 ms; flip angle: 12°; number of averages: 1 , acquisition matrix: 400 x 188 x 100 , duration: 15 min 40 s ) . Pre-contrast images were used to differentiate susceptibility artifacts caused , for example , by tumor microbleedings from vessel signals that were only detectable after contrast administration . Dynamic contrast-enhanced ( DCE ) imaging ( TE: 1 . 8 ms; TR: 16 ms; flip angle: 10°; slice thickness: 700 µm , acquisition matrix: 66 x 128 , 3 slices acquired , number of averages: 1 , 300 repetitions; 700 µm in plane resolution; duration: 10 min , time resolution 2 s ) was used to assess vascular permeability ( Ktrans ) . 0 . 2 mmol/kg Gadodiamide ( Omniscan , Nycomed , Ismaningen , Germany ) was administered by tail vein injection for DCE and post-contrast scans . In five animals , 50 µl Gadodiamide was administered by intraperitoneal injection ( ip ) , which had been determined before to match an iv dose of 0 . 2 mmol/kg . To assess the validity of the T2*-w sequence with an iron-based contrast agent , we performed blood pool imaging pre- and post-iv injections of crossed linked iron oxide nanoparticles ( USPIO , CLIO-FITC , 15 mg/kg , particle size: 31 nm , kind gift by R . Weissleder , MGH , Boston ) . MR imaging was started 7 to 14 days post GL261 tumor cell implantation and repeated weekly up to 5 weeks after tumor cell implantation . For MR imaging , animals were anesthetized with 3% isoflurane . Anesthesia was maintained with 1–2% isoflurane . Animals were kept on a heating pad to keep the body temperature constant . Animal respiration was monitored externally during imaging with a breathing surface pad controlled by an in-house developed LabView program ( National Instruments Corporation ) . DCE imaging was post-processed with OLEA software ( OLEA Medical , La Ciotat , France ) . Pseudocolor images represent vascular permeability maps ( Ktrans ) that were calculated using the Tofts and Kermode model ( Tofts and Kermode , 1991 ) . Quantification of DCE time series was performed in FIJI by ROI analysis of the main tumor slide and encompassing the entire tumor area . A mirror ROI was set in the contralateral hemisphere . Intensity values were measured and processed in Microsoft Excel ( Microsoft ) . Signal ratios ( SR ) were calculated over time ( SR=signaltumor / signalcontralateral site ) and displayed as blood–brain barrier disruption ( BBB-D , arbitrary units , a . u . ) . GL261 glioma-bearing mice were treated with the murine-chimeric anti-VEGF antibody which is based on the humanized anti-VEGF antibody B20-4 . 1 ( Srivastava et al . , 2014 ) or murine IgG control antibody ( Roche , pRED Innovation Center Penzberg ) at a dose of 10 mg/kg every 3 days by ip injection ( four mice per group ) . Treatment was initiated 2 weeks after tumor cell implantation when a solid tumor was present on MRI . Animals were treated for 1 week ( days 0 , 3 , 6 ) and MRI was performed on day 7 ( 3 weeks post inoculation ) to quantify vascularization parameters . After the MR investigation , animals were injected with lectin-FITC or lectin-texas red and sacrificed for UM correlation . Photon irradiation was delivered by XRAD320 X-ray device ( Precision X-Ray , CT , USA ) . Animals were anesthetized by ketamine/xylazine during the procedure . Whole-brain irradiation was performed with a lateral photon beam at a dose of 2 grays per day . Irradiations were performed for four consecutive days starting 2 weeks after GL261 tumor implantation . The body was protected by lead shield to avoid radiation exposure to organs other than the brain . MR measurements were performed before ( week 2 ) and after the completion of irradiation ( week 3 ) . Mice without irradiation served as controls ( n=3 mice per group ) . To quantify the apparent volume loss of brain tissue induced by the clearing procedure , we performed computed tomography ( CT ) scans before and after clearing . Two 360° scans were obtained on a small animal CT ( Quantum FX , Perkin Elmer , Waltham , MA ) using the following parameter: 90 kV , 200 µA , 24 mm field of view , 40 µm isotropic resolution . Image reconstruction was performed using a standard filtered backprojection algorithm implemented in the vendor’s software . The maximum brain dimensions were determined in transversal slices and compared before and after clearing . We found that the brain volumes shrinks by ~40% but that the ratio of maximum length/width stays constant before and after clearing ( length and width pre clearing: 1 . 33 ± 0 . 54 cm x 1 . 05 ± 0 . 38 cm; post clearing: 0 . 79 ± 0 . 24 cm x 0 . 62 ± 0 . 17 cm; ratio: 1 . 25 ± 0 . 01 and 1 . 28 ± 0 . 03 , p>0 . 05 for ratio comparison ) . The vascularized area was quantified in Osirix software ( V . 4 . 12 , Pixmeo , Geneva ) by manually selecting region of interests ( ROIs ) around tubular vessel-like structures on post-contrast T2*-w images . Care was taken to exclude areas that were hypointense on pre-contrast images and most likely represent microbleedings or calcifications . For histogram analysis , the tumor region and an outside region in the frontal white matter were manually segmented . Intensity values for histogram analysis were read out in Matlab ( Release 2015a , The MathWorks , Inc . , Natick , MA ) . Images were normalized using the following formula: ( mean voxel intensitytumor – intensityoutside ) /standard deviation ( SD ) . For correlative optical microscopy of the microvasculature , animals were injected with fluorescent lectins that bind to N-acetyl-β-D-glucosamine oligomers of endothelial cells ( Wälchli et al . , 2015 ) . Injection of isolectin-FITC ( 12 mg/kg , Sigma ) or lectin-texas red from Lycopersicon esculentum ( 12 mg/kg; Vector laboratories ) was performed via the tail vein in 100 µl PBS after MRI . Animals were sacrificed 5 min after lectin injection by a ketamine/xylazine overdose and transcardially perfused with 5 ml PBS followed by 10 ml 4% paraformaldehyde ( Histofix , Carl Roth GmbH , Karlsruhe ) . For the assessment of BBB-D , animals ( n=2 ) were iv injected with 150 µl of 2% Evans blue ( EB , Sigma ) diluted in PBS . Animals were sacrificed 10 min after EB injection and cleared using the FluoClearBABB protocol ( see below ) . Caution must be taken when using EB in UM as prolonged animal perfusion and clearing can result in a drop of fluorescent signal due to a possible washout of the hydrophilic compound . Brains were fixed after perfusion with 4 % buffered formalin for at least 24 hr at 4°C in the dark . For UM-analysis , whole brains were optically cleared using organic solvents . Clearing was performed according to the 3DISCO protocol ( Ertürk et al . , 2011; 2012 ) . Brains were transferred into glass vials for tetrahydrofuran ( THF; Sigma Aldrich ) dehydration . Two milliliter of 50 % THF were gently added using a pipette . Vials were placed into black 50-ml Falcon tubes and then mounted onto an overhead turning wheel ( program C3 , 15 rpm , Neolab , intelli-mixer ) . Clearing was performed at room temperature . After 12 hr , the 50 % THF solution was exchanged by a 70 % THF solution . Vials were again put into black Falcon tubes and mounted onto the turning wheel for another 12 hr . The procedure was repeated with 80% and 100% THF solutions , respectively . Samples were incubated for 12 hr in 100 % THF for three times . The dehydrated brains were placed in benzyl ether for 48 hr in order to clear the samples ( 98 % dibenzyl ether , DBE , Sigma-Aldrich , Steinheim , Germany ) . To avoid degradation of the fluorescent signal , samples were kept in the dark and imaged immediately after the clearing procedure . S24 tumor-bearing brains were cleared with the recently published protocol FluoClearBABB ( Schwarz et al . , 2015 ) . This protocol is based on benzyl alcohol/benzyl benzoate clearing in combination with a basic pH , which is maintained throughout the clearing procedure . The protocol is especially suited for effective clearing of aged mouse brains . Mice were perfused with lectin-FITC as described . After dissection , brains were kept in PBS at 4°C . For the dehydration of brains , analytical grade alcohol ( t-butanol , Sigma ) was diluted with double-distilled water . Brains were dehydrated using t-butanols ranging from 30 to 100 % . The clearing solution BABB was prepared by mixing benzyl alcohol ( Merck , analytical grade ) and benzyl benzoate ( Sigma , 'purissimum p . A . ' grade ) in a 1: 2 volume ratio . The pH levels of dehydration and clearing solutions were adjusted using an InLab Science electrode suited for organic solvents ( Mettler-Toledo ) . pH levels were adjusted with triethylamine ( Sigma-Aldrich ) . The cleared brains were scanned with a light sheet microscope ( LaVision BioTec GmbH , Bielefeld , Germany ) . We used 0 . 63x , 1 . 0x , and 2 . 0x with a 2x objective lens and a white light laser ( SuperK EXTREME 80 mHz VIS; NKT Photonics , Cologne , Germany ) with a wavelength spectrum ranging from 400 to 2400 nm ( pixel size for 0 . 63x: 5 . 16 µm; for 1 . 0x: 3 . 25 µm and for 2 . 0x: 1 . 62 µm ) . For the detection of blood vessels , the following filters were used: lectin-FITC , excitation 470 / 40 nm; emission 525 / 50 nm; lectin-texas red excitation 545 / 25; emission 585 / 40 . Z-stacks with 5 µm step size and a total range of up to 1500 to 2000 µm for the transversal measurement of the whole brain were acquired . Measurements with exposure times of 300 ms per slice resulted in a total acquisition time of ~10 min per brain sample and magnification . Images were exported as tagged image file ( tif ) and further post-processed in the ImageJ package FIJI , version 1 . 49 ( http://fiji . sc/Fiji ) . For the generation of UM movies , the ‘Running Z projector’ plugin ( FIJI ) was used . Representative single slices from light sheet microscopy data sets were scaled in FIJI to a resolution of 0 . 5 x 0 . 5 μm2 . ROIs were manually chosen inside the tumor region and in a region not affected by tumor or vessel alterations ( 'outside' ) . Vessels were identified with the FIJI plugin 'tubeness' that finds linear structures in an image ( Sato et al . , 1998 ) . The resulting images were binarized individually to ensure maximal congruence with the respective vessels on the original image slice . For each binarized ROI , vessel objects were identified and skeletonized using Matlabs built-in functions 'bwmorph' and 'regionprops' . After subtracting vessel branch points of the skeletonized image , the length of each vessel segment was determined . To obtain the vessel diameter and to avoid very small vessel cross-sections with minimal length , only those segments were considered whose lengths were longer than the difference of mean and standard error of the segment length distribution . For each pixel iS of each segment S , the minimum distance to a non-vessel pixel was determined as the pixel-specific segment radius ris . Subsequently , we determined average segment radius rS¯=∑iS∈SriS∑ ( iS∈S ) and overall mean segment radius r~=∑SrS¯/∑S . The tortuosity of the longest 10% of the segments was obtained by dividing segment length by the Euclidean distance between its endpoints . The quantification of the vasculature in the 3D stack was performed in Amira ( FEI , Hilsboro ) . The tumor and an 'outside region' were segmented in UM data sets ( 100–200 images per stack ) . Segmentation of the vasculature was performed semi-automatically based on pixel intensity . Thresholded images were median filtered , skeletonized and vessel length and radius were quantified . To correlate MRI findings to histology 10 µm coronal cryostat sections were cut . Hematoxylin & eosin staining was performed for anatomical assessment . Lectin-FITC or lectin-texas red was injected iv in healthy animals ( n=2 ) and dye labeling of endothelial cells was assessed on cryosections . Immunohistochemistry for endothelium ( CD31 antibody , BD Bioscience ) was performed using standard immunohistochemistry protocols and analyzed on a confocal laser-scanning microscope ( Olympus FV1000 ) . CD31 staining in lectin-FITC injected mice was performed in three GL261 tumor-bearing animals . Data is shown as mean ± SEM . Statistical analyses were performed in PRISM ( GraphPad ) . Two-tailed student’s t-tests were used to compare two groups . One-way ANOVA with Bonferroni’s post hoc testing was used for multiple comparisons . p-Values < 0 . 05 were considered significant . *denotes p<0 . 05; **p<0 . 01; ***p<0 . 001 .
Glioma-bearing mice were monitored weekly with high-field MRI ( 9 . 4 T ) starting 1 week after tumor cell implantation . Tumors could be clearly visualized by Gd-enhanced T1-weighted imaging ( Figure 1a ) . Tumors grew rapidly and untreated mice got symptomatic and were sacrificed 4–5 weeks post implantation when they had developed large intracranial tumors ( Figure 1a ) . To monitor vascularization both in the healthy brain and under pathological conditions , we employed a gradient echo , T2*-w sequence with high spatial resolution ( 80 µm isotropic resolution ) . This sequence allows the visualization of venules due to the BOLD effect of deoxygenized blood and the resulting susceptibility signals before contrast administration ( Figure 1b ) . After administration of a clinically applied Gd-contrast agent , arterioles and venules show strong tubular vascular susceptibility signals that allow the assessment of the vascularization status ( Figure 1b ) . Also , an iron-based experimental contrast agent ( ultrasmall iron oxide nanoparticle , USPIO ) resulted in a good delineation of the vasculature ( Figure 1—figure supplement 1 ) . In the healthy brain , for example , basal ganglia and cortical penetrating vessels can be visualized ( Figure 1b and Figure 1—figure supplement 1 ) . Under pathological conditions arterioles and venules within the tumor bed could be identified starting 2 weeks post-tumor implantation ( Figure 1c , and Figure 1—figure supplement 2 ) . First detectable vessels run both centrally and in the periphery of the developing glioma and could only be visualized after the administration of contrast material as tubular hypointense structures ( Figure 1c ) . 10 . 7554/eLife . 11712 . 003Figure 1 . Imaging tumor vessel development with T2*-w sequences . Time course of tumor development on T1-w post Gd-contrast images ( a ) . T2*-w images ( 80 µm resolution ) . Hypointense tubular structures , most likely venules ( arrowheads ) , are visible due to the BOLD effect on pre-contrast images ( upper row ) . Post-contrast administration arterioles and venules can be visualized ( middle row ) . A minimum intensity projection ( mIP ) is shown in the bottom row ( b ) . T2*-w images two and 3 weeks post-tumor implantation ( c , d ) . Quantification of tumor sizes on T1-w post-contrast images ( e ) . Quantification of the vascularized area on T2* images ( f ) . Histogram analysis of the tumor region on post-contrast images 2 ( green distribution ) and 3 weeks ( red ) after tumor implantation . A significant signal drop ( red distribution ) within the tumor relative to the healthy white matter ( WM ) occurs within 1 week ( p<0 . 001 , g ) . This signal drop is only visible post-contrast administration ( red distribution , p<0 . 001 , h ) . Single plane T2*-w images pre- and post-contrast 2 weeks post-tumor implantation . The tubular vessel is only visible after contrast injection ( i ) . Note the tortuous appearance and the multiple branches of the vessel . Scale bars are 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 00310 . 7554/eLife . 11712 . 004Figure 1—figure supplement 1 . Monitoring vascularization with USPIOs in healthy mice . T2*-w images pre- and post iron oxide nanoparticle ( USPIO ) administration . Post-contrast images were acquired directly after iv injection of USPIO . mIP: minimum intensity projection . Scale bar 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 00410 . 7554/eLife . 11712 . 005Figure 1—figure supplement 2 . Imaging tumor vasculature with USPIOs . T2*-w images pre- and post USPIO administration in a GL261 tumor . Post-contrast images were acquired directly after USPIO injection . Dashed boxes indicate magnified areas ( right side ) . mIP: minimum intensity projection . Scale bar is 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 00510 . 7554/eLife . 11712 . 006Figure 1—figure supplement 3 . Gd contrast kinetics in T2*-w time series Sequential . T2*-w images following Gd-administration . Post-contrast images 15 min after Gd-administration show strong tubular susceptibility changes within the tumor . The contrast agent is subsequently cleared from the circulation . 125 min post-contrast administration the vascular signals have returned to pre levels . Scale bar = 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 006 Within a week from the detection of first neovessels massive additional hypointense vessel-like structures occurred in the entire tumor bed as visualized on T2*-w images ( Figure 1d and Videos 1 , 2 ) . At this 'late' stage , single vessels were hard to differentiate within the tumor core because vascular density became too high for single vessel differentiation . Next to vascular susceptibility signals , also some dot-like susceptibility signals were found before contrast administration , most likely representing tumor microbleedings ( Figure 1d ) . Kinetic studies showed that the Gd contrast agent is cleared from the circulation within ~2 hr after iv injection and vascular signals return to baseline within this time period ( Figure 1—figure supplement 3 ) . Quantification of the tumor size and vascularized area showed an exponential increase of tumor vessels over the time of investigation 1–3 weeks post-implantation ( vascularized area week 1: 0 . 2 mm2 , week 2: 0 . 9 mm2 , week 3: 3 . 5 mm2 , p<0 . 01; Figure 1e , f ) . Histogram analysis of MR images showed a strong signal drop in the tumor area 3 weeks post-tumor cell implantation which was present only after contrast administration ( Figure 1 g , h ) . Interestingly , MRI was sensitive enough to detect single tumor vessels starting 2 weeks post implantation ( Figure 1i ) . The resolution of MRI is , however , intrinsically limited to the µm range which is insufficient to resolve tumor vessels below ~50 µm in size . To validate and correlate our MR results to the cellular level , we performed UM in combination with tissue clearing . 10 . 7554/eLife . 11712 . 007Video 1 . T2*-w image stack pre- and post-Gd contrast from a mouse 3 weeks after GL261 tumor implantation . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 00710 . 7554/eLife . 11712 . 008Video 2 . T2*-w post-contrast image stack 2 and 3 weeks after GL261 tumor implantation . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 008 We adapted recently published light sheet microscopy protocols of cleared whole-brain specimen ( Ertürk et al . , 2012; Schwarz et al . , 2015 ) . Similar to MRI , UM results in 3D datasets but possess a ~10-fold higher resolution ( Keller and Dodt , 2012 ) . For UM , the microvasculature was assessed by lectin-FITC or lectin-texas red . These intravital dyes bind to endothelial glycoproteins in healthy tissue and tumor endothelium ( Figure 2b and Figure 2—figure supplement 1 , Wälchli et al . , 2015 ) . After MR imaging , animals were injected intravenously with lectin and sacrificed after 5 min of dye circulation . Subsequently , brains were explanted , cleared and vascularization was assessed by UM . The pathological neovascularization of the xenografted glioma was clearly visible: Single vessels in the tumor bed could be well delineated on correlated MRI and UM datasets ( MR-UM ) 2 weeks post-tumor implantation ( Figure 2c and Videos 3 , 4 ) . Vessel density massively increased and large pathological vessels occurred 3 weeks post-tumor implantation ( Figure 2d ) . In some larger tumors parts of the tumor core became necrotic and no vessels were present in these areas ( Figure 2d ) . The MR-UM approach can also be employed for additional targets: for example , conventional Gd-enhanced MRI can easily be correlated to optical Evans blue extravasation studies to assess BBB-D and vascular permeability within the tumor ( R2: 0 . 87 , p<0 . 001; Figure 2—figure supplement 2a–c ) . 10 . 7554/eLife . 11712 . 009Figure 2 . Correlated MR-UM provides complementary information of the tumor vascular architecture . Illustration of the mouse brain before and after clearing using the 3DISCO protocol . The brain shrinks by ~40% in size during the clearing protocol ( a ) . Cleared UM images of lectin-FITC stained microvessels . Images show the healthy cortex ( left ) and hippocampus ( middle ) . The glioma-stroma border ( dotted line ) is depicted on the right image ( b ) . T2*-w images and correlative UM images 2 weeks , ( c ) and 3 weeks ( d ) after GL261 tumor implantation . Arrowheads indicate areas of necrosis . T2*-w image and correlative UM images 10 weeks after S24:td-tomato implantation . Inner necrotic tumor areas around the injection track lack fluorescent signal ( e ) . The microvasculature is stained with lectin-FITC . MIP: maximum intensity projection , mIP: minimum intensity projection . Scale bar in ( a ) is 5 mm and 1 mm for MR images . For UM scale bars are 250 µm in ( b ) and 1 mm in ( c-e ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 00910 . 7554/eLife . 11712 . 010Figure 2—figure supplement 1 . Lectin-FITC and lectin-texas red staining in healthy mice . Representative cryosections ( 10 µm thickness ) of microvessels stained with iv lectin-FITC or lectin-texas red . Dashed boxes indicate magnified area ( right side ) . Scale bars are 100 µm on overview images and 25 µm on magnified images . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01010 . 7554/eLife . 11712 . 011Figure 2—figure supplement 2 . Correlated permeability imaging using MR-UM . Two levels of MRI ( upper row ) and UM images ( lower row ) after Gd or evans blue injection . The permeability and disrupted BBB can be compared side-by-side ( a ) . Correlation analysis of BBB-D on MR and UM ( b ) . Evans blue extravasation on cryosection , as assessed by confocal microscopy . The image shows a tile scan of the right hemisphere 2 weeks after GL261 tumor implantation ( c ) . Dashed line indicates the tumor-stroma border . Scale bars = 1 mm in ( a ) and 500 µm in ( c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01110 . 7554/eLife . 11712 . 012Figure 2—figure supplement 3 . Time course of S24 tumor development . T2-w ( upper row ) , T1-w post Gd-contrast ( middle row ) and T2*-w post-contrast images of S24 tumors 4 , 7 , and 10 weeks post-tumor implantation . Dashed outline depicts hyperintense areas . Note the heterogeneous tumor mass/edema ( T2 signal ) and BBB-D ( contrast enhancement on T1-w ) . Arrowheads indicate vascular susceptibility signals . Scale bar = 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01210 . 7554/eLife . 11712 . 013Video 3 . Ultramicroscopy movie from a mouse injected with lectin-FITC and cleared using the FluoClearBABB protocol . Magnified image shows the tumor-stroma border and adjacent cortex ( green box ) from a mouse 2 weeks after GL261 tumor implantation . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01310 . 7554/eLife . 11712 . 014Video 4 . Correlated MR and UM stack 3 weeks after GL261 tumor implantation . Areas of necrosis are visible on MR and UM and are devoid of vessels . Box indicates the region of UM . UM data is shown as maximum intensity projection of six single images . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 014 The comparison of growth and angiogenesis patterns of primary , human S24 tumors with syngeneic mouse GL261 tumors by MR-UM revealed striking differences: S24 tumors grew markedly slower and highly invasive compared to GL261 tumors . Individual S24 tumor cell clusters were found outside of the tumor bulk even in the contralateral hemisphere . BBB-D , an early feature of GL261 tumors , developed in S24 tumors only at 10 weeks post-tumor implantation ( Figure 2—figure supplement 3 ) . At this time point , S24 tumors started to show clear signs of increased vascularity at the tumor stroma border on T2* post-contrast images and correlated UM ( Figure 2e , Figure 2—figure supplement 3 and Video 5 ) . UM also confirmed tumor cell infiltration into the contralateral hemisphere , which induced focal areas of aberrant vessel patterns with increased vessel density and diameter . The stable expression of red-fluorescent protein ( td-tomato ) by the S24 line allowed dual-color UM to assess the relation of cellular growth patterns and microvessel anatomy ( Figure 2e ) . 10 . 7554/eLife . 11712 . 015Video 5 . Correlated MR and UM stack 10 weeks after S24 tumor implantation . UM data is shown as maximum intensity projection of 10 single images . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 015 For the quantification of UM data , vessels were segmented on single image planes and quantified using Matlab ( Figure 3a ) . We found a progressive increase in vessel caliber 3 weeks post-tumor implantation compared to healthy regions ( mean vessel diameter in outside regions: 3 . 2 µm; tumor region week 2: 11 . 5 µm , week 3: 14 . 7 µm; p<0 . 001 , Figure 3b ) . Also , the tortuosity of tumor vessels was higher ( tortuosity in outside regions: 1 . 11; tumor: 1 . 16; p=0 . 05; Figure 3c ) . Vessel density showed a trend toward higher densities within the tumor ( 0 . 16 vs . 0 . 11 , n . s . Figure 3d ) . The assessment of vessel parameters in entire 3D data sets ( 100–200 images/animal ) was performed with Amira software and showed likewise a significant increase of vessel radii and vessel pathlength within the tumor area compared to outside regions ( Figure 3e–h ) . Large , CD31 positive microvessels with increased tortuosity were also confirmed within the tumor by immunohistochemistry ( Figure 3—figure supplement 1a , b ) . In the tumor , however , some CD31-positive vessels were negative for lectin-FITC , indicating non-perfused vessels , whereas outside of the tumor all microvessels were double positive for CD31 and lectin-FITC ( Figure 3—figure supplement 1c ) . 10 . 7554/eLife . 11712 . 016Figure 3 . Quantification of neoangiogenesis by ultramicroscopy . Representative single plane image of a cleared brain in the tumor and an 'outside' region . Illustration of the vessel segmentation using the 'tubeness' plugin . Vessel segments ( magenta ) and vessel outline ( green ) are used to determine the vascular diameter ( a ) . Quantification of the vessel diameter , tortuosity and vessel density ( b–d ) . Illustration of vessel segmentation in 3D ( e , f ) . Vessel radii ( g ) and pathlength ( h ) are shown . Scale bars = 100 µm in ( a ) and 500 µm in ( e ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01610 . 7554/eLife . 11712 . 017Figure 3—figure supplement 1 . Histological assessment . Hematoxylin and eosin staining shows the tumor in the right hemisphere . Magnified image depicts the tumor-stroma border ( a ) . Immunohistochemistry for CD31 shows vascular staining patterns in the tumor and an outside region ( b ) . Representative immunohistochemistry image , stained for CD31 ( red ) from a mouse injected with lectin-FITC ( green ) , ( c ) . Magnified images show the co-localization of CD31 and lectin outside of the tumor and two tumor vessels ( one double positive vessel , middle row; one vessel only positive for CD31 , lower row ) . n=3 mice . Dashed boxes indicate magnified area and dashed line shows the tumor-stroma border . Scale bars = 100 µm in a , b; 200 µm in c and 20 µm in magnified images . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 017 To investigate the power of MR-UM for treatment monitoring , we inhibited the vascular endothelial growth factor with a murine antibody and monitored treatment effects on the vascular compartment . We found that neovessel-formation was partially blocked by VEGF inhibition ( Figure 4a , b and Video 6 ) . Also , the vascularized area and BBB-D decreased as assessed by T2*-w and DCE imaging ( Figure 4c–f ) . UM quantification confirmed a partial block of pathological vessel features ( vessel diameter and density ) by VEGF inhibition ( mean vessel diameter under VEGF treatment: 8 . 3 µm; isotype control: 13 . 2 µm; p<0 . 001; Figure 4g , h ) . The beneficial effect on the vascular compartment was specific for VEGF targeted therapy , as cytotoxic treatment with photon irradiation did not lead to signs of vascular normalization ( Figure 4—figure supplement 1 ) . Despite the efficient blockage of neoangiogenesis , anti-VEGF therapy did not lead to reduced tumor growth ( tumor size 1 week after VEGF inhibition: 0 . 22 mm2 vs 0 . 22 mm2 in the isotype control group; p>0 . 05; four mice per group ) , reflecting current clinical data ( Chinot et al . , 2014; Taal et al . , 2014 ) . 10 . 7554/eLife . 11712 . 018Figure 4 . Monitoring treatment effects of VEGF inhibition on glioma vessels using MR-UM . Experimental outline ( a ) . Single plane , T2*-w images before ( week 2 ) and after VEGF or isotype control treatment ( week 3 ) . Treatment was initiated 2 weeks after tumor implantation when a solid tumor component had formed as confirmed on MRI . Correlative UM is shown of the same animal ( b ) . Permeability ( Ktrans ) maps , calculated from DCE MRI are depicted in ( c ) . Quantification of the vascularized area on T2*-w images ( d ) . Quantification of the blood-brain barrier disruption ( BBB-D ) on DCE images ( e ) . Correlation of BBB-D and the vascularized area ( f ) . Quantification of vessel diameter and vessel density on UM images ( g , h ) . MIP: maximum intensity projection . Scale bars are 1 mm on MR images and 500 µm on UM images . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01810 . 7554/eLife . 11712 . 019Figure 4—figure supplement 1 . Irradiation does not reduce tumor vascularization . T1-w post contrast images show tumor growth in irradiated mice ( upper row ) . T2*-w images post-contrast ( middle and lower row ) 2 ( before treatment ) and 3 weeks post-tumor implantation ( after treatment ) . Irradiation with two gray was performed on 4 consecutive days starting 2 weeks post-tumor implantation . Tumor sizes and vascularization status were assessed 3 weeks post-tumor implantation . mIP: minimum intensity projection . Scale bar = 1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 01910 . 7554/eLife . 11712 . 020Video 6 . Correlated MR and UM stack 3 weeks after tumor implantation . Animal was treated with a mouse VEGF inhibitor every 3 days ( 10 mg / kg ) starting 2 weeks after tumor implantation . DOI: http://dx . doi . org/10 . 7554/eLife . 11712 . 020
In this study , we combined in vivo 9 . 4 Tesla MRI with ex vivo UM and tissue clearing to dynamically resolve tumor arterioles and venules as well as the capillary bed of mouse gliomas at high resolution . Contrast-enhanced T2*-w imaging allowed the visualization of neovessels in the 50–100 µm range . In the GL261 model , angiogenesis started 2 weeks post-tumor implantation and massively increased until week 3 . One limitation of MRI concerns its resolution , which was limited to ~80 µm in our study . To overcome this limitation , we employed correlated 3D UM of cleared specimen . For UM , we applied iv perfusion with fluorescent lectins that bind specifically to endothelial glycoproteins ( Wälchli et al . , 2015 ) and subsequent optical clearing of whole adult mouse brains using the well-established 3DISCO and the recently published FluoClearBABB protocol ( Ertürk et al . , 2012; Schwarz et al . , 2015 ) . We obtained high-quality images with both clearing methods that were used to reconstruct the entire tumor anatomy and blood vessel architecture including bulk mass and invasive zone . Tissue clearing and UM has not been employed for the study of neurooncological disease . Previous UM studies investigated microvessels under physiological conditions ( Jährling et al . , 2009 ) and in the context of solid visceral tumors ( Dobosz et al . , 2014 ) . We have extended this work to the field of neurooncology and combined it with advanced MRI techniques . This bridges the gap between MRI and whole brain optical microscopy , which represent separated domains so far . Optical clearing methods have recently undergone a rapid development and evoked large interest in the neuroscience community as they promise to allow 'connectomic' and circuit reconstruction studies ( Chung et al . , 2013; Dodt et al . , 2007; Ertürk et al . , 2011; Keller and Dodt , 2012; Schwarz et al . , 2015; Susaki et al . , 2014 ) . Clearing techniques are , however , also amenable to the study of pathological states: First reports have highlighted how 3D information obtained by tissue clearing can give new insights into pathophysiology of neuroinflammatory and neurodegenerative disease ( Jährling et al . , 2015; Spence et al . , 2014 ) . The combination with MR , the most widely used clinical imaging technique , offers added value by correlating spatiotemporal information obtained by MR with high-resolution optical imaging . In the preclinical arena MR and UM with tissue clearing appear as natural partners as both produce 3D datasets , which can be compared side-by-side . The advantages of both techniques can be combined: Namely , MR is a versatile tool for longitudinal in vivo studies . UM of cleared specimen can resolve cellular and subcellular processes and utilize the vast toolbox of genetic and chemical fluorescent labels with their high molecular specificity , thus complementing the information gained by MRI . Also , a combination of the two techniques allows cross-correlation of in vivo signals with molecularly defined optical methods . To show the additive power of both methods , we used a neovascularization paradigm since neovessels are hallmarks of many diseases including cancer ( Carmeliet and Jain , 2011; Goveia et al . , 2014 ) . Using cross-correlated MR-UM we achieved single vessel resolution and found pathological vessel permeability , tortuosity , and calibers , all of which could be partially reversed by VEGF treatment . Our data indicate that similar to the human situation , mono-modality antiangiogenic treatment is not sufficient to halt glioma growth . Thus , novel antiangiogenic treatment regimes , possibly in combination with chemotherapy in recurrent disease ( as done with promising results in the Dutch Belob Trial ( Taal et al . , 2014 ) , and the ongoing EORTC 26 , 101 and 26 , 091 trials ) , or in combination with additional treatment modalities ( i . e . immunotherapy , chemotherapy ) are warranted ( Chinot et al . , 2014; Kamoun et al . , 2009 ) . Anti-angiogenic treatment development and clinical trials require imaging biomarkers to assess vascularization status and treatment effects . The developed T2*-w sequence is suitable to map tumor arterioles and venules in a preclinical setting . Future studies at high-field clinical MR systems should address a possible translation of our MRI approach to the clinical arena . One limitation of our study relates to the differentiation of the various hierarchical parts of the vascular system . Tumors harbor a manifold of different vascular compartments ( veins and arteries of different calibers and sizes with highly variable , but mostly pathological perfusion ) . Neither MR nor optical methods are able to capture the entire complexity of these compartments . Also , we found that in the bulk tumor some CD31 positive vessels were negative for lectin-FITC , indicating that some tumor vessels were hypo- or non-perfused . This implies that the visualization and quantification of the microvasculature by lectin-injection might slightly underestimate the vascular tumor compartment . The fact that hypo- or non-perfused vessels with abnormal flow exist in tumors is well established ( Carmeliet and Jain , 2011 ) . Another drawback might be that clearing methods may lead to an alteration in brain size , that is , due to dehydration . For 3DISCO , we quantified the apparent reduction of brain volume using CT imaging before and after clearing . This revealed a volume reduction of ~40% , which however occurred in a uniform fashion and did not change macroscopic tissue proportions . Anatomical structures and substructures of the brain remained fully intact . Vessel diameters obtained by UM could therefore be corrected by a multiplication factor of ~1 . 6 . However , microscopic shrinkage may slightly differ from the macroscopic factor obtained by CT . Also , when using endogenous fluorescent proteins such as GFP or RFP in combination with exogenous fluorescent markers such as FITC , preservation of fluorescent signals during the clearing process might vary and should be carefully controlled for . In summary , the present study shows the additive power of correlative MR and UM for assessing the vascular system and the dynamic changes that occur in tumors over time . The applied principles should be easily transferable to additional targets like the immune cell compartment that are amenable to MR and optical labeling strategies ( Kircher et al . , 2003; Weissleder et al . , 2014 ) . Furthermore , MR-UM should be useful for translational approaches to aid preclinical therapy development . In our study , mono-antiangiogenic treatment with a murine VEGF inhibitor was insufficient to halt tumor growth which mirrors current human studies ( Chinot et al . , 2014 ) . Dual angiogenesis inhibitors are currently being developed and therapeutic effects could be assessed in detail using our toolkit . Thus , we believe that MR-UM can provide a versatile platform for translational research approaches to cross-correlate MR and optical signals . | Blood vessels are the body’s highways that allow blood to transport oxygen , nutrients , hormones and waste products quickly and efficiently around the body . Tumors are made up of particularly active cells and so their growth heavily depends on blood vessels . Indeed , a fundamental hallmark of tumor progression is for nearby blood vessels to form more quickly . Tumor blood vessels also differ in structure from their normal counterparts for reasons that need to be investigated in more detail . Compounds that block the formation of blood vessels have been developed for treating highly malignant brain tumors called gliomas . However , although many of these compounds show promising effects in preclinical trials , clinical trials on humans have been less successful . Having the ability to image the blood vessels in high detail during preclinical trials would help to reveal how treatments that inhibit blood vessel formation work and how tumors might develop resistance to these drugs . However , studying tumor blood vessels remains a challenge due to technical restrictions: techniques that are able to capture how the vessels change over time are unable to show individual cells in much detail , and vice versa . Magnetic resonance imaging is a versatile tool that can monitor how the blood vessel system of a tumor changes over time in living animals . On the other hand , ultramicroscopy is able to determine the structure of single cells of a particular type . By combining these techniques , Breckwoldt , Bode et al . have now developed a imaging platform that allows the formation of tumor blood vessels to be precisely mapped in the setting of a preclinical study . It also enables detailed investigations into how the structure of the blood vessels is altered by treatments that aim to inhibit the formation and growth of new vessels . Using this approach on mice with gliomas , Breckwoldt , Bode et al . demonstrated that drugs that inhibit the formation of the blood vessels that supply tumors also cause the blood vessels to take on a more normal structure . Furthermore , treating the mice with a single inhibitory drug was unable to stop tumor growth , mirroring the situation in humans . Currently , new inhibitors are being developed , offering the possibility of combined treatments that may be more effective than using a single drug on its own . The imaging platform developed by Breckwoldt , Bode et al . will allow the therapeutic effects obtained by these new treatments to be analyzed in detail during preclinical studies . | [
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2016 | Correlated magnetic resonance imaging and ultramicroscopy (MR-UM) is a tool kit to assess the dynamics of glioma angiogenesis |
Naïve human pluripotent stem cells ( hPSCs ) provide a unique experimental platform of cell fate decisions during pre-implantation development , but their lineage potential remains incompletely characterized . As naïve hPSCs share transcriptional and epigenomic signatures with trophoblast cells , it has been proposed that the naïve state may have enhanced predisposition for differentiation along this extraembryonic lineage . Here we examined the trophoblast potential of isogenic naïve and primed hPSCs . We found that naïve hPSCs can directly give rise to human trophoblast stem cells ( hTSCs ) and undergo further differentiation into both extravillous and syncytiotrophoblast . In contrast , primed hPSCs do not support hTSC derivation , but give rise to non-self-renewing cytotrophoblasts in response to BMP4 . Global transcriptome and chromatin accessibility analyses indicate that hTSCs derived from naïve hPSCs are similar to blastocyst-derived hTSCs and acquire features of post-implantation trophectoderm . The derivation of hTSCs from naïve hPSCs will enable elucidation of early mechanisms that govern normal human trophoblast development and associated pathologies .
Mammalian pluripotency spans a continuum of discrete but interconvertible states , each with a distinct set of molecular and functional attributes . Prior to implantation , the pluripotent epiblast compartment within the inner cell mass ( ICM ) of the blastocyst constitutes a naïve or ground state of pluripotency ( Nakamura et al . , 2016; Nichols and Smith , 2009; Stirparo et al . , 2018 ) . This naïve state can be captured in vitro in the form of mouse embryonic stem cells ( mESCs ) . After implantation , transcription factors associated with naïve pluripotency are downregulated and pluripotent cells become primed for differentiation in response to signals from the surrounding extraembryonic tissues . Human embryonic stem cells ( hESCs ) derived under conventional conditions are thought to represent a primed pluripotent state , and were shown to correspond transcriptionally to the late post-implantation epiblast in a non-human primate model ( Nakamura et al . , 2016; Nichols and Smith , 2009 ) . Much effort has been made in recent years to develop strategies for capturing hESCs in a naïve pluripotent state ( Chan et al . , 2013; Gafni et al . , 2013; Hanna et al . , 2010; Qin et al . , 2016; Takashima et al . , 2014; Theunissen et al . , 2014; Ware et al . , 2014; Zimmerlin et al . , 2016 ) . In particular , two transgene-free culture systems , 5i/L/A and t2i/L/Gö , were shown to induce defining transcriptional and epigenetic features of the human pre-implantation epiblast ( Huang et al . , 2014; Liu et al . , 2017; Stirparo et al . , 2018; Takashima et al . , 2014; Theunissen et al . , 2016; Theunissen et al . , 2014 ) . The isolation of naïve hESCs provides a cellular experimental platform to interrogate aspects of human pre-implantation development that are difficult to study in primed hESCs . For example , naïve hESCs have offered insights into the mechanisms governing X-linked dosage compensation ( Sahakyan et al . , 2017 ) , the role of human-specific transposable elements that are expressed in the pre-implantation embryo ( Pontis et al . , 2019; Theunissen et al . , 2016 ) , and the mechanisms leading to activation of naïve-specific enhancers ( Pastor et al . , 2018 ) . Naïve hESCs may also afford a platform for dissecting cell fate decisions in the early human embryo ( Dong et al . , 2019 ) . While naïve hESCs are unresponsive to direct application of embryonic inductive cues , they acquire the capacity to undergo efficient multi-lineage differentiation upon treatment with a Wnt inhibitor , a process called ‘capacitation’ ( Rostovskaya et al . , 2019 ) . This process is thought to reflect the requirement for dismantling of the naïve transcriptional program upon implantation in vivo , and the acquisition of a differentiation-competent formative phase ( Smith , 2017 ) . Molecular profiling of naïve hESCs has suggested that these cells may harbor a predisposition towards human extraembryonic cell fates . Gene expression studies revealed a pronounced upregulation in naïve relative to primed hESCs of trophoblast-associated transcription factors , including ELF3 , GCM1 , and TFAP2C ( Theunissen et al . , 2016 ) . In addition , chromatin accessibility studies indicated that naïve hESCs share a broad panel of open chromatin sites with first-trimester placental tissues ( Pontis et al . , 2019 ) . Intriguingly , embryonic and extraembryonic lineage markers are briefly co-expressed in the late morula and early blastocyst according to single cell RNA-seq ( scRNA-seq ) studies of human pre-implantation embryos ( Petropoulos et al . , 2016 ) . This is precisely the stage of human development that displays the closest correspondence to naïve hESCs based on the expression patterns of transposable elements ( Theunissen et al . , 2016 ) . Thus , we surmised that current methodologies for inducing naïve human pluripotency may yield a pre-implantation identity that is competent for both embryonic and extraembryonic differentiation . Here , using three independent methodologies , we find that naïve hPSCs have enhanced capacity for differentiation along the trophoblast lineage relative to primed hPSCs . In particular , we show that when cultured in human trophoblast stem cell ( hTSC ) media ( Okae et al . , 2018 ) , naïve hPSCs can directly give rise to hTSCs , as confirmed by morphological , molecular , and transcriptomic criteria . We have also profiled the chromatin accessibility landscape of hTSCs for the first time , thus providing a valuable resource to identify potential regulatory elements and transcriptional determinants of human trophoblast development .
As a first step toward examining the trophoblast potential of naïve and primed hESCs , we measured the expression levels of trophoblast-associated markers during embryoid body ( EB ) formation ( Figure 1A ) , which provides a rapid assessment of spontaneous differentiation capacity into early lineages ( Allison et al . , 2018 ) . Previous studies reported limited induction of embryonic lineage markers in EBs formed from naïve hESCs , but did not examine the expression of trophoblast-associated genes ( Liu et al . , 2017; Rostovskaya et al . , 2019 ) . We generated naïve hESCs in 5i/L/A ( Theunissen et al . , 2014 ) from two genetic backgrounds , H9 and WIBR3 , confirmed their upregulation of naïve-specific markers and downregulation of primed-specific markers ( Figure 1—figure supplement 1A ) , and aggregated them to form EBs in growth factor- and inhibitor-free media for 12 days ( Figure 1A; Figure 1—figure supplement 1B ) . The mRNA expression levels of six trophoblast markers , ELF5 , KRT7 , TFAP2C , GATA3 , TEAD4 , and CDX2 ( Hemberger et al . , 2010; Lee et al . , 2016; Ng et al . , 2008; Strumpf et al . , 2005 ) , were measured by quantitative real time PCR ( qRT-PCR ) analysis ( Figure 1B; Figure 1—figure supplement 1C ) . Significantly enriched expression of trophoblast markers was observed in EBs derived from naïve compared to primed hESCs ( Figure 1B; Figure 1—figure supplement 1C ) . Furthermore , many of the examined trophoblast markers were already elevated in naïve versus primed hESCs prior to EB formation ( Figure 1B; Figure 1—figure supplement 1C ) , consistent with our prior transcriptome analysis ( Theunissen et al . , 2016 ) . These findings support the notion that naïve hESCs harbor increased spontaneous trophoblast differentiation potential compared to primed hESCs . We next assessed the trophoblast potential of naïve relative to primed hESCs using a protocol for directed trophoblast differentiation that utilizes a low dose of bone morphogenetic protein 4 ( BMP4 ) ( Horii et al . , 2016 ) . It has long been known that primed hPSCs acquire certain trophoblast characteristics upon stimulation with BMP4 ( Amita et al . , 2013; Horii et al . , 2016; Xu et al . , 2002 ) . However , when naïve hESCs were subjected to a protocol for BMP4-directed differentiation into cytotrophoblast ( CTB ) progenitors ( Horii et al . , 2016 ) , the cells did not survive ( Figure 1C ) . This recalcitrance to BMP4 is reminiscent of the delayed response of naïve hPSCs to embryonic inductive cues ( Liu et al . , 2017; Rostovskaya et al . , 2019 ) . We examined whether naïve hESCs would gain the capacity for BMP4-directed differentiation upon returning to the primed state , a process referred to as ‘re-priming’ ( Theunissen et al . , 2016 ) . Indeed , naïve hESCs treated with StemPro for five passages re-gained competence for BMP4-directed trophoblast differentiation , as shown by the expression of several CTB markers ( Figure 1D , E; Figure 1—figure supplement 2A , B ) . When these CTBs were further differentiated using feeder-conditioned medium supplemented with BMP4 , we observed the expected induction of extravillous trophoblast ( EVT ) and syncytiotrophoblast ( STB ) marker genes , as observed for differentiation from primed hESCs ( Horii et al . , 2016; Figure 1G; Figure 1—figure supplements 1D and 2A , B ) . We also examined whether capacitation of naïve hESCs using the Wnt inhibitor , XAV939 ( Rostovskaya et al . , 2019 ) , would confer responsiveness to BMP4-directed trophoblast differentiation . Capacitated cells could not only readily undergo Step I CTB differentiation , but did so more efficiently than primed hESCs based on analysis of trophoblast-specific transcripts and proteins ( Figure 1F , G; Figure 1—figure supplements 1D and 2C ) . These results indicate that an efficient response to BMP4-directed trophoblast differentiation requires exit from naïve human pluripotency , and appears to be most efficient when initiated from the formative phase . The above experiments indicate that naïve hESCs efficiently upregulate trophoblast markers during spontaneous EB differentiation , but require transition into a formative pluripotent state to become competent for BMP4-directed trophoblast differentiation . This led us to consider whether naïve hESCs might be directly responsive to alternative conditions for trophoblast differentiation that are independent of BMP4 . A recent study reported the derivation of hTSCs from blastocysts and first-trimester placental tissues in the presence of recombinant epidermal growth factor ( EGF ) and Wnt activator ( CHIR ) , and inhibitors of transforming growth factor beta ( TGFβ ) , histone deacetylase ( HDAC ) , and Rho-associated kinase ( ROCK ) ( Okae et al . , 2018 ) . We seeded isogenic lines of naïve and primed hPSCs on Collagen IV and examined their response to hTSC media ( Figure 2A ) . Naïve hPSCs acquired a typical hTSC-like morphology within several passages and could be expanded for at least 20 passages while maintaining a high proliferation rate ( Figure 2B; Figure 2—figure supplement 1A ) . Similar results were obtained using three independent naïve hPSC lines derived from H9 hESCs , WIBR3 hESCs , and AN1 induced pluripotent stem cells ( iPSCs ) . In contrast , the parental primed hPSCs did not acquire an hTSC-like morphology , even after prolonged culture in hTSC media ( Figure 2B ) . These observations indicate that naïve hPSCs are capable of adapting to the specific culture environment of hTSCs , whereas primed hPSCs are not . We proceeded to further characterize the naïve hPSC-derived hTSC-like cells ( from here referred to as naïve hTSCs ) . We found that naïve hTSCs uniformly express ITGA6 and EGFR , two commonly used cell surface markers that mark both CTBs and hTSCs ( Bischof and Irminger-Finger , 2005; Horii et al . , 2016; Okae et al . , 2018 ) , in contrast to primed hPSCs that had been expanded in hTSC media ( Figure 2C; Figure 2—figure supplement 1B ) . The naïve hTSCs also expressed significantly higher levels of ELF5 , KRT7 , GATA3 , TFAP2C , and TEAD4 transcripts ( Figure 2D; Figure 2—figure supplement 1C ) . Notably , naïve hTSCs exhibited almost no CDX2 expression , which is consistent with hTSCs derived from blastocysts or first-trimester placental tissues ( Okae et al . , 2018; Figure 2D; Figure 2—figure supplement 1C ) . Naïve hTSCs also expressed the hTSC markers KRT7 , TEAD4 , and TP63 at the protein level ( Lee et al . , 2016; Li et al . , 2013; Figure 2E; Figure 2—figure supplement 1D ) . Flow cytometry analysis for CD75 and SUSD2 , two naïve-specific cell surface markers ( Bredenkamp et al . , 2019a; Collier et al . , 2017 ) , confirmed the loss of naïve identity in naïve hTSCs ( Figure 2F ) , which was further corroborated by downregulation of KLF17 transcript ( Figure 2G; Figure 2—figure supplement 1E ) . These findings indicate that naïve hPSCs directly give rise to a uniform population of cells that closely resemble hTSCs based on gross morphology , surface markers , and trophoblast-specific gene expression . Conversely , primed hPSCs did not upregulate trophoblast markers when cultured in hTSC media , but instead exhibited increased expression of VIM and PAX6 , suggesting the acquisition of a neuroectodermal fate ( Figure 2H; Figure 2—figure supplement 1F ) . To confirm that the ability of naïve hPSCs to give rise to hTSC-like cells is not specific to the 5i/L/A culture condition , we attempted to derive naïve hTSCs from naïve hPSCs cultured in an alternative naïve medium , PXGL ( Bredenkamp et al . , 2019b; Figure 2—figure supplement 1G ) . Both morphological and molecular analyses indicated that naïve hTSCs derived from PXGL-cultured naïve hPSCs closely resemble those derived from 5i/L/A-cultured naïve hPSCs ( Figure 2—figure supplement 1G–J ) . This suggests that naïve hTSC-derivation is likely an intrinsic property of the naïve state of human pluripotency , regardless of the specific culture condition used . Additionally , we investigated whether naïve hPSCs can give rise to naïve hTSCs on a clonal level , which would preclude that hTSCs arise from a small population of pre-existing trophoblast-like cells in the culture . We therefore picked and expanded three single-cell H9 naïve hPSC clones ( Figure 2—figure supplement 2A ) , and confirmed their naïve molecular characteristics and lack of hTSC marker expression ( Figure 2—figure supplement 2B , C ) . Subsequently , naïve hTSCs were derived from these clonally expanded naïve hPSCs . All of these clonally derived naïve hTSCs exhibit a typical hTSC morphology ( Figure 2—figure supplement 2A ) , and are negative for naïve hPSC markers but positive for hTSC markers ( Figure 2—figure supplement 2B , C ) . These results lend further support to the notion that the ability to give rise to hTSCs is an intrinsic property of naïve hPSCs . Since naïve hTSCs exhibit numerous characteristics of hTSCs derived from blastocysts or placental tissues , we sought to examine their differentiation potential into specialized trophoblast cells . First , we performed directed differentiation of naïve hTSCs into EVTs , which invade the endometrium to increase blood flow between the mother and fetus ( James et al . , 2012; Watson and Cross , 2005 ) . Following application of a culture system for EVT differentiation from hTSCs that contains Neuregulin 1 ( NRG1 ) , the TGF-β inhibitor A83-01 , and Matrigel ( Okae et al . , 2018 ) , naïve hTSCs acquired a characteristic EVT-like morphology ( Figure 3A , B ) . We tested whether the naïve hTSC-derived EVT-like cells expressed two EVT-specific protein markers , HLA-G and MMP2 ( Horii et al . , 2016; Lee et al . , 2016 ) . Flow cytometry analysis demonstrated that about 90% of EVT-like cells were positive for HLA-G ( Figure 3C ) . Similarly , expression and secretion of MMP2 was detected by ELISA and immunofluorescence staining ( Figure 3D; Figure 3—figure supplement 1A ) . The induction of HLA-G and MMP2 in naïve hTSC-derived EVT-like cells was also confirmed at the mRNA level by qRT-PCR ( Figure 3E; Figure 3—figure supplement 1B ) . Finally , since invasiveness is a prominent property of EVTs ( Burton et al . , 2009; James et al . , 2012; McEwan et al . , 2009 ) , we performed a transwell-based Matrigel invasion assay . The results indicate that our EVT-like cells have invasive potential , in contrast to the naïve hTSCs from which they were derived ( Figure 3F; Figure 3—figure supplement 1C ) . Second , we performed directed differentiation of naïve hTSCs into STBs , which are multinucleated cells that produce placental hormones and mediate maternal-fetal communication . We applied two protocols for differentiation of hTSCs into STBs involving either 2D or 3D culture in the presence of Forskolin ( Okae et al . , 2018; Figure 3G ) . The naïve hTSC-derived 2D STB-like cells showed characteristic morphological features , including multinucleation ( Figure 3—figure supplement 1D ) . They also expressed the STB marker human chorionic gonadotropin ( hCG ) ( Figure 3—figure supplement 1E ) . The naïve hTSC-derived 3D STB-like cells exhibited a cyst-like morphology typical for 3D STBs ( Okae et al . , 2018; Figure 3H ) , and expressed the STB markers hCG and SDC1 based on immunofluorescence and qRT-PCR analysis ( Jokimaa et al . , 1998; Strumpf et al . , 2005; Figure 3I–K ) . Finally , qRT-PCR analysis confirmed downregulation of the hTSC-specific marker TEAD4 ( Figure 3L ) , indicating the loss of hTSC identity . These morphological and molecular data demonstrate that hTSCs derived from naïve hPSCs are capable , at least at the population level , of undergoing further differentiation into two specialized trophoblast cell types , EVT and STB . To examine whether naïve hTSCs and their differentiated derivatives possess transcriptomic signatures of the trophoblast lineage , we sequenced total RNA isolated from three types of hPSCs ( naïve , capacitated , and primed ) , blastocyst-derived BT5 hTSCs ( Okae et al . , 2018 ) , naïve hTSCs and their differentiated progeny ( EVT and STB ) , and primed hPSCs cultured in hTSC media . Principal component analysis ( PCA ) revealed that the samples clustered together based on cell type , rather than genetic background ( Figure 4A ) . Pluripotent and trophoblast identities were predominantly divided along principal component 1 ( PC1 ) , which accounted for 41% of the variation in gene expression . In addition , naïve , capacitated , and primed hPSCs formed clearly defined clusters that were separated along PC2 . It is worth noting that naïve hTSCs and BT5 hTSCs clustered together very closely on the PCA plot ( Figure 4A ) . Further examination indicated that naïve hTSCs express key trophoblast marker genes at comparable or even higher levels than the bona fide , BT5 hTSCs ( Figure 4—figure supplement 1A ) . Whereas naïve hTSCs were well-separated from the pluripotent samples , primed hPSCs acquired a very distinct transcriptional profile in hTSC media compared to naïve hTSCs . Naïve hTSCs displayed strong upregulation of many transcripts associated with trophoblast development , whereas primed cells instead acquired neuroectodermal characteristics in hTSC media ( Figure 4B and Supplementary file 1 ) . Hence , exposure to hTSC media induces a drastically different transcriptomic profile when applied to either naïve or primed hPSCs . To further understand gene expression dynamics during naïve hTSC derivation and terminal EVT and STB differentiation , we identified six clusters of genes based on differential expression in naïve hPSCs , naïve hTSCs , and naïve hTSC-derived EVTs and STBs ( Figure 4C ) . Cluster one includes genes that are enriched in naïve hPSCs , such as DPPA5 , KLF4 , NANOG , POU5F1 , SUSD2 , and ZFP42 ( Figure 4D and Supplementary file 2 ) . This cluster is enriched in gene ontology ( GO ) terms associated with transcriptional regulation and embryonic development ( Supplementary file 3 ) . Cluster two includes genes such as LRP5 and TEAD4 that are enriched in both naïve hPSCs and naïve hTSCs ( Figure 4D and Supplementary file 2 ) . Cluster three includes genes that are enriched in naïve hTSCs , such as CCR7 , CTNNB1 , ELF5 , EPCAM , TFAP2C , ITGA6 , OVOL1 , and TP63 ( Figure 4D and Supplementary file 2 ) , and is enriched in GO terms associated with cytoskeletal remodeling ( Supplementary file 3 ) . Cluster four includes genes such as IGF1R , IGF2R , and KRT7 that are enriched in both naïve hTSCs and EVTs ( Figure 4D and Supplementary file 2 ) . Finally , clusters 5 and 6 include genes that are enriched in either EVTs or STBs derived from naïve hTSCs , respectively . EVT-specific genes include FN1 , HLA-G , ITGA1 , ITGA5 , and MMP2 , while STB-specific genes include CGA , CGB1 , INHA , PSG1 , SDC1 , and TBX3 ( Figure 4D and Supplementary file 2 ) . Notably , CGA and CGB1 encode the α and β chains of the placental hormone hCG , while TBX3 was recently identified as a transcriptional regulator of STB differentiation ( Lv et al . , 2019 ) . EVT-specific genes are enriched in GO terms related to cell migration and invasion , while STB-specific genes are enriched in terms related to membrane fusion and interferon signaling ( Supplementary file 3 ) . Many , if not all , of the genes highlighted in these clusters have the same specific gene expression profiles reported for primary hTSCs , EVTs , and STBs derived from human blastocysts ( Okae et al . , 2018 ) . We conclude that the derivation of hTSCs from naïve hPSCs and their subsequent , cell-type-specific differentiation recapitulate the processes of hTSC derivation from the human embryo and its differentiation into EVT and STB cell types . We examined whether hTSCs generated by direct derivation from naïve hPSCs share molecular signatures with human trophectoderm ( TE ) at pre- or post-implantation stages . In a recent study , human embryos were cultured in vitro past implantation , and single cell RNA-seq ( scRNA-seq ) was performed at days 6 , 8 , 10 , and 12 post-fertilization ( Zhou et al . , 2019 ) . This work identified genes that are specifically expressed in the human TE , EPI , and primitive endoderm ( PE ) lineages ( Zhou et al . , 2019 ) . We assayed the expression of these genes in naïve and primary hTSCs , then correlated their expression patterns to that of human TE , EPI , and PE at distinct time points ( Figure 4E; Figure 4—figure supplement 1B ) . We found that hTSCs cluster the most closely amongst themselves , as well as with day 10 and day 12 TE ( Figure 4E ) , while showing the best correlation with day 12 TE ( Figure 4—figure supplement 1B ) . In contrast , the hTSCs correlate poorly with the EPI or PE lineages ( Figure 4E ) . To further uncover the genes and biological processes that contribute to the similarities between naïve hTSCs and post-implantation TE , we selected those genes that are specifically expressed in the TE or EPI at day 10 or day 12 , and examined their expression in naïve hTSCs relative to naïve hPSCs ( Zhou et al . , 2019 ) . We observed significant correlation between naïve hTSC/naïve hPSC in vitro , and TE/EPI at day 10 or day 12 in vivo ( r = 0 . 61 ) ( Figure 4—figure supplement 1C ) . Additionally , we performed GO term analysis on the top 10% most highly upregulated genes in naïve hTSCs relative to naïve hPSCs ( Figure 4—figure supplement 1C and Supplementary file 4 ) . These genes include placenta-associated factors such as CGB7 , EGFR , ITGA2 , LHB and multiple pregnancy-specific glyocoproteins ( PSG2-6 ) , and are enriched in biological processes such as female pregnancy , reproductive processes , and positive regulation of NF-kappa B transcription factor activity ( Figure 4—figure supplement 1D ) . This provides additional validation that the process of hTSC derivation from naïve hPSCs activates genes and pathways that are relevant to human trophoblast biology . We conclude that hTSCs derived from naïve hPSCs are comparable to bona fide , blastocyst-derived hTSCs , and are transcriptionally the most similar to human post-implantation TE at day 12 . Finally , we profiled the chromatin accessibility landscape of two naïve-hPSC derived hTSC lines and a blastocyst-derived hTSC line , BT5 ( Okae et al . , 2018 ) , using the assay for transposase-accessible chromatin followed by sequencing ( ATAC-seq ) . Overall , we found ATAC-seq peaks in all three hTSC lines to be highly correlated globally ( Figure 4F ) , which indicates that derivation of hTSCs from naïve hPSCs induces a similar epigenomic profile as hTSC derivation from the human blastocyst . We identified differentially accessible regions ( DARs ) between naïve hESCs ( 9 , 059 ) and naïve hTSCs ( 12 , 132 ) , and found that naïve hESC-specific open chromatin sites are more enriched over promoter regions ( 1 , 084 ) and tend to be in closer proximity to the nearest transcriptional start site relative to hTSC-specific open chromatin sites ( Figure 4—figure supplement 2A , B ) . In hTSCs , only 208 DARs are located in gene promoter regions , and a higher percentage of hTSC-specific DARs are located in intronic and intergenic regions ( Figure 4—figure supplement 2A , B ) , suggesting the importance of distal regulatory elements during differentiation of naïve hPSCs into hTSCs . We further analyzed the ATAC-seq signal at the promoter regions of genes within differentially expresses genes ( DEG ) cluster 1 and 3 ( Figure 4C ) , which are genes more specifically expressed in naïve hPSCs and hTSCs , respectively . Consistent with DAR analysis , we noticed that genes in clusters 1 and 3 are associated with higher ATAC-seq signals at the promoter regions in naïve hESCs and hTSCs , respectively . However , the ATAC-seq signal at the promoter regions of cluster 3 ( hTSC-specific ) genes was only slightly increased in naïve hTSCs relative to naïve hESCs ( Figure 4—figure supplement 2C ) . This indicates that the promoter regions of hTSC-specific genes are already accessible in naïve hESCs , which is consistent with previous findings that naïve hPSCs share open chromatin regions with first-trimester placental tissues ( Pontis et al . , 2019 ) . To identify transcription factors that may differentially regulate naïve hESCs vs . naïve hTSCs , we examined enrichment for known transcription factor motifs in differentially accessible ATAC peaks . Naïve hESC-specific enrichment was observed for the POU5F1:SOX2 and KLF4 motifs ( Figure 4G ) , consistent with the roles of these transcription factors in regulating naïve pluripotency ( Guo et al . , 2009; Stuart et al . , 2019; Wong et al . , 2016 ) . Intriguingly , we observed significant enrichment for motifs associated with TEAD4 , CEBP , GATA , and TFAP2 over non-promoter regions in naïve hTSCs , while motifs associated with TFAP2 , GATA , SP1 , ELK1 , and GRHL were enriched over promoter regions in naïve hTSCs ( Figure 4G ) . Similar motif enrichments were observed for naïve hESCs vs . BT5 hTSC DARs ( Figure 4G ) . Among these transcription factors , TEAD4 and TFAP2 are known to be expressed in human TE ( Petropoulos et al . , 2016 ) , while CEBP and GATA factors have been implicated as regulators of trophoblast-specific HLA-G expression at the maternal-fetal interface ( Ferreira et al . , 2016 ) . The strong enrichment of the TEAD4 binding motif in hTSCs is especially significant , as it suggests that TEAD4 may play a significant role during TE specification in both mouse and human ( Nishioka et al . , 2009; Nishioka et al . , 2008; Yagi et al . , 2007 ) . To explore the role of TEAD4 in hTSC derivation from naïve hPSCs , we treated the cells with Verteporfin , an inhibitor that disrupts YAP-TEAD4 interactions and was recently shown to prevent cavity formation in blastocyst-like structures generated from mouse extended pluripotent stem ( EPS ) cells ( Li et al . , 2019 ) . Verteporfin treatment induced widespread cell death within several days when applied during hTSC derivation ( Figure 4—figure supplement 2D ) . We then tested the effect of Verteporfin on the maintenance of naïve hPSCs and found that it had a similar detrimental effect on viability . This suggests that YAP-TEAD4 signaling may be important for both the maintenance of naïve hPSCs and their transition into a trophoblast fate . On the other hand , these experiments may also indicate that Verteporfin is simply toxic at the examined concentration . A recent study has pointed to species-specific differences in requirements for YAP-TEAD4 signaling between mouse and human . Unlike in the mouse embryo , where YAP is nuclear in the TE and cytoplasmic in the EPI compartment ( Nishioka et al . , 2009 ) , YAP is nuclear in both the TE and EPI in human blastocysts ( Qin et al . , 2016 ) . We confirmed by immunofluorescence analysis that YAP shows nuclear localization in both naïve hPSCs and hTSCs ( Figure 4—figure supplement 2E ) . Thus , YAP-TEAD4 may contribute to both EPI and TE development in human , potentially by targeting different regulatory elements and transcription factors . In support of this interpretation , we detected ATAC-seq peaks containing TEAD4 transcription factor binding motifs at key pluripotency regulators in naive hESCs , including KLF4 , NANOG , and DPPA2/4 ( Figure 4C; Supplementary file 5 ) . Meanwhile , naïve hTSC-specific DARs containing TEAD4 binding motifs are enriched at loci encoding key trophoblast marker genes such as ITGA2 , KRT7 , and EGFR ( Supplementary file 6 ) , suggesting that TEAD4 is involved in hTSC specification . Hence , the transcription factor binding motifs that become enriched at open chromatin sites during hTSC derivation from naïve hESCs provide a resource for identifying candidate regulators of early trophoblast specification .
In this study , we have demonstrated that naïve , but not primed , hPSCs can give rise to self-renewing hTSCs capable of undergoing further differentiation into specialized trophoblast cell types . Furthermore , by performing RNA-seq , we have shown that naïve hPSC-derived hTSCs - ‘naïve hTSCs’ are comparable with blastocyst-derived hTSCs and share the strongest transcriptional correlation with human TE at day 12 post-fertilization . By performing ATAC-seq analyses on hTSCs for the first time , we established that naïve hTSCs and human blastocyst-derived hTSCs have a similar chromatin accessibility landscape , and identified transcription factor binding motifs at sites that become accessible during hTSC derivation . We believe that this work has three major implications , as outlined below . First , our results give insight into the lineage potential of naïve hPSCs . We previously reported that naïve hPSCs exhibit expression of transcription factors and open chromatin sites associated with the trophoblast lineage ( Pontis et al . , 2019; Theunissen et al . , 2016 ) . Our data now provide functional evidence that naïve hPSCs indeed have enhanced potential to access trophoblast fates during both spontaneous differentiation assays and upon treatment with recently devised conditions for hTSC isolation ( Okae et al . , 2018 ) . While naïve hPSCs can directly give rise to bipotent hTSCs , primed hPSCs instead give rise to cells with a neuroectodermal gene expression profile when exposed to the same signaling milieu . This is reminiscent of the mouse paradigm where CDX2 transgene expression allowed for mouse trophoblast stem cell ( mTSC ) differentiation from naïve mESCs , but not from primed mouse epiblast stem cells ( mEpiSCs ) ( Blij et al . , 2015 ) . We propose that the enhanced , transgene-independent extraembryonic potential of naïve hPSCs is a distinctive functional attribute of naïve human pluripotency , which may reflect the co-expression of embryonic and extraembryonic genes in the human pre-implantation embryo ( Petropoulos et al . , 2016 ) . This conclusion is further corroborated by recent work demonstrating that naïve , but not primed , hPSCs can give rise to human extraembryonic endoderm ( XEN ) cells in response to XEN media ( Linneberg-Agerholm et al . , 2019 ) . It will be instructive to define the exact time window when extraembryonic differentiation potential is acquired or lost during the interconversion between naïve and primed states . The capacity to attain an hTSC-like state does not , however , appear to be restricted to the naïve state . hTSC-like cells have also been obtained from human expanded potential stem cells ( hEPSCs ) ( Gao et al . , 2019 ) , while another study reported that putative hTSCs can be derived from primed hPSCs after transition into a CTB-like state ( Mischler et al . , 2019 ) . It will be important to compare the quality of hTSCs derived from these distinct sources , and define transcriptional and/or epigenomic similarities between the various cell types that are capable of differentiation into hTSCs . Such a comparative analysis may reveal the precise molecular characteristics that impart competence for hTSC derivation . Second , our study provides a cellular model system of early mechanisms governing human trophoblast specification . Previous methodologies for directing human trophoblast differentiation from pluripotent cells have relied on primed hPSCs treated with BMP4 ( Amita et al . , 2013; Horii et al . , 2016; Xu et al . , 2002 ) . However , primed hPSCs are most closely aligned with the late post-implantation epiblast based on scRNA-seq studies ( Nakamura et al . , 2016 ) , when TE and EPI lineages have already segregated . Primed hPSCs also have substantially elevated levels of DNA methylation compared to pre-implantation embryos , naïve hPSCs , or hTSCs ( Okae et al . , 2018; Smith et al . , 2014; Theunissen et al . , 2016; Zhou et al . , 2019 ) . Hence , it has been proposed that putative CTB progenitors derived from primed hPSCs upon BMP4 treatment may actually correspond to an extraembryonic mesoderm identity ( Bernardo et al . , 2011 ) . Recent work from the Parast lab confirms that BMP4 treatment alone induces a mixture of trophoblast and mesoderm fates in primed hPSCs ( Horii et al . , 2019 ) . Our data indicate that naïve hPSCs are unresponsive to BMP4 ( Figure 1C ) , in accordance with recent findings that naïve hPSCs are recalcitrant to direct application of some lineage-inductive cues ( Rostovskaya et al . , 2019 ) , and may in fact suggest that the BMP4-mediated differentiation protocol is primed-state specific . However , unlike primed hPSCs , naïve hPSCs when cultured in hTSC media can directly give rise to hTSCs that share transcriptional signatures with human post-implantation TE ( Figures 2 and 4 ) . We propose that the derivation of hTSCs from naïve hPSCs presents a novel experimental paradigm to identify early determinants of human trophoblast specification . Our analysis of transcription factor binding motifs in open chromatin regions enriched during hTSC derivation provides a blueprint for exploring candidate regulators of this transition . Intriguingly , the TEAD4 binding motif is significantly enriched at the open non-promoter regions specific to naïve and primary hTSCs . This suggests that TEAD4 plays an important role in human TE development . It may also indicate that the HIPPO/YAP signaling pathway , a known regulator of TEAD4 ( Zhao et al . , 2008 ) , is involved in mouse as well as human TE specification ( Nishioka et al . , 2009; Nishioka et al . , 2008; Yagi et al . , 2007 ) . However , unlike in the mouse blastocyst , YAP is also localized to the nucleus in human EPI cells ( Qin et al . , 2016 ) , and we show here that YAP is nuclear in both naïve hPSCs and hTSCs . This points to a possible species-specific difference with HIPPO/YAP/TEAD4 signaling being involved in both TE and EPI development in human . This divergence is not unprecedented: TFAP2C , a trophoblast-specific transcription factor in mouse , was shown to play an essential role in naïve human pluripotency as well ( Pastor et al . , 2018 ) . How HIPPO/YAP/TEAD4 signaling differentially regulates EPI versus TE development in human , and how that differs from the mouse paradigm , will be of interest for future studies . Third , our work provides a robust methodology for a renewable , patient-specific source of hTSCs . Thus far , the isolation of hTSCs has required the use of blastocysts or first-trimester placental tissues ( Okae et al . , 2018 ) , neither of which are readily accessible or amenable to genetic manipulation . In contrast , naïve hPSCs can be conveniently derived from any pre-existing line of hPSCs and can undergo efficient genetic modification ( Yang et al . , 2016 ) . Thus , naïve hPSCs provide a scalable source of cellular material for hTSC derivation . Furthermore , several groups have recently described protocols for direct reprogramming of human somatic cells into naïve pluripotency ( Giulitti et al . , 2019; Kilens et al . , 2018; Liu et al . , 2017; Wang et al . , 2018 ) . The derivation of naïve iPSCs from patient-specific cells and their subsequent differentiation into hTSCs may provide a pathway to uncover the genetic origins of common pathologies afflicting the trophoblast lineage , such as miscarriage , pre-eclampsia , and fetal growth restriction .
The identities of the cell lines used in this study were authenticated using Short Tandem Repeat ( STR ) profiling . The cell culture is regularly tested and negative for mycoplasma contamination . Primed hPSCs were cultured in mTeSR1 ( STEMCELL Technologies , 85850 ) on Matrigel ( Corning , 354277 ) coated wells and passaged using ReLeSR ( STEMCELL Technologies , 05872 ) every 4 to 6 days . Primed hPSCs were cultured in 5% CO2 and 20% O2 . Naive hPSCs were cultured on mitomycin C-inactivated mouse embryonic fibroblast ( MEF ) feeder cells , and were passaged by a brief PBS wash followed by single-cell dissociation using 5 min treatment with TrypLE Express ( Gibco , 12604 ) and centrifugation in fibroblast medium [DMEM ( Millipore Sigma , #SLM-021-B ) supplemented with 10% FBS ( Millipore Sigma , ES-009-B ) , 1X GlutaMAX ( Gibco , 35050 ) , and 1% penicillin-streptomycin ( Gibco , 15140 ) ] . Naive hPSCs were cultured in the 5i/L/A media as previously described ( Theunissen et al . , 2014 ) . 500 mL of 5i/L/A was generated by combining: 240 mL DMEM/F12 ( Gibco , 11320 ) , 240 mL Neurobasal ( Gibco , 21103 ) , 5 mL N2 100X supplement ( Gibco , 17502 ) , 10 mL B27 50X supplement ( Gibco , 17504 ) , 10 μg recombinant human LIF ( made in-house ) , 1X GlutaMAX , 1X MEM NEAA ( Gibco , 11140 ) , 0 . 1 mM β-mercaptoethanol ( Millipore Sigma , 8 . 05740 ) , 1% penicillin-streptomycin , 50 µg/ml BSA Fraction V ( Gibco , 15260 ) , and the following small molecules and cytokines: 1 μM PD0325901 ( Stemgent , 04–0006 ) , 1 μM IM-12 ( Enzo , BML-WN102 ) , 0 . 5 μM SB590885 ( Tocris , 2650 ) , 1 μM WH4-023 ( A Chemtek , H620061 ) , 10 μM Y-27632 ( Stemgent , 04–0012 ) , and 10 ng/mL Activin A ( Peprotech , 120–14 ) . Naïve hPSCs were cultured in 5% O2 , 5% CO2 . For indicated experiments , 1 or 2 μM Verteporfin was added to 5i/L/A . For primed to naïve hPSC conversion , 2 × 105 single primed cells were seeded on a 6-well plate with MEF feeder layer in 2 mL mTeSR1 supplemented with 10 μM Y-27632 . Two days later , medium was switched to 5i/L/A . After 7 to 10 days from seeding , the cells were expanded polyclonally using TrypLE Express on a MEF feeder layer . Tissue culture media were filtered using a 0 . 22 μm filter . Media were changed every 1–2 days . Naïve hPSCs before passage 10 were used for experiments . For naïve hPSC clonal expansion experiments , naïve cells were passaged and seeded at clonal density ( ca . 1 , 000 cells per well of a 6-well plate ) . Single cell clones that exhibit typical dome-shaped morphology were picked and single cell dissociated . They were then seeded and expanded on MEF feeder layer in 5i/L/A as described above . For indicated experiments , naïve hPSCs were cultured in PXGL [N2B27 supplemented with 1 μM PD0325901 , 2 μM XAV939 ( Selleckchem , S1180 ) , 2 μM Gö6983 ( Tocris , 2285 ) , and 10 ng/mL human LIF] as previously described ( Bredenkamp et al . , 2019b ) . 10 μM Y-27632 was added during passaging . Primed hPSCs were first converted to the naïve state using 5i/L/A , then the media was switched to PXGL for eight passages before the cells were used for experiments . Cells were cultured on inactivated MEF feeder layers and in 5% O2 , 5% CO2 . Media were changed every 1–2 days . Capacitation of naïve hPSCs was performed as previously described ( Rostovskaya et al . , 2019 ) . Briefly , around 0 . 5 × 106 naïve hPSCs were seeded on one well of a Geltrex ( Thermo Fisher , A1413201 ) coated 6-well plate . The cells were first cultured in 5i/L/A for 2 days , then cultured in capacitation media [N2B27 supplemented with 2 μM XAV939] for 10 more days . Media were changed every 1–2 days . The cells were passaged once using TrypLE Express during capacitation at a ratio of 1:2 when confluent , usually after 4 or 5 days of culture in capacitation media . 10 μM of Y-27632 was added for 24 hr following passaging . After 10 days of treatment with capacitation media , the cells were used for subsequent analysis . hPSCs were single-cell dissociated using TrypLE , and 3 . 0 × 106 cells were aggregated in EB media [DMEM/F12 supplemented with 20% FBS , 1X GlutaMax , 1X MEM NEAA , 0 . 1 mM β-mercaptoethanol , and 1% penicillin-streptomycin] supplemented with 10 μM Y-27632 in 24-well 800 μm Aggrewell plates ( STEMCELL Technologies , 34811 ) . After 24 hr , the aggregates were flushed from the Aggrewell using EB media , and cultured on ultra-low attachment 6-well plates in EB media only . Media were changed every 2 days . After 12 days of culture , the EBs were collected for downstream analysis . Trophoblast differentiation from hPSCs using BMP4 was performed as previously described ( Horii et al . , 2016 ) . Prior to differentiation , primed or naïve hPSCs were adapted to culture in StemPro ( Thermo Fisher , A1000701 ) medium with 12 ng/mL recombinant basic FGF ( bFGF ) ( Thermo Fisher , PHG0261 ) on Geltrex coated plates . Naïve and capacitated hPSCs were directly subjected to trophoblast differentiation conditions . Briefly , 2 . 0 × 104 to 1 . 0 × 105 primed , re-primed , naïve , or capacitated hPSCs were seeded on Geltrex coated 24-well plates in EMIM medium [DMEM/F12 supplemented with 1X MEM NEAA , 2% BSA ( Sigma-Aldrich , A9418 ) , 2 mM L-Glutamine ( Corning , 25–005 CI ) , 1% ITS ( Gibco , 41400 ) , and 100 ng/mL heparin sulfate ( Stemcell Technologies , 07980] for 2 days . They were cultured in EMIM supplemented with 10 ng/mL human BMP4 ( R and D Systems , 314 BP ) for four more days , at which point they were designated as CTBs . For terminal EVT and STB differentiation , the CTBs were further cultured in FCM [DMEM/F12 supplemented with 1X GlutaMAX , and 1% non-essential amino acid , 0 . 1 mM β-mercaptoethanol; conditioned on irradiated MEFs for 24 hr] supplemented with 10 ng/mL human BMP4 for 6 days . Media were changed every 1–2 days . Cells were cultured in 5% CO2 and 20% O2 . hTSCs were cultured as previously described ( Okae et al . , 2018 ) . Briefly , a 6-well plate was coated with 5 μg/mL Collagen IV ( Corning , 354233 ) at 37°C overnight . Cells were cultured in 2 mL TS medium [DMEM/F12 supplemented with 0 . 1 mM 2-mercaptoethanol , 0 . 2% FBS , 0 . 5% Penicillin-Streptomycin , 0 . 3% BSA , 1% ITS-X ( Gibco , 51500 ) , 1 . 5 μg/ml L-ascorbic acid ( Wako , 013–12061 ) , 50 ng/ml EGF ( Rockland , 009–001 C26 ) , 2 μM CHIR99021 ( Stemgent , 04–0004 ) , 0 . 5 μM A83-01 ( BioVision , 1725 ) , 1 μM SB431542 ( BioVision , 1674 ) , 0 . 8 mM VPA ( Tocris , 2815 ) , and 5 μM Y-27632] and in 5% CO2 and 20% O2 . Media were changed every 2 days , and cells were passaged using TrypLE Express every 3 days at a ratio of 1:4 . Unless otherwise specified , hTSCs between passage 10 and 20 were used for experiments . Naïve and primed hPSCs were single-cell dissociated by TrypLE Express , and 0 . 5–1 . 0 × 106 cells were seeded in a 6-well plated pre-coated with 5 μg/mL Collagen IV and cultured in 2 mL TS medium . Cells were cultured in 5% CO2 and 20% O2 , media was changed every 2 days , and passaged upon 80–100% confluency at a ratio of 1:2 to 1:4 . For indicated experiments , 1 or 2 μM Verteporfin was added to TS medium during the derivation process . For derivation from naïve hPSCs , the cells grew slowly during the initial few passages . Between passage 5 and 10 , highly proliferative hTSCs emerged . In contrast , for derivation from primed hPSCs , the cells did not appear to gain any hTSC characteristic even following 20+ passages . Differentiation of hTSCs into terminal cell types were performed as previously described ( Okae et al . , 2018 ) , with minor modifications . Prior to differentiation , hTSCs were grown to about 80% confluency , and then single-cell dissociated using TrypLE Express . For EVT differentiation , 6-well plates were coated with 1 μg/mL Collagen IV overnight . 0 . 75 × 105 hTSCs were seeded per well in 2 mL EVT basal medium [DMEM/F12 supplemented with 0 . 1 mM β-mercaptoethanol , 0 . 5% penicillin-streptomycin , 0 . 3% BSA , 1% ITS-X , 7 . 5 μM A83-01 , 2 . 5 μM Y27632] supplemented with 4% KSR ( Gibco , 10828 ) and 100 ng/mL NRG1 ( Cell signaling , 5218SC ) . Matrigel was added to a 2% final concentration shortly after resuspending hTSCs in the medium . At day 3 , the media were replaced with 2 mL EVT basal medium supplemented with 4% KSR , and Matrigel was added to a 0 . 5% final concentration . At day 6 , the media were replaced with 2 mL EVT basal medium , and Matrigel was added to a 0 . 5% final concentration . At day 8 , the cells were ready for downstream analysis . For 2D STB differentiation , 6-well plates were coated with 2 . 5 μg/mL Collagen IV overnight . 1 × 105 hTSCs were seeded per well in 2 mL 2D STB medium [DMEM/F12 supplemented with 0 . 1 mM β-mercaptoethanol , 0 . 5% penicillin-streptomycin , 0 . 3% BSA , 1% ITS-X , 2 . 5 μM Y-27632 , 2 μM Forskolin ( Sigma-Aldrich , F3917 ) , and 4% KSR] . The media was changed at day 3 , and at day 6 the cells were ready for downstream analysis . For 3D STB differentiation , 2 . 5 × 105 hTSCs were seeded per well in 3 mL 3D STB medium [DMEM/F12 supplemented with 0 . 1 mM β-mercaptoethanol , 0 . 5% penicillin-streptomycin , 0 . 3% BSA , 1% ITS-X , 2 . 5 μM Y-27632 , 50 ng/ml EGF , 2 μM Forskolin , and 4% KSR] in an ultra-low attachment 6-well plate . At day 3 , another 3 mL of 3D STB medium was added per well . At day 6 , the cells were passed through a 40 μm cell strainer , and the cells remaining on the strainer were collected and used for downstream analysis . Cells were fixed with 4% paraformaldehyde for 20 min at room temperature , then washed with PBS three times . The 3D STBs were resuspended in a small amount of PBS and seeded on plus-charged slides . PBS was allowed to air dry , and immunostaining was performed directly on the slides . For other types of cells , immunostaining was performed directly in the wells . The cells were permeabilized with 0 . 1% Triton X-100 ( Sigma , T8787 ) in PBS for 5 min , then blocked with blocking buffer [PBS supplemented with 0 . 5% BSA and 0 . 1% Triton X-100] for one hour . Cells were then incubated with primary antibodies diluted in the blocking buffer overnight at 4°C . The following primary antibodies were used: anti-KRT7 , 1:100 ( Cell signaling , 4465 ) ; anti-TP63 , 1:20 , ( R and D system , AF1916 ) ; anti-TEAD4 , 1:400 ( Abcam , ab58310 ) ; anti-SDC1 , 1:100 ( Abcam , ab34164 ) ; anti-ZO1 , 1:100 ( Invitrogen , 33–9100 ) ; anti-hCG , 1:100 ( Invitrogen , 14650882 ) ; anti-MMP2 , 1:800 ( Cell Signaling , 40994 ) ; anti-YAP , 1:100 ( Cell signaling , 14074 ) . The cells were washed 3 times in PBS , then incubated with secondary antibodies diluted in blocking buffer for 1 hr at room temperature . The following secondary antibodies were used: anti-mouse-Alexa 488 , 1:500 ( Invitrogen , A-21202 ) ; anti-rabbit-Alexa 647 , 1:500 ( Invitrogen , A-31573 ) ; anti-goat-Alexa 555 , 1:500 ( Invitrogen , A-21432 ) . The nuclei were stained with Hoechst 33258 ( Thermo Fisher , H3569 ) . Cells were washed 3 times in PBS , then imaged with a Leica DMi-8 fluorescence microscope . Some images were globally adjusted for brightness and/or contrast . Cells were single-cell dissociated using TrypLE Express and washed once in FACS buffer [PBS supplemented with 5% FBS] . The cells were then resuspended in 100 μL fresh FACS buffer , and incubated with antibodies for 30 min on ice . The following antibodies were used: anti-SUSD2-PE , 1:100 ( BioLegend , 327406 ) ; anti-CD75-eFluor 660 , 1:100 ( Thermo Fisher , 50-0759-42 ) ; anti-ITG6-FITC , 1:100 ( Miltenyi , 130-097-245 ) ; anti-EGFR-APC , 1:20 ( BioLegend , 352905 ) ; anti-HLA-G-PE , 1:100 ( Abcam , ab24384 ) . Following antibody incubation , the cells were washed once with FACS buffer , resuspended in fresh FACS buffer , and passed through a cell strainer . Unstained cells that have undergone the same procedures were used as controls . Flow cytometry was performed using a BD LSRFortessa X-20 , and the data were analyzed using the FlowJo software . At day 8 of EVT differentiation , the media supernatants which had been in culture for 2 days were collected , and stored at −80°C . As controls , 0 . 75 × 105 hTSCs were seeded per well in 2 mL TS medium . After 2 days of culture , the media supernatants were collected and stored at −80°C . The amount of secreted MMP2 was measured using a commercial MMP2 ELISA kit ( Abcam , ab100606 ) . Each experiment was performed with two biological replicates , and each biological replicate contains two technical replicates . Total RNA was isolated using the E . Z . N . A . total RNA kit I ( Omega , D6834 ) , and cDNA synthesis was performed from total RNA using the high capacity cDNA reverse transcription kit ( Applied Biosystems , 4368814 ) . Real-time PCR was performed using PowerUp SYBR green master mix ( Applied Biosystems , A25743 ) on the StepOnePlus Real-Time PCR System ( Applied Biosystems ) . All analyses were done in triplicate . Gene expression was normalized to RPLP0 . Error bars represent the standard deviation ( SD ) of the mean of triplicate reactions . Student’s t test was performed for statistical analysis . Primer sequences are included in the following primer table: GenePrimer sequence ( 5’- 3’ ) RPLP0-FGCTTCCTGGAGGGTGTCCRPLP0-RGGACTCGTTTGTACCCGTTGKLF17-FCTGCCTGAGCGTGGTATGAGKLF17-RTCATCCGGGAAGGAGTGAGAPax6-FCTTTGCTTGGGAAATCCGAGPax6-RAGCCAGGTTGCGAAGAACTCVIM-FTGTCCAAATCGATGTGGATGTTTCVIM-RTTGTACCATTCTTCTGCCTCCTGSox17-FCGCACGGAATTTGAACAGTASox17-RGGATCAGGGACCTGTCACACCDX2-FTTCACTACAGTCGCTACATCACCCDX2-RTTGATTTTCCTCTCCTTTGCTCTEAD4-FCAGGTGGTGGAGAAAGTTGAGATEAD4-RGTGCTTGAGCTTGTGGATGAAGTFAP2C-FTCTTGGAGGACGAAATGAGATGGTFAP2C-RGGGCTTCTTTGATGTAGTTCTGCELF5-FAGTCTGCACTGACATTTTCTCATCELF5-RCAGAAGTCCTAGGGGCAGTCKRT7-FAGGATGTGGATGCTGCCTACKRT7-RCACCACAGATGTGTCGGAGAGATA3-FTGCAGGAGCAGTATCATGAAGCCTGATA3-RGCATCAAACAACTGTGGCCAGTGAMMP2-FTGGCACCCATTTACACCTACACMMP2-RATGTCAGGAGAGGCCCCATAGAHLA-G-FCAGATACCTGGAGAACGGGAHLA-G-RCAGTATGATCTCCGCAGGGTCGB-FACCCTGGCTGTGGAGAAGGCGB-RATGGACTCGAAGCGCACASDC1-FGCTGACCTTCACACTCCCCASDC1-RCAAAGGTGAAGTCCTGCTCCC In vitro invasion assay was performed in Matrigel-coated transwell inserts with 8 . 0 μm pores ( Corning , 354480 ) . EVTs or hTSCs were single cell dissociated , and seeded at density of 2 × 105 cells per well into the upper chamber of Matrigel-coated transwells in 200 μl EVT basal medium or TS medium . The lower chamber was filled with 800 μl of the same type of medium containing 20% FBS . Cells were cultured at 37°C in 5% CO2 and 20% O2 . After 36 hr , cells on the upper chamber were carefully removed with a cotton swab . The lower chamber was fixed with 4% paraformaldehyde , washed with PBS , and then stained with crystal violet . Invaded cells were imaged on a Leica DMi1 microscope . Thereafter , the stained cells from five random fields were counted to calculate the relative fold change in the number of invading cells . Student’s t test was performed for statistical analysis . Each experiment was performed in triplicate . Total RNA was isolated using the E . Z . N . A . total RNA kit I . Library construction was performed using the SMARTer Directional cDNA Library Construction Kit ( Clontech , 634933 ) . Libraries were sequenced on an Illumina HiSeq3000 1 × 50 platform . RNA-seq reads were aligned to the human genome hg38 with STAR version 2 . 5 . 4b ( Dobin et al . , 2013 ) . Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount ( Liao et al . , 2013 ) , version 1 . 4 . 6 , with hg38 gene annotation ENCODE V27 ( Harrow et al . , 2012 ) . All gene-level transcript counts were then imported into the R/Bioconductor package DESeq2 ( Love et al . , 2014 ) . Transcripts with CPM > 1 . 0 were converted into a DESeq2 dataset and then regularized log transformed using the rlog function from the DESeq2 package . Adjusted p-values for DGE were determined by DESeq2 using the R stats function p . adjust using the Benjamini and Hochberg correction to determine the false discovery rate with a 2- fold expression change and FDR < 0 . 05 required to consider a gene differentially expressed . Principal Component Analysis was performed using plotPCA also from the DESeq2 package and plotted using ggplot2 . The differentially expressed gene lists from pairwise comparisons between conditions were merged along with their RPKMs . RPKMs were z-score scaled for each gene , and k-means clustering was performed using the kmeans function from the R stats package to separate genes into six clusters with options ‘iter . max=1000’ and ‘nstart = 1000’ specified to maximize reproducibility of clustering . Human pre- and post-implantation embryo single cell gene expression data ( transcript per million , TPM ) were downloaded from GEO with accession number GSE109555 ( Zhou et al . , 2019 ) . Gene expressions of single cells in the TE , EPI , and PE lineages were further averaged for days 6 , 8 , 10 , and 12 based on cell identification . The bulk RNA-seq gene expressions data of BT5 hTSC , naïve hTSC , and naïve hPSC were scaled to TPM and combined with the human embryo single cell gene expression data based on the common gene symbols . The expression of TE , EPI , and PE marker genes ( from Zhou et al . supplemental table 5 ) in all the samples was isolated to calculate the Pearson correlation between all sample pairs . The correlation coefficients were plotted in R using gplots heatmap . 2 function . The gene expression ratios of TE and EPI marker genes in naïve hTSC/naïve hPSC and TE/EPI ( days 10 and 12 ) were calculated and plotted in R using gplots heatmap . 2 . ATAC-seq was performed as previously described with minor modifications ( Corces et al . , 2017 ) . Cells were harvested by TrypLE Express dissociation and centrifuged at 500 RCF for 5 min at 4°C . The supernatant was aspirated . Cells were washed once with cold PBS . Cell pellets were then lysed in 100 µL ATAC-seq RSB [10 mM Tris pH 7 . 4 , 10 mM NaCl , 3 mM MgCl2] containing 0 . 1% NP40 , 0 . 1% Tween-20 , and 0 . 01% Digitonin by pipetting up and down and incubating on ice for 3 min . 1 mL of ATAC-seq RSB containing 0 . 1% Tween-20 was added and mixed with the lysis reaction . Nuclei were then pelleted by centrifuging at 800 RCF for 5 min at 4°C . Supernatant was removed , and the nuclear pellets were resuspended in 20 µL 2x TD buffer [20 mM Tris pH 7 . 6 , 10 mM MgCl2 , 20% Dimethyl Formamide] . Nuclei were counted , and 50 , 000 counted nuclei were then transferred to a tube with 2x TD buffer filled up to 25 µL . 25 µL of transposition mix [2 . 5 µL Transposase ( 100 nM final ) ( Illunina , 20034197 , 16 . 5 µL PBS , 0 . 5 µL 1% Digitonin , 0 . 5 µL 10% Tween-20 , 5 µL H2O ) was then added . Transposition reactions were mixed and incubated at 37°C for 30 min with gentle tapping every 10 min . Reactions were cleaned up with the Zymo DNA Clean and Concentrator-5 kit ( Zymo Research , D4014 ) . The ATAC-seq library was prepared by amplifying for nine cycles on a PCR machine . The PCR reaction was purified with Ampure XP beads ( Beckman Coulter , A63880 ) using double size selection following the manufacturer’s protocol , in which 27 . 5 µL beads ( 0 . 55X sample volume ) and 50 µL beads ( 1 . 55X sample volume ) were used based on 50 µL PCR reaction . The ATAC-seq libraries were quantitated by Qubit assays and sequenced on an Illumina NextSeq platform . QC and analysis on ATAC-seq libraries was performed using AIAP ( Liu et al . , 2019 ) . The generated peaks files for each library were merged using bedtools merge , and counts on each peak were quantified for all libraries using bedtools coverage . Differentially accessible region ( DAR ) analysis was performed as described above for DEG analysis on RNA-seq libraries with the following modifications: peaks with an average read density < 5 CPM across all libraries were excluded and a significance cutoff of |L2FC| > 1 and padj < 1×10−3 was required for a region to be considered differentially accessible . DARs were split based on the direction of increased accessibility and annotated using the annotatePeaks . pl function from HOMER ( Heinz et al . , 2010 ) . Annotated DARs were classified as promoter or non-promoter peaks DARs on their distance to the nearest transcription start site ( TSS ) using a cutoff of 2 kb from a TSS . Motif analysis was performed on these DARs using the findMotifGenome . pl function from HOMER with the option ‘-size given’ . The transcription factor ( TF ) motifs were selected if the motif can be identified in at least 10% of input DARs and the matching score of TF motif is > 0 . 9 . The motif occupancy heatmap was plotted in R using pheatmap function . Promoter regions for genes from the RNA-seq k-means clustering were determined by their genomic location in the GENCODE gene annotation file ( hg38 , V27 ) . | The placenta is one of the most important human organs , but it is perhaps the least understood . The first decision the earliest human cells have to make , shortly after the egg is fertilized by a sperm , is whether to become part of the embryo or part of the placenta . This choice happens before a pregnancy even implants into the uterus . The cells that commit to becoming the embryo transform into ‘naïve pluripotent’ cells , capable of becoming any cell in the body . Those that commit to becoming the placenta transform into ‘trophectoderm’ cells , capable of becoming the two types of cell in the placenta . Placental cells either invade into the uterus to anchor the placenta or produce hormones to support the pregnancy . Once a pregnancy implants into the uterus , the naïve pluripotent cells in the embryo become ‘primed’ . This prevents them from becoming cells of the placenta , and it poses a problem for placental research . In 2018 , scientists in Japan reported conditions for growing trophectoderm cells in the laboratory , where they are known as “trophoblast stem cells” . These cells were capable of transforming into specialized placental cells , but needed first to be isolated from the human embryo or placenta itself . Dong et al . now show how to reprogram other pluripotent cells grown in the laboratory to produce trophoblast stem cells . The first step was to reset primed pluripotent cells to put them back into a naïve state . Then , Dong et al . exposed the cells to the same concoction of nutrients and chemicals used in the 2018 study . This fluid triggered a transformation in the naïve pluripotent cells; they started to look like trophoblast stem cells , and they switched on genes normally active in trophectoderm cells . To test whether these cells had the same properties as trophoblast stem cells , Dong et al . gave them chemical signals to see if they could mature into placental cells . The stem cells were able to transform into both types of placental cell , either invading through a three-dimensional gel that mimics the wall of the uterus or making pregnancy hormones . There is a real need for a renewable supply of placental cells in pregnancy research . Animal placentas are not the same as human ones , so it is not possible to learn everything about human pregnancy from animal models . A renewable supply of trophoblast stem cells could aid in studying how the placenta forms and why this process sometimes goes wrong . This could help researchers to better understand miscarriage , pre-eclampsia and other conditions that affect the growth of an unborn baby . In the future , it may even be possible to make custom trophoblast stem cells to study the specific fertility issues of an individual . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2020 | Derivation of trophoblast stem cells from naïve human pluripotent stem cells |
During cell division , progression through mitosis is driven by a protein phosphorylation wave . This wave namely depends on an activation-inactivation cycle of cyclin B-dependent kinase ( Cdk ) 1 while activities of major protein phosphatases , like PP1 and PP2A , appear directly or indirectly repressed by Cdk1 . However , how Cdk1 inactivation is coordinated with reactivation of major phosphatases at mitosis exit still lacks substantial knowledge . We show here that activation of PP2A-B55 , a major mitosis exit phosphatase , required the phosphatase Fcp1 downstream Cdk1 inactivation in human cells . During mitosis exit , Fcp1 bound Greatwall ( Gwl ) , a Cdk1-stimulated kinase that phosphorylates Ensa/ARPP19 and converts these proteins into potent PP2A-B55 inhibitors during mitosis onset , and dephosphorylated it at Cdk1 phosphorylation sites . Fcp1-catalyzed dephosphorylation drastically reduced Gwl kinase activity towards Ensa/ARPP19 promoting PP2A-B55 activation . Thus , Fcp1 coordinates Cdk1 and Gwl inactivation to derepress PP2A-B55 , generating a dephosphorylation switch that drives mitosis progression .
We recently reported a critical , transcription-independent , role for the essential RNA polymerase II-carboxy-terminal domain ( RNAP II-CTD ) phosphatase Fcp1 in Cdk1 inactivation at the end of mitosis ( Visconti et al . , 2012 ) . Indeed , depleting Fcp1 from living HeLa cells as well as from mitotic HeLa cell extracts , that are nuclei-free thus non-transcribing , substantially impaired cyclin B degradation , Cdk1 inactivation and mitosis exit ( Visconti et al . , 2012 ) . In that study , we also noticed that Fcp1 depletion impaired bulk mitotic protein dephosphorylation also upon chemical inhibition of Cdk1 in non-transcribing mitotic cell extracts ( Visconti et al . , 2012 ) . This observation suggested that Fcp1 was required for crucial mitosis exit dephosphorylations even downstream Cdk1 inactivation in a transcription-independent manner . However , bulk dephosphorylation at mitosis exit are likely due to action of major phosphatases like PP1 or PP2A , rather than Fcp1 itself ( Ferrigno et al . , 1993; Qian et al . , 2013 ) . The PP2A-B55 isoform , in particular , has relevant roles for late mitotic events like spindle breakdown , chromatin decondensation , nuclear membrane and Golgi reassembly and cytokinesis ( Schmitz et al . , 2010; Cundell et al . , 2013 ) . In addition , PP2A-B55 has been shown to reverse bulk mitotic phosphorylations detectable by a commercially available anti-Cdk1 substrate antibody , recognizing the K/HpSP motif ( where pS is phosphorylated Ser ) , and the phosphorylation of PRC1 , a crucial cytokinesis protein , at T481 ( Schmitz et al . , 2010; Cundell et al . , 2013; Qian et al . , 2013 ) . In preliminary experiments , in which Fcp1 expression was downregulated in HeLa cells by small interfering RNAs ( siRNAs ) , we found that dephosphorylations of bulk K/HpSP motif and pT481-PRC1 at mitosis exit were indeed dependent on Fcp1 ( Figure 1—figure supplement 1 ) . However , given the role for Fcp1 in inactivation of the spindle assembly checkpoint and of Cdk1 ( Visconti et al . , 2012; Visconti et al . , 2013 ) , delayed dephosphorylations could be due to persistance of Cdk1 kinase activity rather than impaired PP2A-B55 phosphatase activation at the end of mitosis . To know whether Fcp1 controlled PP2A-B55 activation downstream Cdk1 inactivation , we determined whether Fcp1 depletion impaired bulk K/HpSP motif and pT481-PRC1 dephosphorylation upon chemical inhibition of Cdk1 activity in mitotic cells and cell extracts . Control and Fcp1 siRNAs-depleted , as well as Fcp1 depleted complemented with siRNAs-resistant wild type Fcp1 ( Fcp1WT ) expression vector , HeLa cells were arrested at pro-metaphase and further treated with the Cdk1 inhibitor RO-3306 ( Figures 1A , B ) . Nuclei-free , mitotic HeLa cell extracts were , instead , either mock immunodepleted , as control , or immunodepleted of Fcp1 or immunodepleted of Fcp1 and reconstituted with purified , active , Fcp1 wild type ( Fcp1WT ) protein before treatment with RO-3306 ( Figures 1C , D ) . The results showed that , in cells and cell extracts , Fcp1 was indeed required for timely PP2A-B55-dependent dephosphorylations following Cdk1 inactivation ( Figure 1A , C ) . Fcp1 is relatively resistant to the potent PP2A inhibitor okadaic acid ( OA ) ; nevertheless , we confirmed that pT481-PRC1 and bulk K/HpSP motif dephosphorylations were OA sensitive ( Figure 1—figure supplement 2 ) ( Schmitz et al . , 2010; Cundell et al . , 2013 ) . 10 . 7554/eLife . 10399 . 003Figure 1 . Fcp1 affects PP2A-B55-dependent dephosphorylations at mitosis exit . ( A , B ) Control ( Cont . ) or Fcp1-depleted ( Fcp1 ) by siRNAs , as well as Fcp1-depleted complemented with wild type Fcp1 ( Fcp1WT ) , HeLa cells were arrested at prometaphase . ( A ) Cells were collected and split into two samples , one sample received vehicle ( - ) , the other RO3306 ( + ) , and they were further incubated for 15 min , lysed and lysates probed for indicated antigens . ( B ) Cells were lysed and lysates probed for indicated antigens . ( C , D ) Mitotic HeLa cell extracts were mock-depleted ( Cont . dep . ) , Fcp1-depleted ( Fcp1 dep . ) or Fcp1-depleted and reconsituted with Fcp1WT . ( C ) Each set was split into two portions , one received just vehicle ( - ) , the other RO3306 ( + ) , and they were incubated for 30 min at 23°C and probed for indicated antigens . ( D ) Before incubation , extracts samples were also probed for Fcp1 and Cdk1 . The data shown are representative of three independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 00310 . 7554/eLife . 10399 . 004Figure 1—figure supplement 1 . Fcp1-dependency of mitotic exit dephosphorylations . Control ( Cont . ) or Fcp1-depleted ( Fcp1 ) , as well as Fcp1-depleted complemented with wild type Fcp1 ( Fcp1WT ) , HeLa cells were arrested at prometaphase . Lysates from cells released from prometaphase arrest and sampled at indicated time points of incubation were probed for indicated antigens . Fcp1-depletion efficiency is depicted in Figure 1B . The data shown are representative of three independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 00410 . 7554/eLife . 10399 . 005Figure 1—figure supplement 2 . Okadaic acid-sensitive mitotic exit dephosphorylations . Prometaphase-arrested HeLa cells were collected and , immediately after taking the time 0 sample , split into two portions , one portion received vehicle ( Cont . ) and the other 500 nM okadaic acid ( OA ) ; then , samples were taken at indicated time points of further incubation . Cell lysates were separated on SDS/PAGE and probed for indicated antigens . Data shown are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 00510 . 7554/eLife . 10399 . 006Figure 1—figure supplement 3 . PRC1 localization in Fcp1-depleted cells . ( A ) Asynchronously growing control ( Cont . siRNAs ) or Fcp1-siRNAs-treated ( Fcp1 siRNAs ) HeLa cells were fixed and stained for PRC1 ( red ) , α-tubulin ( green ) and DNA ( blue ) . ( B ) Control ( Cont . siRNAs ) or Fcp1-siRNAs-treated ( Fcp1 siRNAs ) HeLa cells , previously arrested at prometaphase , were released into fresh medium in the presence of the proteasome inhibitor MG132 . After 30 min incubation , the Cdk1 inhibitor RO3306 was added , cells were fixed and stained for PRC1 ( red ) , α-tubulin ( green ) and DNA ( blue ) at indicated time points of further incubation . Scale bars , 10 μm . Data shown are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 00610 . 7554/eLife . 10399 . 007Figure 1—figure supplement 4 . Mitotic phenotypes in Fcp1-depleted cells . Asynchronously growing control-siRNAs- ( Cont . siRNAs ) or Fcp1-siRNAs-treated ( Fcp1 siRNAs ) HeLa cells were taken at the indicated time points ( hours , hrs ) from the siRNAs treatment . Upper left panels , cell lysates were separated on SDS/PAGE and probed for indicated antigens including clived caspase 3 ( Clived Casp 3 ) . Lower left panels , cells were fixed and stained for DNA ( blue ) and α-tubulin ( green ) . Rigth graphs , percentage of mitotic , binucleated and multinucleated cells , scored visually from three fields per sample after staining as described for the lower left panels ( about 100 cells per field; cells were scored as mitotic from late prophase to anaphase-telophase ) and of death cells , scored visually by trypan blue inclusion from triplicates per sample . Scale bars , 10 μm . Data shown are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 007 Considering that pT481-PRC1 and bulk K/HpSP motif dephosphorylations at mitosis exit have been shown to depend on PP2A-B55 activity ( Schmitz et al . , 2010; Cundell et al . , 2013 ) , together , our data suggested that PP2A-B55 activation might require Fcp1 downstream Cdk1 inactivation . Whether pT481-PRC1 dephosphorylation is necessary for PRC1 localization and spindle midzone organization is debated ( Cundell et al . , 2013; Hu et al . , 2012 ) . Nevertheless , we found that in Fcp1-depleted cells PRC1 poorly concentrated at spindle midzone , even upon Cdk1 inhibition ( Figure 1—figure supplements 3A , 3B ) , and that Fcp1 depletion induced several mitotic phenotypes in asynchronously growing cells , in addition to cell death , including accumulation of binucleated cells , suggesting that Fcp1 was indeed required also for proficient cytokinesis ( Figure 1—figure supplement 4 ) . During mitosis onset , PP2A-B55 is inhibited by a recently elucidated pathway: Cdk1 phosphorylates and stimulates Gwl kinase that , in turn , represses PP2A-B55 activity by phosphorylating Ensa/ARPP19 and converting these proteins into potent PP2A-B55 inhibitors ( Schmitz et al . , 2010; Castilho et al . , 2009; Vigneron et al . 2009; Lorca and Castro , 2013; Mochida and Hunt , 2012 ) . How this condition is reversed at the end of mitosis is still unclear . The Fcp1 phosphatase has already been called in the mechanism for PP2A-B55 reactivation at mitosis exit ( Hegarat et al . , 2014 ) . Indeed , Fcp1-depleted , mitotic cells were found to maintain Gwl-dependent phosphorylation of Ensa ( at S67 in human Ensa; pS67-Ensa ) upon Cdk1 inhibition , and some evidence was provided that Fcp1 directly dephosphorylated pS67-Ensa to allow PP2A-B55 reactivation ( Hegarat et al . , 2014 ) . However , this evidence was subsequently challenged by Williams and co-workers who showed that PP2A-B55 itself , rather than Fcp1 , dephosphorylated pS67-Ensa and autoactivated , although this slow reaction could be rapidly reversed as long as Gwl stays active ( Williams et al . , 2014 ) . We tested the ability of Fcp1 to dephosphorylate specifically pS67-Ensa in vitro and found that , in agreement with Williams and coworkers conclusions ( Williams et al . , 2014 ) , Fcp1 was unable to do so ( Figure 2A ) . 10 . 7554/eLife . 10399 . 008Figure 2 . Fcp1-dependency of Gwl dephosphorylation at S90 and S453 during mitosis exit . ( A ) Ensa IP from prometaphase-arrested HeLa cells , previously transfected with a Flag-hEnsa expression vector , was divided into four sets and incubated as substrate in phosphatase reactions with either just buffer ( Mk ) or with purified active Fcp1 ( WT ) , or an inactive , catalytic dead , Fcp1 version ( CD ) , or active PP2A for 60 min at 30°C ( lanes 1 , 2 , 3 and 4 , respectively ) . The reactions were then probed for pS67Ensa and Ensa content ( the Fcp1WT and Fcp1CD protein preparations were the same used in the experiment depicted in Figure 3C; see below ) . ( B ) Prometaphase-arrested HeLa cells were released into fresh medium and sampled at indicated time points of further incubation , lysed and processed for mock ( Mk ) or Gwl IP . IPs were separated in parallel SDS/PAGE and probed for indicated antigens . ( C ) Prometaphase-arrested HeLa cells were released into fresh medium and split into two portions; then , immediately after taking the time 0 sample , vehicle ( Control ) or 2 μM okadaic acid ( + OA ) were added . Cells were further sampled at indicated time points of incubation . Upper panels , cell lysates were processed for mock ( Mk ) or Gwl IP that were probed for indicated antigens; lower panels , total cell lysate samples ( Tot . ) were probed for indicated antigens . ( D ) Control ( Cont . ) or Fcp1-depleted ( Fcp1 ) , as well as Fcp1-depleted complemented with wild type Fcp1 ( Fcp1WT ) , HeLa cells were arrested at prometaphase , released and sampled at the indicated time points of incubation . Upper panels , cell lysates were processed for mock ( Mk ) or Gwl IP that were probed for indicated antigens; lower panels , total cell lysate samples ( Tot . ) were probed for indicated antigens . ( E ) Control ( Cont . ) or Fcp1-depleted ( Fcp1 ) , as well as Fcp1-depleted complemented with wild type Fcp1 ( Fcp1WT ) , HeLa cells were arrested at prometaphase and split into two samples , one sample received vehicle ( - ) the other RO3306 ( + ) , further incubated for 15 min . Left panels , cell lysates were processed for mock ( Mk ) or Gwl IP that were probed for indicated antigens; right panels , total cell lysate samples ( Tot . ) were probed for indicated antigens . The levels of Fcp1 depletion and complementation were similar to those shown in Figure 1B . Data shown are representative of at least four independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 00810 . 7554/eLife . 10399 . 009Figure 2—figure supplement 1 . Cdk1-dependent phosphorylations of Gwl . ( A ) Mock- ( Mk ) , V5-GwlWT- , V5-GwlS90A- and V5-GwlS453A-transfected HeLa cells were arrested at prometaphase , lysed and processed for V5 IP . IPs were separated in parallel SDS/PAGE and probed for indicated antigens . ( B ) . Mock- ( Mk ) and V5-GwlWT-transfected HeLa cells were arrested at prometaphase and released . Cells were taken after 120 min of incubation , to allow complete transition into the G1 cell cycle phase , and lysates processed for V5 IP . IPs were split into two portions and incubated in kinase reactions at 37°C for 20 min – or + active Cdk1 ( Cyclin A-Cdk1 , 126 ng per reaction ) , before being separated in parallel SDS/PAGE and probed for indicated antigens . Data shown are representative of at least four independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 00910 . 7554/eLife . 10399 . 010Figure 2—figure supplement 2 . Effect of prolonged Cdk1 inhibition on Gwl dephosphorylation in Fcp1-depleted cells . Control ( Cont . ) or Fcp1-depleted ( Fcp1 ) HeLa cells were arrested at prometaphase , split into two samples , one sample receiving vehicle ( - ) , the other RO3306 ( + ) , and further incubated for 30 min . Cell lysates were processed for mock ( Mk ) or Gwl IP that were then probed for indicated antigens . Data shown are representative of at least four independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 010 To address mechanistically the Fcp1 dependency of PP2A-B55 activation , we hypothesized that Fcp1 was required to inactivate the Ensa/ARPP19 kinase ability of Gwl and allow PP2A-B55 autoactivation . Gwl is higly phosphorylated in mitosis and several observations indicate that Cdk1 directly phosphorylates Gwl stimulating its kinase activity ( Vigneron et al . , 2011; Blake-Hodek et al . , 2012; Dephoure et al . , 2008 ) . Once activated by Cdk1 , Gwl can autophosphorylate and autophosphorylation appears to contribute to its own overall activity ( Blake-Hodek et al . 2012 ) . A study in Xenopus laevis egg extracts has very recently provided compelling evidence that PP1 is the phosphatase that dephosphorylates Gwl at autophosphorylation sites , contributing this way to Gwl inactivation at the end of mitosis . However , the same study showed that dephosphorylation of Gwl at other sites , most likely also those phosphorylated by Cdk1 , is PP1-independent ( Heim et al . , 2015 ) . As Fcp1 is known to be able to reverse Cdk-dependent phosphorylation ( Ghosh et al . , 2008; Visconti et al . , 2012 ) , we hypothesised that Fcp1 was required for Gwl inactivation by removing Cdk1-dependent , activatory , phosphorylations of Gwl at the end of mitosis . First , we established a way to monitor changes at potential Cdk1-dependent Gwl phosphorylation sites during mitosis exit . Two commercially available anti-Cdk1 substrate antibodies , the previously mentioned anti-K/HpSP motif and an anti-pSPXR/K ( where pS is phosphorylated serine and X any aminoacid ) motif , can in principle recognize serine 90 and serine 453 in human Gwl ( S90-Gwl and S453-Gwl ) , respectively , being S90-Gwl ( 89-KSP-91 ) and S453-Gwl ( 453-SPCK-456 ) , the only human Gwl serine residues in those contexts . While S453-Gwl has been shown to be specifically phosphorylated in mitosis by proteomic approaches , S90-Gwl has not ( Dephoure et al . , 2008; Blake-Hodek et al . , 2012 ) . Nevertheless , both antibodies reacted against V5-tagged wild-type Gwl ( V5-GwlWT ) but not against V5-Gwl versions in which serine 90 and serine 453 were respectively mutated into non-phosphorylatable alanine ( V5-GwlS90A; V5-GwlS453A ) when the tagged proteins were isolated from transfected , mitotic , HeLa cells , indicating that both residues are phosphorylated in mitosis ( Figure 2—figure supplement 1A ) . In addition , these antibodies did not react against V5-GwlWT isolated from HeLa cells in G1 , unless it was treated with purified , active , Cdk1 in vitro ( Figure 2—figure supplement 1B ) . To analyse potential changes in Gwl phosphorylation at S90 and S453 during mitosis exit , we probed endogenous Gwl isolated from HeLa cells taken at various time points during mitosis exit with anti-K/HpSP and pSPXR/K antibodies ( Figure 2B ) . PS90- and pS453-Gwl signals were readly detected in prometaphase but were progressively lost as cells transited out of mitosis ( Figure 2B ) . Importantly , dephosphorylation at both sites was resistant to OA at a dose ( 2 μM ) that potently inhibited PP1 ( Figure 2C ) , as indicated by persistance , despite Cdk1 inactivation by cyclin B degradation , of Cdk1-dependent inhibitory phosphorylation of PP1 catalytic subunit a ( PP1cα ) at T320 ( pT320-PP1cα ) , a site that PP1 autodephosphorylates upon Cdk1 inactivation ( Qian et al . , 2013; Heim et al . , 2015 ) ( Figure 2C ) . However , OA significantly delayed the kinetics of Gwl migration shift on SDS/PAGE , in agreement with the notion that also OA-sensitive phosphatase ( s ) dephosphorylate Gwl at several other sites during mitosis exit ( Hegarat et al . , 2014; Williams et al . , 2014; Heim et al . , 2015 ) . Conversely , depletion of Fcp1 delayed Gwl dephosphorylation at both S90 and S453 as well as Gwl migration shift and pS67-Ensa dephosphorylation ( Figure 2D ) . In addition , a 15-min treatment with RO-3306 promptly induced pS90- and pS453-Gwl dephosphorylation and Gwl downshift in prometaphase-arrested control and Fcp1 re-expressing cells but not in prometaphase-arrested Fcp1-depleted cells , indicating that Fcp1 was required for these dephosphorylations downstream Cdk1 inactivation ( Figure 2E ) . Prolonging Cdk1 inhibitor treatment up to 30 min resulted in some Gwl dephosphorylation also in Fcp1 siRNAs-treated cells ( Figure 2—figure supplement 2 ) ; however , we could not rule out whether this was caused by the action of other phosphatases or of residual Fcp1 after substantial time from Cdk1 inactivation . Nevertheless , the 15-min treatment with RO-3306 was able to potently induce autoactivatory pT320-PP1cα dephosphorylation in Fcp1-depleted as in control cells ( Figure 2E ) . Thus , Fcp1 is required for timely dephosphorylation of at least two Cdk1-dependent sites of Gwl , S90 and S453 , at mitosis exit . We set out to determine whether Fcp1 directly dephosphorylated Gwl at S90 and S453 . As Fcp1 can be found in complexes with its substrates ( Visconti et al . , 2015 ) , we first asked whether Fcp1 and Gwl physically interacted during mitosis exit . Indeed , by co-immunoprecipitation ( coIP ) of endogenous proteins , we found that Gwl transiently bound Fcp1 during mitosis exit ( Figure 3A ) . Binding was lower in prometaphase , increased during the period of spindle assembly ( 20-30 min ) to decrease thereafter ( Figure 3A; Fcp1 and Gwl appear to be similarly abundant proteins in HeLa cells and we estimate that approximately 8-14% of Fcp1 interacts with Gwl in cells at the peak of interaction as , routinely , ~3-5% of total lysate Fcp1 was found in Gwl IP that contained ~30-40% of total lysate Gwl ) . In Fcp1-depleted cells complemented with exogenous Fcp1 , Gwl interacted with the exogenous protein with similar kinetics observed with the endogenous Fcp1 ( Figure 3B ) . Transient Gwl-Fcp1 interaction was also detected during mitosis exit in non-transformed , telomerase-immortalized , human retinal pigment epithelium cells hTERT-RPE1 ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 10399 . 011Figure 3 . Fcp1 binds and dephosphorylates Gwl during mitosis exit . ( A ) Prometaphase-arrested HeLa cells were released into fresh medium and sampled at indicated time points of incubation . Upper panels , cell lysates were processed for mock ( Mk ) or Gwl IP that were probed for indicated antigens; lower panels , total cell lysate samples ( Tot . ) were probed for indicated antigens . ( B ) Prometaphase-arrested , Fcp1-siRNAs-transfected and complemented with siRNAs-resistant Fcp1WT ( Fcp1-siRNAs +Fcp1 comp . ) HeLa cells were released into fresh medium and sampled at indicated time points of incubation . Total lysates ( Tot . ) , mock ( Mk ) or Gwl IPs were probed for indicated antigens . Data shown are representative of at least three independent experiments per type . ( C ) Gwl IP from prometaphase-arrested HeLa cell lysates was divided into three sets and incubated in phosphatase reactions with either just buffer ( Cont . ) , Fcp1WT or Fcp1CD proteins ( Mk IP; 1/3 mock IP incubated with buffer ) for 60 min at 30°C . Then , IPs were washed and probed for indicated antigens . Data shown are representative of three independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 01110 . 7554/eLife . 10399 . 012Figure 3—figure supplement 1 . Fcp1-Gwl interaction in hTERT-RPE1 cells . hTERT-RPE1 cells were arrested at prometaphase , collected , released from arrest and sampled at the indicated time points during further incubation in fresh medium . Mock ( Mk ) or Gwl immunoprecipitates ( IPs ) from cell lysates and total cell lysate ( Tot . ) samples were separated in parallel SDS/PAGE and probed for indicated antigens . The data shown are representative of two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 012 Taken together , these data suggest that Fcp1 bound and dephosphorylated Gwl at S90 and S453 , and possibly at other Cdk1-dependent sites , during mitosis exit and that Fcp1-catalyzed dephosphorylation lowered Gwl activity towards Ensa/ARPP19 , allowing PP2A-B55 to autoactivate . To investigate this hypothesis , we first asked whether purified Fcp1 could dephosphorylate Gwl at S90 and S453 in vitro . Indeed , purified active Fcp1 ( Fcp1WT ) , but not an inactive , catalytic dead , Fcp1 version ( Fcp1CD ) , was able to dephosphorylate Gwl isolated from mitotic cells at both sites ( Figure 3C ) . Next , we asked whether Fcp1-dependent dephosphorylation of Gwl in vitro lowered its kinase activity towards Ensa/ARPP19 . To this end , endogenous Gwl , isolated from mitotic cells , was mock-treated or treated with active Fcp1WT or inactive Fcp1CD proteins as described in Figure 3C . Subsequently , Gwl activity was assessed on recombinant X . laevis Ensa protein as substrate ( Figure 4A ) . The results clearly show that pre-treatment of mitotic Gwl with Fcp1WT , but not with Fcp1CD , substantially reduced Gwl ability to phosphorylate S67-Ensa ( Figure 4A ) . Similar results were obtained with V5-GwlWT isolated from transfected , mitotic , cells or using ARPP19 as Gwl substrate ( Figure 4—figure supplement 1A , 1B ) . We cannot exclude that , in addition to S90 and S453 , other Cdk1 phosphorylation sites in Gwl are dephosphorylated by Fcp1; nevertheless , assaying S67-Ensa kinase activity of V5-GwlS90A and V5-GwlS453A mutant proteins , isolated from transfected and prometaphase-arrested HeLa cells , revealed that both mutants had significantly reduced S67-Ensa kinase activity compared to V5-GwlWT ( Figure 4B ) . The S90A mutation of human Gwl showed highest kinase reduction similarly to what reported for a non-phosphorylatable mutant at equivalent residues in X . laevis Gwl ( Blake-Hodek et al . , 2012 ) . Moreover , a non-phosphorylatable Gwl mutant at S452 ( V5-GwlS452A ) , just preceding serine S453 , that is not in a Cdk1 phosphorylation consensus but is found phosphorylated in mitotic cells ( Dephoure et al . , 2008 ) , had S67-Ensa kinase activity similar to GwlWT ( Figure 4—figure supplement 1C ) . In addition , we assessed the S67-Ensa kinase activity of Gwl isolated from the same lysates of the experiment described in Figure 2E in which control , Fcp1-depleted and Fcp1-depleted and complemented prometaphase-arrested cells were treated with the Cdk1 inhibitor for 15 min , and found that while this treatment completely abolished Gwl activity in control cells it had negligible effects in Fcp1-depleted cells ( Figure 4C; see also Figure 2E ) . Like for Gwl dephosphorylation , prolonging Cdk1 inhibitor treatment up to 30 min resulted in a reduction of Ensa kinase activity of Gwl also in Fcp1-depleted cells and , again , we cannot rule out whether this is due to the fact that other phosphatases or residual Fcp1 dampened Gwl activity after substantial time from Cdk1 inactivation ( Figure 4—figure supplement 2 ) . Nevertheless , it is important to remark that the 15-mintreatment with RO-3306 induced full pT320-PP1cα dephosphorylation in Fcp1-depleted cells without significantly affecting Gwl kinase activity towards Ensa ( Figures 2E and 4C ) . 10 . 7554/eLife . 10399 . 013Figure 4 . Fcp1-dependent dephosphorylation reduces Gwl kinase activity towards Ensa . ( A ) Gwl IP from prometaphase-arrested HeLa cell lysates were divided into three sets and incubated for phosphatase reactions with either just buffer ( Cont . ) , Fcp1WT or Fcp1CD proteins ( Mk IP; 1/3 mock IP incubated with buffer ) for 60 min at 30°C . Then , each IP set was split into three portions and incubated in kinase reactions with recombinant Ensa protein . ( B ) V5 IP from lysates of prometaphase-arrested HeLa cells , previously transfected with V5-GwlWT and V5-GwlS90A or V5-GwlS453A , were split into three portions and incubated at 37°C in kinase reactions with recombinant Ensa protein . ( C ) Gwl IPs from the same cell lysates of the experiment described in Figure 2E , in which prometaphase-arrested control ( Cont . ) and Fcp1 siRNAs-treated ( Fcp1 ) , as well as Fcp1 siRNAs-treated complemented with wild type Fcp1 ( Fcp1WT ) , HeLa cells were treated – or + RO3306 for 15 min , split into three portions and incubated in kinase reactions at 37°C and with recombinant Ensa protein . Kinase reactions were stopped at indicated time points of incubation and probed for indicated antigens ( Mk IPs were incubated for 15 min ) . Graphs show quantitation ( arbitrary units ) of pS67-Ensa optical density . Data shown are representative of three independent experiments per type . ( D ) Schematic for the proposed model of PP2A-B55 control . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 01310 . 7554/eLife . 10399 . 014Figure 4—figure supplement 1 . Fcp1 affects Ensa/ARPP19 kinase ability of Gwl . ( A ) V5 IP from V5-GwlWT-transfected , prometaphase-arrested , HeLa cells was divided into three sets and incubated for phosphatase reactions with buffer ( Cont . ) , Fcp1WT or Fcp1CD proteins ( Mk IP; 1/3 mock IP incubated with buffer ) . IPs were washed , each set split into three portions and incubated in kinase reactions at 37°C with recombinant Ensa protein . ( B ) Gwl IP from prometaphase-arrested HeLa cell lysates was divided into three sets . Each set was incubated in phosphatase reactions containinig either just buffer ( Cont . ) , Fcp1WT or Fcp1CD proteins ( Mk IP; 1/3 mock IP incubated with buffer ) . After phosphatase reaction , IPs were washed , each set split into three portions and incubated in kinase reactions at 37°C with recombinant ARPP19 protein . ( C ) V5 IP from lysates of prometaphase-arrested HeLa cells , previously transfected with V5-GwlWT , V5-GwlS453A or V5-GwlS452A , were split into three portions and incubated in kinase reactions at 37°C with recombinant Ensa protein . Kinase reactions were stopped at indicated time points and probed for indicated antigens ( Mk IPs were incubated for 15 min ) . Graphs show quantitation ( arbitrary units ) of pS67-Ensa and of pS62-ARPP19 optical density . Data shown are representative of three independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 01410 . 7554/eLife . 10399 . 015Figure 4—figure supplement 2 . Effects of prolonged Cdk1 inhibition on Gwl activity in Fcp1-depleted cells . Gwl IPs from cell lysates of prometaphase-arrested Fcp1 siRNAs-treated ( Fcp1 ) HeLa cells treated – or + RO3306 for 30 min , the same cell lysates of the experiment described in Figure 2—figure supplement 2 , were each split into three portions and incubated in kinase reactions at 37°C with recombinant Ensa protein . Kinase reactions were stopped at indicated time points of incubation and probed for indicated antigens ( Mk IPs were incubated for 15 min ) . Graphs show quantitation ( arbitrary units ) of pS67-Ensa optical density . Data shown are representative of three independent experiments per type . DOI: http://dx . doi . org/10 . 7554/eLife . 10399 . 015 Together , these data indicate that Fcp1-dependent dephosphorylation greatly reduces S67-Ensa kinase activity of Gwl and that , downstream inactivation of Cdk1 , Fcp1 deficit substantially blunts inactivation of Gwl . We previosly reported that the Fcp1 phosphatase is required to perform dephosphorylations that ultimately bring about Cdk1 inactivation at the end of mitosis ( Visconti et al . , 2012 ) . We report now that , downstream Cdk1 inactivation , Fcp1 dephosphorylates Gwl at sites probably phosphorylated by Cdk1 and downregulates Gwl kinase activity towards Ensa/ARPP19 ( Figure 4D ) . This ultimately leads PP2A-B55 to take the upper hand in dephosphorylating and releasing Ensa/ARPP19 as competitive inhibitors , getting free to dephosphorylate other substrates to complete mitosis ( Williams et al . , 2014 ) . Recently , PP1 has been involved in the mechanism of Gwl inactivation at the end of mitosis by reversing Gwl activatory autophosphorylation ( Heim et al . , 2015 ) . We found that , upon Cdk1 inactivation , Gwl inactivation is strongly delayed if Fcp1 is downregulated , despite potentially active PP1 ( see Figures 2E and 4C ) . As Cdk1-dependent phosphorylation stimulates Gwl activity not only towards Ensa/ARPP19 but also towards Gwl itself at autoactivatory sites ( Blake-Hodek et al . , 2012 ) , by reversing Cdk1-dependent phosphorylation Fcp1 could also reduce Gwl autoactivatory strength , favouring this way PP1 action to stably switch off Gwl autoactivation and , along with directly reducing Gwl activity towards Ensa/ARPP19 , allow PP1 and PP2A-B55 to shut Gwl activity off definitively ( Hegarat et al . , 2014; Williams et al . , 2014; Heim et al . , 2015 ) . Thus , by controlling Cdk1 and Gwl inactivation , Fcp1 appears to be at the apex of a phosphorylation cascade required to exit mitosis ( Figure 4D ) . Along with other recently described phosphatase activation networks ( Lorca and Castro , 2013; Porter et al . , 2013; Nijenhuis et al . , 2014; Grallert et al . , 2015 ) , this pathway contributes to ensure coordinated reversal of mitotic phosphorylations to grant correct completion of mitosis .
HeLa and hTERT-RPE1 cells were grown and maintained as previously described ( Visconti et al . , 2012; Visconti et al . , 2010 ) . Prometaphase-arrested cells were obtained by performing a double thymidine ( 4 mM; Sigma-Aldrich , St . Louis , MO ) block ( 18 hr each , separated by a 6 hr incubation in fresh medium ) followed by release into fresh medium containing nocodazole ( 500 nM; Calbiochem , Billerica , MA ) and incubation for 12 or 14 hr for HeLa and hTERT-RPE1 , respectively . Release from prometaphase arrest was obtained by washing detached cells twice with PBS and twice with fresh medium , followed by incubation in fresh medium . Cells in G1 were obtained after 120 min incubation from prometaphase release . For asynchronous siRNAs treatment , Hela cells were transfected with non-targeting or human Fcp1 3’ UTR-targeting ( 5’-guaagugacagguguuaaa-3’ ) siRNAs ( Dharmacon Inc . , Lafayette , CO ) . For siRNAs treatment and complementation experiments , HeLa cells were first transfected with 3XFlag-Fcp1WT expression vector ( or empty vector for mock transfections; Visconti et al . , 2012 ) . Eight hours post transfection , cells were treated with thymidine ( 4 mM; Sigma-Aldrich ) for 18 hr . Cells were released from thymidine block into fresh medium and transfected with non-targeting or human Fcp1-targeting siRNAs as above . Cells were treated again with thymidine 6 hr after siRNAs transfections , and incubated for further 18 hr . Cells were then released from the second thymidine block into fresh medium containing nocodazole ( 500 nM; Calbiochem ) for 12 hr . Mitotic extracts from prometaphase-arrested HeLa cells ( checkpoint extracts ) were produced , Fcp1-immunodepleted and complementated exactly as previously described ( Visconti et al . , 2012 ) . Recombinat Fcp1WT and Fcp1CD proteins were produced and stored in EXB ( 20 mM HEPES pH 7 . 6 , 5 mM KCl , 1mM DTT , 100 μg/ml 3XFLAG peptide , 10% glycerol; Sigma-Aldrich ) as previously described ( Visconti et al . , 2012 ) . V5-GwlWT expression vector was obtained by subcloning pENTR221-Gwl clone into V5-tagged pcDNA3 . 1 vector ( Invitrogen , Carlsbad , CA ) . To generate the V5-Gwl-S90A mutant , V5-Gwl-S453A mutant and V5-Gwl-S452A mutant , serine 90 , serine 453 and serine 452 of human Gwl were mutagenized into alanine by QuikChange II XL site-directed mutagenesis kit ( Agilent Technologies , Santa Clara , CA ) using the V5-GwlWT expression construct as template . 3XFlag-Fcp1WT or 3XFlag-Fcp1CD expression vectors have been previously described ( Visconti et al . , 2012 ) . Flag-hEnsa expression vector was purchased from Origene ( Rockville , MD ) . Eukaryotic expression vectors transfections were performed using Linear Polyethlenimine ( Polysciences Inc . , Warrington , PA ) . Recombinant 6His-tagged X . laevis Ensa and ARPP19 proteins were expressed in BL21 E . coli cells and purified using Ni-NTA agarose kit ( Qiagen , Germany ) according to the manufacturer’s instructions . The Cdk1 inhibitor RO3306 ( Calbiochem ) was used at 5 and 50 μM in cells and mitotic cell extracts , respectively . Okadaic acid ( Calbiochem ) was used at 500 nM or 2 μM as indicated , MG132 ( Calbiochem ) at 10 μM . Cell viability was assessed by trypan blue ( EuroClone , Italy ) exclusion and immunoblots for cleaved caspase 3 . Anti-phospho-serine Cdk Substrates ( P-S2-100; recognizing K/HpSP ) , anti phosphorylated MAPK/CDK substrates ( recognizing PXpSP or pSPXK/R ) and anti phospho-ENSA/ARPP19 ( pS67/pS62 ) , anti phosphoT320-PP1cα and Cleaved Caspase-3 antibodies were purchased from Cell Signaling Technology ( Danvers , MA ) ; anti-MASTL antibodies from Bethyl Laboratories ( Montgomery , TX ) and NOVUS ( Littleton , CO ) ; anti-Fcp1 antibodies from Bethyl Laboratories and Santa Cruz Biotechnology ( Dallas , TX ) . Other antibodies were from Santa Cruz Biotechnology . Immunoprecipitations and immunoblots were performed as previously described ( Visconti et al . , 2012 ) . For immunofluorescence , cells were grown or spun on microscopy slides , washed in PBS , fixed with 4% formaldehyde in PBS for 10 min and permeabilized with 0 . 2% Triton X-100 in PBS for further 10 min . After blocking with 3% BSA in PBS for 1 hr , samples were incubated with primary antibodies in PBS + 1% BSA for 3 hr . After 3 PBS washes , samples were incubated with secondary antibodies ( Jackson ImmunoResearch Laboratories Inc . , Westgrove , PA ) in PBS + 1% BSA for 1 hr at room temperature . DNA was stained by incubation with Hoechst 33258 ( 10 µg/ml; Santa Cruz Biotechnology ) in PBS . Samples were observed and photographed using an Axiovert 200M inverted microscope equipped with the Apotome slider module with 63X or 40X objectives ( Zeiss , Germany ) . For in vitro dephosphorylation assays , endogenous Gwl IP or V5 IP or Flag IP , from previously V5-GwlWT- or Flag-hEnsa-transfected cells respectively , from 3 ml lysates , of ~1 . 5 mg/ml of protein concentration , of prometaphase-arrested cells were washed in phosphatase assay buffer ( PAB: 20 mM HEPES , pH 7 . 6 , 10 mM MgCl2 , 1 mM dithiothreitol ) , split into three portions , each containing approximately 500 ng of Gwl , and incubated at 30°C for 1 hr in 10 μl of either PAB + 1/10 volume of EXB , as control , PAB + 1/10 volume of Fcp1WT ( 50 ng/μl; final protein conc . ) or PAB + 1/10 volume of Fcp1CD ( 50 ng/μl; final protein conc . ) or PAB + 1/10 volume of purified PP2A ( 0 . 1 unit per reaction; Merck Millipore , Billerica , MA ) . After phosphatase reaction , samples were separated on SDS/PAGE and probed for the indicated antigens or , where indicated , further processed for Gwl kinase activity assays . For Gwl kinase , after phosphatase reactions , each IP set was washed with EB ( 80 mM β-glycerophosphate , 10 mM MgCl2 and 20 mM EGTA ) , divided into three aliquots and incubated for indicated time points in EB buffer supplemented with 1 mM ATP , 10 mM phosphocreatine , 0 . 1 mg/ml creatine phosphokinase ( kinase buffer , KB ) and recombinant X . laevis , Ensa or ARPP19 proteins ( 1 μg per sample ) . One-tenth of each reaction was probed on separate blots for total Ensa or ARPP19 proteins , the remaining was probed for pS67/62-Ensa/ARPP19 and Gwl . Total Ensa was also visualized by re-probing blots previosly probed for pS67-Ensa . For in vitro Gwl rephosphorylation assays , V5 IPs from mock- or V5-GwlWT-transfected , G1-synchronised , HeLa cells were split in two portion and incubated at 37°C for 20 min in KB − or + active Cyclin A2-CDK1 ( 126 ng/reaction , ProQuinase , Germany ) . | Cells multiply through a cell division cycle that has distinct phases . In a phase called mitosis , a cell splits its genetic material , which was duplicated in a preceding phase , into two identical sets . Each of these sets will form the genetic material of daughter cells . If this process goes wrong , then cells can die or become cancerous , and so cells have evolved a complex regulatory process to ensure that mitosis begins and ends at the correct time . For mitosis to begin , an enzyme adds tags called phosphate groups to hundreds of target proteins . These phosphate groups are then removed again to end mitosis . PP2A-B55 is an enzyme that removes these phosphate groups and is needed to complete mitosis , but must remain inactive before this point . This inactivation occurs because a protein called Greatwall activates two other proteins that inhibit PP2A-B55 . To reactivate PP2A-B55 at the end of mitosis , Greatwall must be inactivated , but it was not known how cells do this . Della Monica , Visconti et al . have now investigated this process in human cells . The experiments show that towards the end of mitosis , another enzyme called Fcp1 inactivates Greatwall by removing phosphate groups from it . This allows PP2A-B55 to reactivate . These studies reveal that Fcp1 is a key factor that is needed to complete mitosis . The next challenge is to determine how Fcp1 activity is regulated at the end of mitosis . | [
"Abstract",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"short",
"report",
"cell",
"biology"
] | 2015 | Fcp1 phosphatase controls Greatwall kinase to promote PP2A-B55 activation and mitotic progression |
The surface of insects is coated in cuticular hydrocarbons ( CHCs ) ; variations in the composition of this layer affect a range of traits including adaptation to arid environments and defence against pathogens and toxins . In the African malaria vector , Anopheles gambiae quantitative and qualitative variance in CHC composition have been associated with speciation , ecological habitat and insecticide resistance . Understanding how these modifications arise will inform us of how mosquitoes are responding to climate change and vector control interventions . CHCs are synthesised in sub-epidermal cells called oenocytes that are very difficult to isolate from surrounding tissues . Here we utilise a transgenic line with fluorescent oenocytes to purify these cells for the first time . Comparative transcriptomics revealed the enrichment of biological processes related to long chain fatty acyl-CoA biosynthesis and elongation of mono- , poly-unsaturated and saturated fatty acids and enabled us to delineate , and partially validate , the hydrocarbon biosynthetic pathway in An . gambiae .
The cuticle , also known as the exoskeleton , is the outermost part of the insect body and plays a pivotal role in its physiology and ability to adapt and survive in terrestrial environments . The cuticle consists of multiple layers with different composition and properties . The thickest layer , the procuticle , is divided into the endo- and exo-cuticle , both of which are rich in chitin and cuticular proteins . The outer layer , or epi-cuticle , is mainly composed of lipids and hydrocarbons ( Lockey , 1988 ) . Cuticular hydrocarbons ( CHCs ) are relatively simple molecules but form complex and varied mixtures of n-alkanes , unsaturated hydrocarbons ( alkenes ) , and terminally and internally methyl-branched alkanes/alkenes . These mixtures of CHCs protect insects from desiccation , are the first barrier to infections from microorganisms and can act as mating recognition signals ( pheromones ) ( Blomquist , 2010 ) . The cuticle composition has also been associated with resistance to insecticides , via reduced penetration , in several insect species ( reviewed in Balabanidou et al . , 2018 ) . Anopheles mosquitoes are intensively studied because of their importance as vectors of malaria and lymphatic filariasis that together affect millions of people every year causing intolerable levels of mortality and morbidity . Recently it was shown that populations of the major African malaria vector Anopheles gambiae have developed a thicker cuticle with elevated amounts of hydrocarbons and this is associated with a reduction in the penetration rate of pyrethroid insecticides contributing to the high levels of resistance observed ( Balabanidou et al . , 2016 ) . The emergence of pyrethroid resistance is a major concern for vector control strategies as it threatens the efficiency of the insecticide treated nets , all of which contain this insecticide class , that have proven so successful in reducing the malaria burden in Africa ( Bhatt et al . , 2015 ) . CHCs are also important in conferring desiccation tolerance in An . gambiae , which may be vital in adaptation to arid conditions and survival during the dry season ( Reidenbach et al . , 2014; Arcaz et al . , 2016 ) . Cuticular hydrocarbons are synthesised in oenocytes which are secretory cells of ectodermal origin found in most , if not all , pterygote insects ( Makki et al . , 2014 ) . In adult mosquitoes oenocytes are found in characteristic , predominantly ventral , subcuticular clumps that form rows in each segment , while in larval stages they are located in small groups underneath each of the abdominal appendages ( Lycett et al . , 2006 ) . The biosynthesis of hydrocarbons has been studied using radiolabelled precursors ( Dillwith et al . , 1981 ) and the biochemical steps of their biosynthetic pathway have been established ( Blomquist , 2010; Chung and Carroll , 2015 ) . The pathway starts with a fatty acid synthase ( FAS ) that uses malonyl-CoA to generate a fatty acyl-CoA . In the case of methyl-branched hydrocarbons propionyl-CoA groups ( as methyl-malonyl-CoA ) are also incorporated in the growing fatty acyl-CoA chain . The fatty acyl-CoA chain is further extended by elongases , which extend the chain to different lengths depending on their specificity . Desaturases introduce double bonds , contributing to the generation of unsaturated hydrocarbons , and reductases convert the generated acyl-CoA to aldehydes . These aldehydes serve as substrates for the final step of the pathway , which involves a single carbon chain-shortening conversion to hydrocarbons catalysed by P450 enzymes ( Qiu et al . , 2012 ) . Only this latter step has been delineated in Anopheles mosquitoes with two P450 decarbonylases identified , Cyp4G16 and Cyp4G17 ( Balabanidou et al . , 2016; Kefi et al . , 2019 ) . Only a subset of the large number of lipid metabolic enzymes encoded in the genome are likely to be significant players in CHC synthesis , but we hypothesised that transcripts from these genes will be specifically enriched in oenocytes to enable this function . Here we report the isolation of oenocytes from adult An . gambiae mosquitoes using a transgenic line with fluorescently tagged oenocytes ( Lynd et al . , 2019 ) . RNAseq of the isolated oenocytes identified the key biological processes enriched in these cells and revealed candidate genes for each step of the CHC biosynthetic pathway . A member of the putative pathway was validated by perturbing expression of the AGAP001899 fatty acid synthase ( hereafter called FAS1899 ) . The elucidation of this pathway is a major milestone in delineating the role of variable hydrocarbon composition on key traits that impact vectorial capacity of these important vectors of human disease .
To tag adult An . gambiae oenocytes , we expressed the red fluorescent marker m-cherry specifically in these cells using the GAL4/UAS system ( Lynd and Lycett , 2012 ) . Two transgenic lines were crossed: 1 ) a homozygous UAS-mCD8: mCherry responder line ( Adolfi et al . , 2018 ) with 2 ) a homozygous oenocyte enhancer-GAL4 driver line ( Oeno-Gal4 ) ( Lynd et al . , 2019 ) . Progeny of this cross had the expected m-cherry fluorescent oenocytes throughout their development ( Lynd et al . , 2019 ) . To purify adult oenocytes , mosquitoes were dissected to expose the oenocytes that are dispersed throughout tissues attached to the ventral abdominal integument . Their release was facilitated using trypsin and mechanical homogenisation of the tissue ( Figure 1A ) and subsequent isolation with Fluorescent Activated Cell Sorting ( FACS ) ( Figure 1B ) . Tagged cells corresponded to 1–5% of the total events counted during the FACS sorting and their morphology was consistent with oenocytes by microscopic inspection of sorted cells ( Figure 1—figure supplement 1 ) . Triplicate RNAseq libraries were generated using mRNA from isolated tagged cells and total cell populations ( cell preparation before FACS , referred herein as carcass cells ) from female and male mosquitoes , barcoded and run on the same lane of an Illumina HiSeq sequencer ( CGR University of Liverpool ) . Paired end reads were processed to remove Illumina adapter sequences and low-quality reads . 97 . 12% of reads passed the quality control and generated a total of 425 million reads , of which 58 . 1% ( + / - 0 . 89% standard error ) were successfully mapped to the annotated transcripts of An . gambiae ( Vector Base AgamP4 . 9 ) . To visualise how gene expression varied in the different samples we performed a principal component analysis ( PCA ) using the normalised gene counts of each sample . The first component accounted for 30 . 1% of the variance in gene expression and separated oenocyte from carcass samples , whereas the second component accounted for 25 . 9% of variance and reflected differences between females and males . All three replicates of each condition ( total female carcass cells , total male carcass cells , female oenocytes , male oenocytes ) clustered together ( Figure 1—figure supplement 2 ) providing support for robustness of replication between samples . We next identified transcripts significantly over-expressed [log2 ( Fold Change ) >1 , Benjamini-Hochberg adjusted pvalue <0 . 001 , from a Wald test ) in oenocytes compared to total ( pre-sorted ) carcass cells . Our analysis of differential expression identified 1123 genes over-expressed in male oenocytes compared to male carcass cells and 718 genes over-expressed in female oenocytes compared to female carcass cells . From all over-expressed genes 472 were commonly over-expressed in both female and male oenocytes ( Figure 1C and Supplementary file 1 ) . Gene Ontology enrichment analysis for these 472 genes showed an enrichment in biological processes related to sphingolipid biosynthesis , long chain fatty acyl-CoA biosynthesis and elongation of mono- , poly- unsaturated and saturated fatty acids ( Figure 1C ) , supporting the role of oenocytes in lipid and hydrocarbon biosynthesis . Other biological processes enriched in the oenocyte samples included endocytic recycling , synaptic vesicle coating and docking , and transmission of nerve impulses . Enrichment analysis of Pfam protein domains showed the over-representation of the ELO family that consists of integral membrane proteins involved in the elongation of fatty acids ( Figure 1C ) . We also investigated whether specific gene isoforms are differentially expressed in oenocytes ( at p<0 . 05 , obtained from an empirical cumulative distribution of isoform frequency changes ) . 672 genes had at least one isoform differentially expressed in female oenocytes compared to female total carcass cells and 752 have at least one isoform differentially expressed in male oenocytes compare to male total carcass cells . The same analysis was performed for female and male oenocytes showing 578 genes to have at least one isoform differentially expressed between sexes ( Appendix 1 , Appendix 1—figure 1 ) . We next examined which transcripts from members of the six gene families ( propionyl-CoA synthases , fatty acid synthetases , elongases , desaturases , reductases and P450 decarbonylases ) having roles in the hydrocarbon biosynthetic pathway ( Figure 2 ) are differentially expressed in oenocytes . The two P450s , Cyp4G16 ( AGAP001076 ) and Cyp4G17 ( AGAP000877 ) , that catalyse the last step in the production of cuticular hydrocarbons , plus the P450 reductase ( CPR ) that supplies electrons to all P450 monooxygenation reactions , were among the significantly enriched genes ( Supplementary file 1 ) . Immunolocalisation experiments have previously shown these genes to be highly expressed in An . gambiae oenocytes ( Balabanidou et al . , 2016; Lycett et al . , 2006 ) , lending confidence that our experimental design detects oenocyte enriched genes . The single propionyl-CoA synthase , AGAP001473 , likely responsible for the generation of precursor molecules for the synthesis of methyl-branched hydrocarbons ( Blomquist , 2010 ) was enriched in oenocytes . Of the four remaining gene families , specific members were found to be oenocyte enriched; these consisted of three of the four fatty acid synthases ( AGAP001899 , AGAP08468 , AGAP028049 ) , nine of the 20 elongases ( AGAP013219 , AGAP004372 , AGAP001097 , AGAP003196 , AGAP005512 , AGAP007264 , AGAP013094 , AGAP003195 , AGAP003197 ) , one desaturase ( AGAP003050 out of nine in the genome ) and five of the 17 reductases ( AGAP005986 , AGAP004787 , AGAP005984 , AGAP004784 , AGAP005985 ) ( Figure 3 ) . In addition , the fatty acid transporter AGAP001763 , the ortholog of the Drosophila melanogaster Fatp ( CG7400 ) functionally implicated in CHC biosynthesis , ( Chiang et al . , 2016 ) was also enriched in the An . gambiae oenocyte transcriptome . The majority of these genes were highly expressed in oenocytes ( among the top 200 most highly expressed ) , with Cyp4G16 , Cyp4G17 and FAS1899 ( AGAP001899 ) being in the top ten , followed by the elongase AGAP007264 ( Supplementary files 1 , 2 and 3 ) . Interestingly , several of these genes have highly correlated expression . A meta-analysis of 48 transcriptomic datasets from insecticide resistant and susceptible Anopheles populations ( Ingham et al . , 2018 ) identified 44 transcripts co-regulated with Cyp4G16 , eight of which were predicted to be part of the CHC pathway . All these eight transcripts , with at least one from each of the six gene families , were present in our experimentally determined CHC synthesising candidate gene list ( Figure 2 ) . Notably expression of four genes with a lipid synthesising role was significantly reduced in oenocytes ( depicted on Figure 4 ) . These include the fatty acid synthase AGAP009176 , the desaturase AGAP004572 , the reductase AGAP003606 and the elongase AGAP003600 . Thus , these genes may be involved in the synthesis of Long Chain Fatty Acids ( LCFA ) in other tissues , most likely in the fat body , and not specific to the CHC biosynthetic pathway . We next compared the transcription profile of isolated oenocytes from female versus male mosquitoes . Out of 216 genes that were differentially expressed , 72 were over-expressed in female oenocytes and 144 in male oenocytes . Three genes expressing cuticular proteins ( CPR130 , CPR25 and CPR26 ) were significantly and highly ( log2FC > 3 . 9 ) over-expressed in female oenocytes . However , with the strict criteria we used for the differential expression analysis ( log2FC > 1 , BH adjusted p-value<0 . 001 in all three replicates ) we did not find any gene belonging to the hydrocarbon biosynthetic gene families to be differentially expressed between sexes ( Supplementary file 4 ) . Phylogenetic trees were constructed for the An . gambiae , Ae . aegypti and D . melanogaster gene families of fatty acid synthases , elongases , desaturases and reductases to provide further insights into gene function , in cases where Drosophila orthologs have been characterised , and to identify priority candidates for further study in all three species . The fatty acid synthases AGAP001899 and AGAP009176 cluster closely with three Drosophila FAS genes ( Figure 4A and Figure 4—figure supplement 1 ) , two of which have been shown by in situ hydridisation to be expressed in oenocytes ( Chung et al . , 2014 ) . AGAP001899 is phylogenetically closest to CG17374 ( FASN3 ) known to be expressed in Drosophila oenocytes whereas AGAP009176 , the only An . gambiae FAS down-regulated in oenocytes , is most closely related to CG3523 ( FASN1 ) , which is expressed in the Drosophila fat body . AGAP003050 is the only desaturase enriched in both female and male oenocytes and is a clear 1:1 ortholog of D . melanogaster CG15531 ( Figure 4B and Figure 4—figure supplement 2 ) with a predicted stearoyl-CoA 9-desaturase activity , and AAEL003611 ( also found expressed in Ae . aegypti oenocytes [Martins et al . , 2011] ) . AGAP001713 and AGAP012920 , the paralog of the three Drosophila desaturases Desat1 ( CG5887 ) , Desat2 ( CG5925 ) and Fad 2 ( CG7923 ) involved in the production of unsaturated hydrocarbons ( Dallerac et al . , 2000; Chertemps et al . , 2006 ) , some of which act as pheromones , were not among the oenocyte enriched genes . The elongase family appears to have radiated further after evolutionary separation of Drosophila and mosquitoes . Five out of the nine An . gambiae elongases that are enriched in oenocytes ( Figure 4C and Figure 4—figure supplement 3 ) , ( AGAP001097 , AGAP003195 , AGAP003196 , AGAP003197 , AGAP013219 ) form a cluster of paralogs phylogenetically related to a single Drosophila elongase , CG6660 , a gene over-expressed in adult oenocytes ( Huang et al . , 2019 ) . Two of these paralogs ( AGAP003196 and AGAP013219 ) are closely related to the Ae . aegypti AAEL013542 elongase , which is also expressed in pupae oenocytes ( Martins et al . , 2011 ) . AGAP013094 , another oenocyte enriched elongase is the single An . gambiae gene in a cluster of D . melanogaster paralogs with known functions in CHC biosynthesis , such as eloF ( CG16905 ) ( Chertemps et al . , 2007 ) , CG30008 , CG18609 and CG9458 ( Dembeck et al . , 2015 ) . Contrary to the other gene families , most fatty acid reductases enriched in oenocytes did not have clear orthology relationships with functionally characterised D . melanogaster genes ( Figure 4D and Figure 4—figure supplement 4 ) . For example , AGA005984 , AGA005985 and AGA005986 clustered in a culicine-specific group of paralogs with no Drosophila orthologs . Similarly , most of the fly genes that are functionally linked to CHC profiles ( Dembeck et al . , 2015 ) , such as CG13091 , CG10097 , CG17562 and CG18031 , form a cluster of paralogs with no one-to-one orthologs in An . gambiae . The fatty acid synthase , FAS1899 and the desaturase , Desat3050 , both of which were significantly enriched in oenocytes , were selected for functional validation . We knocked-down their expression through oenocyte specific RNAi and examined the effect on the CHC profile . Firstly UAS-regulated responder lines carrying FAS1899 and Desat3050 hairpin RNAi constructs were established . Crossing the responder lines with the oenocyte specific-Gal4 promoter line ( Oeno-Gal4 ) ( Lynd et al . , 2019 ) resulted in ~80% knock down for the FAS1899 and ~26% for the Desat3050 ( in L2 larvae ) . In both cases oenocyte specific RNAi suppression was lethal at the L2/L3 larvae stages . Subsequently we crossed the two responder lines with the Ubi-A10 Gal4 line ( marked by CFP ) ( Adolfi et al . , 2018 ) which directs widespread tissue expression , but at lower levels in oenocytes compared to the oeno-Gal4 line . The majority of progeny from these crosses expressing dsRNA for FAS1899 and Desat3050 reached the pupae stage , but 70–80% died either as mid to late pupae or during adult emergence ( Figure 5—figure supplement 1 ) . QPCR analysis in whole adults indicated a ~ 26% knock down of FAS1899 transcripts , but no significant difference in Desat3050 knockdown . GC-MS analysis of the hexane extracted hydrocarbons revealed the presence of at least 60 CHC peaks in all samples; 15 of which were alkanes , five unsaturated alkanes and 40 methyl-branched alkanes . While 19 of the CHC peaks had an abundance of ≥1% , the alkanes C29 , C27 and C31 , and the methyl-branched methyl-C31 were consistently among the most abundant accounting for approximately half of the total CHCs ( Figure 5—source data 1 ) . The CHC profile of surviving FAS1899 and Desat3050 knock down adults was compared to control siblings . A significant ( Student’s t-test p-value≤0 . 05 ) 25% reduction in the total amount of hydrocarbons was observed in both female and male FAS1899i mosquitoes . The proportion of the different CHC categories also changed significantly ( Student’s t-test p-value≤0 . 05 ) in the FAS1899i individuals , as the relative abundance of methyl-branched hydrocarbons decreased while the relative abundance of unsaturated and n-alkanes increased ( Figure 5 ) . No difference in the total amount of CHCs , nor of the % of unsaturated CHCs , was observed for the surviving Desat3050 knock down adults ( Figure 5—source data 1 ) .
CHCs affect key traits in Anopheles mosquitoes that determine their fitness and thus vectorial capacity . The difficulties in isolating the CHC synthesising cells in adult mosquitoes , due to their close association with fat body cells within the abdomen , and the absence of clear one to one orthologs with Drosophila in some families ( Figure 4 ) , has hindered the identification of genes involved in mosquito CHC production . In this study we describe the FACS purification of fluorescently tagged oenocytes from adult An . gambiae mosquitoes , and the subsequent transcriptomic analysis of the purified cells which enabled us to identify key candidate genes in the CHC biosynthetic pathway . The samples analysed consisted of total cells recovered from dissected abdomen integument , containing ~12% of tagged oenocyte cells , which were then compared to purified oenocyte cells isolated by passage through the FACS . The abdomen tissue is mainly composed of fat body and epithelial cells , neither of which are expected to synthesise hydrocarbons . Fat bodies do however have a primary role in lipid biosynthesis , which has several steps in common with the CHC biosynthetic pathway , both utilising fatty acid synthases , elongases and desaturases . The analysis pathway was purposively designed to reveal genes and isoforms that are predominantly enriched in oenocytes and thus likely to be involved in CHC biosynthesis but a limitation , in our goal to delineate the entire CHC pathway , is that it will likely fail to detect genes that are expressed at similar levels in fat bodies and oenocytes and are involved in both CHC and lipid biosynthesis ( Wicker-Thomas et al . , 2015 ) . Our data set is the first transcriptomic data for adult mosquito oenocytes . Limited depth transcriptional analysis of larval Ae . aegypti oenocytes that persist during early pupal development , and are relatively easily dissected in pure form due to their distinct large size and loose attachment as clumps of cells to the integument ( Makki et al . , 2014 ) , has previously been performed ( Martins et al . , 2011 ) . Comparison of the partial oenocyte Aedes transcriptome with our adult Anopheles oenocyte data set provides insights into key genes potentially involved in CHC synthesis throughout development . Seven genes involved in lipid biosynthesis were detected in Aedes larval oenocytes , including one acetyl-coA synthetase ( AAEL007283 ) , two elongases ( AAEL008219 and AAEL013542 ) , two desaturases ( AAEL003611 and AAE004278 ) and the two orthologs of Cyp4G16 and Cyp4G17 ( AAEL004054 and AAEL006824 ) . Clear orthologs for five of these larval oenocyte expressed genes ( except for the desaturase AAEL004278 ) were present in our An . gambiae adult oenocyte transcriptome ( Figure 4 ) . Further work to characterise Anopheles oenocyte transcriptomes at earlier life stages will be facilitated by this FACS approach to enable functional analysis of these cells during mosquito development . In addition to genes involved in lipid and hydrocarbon biosynthesis , genes associated with the biological processes of synaptic vesicle coating and docking , and nerve impulse transmission were found enriched in the oenocyte transcriptome . The Oeno-Gal4 driver line used to generate the mosquito population with fluorescent oenocytes has a red fluorescent marker ( dsRed ) under the control of the 3xP3 promoter that drives expression in the eyes and nerve cord . A small contamination of the FACS isolated oenocytes with cells of the nerve cord could be speculated , although nerve cells were not observed when visually observing the isolated cells with confocal microscopy . Moreover , oenocytes have been reported to play a role in the neuronal processes during D . melanogaster embryogenesis through the secretion of semaphorin ( Sema2a ) , a peptide that drives axon elongation; ablation of oenocytes results in sensory axon defects similar to the sema2a mutant phenotype ( Bates and Whitington , 2007 ) . In addition , the development of oenocytes and of sensory organ precursors ( SOPs ) in the peripheral nervous system is intimately linked . A previous study has shown that in D . melanogaster embryos primary SOPs signal , via the EGFR pathway , to the overlying ectoderm . This results in the differentiation of signal receiving cells into oenocytes , in the presence of the Sal transcription factor , or into secondary SOPs in the absence of Sal ( Rusten et al . , 2001 ) . Thus , oenocytes likely have a variety of currently unexplored functions , which is also supported by our gene enrichment analysis . A further interesting aspect is their potential role in lipid synthesis , processing and secretion ( signalling ) , as reviewed in Makki et al . , 2014 and suggested by the enrichment in our dataset of sphingolipid and fatty acid biosynthesis and endocytic recycling . Further studies are needed to explore these functions in more detail and to investigate the potential cross talk ( regulation ) of oenocytes with other tissues . For example D . melanogaster larval oenocytes are thought to produce a VLCFA dependent signal that controls remotely the water-tightness of the respiratory system ( Parvy et al . , 2012 ) . Thus , oenocyte regulated lipid signalling under normal and stressful developmental conditions would be an interesting area of future research . We functionally validated the role of the fatty acid synthase FAS1899 in CHC biosynthesis , by stably knocking down its expression during mosquito development . Oenocyte specific knock down of FAS1899 was lethal at the L2/L3 larvae stages , showing its important role for the normal mosquito development , possibly by synthesising Very Long Chain Fatty Acids ( VLCFA ) that are utilised at the larvae stage either for waterproofing the respiratory system ( Parvy et al . , 2012 ) or for other metabolic purposes . Lethality was also reported for the RNAi-mediated knock down of its ortholog ( CG17374 ) in D . melanogaster before adult eclosion ( Chung et al . , 2014 ) . Silencing of the FAS1899 expression using the polyubiquitin ( Ubi ) promoter also resulted in high levels of mortality ( 70–80% ) , but this time at the pupae stage and during adult emergence . This milder phenotype could be explained by the fact that the Ubi promoter drives lower levels of expression in oenocytes , which is supported also by the quantitative real time PCR data ( 26% knock down of the FAS1899 in adult progeny of the UAS-FAS1899i x Ubi-A10 Gal4 cross compared to the 70% of knock down seen in L2/L3 progeny of the UAS-FAS1899i x Oeno-Gal4 cross ) . The relative expression levels of FAS1899 affect both the quantity and composition of CHCs produced in adult oenocytes . A 25% reduction in the total amount of hydrocarbons was observed for adults surviving knock down of FAS1899 and the CHC profile showed a decrease in the total proportion of methyl branched CHCs and an increase in saturated and un-saturated straight-chain hydrocarbons . Silencing Cyp4G16 or Cyp4G17 transcript levels in An . gambiae oenocytes by approximately 90% resulted in high mortality in late pupae , pharate adults and during adult emergence and , in surviving adults , a 50% reduction in the total amount of CHCs was observed ( Lynd et al . , 2019 ) . The Cyp4G16 and Cyp4G17 P450s catalyse the final decarbonylation step in the cuticular hydrocarbon synthetic pathway , while FAS1899 is thought to catalyse the first step using acetyl-CoA to generate and elongate a fatty acyl-CoA chain . Thus , perturbing both extremes of the pathway can influence the final amount of synthesised hydrocarbons . Partial knock down of the Desat3050 transcripts in larval oenocytes was correlated with larval lethality , similar to FAS1899 knock down . High levels of mortality were also observed when using the weaker oenocyte line ( but more widespread driver line ) . However , no qualitative or quantitative differences in the CHC profile were observed in surviving adults . Further work is required , but it may indicate that Desat3050 catalyses the formation of unsaturated lipids that are not converted to hydrocarbons but are important in development , such that even a slight perturbance in the expression levels of this gene can have severe developmental effect . In D . melanogaster , either deletion or strong over-expression of the desaturase , desat1 , result in larval mortality , showing that correct regulation of this gene is critical for development ( Köhler et al . , 2009 ) . In addition desat1 was shown to affect not only the biosynthesis of unsaturated lipids , but also the availability of saturated lipids , as a reduction in its activity results in decreased amounts of both unsaturated and saturated fatty acids ( Ueyama et al . , 2005 ) . Thus , a perturbance in the function of a desaturase enzyme can have a broader effect on lipid metabolism potentially leading to developmental abnormalities or lethality . Variations in the relative abundance of CHCs on the cuticular surface have been correlated in Anopheles mosquitoes with species , karyotype , age and mating status ( Caputo et al . , 2005; Polerstock et al . , 2002 ) . Sex specific differences in the relative abundance of some CHC compounds have also been reported in An . gambiae ( Caputo et al . , 2005 ) , but in contrast to other insects like Drosophila melanogaster ( Coyne and Oyama , 1995 ) , sexual dimorphism in CHCs in mosquitoes has not been reported . This lack of sex specificity is reflected in the absence of sex specific expression of CHC synthesising genes in our analysis . However , interestingly we did identify some splice variants of Cyp4G16 , encoding for a different C-terminus , to be differentially expressed between male and female oenocytes , but further work is needed to validate this observation . A change in C- terminus is likely to alter the intracellular location of proteins through removal of the ER retention signal . Previous work on females has demonstrated enriched localisation of CYP4G16 on the oenocyte plasma membrane surface ( Balabanidou et al . , 2016 ) . It would be interesting to examine males in comparison . Variation in the abundance of CHCs has been associated in An . coluzzii with insecticide resistance; a 30% increase in CHC content has been correlated with a decrease in the penetration rate of pyrethroid insecticides ( Balabanidou et al . , 2016 ) . Several of the genes implicated in CHC biosynthesis from the results of the current study are expressed at elevated levels in pyrethroid resistant mosquitoes and may provide useful genetic markers for detecting this emerging resistance phenotype . For example FAS1899 is a member of the Cyp4G16 correlation network and is over-expressed in pyrethroid insecticide resistant An . gambiae and An . coluzzii populations from Burkina Faso and Côte d’Ivoire ( data from the IR-TEx web-based application [Ingham et al . , 2018] ) . Thus , this gene could be implicated in cuticular resistance , through the production of a thicker cuticle with more hydrocarbons . In addition to insecticide exposure , environmental factors can also select for changes in the CHC profile; relative proportions of unsaturated and methyl-branched CHCs altered following exposure to arid conditions in the insectary ( Reidenbach et al . , 2014 ) and these arid conditions were also associated with an enrichment of genes involved in lipid biosynthesis , including six elongases ( Cheng et al . , 2018 ) , four of which overlap with the oenocyte enriched elongases identified in this study . The pleiotropic effect of alterations in CHC composition has important implications . Selection pressures that alter the CHC composition , for example the extensive use of insecticides , or an increase in aridity due to climate change , could have multiple effects on mosquito fitness and impacts on disease transmission . Investigating how the different traits influence one another and how this is regulated by the CHC composition is a key next step to understand how mosquitoes adapt and survive in a changing environment and in response to disease control interventions .
An . gambiae mosquitoes were reared at 28°C under 80% humidity and at a 12/12 h day/night cycle . Larvae were fed with fish food ( TetraMin , Tetra GmbH ) , and adult mosquitoes were fed ad libitum with 10% sugar . To generate mosquitoes with fluorescent oenocytes we crossed males from the UAS-mCD8: mCherry responder line ( Adolfi et al . , 2018 ) with virgin females of the oeno-Gal4 driver line ( Lynd et al . , 2019 ) . Adult progeny ( 2–4 days old ) were collected , anesthetised on ice and dissected in 1X PBS . The head , thorax and internal tissues ( midgut , malpigian tubules and reproductive tissues ) were removed and the remaining integument ( carcass ) was cut open . Each sample ( N = 12 in total , Supplementary file 5 ) consisted of 30 carcasses . Samples were washed twice with 1X PBS and incubated for 30 min at 37°C with 0 , 25% trypsin in 1X PBS . After incubation tissues were washed twice with 1X PBS and homogenised by pipetting up and down in 1X PBS containing 1% fetal bovine serum . Dissociated cells were filtered through a plastic filter mesh ( ThermoFisher 70 µm Nylon Mesh ) . For samples used to isolate oenocytes ( N = 6 ) , cells were immediately used for FACS sorting . In the case of total carcass cells ( N = 6 ) total RNA was extracted after filtering using the Arcturus PicoPure RNA extraction kit . For oenocyte isolation the BD ARIA III Cell Sorter ( BD Biosciences ) equipped with lasers at 405 and 561 nm was used . Cells were gated based on the m-Cherry fluorescence . A sample of cells from wild type G3 mosquitoes with no fluorescence was used as control to define the threshold of fluorescence for isolation . All samples were acquired in Facsdiva software version 8 . 1 ( BD Biosciences ) . All debris doublets were removed from the analysis . The purity of isolation was initially assessed by visualisation of isolated cells using a Zeiss LSM 880 confocal microscope . Oenocytes were directly sorted in the extraction buffer of the Arcturus PicoPure RNA extraction kit . Total RNA was extracted based on the manufacturer’s instructions , including treatment with DNAse . Generation and amplification ( 11 cycles ) of c-DNA from all samples was done in the Center for Genome Research ( University of Liverpool ) using the SMART-Seq v4 Ultra Low Input RNA Kit , according to manufacturer’s instructions . The cDNA samples were purified using AMPure XP beads ( Beckman Coulter ) and their concentration and quality determined using the Agilent 2100 Bioanalyzer and Agilent’s High Sensitivity DNA Kit . Libraries were constructed with a total of 1 ng of Smarter amplified material and amplified using 12 cycles of PCR . Quality control was performed by running 1 µl undiluted library on an Agilent Technology 2100 Bioanalyzer ( RRID:SCR_018043 ) using a High Sensitivity DNA kit . Samples were run on a Illumina HiSeq 4000 ( RRID:SCR_016386 ) . Illumina adapter sequences were removed from the read files ( 24 fastq files in total: 12 RNA-seq runs with right and left reads ) using cutadapt 1 . 2 . 1 ( Martin , 2011 ) ( flag -O 3 ) . Low-quality reads were removed using Sickle 1 . 200 ( minimum window quality score of Phread = 20 , removing reads shorter than 20 bp ) ( Joshi and Fass , 2011 ) , retaining only read pairs in which both left and right reads passed quality filters . These steps were performed by the Liverpool University CGR sequencing facility . Each read file was analysed with fastqc 0 . 11 . 5 ( Andrews , 2014 ) to confirm the absence of adapters sequences . Overall , 97 . 12% of reads passed the quality control process ( Supplementary file 5 ) . The reference gene annotation and assembly of An . gambiae was obtained from VectorBase ( Giraldo-Calderón et al . , 2015 ) ( GFF and FASTA formats , version AgamP4 . 9 ) . We obtained the predicted peptides of each gene using gffread ( Geo , 2019 ) . Then , we annotated their Gene Ontology functional annotations using eggNOG emapper 1 . 0 . 3 ( Huerta-Cepas et al . , 2017 ) ( HMM mode , which uses hmmscan from HMMER 3 . 2 . 1 ( HMMER 2015 ) ) with the euNOG database of eukaryotic protein annotations ( Huerta-Cepas et al . , 2016 ) ( eggNOG version 4 . 5 ) as a reference . In parallel , we annotated the protein domains using Pfamscan , based on version 31 of the Pfam database ( Punta et al . , 2012 ) . We quantified gene expression using the trimmed , clean reads . Specifically , we used Salmon 0 . 10 . 2 ( Patro et al . , 2017 ) to build an index of transcripts ( salmon index; using the longest isoform per gene as a reference ) , using the quasi-mapping procedure ( --type quasi flag ) and k-mers of length 31 ( -k 31 ) ; and then quantified transcript abundance ( salmon quant ) in each sample using the paired-end read files ( using automated library type inference , -l A flag ) , in order to obtain TPM ( transcripts per million ) values for each gene . Then , we performed a differential expression analysis between sample groups ( female oenocytes vs female carcass cells , male oenocytes vs male carcass cells and female oenocytes vs male oenocytes ) using the R DESeq2 library 1 . 24 . 0 ( Love et al . , 2014 ) . First , we imported the transcript quantification values from Salmon ( see above ) using the tximport library 1 . 12 . 0 ( Soneson et al . , 2015 ) . Then , we performed targeted differential expression analyses between groups of samples using the DESeq function from DESeq2 ( using the Wald procedure for significance testing ) , produced a table of normalised gene counts per sample using the counts function ( using DESeq2 normalisation factors ) , and obtained the fold changes and p-values from a Wald test for each gene , using the results command ( using a Benjamini-Hochberg [FDR] p-value correction ( Benjamini and Hochberg , 1995 ) and an alpha threshold = 0 . 001 , and all combinations of samples from Supplementary Table 1 to define the contrast parameter ) . The log-fold change values were corrected ( shrunken ) with lfcShrink and the apeglm algorithm ( Zhu et al . , 2019 ) . We defined a gene as being differentially expressed in a given comparison if the adjusted p<0 . 001 and the absolute shrunken log-fold change >1 ( i . e . absolute fold change >2 ) . We explored the variation in gene expression across samples using the normalised gene counts ( log-transformed , and standardised to mean = 0 and standard deviation = 1 using the scale R function ) . First , we performed a Principal Components analysis ( PCA ) using the normalised gene counts of each sample ( prcomp function of the R stats library ) . To visualise changes in expression for genes involved in CHC biosynthesis , we produced heatmaps of gene expression by plotting the normalised gene counts of each gene in each sample ( pheatmap function from the pheatmap 1 . 012 R library ( Kolde 2019 ) , using Pearson correlation values to set the order of genes ) . We used SUPPPA2 ( Trincado et al . , 2018 ) to generate a set of alternative splicing events from the annotated isoforms in the An . gambiae genome ( GFF file from Vectorbase , AgamP4 . 9 ) , using the generateEvents mode to detect retained introns , skipped exons , and alternative first or last exons , and mutually exclusive exons ( -e SE MX RI SS FL ) , with 10 bp as the minimum exon length ( -l 10 ) . We also calculated the expression levels at the isoform level using Salmon 0 . 10 . 2 ( Patro et al . , 2017 ) ( output in TPM ) . Then , we used SUPPA2 psiPerIsoform mode to calculate the inclusion rates of each isoform ( PSI: percentage spliced-in ) in each sample , using the expression levels of each isoform ( obtained from Salmon ) as a reference . Differential splicing was quantified by calculating the calculating the average difference in PSI values between each sample group ( male/female oenocytes and carcasses ) , and p-values were obtained using the empirical significance calculation method described in SUPPA2 ( Trincado et al . , 2018 ) . The PSI values of selected differentially spliced genes ( p<0 . 05 ) belonging to the biosynthesis pathway were reported using a heatmap table ( pheatmap function from the pheatmap 1 . 012 R library ) . Gene Ontology enrichments based on the GOs annotated with eggNOG mapper ( see above ) were computed using the topGO R library ( 2 . 34 ) ( Alexa and Rahnenfuhrer , 2020 ) . Specifically , we computed the functional enrichments based on the counts of genes belonging to the group of interest relative to all annotated genes , using Fisher’s exact test and the elim algorithm for GO graph weighting ( Alexa et al . , 2006 ) . Functional enrichment tests of Pfam domain annotations were performed using hypergeometric tests as implemented in the R stats 3 . 6 library ( phyper ) ( R Development Core Team , 2017 ) , comparing the frequencies of presence of Pfam domains in a list of genes of interest to the same frequencies in the whole gene set ( using unique domains per gene ) . We adjusted p values using the Benjamini-Hochberg procedure . We retrieved genes belonging to gene family-members of the fatty acid biosynthesis pathway from the proteomes of An . gambiae ( Vectorbase , AgamP4 . 9 annotation ) , Ae . aegypti ( Vectorbase LVP_AGWG AaegL5 . 1 annotation ) and D . melanogaster ( Flybase r6 . 21 annotation ) . Specifically , we defined the list of candidate genes for phylogenetic analysis according to the presence of the following catalytic Pfam domains: FA_desaturase ( PF00487 ) for desaturases ( totalling 29 individual domains ) , ELO ( PF01151 ) for elongases ( 62 ) , NAD_binding_4 ( PF07993 ) for reductases ( 61 ) , ketoacyl-synt ( PF00109 ) for synthases ( 16 ) . Pfam annotations were obtained from Pfamscan as described above . Functional domain sequence sets were aligned using MAFFT 7 . 310 ( 1 , 000 rounds of iterative refinement , L-INS-i algorithm ) ( Katoh and Standley , 2013 ) , and later trimmed position-wise using trimAL 1 . 4 ( automated1 procedure ) ( Capella-Gutiérrez et al . , 2009 ) . The trimmed alignments were used to build maximum-likelihood phylogenetic trees for each gene family , using IQ-TREE 1 . 6 . 10 ( Nguyen et al . , 2015 ) . The best-fitting evolutionary model ( LG substitution matrix [Le and Gascuel , 2008] with four Γ categories and accounting for invariant sites , or LG+I+G4 ) was selected for each gene family according to the BIC criterion . Phylogenetic statistical supports were calculated using the UF bootstrap procedure ( 1000 replicates ) ( Hoang et al . , 2018 ) . The resulting phylogenetic trees were mid-point rooted using the R phangorn 2 . 53 library ( Schliep , 2011 ) , and visualisations were produced using the phytools 0 . 6–60 ( Revell , 2012 ) and ape 5 . 3 libraries ( plot . phylo ) ( Paradis and Schliep , 2019 ) . A UAS responder plasmid was generated for the expression of dsRNA targeting the third exon of the AGAP001899 gene and the first exon of the AGAP003050 gene . Specifically 200 bp inverted repeats separated by the 203 bp fourth intron of the Drosophila melanogaster white eye gene ( CG2759 ) were synthesised by GeneScript and cloned into the YFP-marked responder plasmid pSL*attB:YFP:Gyp:UAS14i:Gyp:attB ( Lynd et al . , 2019 ) downstream of the UAS using EcoRI/NheI restriction enzymes . The intron of the Drosophila white eye gene was used because all internal introns of the AGAP001899 gene were shorter than 100 bp , as well as the first intron of AGAP003050 , making the synthesis of the 200 bp inverted repeats impossible . Embryo injections were performed using the A11 docking line ( Lynd et al . , 2019 ) , which carries two inverted attP sites and is marked with 3xP3-driven CFP . 350 ng/μL of the responder plasmid and 150 ng/μL of the integrase helper plasmid pKC40 encoding the phiC31 integrase ( Ringrose , 2009 ) were injected as described in Pondeville et al . , 2014 . Emerging F0 individuals were outcrossed with wild type G3 individuals of the opposite sex . The F1 generation was screened for the expression of the YFP marker in the eyes and nerve cord and the absence of the CFP marker , indicating the successful cassette exchange . The direction of the cassette exchange was determined as described in Adolfi et al . , 2019 and shown to be of the A orientation . The FAS1899 RNAi and Desat3050 RNAi responder lines that were established were kept as a mix of homozygous and heterozygous individuals so as to obtain Gal4/+ progeny after crossing with the Gal4 driver lines and obtain siblings that serve as transgenic blank controls . Crosses were performed between the responder lines UAS-FAS1899i , UAS-Desat3050i and the two Gal4 lines: oenocyte specific-GAL4 ( Oeno-Gal4 ) ( Lynd et al . , 2019 ) and Ubi-A10 Gal4 line ( Adolfi et al . , 2018 ) . Progeny ( at least 10 individuals for each group , pooled in 2–3 biological replicates ) of these crosses , YFP marked and blank ( control siblings ) , were collected either at the L2-L3 stage ( for the cross with the oeno-Gal4 line ) or at the adult stage ( for the cross with the Ubi-A10 line ) and used to extract RNA with the PicoPure RNA isolation kit ( Thermo Fisher Scientific ) and treated with DNase using the Qiagen RNase-free DNase kit . 2ugr of RNA were reverse transcribed using SuperScript III ( Invitrogen ) and oligo ( dT ) 20 primers to produce cDNA . Expression of AGAP001899 ( FAS1899 ) and AGAP003050 ( Desat3050 ) was validated by qPCR using the following primers: ( FAS1899 Forward: 5’-AGCGATCTGCGTGATGTACC-3’ , FAS1899 Reverse: 5’-GCCTTCCTCCTTAAACCCGTC-3’ , Desat3050 Forward: 5’ CCGTACTACAGCGACAAGGAC-3’ , Desat3050 Reverse 5’- GAACATCACAATACCGTCCGC-3’ ) and reference gene for normalisation the Ribosomal S7 ( AGAP010592 ) ( Forward: 5’-AGAACCAGCAGACCACCATC-3’ Reverse: 5’-GCTGCAAACTTCGGCTATTC-3’ ) . Expression analysis was performed according to Pfaffl , 2001 . CHCs were extracted from pools of adult ( 3–5 days old ) mosquitoes ( each pool consisted of 2–5 mosquitoes depending on availability , at least three pools per condition ) by immersing them and gently agitating them , for 10 min at room temperature , in 200 μl of hexane ( Sigma-Aldrich ) spiked with 1 ng/ml of octadecane ( Sigma-Aldrich ) as internal standard . Hexane extracts were concentrated under a N2 stream and 2 μl injected in a Waters GCT gas chromatograph-mass spectrometer . The GC column was a 30 m long , 0 . 25 mm internal diameter , 0 . 25 μm film thickness BPX5 ( SGE ) . The oven temperature gradient was 50°C to 370°C at 10 °C/minute and the carrier gas was helium ( BOC ) at a flow rate of 1 ml/minute . The scan range was m/z 40 to 450 Da in scan time 0 . 9 s . Compounds were identified based on their mass spectra in comparison to those of an alkane standard mixture ( C10-C40 , Merck 68281–2 ML-F ) , by comparison of their retention times and fragmentation patterns to published Anopheles gambiae CHC mass spectra ( Balabanidou et al . , 2016 ) and searches of the NIST mass spectrum library supplied with Waters MassLynx software ( RRID:SCR_014271 ) . Peak areas were measured manually using the peak integration tool in the Waters MassLynx software . The total amount of hydrocarbon present was calculated by summing all the peak areas measured relative to the area of the internal standard . Student’s t-test was performed for the statistical analysis of differences in total CHC amount and relative abundance of CHC categories . Transcriptome sequencing has been deposited in the European Nucleotide Archive ( ENA ) , under PRJEB37240 project . All transgenic lines produced in this study will be provided by L . G upon request . All data and code ( in R ) required to perform the differential expression , alternative splicing and phylogenetic analyses in this paper is available in the following Github repository: https://github . com/xgrau/oenocytes-agam ( Grau-Bové , 2020; copy archived at https://github . com/elifesciences-publications/oenocytes-agam ) . | The bodies of insects are encased in an exoskeleton or cuticle that is key for their survival . The cuticle helps protect insects against damage , prevents water loss and can defend against pesticides . A better understanding of the role of the cuticle for survival in mosquitoes and other insects could lead to new ways to prevent the spread of diseases such as malaria . The cuticle is coated with various molecules from a group of chemicals called hydrocarbons . This coating is made by specialized cells called oenocytes and helps to protect insects . Hydrocarbons can also influence communications between certain insects by acting as recognition signals . In mosquitoes , oenocytes make several hydrocarbons using a set of processes that are not well understood , and the types of hydrocarbons they make can vary between individuals of the same species . It is unclear how this mixture of hydrocarbons is generated and how differences in the mixture can determine how mosquitoes adapt to their surroundings . Grigoraki et al . studied the genes that were active in isolated oenocytes from the mosquito Anopheles gambiae , which carries the parasite that causes malaria . The study revealed a set of genes which are highly active in oenocytes and control the production of fatty acids , a group of molecules used to make hydrocarbons . Other genes involved in creating hydrocarbons were also found . Grigoraki et al . further investigated a specific gene called FAS1899 and showed that loss of this gene reduces overall hydrocarbon production by 25% . Additionally , genes for transporting and recycling molecules and for producing fats were also shown to be active , which may indicate that oenocytes have a variety of unexplored roles besides making hydrocarbons . Grigoraki et al . identify the genes involved in producing the hydrocarbon coating of mosquitoes and demonstrate their significance . Further work is needed to understand the precise roles of each of these genes and how they are regulated to adapt the hydrocarbon coating to different situations . This can help explain how the hydrocarbon coating changes in mosquitoes , for example in response to the use of insecticides or climate change . This information is important to adapt and develop new tools to improve mosquito control . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"genetics",
"and",
"genomics"
] | 2020 | Isolation and transcriptomic analysis of Anopheles gambiae oenocytes enables the delineation of hydrocarbon biosynthesis |
A critical event in ischemia-based cell death is the opening of the mitochondrial permeability transition pore ( MPTP ) . However , the molecular identity of the components of the MPTP remains unknown . Here , we determined that the Bcl-2 family members Bax and Bak , which are central regulators of apoptotic cell death , are also required for mitochondrial pore-dependent necrotic cell death by facilitating outer membrane permeability of the MPTP . Loss of Bax/Bak reduced outer mitochondrial membrane permeability and conductance without altering inner membrane MPTP function , resulting in resistance to mitochondrial calcium overload and necrotic cell death . Reconstitution with mutants of Bax that cannot oligomerize and form apoptotic pores , but still enhance outer membrane permeability , permitted MPTP-dependent mitochondrial swelling and restored necrotic cell death . Our data predict that the MPTP is an inner membrane regulated process , although in the absence of Bax/Bak the outer membrane resists swelling and prevents organelle rupture to prevent cell death .
Mitochondria have emerged as critical regulators of apoptotic and necrotic cell death , with distinct molecular effectors underlying each process ( Tait and Green , 2010 ) . Apoptosis is both an ATP- and caspase-dependent event characterized by chromatin condensation , cell shrinkage , and plasma membrane blebbing ( Danial and Korsmeyer , 2004 ) . Necrosis , however , is an ATP-independent event characterized by organelle swelling and early plasma membrane rupture ( Danial and Korsmeyer , 2004 ) . One key difference between apoptosis and necrosis is the type of events that occur in the mitochondria and the distinct pores that are formed . During apoptosis , the mitochondrial outer membrane is permeabilized through the action of the Bcl-2 family members Bax/Bak that generate large pores that allow for the release of cytochrome c and the subsequent activation of the caspase cascade . In contrast , during forms of cellular necrosis , one key regulated event is the opening of the mitochondrial permeability transition pore ( MPTP ) , a protein complex that was proposed to span the inner and outer mitochondrial membranes in facilitating loss of the inner membrane potential , swelling , and eventual rupture of the organelle ( Halestrap , 2009; Tait and Green , 2010 ) . While the identity of the outer and inner membrane components of the MPTP remains elusive , cyclophilin D ( CypD ) , a peptidylprolyl isomerase located in the mitochondrial matrix , is known to bind and regulate the inner membrane complex . Indeed , deletion of the gene encoding CypD renders mitochondria resistant to Ca2+ overload–induced swelling and the heart and brain partially resistant to cell death due to ischemic injury ( Baines et al . , 2005; Nakagawa et al . , 2005; Schinzel et al . , 2005 ) . Apoptosis at the level of the mitochondria is absolutely dependent on Bax/Bak , as deficiency in the genes encoding these two proteins renders cells resistant to apoptosis and concomitant release of cytochrome c through the outer mitochondrial membrane ( Tait and Green , 2010 ) . Previous studies have suggested a role for Bax/Bak in the regulation or formation of the MPTP ( Marzo et al . , 1998; Narita et al . , 1998 ) , although more recent studies have disputed such a direct role , suggesting that Bax/Bak only have a secondary effect , such as by regulating mitochondrial fission/fusion or no effect whatsoever , or just a secondary effect due to organelle rupture ( De Marchi et al . , 2004; Vaseva et al . , 2012; Whelan et al . , 2012 ) . Here , we show that Bax/Bak are required for mitochondrial permeability pore-dependent cell death by serving as a necessary functional component of the MPTP within the outer mitochondrial membrane in a manner that is distinct from their more active mechanism of oligomerization during apoptosis , placing Bax/Bak at the bifurcation point of mitochondrial-dependent apoptosis and necrosis .
We first evaluated the ability of wild type ( Wt ) and Bax/Bak1-deficient ( Bak1 gene encodes Bak protein ) mouse embryonic fibroblasts ( MEFs ) to die by apoptosis or mitochondrial-dependent necrosis . As previously reported , Bax/Bak1 double-knockout ( DKO ) MEFs were highly resistant to cell death by the apoptotic inducer staurosporine ( Wei et al . , 2001 ) , while Wt MEFs showed high levels of killing ( Figure 1A ) . However , DKO MEFs were also resistant to H2O2 , ionomycin , and DNA alkylation with methyl methanesulfonate ( MMS ) that appeared to induce a necrotic phenotype ( Figure 1B–D ) . Indeed , electron microscopy verified that staurosporine-treated Wt cells , but not DKO cells , exhibited hallmarks of apoptosis including nuclear condensation with maintenance of plasma membrane integrity , while ionomycin induced a purely necrotic phenotype that showed rupture of the plasma membrane and dispersion of the nucleus , which was not seen in DKO cells ( Figure 1E ) . By comparison , analysis of apoptotic cell death with annexin V ( phosphatidylserine externalization in the plasma membrane ) and necrotic death with propidium iodide ( PI , plasma membrane opening/rupture ) staining in cultured MEFs showed an apoptotic profile with staurosporine treatment , while ionomycin induced a necrotic phenotype; H2O2 had an intermediate profile of both forms ( Figure 1—figure supplement 1A–D ) . Apoptosis leads to annexin V staining prior to PI staining , while necrosis causes simultaneous annexin V and PI staining . Molecular markers also confirmed a uniform apoptotic profile in Wt MEFs treated with staurosporine , such that caspase cleavage was observed and cell death was inhibited with z-Vad ( pan-caspase inhibitor ) but not by disabling the MPTP in MEFs from Ppif ( CypD ) -null mice ( Figure 1—figure supplement 1E–H ) . By comparison , ionomycin-induced cell death in Wt MEFs did not appreciably involve caspase activation nor was it affected by caspase inhibition , while inhibiting MPTP function by deletion of the Ppif gene–reduced cell death , suggesting mitochondrial-dependent necrosis ( Figure 1—figure supplement 1E–H ) . Using these conditions , DKO MEFs were completely refractory to H2O2- and ionomycin-induced necrosis as assessed with annexin V and PI labeling ( Figure 1F , G; compare with Figure 1—figure supplement 1C , D ) , but all deaths were restored in DKO MEFs containing a stably integrated Bax or Bak viral–based expression cassette , or acutely by infection with a recombinant Bax encoding adenovirus , collectively indicating that there is not an unrelated defect in the DKO MEFs ( Figure 1H , I ) . Reconstitution with Bax and Bak produced an equal or slightly lower level of protein compared with endogenous protein content in Wt cells ( data not shown ) . 10 . 7554/eLife . 00772 . 003Figure 1 . Bax/Bak1 DKO MEFs are resistant to a necrotic-like cell death . ( A ) – ( C ) MultiTox-Fluor multiplex cytotoxicity assay , which measures membrane ( mem ) integrity loss induced death , for different time points in cultures of Wt and Bax/Bak1 DKO MEFs treated with staurosporine ( St ) , H2O2 , ionomycin ( Iono ) for different time points . ( D ) Propidium iodide ( PI ) inclusion to assess membrane integrity following methyl methanesulfonate ( MMS ) for different time points . ( E ) Transmission electron microscopy of Wt and DKO MEFs treated with the indicated agents for 10 hr ( St , 200 nM ) or 20 hr ( Iono , 20 μM ) . Magnification is ×10 , 000 for all panels . ( F ) and ( G ) FACS quantitation of annexin V and PI staining of DKO MEFs treated with H2O2 ( 800 μM ) or ionomycin ( 20 μM ) for the indicated time points . Apoptotic and necrotic killing was almost nonexistent in DKO MEFs compared with Wt MEFs shown in Figure 1—figure supplement 1 . ( H ) PI inclusion rates for cell death assessment in DKO MEFs or DKO MEFs expressing a stable cDNA for Bax or Bak with the indicated death inducing agents for 20 hr . #p<0 . 05 vs DKO alone . ( I ) PI inclusion rates for cell death assessment in DKO MEFs and DKO MEFs infected with a Bax expression adenovirus following stimulation with the indicated death-inducing agents for 12 ( St ) or 24 hr ( H2O2 , Iono ) . #p<0 . 05 vs DKO alone . ( J ) and ( K ) Oxygen consumption rates ( OCR ) in cultures of Wt ( J ) or DKO ( K ) MEFs treated with St ( 200 nM ) or Iono ( 20 μM ) for the indicated time points . Rates are expressed as pmol/min per 1000 cells in a well of a 24-well dish . All assays were performed in duplicate and averaged from three independent experiments . *p<0 . 05 vs 0 time point; #p<0 . 05 vs no treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 00310 . 7554/eLife . 00772 . 004Figure 1—figure supplement 1 . Conditions whereby staurosporine induces only apoptosis while ionomycin induces necrosis in cultured Wt MEFs . ( A ) . Immunocytochemistry for propidium iodide ( PI , red ) , annexin V ( green ) , or a merged channel with Hoechst ( blue for nuclei ) in Wt MEFs treated with staurosporine ( St ) for 6 hr , H2O2 for 12 hr , or ionomycin ( Iono ) for 20 hr ( B ) – ( D ) FACS analysis of Wt MEFs treated with St ( 200 nM ) , H2O2 ( 800 μM ) , or Iono ( 20 μM ) for the indicated times , after which cultures were lifted from the plates and sorted for annexin V ( AnxV ) or annexin V with PI . Annexin V only cells are apoptotic , whereas annexin V with PI labeled cells are necrotic . *p<0 . 05 vs PI + AnxV in ( B ) or vs AnxV in ( D ) . Results were averaged from three independent experiments . ( E ) Western blot quantitation for cleaved caspase 3 from Wt MEFs treated with the agents shown under conditions shown in ( A ) . Cleaved caspase 3 is a reflection of apoptosis induction , which is highly correlated with St ( 200 nM , 8 hr ) treatment but not Iono ( 20 μM , 12 hr ) or H2O2 ( 800 μM , 12 hr ) . ( F ) Cell death assay for PI positivity in Wt MEFs in the presence of the indicated pro-death agents for the times shown in ( A ) with or without caspase inhibition with z-Vad . Caspase inhibition only reduced apoptotic death induced by St . *p<0 . 05 vs none ( n = 3 experiments ) . St ( 200 nM , 12 hr ) , Iono ( 20 μM , 16 hr ) , and H2O2 ( 800 μM , 12 hr ) . ( G ) Cell death assay for PI positivity in Wt and Ppif null MEFs with the indicated pro-death agents . Ppif null MEFs reduced cell death with H2O2 or ionomycin necrotic stimulation but not with the apoptotic stimuli staurosporine ( St ) . *p<0 . 0001 vs Wt control , n = 3 independent experiments . St ( 200 nM , 24 hr ) , Iono ( 20 μM , 24 hr ) , H2O2 ( 800 μM , 24 hr ) . ( H ) Western blot showing cyclophilin D and GAPDH ( control ) protein levels from Wt and Ppif knockout MEFs . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 00410 . 7554/eLife . 00772 . 005Figure 1—figure supplement 2 . Bax/Bax1 null MEFs ( DKO ) have normal mitochondrial protein content except for a slight upregulation of Bim and BNip3 . Western blotting of protein extracts fractionated as cytosol or purified mitochondria for the indicated proteins . Metabolic proteins were unchanged in Bax/Bak1 null MEFs vs Wt MEFs , as were cyclophilin D ( CypD ) and VDAC1 . Other Bcl-2 family members were unchanged apart from a slight increase in Bim and Bnip3 . Bax and Bak protein was missing . β-tubulin shows successful fractionation of just the cytosol . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 00510 . 7554/eLife . 00772 . 006Figure 1—figure supplement 3 . Bax/Bak1-deficient mitochondria remain coupled for ATP generation after ionomycin treatment . ( A ) and ( B ) Oxygen consumption rates ( OCR ) in 24-well dish cultures of Wt ( A ) or DKO ( B ) MEFs treated with St ( 200 nM ) or Iono ( 20 μM ) for 12 hr and then sequentially treated with 2 μM oligomycin ( oligo ) , 3 μM FCCP or 4 μM Antimycin A ( ANT A ) . Rates are expressed as pmoles/min per 1000 cells . All assays were performed in duplicate and averaged from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 006 Apoptotic cell death proceeds with ATP generation and cellular respiration , as the inner mitochondrial membrane typically remains intact for a period of time after the outer membrane is permeabilized . Necrosis more immediately extinguishes oxidative phosphorylation as the inner mitochondrial membrane is opened and the proton gradient is lost . Indeed , Wt MEFs treated with staurosporine maintained respiration considerably longer than observed with ionomycin treatment even though staurosporine-treated Wt MEFs died considerably faster than those with ionomycin ( Figure 1J ) . However , DKO MEFs showed no reduction in oxygen consumption with ionomycin over 24 hr , providing yet another line of evidence that MPTP-dependent mitochondrial dysfunction does not occur in the absence of Bax/Bak protein ( Figure 1K ) . Other mitochondrial proteins involved in cell death , the MPTP , and respiration all appeared normal in DKO MEFs , except for a mild upregulation of Bim and Bnip3 ( Figure 1—figure supplement 2 ) . We also verified that ionomycin treated Bax/Bak1 null MEFs still had functionally competent mitochondria as compared to ionomycin treated Wt MEFs by measuring respiration in response to oligomycin , FCCP , and antimycin A . Both oligomycin ( complex V inhibitor ) and antimycin A ( complex III inhibitor ) inhibited oxygen consumption , suggesting that respiration observed was due to mitochondrial oxygen consumption as opposed to extra-mitochondrial sources . Further , treatment with FCCP increased respiration in the Bax/Bak1 null but not in the Wt MEFs , suggesting that mitochondrial respiration was still coupled to ATP synthesis in ionomycin treated DKO cells ( Figure 1—figure supplement 3 ) . Thus , mitochondria from Bax/Bak1 null MEFs were fully functional after ionomycin and responded to the drugs similar to untreated mitochondria , showing their ability to maintain respiration and generate ATP long after Wt mitochondria have lost such ability with ionomycin treatment ( Figure 1—figure supplement 3 ) . Finally , and consistent with results in DKO MEFs , cardiac-specific deletion of Bax/Bak1 significantly protected the heart from ischemia–reperfusion ( I-R ) injury and reduced lethality in mice subjected to permanent myocardial infarction injury ( Figure 2—figure supplement 1 ) . These results indicate that loss of Bax/Bak protein protects from ischemia-induced necrotic cell death in vivo due to a disease relevant stimuli . In response to ischemia and Ca2+ overload , CypD-regulated opening of the MPTP leads to mitochondrial swelling ( Halestrap , 2009 ) . Examination of this effect in purified mitochondria isolated from Wt MEFs showed swelling with acute Ca2+ or atractyloside administration , which was blocked with cyclosporine ( CsA ) , a CypD inhibitor ( Figure 2A , C ) . However , purified mitochondria from DKO MEFs failed to show Ca2+ or atractyloside-induced swelling under the same conditions ( Figure 2B , D ) . This difference in Ca2+-induced swelling is even more pronounced in mitochondria freshly isolated from Wt or single Bak1−/− mouse hearts or livers , which showed abundant swelling ( Figure 2E , F , H , I ) . However , mitochondria purified from mouse hearts or livers of Bak1/Bax-loxP targeted mice subjected to heart- or liver-specific deletion with the appropriate Cre-expressing transgenes ( Figure 2—figure supplement 1 and 2 ) were resistant to Ca2+-induced swelling ( Figure 2G , J ) . Consistent with these results , purified DKO MEF mitochondria continue to take-up exogenous Ca2+ while mitochondria isolated from Wt MEFs show gradual inhibition of Ca2+ uptake at the 50- and 100-μM pulses , such that the MPTP opens and the mitochondria achieve equilibrium with the test solution at less cumulative Ca2+ levels ( Figure 2K , L ) . Direct inhibition of the MPTP with CsA enhanced Wt mitochondrial Ca2+ uptake because the MPTP remains closed through a higher range of Ca2+ , and by comparison , Bax/Bak1 deficiency was also highly potent in allowing progressive Ca2+ uptake without engaging the MPTP ( Figure 2L ) . Indeed , mass spectrometry analysis of Ca2+ and other cations showed that basal Ca2+ was enhanced in Bax/Bak-deficient mitochondria compared with Wt , and that with Ca2+ addition , Bax/Bak-deficient mitochondria maintained more of this ion beyond any additional effect with CsA ( Figure 2—figure supplement 3 ) . Since MPTP formation has been proposed to function as a Ca2+ release mechanism ( Elrod et al . , 2010; Bernardi and von Stockum , 2012 ) , the observed augmentation in baseline Ca2+ in Bax/Bak-deficient mitochondria , and their greater uptake capacity further suggest that Bax/Bak are required for MPTP-dependent Ca2+ release . 10 . 7554/eLife . 00772 . 007Figure 2 . Bax/Bak-deficient mitochondria are resistant to swelling and MPTP formation . ( A ) – ( D ) Mitochondrial absorbance change swelling assay upon Ca2+ ( A and B ) or atractyloside ( Atr . C and D ) addition ( arrowhead ) in mitochondria purified from Wt or DKO MEFs . CsA ( 2 μM ) is given to desensitize MPTP-induced swelling as a control . ( E ) – ( G ) Swelling assays in mitochondria purified from mouse hearts of the indicated genotypes . Ca2+ is added at a given time ( arrowhead ) , and CsA ( 2 μM ) is a control that desensitizes MPTP formation and swelling . ( H ) – ( J ) Swelling assays in mitochondria purified from mouse livers of the indicated genotypes . Ca2+ is added at a given time ( arrowhead ) , and CsA ( 2 μM ) is a control that desensitizes MPTP formation and swelling . ( K ) and ( L ) Ca2+ uptake capacity assay with the external Ca2+ indicator dye calcium green-5N with purified mitochondria from Wt or DKO MEFs . 50 ( K ) or 100 μM ( L ) cumulative Ca2+ additions are shown at each arrowhead . Fluorescence diminishes as the mitochondria remove the Ca2+ from the solution until the MPTP opens and Ca2+ is no longer sequestered . CsA is given to Wt mitochondria as a control to show the closed state of the MPTP . The swelling and Ca2+ uptake assays were performed in three independent experiments , although representative tracings are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 00710 . 7554/eLife . 00772 . 008Figure 2—figure supplement 1 . Cardiac-specific Bax/Bak1 deficiency renders the heart partially resistant to cell death after ischemic reperfusion ( I-R ) injury . ( A ) Western blotting for Bak and Bax protein from the heart of Wt mice containing the αMHC-Cre transgene , Bak1−/− only mice , and Bak1−/− mice with two Bax-loxP targeted alleles with the αMHC-Cre transgene , the later of which deletes greater than 90% of all Bax protein only in cardiac myocytes . ( B ) Survival of mice after myocardial infarction ( MI ) injury in which the left coronary artery was permanently ligated . Sham mice of all genotypes showed no lethality , but the WtαMHC-Cre control mice showed 80% lethality after MI , which was significantly reduced in mice lacking Bax/Bak protein in the heart , showing protection from MI-related death of the mice . ( C ) Histological analysis of mouse hearts from the indicated genotypes after ischemia for 1 hr and reperfusion for 24 hr stained with 2 , 3 , 5-triphenyltetrazolium chloride ( TTC ) to show area of infarction ( white area did not stain and is dead tissue ) , which was much larger in Wt controls ( WtαMHC-Cre ) than in Bax/Bak1-deleted mice ( the hashed area outlines the area of injury in the heart ) . ( D ) Quantitation of the data shown in ( C ) . #p<0 . 05 vs Wt control mice; n = >8 mice in each group . IA/AAR = ischemic area/area at risk . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 00810 . 7554/eLife . 00772 . 009Figure 2—figure supplement 2 . Liver-specific Bax/Bak1 deletion with the albumin-Cre transgene . The Cre allele deletes approximately 90% of all protein from the liver as shown by Western blotting for Bak and Bax protein from the liver of Wt mice ( WtAlb-Cre ) , Bak1−/− only , or Bak1−/− mice with two Bax-loxP targeted alleles with the Alb-Cre transgene . GAPDH is a control for protein loading . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 00910 . 7554/eLife . 00772 . 010Figure 2—figure supplement 3 . Atomic absorption mass spectrometry from purified mitochondria from Wt or Bak1−/− BaxAlb-Cre livers . Mitochondria were treated with Ca2+ for 5 min or Ca2+ with cyclosporine A ( CsA , 2 μM ) . The data show that mitochondria from DKO MEF mitochondria hold more Ca2+ at baseline ( compare black and gray bars ) and after Ca2+ challenge compared with Wt mitochondrial ( compare red and orange bars ) . CsA did not increase the amount of Ca2+ that DKO mitochondria could absorb , but it did increase Ca2+ uptake in Wt mitochondria , suggesting that loss of Bax/Bak functions in concert with the MPTP . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 01010 . 7554/eLife . 00772 . 011Figure 2—figure supplement 4 . Bax/Bak-deficient mitochondria are resistant to Ca2+-induced swelling but are otherwise capable of nonspecific swelling . ( A ) – ( H ) Mitochondrial absorbance change as assessment of swelling with addition of Ca2+ and 40 μM alamethicin ( A , B , E , and F ) or the indicated concentrations of KCl ( C , D , G , and H ) to affect baseline swollen state . Mitochondria were purified from Wt or DKO MEFs ( A–D ) or livers ( E–H ) . ( A and E ) are representative traces graphed using raw absorbance values and no normalization , while ( B and F ) are the corresponding traces graphed as relative absorbance to normalize the baselines . The swelling assays were performed in three independent experiments , although representative tracings are shown . The data show that normalization does not affect interpretation of results and that KCl concentration differences have a similar effect on baseline swelling of Wt and DKO mitochondria from both MEFs and livers . The data also show that while mitochondria from DKO MEFs and livers remain resistant to calcium-induced swelling , the non-specific permeabilizing agent alamethicin can still cause swelling . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 011 We instituted additional control experiments to further verify that mitochondria from either Bax/Bak1 null MEFs or deficient livers were otherwise uncompromised . Alamethicin , which nonspecifically permeabilizes mitochondria , showed equal swelling between mitochondria from both DKO and Wt MEFs or livers ( Figure 2—figure supplement 4A , B , E , F ) . These results indicate that mitochondria lacking Bax/Bak are still capable of swelling if uniformly permeabilized in a non-MPTP-dependent process . However , in these same experiments , DKO mitochondria were still highly resistant to Ca2+-induced swelling , whether viewed as the raw tracings or when normalized . Finally , different concentrations of KCl buffer were also used to generate a range of mitochondrial osmotic states , which showed no difference in absorbance between mitochondria isolated from Wt and DKO MEFs or livers , hence loss of Bax/Bak did not adversely affect mitochondrial swelling in this manner , suggesting that despite being insensitive to Ca2+-induced MPTP opening , they are otherwise normal ( Figure 2—figure supplement 4C , D , G , H ) . Upon apoptotic stimulation , Bax/Bak become activated and oligomerize in the outer mitochondrial membrane to generate large pores that release cytochrome c ( Tait and Green , 2010 ) . Here , we investigated Bax or Bak activation and oligomerization using our conditions for apoptotic ( staurosporine ) or necrotic ( ionomycin ) killing . Staurosporine caused the known conformational shift in both Bax and Bak in their N-termini that facilitates activation , but ionomycin caused no such effect ( Figure 3A , B ) . Consistent with these results , staurosporine but not ionomycin caused oligomerization of Bax into larger complexes assessed by gel filtration ( Figure 3C ) . 10 . 7554/eLife . 00772 . 012Figure 3 . Nonoligomerized Bax/Bak mediate mitochondrial swelling and necrosis . ( A ) Quantitation by FACS analysis of Wt MEFs treated previously with the indicated stimuli . Sorting was with the activated Bax epitope mAB 6A7 . The results were averaged from three independent experiments . ( B ) Western blotting for activated Bak from fixed cells ( DSS crosslinker ) that were previously stimulated with staurosporine ( 200 nM for 10 hr ) or ionomycin ( 20 μM for 20 hr ) . ( C ) Western blotting for Bax after gel filtration chromatography to show increasing molecular weights of complexes in cells stimulated previously with staurosporine ( 200 nM for 10 hr ) or ionomycin ( 20 μM for 20 hr ) . One of three independent experiments is shown , all with similar results . ( D ) PI incorporation cell death assay in DKO MEFs at baseline ( control ) , DKO MEFs reconstituted with Wt Bax , or DKO MEFs reconstituted with Bax mutants ( amino acids 63–65 or 92–94 were mutated to alanines ) that cannot oligomerize and generate apoptotic pores in the outer mitochondrial membrane . MEFs were stimulated with staurosporine ( 200 nM for 24 hr ) or ionomycin ( 20 μM for 24 hr ) . The results were averaged from three independent experiments . ( E ) Fluorescence reading of Ca2+ measured with calcium green-5N indicator in solution in the presence of purified mitochondria from the indicated MEFs . Cumulative Ca2+ additions are shown at each arrowhead . The assay was performed in three independent experiments , although representative tracings are shown . *p<0 . 05 vs none . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 012 Our results suggest that the presence of Bax or Bak in the outer membrane , even if inactive as defined for apoptotic stimuli , is permissive for mitochondrial swelling by facilitating permeability . Indeed , Bcl-2 family members are known to alter membrane characteristics and have channel-like activities of multiple conductance states ( Antonsson et al . , 1997; Minn et al . , 1997; Schlesinger et al . , 1997 ) . Here , we examined the ability of two nonoligomerizing mutants of Bax ( George et al . , 2007 ) to mediate cell death in the DKO MEF background . As a control , DKO MEFs reconstituted with a retrovirus expressing Wt Bax showed restoration of both staurosporine-mediated ( apoptotic ) and ionomycin-mediated ( necrotic ) killing ( Figure 3D ) . However , restoration with either of the two mutants of Bax that cannot oligomerize , although both still localize to the mitochondria ( George et al . , 2007; Hoppins et al . , 2011 ) , restored ionomycin-induced killing but not apoptotic killing with staurosporine ( Figure 3D ) . More importantly , purified mitochondria from these DKO MEFs reconstituted with Wt or the nonoligomerizing Bax mutants ( Hoppins et al . , 2011 ) each showed partial restoration of MPTP function in the Ca2+ uptake assay ( Figure 3E ) . Specifically , purified mitochondria from DKO MEFs pulsed with 75 μM Ca2+ continued to show progressive Ca2+ uptake , while mitochondria from Wt or the two nonoligomerizing Bax mutants each showed early saturation with Ca2+ ( Figure 3E ) . These results indicate that Bax can reconstitute MPTP function and permit nonapoptotic death in a nonoligomerized conformation , likely by affecting the permeability characteristics of the outer mitochondrial membrane . To examine the structural basis for how Bax/Bak might control MPTP-dependent cell death through the outer mitochondrial membrane , we first performed electron microscopy on isolated mitochondria . Purified Wt liver mitochondria treated with Ca2+ show a characteristic swelling profile and a loss of inner membrane cristae , which was prevented by inhibition of the MPTP with CsA ( Figure 4A , B ) . Purified Bax/Bak1-deficient liver mitochondria appeared morphologically normal at baseline , although in response to a Ca2+ challenge , they did not swell , despite inner membrane cristae reorganization that was still inhibited with CsA ( Figure 4A , B ) . This observation suggests that in the absence of Bax/Bak , the inner membrane still undergoes MPTP-associated cristae reorganization , but mitochondrial swelling and rupture are prevented . 10 . 7554/eLife . 00772 . 013Figure 4 . Bax/Bak-deficient mitochondria are defective in outer membrane permeability and associated swelling . ( A ) Transmission electron microscopy ( EM ) of purified Wt and Bax/Bak1-deficient ( DKO ) liver mitochondria at baseline ( none ) or with Ca2+ with or without CsA ( 2 μM ) for 5 min prior to fixation . The arrows show swollen and rupturing mitochondria . Magnification is ×40 , 000 for all panels . ( B ) Quantitation of mitochondrial cross-sectional area in different quartiles from the type of EM data shown in ( A ) . ( C ) Absorbance reading for swelling in liver-derived DKO mitochondria treated with the indicated conditions . The gray lines are the controls that represent all five Ca2+ stimulated conditions with CsA , or DKO mitochondria not stimulated with Ca2+ . ( D ) Quantitation of the change in absorbance tracings shown in ( C ) for mitochondrial swelling under the indicated conditions . Four independent swelling experiments were tabulated . *p<0 . 05 vs DKO + Ca2+ only . ( E ) Absorbance reading for swelling in liver-derived Wt mitochondria treated with Ca2+ and increasing concentration of poloxamer 188 ( P-188 ) . ( F ) EM of purified Wt liver mitochondria with and without Ca2+ , with and without 5% P-188 . The arrows show swollen and rupturing mitochondria . Magnification is ×40 , 000 for all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 01310 . 7554/eLife . 00772 . 014Figure 4—figure supplement 1 . Bax/Bak1 DKO MEF are fully susceptible to inner membrane permeability , but resist full MPTP with Ca2+ overload . ( A ) Calcein–CoCl2 staining of Wt and Bax/Bak1 DKO MEFs treated with 200 nM Staurosporine ( St ) or 20 μM ionomycin ( iono ) for 15 min . Both are equally susceptible to inner membrane opening under living conditions . ( B ) Western blots from purified cytoplasm ( without mitochondria ) showing cytochrome c release and cleaved caspase 3 from Wt and DKO MEFs that were treated with 200 nM St , 800 μM H2O2 , or 20 μM Iono for 6 hr . DKO MEFs do not form pores to release cytochrome c with staurosporine or rupture after H2O2 or ionomycin treatment to release cytochrome c . ( C ) Relative oxygen consumption rates from mitochondria isolated from Wt or DKO MEFs that were subjected to four sequential additions of 200 μM Ca2+ . There was a 5-min incubation time after each injection of Ca2+ . DKO mitochondria , despite transient inner membrane opening , are able to maintain respiration , likely because the lack of outer membrane permeability keeps the mitochondria intact so the inner membrane can reestablish electron transport . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 014 We more directly examined the manner in which the outer membrane could affect the inner membrane and MPTP function . Specifically , in the absence of Bax/Bak protein , mitochondrial inner membrane opening still occurred in response to ionomycin as measured directly with the calcein–CoCl2 assay , similar to Wt MEFs ( Figure 4—figure supplement 1A ) . This later observation is consistent with ultrastructual data discussed above in which the inner membrane can still reorganize without Bax/Bak in response to necrotic stimuli . Notably , purified mitochondria from Wt MEFs showed release of cytochrome c with ionomycin stimulation and loss of respiration with serial addition of Ca2+ , while mitochondria lacking Bax and Bak protein did not show cytochrome c release or loss of respiration ( Figure 4—figure supplement 1B , C ) . These results suggest that the inner membrane component of the MPTP still functions normally in Bax/Bak1 DKO MEFs , but without sufficient outer membrane permeability , swelling/rupture is prevented and the mitochondria can reestablish respiration capacity . To more carefully examine this concept , we attempted to reconstitute the outer membrane function of Bax/Bak with other pore-forming entities or nonoligomerized recombinant Bax ( rBax ) . Purified liver DKO mitochondria stimulated with Ca2+ were still protected from swelling ( Figure 4C , D ) . However , treatment of DKO mitochondria with rBax at concentrations that did not cause cytochrome c release ( data not shown ) fully restored Ca2+-induced mitochondrial swelling , which was still inhibited with CsA ( Figure 4C , D ) . Moreover , treatment with gossypol , a BH3-mimetic compound that can convert Bcl-2 and Bcl-xl into pore-forming pro-cell death agents ( Lei et al . , 2006 ) restored mitochondrial swelling in the absence of Bax/Bak protein , which was also inhibited with CsA ( Figure 4C , D ) . Generation of nonspecific pores that only affected the outer mitochondrial membrane with either digitonin or tetanolysin also restored Ca2+-induced swelling in DKO mitochondria , which was again inhibited with CsA ( Figure 4C , D ) . These results suggest that Bax/Bak permit mitochondrial swelling through a permeabilization activity in the outer mitochondrial membrane that functions independently of the inner membrane CypD-regulated component of the MPTP . Indeed , increasing concentrations of poloxamer-188 ( P-188 ) , which inherently reduces membrane permeability characteristics ( Wu et al . , 2004 ) , prevented mitochondrial swelling induced by Ca2+ as analyzed in purified mitochondria in solution or by transmission electron microscopy ( Figure 4E , F ) . These observations suggest that outer membrane permeability plays a necessary and permissive role in MPTP-dependent swelling , rupture , and cellular necrosis . We also performed patch clamping of the outer mitochondrial membrane from Wt and DKO purified mitochondria to directly examine if Bax/Bak were generating a pore-like activity in the outer membrane . The MPTP has been well characterized by direct patch-clamping of mitochondria and shown to have a specific conductance profile that is distinct from any other channels or permeating activities ( Kinnally and Antonsson , 2007 ) . The data show baseline conductance of approximately 750 pS in Wt mitochondria , which was significantly reduced in DKO mitochondria to approximately 300 pS ( Figure 5A , C ) . Addition of the same non-oligomerizing rBax restored this baseline permeability and even enhanced it slightly ( Figure 5A , C ) . This low-conductance channel activity with rBax was slightly cation selective but was not voltage dependent ( ±50 mV ) ( data not shown ) . These results suggest that the presence of Bax/Bak impart a different permeability characteristic to the outer mitochondrial membrane that relates to its necessary but permissive role in MPTP function . Because Bax/Bak could be acting through another protein in the outer mitochondrial membrane to impart this increase in permeability , we also used a reconstitution assay in liposomes ( Figure 5B , C ) . Remarkably , addition of monomeric rBax significantly increased the basal permeability of liposomes , but to a level , that was still more than 10× less than when rBax was forced to oligomerize with recombinant tBid treatment ( Figure 5B , C ) . The high resistance and low permeability of the outer mitochondrial membrane with Bax/Bak deficiency reported here and previously ( Martinez-Caballero et al . , 2009 ) suggest that this membrane is not a sieve but rather a membrane whose permeability is tightly regulated , despite the presence of other channels such as voltage-dependent anion channel ( VDAC , comprising as many as three gene products , Vdac1 , Vdac2 , and Vdac3 ) and TOM ( Jonas et al . , 2004; Martinez-Caballero et al . , 2005 ) . 10 . 7554/eLife . 00772 . 015Figure 5 . Bax/Bak regulate outer mitochondrial membrane permeability directly . ( A ) and ( B ) Patch clamp current traces ( 2 s , 5 KHz sampling , 1 KHz filter in 0 . 15 M KCl media ) from Wt and DKO whole mitochondria and liposomes are shown after excised patches were voltage clamped at −20 to 40 mV and reveal small variable-sized transitions , typically approximately 50–100 pS . Recombinant Bax ( rBax at 50–100 ng/µl ) was included in the patch pipette as indicated . Current trace recorded from a liposome with a patch pipette backfilled with media containing 10 nM rBax + 10 nM tBid reveals up to 3000 pS transitions , which are not seen with rBax alone . ( C ) Graph of multiple independent patch recordings of the representative traces shown in ( A and B ) . Large channel activities , such as from TOM , were excluded from the Wt and DKO mitochondrial data set . *p<0 . 05 vs Wt mitochondria; #p<0 . 05 vs untreated liposomes . ( D ) Patch clamp current traces from mitoplasts ( inner membrane ) were recorded as in ( A ) from Wt and DKO mitochondria at −60 and −20 mV , respectively . ( E ) Graphs showing comparisons of MPTP single-channel characteristics recorded from the indicated number of independent patches of mitoplasts isolated from Wt and DKO cells . MPTP was scored present from mitoplast recordings if the peak conductance was ≥1 nS , transition sizes ≥0 . 3 nS , voltage dependence , and when possible , cation selectivity . ( F ) Western blot for the outer and inner mitochondrial membrane proteins Sam-50 and complex V/III , respectively . Samples were normalized for complex V/III to show differences in Sam-50 and proper enrichment of inner membrane in the mitoplast preparation vs whole mitochondria . ( G ) Absorbance reading for swelling in liver-derived Wt and DKO mitoplasts ( outer membranes removed ) treated with the indicated conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 015 Importantly , direct patching of the inner mitochondrial membrane in mitoplasts ( outer mitochondrial membranes are removed ) showed that Bax/Bak do not directly affect the MPTP at the level of the inner membrane ( Figure 5D , E ) . Indeed , the MPTP from the inner membrane of Wt mitochondria showed no difference in frequency , conductance peak , or reversal potential with Bax/Bak null mitochondria ( Figure 5D , E ) . Direct analysis of Ca2+-induced swelling of purified mitoplasts ( outer membranes removed ) showed that both Wt and DKO preparations swelled to the same extent in a manner that was still inhibited with CsA ( Figure 5G ) . Western blotting confirmed the integrity of the mitoplast preparation ( Figure 5F ) . These final observations further indicate that Bax and Bak do not directly regulate the inner membrane aspects of the MPTP , further supporting their more dedicated function within the outer mitochondrial membrane in permitting mitochondrial swelling and end-stage MPTP that leads to organelle rupture and cell death . Bax/Bak1 DKO cells were previously shown to have reduced Ca2+ levels in the endoplasmic reticulum ( ER ) as a protective mechanism against cell death ( Scorrano et al . , 2003 ) . We also examined this effect to determine if it might secondarily influence necrotic cell death in the DKO MEFs , or in purified mitochondria from these cells . Total intracellular Ca2+ levels were similar in DKO MEFs compared with Wt MEFs , as was ER Ca2+ levels measured with two different agonists ( thapsigargin or ATP ) , although total mitochondrial Ca2+ levels were significantly elevated in DKO MEFs ( Figure 6A–D ) . It is uncertain why we failed to observe a decrease in ER Ca2+ in Bak/Bak1 DKO MEFs , although our conditions were different from those previously reported , as we measured individual living cells in adherent cultures and used fivefold higher levels of thapsigargin . To unequivocally rule out a secondary effect due to lower levels of ER Ca2+ in DKO MEFs , we also used a recombinant adenovirus expressing the sarcoplasmic reticulum Ca2+ ATPase 1 ( SERCA1 ) to load the ER with even more Ca2+ , as previously shown ( Scorrano et al . , 2003 ) . However , AdSERCA1 infection neither restored cell death with ionomycin or staurosporine administration in DKO MEFs nor did it increase cell death in Wt MEFs , although it did load cells with significantly more Ca2+ ( Figure 6E and data not shown ) . Hence , we do not believe that DKO cells are resistant to mitochondrial pore-dependent cell death due to the previously reported mechanism of reduced ER Ca2+ levels . Indeed , the initial concept that increased Bcl-2 activity/expression decreases ER Ca2+ load by increasing leak , thereby protecting from cell death , is controversial because a number of reports failed to observe any such decrease in ER Ca2+ ( Distelhorst and Shore , 2004 ) . 10 . 7554/eLife . 00772 . 016Figure 6 . Assessment of Ca2+ dynamics in Bax/Bak1-deficient MEFs . ( A ) Measurement of total cellular Ca2+ in Wt and DKO MEFs using the ratiometric Ca2+ indicator Fura-2 ( read as ratio difference F340/F380 ) . The arrowhead shows where thapsigargin , FCCP , EDTA , and ionomycin are added to release all intracellular Ca2+ from the ER and mitochondria . Two hundred Wt and 320 DKO cells were analyzed . ( B ) and ( C ) Same measurements as in ( A ) , except that the SERCA1 inhibitor thapsigargin ( B ) or ATP ( C ) are used to release ER Ca2+ over time . The Ca2+ signal from DKO MEFs is not significantly different from Wt MEFs . Individual cells were measured on the dish while still alive . One hundred and eighty-five Wt and 282 DKO cells were analyzed for Tg , and 88 Wt and 181 DKO were analyzed for ATP . ( D ) Same measurements as in ( A ) except that only the mitochondrial Ca2+ liberating agent FCCP is given . DKO MEFs on a culture dish have greater Ca2+ release from their mitochondria than do Wt MEFs ( p<0 . 05 ) , suggesting greater content in the mitochondria at rest . Twenty-six Wt and 44 DKO cells were analyzed . ( E ) Assessment of SERCA1 overexpression in Bax/Bak1 DKO MEFs and cell death ( PI positive ) . The data show that SERCA1 overexpression in Wt or DKO MEFs does not sensitize to cell death with ionomycin or staurosporine ( N = 3 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 016 Another issue to address is the potential compensatory role that the prosurvival Bcl-2 family members might play in the absence of Bax/Bak protein in affecting the MPTP-dependent cell death . To examine this issue , we used the Bcl-2/Bcl-xl inhibitor ABT-737 in Bax/Bak1 DKO and Wt MEFs and examined killing ( Figure 7A ) . ABT-737 administration did not restore necrotic killing with ionomycin in Bax/Bak1 DKO MEFs , although it did increase killing in Wt MEFs ( Figure 7A ) . However , gossypol , which can convert endogenous Bcl-2 proteins into a pro-death configuration and increase membrane permeability ( de Peyster et al . , 1986 ) , enhanced or induced ionomycin-dependent cell death in Wt and DKO MEFs , respectively ( Figure 7A ) . This result suggests that increasing outer membrane permeability through Bcl-2 family members enhances mitochondrial-dependent necrotic cell death in vivo ( Figure 7A ) . 10 . 7554/eLife . 00772 . 017Figure 7 . Protective Bcl-2 family members are not responsible for the protection observed with Bax/Bak1 DKO and gossypol restores Ca2+-induced killing in DKO cells . ( A ) PI incorporation cell death assay in Wt and DKO MEFs without treatment ( none ) or with staurosporine ( St ) or ionomycin ( Iono ) with or without the Bcl-2/Bcl-xl inhibitor ABT-737 and or gossypol . St was used at 200 nM for 12 hr , ionomycin was used at 20 μM for 24 hr , ABT-737 was used at 20 μM , and gossypol was used at 10 μM . The results were averaged from three independent experiments . *p<0 . 05 vs untreated; #p<0 . 05 vs no ABT-737 or gossypol in St or Iono treated Wt MEFs; †p<0 . 05 vs ionomycin-treated DKO MEFs . ( B ) Ca2+ uptake capacity assay with the external Ca2+ indicator dye calcium green-5N and purified mitochondria from Wt or DKO MEFs . 75 μM Ca2+ additions are shown at each arrowhead . Fluorescence in the supernatant diminishes as the mitochondria remove the Ca2+ from the solution , until the MPTP opens and the Ca2+ is no longer sequestered . The swelling and Ca2+ uptake assays were performed over three independent experiments , although representative tracings are shown . Gossypol was given as a control for an agent that can increase the permeability of the outer mitochondrial membrane in the absence of Bax/Bak . ( C ) Assay similar to that shown in ( B ) for Ca2+ release and MPTP activity under the indicated conditions in purified Wt or DKO mitochondria . The assay was recorded continuously while 50 μM Ca2+ was given in two spikes over time , followed by treatment with ABT-737 or gossypol ( given at the arrowheads ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 017 We also analyzed Ca2+ uptake capacity in purified mitochondria in the presence of ABT-737 and gossypol to more specifically address the effect on MPTP activity through Bcl-2 family members . ABT-737 did not restore the lack of Ca2+ release induced by the MPTP in Bax/Bak1 DKO mitochondria , while it did hasten MPTP-dependent Ca2+ release in Wt mitochondria , collectively suggesting that the prosurvival Bcl-2 family members are not gaining a new protective function in the absence of Bax/Bak protein that might antagonize the MPTP , but that they are working through Bax/Bak ( Figure 7B ) . Consistent with the results presented in Figure 7A , addition of gossypol to DKO mitochondria restored MPTP function with Ca2+ addition , again suggesting that permeabilization of the outer mitochondrial membrane can suffice for Bax/Bak in mediating the final aspects of MPTP-dependent mitochondrial swelling/rupture ( Figure 7B ) . We probed this final conclusion in greater mechanistic detail by performing a comparative analysis of Ca2+ release from Wt and DKO mitochondria with various agents . After two dosages of Ca2+ , subsequent gossypol addition induced full Ca2+ release in both Wt and DKO mitochondria ( Figure 7C , top two green colored traces ) . DKO mitochondria with or without ABT-737 were resistant to Ca2+ loss , whereas Wt mitochondria treated with this agent showed partial Ca2+ release ( Figure 7C , compare blue colored traces ) . This later result shows that by inhibiting the Bcl-2 proteins in Wt mitochondrial , Bax and Bak have even greater activity in the outer mitochondrial membrane leading to enhanced permeability that is more likely to cause swelling with reduced Ca2+ release . The fact that DKO mitochondria do not respond whatsoever to ABT-737 demonstrates that the effect is fully dependent on Bax and Bak and their function at the outer membrane ( Figure 7C ) .
It is intriguing that loss of Bax/Bak protein dramatically desensitizes the mitochondrial permeability pore , in a manner similar to CsA . Over a decade ago , several reports suggested that Bax and/or Bak directly interacted with presumed components of the MPTP , the adenine nucleotide translocator ( ANT ) in the inner membrane and the VDAC in the outer membrane ( Marzo et al . , 1998; Narita et al . , 1998; Shimizu et al . , 1999 ) . However , while the mitochondrial swelling/permeability pore data reported in these earlier manuscripts are still valid , the conclusion that Bax/Bak function merely to regulate the MPTP by binding VDAC or ANT is not consistent with the data from Slc25a4/Slc25a5 ( ANT1/2 protein ) and Vdac1/2/3 gene–deleted mice , which showed that neither protein is directly required for MPTP formation ( Kokoszka et al . , 2004; Baines et al . , 2007 ) . Our data support a model whereby Bax and/or Bak are part of the outer membrane component of the MPTP in a ‘resting’ nonapoptotic state . Bax has been shown to adopt two types of channel activities , one of which has a small pore size ( 0 . 9 nm ) and is homogenous in the outer mitochondrial membrane , which could fulfill the permeability function we observed here to permit nonapoptotic cell death ( Lin et al . , 2011 ) . As another consideration , the α5/α6 domain of Bax is known to have membrane permeation activity and to permit pore formation on its own , the permeation size of which appears to reach an equilibrium of less than 10 kDa ( Fuertes et al . , 2010 ) . Bcl-2 family members appear to generally alter the permeability and rigidity of lipid membranes , and even the Bcl-2 homologue from Caenorhabditis elegans , CED-9 can induce leakiness of reconstituted lipid membranes ( Tan et al . , 2011 ) . The exact mechanism whereby Bcl-2 family members might increase membrane permeability and small molecular permeation in their nonoligomerized state is unknown , although it may be related to the distinct lipid environment in the mitochondria and how it interacts with the select hydrophobic α-helixes in the core of the Bcl-2 proteins . The most straightforward model that is consistent with our data and the literature describing the biophysical properties of Bcl-2 family members is that Bax and Bak function as the outer membrane activity of the MPTP . Once the inner membrane pore opens through a CypD-regulated event , the outer membrane is already poised to complete the process , given the permeability characteristics imparted by Bax/Bak in their monomeric states ( Figure 8 ) . Indeed , liposome reconstitution with recombinant nonactivated Bax directly increased membrane permeability , and rBax immediately restored swelling in isolated mitochondria . Moreover , increasing the activity of Bax/Bak with ABT-737 sensitized the MPTP to opening with mild Ca2+ stimulation in Wt but not DKO mitochondria , and nonspecific increases in outer mitochondrial membrane permeability with gossypol or other agents fully restored mitochondrial swelling and cell death in the absence of Bax/Bak . Direct patching of mitochondria outer vs inner membranes showed that Bax/Bak permit MPTP-dependent mitochondrial swelling by only enhancing conductance of the outer membrane . Our results also suggest that the outer mitochondrial membrane does not directly induce the inner membrane to undergo MPTP , and that initiation of this process is an inner membrane-regulated phenomenon ( Figure 8 ) . However , lack of sufficient outer membrane permeability , as observed in the absence of Bax/Bak protein or with addition of poloxamer-188 in Wt mitochondria , sufficiently restrains swelling and organelle rupture to allow reestablishment of inner membrane potential and continued cell survival . Thus , Bax and Bak are necessary for MPTP-dependent cell death by functioning exclusively within the outer membrane as permeability factors that facilitate an irreversible swelling threshold leading to rupture of the mitochondrion and cell death , once the inner membrane component is fully engaged . 10 . 7554/eLife . 00772 . 018Figure 8 . Schematic representation of how Bax and Bak influence MPTP-dependent mitochondrial swelling and organelle rupture . The model shows mitochondria undergoing MPTP opening ( blue ) via CypD ( yellow ) in the presence and absence of Bax/Bak ( red ) . When Bax/Bak are present on the outer membrane and the MPTP opens in response to Ca2+ , it causes the mitochondrial inner membrane to dissipate its electrochemical gradient leading to additional swelling and eventually rupture of the outer membrane and entire organelle . When Bax/Bak are absent , the outer membrane has lower permeability , which prevents swelling and rupture and subsequent necrosis even though the inner membrane has undergone MPTP opening . DOI: http://dx . doi . org/10 . 7554/eLife . 00772 . 018 While the role of Bax/Bak might appear to be passive in permitting mitochondrial swelling and MPTP-dependent death by augmenting the permeability characteristics of the outer membrane , the entire process is clearly not passive and should still be subject to acute regulation through other Bcl-2 family members that affect Bax/Bak . Indeed , ABT-737 treatment increased cellular necrosis , mitochondrial swelling , and enhanced Ca2+ release in isolated mitochondria by augmenting the amount/activity of Bax/Bak in the outer mitochondrial membrane . While we cannot rule out other potential mechanisms of action for ABT-737 , our results are consistent with the hypothesis that acute alterations in the activity of select Bcl-2 family members could affect cellular necrosis by acutely regulating the activity , localization , or function of Bax/Bak . Indeed , Nix/BNip3L , which is induced by disease/developmental signaling pathways that enhance cell death , can alter the activity and pro-death characteristics of Bax/Bak as well as alter MPTP-dependent cell death ( Chen et al . , 2010 ) . With respect to medical relevance , our results are important because they suggest that Bax/Bak are a common nodal point in both apoptotic and necrotic cell death . Thus , inhibition of Bax/Bak should be the most therapeutically potent means of antagonizing acute cell death following ischemic injury in vivo or in adult onset degenerative diseases , as it would block both mitochondrial-dependent processes . Moreover , other types of regulated necrotic cell death might also be inhibited in the absence of Bax/Bak , such as necroptosis , also referred to as extrinsic necrosis or RIP-dependent necrosis ( Degterev et al . , 2005; Hitomi et al . , 2008; Wang et al . , 2012 ) . Indeed , we and others have observed that Bax/Bak1 null MEFs were resistant to this form of cell death ( data not shown and Irrinki et al . , 2011 ) . Necroptosis is induced in cells when stimulated with tumor necrosis factor α ( TNFα ) in the presence of caspase inhibitors , such as zVAD-fmk ( Degterev et al . , 2005 ) . Necroptosis is inhibited with by necrostatins , a series of serine–theronine kinase inhibitors that were later shown to block RIP1 kinase in preventing TNFα- and ZVAD-dependent necroptosis in culture . RIP1 is able to interact with RIP3 in controlling necroptosis ( Cho et al . , 2009; He et al . , 2009; Zhang et al . , 2009 ) . Recently , mixed lineage kinase domain–like protein ( MLKL ) and phosphoglycerate mutase family member 5 ( PGAM5 ) were shown to be downstream of the RIP proteins ( Sun et al . , 2012; Wang et al . , 2012 ) . Knockdown of either MLKL or PGAM5 resulted in protection against RIP-mediated necroptosis , as well as protection against ROS or Ca2+ overload–mediated necrosis ( Wang et al . , 2012 ) . This duality led the authors to recognize the existence of two arms of the necrotic pathway , the extrinsic arm ( TNF + zVAD–induced necrosis ) and the intrinsic arm that responds to ROS or Ca2+ overload–induced necrosis , which is the form of necrosis that we investigated here as affected by Bax/Bak ( Wang et al . , 2012 ) .
Wt and Bax and Bak1 DKO SV40 immortalized MEFs , and all variations were cultured in IMDM medium supplemented with 10% bovine growth serum , antibiotics , and nonessential amino acids . DKO MEFs expressing Wt and mutant Bax-GFP were previously described ( Wei et al . , 2001; Hoppins et al . , 2011 ) . DKO MEFs with reconstituted Bax and Bak were also previously described ( Kim et al . , 2009 ) . DKO MEFs were infected with a binary adenovirus expressing Bax ( Ad-Bax ) was previously described and obtained from Dr Bingliang Fang ( Kagawa et al . , 2000 ) . DKO MEFs were also infected with an adenovirus expressing β-galactosidase ( Adβgal , control ) or sarco/endoplasmic reticulum Ca2+-ATPase ( AdSerca1 ) ( Vector BioLabs , Philadelphia , PA ) . Ppif null MEFs were described previously ( Baines et al . , 2005 ) , although for the current experiments , the cells were subjected to SV40-mediated immortalization . At 80% confluence , MEFs were treated with 200 nM staurosporine for 2–24 hr , 20 μM ionomycin for 4–24 hr , 750–800 μM H2O2 for 2–24 hr , 750 μM MMS for 2–24 hr , or 60–90 ng TNFα and 40 μM caspase inhibitor z-Vad-FMK ( z-Vad; Promega , Madison , WI ) for 24 hr ( in some experiments , cells were pretreated with 40 μM z-Vad , 20 μM ABT-737 , or 10 μM Gossypol ) . Cell death and viability was determined by PI uptake and annexin V ( AV ) positivity ( BioVision , Milpitas , CA ) . Briefly , cells were trypsinized and washed twice and incubated with AV and PI for 10 min . The cells were then quantified for PI and AV positivity at 10 , 000 counts per sample by using a Cell Lab Quanta SC flow cytometer ( Beckman Coulter , Indianapolis , IN ) . In some experiments , only PI was used and AV was replaced with PBS . Cell death and viability were also measured using the MultiTox-Fluor Multiplex Cytotoxicity Assay ( Promega ) . Briefly , cells were cultured in 96-well plates and treated for given times and dosages , after which they were incubated with 2× MultiTox-Fluor Multiplex Cytotoxicity Assay reagent for 30 min at 37°C . The assay quantifies dead/live ratios by measuring membrane rupturing or opening , such as during necrosis or apoptosis , which allows access of proteases to the fluorogenic peptide substrate that was quantified by using Synergy 2 Multi-Mode Microplate Reader ( BioTek , Winooski , VT ) . Oxygen consumption was measured with an XF extracellular flux analyzer ( Seahorse Bioscience , North Billerica , MA ) . Cells were plated to 100% confluence on XF24 cell culture microplates . After an initial measurement , the wells were injected with St , Iono , or vehicle , and oxygen consumption was measured every 6 hr for 24 hr . Liver , heart , and MEF mitochondria were isolated by homogenization followed by differential centrifugation . Livers and hearts were prepared with a Teflon homogenizer , while cells were disrupted with a glass homogenizer . The isolation buffer consisted of 250 mM sucrose and 10 mM Tris pH 7 . 4 . Mitoplasts were isolated by incubating the isolated mitochondria in a hypotonic buffer , which consisted of 10 mM KCl , 10 mM Tris pH 7 . 4 for 20 min followed by gentle agitation with a pipette , and a low-speed centrifugation . Mitochondrial swelling was performed on either 0 . 5 or 1 mg of mitochondria in 1 ml with light scattering measured at 540 nm . The mitochondrial swelling buffer consisted of 120 mM KCl , 10 mM Tris pH 7 . 4 , 5 mM KH2PO4 , 7 mM pyruvate , 1 mM malate , and 10 μM EDTA . Swelling was induced by either 400 or 800 μM CaCl2 or 200 μM atractyloside in EDTA-free buffer . In some experiments , mitochondria were also treated with 10 μM gossypol , 9 nM tetanolysin , 25 μM digitonin , 1–10% poloxamer-188 , 2 μM CsA , or 5 μg of monomeric recombinant Bax-GST ( rBax ) . Mitoplast swelling was performed identically to mitochondrial swelling except that the swelling buffer consisted of 10 mM KCl . We determined that the rBax was not causing cytochrome c release by treating mitochondria from Bax-loxP ( fl ) Bak1−/− Albumin-Cre livers with 5 μg rBax alone or with 2 . 5 μg recombinant tBid and 5 μg rBax as a positive control or vehicle . The mitochondria were incubated in swelling buffer containing 100 mM KCl for 10 min at room temperature to emulate the swelling assay . After incubation , the mitochondria were centrifuged and the supernatant was subjected to Western blot analysis for cytochrome c . Cytochrome c release was detected from whole cells treated with 200 nM St , 800 μM H2O2 , or 20 μM Iono for 6 hr by Western blot analysis . Cytosolic fractions were obtained by homogenizing the treated cells in mitochondrial isolation buffer containing 100 mM KCl . Homogenates were then centrifuged at 14 , 000×g for 10 min; the cytosolic fractions were collected and subjected to an additional spin and then Western blotting . Mitochondrial Ca2+ uptake was measured with Calcium Green-5N ( Invitrogen , Grand Island , NY ) as previously described ( Kwong et al . , 2007 ) . Briefly , MEF mitochondria were isolated in MS-EGTA buffer ( 225 mM mannitol , 75 mM sucrose , 5 mM HEPES , and 1 mM EGTA , pH 7 . 4 ) . Mitochondria ( 150 or 200 μg for 96-well format or 500 μg for single cuvette format ) were then incubated in KCl buffer ( 125 mM KCl , 20 mM HEPES , 2 mM MgCl2 , 2 mM KH2PO4 , and 40 μM EGTA , pH 7 . 2 ) containing 200 nM Calcium Green-5N , 7 mM pyruvate , and 1 mM malate . Mitochondria were treated with sequential additions of CaCl2 ( 50–100 μM ) . Fluorescence was quantified using a Synergy 2 Multi-Mode Microplate Reader ( BioTek ) or by cuvette-based fluorometric analysis using a fluorometer ( Photon Technology International [PTI] , Birmingham , NJ ) . For some experiments , mitochondria were treated with 10 μM gossypol , 20 μM ABT-737 , or 1–10% poloxamer-188 . Inner mitochondrial membrane permeability was assessed by using the calcein/cobalt assay . Briefly , cells were incubated with 1 μM Calcein-AM and 8 mM CoCl2 for 10 min and then treated with 200 nM St or 20 μM Iono for 15 min . Cells were imaged by confocal microscopy . Mitochondrial oxygen consumption was measured with an XF extracellular flux analyzer . Mitochondria ( 150 μg ) were incubated in KCl buffer containing 7 mM pyruvate and 1 mM malate . Mitochondria were treated with four sequential additions of CaCl2 ( 200 μM ) . Oxygen consumption was measured 5 min after every injection . Bax-loxP ( fl ) Bak1−/− mice were described previously ( Takeuchi et al . , 2005 ) . To create heart- and liver-specific Bax and Bak1 knockout mice , Baxfl/fl Bak1−/− mice were crossed with α-myosin heavy chain ( αMHC ) -Cre and albumin-Cre transgenic lines ( Postic et al . , 1999; Oka et al . , 2006 ) . IR injury was performed as previously described ( Kaiser et al . , 2004 ) . Mice were also subjected to permanent ligation of the left anterior descending coronary artery to induce myocardial infarctions . These mice were monitored daily for 8 weeks , and death events were recorded . Electron microscopy was performed on MEFs and mitochondria isolated from Wt Albumin-Cre and Baxfl/fl Bak1−/− Albumin-Cre mouse livers . Prior to fixation , the mitochondria were subjected to the mitochondrial swelling assay . Samples were then fixed in glutaraldehyde and cacodylate , embedded in epoxy resin , and sectioned . Sections were counterstained with uranyl acetate and lead citrate . Mitochondrial cross-sectional area was quantified by using image-J software ( NIH ) . Histological analysis of the injury induced by myocardial ischemia/reperfusion was described previously ( Kaiser et al . , 2004 ) . Livers , hearts , and MEFs were homogenized in RIPA buffer containing protease inhibitor cocktails ( Roche , Indianapolis , IN ) . The following antibodies were used: Bax ( Santa Cruz , Santa Cruz , CA ) , active Bax ( Exalpha Biologicals , Shirley , MA ) , Bak ( Millipore , Billerica , MA ) , Bcl-2 ( Santa Cruz ) , Bcl-xl ( Santa Cruz ) , Bid ( Santa Cruz ) , Bim ( BD Biosciences , San Jose , CA ) , BNip3 ( Cell Signaling , Danvers , MA ) , VDAC 1 ( MitoSciences , Eugene , OR ) , ANT 1 ( EMD4Biosciences , Billerica , MA ) , CypD ( MitoSciences ) , cytochrome c ( BD biosciences ) , complex I-V ( MitoSciences ) , Sam50 ( Sigma , St . Louis , MO ) , and β-tubulin ( Santa Cruz ) . Mitochondria from Wt and DKO MEFs were lysed in IP buffer ( 20 mM Tris pH 7 . 4 , 150 mM NaCl , 1% triton X-100 , 0 . 3 mM PMSF , 0 . 5 mM DTT , and 1× phosphatase inhibitors ) . Lysates were incubated with primary antibody for Bax and protein A/G plus agarose ( Santa Cruz ) beads for 12 hr at 4°C . The lysates were centrifuged and washed three times and boiled in SDS sample buffer . Flow cytometry analysis of active Bax has been previously described ( Panaretakis et al . , 2002 ) . Briefly , cells were harvested and fixed in 0 . 25% paraformaldehyde for 5 min , washed three times in PBS , and incubated with primary antibody diluted at 1:50 in PBS-containing digitonin ( 100 μg/ml ) for 30 min . The cells were then washed three times in PBS and incubated with FITC-labeled secondary antibody for 30 min , washed , and resuspended in PBS and analyzed with a flow cytometer ( 10 , 000/sample ) . Activation of Bak was determined by cross-linking isolated mitochondria ( 50 μg ) from cells treated with 10 mM disuccinimidyl suberate ( DSS ) at room temperature for 30 min . The mitochondria were then suspended in SDS sample buffer , and a Western blot was performed . Wt MEFs were treated with or without St and Iono for 12 hr in the presence of 40 μM z-Vad . After 12 hr , they were washed with ice-cold PBS and lysed in HNC buffer ( 25 mM HEPES , pH 7 . 5 , 300 mM NaCl , 1 mM DTT , and 2% CHAPS ) . The lysates were loaded onto a Superdex 200 HR 10/30 column ( GE , Pittsburgh , PA ) . The column was pre-equilibrated with a buffer consisting of 25 mM HEPES , pH 7 . 5 , 300 mM NaCl , 0 . 2 mM DTT , and 2% CHAPS and calibrated with ferritin ( 440 kDa ) , B amylase ( 200 kDa ) , alcohol dehydrogenase ( 150 kDa ) , BSA ( 66 kDa ) , carbonic anhydrase ( 29 kDa ) , and cytochrome c ( 12 kDa; Sigma ) . The proteins were collected in 0 . 6 ml fractions every minute . The fractions were then subjected to TCA precipitation to concentrate the protein before Western blotting . MEFs were plated on glass bottom dishes ( MatTek , Ashland , MA ) . Cells were loaded in filtered Ringer’s solution consisting of 145 mM NaCl , 5 mM KCl , 2 mM CaCl2 , 1 mM MgCl2 , and 10 mM HEPES with 5 μM Fura-2 AM for 30 min at room temperature , washed twice , and incubated for an additional 30 min before beginning the experiment . Experiments were performed on a Nikon Eclipse Ti-U inverted microscope equipped with a Delta Scan dual-beam spectrofluorophotometer ( PTI ) set to excite at 340 and 380 nm . A Photometrics CoolSnap ES2 camera was used to acquire images at 510 nm . EasyRatioPro software was used to gather and analyze the data . Experiments were performed by imaging cells at 1 Hz for 2 min in calcium- and magnesium-free ( CMF ) solution to obtain a basal calcium reading . After 2 min , the cells were treated with the following: 1 μM thapsigargin ( Tg ) , 100 μM ATP , 2 μM FCCP , or a combination of 5 μM Iono , 1 μM Tg , 2 μM FCCP , and EDTA . Imaging continued at 1 Hz for 5 min . Mitochondria from Wt and DKO MEFs were subjected to the mitochondrial swelling assay for 5 min . They were then centrifuged and were analyzed for 39K , 43Ca , 23Na , 31P , and 24Mg . Each sample was analyzed with a 100× dilution . Typical standards ranging from 0 to 5000 ppb were used for the calibration curves , and most curves were determined to be R2 = 0 . 998–0 . 999 . Internal standards 9Be and 89Y were used during analysis . An Agilent ICPMS 7700× ( Agilent Technologies , Santa Clara , CA ) was used for the element-specific detection . The ICPMS was equipped with a microconcentric nebulizer supplied by Glass Expansion ( Pocasset , MA ) , a Scott-type double-channel spray chamber ( cooled to 2°C ) , a shield torch , an octopole collision/reaction cell with pressurized helium gas ( purity of 99 . 999% ) , a quadrupole mass analyzer , and an electron multiplier . Instrumental parameters: RF forward power = 1500 W , plasma Ar gas flow rate = 15 . 0 l/min , carrier Ar gas flow rate = 1 . 07 l/min , and He collision gas = 4 . 5 ml/min . Resulting data were analyzed using Agilent Mass Hunter ICPMS software . Cells at 80% confluence were harvested and mitochondria were isolated as previously described ( Murphy et al . , 2001 ) . Membrane patches of isolated mitochondria were excised after formation of a seal using micropipettes with approximately 0 . 3-µm tips and resistances of 10–30 MΩ at room temperature . Patching media was 150 mM KCl , 5 mM HEPES , 0 . 23 mM CaCl2 , 1 mM EGTA , pH 7 . 4 . Voltages were clamped with an Axopatch 200 amplifier and reported as pipette potentials . Permeability was typically determined from stable current levels and/or total amplitude histograms of 30 s of data at +20 mV . pClamp version 8 ( Axon Instruments , Sunnyvale , CA ) and WinEDR v2 . 3 . 3 ( Strathclyde Electrophysiological Software , Glasgow , UK ) were used for current analysis as previously described ( Martinez-Caballero et al . , 2009 ) . Sample rate was 5 kHz with 1–2 kHz filtration . The channel activity of rBax was characterized in DKO mitochondria and liposomes devoid of other proteins with and without its activator tBid . Micropipette tips were filled with media containing 10–100 ng/µl monomeric rBax and then backfilled with patching media . Hence , the actual Bax concentration was lower than that loaded in the tips . Seals were formed with these micropipettes on giant liposomes or mitochondria prepared as described previously ( Martinez-Caballero et al . , 2009 ) . Ion selectivity was determined through reversal potentials after a 150:30 mM KCl gradient was established across the patch as previously reported ( Pavlov et al . , 2001 ) . | In all multicellular plants and animals , cells are continuously dying and being replaced . There are a number of different types of cell death , but two of the best studied are apoptosis and necrosis . Apoptosis , sometimes referred to as ‘cell suicide’ , is a form of programmed cell death that is generally beneficial to the organism . Necrosis , however , occurs whenever cells are damaged—for example , due to a lack of oxygen—and can trigger harmful inflammation in surrounding tissue . Although the processes leading up to apoptosis and necrosis are very different , they both involve regulated changes in mitochondria—the organelles that supply cells with chemical energy . Mitochondria have a distinctive appearance , being enclosed by two membranes , the innermost of which is highly folded . During apoptosis , large pores form in the outer membranes of mitochondria . These pores are generated by two proteins—Bax and Bak—and they enable the mitochondrion to release proteins that activate processes involved in apoptosis . Pores also form in the mitochondrial membrane during necrosis . However , these mitochondrial permeability transition pores ( MPTPs ) occur simultaneously in both the inner and outer membranes and are thought to lead to swelling and rupture of mitochondria . Now , Karch et al . have shown that Bax and Bak are also involved in the formation of these permeability pores that underlie necrosis . When mouse cells that had been genetically modified to lack Bak and Bax were grown in cell culture , they were found to be resistant to substances that normally induce necrosis . Instead , their mitochondria continued to function normally , suggesting that MPTPs cannot form in the absence of Bak and Bax . Karch et al . then generated mice with heart cells that lack Bax and Bak , and deprived their hearts of oxygen to simulate a heart attack . Compared to normal mice , the genetically modified animals experienced less damage to their heart muscle , suggesting that the absence of Bax and Bak prevents cell death due to necrosis . If Bax and Bak are involved in both apoptosis and necrosis , inhibiting them could be a powerful therapeutic approach for preventing all forms of cell death during heart attacks or in certain degenerative diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2013 | Bax and Bak function as the outer membrane component of the mitochondrial permeability pore in regulating necrotic cell death in mice |
The ability to distinguish males from females is essential for productive mate selection and species propagation . Recent studies in Drosophila have identified different classes of contact chemosensory neurons that detect female or male pheromones and influence courtship decisions . Here , we examine central neural pathways in the male brain that process female and male pheromones using anatomical , calcium imaging , optogenetic , and behavioral studies . We find that sensory neurons that detect female pheromones , but not male pheromones , activate a novel class of neurons in the ventral nerve cord to cause activation of P1 neurons , male-specific command neurons that trigger courtship . In addition , sensory neurons that detect male pheromones , as well as those that detect female pheromones , activate central mAL neurons to inhibit P1 . These studies demonstrate that the balance of excitatory and inhibitory drives onto central courtship-promoting neurons controls mating decisions .
Across the animal kingdom , the ability to distinguish males from females is critical to select among potential mates . The specificity of mating decisions is exemplified by the Drosophila courtship ritual , in which males follow , sing to , and copulate with females but not males . Although much progress has been made in identifying the circuits that underlie mating decisions in the male fly brain , the sensory neurons that detect sex-specific cues and the pathways that they activate to generate sex-specific behaviors are incompletely understood . A major advance in elucidating the neural circuits that govern male mating decisions has come from the discovery that a male-specific splice form of the Fruitless trancriptional regulator ( FruM ) is expressed in peripheral and central neurons that drive courtship behavior ( Manoli et al . , 2005; Stockinger et al . , 2005 ) , arguing that FruM marks neural circuits for courtship . Studies of the function of FruM-positive neurons has led to the identification of olfactory and gustatory neurons that detect pheromones , as well as central neurons that drive behavioral subprograms of courtship ( Datta et al . , 2008; Ha and Smith , 2006; Kurtovic et al . , 2007; Lu et al . , 2012; Ruta et al . , 2010; Thistle et al . , 2012; Toda et al . , 2012; von Philipsborn et al . , 2011 ) . One set of neurons that has emerged as a central driver of male mating behavior is the group of P1 ( a . k . a . pMPe , pMP4 or pC1 ) neurons in the protocerebrum ( Cachero et al . , 2010; Kimura et al . , 2008; Lee et al . , 2002; Yu et al . , 2010 ) . Inducible activation of these neurons leads to sustained male courtship behaviors ( Inagaki et al . , 2014; Kohatsu et al . , 2011; Pan et al . , 2012; von Philipsborn et al . , 2011 ) . Moreover , these neurons are activated by female pheromones and this activation is partially inhibited in the presence of the male inhibitory pheromone , cis vaccenyl actate ( cVA ) ( Kohatsu et al . , 2011 ) . These data show that P1 neurons receive sensory cues signaling females or males and drive courtship decisions , suggesting that they may be command neurons for courtship behaviors . The sensory pathways that converge onto P1 neurons are poorly defined . Diverse sensory stimuli contribute to courtship decisions , including visual , auditory , and chemosensory cues . Important sensory cues detected primarily by contact chemosensory neurons are sex-specific cuticular hydrocarbons that act as pheromones . Multiple gustatory receptors and neurons have been implicated in pheromone detection ( Bray and Amrein , 2003; Koh et al . , 2014; Miyamoto and Amrein , 2008; Moon et al . , 2009; Watanabe et al . , 2011 ) . We and others recently showed that leg chemosensory neurons expressing the PPK23 pickpocket ion channel detect pheromones ( Lu et al . , 2012; Thistle et al . , 2012; Toda et al . , 2012 ) . PPK23 is expressed in sensory neurons of many leg chemosensory bristles , with generally two PPK23 cells per bristle . One cell responds selectively to male pheromones ( M cells ) and the other cell to female pheromones ( F cells ) ( Pikielny , 2012; Thistle et al . , 2012 ) . In contrast , the PPK25 channel is expressed in one of the two PPK23-positive cells per bristle , and PPK25 is required for cellular and behavioral responses to female pheromones , arguing that it selectively labels F cells ( Starostina et al . , 2012; Vijayan et al . , 2014 ) . Unlike other classes of gustatory neurons implicated in pheromone detection , PPK23 cells are Fruitless-positive ( Lu et al . , 2012; Thistle et al . , 2012; Toda et al . , 2012 ) . This suggested that it may be possible to trace pheromonal pathways from PPK23 cells in the periphery to the central nervous system by using FruM neurons as a guide . Here , we examine sensory pathways in the male brain , from pheromone-sensing cells on the legs to the ventral nerve cord to the protocerebrum , in order to elucidate the neural circuits that allow the male fly to distinguish between appropriate and inappropriate mates . These studies define sensory pathways that act as excitatory and inhibitory drives onto P1 , providing insight into the functional connectivity of the courtship circuit .
To examine pathways activated by female excitatory pheromones and male inhibitory pheromones , we focused on different subpopulations of PPK23 cells as specific sensory inputs . By GCaMP6s calcium imaging ( Chen et al . , 2013 ) of PPK23 cells in a background in which PPK25 cells were independently labeled , we first confirmed that the PPK25-positive cells ( F cells ) are tuned to female pheromones and the PPK25-negative cells ( M cells ) to male pheromones ( Figure 1A , B; Table 1 contains genotypes of flies used for all experiments ) . In addition , we found that F cells are the only leg neurons that express the vesicular glutamate transporter-Gal4 driver ( vGlut-Gal4 ) ( Daniels et al . , 2008 ) , suggesting that the two classes differ in their neurotransmitter profiles and providing an additional marker that selectively labels F cells ( Figure 1C , D ) . F cells and M cells also differ in their axonal projection patterns: F cells terminate in the ventral nerve cord ( VNC ) whereas M cells also have fibers that project to the subesophageal zone ( SEZ ) of the central brain ( Figure 1E , G ) . 10 . 7554/eLife . 11188 . 003Figure 1 . F and M cells comprise distinct chemosensory neuron classes . ( A ) F cells ( PPK23+ PPK25+ ) respond to female pheromones whereas M cells ( PPK23+ PPK25- ) respond to male pheromones; n = 7 bristles . The female pheromone mix contained 7 , 11-heptacosadiene and 7 , 11-nonacosadiene . The male pheromone mix contained 7-tricosene and cis-vaccenyl acetate . ( B ) GCaMP6s marks two PPK23 cells per bristle ( left ) . CD8::tdTomato marks the PPK25 cell ( middle ) . Maximum ΔF of both PPK23 cells to female ( red ) or male pheromones ( blue ) ( right ) . ( C ) vGlut ( magenta ) is expressed in one PPK23 cell ( green ) under each bristle . Transgenic flies with ppk23-LexA , lexAop-CD2::GFP , vGlut-Gal4 , UAS-CD8::tdTomato were used for cell labeling . ( D ) vGlut ( green ) and PPK25 ( magenta ) are expressed in the same cell under each bristle . Flies contained vGlut-LexA::VP16 , lexAop-CD2::GFP , ppk25-Gal4 , UAS-CD8::tdTomato . ( E ) Axons from F cells in the legs ( green ) do not project to the central brain ( left ) but instead terminate in the six leg neuromeres of the ventral nerve cord ( VNC , right ) . Brains are counterstained with nc82 ( magenta ) to show neuropil . ( F ) Expression of dTRPA1 in F cells promoted male-female courtship upon heat-evoked neural activation; n = 11–21/condition . Number of unilateral wing extensions per 10-minute trial was recorded . ( G ) M cells from the legs project to the SEZ in the brain ( arrow shows fibers entering from the cervical connective ) and VNC . Other SEZ axons come from the proboscis , with fibers entering from the labellar nerve . ( H ) dTRPA1-mediated activation of M cells suppressed male-female courtship . n = 10–12/condition . Number of unilateral wing extensions per 10-minute trial was recorded . Scale bars , 5 ( B ) , 10 , ( C , D ) 25 ( G , SEZ ) or 50 μm ( E , G , VNC ) . Data are Mean ± SEM . Kruskal-Wallis test , Dunn’s post-hoc ( A ) or 2-way ANOVA , Bonferroni post-hoc ( F , H ) . **p< 0 . 01 . See also Figure 1—figure supplement 1 , on selectively targeting F or M cells . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 00310 . 7554/eLife . 11188 . 004Figure 1—figure supplement 1 . Approach to selectively target F or M cells . ( A ) vGlut-QF ( green ) is expressed in PPK25 cells ( magenta ) , similar to vGlut-LexA in Figure 1D . Flies were vGlut-QF2 , QUAS-Gal80 , ppk25-Gal4 , UAS-CD8::tdTomato . ( B ) ( left ) , Ablating F cells ( ppk25-Gal4 ) using diphtheria toxin leaves M cells intact , with one GFP-positive cell/bristle rather than two . This strategy was used to label M cells in Figure 1G . ( middle ) , Suppressing Gal4-dependent expression in F cells ( ppk23-Gal4 , UAS-CD8::tdTomato , vGlut-LexA , lexAop-Gal80 ) leaves expression intact in M cells . This strategy was used to activate M cells in Figure 1H . ( right ) , Suppressing LexA-dependent expression in F cells ( ppk23-LexA , lexAop-CD2::GFP , vGlut-QF2 , QUAS-Gal80 ) leaves expression intact in M cells . This strategy was used to activate M cells in Figures 2B , 4B , and 5E . Different approaches were used to accommodate chromosome locations of transgenes . Scale bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 00410 . 7554/eLife . 11188 . 005Table 1 . Genotypes of flies used for experiments in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 005Figure panelGenotypeFigure 1A w-/y; ppk23-LexA/ppk25-Gal4; lexAop-GCaMP6S/UAS-CD8::tdTomato Figure 1B same as Figure 1A Figure 1C w-/y; ppk23-LexA/vGlut-Gal4; lexAop-CD2::GFP/UAS-CD8::tdTomato Figure 1D w-/y; vGlutMI04979-LexA::QFAD/ppk25-Gal4; lexAop-CD2::GFP/UAS-CD8::tdTomato Figure 1E w-/y; UAS-CD8::GFP/+; ppk25-Gal4/+Figure 1F see FigureFigure 1G w-/y; ppk23-LexA/ppk25-Gal4; lexAop-CD2::GFP/UAS-DTI Figure 1H see FigureFigure 1—figure supplement 1A w-/y; vGlutMI04979-QF2/ppk25-Gal4; QUAS-mCD8::GFP/UAS-CD8::tdTomato Figure 1—figure supplement 1B ( left ) w-/y; ppk23-LexA/ppk25-Gal4; lexAop-CD2::GFP/UAS-DTI ( middle ) w-/y; vGlutMI04979-LexA::QFAD/ppk23-Gal4; QUAS-Gal80/UAS-CD8::tdTomato ( right ) w-/y; vGlutMI04979-QF2/ppk23-LexA; QUAS-Gal80/lexAop-CD2::GFP Figure 2A w-/y; P1-Gal4-AD/UAS-CD8::GFP;P1-Gal4-DBD/+ Figure 2B F cell stim: w-/y; vGlutMI04979-LexA::QFAD/lexAop-P2X2; R71G01-Gal4/UAS-GCaMP6S M cell stim: w-/y; ppk23-LexA , lexAop-P2X2/ vGlutMI04979-QF2; R71G01-Gal4 , UAS-GCaMP6S/QUAS-Gal80 Figure 3A w-/y; UAS-CD8::GFP/+; R56C09-Gal4/+Figure 3B–C w-/y; +/+; R56C09-Gal4/UAS-DenMark , UAS-synaptotagmin-GFP Figure 3D see FigureFigure 3E w-/y; R56C09-LexA/ppk25-Gal4; lexAop-CD2::GFP/UAS-CD8::tdTomato Figure 3F w- , UAS-CD8::tdTomato/y; P1-Gal4-AD/R56C09-LexA; P1-Gal4-DBD/lexAop-CD2::GFP Figure 4A w-/y; UAS-CD8::GFP/+; R56C09-Gal4/+Figure 4B F cell stim: UAS-CD8::tdTomato/y; vGlutMI04979-LexA::QFAD/lexAop-P2X2; R56C09-Gal4/UAS-GCaMP6S M cell stim: UAS-CD8::tdTomato/y; ppk23-LexA , lexAop-P2X2/ vGlutMI04979-QF2; R56C09-Gal4 , UAS-GCaMP6S/QUAS-Gal80 F+M cell stim: UAS-CD8::tdTomato/y; ppk23-LexA/lexAop-P2X2; R56C09-Gal4/UAS-GCaMP6S Figure 4C w- , UAS-CD8::tdTomato/y; P1-Gal4-AD/R56C09-LexA; P1-Gal4-DBD/lexAop-CD2::GFP Figure 4D w-/y; UAS-GCaMP6S/R56C09-LexA; R71G01-Gal4/lexAop-Chrimson Figure 5A w-/y; ppk23-LexA/lexAop-P2X2; fru-Gal4/UAS-GCaMP6S Figure 5B w-/y; UAS>stop>CD8::GFP/lexAop-FLPL; R43D01-Gal4/fru-LexA Figure 5C see FigureFigure 5D see FigureFigure 5E F cell stim: w-/y; vGlutMI04979-LexA::QFAD/lexAop-P2X2; R43D01-Gal4/UAS-GCaMP6S M cell stim: w-/y; ppk23-LexA , lexAop-P2X2/ vGlutMI04979-QF2; R43D01-Gal4 , UAS-GCaMP6S/QUAS-Gal80 Figure 5F same as Figure 5e and F+M cell stim: ppk23-LexA/lexAop-P2X2; R43D01-Gal4/UAS-GCaMP6S Figure 5—figure supplement 1A w-/y; UAS-CD8::GFP/+; R43D01-Gal4/+Figure 5—figure supplement 1B w-/y; UAS-CD8::GFP/+; R43D01-Gal4/+Figure 5—figure supplement 1C w-/y; UAS>stop>nsyb-GFP 19a/lexAop-FLPL; R43D01-Gal4/fru-LexA Figure 5—figure supplement 1D w-/y; UAS>stop>Dscam17 . 1-GFP 19a/lexAop-FLPL; R43D01-Gal4/fru-LexA Figure 5—figure supplement 1E w-/y; UAS-CD8::GFP/ppk23-LexA; R43D01-Gal4/lexAop-myr::mCherry Figure 5—figure supplement 1F w-/y; ppk23-LexA/lexAop-P2X2; R43D01-Gal4/R71G01-Gal4 , UAS-GCaMP6S Figure 5—figure supplement 1G w-/y; R43D01-LexA/+; lexAop-CD2::GFP/+Figure 6A w- , UAS-CD8::tdTomato/y; P1-Gal4DBD/R43D01-LexA; P1-Gal4AD/lexAop-CD2::GFP Figure 6B see figure , genotype includes UAS-Dicer ( X ) Figure 6C see FigureFigure 6D w-/y; ppk23-LexA/lexAop-P2X2; R71G01-Gal4 , UAS-GCaMP6S/R43D01-Gal4 Figure 6E w-/y; R43D01-LexA/UASArcLight; R71G01-Gal4/lexAop-Chrimson Figure 6F same as Figure 6E To ask whether F or M cell activation is sufficient to modify courtship behavior , we used genetic strategies to express the heat-activated cation channel dTRPA1 ( Hamada et al . , 2008 ) selectively in F or M cells . A single male was placed in a chamber with a virgin female and number of single wing extensions by the male was monitored , as this motor subprogram occurs specifically during courtship song production . Males expressing dTRPA1 in F cells tested at 30°C ( a temperature that activates dTRPA1 ) had a significantly higher wing extension rate than genetically identical flies tested at the non-activating temperature of 23°C or control flies at either temperature ( Figure 1F ) . In contrast , males expressing dTRPA1 in M cells had a significantly lower wing extension rate at 30°C compared to genetic and temperature controls ( Figure 1H ) . Thus , F and M cells comprise genetically and anatomically distinct chemosensory neuron classes that respond to female or male pheromones and promote or inhibit courtship . The ability to genetically access and specifically activate cells responding to female or male pheromones provided the opportunity to trace sex-selective pathways in the male brain and examine the neural substrates for courtship decisions . We genetically accessed F cells with the ppk25-Gal4 , vGlut-LexA and vGlut-QF2 drivers , all of which specifically label F cells in the legs ( Figure 1C , D , and Figure 1—figure supplement 1 ) . To selectively access M cells ( PPK23-positive , PPK25-negative cells ) , we used a driver that labels both F and M cells ( ppk23-LexA ) while driving Gal80 with F cell drivers ( Figure 1—figure supplement 1 ) . The ATP-gated cation channel P2X2 ( Lima and Miesenbock , 2005 ) was selectively expressed in these sensory classes and ATP was applied to the legs for robust cell-specific activation . Male-specific , Fru-positive P1 neurons ( Cachero et al . , 2010; Kimura et al . , 2008; Lee et al . , 2002; Yu et al . , 2010 ) are prominent courtship-promoting neurons in the protocerebrum that trigger sustained courtship behaviors upon ectopic activation ( Inagaki et al . , 2014; Kohatsu et al . , 2011; Pan et al . , 2012; von Philipsborn et al . , 2011 ) . As P1 neurons have been shown to respond to hydrocarbon extracts from female and male abdomens ( Kohatsu et al . , 2011 ) , we tested whether F and M cells provided specific sensory inputs onto P1 neurons . P1 activity was monitored by GCaMP6s calcium imaging in live flies expressing P2X2 in both F and M cells ( F+M ) , F cells , or M cells while ATP was applied to the legs ( Figure 2A , B ) . Activation of F cells triggered calcium increases in P1 , demonstrating that sensory neurons that detect female pheromones activate courtship-promoting P1 neurons . In contrast , we observed no significant calcium response in P1 neurons upon M cell activation . Activating F+M cells using the same ppk23-LexA driver ( without expression of Gal80 in F cells ) caused robust calcium responses in P1 neurons . These experiments argue that F cells , but not M cells , activate P1 neurons . This is in contrast to a previous study that observed calcium increases in P1 neurons in response to male abdomens ( Kohatsu et al . , 2011 ) ; however , male abdomens may activate other sensory neurons in addition to M cells , such as Fru-negative pheromone-sensing neurons , gustatory neurons , or mechanosensory neurons . The selective activation of single classes of sensory cells allows us to disambiguate sensory cues and determine the contribution of specific sensory inputs . 10 . 7554/eLife . 11188 . 006Figure 2 . F cells activate courtship-promoting P1 neurons . ( A ) Male-specific P1 neurons ( green ) are located in the protocerebrum . Scale bar , 50 μm . Flies contained P1-Gal4DBD , P1-Gal4AD , UAS-CD8::GFP . ( B ) ATP-mediated stimulation of F+M cells ( ppk23-LexA , lexAop-P2X2 ) , F cells ( vGlut-LexA ) but not M cells ( ppk23-LexA , lexAop-P2X2 , vGlut-QF2 , QUAS-Gal80 ) triggered calcium increases in P1 neurons; n = 5–8/condition . Mock is no P2X2 . Traces on the left show averaged △F/F with mean in black and SEM shaded . Arrows indicate stimulus . Schematics show cells monitored with GCaMP6s ( green ) and connections tested . Data are also displayed as bar graph ( right ) . Differences in expression levels of ppk23-LexA and vGlut-LexA may contribute to different response magnitudes of F+M versus F cell stimulation . Mean ± SEM of maximum △F/F . Kruskal-Wallis test , Dunn’s post-hoc to mock , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 006 As F cells terminating in the VNC do not directly contact P1 , we screened existing Gal4 collections ( Gohl et al . , 2011; Jenett et al . , 2012 ) for neurons that might contact F cell projections and project to higher brain regions . We identified a pair of neurons with dendrites in the VNC and axons in the protocerebrum , marked specifically by R56C09-Gal4 , which we name Pheromone Projection Neuron class 1 ( PPN1 ) ( Figure 3A–C ) . PPN1 neurons have cell bodies on the dorsal surface of the third leg neuromere , send projections to all leg and wing neuromeres in the VNC , and terminate in the ventrolateral protocerebrum . PPN1 dendrites are in close proximity to PPK25 axons and PPN1 axons overlap with P1 fibers , as shown by double labeling experiments ( Figure 3E , F ) . To test whether PPN1 is involved in courtship behavior , we expressed dTRPA1 in PPN1 neurons and monitored male courtship towards females upon heat-induced neural activation . Consistent with a role in promoting courtship , activating PPN1 with dTRPA1 increased male courtship behavior toward females , as measured by unilateral wing extension rate ( Figure 3D ) . Unlike other neurons of the courtship circuit , PPN1 is not Fru-positive based on intersectional approaches with Fru-Flp and Fru-LexA and PPN1 projections are not sexually dimorphic ( data not shown ) . Nonetheless , the anatomical and behavioral studies suggest that PPN1 might transmit F cell activation to P1 to promote courtship . 10 . 7554/eLife . 11188 . 007Figure 3 . PPN1 neurons are courtship-promoting neurons in proximity to PP23 axons and P1 fibers . ( A ) PPN1 neurons have cell bodies in the third leg neuromere of the VNC and send projections to the six leg neuromeres and wing neuromere of the VNC and to the ventrolateral protocerebrum of the brain . R56C09-Gal4 drives expression of UAS-CD8::GFP exclusively in the pair of PPN1 neurons . ( B-C ) PPN1 has dendrites in the VNC ( B , DenMark , magenta ) and axons in the ventrolateral protocerebrum ( C , syt-GFP , green ) . B and C are from the same animal containing R56C09-Gal4 , UAS-DenMark , UAS-synaptotagmin-GFP . ( D ) Activation of PPN1 with dTRPA1 causes increased male-female courtship at 30°C; mean ± SEM , n = 16–30/condition , **p<0 . 01 ( 2-way ANOVA , Bonferroni post-hoc ) . ( E ) Overlap is observed in the VNC between PPN1 dendrites ( green ) and incoming PPK25 axons ( magenta ) . R56C09-LexA , lexAop-CD2::GFP , ppk25-Gal4 , UAS-CD8::tdTomato flies were used . F . Overlap between PPN1 ( green ) and P1 ( magenta ) in the anterior ventrolateral protocerebrum ( 50 μm collapsed Z-stack ) . Scale bars , 25 μm ( F ) 50 μm ( A-E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 007 To test whether PPN1 receives pheromonal signals , we stimulated F cells or M cells by ATP-mediated activation of P2X2 while monitoring calcium changes in PPN1 axons in the higher brain . These studies revealed calcium increases in PPN1 axons upon F cell stimulation but not M cell stimulation ( Figure 4A ) . As with P1 , activating F+M cells using the same ppk23-LexA driver ( without expression of Gal80 in F cells ) triggered robust calcium responses in PPN1 , further arguing that PPN1 neurons are downstream of F cells and not M cells . To more directly test whether F cell activity is transmitted to P1 by PPN1 , we expressed the red-shifted opsin Chrimson ( Klapoetke et al . , 2014 ) in PPN1 and monitored calcium changes in P1 . Activation of PPN1 with red light generated calcium responses in P1 ( Figure 4B ) . Red light had no effect on P1 activity in control animals not fed the essential cofactor retinal . Together , these experiments argue that female pheromones activate a neural pathway from F cells to PPN1 to P1 to drive courtship behavior . 10 . 7554/eLife . 11188 . 008Figure 4 . F cells activate courtship-promoting PPN1 neurons which activate P1 . ( A ) Calcium imaging of PPN1 axons while activating different sensory classes revealed that PPN1 is activated by F+M and F cell stimulation but not M cell stimulation; n = 5–7/condition . Mock is no P2X2 . Arrows indicate stimulus . ( B ) Chrimson-mediated activation of PPN1 triggers calcium increases in P1 by GCaMP6s calcium imaging; n = 5–6/condition . Arrows indicate stimulus . Schematics show cells monitored with GCaMP6s ( green ) and connections tested . Data are Mean ± SEM . Kruskal-Wallis test , Dunn’s post-hoc to mock ( A ) or Mann-Whitney test ( B ) . *p<0 . 05 , **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 008 As M cell stimulation did not activate P1 , we searched for other targets of F and M cells by monitoring calcium responses in all Fru neurons upon F+M cell stimulation , observing activity throughout the brain using spinning disk confocal microscopy . One set of neurons was prominently activated by F+M stimulation ( Figure 5A ) . These were anatomically identifiable as mAL neurons ( a . k . a . aDT2 , aDTb ) by their distinct arborization patterns ( Cachero et al . , 2010; Ito et al . , 2012; Kimura et al . , 2005; Yu et al . , 2010 ) . Based on their anatomy and neurotransmitter profile , mAL neurons have been proposed to be sexually dimorphic GABAergic interneurons that convey inhibitory courtship signals to the higher brain in males ( Koganezawa et al . , 2010 ) . However , the behavioral role of mAL neurons in courtship , the sensory stimuli that activate mAL , and the relationship between mAL and other components of the courtship circuit have not been determined . To examine these questions , we visually screened Gal4 lines ( Jenett et al . , 2012 ) and identified R43D01-Gal4 , which includes Fru-positive , GABAergic mAL neurons ( Figure 5B and Figure 5—figure supplement 1 ) . Consistent with the notion that mAL might be downstream of pheromone-sensing sensory neurons , we found that M cell axons and mAL dendrites overlap in the SEZ ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 11188 . 009Figure 5 . M cells and F cells activate courtship-suppressing mAL neurons . ( A ) Example image of a GCaMP6s ΔF heat map in fru-LexA neurons upon P2X2-mediated activation of F+M cells . ( B ) mAL neurons labeled by the intersection of R43D01-Gal4 and fru-LexA connect the SEZ and protocerebrum . ( C ) Activating mAL neurons with dTRPA1 suppresses courtship toward females; n = 10/condition . ( D ) Silencing mAL neurons with tetanus toxin or knocking down vGAT with R43D01-Gal4 induces male-male chaining; n = 8–10 groups/condition , 6–9 males per group . For C and D , >> means >stop> . Chaining index represents the fraction of time 3 or more males were courting over the 10-min trial . ( E ) P2X2-mediated stimulation of either F or M cells activates mAL neurons by GCaMP calcium imaging . Arrows indicate stimulus . ( F ) Maximum ΔF/F in mAL cell bodies; n = 5–9/condition . Mock is no P2X2 . Scale bars , 50 μm ( A , B ) . Data are Mean ± SEM , 2-way ANOVA , Bonferroni post-hoc ( C ) , Mann-Whitney test ( D ) , or Kruskal-Wallis test , Dunn’s post-hoc to mock ( F ) . *p<0 . 05 , ***p<0 . 001 . See also Figure 5—figure supplement 1 , for anatomical characterization of mAL neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 00910 . 7554/eLife . 11188 . 010Figure 5—figure supplement 1 . Characterization of mAL neurons . ( A ) mAL neurons labeled by R43D01-Gal4 ( green , CD8::GFP ) are immunopositive for GABA ( magenta ) . ( B ) mAL neurons labeled by R43D01-Gal4 ( green , CD8::GFP ) express FruM ( magenta ) . ( C ) mAL neurons have presynaptic termini ( green , nsyb-GFP ) in the higher brain and SEZ . ( D ) mAL neurons have dendrites ( green , Dscam17 . 1-GFP ) primarily in the SEZ . ( E ) M cell axons ( magenta , myr::mCherry ) interdigitate with mAL fibers ( green , CD8::GFP ) in the SEZ . ( F ) A representative image of the mAL axonal tract in the higher brain before and after 2-photon-mediated lesioning . ( G ) R43D01-LexA labels mAL neurons and some olfactory sensory neurons ( green , CD2::GFP ) . This line was used in Figure 6E . Scale bars , 5 μm ( A ) , 10 μm ( B , F ) , 25 μm ( E ) , 50 μm ( C , D , G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 010 To test whether mAL neurons participate in courtship behavior , we conditionally activated or inactivated them using the genetic intersection of R43D01-Gal4 and Fru-LexA and monitored courtship behavior . Activation of mAL neurons with dTRPA1 greatly suppressed courtship toward females compared to controls ( Figure 5C ) , as measured by unilateral wing extension rate . Inactivation of mAL neurons with tetanus toxin caused robust male-male chaining , a behavior in which three or more males serially court each other ( Figure 5D ) , and which was almost never observed in control animals . Furthermore , expression of RNAi against the vesicular GABA transporter ( vGAT ) using R43D01-Gal4 also caused male-male chaining ( Figure 5D ) , arguing that GABA release from mAL inhibits courtship . Thus , activation of mAL inhibits courtship , whereas inactivation enhances courtship , demonstrating an inhibitory role for mAL in courtship decisions . Finally , we tested the specificity of the response of mAL neurons to pheromone sensory cell stimulation , using the R43D01-Gal4 line to express GCaMP6s and the same drivers used for the PPN1 and P1 experiments to express P2X2 in M and F cells ( Figure 5E , F ) . Surprisingly , both M cells and F cells activated mAL , arguing that detection of female as well as male pheromones provides an inhibitory courtship drive . As F cells additionally activated PPN1 and P1 courtship-promoting neurons , whereas M cells did not , this suggests that the balance between excitation and inhibition underlies the decision to court . How do mAL neurons inhibit courtship ? P1 neurons are in close proximity to mAL termini ( Figure 6A ) , suggesting that they may be candidate targets of GABAergic mAL neurons . To test whether P1 neurons receive inhibitory signals , we generated flies containing an RNAi against the GABA-A receptor subunit Resistant to dieldrin ( Rdl ) ( Ffrench-Constant et al . , 1991 ) in P1 neurons and examined the behavioral consequence . Males expressing Rdl RNAi in P1 neurons displayed increased courtship toward other males , arguing that P1 neurons receive GABAergic inhibition ( Figure 6B ) . In addition , whereas Chrimson-mediated activation of P1 neurons induced courtship behavior in solitary males , co-activation of P1 and R43D01-Gal4 neurons suppressed this behavior ( Figure 6C ) , arguing that mAL suppresses P1-mediated courtship . To more directly test whether inhibitory signals from mAL impinge on P1 , we simultaneously stimulated F and M cells and monitored activity in P1 before and after 2-photon guided lesioning of mAL axons ( Figure 6D and Figure 5—figure supplement 1 ) . Indeed , simultaneous stimulation of F and M cells caused transient activation of P1 and lesioning mAL axons significantly increased P1 activation . 10 . 7554/eLife . 11188 . 011Figure 6 . mAL neurons functionally and behaviorally inhibit P1 neurons . ( A ) Overlap between mAL ( green ) and P1 ( magenta ) in the superior lateral protocerebrum ( collapsed 97-μm stack ) . Scale bar is 25 μm . ( B ) Knockdown of GABAA receptor Rdl in P1 neurons induces male-male chaining; n = 10/condition . ( C ) Chrimson-mediated activation of P1 neurons causes wing extension ( wing ext . ) in solitary males , which is suppressed by co-activation of neurons expressing R43D01-Gal4 . ( D ) . Lesioning mAL axons increases GCaMP response in P1 upon P2X2-mediated stimulation of F+M cells . ( E ) Chrimson-mediated activation of mAL neurons causes P1 hyperpolarization , as detected by increased ArcLight fluorescence . Arrow indicates laser . F . Maximum ΔF/Fin P1; n = 10–11/condition . Data are Mean ± SEM . Mann-Whitney test ( B , F ) , Fisher’s exact test ( C ) , or paired t-test ( D ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 011 These data suggest that mAL neurons inhibit P1 neurons via GABA-A receptors . To test directly whether mAL input onto P1 causes hyperpolarization , we expressed ArcLight ( Cao et al . , 2013 ) , a fluorescent voltage sensor , in P1 neurons and monitored its fluorescence using 2-photon imaging ( Figure 6E ) . Chrimson-mediated activation of mAL neurons , using R43D01-LexA ( Figure 5—figure supplement 1 ) , evoked a rapid increase in ArcLight fluorescence in P1 neurons , indicating hyperpolarization . No fluorescence change was observed in control flies lacking the R43D01-LexA transgene . These experiments demonstrate that mAL neurons provide an inhibitory drive onto P1 neurons .
This work identifies pheromone-responsive neural circuits underlying mating decisions ( Figure 7 ) . We used cell-specific activation to determine sensory pathways that impinge on courtship-promoting P1 neurons , providing insight into the functional connectivity of the courtship circuit . Female pheromones trigger neural pathways that excite and inhibit P1 , with PPN1 providing an excitatory drive onto P1 and mAL providing an inhibitory drive . F cell activation leads to P1 activation and increased courtship behavior , arguing that the sum of inputs onto P1 produces excitation or that the sequence of inputs , i . e . , fast excitation followed by inhibition , provides a temporal window for excitation . In contrast , M cells activate mAL neurons to inhibit P1 and inhibit courtship behavior . These studies argue that the balance of excitation and inhibition onto P1 neurons is different following F cell or M cell activation: F cell activation leads to overall P1 activation whereas M cell activation leads to overall P1 inhibition . 10 . 7554/eLife . 11188 . 012Figure 7 . Schematic of courtship-promoting and courtship-inhibiting circuits activated by F and M cells . F cells on the leg express PPK23 , PPK25 , and VGlut , and respond to female pheromones . M cells on the leg express PPK23 and respond to male pheromones . The M cell neurotransmitter is unknown . F cells activate PPN1 , a class of projection neuron with cell bodies and dendrites in the VNC and long-range axonal projections to the ventrolateral protocerebrum . PPN1 axons are in close proximity to P1 fibers , and PPN1 activation causes activation of P1 . M cells on the leg activate GABAergic mAL neurons , which connect the SEZ and superior lateral protocerebrum . mAL axons interdigiate with P1 fibers , and mAL acitvation causes hyperpolarization of P1 , likely through the GABA-A receptors containing the Rdl subunit . F cells also provide an inhibitory drive onto P1 via mAL . The contact between F cells and mAL is not direct ( dotted line ) . Other connections may not be monosynaptic . DOI: http://dx . doi . org/10 . 7554/eLife . 11188 . 012 Our calcium imaging studies provide strong support that the pathways identified in this study are activated by F and M cell stimulation , and our behavioral experiments demonstrate that activation of these neurons directly contributes to courtship decisions . However , we do not exclude the possibility that additional intermediary neurons may also be activated by F or M cells . Indeed , a recent study identified vAB3 neurons , with projections in the first leg neuromere in the VNC , the SEZ and protocerebrum , as activated by female but not male pheromones ( Clowney et al . , 2015 ) . This study also showed that mAL neurons are activated by male and female abdomens and provided evidence that mAL inhibits P1 . Our independent observations are consistent with this work and extend the findings by providing behavioral evidence that each identified component of the circuit promotes or inhibits courtship as predicted by its response properties , by using cell-type specific activation strategies coupled with cell-type specific imaging studies to directly test potential connections , and by identifying PPN1 as a novel neural component of the courtship circuit . Together , the studies argue that pheromones activate multiple excitatory and inhibitory interneurons that impinge on P1 . A caveat of our studies is that they rely on ectopic expression of reporters , and expression levels of P2X2 and GCaMP6s may influence the ability to detect responding cells . Nevertheless , the same driver was used to express P2X2 in M cells in all calcium-imaging experiments , but application of ATP only elicited calcium responses in mAL neurons . Furthermore , the ppk23-LexA driver ( in F+M cells ) that we used for M cell activation was sufficient to produce ATP-mediated activation of P1 and PPN1 , and these responses were abolished in the presence of Gal80 in F cells . Thus , the observation that M cells activate mAL but not P1 or PPN1 is unlikely due to technical limitations . This study demonstrates a specific computational logic used by the nervous system to integrate different sensory inputs . Pheromones provide excitatory or inhibitory drives onto P1 , such that P1 activity reflects the integration of positive and negative sensory inputs , with female pheromones causing net excitation and male pheromones causing net inhibition . P1 also integrates inputs from other sensory systems , as P1 neurons respond to visual stimuli ( Kohatsu and Yamamoto , 2015 ) and olfactory pheromones ( Clowney et al . , 2015; Kohatsu et al . , 2011 ) . Thus , diverse sensory stimuli may alter the weight of excitation versus inhibition onto P1 and bias the decision to court . Altering the weights of excitation or inhibition through experience or evolution is an appealing strategy to dynamically modulate the response to potential mates or to tune attraction to conspecifics . Taken together , these studies reveal an elegant strategy used by the nervous system in which excitatory and inhibitory inputs directly converge onto a common output to control a significant behavioral decision .
Single bristle imaging was performed as previously described ( Thistle et al . , 2012 ) . Flies containing ppk23-LexA , LexAop-GCaMP6S , ppk25-Gal4 , UAS-CD8::tdTomato were placed in a custom imaging chamber and their forelegs were secured with wax . Female ( 7 , 11-heptacosadiene and 7 , 11-nonacosadiene ) or male ( 7-tricosene and cis-vaccenyl acetate ) pheromone mixes ( 100 ng/μL , Cayman Chemical , Ann Arbor , MI ) were applied to single bristles on the three distal leg segments for 30 s . GCaMP6s responses were captured using a 3i spinning disk confocal system equipped with a 20x air objective and 1 . 6x optical zoom . Responses in the two PPK23 cells under each bristle were analyzed based on their expression of CD8::tdTomato; M cells were tdTomato-negative , and F cells were tdTomato-positive . The change in fluorescence in each cell was calculated as follows: 100* ( ( Ft-F0 ) /F0 ) , where F0 is the mean fluorescence intensity during the 4 s prior to stimulation . The heat map in Figure 1B was created in Fiji . The average fluorescence intensity of the 10 frames preceding each stimulation ( female and male mixes ) was subtracted from the frame at which the fluorescence intensity of the responding cell was at a maximum . The resulting images from each stimulation were then merged into a two-channel image . The channels corresponding to the female or male pheromone stimulation were pseudocolored red and blue , respectively . GCaMP imaging experiments were performed as previously described ( Harris et al . , 2015 ) . Virgin males were collected at eclosion and were aged in isolation for 2 to 6 days before imaging . They were briefly anaesthetized with CO2 and placed into a small slit on a custom-built plastic mount at the neck such that the head was isolated from the rest of the body . The head was then immobilized using nail polish . Two small pieces of plastic were affixed with nail polish to the underside of the plastic mount on either side of the thorax , such that the legs were forced into a forward-facing position . The proboscis was covered with wax to prevent labellar taste input . The head cuticle was dissected with fine forceps in ice-cold adult hemolymph-like solution ( AHL ) ( Wang et al . , 2003 ) , and obscuring air sacs and other debris were removed . Eyes were damaged or removed to minimize visual input from the imaging laser . A coverglass was placed at a 45-degree angle to the plane of the plastic mount such that the head was isolated from the rest of the body . GCaMP6s responses were captured using a fixed-stage 3i spinning disk confocal system equipped with a 20x water objective and 1 . 6x ( mAL and P1 ) or 2 . 5x ( PPN1 ) optical zoom and a 488 nm laser . During the stimulation , stacks of 15–20 Z-slices ( 1–1 . 5 μm/Z-slice , for mAL and P1 ) or 8–12 Z-slices ( 0 . 5–0 . 8 μm/Z-slice , for PPN1 ) were obtained with a 100-ms exposure per Z-slice , resulting in each imaging volume/timepoint being acquired every 1 . 7–3 . 9 s . For each trial , 20 imaging timepoints were acquired . For P2X2-mediated stimulation of M and/or F cells , ~4 μL of 100 mM ATP ( adjusted to pH 7 ) was pipetted onto a small cube of 2% agar , which was placed on the end of a glass capillary ( OD 1 . 0 mm , ID 0 . 78 mm ) . The capillary was placed into an electrode holder that was secured to a micromanipulator . The agar cube was advanced such that it touched the flies’ legs at timepoint 6 of 20 . The cube was left within reach of the flies’ legs for 3 timepoints before being removed . F and M cells were exogenously activated to ensure specific stimulation of different sensory classes . Delivery of synthetic pheromones requires solubilization in ethanol or hexane , solvents that dissolve the fly’s endogenous cuticular hydrocarbons , which may independently activate F or M cells . It was possible to apply pheromone solutions to the tip of a single sensory bristle without detriment ( Figure 1A ) . For PPN1 imaging , because GCaMP6s baseline fluorescence was very low , a red reporter ( UAS-CD8::tdTomato ) was included in order to visualize axonal endings . One stack of 15–20 Z-slices from the 561 nm laser line was obtained at the beginning of each imaging session . These were later used to define regions of interest for imaging analysis . For PPN1 activation experiments , we fed isolated adult male flies for a minimum of 3 days on standard fly food supplemented with all-trans retinal ( final concentration 400 μM ) . These flies , as well as control males fed normal food , were kept in constant darkness until the experiment . Imaging was performed similarly to above ( “GCaMP6s imaging” ) except that a 635 nm laser ( Laserglow , Canada ) was directed at the thorax . The laser was turned to its highest power ( ~0 . 01 mW/mm2 ) but was in standby mode until frame 6 of 20 , at which time the key was turned to open the shutter . The laser was left on for 6 frames . During imaging , a 525/45 bandpass emission filter ( Semrock , Rochester , NY ) was used to prevent the 635 nm laser light from interfering with the GCaMP signal . In some cases , we found that the 488 nm imaging laser was sufficient to activate Chrimson and trigger calcium responses in downstream neurons ( i . e . , ppk23>Chrimson , P1>GCaMP6s ) . In addition , we found that Chrimson was weakly activated in the absence of retinal ( see ArcLight imaging below ) . However , these phenomena were not observed in PPN1>Chrimson flies , likely due to weak PPN1 or Chrimson transgene expression . For Figures 2 , 4 and 6D , calcium imaging data were processed in Fiji . For PPN1 , a red anatomy scan ( with 561nm laser ) measuring CD8::tdTomato fluorescence was taken prior to GCaMP calcium imaging . A maximum intensity Z-projection for the anatomy scan and each GCaMP timepoint was used for analysis . The anatomy projection was used to draw an ROI covering the axonal region in the ventrolateral protocerebrum ( “anatomical ROI” ) . A second ROI was drawn in a region lacking both tdTomato and G-CaMP signal ( “background ROI” ) . Mean fluorescence levels from the background ROI was subtracted from the anatomical ROI at each GCaMP timepoint resulting in the fluorescence trace over time: Ft . △F/F ( % ) was calculated as follows: 100* ( ( Ft-F0 ) /F0 ) , where F0 is the mean fluorescence intensity during time points 2 to 5 . For P1 fibers , △F/F ( % ) was calculated in the same way , except that in place of an anatomical CD8::tdTomato scan , “anatomical ROIs” covering P1 commisural fibers were drawn using the maximum projection across time of the GCaMP signal . Maximum △F/F ( % ) was calculated by subtracting the average △F/F ( % ) of the 3 timepoints preceding the stimulation from the maximum △F/F ( % ) of the 4 timepoints following the stimulation . Due to unavoidable differences in the background fluorescence between pre- and post-ablation imaging scans , the △F/F values presented in Figure 6D were calculated without background subtraction . For Figure 5 , calcium imaging data were processed in Matlab . ROIs were drawn around mAL cell bodies in single slices . △F/F ( % ) was calculated as follows: 100* ( ( Ft-F0 ) /F0 ) , where F0 is the mean fluorescence intensity during time points 2 to 5 and Ft is the fluorescence at each timepoint . Maximum △F/F ( % ) was calculated by subtracting the average △F/F ( % ) of the 3 timepoints preceding the stimulation from the maximum △F/F ( % ) of the 4 timepoints following the stimulation . For the heatmap in Figure 5A , △F values were calculated for each pixel in each slice at each timepoint , generating a 4-dimensional data set . These data were collapsed spatially into a 3-dimensional data set using a maximum intensity projection in the Z dimension . The heat map represents the maximum △F values that occurred during stimulus ( timepoints 6–9 ) . This heatmap was overlaid on a grayscale image that is the maximum intensity projection of the average baseline fluorescence ( timepoints 2–5 ) . The color bar scale represents the minimum ( blue ) to maximum ( dark red ) △F . For all GCaMP data , averaging the △F/F ( % ) traces across animals required re-sampling the individual △F/F ( % ) traces at 10 Hz ( completed with Matlab using a linear interpolation ) , due to the variable duration of timepoints between animals . Ablations were performed on a Zeiss LSM 780 NLO AxioExaminer microscope . Flies expressed GCaMP6s in both mAL and P1 neurons , visualized using 488 nm light . A rectangular ROI ( approximately 5 μm x 15 μm ) was drawn to cover the width of the tract carrying mAL axons in a single z-section located in the middle of the tract . We then scanned the ROI 10 times ( 3 . 15 μs pixel dwell time ) with intense 760 nm light ( ~50 mW at the front lens ) . Lesions were considered successful when the mAL axonal tract became discontinous . Mock ablations were performed identically except that the ROI was moved lateral to the mAL axonal tract ( at the edge of the optic lobe ) . All ablations were performed bilaterally . Pre- and post-ablation stimulation of F and M cells were performed on a spinning disk microscope as described above . We waited 10–20 min after the ablation to stimulate the fly with ATP ( “post-ablation” condition ) . Isolated adult male flies were fed standard fly food supplemented with all-trans retinal ( final concentration 400 μM ) for a minimum of 2 days and were kept in constant darkness until the experiment . Three- to 5-day-old flies were prepared as described above ( GCaMP6s imaging ) . Imaging was performed on a Zeiss LSM 780 NLO AxioExaminer microscope . To find the region of greatest mAL-P1 overlap , Chrimson . mVenus in mAL neurons was briefly imaged with low-intensity 514 nm light , and the ROI to be scanned ( approximately 30 μm x 30 μm ) was drawn around the distal mAL axons where they interdigitate with P1 fibers ( “signal ROI” ) . A second region of interest ( “background ROI” ) was drawn in an area lacking ArcLight or Chrimson . mVenus fluorescence . ArcLight was excited with 925 nm light and scanned at approximately 15 Hz . To activate mAL neurons , flies were stimulated 5 times ( ~5 s/stimulation , ~30 s between stimulations ) with a 635-nm laser ( Laserglow , Canada , ~0 . 01 mW/mm2 ) . We observed weak responses in flies expressing Chrimson in mAL but not fed retinal . These responses were significantly smaller than the responses in flies fed retinal . To calculate △F/F of P1 ArcLight signal , the background ROI intensity trace was first subtracted from the signal ROI intensity trace , resulting in F . For each fly , the 5 laser stimulations were then aligned such that laser onset for each stimulation was t = 0 and the average was taken . △F/F ( % ) for each animal was calculated as follows: 100* ( ( F-F0 ) /F0 ) , where F0 is the mean fluorescence intensity over the period from 0 . 7 to 2 . 6 s preceding the stimulation . Maximum △F/F ( % ) was calculated by subtracting the average △F/F ( % ) of 2 s preceding the stimulation from the maximum △F/F ( % ) of 2 s following the stimulation . Courtship behavior experiments were performed essentially as described ( Thistle et al . , 2012 ) , with the following modifications: male-female assays were recorded for 10 min; assays involving UAS-dTRPA1 were performed at room temperature ( ~23°C ) and 30°C; for Figure 3E , due to the relative weakness of UAS>stop>TRPA1myc , male flies were pre-incubated at 30°C for 5 min before being presented with a female; assays involving Chrimson were performed under white light ( ~0 . 05 mW/mm2 ) with male flies fed 400 μM all-trans retinal for a minimum of 3 days in the dark . We found that bright white light was sufficient to activate Chrimson and cause behavioral phenotypes . Antibody staining and immunohistochemistry were performed as previously described ( Wang et al . , 2004 ) . The following primary antibodies were used: rabbit anti-GFP ( Invitrogen , Carlsbad , CA , 1:1 , 000 ) , mouse anti-GFP ( Invitrogen , Carlsbad , CA , 1:1 , 000 ) , mouse anti-nc82 ( Hybridoma Bank , Iowa City , IA , 1:500 ) , rabbit anti-RFP ( Clontech , Mountain View , CA , 1:500 ) ; rabbit anti-GABA ( Sigma-Aldrich , St . Louis , MO , 1:1 , 000 ) ; rabbit anti-FruM ( 1:100 ) . For GRASP experiments , we used a mouse monoclonal antibody that specifically recognizes reconstituted GFP ( Sigma-Aldrich , St . Louis , MO , 1:200 ) . Secondary antibodies were Alexa Fluor goat anti-mouse 488 , goat anti-rabbit 488 , goat anti-mouse 568 , goat anti-rabbit 568 ( Life Technologies , Carlsbad , CA , 1:100 ) . To generate R43D01-LexA , 1157 bp fragment from genomic DNA , including the entire R43D01 tile from the FlyLight collection ( Pfeiffer et al . , 2012 ) , was amplified using the primers ttgagcacggatttcagcag and ggggtcctcaaatgtgtcgatttgt . This fragment was recombined into the pBPLexA::p65Uw plasmid ( Pfeiffer et al . , 2010 ) , and inserted into the VK00018 landing site ( Venken et al . , 2006 ) . | Courtship displays are seen throughout the animal kingdom . For example , male birds-of-paradise are perhaps best known for the elaborate dances they use to attract a mate . Male fruit flies , belonging to the species Drosophila melanogaster , also perform courtship toward female flies . However , male flies do not court other males . Previous studies have shown that sex-specific chemical signals ( or pheromones ) are important cues that males use to direct courtship towards females . Researchers have previously identified two sets of sensory neurons that detect pheromones: one set detects female pheromones and promotes courtship , while the other detects male pheromones and inhibits courtship . However it was unclear how these sensory neurons controlled courtship behavior . Now , Kallman et al . have studied the circuits of neurons in the fruit fly that promote or inhibit courtship when a fly detects a pheromone . The experiments identified several pathways of neurons in the brain of male Drosophila that respond to female and male pheromones . These pathways send signals that either excite or inhibit a central target , called P1 neurons . Female pheromones activated a pathway that activates the P1 neurons , whereas male pheromones activate another pathway that inhibits the P1 neurons . Kallman et al . suggest that the balance of these excitatory and inhibitory signals controls a fly’s decision to court . Following on from this work one of the next challenges will be to identify the neural circuits that act downstream of the P1 neurons to control courtship . Future studies could also explore how P1 neurons integrate signals from different senses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Excitation and inhibition onto central courtship neurons biases Drosophila mate choice |
Cholera toxin ( CT ) enters and intoxicates host cells after binding cell surface receptors using its B subunit ( CTB ) . The ganglioside ( glycolipid ) GM1 is thought to be the sole CT receptor; however , the mechanism by which CTB binding to GM1 mediates internalization of CT remains enigmatic . Here we report that CTB binds cell surface glycoproteins . Relative contributions of gangliosides and glycoproteins to CTB binding depend on cell type , and CTB binds primarily to glycoproteins in colonic epithelial cell lines . Using a metabolically incorporated photocrosslinking sugar , we identified one CTB-binding glycoprotein and demonstrated that the glycan portion of the molecule , not the protein , provides the CTB interaction motif . We further show that fucosylated structures promote CTB entry into a colonic epithelial cell line and subsequent host cell intoxication . CTB-binding fucosylated glycoproteins are present in normal human intestinal epithelia and could play a role in cholera .
The bacterium Vibrio cholerae is the etiological agent of cholera ( Foster and Baron , 1996 ) . Cholera toxin ( CT ) is secreted by V . cholerae and is the direct cause of the profuse , watery diarrhea that characterizes fatal cholera . CT is a heterohexamer comprising one copy of cholera toxin subunit A ( CTA ) and five copies of subunit B ( CTB ) . Mechanistic studies have yielded the following model for how CT intoxicates host cells ( Sánchez and Holmgren , 2008; Lencer , 2003 ) . The CTB subunits of the holotoxin bind receptors on the surface of host enterocytes , enabling endocytosis of CT . CT follows a retrograde trafficking pathway to the ER where it is disassembled to release CTA . CTA enters the cytoplasm and catalyzes ADP-ribosylation of the α-subunits of heterotrimeric GTP-binding proteins ( Gαs ) . The resulting extended activation of Gαs leads to increased activity of adenylate cyclase , raising intracellular cAMP levels . Elevated cAMP causes activation of chloride channels and chloride efflux , followed by massive secretion of water and ions into the intestinal lumen . Affected individuals can experience rapid and severe dehydration , sometimes leading to death ( Foster and Baron , 1996 ) . The initial and required step in host cell intoxication is recognition of cell surface receptors by CT . In the 1970s , the ganglioside GM1 was identified as a host cell receptor for CT . A role for gangliosides was first postulated when Van Heyningen et al . discovered that a lipid extract from the brain inhibited CT activity ( van Heyningen et al . , 1971 ) ; subsequently , multiple groups showed that purified gangliosides inhibited CT binding , with GM1 the most potent inhibitor ( Cuatrecasas , 1973; Holmgren et al . , 1973; King and van Heyningen , 1973 ) . To test whether GM1 could function as a receptor , exogenous GM1 was incorporated into host cell membranes , where it was shown to increase sensitivity to toxin , ( Cuatrecasas , 1973 ) even sensitizing toxin-resistant cells ( Moss et al . , 1976 ) . Holmgren and co-workers examined intestinal mucosa from several species and found that the extent of CT binding correlated with GM1 content ( Holmgren et al . , 1975 ) . Further , addition of exogenous GM1 to intestinal mucosa resulted in increased secretory activity in response to CT stimulation , implying that GM1 serves as a functional receptor . Recognition of GM1 occurs exclusively through the CTB subunit . Indeed , the high affinity CTB-GM1 interaction has been extensively characterized through binding assays ( Kuziemko et al . , 1996 ) and x-ray crystallography analysis ( Merritt et al . , 1994 ) . CTB is closely related to the B subunit of E . coli heat-labile toxin ( LTB ) at the levels of sequence , ( Dallas and Falkow , 1980 ) structure , ( Sixma et al . , 1991 ) and function ( Spangler , 1992 ) . While LTB is known to bind both GM1 and glycoprotein receptors , GM1 is commonly described to be the sole host cell receptor recognized by CTB ( Foster and Baron , 1996 ) . However , a variety of experimental approaches have pointed to the possibility that CTB may also recognize glycoproteins present on mammalian cells ( Morita et al . , 1980; Monferran et al . , 1990; Balanzino et al . , 1994; Platt et al . , 1997; Hansen et al . , 2005; Blank et al . , 2007; Day et al . , 2012 ) . Indeed , CTB binding to cells does not uniformly parallel GM1 levels , implying the existence of additional CTB-binding molecules ( Platt et al . , 1997; Yanagisawa , 2006 ) . Moreover , GM1 binding does not always correlate with intoxication . For example , treatment of intestinal mucosa with V . cholerae sialidase yielded more GM1 but had no effect on toxin sensitivity ( Holmgren et al . , 1975 ) . Also , a point mutant of CTB ( H57A ) was shown to maintain GM1 binding but the corresponding holotoxin did not intoxicate host cells ( Aman et al . , 2001 ) . Finally , a recent analysis of a normal human intestinal epithelia found that GM1 comprises only 0 . 01% of the glycosphingolipid content , raising the question of whether its concentration in enterocytes is sufficient to account for intoxication by CT ( Breimer et al . , 2012 ) . Here we report that fucosylated molecules and glycoproteins are recognized by CTB and can function as receptors in host cell intoxication . Glycoproteins , not gangliosides , are responsible for the majority of CTB binding to human colonic epithelial cell lines . Using a metabolically incorporated photocrosslinking sugar analog , we isolated and identified one CTB-interacting glycoprotein , CEACAM5 . The carbohydrate portion of the glycoprotein , not the amino acids , provides the CTB interaction motif . We show that fucose-containing glycans are recognized by CTB , resulting in internalization , the first step in host cell intoxication . Finally , we report evidence suggesting that fucosylated glycoconjugates recognized by CTB are present in normal human tissue . These results shed new light on mechanisms by which CT can enter and intoxicate host cells . In addition , the demonstration that CTB recognizes molecules other than GM1 has important implications for the interpretation of experiments where CTB is used to study the organization of lipids in the plasma membrane . Overall , these observations reveal a previously unrecognized , and potentially physiologically relevant , molecular mechanism for CT entry into epithelial cells .
Previously , we reported a cell-permeable precursor sugar ( Ac4ManNDAz ) that can be metabolized to a photocrosslinking sialic acid analog ( SiaDAz ) and incorporated into glycoconjugates – both glycoproteins and glycolipids – in place of natural sialic acids ( Figure 1A ) ( Tanaka and Kohler , 2008 ) . Culturing Jurkat cells , a human T cell line , with Ac4ManNDAz results in production of SiaDAz-modified gangliosides , including SiaDAz-modified GM1 ( Bond et al . , 2010 ) . After adding CTB to these SiaDAz-producing cells and applying UV radiation , we observed crosslinking of CTB to GM1 , consistent with the idea that GM1 is the CT receptor ( Bond et al . , 2010 , 2011 ) . Here we repeated the CTB crosslinking experiment in Jurkat cells , and also in two colonic epithelial cell lines , T84 and Colo205 . We chose T84 cells because they are commonly used in studies of host cell intoxication by CT , ( Lencer , 1992 ) and Colo205 cells as a second model of human colonic epithelia . By anti-CTB immunoblot analysis of Jurkat cell lysates , we confirmed detection of a CTB-containing species whose apparent mass ( ∼13 kDa ) matches the molecular weight of a CTB-GM1 complex ( Figure 1B ) . In contrast , the CTB-GM1 complex was absent in lysates from both colonic epithelial cell lines ( Figure 1B ) . 10 . 7554/eLife . 09545 . 003Figure 1 . Products of SiaDAz-mediated crosslinking of CTB depend on cell type . ( A ) Photocrosslinking sialic acid ( SiaDAz ) is produced by culturing cells with Ac4ManNDAz . SiaDAz is incorporated into glycolipids and glycoproteins that are displayed on the cell surface . CTB is added to cells . Application of 365 nm radiation causes activation of the diazirine crosslinker and results in covalent crosslinking between CTB and neighboring SiaDAz-modified glycoconjugates . Crosslinked complexes can be observed by immunoblot , or purified and characterized by LC-MS/MS analysis . ( B ) Jurkat , T84 , and Colo205 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Lysates were analyzed by 15% SDS-PAGE immunoblot with anti-CTB antibody . Red boxes highlight crosslinked complexes not present in control lanes . ( C ) Jurkat , T84 , and Colo205 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Lysates were analyzed by 6% SDS-PAGE immunoblot with anti-CTB antibody ( for Jurkat and T84 samples ) or anti-CT antibody ( for Colo205 samples ) . Red boxes highlight crosslinked complexes not present in control lanes . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 00310 . 7554/eLife . 09545 . 004Figure 1—figure supplement 1 . SiaDAz-mediated crosslinking of CTB in additional cell types . ( A ) hCMEC/D3 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Lysates were analyzed by 13% SDS-PAGE immunoblot with anti-CTB antibody . ( B ) HBEC cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Lysates were analyzed by 16% SDS-PAGE immunoblot with anti-CTB antibody . Images were acquired on a Bio-Rad Chemi-Doc MP Imaging system . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 004 By reanalyzing the crosslinked lysates using a lower percent gel , we discovered additional CTB-reactive bands at much higher molecular weights in lysates from all three cell lines ( Jurkat , T84 and Colo205; Figure 1C ) . Appearance of the high molecular weight bands was dependent on both inclusion of Ac4ManNDAz and UV irradiation , suggesting that these bands also represent CTB crosslinked to sialylated molecules , but of much larger molecular weight than GM1 . Surprised by the difference we observed between Jurkat and colonic epithelial cell lines , we also examined SiaDAz-mediated CTB crosslinking in two additional cell types , a human brain capillary endothelial cell line ( hCMEC/D3 ) ( Weksler , 2005 ) and a human bronchial epithelial cell line ( HBEC ) ( Ramirez , 2004 ) . In hCMEC/D3 cells , we observed both a complex with molecular weight consistent with CTB crosslinked to GM1 , as well as a high molecular weight complex ( Figure 1—figure supplement 1A ) . In HBECs , we also observed a complex consistent with CTB-GM1 crosslinking , as well as a faint higher molecular complex ( Figure 1—figure supplement 1B ) . Thus , CTB crosslinking patterns are cell type dependent , and CTB crosslinking to GM1 is not observed in all cell types . We considered two possible explanations for the high molecular weight CTB crosslinked complex . One possibility was that the complex represented CTB crosslinked to GM1 , or another glycolipid , but behaving as an aggregate in the SDS-PAGE analysis . The second possibility was that the complex represented CTB crosslinked to a sialylated glycoprotein . To distinguish between these possibilities and to determine which class of glycoconjugates was required for formation of the high molecular weight complex , we made use of small molecule inhibitors and a decoy substrate that specifically interfere with production of different classes of glycoconjugates . The colonic epithelial T84 ( Figure 2A , B ) or Colo205 ( Figure 2C ) cell lines were therefore cultured with both Ac4ManNDAz and an inhibitor of glycosylation . CTB was added to the cells and crosslinking was performed . Lysates were examined by immunoblot using an anti-CT or anti-CTB antibody , and the effect of the various inhibitors on the intensity and molecular weight of the crosslinked band was determined . 10 . 7554/eLife . 09545 . 005Figure 2 . CTB recognizes glycoproteins on human colonic epithelial cell lines . ( A ) T84 cells were cultured with Ac4ManNDAz and a glycosylation inhibitor , incubated with CTB , and UV irradiated . NB-DGJ interferes with ganglioside biosynthesis; benzyl-α-GalNAc competitively inhibits GalNAc-type O-linked glycosylation . Lysates were analyzed by 7 . 5% SDS-PAGE immunoblot with anti-CTB antibody . The asterisk indicates a SiaDAz-independent band that is observed with benzyl-α-GalNAc treatment . ( B ) T84 cells were cultured with Ac4ManNDAz and a glycosylation inhibitor , incubated with CTB , and UV irradiated . Deoxymannojirimycin and kifunensine both interfere with maturation of N-linked glycans . Lysates were analyzed by 7 . 5% SDS-PAGE immunoblot with anti-CTB antibody . ( C ) Colo205 cells were cultured with Ac4ManNDAz and a glycosylation inhibitor , incubated with CTB , and UV irradiated . Lysates were analyzed by 6% SDS-PAGE immunoblot with anti-CT antibody . The red asterisk indicates a SiaDAz-independent band that is observed with benzyl-α-GalNAc treatment . ( D ) Wild-type or C1GALT1C1 KO Colo205 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Lysates were analyzed by 6% SDS-PAGE immunoblot with anti-CT antibody . In all panels , red boxes highlight CTB crosslinked complexes observed in cells cultured with Ac4ManNDAz and treated with UV radiation . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 00510 . 7554/eLife . 09545 . 006Figure 2—figure supplement 1 . Effectiveness of N-linked glycosylation inhibitors in human colonic epithelial cell lines . ( A ) T84 cells were cultured with inhibitors of N-linked glycosylation . Lysates were analyzed by immunoblot with an anti-LAMP1 antibody . ( B ) Colo205 cells were cultured with inhibitors of glycosylation , then ConA binding was analyzed by flow cytometry . ConA binds high mannose structures that are produced when N-linked maturation is blocked . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 00610 . 7554/eLife . 09545 . 007Figure 2—figure supplement 2 . Effectiveness of O-linked glycosylation inhibitor in human colonic epithelial cell lines . ( A ) T84 cells were cultured with benzyl-α-GalNAc , an inhibitor of O-linked glycosylation . Lysates were analyzed by immunoblot with an anti-CD44 antibody . ( B ) Colo205 cells were cultured with benzyl-α-GalNAc , an inhibitor of O-linked glycosylation , then PNA binding was analyzed by flow cytometry . PNA binds the T-antigen , which is produced , but not elaborated , when cells are cultured with benzyl-α-GalNAc . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 007 The first inhibitor of glycosylation we used was NB-DGJ , a compound that interferes with glucosylation of ceramide , an early step in ganglioside biosynthesis ( Andersson et al . , 2000 ) . We found that the CTB complexes detected by immunoblot in T84 and Colo205 cells were unaffected by culturing the cells with NB-DGJ ( Figure 2A , C ) . In contrast , culturing Jurkat cells with NB-DGJ completely eliminates formation of the CTB-GM1 crosslinked complex ( Bond et al . , 2010 ) . These data imply that the high molecular weight band does not represent an aggregate of the CTB-GM1 crosslinked complex . Next , we examined inhibitors of protein glycosylation . To test if N-linked protein glycosylation is required for CTB crosslinking , cells were cultured with either deoxymannojirimycin or kifunensine , small molecules that interfere with the maturation of N-linked glycans ( Fuhrmann et al . , 1984; Elbein et al . , 1990 ) . Immature N-linked glycans will not contain SiaDAz and will be unable to engage in crosslinking . If the CTB crosslinked complex depends on N-linked protein glycosylation , then culturing cells with these inhibitors should reduce or eliminate the complex . We first confirmed the effectiveness of these inhibitors in T84 cells by observing a reduction in the apparent molecular weight of LAMP1 , a protein with multiple sites of N-linked glycosylation ( Figure 2—figure supplement 1A ) . Likewise , in Colo205 cells , we observed increased binding of lectin concanavalin A ( ConA ) , reflecting accumulation of immature high mannose structures ( Figure 2—figure supplement 1B ) . Both deoxymannojirimycin and kifunensine caused subtle effects on the high molecular weight crosslinked CTB complexes , altering the intensities and increasing the apparent molecular weight ( Figure 2B , C ) . These results suggested that proteins with N-linked glycosylation play a role in the formation of crosslinked CTB complexes , but that N-linked glycosylation is not the sole contributor . To test if O-linked protein glycosylation is required for CTB crosslinking , cells were cultured with benzyl-α-GalNAc , a decoy substrate that competitively inhibits GalNAc-type O-linked glycosylation ( Kuan et al . , 1989 ) . We first confirmed the effectiveness of benzyl-α-GalNAc in T84 cells by showing that it caused a reduction in the apparent molecular weight of CD44 , a protein with multiple sites of O-linked glycosylation ( Figure 2—figure supplement 2A ) . Benzyl-α-GalNAc also inhibited maturation of O-linked glycans in Colo205 cells , demonstrated by the observed increase in binding of the lectin peanut agglutinin ( PNA ) , which binds to the T-antigen disaccharide ( Figure 2—figure supplement 2B ) . Culturing either T84 or Colo205 cells with benzyl-α-GalNAc resulted in dramatic reductions in the intensity of the high molecular weight crosslinked complexes , suggesting that CTB crosslinks to glycoproteins bearing O-linked glycans ( Figure 2A , C ) . However , use of benzyl-α-GalNAc also resulted in the appearance of a new CTB-containing species at lower apparent molecular weight ( ∼100 kDa ) . Because appearance of the 100 kDa band was not dependent on the addition of Ac4ManNDAz , it does not represent crosslinking through SiaDAz , and may relate to the UV absorbance properties of the benzyl group in benzyl-α-GalNAc . Overall , the results of the inhibition experiments demonstrate that CTB crosslinks to glycoproteins , with both N-linked and O-linked glycans playing roles . Because of the potential for small molecule inhibitors to exert unanticipated effects , we sought a second approach to gain insight into the role of glycoproteins in CTB crosslinking . We used Colo205 cells in which elongation of GalNAc-type O-linked glycans is blocked due to zinc finger nuclease ( ZFN ) targeting of C1GALT1C1 , which encodes a chaperone required for biosynthesis of GalNAc-type O-linked glycans ( Steentoft et al . , 2011 ) . While wild-type Colo205 cells produced the high molecular weight crosslinked CTB complex , the intensity of this band was dramatically reduced in cells lacking C1GALT1C1 activity ( Figure 2D ) . We conclude that GalNAc-type O-linked glycosylation of proteins plays an important role in formation of high molecular weight crosslinked CTB complexes in Colo205 cells . Taken together , the crosslinking data indicate that CTB crosslinks to glycoproteins in colonic epithelial cells , with both N-linked and O-linked glycosylation of proteins making contributions . In contrast , no evidence pointed to CTB crosslinking to GM1 or other gangliosides in either colonic epithelial cell line . Our inability to detect CTB crosslinking to GM1 in either colonic epithelial cell line stimulated us to evaluate the ganglioside content of these cells . Using Soxhelt extraction , lipids were extracted from T84 , Colo205 , and Jurkat cells cultured with either vehicle or the glycosphingolipid inhibitor NB-DGJ . Alkali-labile phospholipids were removed by mild alkaline hydrolysis , and some non-polar compounds were removed by silicic acid chromatography . When analyzed by thin-layer chromatography and resorcinol staining ( Figure 3A ) , the fraction from untreated Jurkat cells yielded several bands migrating with mobilities similar to and slower than the GM3 ganglioside ( Figure 3A , lane 5 ) . In contrast , in the fraction from Jurkat cells cultured with NB-DGJ , only the most slow-migrating band was present ( Figure 3A , lane 6 ) . The fraction from T84 cells produced two bands , one migrating between GM3 and GM1 gangliosides , and one migrating similarly to GM1 ( Figure 3A , lane 1 ) . Finally , the fraction from Colo205 cells was only faintly stained ( Figure 3A , lane 3 ) . 10 . 7554/eLife . 09545 . 008Figure 3 . HP-TLC analysis of glycosphingolipids from T84 , Colo205 , and Jurkat cells . Partially purified glycosphingolipid fractions isolated from T84 , Colo205 and Jurkat cells cultured with either vehicle or the glycosphingolipid inhibitor NB-DGJ were separated on aluminum-backed silica gel plates using chloroform/methanol/water ( 60:35:8 , by volume ) as solvent and stained with resorcinol ( A ) . Chromatograms with separated glycosphingolipids were incubated with 125I-labeled CTB ( B ) or lectin from Erythrina cristagalli ( C ) , followed by autoradiography for 12 hr . In ( C ) , the asterisk ( * ) highlights the putative neolactotetraosylceramide band . Additional detail about samples analyzed is provided in the methods section . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 008 To further identify these species , the partially purified glycosphingolipid fractions were probed for binding to 125I-labeled CTB , also in thin-layer chromatography format . CTB bound strongly to the fraction isolated from untreated Jurkat cells ( Figure 3B , lane 5 ) , and also more weakly to the fraction isolated from Jurkat cells cultured with NB-DGJ ( Figure 3B , lane 6 ) . The CTB-binding material from Jurkat cells co-migrated with GM1 , and appeared as a doublet , likely corresponding to GM1 species with different ceramide components . No binding of 125I-labeled CTB to crude glycosphingolipid fractions from T84 or Colo205 cells was observed ( Figure 3B , lanes 1 and 3 ) , not even when high concentrations of the glycosphingolipid fractions and high concentrations of 125I-CTB were used . Based on the amount of cells analyzed and the sensitivity of detection , we estimate that T84 and Colo205 cells contain no more than 5 ng of GM1 per million cells . To evaluate the effect of NB-DGJ on glycosphingolipid production , binding of the Galβ4GlcNAc-binding lectin from E . cristagalli ( Teneberg et al . , 1994 ) to the partially purified glycosphingolipid fractions was tested . For the crude glycosphingolipid fraction from T84 cells , binding in the tetraglycosylceramide region , most likely to neolactotetraosylceramide ( Galβ4GlcNAcβ3Galβ4Glcβ1Cer ) , was apparent ( Figure 3C , lane 1 ) , but this binding was not visible in the fraction from T84 cells cultured with NB-DGJ ( Figure 3C , lane 2 ) . We conclude that NB-DGJ effectively inhibits production of glucosylceramide glycolipids in T84 cells . Thus , if a low , undetectable amount of GM1 is present in the intestinal epithelial cell lines , it is reasonable to assume that this level is further reduced by culturing the cells with NB-DGJ . With the insight that CTB can recognize glycoproteins displayed on the colonic epithelial cells , we next assessed the relative contributions of different glycoconjugates to overall CTB binding in different cell types . First we examined Jurkat cells , where we had observed multiple CTB crosslinked species . Jurkat cells were cultured with inhibitors of glycosylation to prevent production of specific classes of glycoconjugates , then CTB binding was measured by flow cytometry ( Figure 4A ) . The ganglioside biosynthesis inhibitor NB-DGJ reduced GM1 production in Jurkat cells ( Figure 3B ) and also resulted in a decrease in CTB binding ( Figure 4A ) . In contrast , culturing Jurkat cells with benzyl-α-GalNAc or kifunensine had no significant effect on CTB binding ( Figure 4A ) . The inability of benzyl-α-GalNAc to affect CTB binding to Jurkat cells is consistent with the known O-linked glycosylation defect in these cells , caused by a frame-shift mutation in the gene encoding chaperone C1GALT1C1 ( Figure 4—figure supplement 1A ) ( Ju and Cummings , 2002 ) . Jurkat cells do produce N-linked glycans , as evidenced by the enhancement of ConA binding to kifunensine-treated cells ( Figure 4—figure supplement 1B ) , but since kifunensine treatment does not affect CTB binding to Jurkat cells ( Figure 4A ) , N-linked glycans do not appear to be major contributors to CTB binding . Thus , the inhibition studies indicate that gangliosides are the dominant binding partners for CTB on Jurkat cells . 10 . 7554/eLife . 09545 . 009Figure 4 . CTB binding to human colonic epithelial cell lines depends on protein glycosylation . ( A ) Jurkat cells were cultured with inhibitors of glycosylation . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from two independent experiments . ( B ) T84 cells were cultured with inhibitors of glycosylation , then incubated with increasing concentrations of CTB . Binding of CTB was measured by ELISA . Data presented are the mean values for duplicate samples with error bars indicating the standard deviation . A replicate experiment yielded similar results . ( C ) T84 cells were cultured with inhibitors of glycosylation . Binding of Alexa Fluor 647-CTB was measured by fluorescence microscopy . ( D ) Colo205 cells were cultured with inhibitors of glycosylation . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from two independent experiments . ( E ) Binding of CTB to wild-type or C1GALT1C1 KO Colo205 cells was measured by flow cytometry . Data shown are a single representative trial from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 00910 . 7554/eLife . 09545 . 010Figure 4—figure supplement 1 . Characterization of protein glycosylation in Jurkat T cells . ( A ) The C1GALT1C1 gene was amplified from genomic DNA of T84 and Jurkat cells by PCR , then sequenced . Jurkat C1GALT1C1 was confirmed to contain a frame shift mutation that results in the expression of a truncated form of the protein ( Ju and Cummings , 2002 ) . ( B ) Jurkat T cells were cultured with inhibitors of glycosylation , then ConA binding was analyzed by flow cytometry . ConA binds high mannose structures that are produced when N-linked maturation is blocked . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 01010 . 7554/eLife . 09545 . 011Figure 4—figure supplement 2 . Representative fluorescence microscopy images of Alexa Fluor 647-CTB binding to T84 cells cultured with glycosylation inhibitors . T84 cells were cultured with inhibitors of glycosylation . Binding of Alexa Fluor 647-CTB and DAPI staining were measured by fluorescence microscopy . Quantification of imaging data is presented in Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 011 Next , we turned attention to the colonic epithelial cell lines . Because of difficulties associated with performing flow cytometry on the highly adherent T84 cells , CTB binding to the surface of T84 cells was measured by other methods . In the first approach , binding of biotin-labeled CTB to T84 monolayers was measured by an ELISA method ( Figure 4B ) . In the second approach , binding of Alexa Fluor 647-labeled CTB to clusters of cells was quantified by fluorescence microscopy ( Figure 4C and Figure 4—figure supplement 2 ) . In the fluorescence microscopy approach , fluorescence was observed primarily to the outer surface of cell clusters , suggesting that Alexa Fluor 647-labeled CTB has limited access to the interior of the cell clusters , or that the CTB ligand is not displayed on the interior surface . Despite the differences in format , the results obtained by the ELISA and fluorescence microscopy methods were in agreement . NB-DGJ , the inhibitor of ganglioside biosynthesis , had no effect on CTB binding . Kifunensine , an inhibitor of N-linked glycan maturation , resulted in only small effects on CTB binding . In contrast , culturing cells with benzyl-α-GalNAc , which blocks elaboration of O-linked glycans , resulted in reductions in CTB binding in both assays . We conclude the glycoproteins are the dominant CTB binding partners in T84 cells and that gangliosides do not make a large contribution to CTB binding to T84 cells . To measure CTB binding to Colo205 cells , we used flow cytometry . NB-DGJ treatment did not affect CTB binding to Colo205 cells ( Figure 4D ) . However , in contrast to the results observed for T84 cells , benzyl-α-GalNAc did not cause a reduction in CTB binding to Colo205 cells ( Figure 4D ) , nor did Colo205 cells lacking C1GALT1C1 show reduced CTB binding ( Figure 4E ) . Instead , Colo205 cells cultured with the N-linked inhibitor kifunensine exhibited enhanced CTB binding ( Figure 4D ) , consistent with a small increase in SiaDAz crosslinking that kifunensine causes in these cells ( Figure 2C ) . We considered the possibility that CTB binds to multiple classes of glycoconjugates in Colo205 cells . We cultured Colo205 cells with pairs of inhibitors and measured CTB binding . The only case where we observed decreased CTB binding was when the two protein glycosylation inhibitors – kifunensine and benzyl-α-GalNAc – were used together ( Figure 4D ) . In no case did culturing cells with NB-DGJ result in decreased CTB binding . Based on these data , we propose that CTB binds primarily to glycoproteins on Colo205 cells , with contributions from both N-linked and O-linked glycans . Gangliosides do not make a large contribution to CTB binding to Colo205 cells . To demonstrate conclusively that CTB binds glycoproteins , we isolated crosslinked CTB complexes and used mass spectrometry to identify one of the crosslinked glycoproteins . T84 cells were cultured with Ac4ManNDAz , and then incubated with biotin-CTB . After UV irradiation , the cells were lysed and the membrane fraction isolated . Biotinylated , crosslinked complexes were purified on streptavidin-agarose and subjected to trypsin digest , followed by LC-MS/MS analysis . Alternatively , purified crosslinked material was loaded onto an SDS-PAGE gel . Then the CTB-glycoprotein crosslinked material was excised , trypsin digest was performed , and the released peptides were analyzed by LC-MS/MS . We focused attention on proteins that were detected with a spectral count higher than 3 in either crosslinked sample , and not detected in the corresponding control sample ( Table 1 ) . 10 . 7554/eLife . 09545 . 012Table 1 . Proteomics analyses of CTB crosslinked complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 012In gelIn solutionProtein symbolProtein namePeptide sequences% coverageSpectral countPeptide sequences% coverageSpectral countITGB4Integrin beta-4116 . 514NANANASLC12A2Solute carrier family 12 ( Sodium/potassium/chloride transporters ) , member 2 ( isoform CRA ) 66 . 710NANANACD44CD44 antigen528 . 67 . 92NANANAPLXNB2Plexin-B243653 . 53 . 00CEACAM5Carcinoembryonic antigen-related cell adhesion molecule 536 . 3523 . 24 . 00CTBCholera toxin subunit B215 . 34853 . 2113 . 00LY75Lymphocyte antigen 75 ( isoform 4 ) 32 . 1396 . 311 . 00COPACoatomer subunit alpha333NANANASPTB2Spectrin beta chain , brain 141 . 83NANANAITGA6Integrin alpha-6 ( isoform Alpha-6X1A ) NANANA1312 . 714 . 83EGFREpidermal growth factor receptor ( isoform 1 ) NANANA1211 . 511 . 92MUC13Mucin-13NANANA615 . 69 . 00ITGB1Integrin beta-1 ( isoform Beta-1A ) NANANA79 . 37 . 96DPP4Dipeptidyl peptidase 4NANANA810 . 17 . 00CDCP1CUB domain-containing protein 1 ( isoform 1 ) NANANA66 . 36 . 99PLXNA1Plexin-A1NANANA84 . 26 . 98SPINT1Kunitz-type protease inhibitor 1 ( isoform 2 ) NANANA611 . 36 . 00PARP4Poly [ADP-ribose] polymerase 4NANANA63 . 86 . 00ITGAVIsoform 1 of Integrin alpha-V ( isoform 1 ) NANANA55 . 15 . 00ATP1B3Sodium/potassium-transporting ATPase subunit beta-3NANANA417 . 24 . 97PTGFRNPTGFRN protein ( Fragment ) NANANA610 . 64 . 00SCARB1Scavenger receptor class B member 1 ( isoform 1 ) NANANA49 . 34 . 00PGRMC1Membrane-associated progesterone receptor component 1NANANA315 . 44 . 00DSG2Desmoglein-2NANANA44 . 44 . 00PVRPoliovirus receptor ( isoform beta ) NANANA415 . 74 . 00COPB1Coatomer subunit betaNANANA45 . 44 . 00PDIA4Protein disulfide-isomerase A4NANANA46 . 23 . 00ST14Suppressor of tumorigenicity 14 proteinNANANA34 . 93 . 00RP2Protein XRP2NANANA37 . 43 . 00CA12Carbonic anhydrase 12 ( isoform 1 ) NANANA515 . 23 . 00EPHA2Ephrin type-A receptor 2NANANA46 . 73 . 00MFI2Melanotransferrin ( isoform 1 ) NANANA23 . 53 . 00CD47Leukocyte surface antigen CD47 ( isoform OA3-293 ) NANANA39 . 63 . 00CTB crosslinked complexes were isolated on streptavidin-agarose . The complexes were eluted from streptavidin-agarose and applied to SDS-PAGE prior to trypsin digest ( in-gel ) or trypsinized directly on streptavidin-agarose ( in-solution ) . Tryptic fragments were analyzed by LC-MS/MS to identify proteins crosslinked to CTB . This table lists proteins with spectral counts ≥ 3 that were present in a crosslinked sample but absent from the corresponding control sample . CEACAM5 ( also known as CEA ) and CD44 were among the top hits from the proteomics analysis and are both known to be highly glycosylated . We tested whether either of these proteins were present in the CTB-glycoprotein crosslinked complex . Cells were cultured with Ac4ManNDAz , then incubated with CTB . After UV irradiation , the cells were lysed . Immunoprecipitation of CEACAM5 resulted in purification of the CTB crosslinked complex ( Figure 5A ) . Similarly , immunoprecipitation performed with anti-CD44 resulted in purification of the CTB crosslinked complex ( Figure 5B ) . We also performed immunoprecipitation with anti-CTB , which resulted in efficient purification of CEACAM5 , but only weak purification of CD44 ( Figure 5C ) . Notably , anti-CTB immunoprecipitated CEACAM5 even when the Ac4ManNDAz crosslinker was omitted , suggesting that the CTB-CEACAM5 interaction is relatively strong ( Figure 5C , D ) . Furthermore , the material purified by anti-CTB and recognized by anti-CEACAM5 had a slightly higher apparent molecular weight when the crosslinker was included than when it was omitted ( Figure 5C ) . Thus , we observe direct crosslinking between CTB and CEACAM5 . Additionally , CD44 co-purifies with the CTB-glycoprotein crosslinked complex , but is not directly crosslinked to CTB . 10 . 7554/eLife . 09545 . 013Figure 5 . Protein glycosylation is required for a CTB-glycoprotein interaction . ( A ) T84 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Immunopurification was performed with control IgG or anti-CEACAM5 . 7 . 5% SDS-PAGE immunoblots were performed with anti-CEACAM5 and anti-CTB . ( B ) T84 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Immunopurification was performed with control IgG or anti-CD44 . 7 . 5% SDS-PAGE immunoblots were performed with anti-CD44 and anti-CTB . ( C ) T84 cells were cultured with Ac4ManNDAz , incubated with CTB , and UV irradiated . Immunopurification was performed with control IgG or anti-CTB . 6% SDS-PAGE immunoblots were performed with anti-CTB , anti-CEACAM5 , and anti-CD44 . ( D ) T84 cells were cultured with inhibitors of glycosylation , then incubated with or without CTB . Immunopurification was performed with control IgG or anti-CTB . 6% SDS-PAGE immunoblots were performed with anti-CEACAM5 . ( E ) T84 cells were cultured with Ac4ManNDAz and with or without kifunensine , incubated with CTB , and UV irradiated . Immunopurification was performed with control IgG or anti-CEACAM5 . 7 . 5% SDS-PAGE immunoblots were performed with anti-CEACAM5 and anti-CTB . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 013 We performed additional experiments to characterize the role of protein glycosylation in the CTB-CEACAM5 interaction . We cultured cells with inhibitors of glycosylation and assessed the effect on the noncovalent CTB-CEACAM5 interaction ( Figure 5D ) . Both benzyl-α-GalNAc and kifunensine treatment caused reductions in the apparent molecular weight of CEACAM5 , suggesting that this protein bears both N-linked and O-linked glycans . Treatment with benzyl-α-GalNAc enhanced the CTB-CEACAM5 interaction , while treatment with kifunensine abrogated the interaction , and treatment with NB-DGJ had no effect . Inhibition of N-linked glycan maturation also blocked crosslinking between CTB and CEACAM5 ( Figure 5E ) . Thus , protein glycosylation regulates the ability of CEACAM5 to interact with CTB . Taken together , these data demonstrate a direct interaction between CTB and at least one glycoprotein , CEACAM5 . Recognition of CEACAM5 by CTB depends on protein glycosylation , suggesting that the glycan , and not the protein , may provide the recognition determinant . While culturing cells with kifunensine abolished the CEACAM5-CTB interaction ( Figure 5E ) , overall CTB crosslinking in T84 cells was only slightly affected by kifunensine treatment ( Figure 2B ) . Thus , additional CTB-crosslinking proteins exist in T84 cells and remain to be identified . Because glycan identity , not protein identity , appeared to be essential to CTB binding , our next goal was to gain greater insight into glycan structures recognized by CTB . Epidemiological studies implicate variation among fucose-containing glycoconjugates in human susceptibility to cholera ( Barua and Paguio , 1977; Swerdlow et al . , 1994; Harris et al . , 2005; Holmner et al . , 2010 ) . In addition , some glycan array binding experiments performed by the Consortium for Functional Glycomics ( CFG ) suggest that fucosylated glycans can be recognized by CTB , albeit with lower avidity than GM1 ( Consortium for Functional Glycomics , 2010 ) . Importantly , fucose is not present in GM1 , which is recognized by CTB through interactions with its terminal Neu5Ac and galactose residues ( Merritt et al . , 1994 ) . We therefore tested whether monosaccharides , including fucose , Neu5Ac , and galactose , could competitively inhibit CTB binding to different cell lines . We found that 100 mM Neu5Ac potently blocked CTB binding to Jurkat cells , while 100 mM galactose inhibited slightly ( Figure 6A , B ) . Other monosaccharides had no effect . In particular , 100 mM fucose did not interfere with CTB binding to Jurkat cells ( Figure 6B ) , even though binding of a fucose-recognizing lectin , Aleuria aurantia lectin ( AAL ) , was completely inhibited by the same concentration of free fucose ( Figure 6—figure supplement 1 ) . These results are consistent with CTB binding to Jurkat cells via recognition of GM1 . 10 . 7554/eLife . 09545 . 014Figure 6 . Fucose blocks binding of CTB to human colonic epithelial cell lines . ( A ) Jurkat cells were incubated with 4 µg/mL of CTB in the presence of 100 mM of free sugar . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from three independent experiments . ( B ) The median fluorescence intensity ( MFI ) for the no sugar treatment sample presented in panel A was normalized to 100% bound . Data shown represent an average of three independent trials and their standard deviations . ( C ) T84 cells were incubated with 200 mM of free sugar and variable concentrations of CTB . Binding of CTB was measured by ELISA . Data presented are the mean values for duplicate samples with error bars indicating the standard deviation . A replicate experiment yielded similar results . ( D ) T84 cells were incubated with variable free sugar concentrations and 10 µg/mL of CTB . Binding of CTB was measured by ELISA . Data presented are the mean values for duplicate samples with error bars indicating the standard deviation . A replicate experiment yielded similar results . ( E ) T84 cells were incubated with 100 mM fucose or 100 mM glucose in the presence of Alexa Fluor 647-CTB . Binding of Alexa Fluor 647-CTB was measured by fluorescence microscopy . ( F ) Colo205 cells were incubated with 10 µg/mL of CTB in the presence of 100 mM of free sugar . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from three independent experiments . ( G ) The median fluorescence intensity ( MFI ) for the no sugar treatment sample presented in panel F was normalized to 100% bound . Data shown represent an average of three independent trials and their standard deviations . ( H ) Colo205 cells were incubated with variable free fucose concentrations and 10 µg/mL of CTB . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 01410 . 7554/eLife . 09545 . 015Figure 6—figure supplement 1 . Effects of free sugars on lectin binding to Jurkat cells . ( A ) Jurkat T cells were incubated with 100 mM fucose or Neu5Ac and 4 µg/mL of biotin-AAL . Binding of AAL was measured by flow cytometry . ( B ) Jurkat T cells were incubated with 100 mM fucose or Neu5Ac and 4 µg/mL of biotin-MAL-II . Binding of MAL-II was measured by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 01510 . 7554/eLife . 09545 . 016Figure 6—figure supplement 2 . Representative fluorescence microscopy images showing effects of free sugars on binding of Alexa Fluor 647-CTB to T84 cells . Binding of Alexa Fluor 647-CTB and DAPI staining in the presence of 100 mM fucose or 100 mM glucose were measured by fluorescence microscopy . Quantification of imaging data is presented in Figure 6E . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 01610 . 7554/eLife . 09545 . 017Figure 6—figure supplement 3 . Effects of free sugars on lectin binding to Colo205 cells . ( A ) Colo205 cells were incubated with 100 mM fucose or Neu5Ac and 10 µg/mL of biotin-AAL . Binding of AAL was measured by flow cytometry . ( B ) Colo205 cells were incubated with 100 mM fucose or Neu5Ac and 10 µg/mL of biotin-MAL-II . Binding of MAL-II was measured by flow cytometry . ( C ) Colo205 cells were incubated with variable free fucose concentrations and 10 µg/mL of biotin-AAL . Binding of AAL was measured by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 017 Competitive inhibition of CTB binding by monosaccharides was also assessed on colonic epithelial cell lines . T84 cells were incubated with increasing amounts of biotin-CTB in a buffer containing 200 mM free monosaccharide . Binding of biotin-CTB was measured by an ELISA method ( Figure 6C ) . Fucose and Neu5Ac , but not other sugars , effectively prevented binding of all concentrations of CTB . Further , both fucose and Neu5Ac inhibited biotin-CTB binding to T84 cells in a concentration-dependent manner ( Figure 6D ) . Free fucose also blocked binding of Alexa Fluor 647-CTB to T84 cells , as measured by fluorescence microscopy ( Figure 6E and Figure 6—figure supplement 2 ) . The ability of monosaccharides to inhibit binding of biotin-CTB to Colo205 cells was measured by flow cytometry ( Figure 6F ) . Fucose was the most effective inhibitor , while galactose and Neu5Ac each showed moderate inhibition ( Figure 6G ) . While as little as 10 mM fucose could interfere with biotin-CTB binding to Colo205 cells , glucose did not affect biotin-CTB binding even at concentrations as high as 200 mM ( Figure 6H ) . Similarly , free fucose , but not free Neu5Ac or glucose , blocked binding of fucose-recognizing AAL to Colo205 cells ( Figure 6—figure supplement 3 ) . A theme emerging from the monosaccharide competition experiments was that fucose specifically inhibits CTB binding to colonic epithelial cell lines , but not to Jurkat cells . In addition , Neu5Ac and galactose each displayed some ability to block binding of CTB to multiple cell lines . Based on the results of the monosaccharide competition experiments , we used lectins as blocking reagents to assess whether fucosylated and/or sialylated structures are recognized by CTB . T84 cells were incubated with lectin , then variable concentrations of biotin-CTB were added and biotin-CTB binding was measured by ELISA ( Figure 7A ) . Treatment with Ulex europaeus agglutinin I ( UEA-1 ) or Lotus tetragonolobus lectin ( LTL ) , each of which recognize specific fucosylated structures , had no effect on biotin-CTB binding . We also examined AAL and Lens culinaris agglutinin ( LCA ) , which both recognize α1-6-fucosylated structures , although AAL has a broader substrate scope and also binds fucose in other linkages ( Kochibe and Furukawa , 1980; Matsumura et al . , 2007; Yu et al . , 2012 ) . AAL was able to block biotin-CTB binding , but LCA was not . We titrated the amount of lectins used in this blocking assay , while holding the biotin-CTB concentration constant ( Figure 7B ) . AAL , but not other lectins , blocked biotin-CTB binding to T84 cells in a concentration-dependent way . We used flow cytometry to test the ability of lectins to block CTB binding to Colo205 cells . Colo205 cells were first incubated with lectin , then CTB was added . Similar to what we observed for T84 cells , AAL effectively blocked CTB binding , while other lectins had no effect ( Figure 7C ) . AAL blocked CTB binding to Colo205 cells in a concentration-dependent way ( Figure 7D ) , and when free fucose was included during the AAL pre-incubation , AAL was no longer able to block CTB binding ( Figure 7E ) . While AAL effectively blocked CTB binding to both T84 and Colo205 cells , this fucose-recognizing lectin was not able to block biotin-CTB binding to Jurkat T cells , even when used in excess over toxin concentration ( Figure 7F ) . Taken together , results from lectin blocking experiments are consistent with the idea that CTB binds fucosylated structures on the surface of colonic epithelial cell lines , while fucosylated structures do not contribute significantly to CTB binding to Jurkat cells . In contrast , neither MAL-II , which recognizes α2-3-linked sialic acid , nor Sambucus nigra lectin ( SNA ) , which recognizes α2-6-linked sialic acid , affected biotin-CTB binding to colonic epithelial cell lines ( Figure 7A–C ) . 10 . 7554/eLife . 09545 . 018Figure 7 . Aleuria aurantia lectin ( AAL ) blocks CTB binding to human colonic epithelial cell lines . ( A ) T84 cells were incubated with 40 µg/mL of the indicated lectin , then variable concentrations of CTB were added . Binding of CTB was measured by ELISA . ( B ) T84 cells were incubated with variable lectin concentrations , then 10 µg/mL of CTB was added . Binding of CTB was measured by ELISA . Data presented are the mean values for duplicate samples with error bars indicating the standard deviation . A replicate experiment yielded similar results . ( C ) Colo205 cells were incubated with 100 µg/mL of the indicated lectins , then 10 µg/mL of CTB was added . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from three independent experiments . ( D ) Colo205 cells were incubated with variable AAL lectin concentrations , then 10 µg/mL of CTB was added . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from two independent experiments . ( E ) Colo205 cells were incubated with 40 µg/mL of AAL lectin in the presence of 200 mM free sugar ( fucose or glucose ) , then 10 µg/mL of CTB was added . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from two independent experiments . ( F ) Jurkat T cells were incubated with 40 µg/mL AAL lectin , then 4 µg/mL of CTB was added . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 018 We tested whether inhibition of fucosylation affected production of CTB-glycoprotein crosslinked complexes . T84 or Colo205 cells were cultured with Ac4ManNDAz and increasing concentrations of peracetylated 2-fluorofucose ( 2F-Fuc ) , a metabolic inhibitor of fucosylation ( Rillahan et al . , 2012 ) . As the 2F-Fuc concentration increased , reduced amounts of the CTB-glycoprotein crosslinked complex were observed ( Figure 8A , B ) , indicating that fucosylation is required for SiaDAz-dependent CTB crosslinking to glycoproteins . Next , we assessed whether fucosylation was required for the direct interaction between CTB and CEACAM5 . T84 cells were cultured with Ac4ManNDAz and 2F-Fuc , then crosslinking to CTB was performed . CEACAM5 was isolated by immunoprecipitation and probed with anti-CTB to assess CTB-CEACAM5 crosslinking ( Figure 8C ) . Less crosslinked CTB-CEACAM5 was observed when fucosylation was inhibited . One possible explanation for this result is that the CEACAM5-CTB interaction is mediated by a fucosylated glycan . Additionally , the decreased crosslinking may reflect reduced overall binding of CTB to the cell surface when fucosylation is inhibited . 10 . 7554/eLife . 09545 . 019Figure 8 . Inhibition of fucosylation reduces CTB crosslinking and binding to human colonic epithelial cell lines . ( A ) T84 cells were cultured with Ac4ManNDAz and increasing concentrations of 2F-Fuc , incubated with CTB , and UV irradiated . Lysates were analyzed by 7 . 5% SDS-PAGE immunoblot with anti-CTB antibody . ( B ) Colo205 cells were cultured with Ac4ManNDAz and increasing concentrations of 2F-Fuc , incubated with CTB , and UV irradiated . Lysates were analyzed by 7 . 5% SDS-PAGE immunoblot with anti-CT antibody . ( C ) T84 cells were cultured with Ac4ManNDAz and with or without 2F-Fuc , incubated with CTB , and UV irradiated . Immunopurification was performed with IgG or anti-CEACAM5 . 7 . 5% SDS-PAGE immunoblots were performed with anti-CEACAM5 and anti-CTB . ( D ) T84 cells were cultured with 2F-Fuc , then incubated with increasing concentrations of CTB . Binding of CTB was measured by ELISA . Data presented are the mean values for duplicate samples with error bars indicating the standard deviation . A replicate experiment yielded similar results . ( E ) T84 cells were cultured with 2F-Fuc . Binding of Alexa Fluor 647-CTB was measured by fluorescence microscopy . ( F ) Colo205 cells were cultured with 2F-Fuc or 3F-NeuAc . Binding of CTB was measured by flow cytometry . Data shown are a single representative trial from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 01910 . 7554/eLife . 09545 . 020Figure 8—figure supplement 1 . Culturing T84 cells with 2F-Fuc causes decreased cell surface fucosylation . ( A ) Culturing T84 cells with 100 µM 2F-Fuc results in reduced biotin-UEA-I binding by flow cytometry . ( B ) Culturing T84 cells with 100 µM 3F-NeuAc does not result in reduced biotin-MAL-II binding by flow cytometry . ( C ) Culturing T84 cells with 100 µM 3F-NeuAc does not result in reduced biotin-SNA binding by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 02010 . 7554/eLife . 09545 . 021Figure 8—figure supplement 2 . Representative fluorescence microscopy images of Alexa Fluor 647-CTB binding to T84 cells cultured with 2F-Fuc . T84 cells were cultured with 2F-Fuc . Binding of Alexa Fluor 647-CTB and DAPI staining were measured by fluorescence microscopy . Quantification of imaging data is presented in Figure 8E . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 02110 . 7554/eLife . 09545 . 022Figure 8—figure supplement 3 . Culturing Colo205 cells with 2F-Fuc and 3F-NeuAc causes decreased cell surface fucosylation and sialylation , respectively . ( A ) Culturing Colo205 cells with 200 µM 2F-Fuc results in reduced biotin-UEA-I binding by flow cytometry . ( B ) Culturing Colo205 cells with 200 µM 3F-NeuAc results in reduced biotin-MAL-II binding by flow cytometry . ( C ) Culturing Colo205 cells with 200 µM 3F-NeuAc results in reduced biotin-SNA binding by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 022 We tested whether inhibiting fucosylation affected overall binding of CTB to colonic epithelial cell lines . First , we confirmed the effectiveness of 2F-Fuc to reduce cell surface fucose in T84 cells by showing that it caused a reduction in binding of UEA-1 ( Figure 8—figure supplement 1 ) . Then , T84 cells were cultured with or without inhibitory concentrations of 2F-Fuc , followed by incubation with increasing concentrations of biotin-CTB . The amount of biotin-CTB bound to the cell surface was measured by an ELISA method ( Figure 8D ) . Inhibition of fucosylation reduced the plateau value for biotin-CTB binding to T84 cells . We also used fluorescence microscopy to measure binding of Alexa Fluor 647-labeled CTB to T84 cells cultured with or without 2F-Fuc . Cells cultured with the fucosylation inhibitor displayed a reduction in Alexa Fluor 647-CTB binding ( Figure 8E and Figure 8—figure supplement 2 ) . Similarly , Colo205 cells were cultured with 2F-Fuc and CTB binding was measured by flow cytometry ( Figure 8F ) . Inhibition of fucosylation reduced CTB binding to Colo205 cells , too . In contrast , Colo205 cells cultured with a sialylation inhibitor , 3F-NeuAc , ( Rillahan et al . , 2012 ) displayed enhanced binding to CTB , consistent with a small increase in binding of a fucose-recognizing lectin that also occurs with inhibition of sialylation ( Figure 8F and Figure 8—figure supplement 3 ) . We conclude that recognition of fucosylated structures is an important and general mechanism for CTB binding to colonic epithelial cells . The data regarding sialylation are less clear-cut and additional experiments will be required to determine whether sialic acids are components of the glycan motifs that CTB recognizes on colonic epithelial cells . Having discovered that fucosylation and protein glycosylation are important factors mediating CTB binding on the surface of colonic epithelial cell lines , we next tested whether fucosylated glycoproteins were involved in the mechanism by which CT intoxicates host cells . We used an in-cell ELISA method to measure the effects of inhibitors of glycosylation on the uptake of biotin-CTB by T84 cells . Cells were first cultured with glycosylation inhibitors , as described above . To quantify internalized biotin-CTB , cells were incubated in the presence of biotin-CTB at 37°C for the indicated times , then unbound biotin-CTB was washed away and the cells were rapidly cooled to 4°C to arrest internalization . Remaining surface-bound biotin-CTB was blocked with avidin , and also removed by acid washing . After permeabilizing cells , internalized biotin-CTB was measured by ELISA ( Figure 9A ) . Total surface bound biotin-CTB was simultaneously assessed by maintaining control cells at 4°C ( no endocytosis ) ( Figure 9—figure supplement 1 ) and yielded values consistent with other measurements ( Figures 4B , C , Figure 8D , E ) . The amount of biotin-CTB internalized by cells treated with NB-DGJ was indistinguishable from that observed for control cells ( Figure 9A ) , suggesting that gangliosides do not mediate the majority of biotin-CTB internalization . Kifunensine caused a slight enhancement in internalization , suggesting that N-linked glycosylation of proteins might serve as an impediment to CTB internalization . Culturing cells with either benzyl-α-GalNAc or 2F-Fuc each nearly abolished internalization , implying that fucosylated and/or O-linked glycoproteins may play functional roles in biotin-CTB internalization . 10 . 7554/eLife . 09545 . 023Figure 9 . Fucosylation mediates CT internalization and intoxication in a human colonic epithelial cell line . ( A ) Internalized biotin-CTB for vehicle- , NB-DGJ- , and kifunensine-treated cells and vehicle- , benzyl-α-GalNAc- , and 2F-Fuc-treated cells are displayed as a function of time . Cells were incubated with biotin-CTB at 37°C for the indicated amount of time to allow internalization to occur . The amount of internalized biotin-CTB was measured by ELISA . Data presented are the mean values for duplicate samples with error bars indicating the standard deviation . Replicate experiments yielded similar results . ( B ) T84 cells were cultured with inhibitors of glycosylation . Cells were exposed to variable concentrations of CT holotoxin for 1 hr . cAMP levels were measured by ELISA . The cAMP levels for experimental samples are reported relative to the corresponding vehicle-treated control cells . Data shown represent an average of four independent trials . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 02310 . 7554/eLife . 09545 . 024Figure 9—figure supplement 1 . Total surface-bound biotin-CTB for vehicle- , NB-DGJ- , and kifunensine-treated cells and vehicle- , benzyl-α-GalNAc- , and 2F-Fuc-treated cells . Cells were incubated with biotin-CTB at 4°C . Surface-bound biotin-CTB was measured by ELISA . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 02410 . 7554/eLife . 09545 . 025Figure 9—figure supplement 2 . The inhibitors of glycosylation kifunensine and benzyl-α-GalNAc exhibit off-target effects on cAMP production . ( A ) T84 cells were cultured with inhibitors of glycosylation . Cells were exposed to 10 μM forskolin for 1 hr at 37°C . cAMP levels were measured by ELISA . The cAMP levels are reported relative to control cells induced with forskolin . Data shown represent an average of two independent trials . ( B ) T84 cells were cultured with inhibitors of glycosylation . Cells were exposed to 1 μM VIP ( vasoactive intestinal peptide ) for 1 hr at 37°C . cAMP levels were measured by ELISA . The cAMP levels are reported relative to control cells induced with VIP . Data shown represent an average of two independent trials . ( C ) T84 cells were cultured with inhibitors of glycosylation . Cells were exposed to variable concentrations of CT holotoxin for 1 hr . cAMP levels were measured by ELISA . The cAMP levels for experimental samples are reported relative to the corresponding vehicle-treated control cells . Data shown represent an average of four independent trials . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 02510 . 7554/eLife . 09545 . 026Figure 9—figure supplement 3 . Brefeldin A blocks CT-induced cAMP accumulation . T84 cells were cultured with or without the ganglioside inhibitor NB-DGJ . Cells were exposed to 1 μg/mL brefeldin A at 37°C for 30 min , then 100 nM of CT holotoxin for 1 hr . cAMP levels were measured by ELISA . The cAMP levels are reported relative to control cells induced with CT in the absence of brefeldin A treatment . Data shown represent an average of three independent trials . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 026 Next , we examined the effects of inhibitors of glycosylation on CT-induced elevation of cAMP , a later step in host cell intoxication . T84 cells were cultured in monolayers in the presence of glycosylation inhibitors . First , we evaluated whether inhibitors of glycosylation caused off-target effects on adenylate cyclase activity . Indeed , both benzyl-α-GalNAc and kifunensine caused reductions in the amount of cAMP that accumulated in response to stimulation with forskolin ( Figure 9—figure supplement 2A ) , and with vasoactive intestinal peptide ( VIP; Figure 9—figure supplement 2B ) , and similarly reduced cAMP accumulation in response to CT ( Figure 9—figure supplement 2C ) . While the effects of benzyl-α-GalNAc and kifunensine on adenylate cyclase activity make it difficult to interpret how these inhibitors modulate CT intoxication , interpreting the effects of NB-DGJ and 2F-Fuc is more straightforward since neither affected cAMP accumulation in response to forskolin or VIP ( Figure 9—figure supplement 2A , B ) . CT holotoxin was incubated with cells at 37°C for 1 hr , then cAMP accumulation was measured . 2F-Fuc treatment resulted in dramatically decreased accumulation of cAMP , while treatment with NB-DGJ had a more moderate effect ( Figure 9B ) . Thus , host cell fucosylation is important for intoxication by CT . Since 2F-Fuc treatment also causes reduced CTB binding and cell entry , the most likely explanation for this result is that reduced fucosylation leads to less CT entering cells , and thereby reduces host cell intoxication . In contrast , NB-DGJ does not dramatically affect CTB binding or internalization , so the explanation for this result is less clear . One possibility is that there is a small amount of GM1 in the cells , which is capable of mediating host cell intoxication via pathway that operates in parallel to the fucosylated glycoproteins . A second possibility is that GM1 is not required for the initial steps of CT binding and internalization , but is required for later steps in intoxication . A final possibility is that GM1 is not essential , but other glucosylceramide glycolipids , which are also reduced due to NB-DGJ treatment , play roles in the trafficking and intoxication process . Notably , CT intoxication appeared to be sensitive to brefeldin A in both untreated cells and in cells that were cultured with NB-DGJ ( Figure 9—figure supplement 3 ) , implying that in both cases , CT intoxication of host cells occurs via retrograde transport through the secretory pathway , consistent with other studies ( Lencer , 2003 ) . The results implicating fucosylation and protein glycosylation in CTB binding and internalization are unexpected since the ganglioside GM1 is generally accepted to be the sole receptor for CT . The experiments reported here were performed in T84 and Colo205 cell lines , both colorectal cancer cell lines . While T84 cells are widely used as a model for host cell intoxication by CT , we wondered whether fucosylation and protein glycosylation might also function in CTB binding to normal colonic epithelial cells . To investigate , we used a human colonic epithelial cell line ( HCEC ) derived from normal human colon cells and immortalized by expression of Cdk4 and hTERT ( Roig et al . , 2010 ) . HCEC cells were cultured with inhibitors of glycosylation ( NB-DGJ , kifunensine , benzyl-α-GalNAc , or 2F-Fuc ) , then CTB binding was measured by an in-cell ELISA method ( Figure 10A ) . CTB binding to HCECs was low overall and unaffected by culturing the cells with either kifunensine or NB-DGJ . Culturing HCECs with benzyl-α-GalNAc resulted in an increase in CTB binding , while culturing HCECs with 2F-Fuc resulted in a decrease in CTB binding . These results suggest that GM1 is not an important determinant for CTB binding to HCECs , and point to a contribution from fucosylated structures , potentially displayed on glycoproteins . 10 . 7554/eLife . 09545 . 027Figure 10 . CTB binds fucosylated glycoproteins present in normal gut epithelia . ( A ) HCEC cells were cultured with inhibitors of glycosylation . Binding of CTB was measured by ELISA . Data presented are the mean values for triplicate samples with error bars indicating the standard deviation . Two additional replicate experiments yielded similar results . ( B ) Normal human intestinal epithelial lysate from three parts of small intestine ( duodenum , jejunum , and ileum ) , normal human colon epithelial lysate , T84 cell lysate , and Colo205 cell lysate were separated by SDS-PAGE and probed with CTB-HRP . ( C and D ) CTB binding to freshly isolated human colonic epithelial cells ( LIVE/DEAD-CD45- EpCAM+ ) assessed by flow cytometry . Cells were blocked with AAL at the indicated concentrations before addition of CTB – Alexa Fluor 647 and antibodies ( C ) or stained with CTB – Alexa Fluor 647 and antibodies in buffer containing the indicated concentrations of fucose ( D ) . The level of blocking was calculated as percent CTB – Alexa Fluor 647 mean fluorescence intensity ( MFI ) compared to samples without sugar or lectin block . Representative flow cytometry data are shown . For bar graphs , results are pooled from 2–4 independent experiments; each symbol represents one patient *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 and ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09545 . 027 While colonic epithelial cells are commonly used as a model system for CT studies , the primary physiological site for CT action is the small intestine . We assessed whether CTB-binding glycoproteins were present in epithelial tissue from different parts of the normal gut . Lysates from T84 cells , Colo205 cells , human small intestine ( duodenum , ileum , and jejunum ) , and human colon were separated by SDS-PAGE and probed with CTB-HRP ( Figure 10B ) . Discrete CTB-reactive bands were found in all samples , with the apparent mobility of these species depending on the lysate source . Thus , CTB-binding glycoproteins are present in normal human epithelial tissue , both from the small intestine and the colon . Finally , we assessed whether fucosylation plays a role in CTB binding to freshly isolated colonic epithelium . Epithelial cells were obtained from unaffected mucosa of colon adenocarcinoma patients undergoing curative tumor resection . Binding of CTB to colonic epithelial cells ( EpCAM+ CD45- ) was significantly blocked by AAL ( Figure 10C ) and free fucose ( Figure 10D ) , but not by glucose or galactose , similar to the results observed with T84 and Colo205 cells . We conclude that fucosylated structures make a significant contribution to CTB binding to normal colonic epithelial tissue .
In this study we show that glycoproteins mediate a large fraction of CTB binding to colonic epithelial cell lines . In contrast , in the Jurkat T cell line , GM1 is the main CTB binding ligand . Further , the results presented here implicate glycoproteins as functional receptors that can mediate toxin internalization . Our findings are in conflict with dogma , ( Foster and Baron , 1996 ) but not inconsistent with previous results . The earliest studies identifying GM1 as the CT receptor were based on inhibition assays and gain-of-function experiments . Indeed , CTB binds GM1 with high affinity , ( Kuziemko et al . , 1996 ) so it is not surprising that GM1 is an effective inhibitor of CTB binding and that adding GM1 to cells leads to enhanced CTB binding . However , experimentally addressing how loss of GM1 affects CTB binding and host cell intoxication has been more difficult . Use of selective inhibitors of glycosphingolipid biosynthesis allowed two groups to observe that a portion of CTB binding is GM1-independent ( Platt et al . , 1997; Blank et al . , 2007 ) . Our data extend this observation , showing that CTB binding is mostly GM1-independent in cell types more closely related to the physiological site of CT action . Further , we report evidence that CTB binds directly to glycoproteins , relying on glycan recognition motifs . Our observation that inhibition of ganglioside biosynthesis does not affect CTB binding to colonic epithelial cell lines is consistent with the low level of GM1 in these cells . Similarly , the normal human intestinal epithelia contains very little GM1 , ( Breimer et al . , 2012 ) suggesting that glycoproteins could be important contributors to CTB binding during Vibrio cholerae infection . In addition to its critical role in human disease , host cell intoxication by CT is an important model system for studying endocytic mechanisms ( Chinnapen et al . , 2007 ) . CT has been shown to enter cells through both clathrin- and caveolin-dependent mechanisms , as well a clathrin- and caveolin-independent mechanism ( Orlandi , 1998; Torgersen et al . , 2001; Massol , 2004; Howes et al . , 2010 ) . Studies to identify the dominant endocytic mechanism have yielded variable results , raising the possibility that CT uses different endocytic mechanisms to enter different cell lines . The existence of multiple glycoprotein receptors provides a plausible explanation for the use of multiple endocytic pathways and for cell type differences . The exact structure of the glycan might also play a functional role in the endocytic mechanism , as modification of proteins with a GM1 ganglioside is sufficient to engender CT binding , but not host cell intoxication ( Pacuszka and Fishman , 1990 ) . In addition , a long-standing conundrum about CT internalization is how binding to GM1 located in the outer leaflet of the plasma membrane can trigger endocytosis . Our results show that CTB can bind glycoproteins , offering a possible mechanism for receptor-mediated endocytosis . While we show that CEACAM5 binds directly to CTB through a glycan-dependent interaction , it seems unlikely that CEACAM5 is an important physiological receptor for CT . Our detection of CEACAM5 in T84 cells likely reflects the increased CEACAM5 expression and altered CEACAM5 glycosylation often found in colorectal cancer ( Saeland et al . , 2012 ) . In addition , while N-linked glycosylation is critical for the CTB-CEACAM5 interaction , the majority of CTB binding to T84 cells depends on O-linked glycosylation of proteins . Another protein previously reported to interact with CTB in pig enterocytes is sucrase-isomaltase ( SI ) ( Hansen et al . , 2005 ) ; however , we did not detect any peptides corresponding to SI in our proteomics analysis of the CTB crosslinked complex . Intriguingly , though , SI is subject to GalNAc-type O-linked glycosylation , which controls its association with detergent-insoluble membrane microdomains ( Alfalah et al . , 1999 ) . We speculate that certain GalNAc-type O-linked glycans may serve as determinants for both CTB recognition and membrane microdomain targeting , potentially explaining why CTB binding fractionates with detergent-insoluble material , an observation that is typically interpreted to result from the binding of CTB to GM1 . The demonstration that glycoproteins can be important contributors to CTB binding has implications for the use of CTB to study the organization of lipids in the plasma membrane . Indeed , differences in observed diffusion rates for bound CTB may reflect the identity of the CTB ligand , rather than the fluidity of the membrane in which it resides . Day and Kenworthy noticed that CTB bound to COS-7 cells diffused surprisingly slowly for a lipid-bound protein and speculated that an interaction with a protein might slow its diffusion ( Day et al . , 2012 ) . Our results predict that CTB diffusion rates will differ in different cell types , with rapid diffusion in GM1-rich cell lines , like Jurkat cells , and slower diffusion in GM1-deficient cell lines , such as T84 and Colo205 cells . The functional assays reported here do not directly address whether glycoproteins serve as functional CT receptors in the normal human gut . However , a consistent theme among the results reported here is that fucose plays an important role in CTB binding , both to cell lines and ex vivo to freshly isolated human epithelial cells . While the lectin blocking studies demonstrate overlap among the sets of glycans recognized by AAL and CTB , additional work is needed to determine the exact structure of the fucosylated glycans to which CTB prefers to bind . More broadly , identification of fucose as an important determinant of CTB binding adds to a growing body of literature that implicates fucose in host-microbe discourse in the gut ( Pacheco et al . , 2012; Pickard et al . , 2014 ) . In summary , we demonstrate that fucosylated glycoproteins mediate a large portion of CTB binding to human colonic epithelial cell lines , that fucosylated glycoproteins play an important role in the mechanism by which CTB enters T84 cells , and that entry of CT into T84 cells via a fucose-dependent mechanism is on-pathway to host cell intoxication . These findings raise the possibility that fucose-containing or -mimicking molecules may have utility in cholera therapy . In addition , the observation that CTB binds to cell surface molecules other than GM1 implies that caution should be applied in the interpretation of experiments where CTB is used to visualize membrane microdomain structures .
For photocrosslinking of CTB to Jurkat T cells ( a suspension cell line ) , 2 million cells were first seeded in 10 mL media into 10-cm tissue culture plates treated with either vehicle ( evaporated ethanol ) or 100 μM Ac4ManNDAz . After culturing for 72 hr , cells were counted , re-suspended in fresh media to a concentration of 5 million cells/mL , transferred to two separate multiwell tissue culture plates ( for –/+ UV ) , and CTB ( Sigma ) was added at a concentration of 2 . 5 μg/mL . The toxin was allowed to bind to the cell surface for 45 min at 4°C in the dark . The cells were then either kept at 4°C for an additional 45 min ( for the – UV samples ) or irradiated on an ice/water bath for 45 min ( for + UV samples ) at 365 nm ( UVP , XX-20BLB lamp ) . The cells were then collected , washed with Dulbecco’s Phosphate Buffered Saline ( DPBS ) , and lysed on ice for 30–60 min in RIPA buffer ( 50 mM TrisHCl , pH 8 . 0 , 150 mM NaCl , 1% ( v/v ) NP-40 , 0 . 5% ( v/v ) sodium deoxycholate , 0 . 1% ( w/v ) sodium dodecyl sulfate ( SDS ) and a protease inhibitor cocktail ( Roche , Indianapolis , IN ) ( catalog no . 11836170001 ) ) . The lysate was pelleted at 20 , 817g for 10 min at 4°C to clear the insoluble debris , and the supernatant was retained for further immunoblot analysis . Protein content was quantified with a BCA assay kit ( Thermo Scientific Pierce Protein Biology , Waltham , MA ) against a BSA standard curve for normalization . For resolution on a high percentage ( 15% ) Tris-glycine gel , 9 μg of lysate was denatured in 2X SDS loading dye ( 100 mM TrisHCl pH 6 . 8 , 4% ( w/v ) SDS , 0 . 04% ( w/v ) bromophenol blue , 20% ( v/v ) glycerol , and 10% ( v/v ) 2-mercaptoethanol ) for 5 min at 90°C . For resolution on a lower percentage ( 6% ) Tris-glycine gel , 12 μg of lysate was denatured in 4X SDS loading dye ( 200 mM TrisHCl pH 6 . 8 , 8% ( v/v ) SDS , 0 . 08% ( w/v ) bromophenol blue , 40% ( v/v ) glycerol , and 40 mM DTT ) for 5 min at 90°C . The samples were separated by SDS-PAGE and transferred to a PVDF membrane ( EMD Millipore , catalog no . IPVH00010 ) . The blots were probed overnight at 4°C for anti-CTB ( Abcam , 1:10 , 000 dilution ) in 5% ( w/v ) non-fat milk in TBST . The blots were then probed with a goat anti-rabbit HRP conjugated secondary antibody ( 1:5000 dilution ) for 1 hr at room temperature , and developed using the SuperSignal West Pico Chemiluminescent Substrate ( Thermo Scientific , catalog no . 34080 ) and X-ray film . Membranes were stripped in mild stripping buffer ( 200 mM glycine , 0 . 1% ( w/v ) SDS , 1% ( v/v ) Tween-20 , pH 2 . 2 ) for 45 min at 37°C before re-probing for the loading control anti-α-tubulin ( Sigma , 1:10 , 000 dilution ) for 1 hr at room temperature . The blots were then probed with a goat anti-mouse HRP conjugated secondary antibody ( 1:5000 dilution ) for 1 hr at room temperature , and developed using the SuperSignal West Pico Chemiluminescent Substrate and X-ray film . For photocrosslinking of CTB to human colonic epithelial cell lines ( both adherent cell lines ) , either 250 , 000 T84 cells or 500 , 000 Colo205 cells were seeded in 2 mL of media into two separate 6-well tissue culture plates ( for –/+ UV ) that were treated with either vehicle ( evaporated ethanol ) or Ac4ManNDAz to a final concentration of 100 μM . When used , glycosylation inhibitors were also added at the time of seeding to achieve these final concentrations: 50 µM NB-DGJ , 2 mM deoxymannojirimycin , 1 µg/mL kifunensine , 2 mM benzyl-α-GalNAc , and 5 – 200 μM 2F-Fuc . Experimental samples were compared to the appropriate vehicle-only control: water for NB-DGJ , deoxymannojirimycin , and kifunensine; DMSO for benzyl-α-GalNAc and 2F-Fuc . After culturing for 72 hr , the media in each well was replaced with 1 mL fresh media containing approximately 4 . 5 μg of CTB ( Sigma ) . The toxin was allowed to bind to the cell surface for 45 min at 4°C in the dark . The cells were then either kept at 4°C for an additional 45 min ( for – UV samples ) or irradiated on an ice/water bath for 45 min ( for + UV samples ) at 365 nm . Cells were then washed with DPBS , collected into RIPA lysis buffer with a cell scraper , and incubated on ice for 30 – 60 min . The lysate was centrifuged at 21 , 000g for 10 min at 4°C to remove insoluble debris , and the supernatant was retained for separation on both a higher ( 15% ) and lower ( ranging from 6 – 7 . 5% ) percentage gel , as described previously for Jurkat T cells . The samples were then transferred to a PVDF membrane , and the blots were probed overnight at 4°C for either anti-CTB ( Abcam , 1:10 , 000 dilution ) or anti-CT ( Sigma , 1:10 , 000 dilution ) as indicated in the figure labels . Membranes were stripped and re-probed for the loading control anti-α-tubulin as described previously . To ensure the glycosyltransferase inhibitors were effective in T84 cells , samples were probed for 1 hr at room temperature for either anti-LAMP-1 ( BD Biosciences , 1:5000 dilution ) or anti-CD44 ( Cell Signaling Technology , 1:2500 dilution ) . Photocrosslinking of CTB to hCMEC/D3 or HBEC cell lines was carried out in an analogous manner , with the exception that 200 , 000 cells were seeded into 6-well tissue culture plates , and approximately 1 . 1 μg of CTB ( Sigma ) was added to each well . For crosslinking for immunopurification ( IP ) experiments , 750 , 000 T84 cells were seeded into 6-cm tissue culture plates in 6 mL of media containing 100 μM Ac4ManNDAz; if required , a glycosylation inhibitor was added at this time . After culturing for 72 hr , the cells were replenished with 1 . 5 mL of fresh media containing 4 μg/mL CTB ( Sigma ) ( i . e . , 6 μg per plate ) . The toxin was allowed to bind to the cell surface for 45 min at 4°C in the dark . The cells were then either kept at 4°C for an additional 45 min ( for the – UV plates ) or irradiated on an ice/water bath for 45 min ( for + UV plates ) at 365 nm . The cells were then washed with DPBS , collected into RIPA lysis buffer with a cell scraper , and incubated on ice for 30–60 min . The lysate was pelleted at 21 , 000g for 10 min at 4°C to clear the insoluble debris , and the supernatant was retained; protein content was quantified with a BCA assay kit ( Pierce ) using a BSA standard curve . For IP with the anti-CEACAM5 and anti-CD44 antibodies , the crosslinked lysates were diluted to 1 . 5 mg/mL in RIPA buffer , and then 270 μg was added to a 1 . 5 mL microcentrifuge tube . A 1:50 dilution of normal mouse IgG ( EMD Millipore , catalog no . NI03 ) , anti-CEACAM5 antibody , or anti-CD44 antibody was added , and the samples were mixed by end-over-end rotation overnight at 4°C . The lysate/Ab mixture was then added to 10 μL of Protein G sepharose ( Sigma , catalog no . P3296 ) and mixed by end-over-end rotation for 2 . 5 hr at 4°C . The beads were washed three times with 200 µL RIPA buffer , and eluted with 10 μL 2X SDS loading dye ( 100 mM TrisHCl pH 6 . 8 , 4% ( w/v ) SDS , 0 . 04% ( w/v ) bromophenol blue , 20% ( v/v ) glycerol , and 20 mM DTT ) for 5 min at 90°C . The supernatant was collected and loaded into a single well of a 7 . 5% Tris-glycine gel . Two sets of samples were analyzed at a time: the first set to ensure that the IP of the target protein was successful ( i . e . , probing for CEACAM5 or CD44 ) , and a second set to probe for association of the target protein with CTB . Therefore , after separation and transfer to a PVDF membrane , the blots were probed overnight at 4°C for either CEACAM5 ( 1:5000 dilution ) , CD44 ( 1:2500 dilution ) , or CTB ( 1:10 , 000 dilution ) in 5% ( w/v ) non-fat milk in TBST . The blots were then probed with a HRP conjugated secondary antibody ( 1:5000 dilution ) for 1 hr at room temperature , and developed using the SuperSignal West Pico Chemiluminescent Substrate and X-ray film . For IP with the anti-CTB antibody , the crosslinked lysates were diluted to 1 . 5 mg/mL in RIPA buffer , and then 300 μg was added to a 1 . 5 mL microcentrifuge tube . Then , 0 . 5 μg of either ant-CTB antibody or normal rabbit IgG ( EMD Millipore , catalog no . NI01 ) was added , and the samples were mixed by end-over-end rotation overnight at 4°C . The lysate/Ab mixture was then added to 10 μL of TrueBlot anti-rabbit Ig IP beads ( Rockland Immunochemicals , Limerick , PA ) ( catalog no . 00-8800-25 ) and mixed by end-over-end rotation for 2 . 5 hr at 4°C . The beads were washed four times with 200 µL RIPA buffer , and eluted with 10 μL 2X SDS loading dye for 5 min at 90°C . The supernatant was collected and loaded into a single well of a 6% Tris-glycine gel . After separation and transfer to PVDF membranes , the blots were probed for CTB , CEACAM5 , and CD44 as described above . For co-IP of CTB and CEACAM5 within non-crosslinked lysates , 750 , 000 T84 cells were seeded in 5 mL of media into 6-cm dishes , and glycosylation inhibitors were added at this time . After culturing for 72 hr , the cells were replenished with 1 . 5 mL of fresh media containing 4 μg/mL CTB ( Sigma ) ( i . e . , 6 μg per plate ) and kept at 4°C for 1 . 25 hr . The cells were then washed with DPBS , collected into RIPA lysis buffer with a cell scraper , and incubated on ice for 30–60 min . The lysate was pelleted at 21 , 000g for 10 min at 4°C to clear the insoluble debris , and the supernatant was retained; protein content was quantified with a BCA assay kit ( Pierce ) using a BSA standard curve for normalization . For IP with the anti-CTB antibody , the lysates were diluted to 1 . 4 mg/mL in RIPA buffer , and then 350 μg was added to a 1 . 5 mL microcentrifuge tube . Then , 0 . 5 μg of either anti-CTB antibody or normal rabbit IgG was added , and the samples were mixed by end-over-end rotation overnight at 4°C . The lysate/Ab mixture was then added to 10 μL of TrueBlot anti-rabbit Ig IP beads and mixed by end-over-end rotation for 2 . 5 hr at 4°C . The beads were then washed , eluted , loaded onto a 6% Tris-glycine gel for separation , and probed on a PVDF membrane for CEACAM5 as described previously for the anti-CTB IP of crosslinked samples . CTB internalization was measured by the in-cell ELISA , described above , with the following adaptations . The biotin-CTB concentration was 1 µg/ml . During the experiment , control samples ( to measure total surface bound biotin-CTB ) were washed three times with ice-cold PBS , then kept on ice awaiting analysis . Experimental samples were warmed to 37°C for the indicated times to allow endocytic uptake , then endocytosis was halted by returning cells to ice and washing three times with cold PBS . Non-internalized biotinylated CTB was masked by successive treatment with 50 µg/ml of avidin ( Sigma-Aldrich ) for 1 hr on ice , followed by three 1-min cold acid washes ( 0 . 2 M acetic acid/0 . 2 M NaCl ) . Cells were fixed with 4% paraformaldehyde and further permeabilized with 0 . 1% ( v/v ) Triton X-100 in PBS . Cells were then incubated at room temperature for 1 hr in streptavidin-HRP conjugate diluted in Q-PBS ( PBS supplemented with 0 . 01% ( w/v ) saponin , 2% ( w/v ) BSA , and 0 . 1% ( w/v ) lysine , pH 7 . 4 ) . Reactive aldehydes and nonspecific binding sites were quenched with Q-PBS . HRP activity was measured as described above . 350 , 000 T84 cells were cultured for 8–10 d in the absence or presence of each glycosylation inhibitor or the corresponding vehicle control , to achieve the following final concentrations: 50 µM NB-DGJ , 1 µg/ml kifunensine , 2 mM benzyl-α-GalNAc , or 100 µM 2F-Fuc . Stock solutions of NB-DGJ and kifunensine were dissolved in water and benzyl-α-GalNAc and 2F-Fuc were dissolved in DMSO . Cells were cultured with inhibitors on 4 . 67-cm2 Transwell inserts ( Costar Laboratories , Cambridge , Mass ) . Monolayer integrity was evaluated by transepithelial electrical resistance ( TEER ) measurements , which consistently increased over the 8–10 d culturing period . Typical TEER values for cells treated with the water control , NB-DGJ , and 2F-Fuc were 800–900 Ω cm2; for cells treated with kifunensine were 700 Ω cm2; for cells treated with DMSO were 1400 Ω cm2; for cells treated with benzyl-α-GalNAc were 1800 Ω cm2 . Cells were incubated with forskolin ( 10 µM ) , VIP ( 1 µM ) , or cholera holotoxin ( 0 , 1 , or 100 nM ) in culture media with 5% CO2 at 37°C for 1 hr , followed 2 washes with cold PBS , then lysed and flash frozen in liquid N2 . Accumulated cAMP was measured using Direct Biotrak EIA kit ( GE Healthcare ) , according manufacturer’s instructions and normalized by protein content ( bicinchoninic acid assay , Pierce BCA protein assay kit , Pierce ) . The study was approved by the Regional Board of Ethics in Medical Research in west Sweden , and all volunteers gave a written informed consent before participation . Five individuals undergoing curative resection of colon tumors at the Sahlgrenska University Hospital were included in the study ( 3 males and 2 females , aged 67 to 81 ) . Immediately after surgery , a section of the unaffected mucosa was collected from the resection border located at least ten centimeters away from the tumor , and transported in ice-cold PBS before isolation of epithelial cells within less than 2 hr . Tissue specimens were washed with PBS and excess mucus and connective tissue underlying the mucosa was removed . The tissue was cut into 5 mm pieces and epithelial cells released by two cycles of treatment with 1 mM EDTA and 1 mM dithiothreitol ( DTT ) in HEPES-buffered Hank’s balanced salt solution containing 2% of fetal calf serum at 37°C for 15 min with slow stirring ( Lundgren et al . , 2005 ) . Released epithelial cells were collected by filtration , and pooled with epithelial cell fractions from two subsequent EDTA/DTT cycles for use in analysis of CTB binding by flow cytometry . The cells were first stained with Live/Dead aqua ( L34957 , Invitrogen ) . After washing , the cells were blocked with various concentrations of AAL-biotin ( B-1395 , Vector Labs ) or kept in buffer prior to addition of antibodies . The antibodies used were anti-CD45 - APC-H7 ( clone 2D1 , BD ) , anti-EpCAM - Alexa Flur 488 ( clone 9C4 , BioLegend , San Diego , CA ) , and the protein CTB - Alexa Fluor 647 ( C34778 , Invitrogen ) . Some samples were stained in the presence of l-fucose ( F2252 , Sigma-Aldrich ) , d-galactose ( G0750 , Sigma-Aldrich ) , or d-glucose ( G8270 , Sigma-Aldrich ) . The staining cocktail was then preincubated for 5 min before addition to the cells . After staining , the cells were washed and analyzed on a flow cytometer ( LSR-II , BD ) and the results were analyzed using the FlowJo software . | Cholera is a serious diarrheal disease that can be deadly if left untreated . It is caused by eating food , or drinking water , contaminated by the bacterium Vibrio cholerae . This bacterium can survive passage through the acidic conditions of the stomach . Inside the small intestine , V . cholerae attaches to the intestinal wall and starts producing cholera toxin . The toxin enters intestinal cells , causing them to release water and ions , including sodium and chloride ions . The salt-water environment created inside the intestine can , by osmosis , draw up to a further six liters of water into the intestine each day . This results in the copious production of watery diarrhea and severe dehydration . Cholera toxin is composed of six protein subunits , including five copies of cholera toxin subunit B ( CTB ) . CTB subunits help the uptake of the toxin by intestinal cells , and it has long been reported that CTB subunits attach to intestinal cells by binding to a cell surface molecule called GM1 . CTB subunits have a high affinity for GM1 , yet recent work suggests CTB may not bind exclusively to GM1; one or more additional cell surface molecules may be directly involved in cholera toxin uptake . Wands et al . now reveal that numerous cell surface molecules are recognized by CTB , and that these molecules can assist cholera toxin uptake by host cells . Glycoproteins , proteins that are marked with sugar molecules , were shown to be the primary CTB binding sites on human colon cells , and it was the glycoprotein’s sugar component , not the protein itself , that interacted with CTB . Wands et al . discovered that in particular glycoproteins containing a sugar called fucose were largely responsible for CTB binding and toxin uptake . Together these findings reveal a previously unrecognized mechanism for cholera toxin entry into host cells , and suggest that fucose-containing or fucose-mimicking molecules could be developed as new treatments for cholera . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | Fucosylation and protein glycosylation create functional receptors for cholera toxin |
Antimicrobial resistance in Gram-negative bacteria is one of the greatest threats to global health . New antibacterial strategies are urgently needed , and the development of antibiotic adjuvants that either neutralize resistance proteins or compromise the integrity of the cell envelope is of ever-growing interest . Most available adjuvants are only effective against specific resistance proteins . Here , we demonstrate that disruption of cell envelope protein homeostasis simultaneously compromises several classes of resistance determinants . In particular , we find that impairing DsbA-mediated disulfide bond formation incapacitates diverse β-lactamases and destabilizes mobile colistin resistance enzymes . Furthermore , we show that chemical inhibition of DsbA sensitizes multidrug-resistant clinical isolates to existing antibiotics and that the absence of DsbA , in combination with antibiotic treatment , substantially increases the survival of Galleria mellonella larvae infected with multidrug-resistant Pseudomonas aeruginosa . This work lays the foundation for the development of novel antibiotic adjuvants that function as broad-acting resistance breakers .
Antimicrobial resistance ( AMR ) is one of the most important public health concerns of our time ( Rochford et al . , 2018 ) . With few new antibiotics in the pharmaceutical pipeline and multidrug-resistant bacterial strains continuously emerging , it is more important than ever to develop novel antibacterial strategies and find alternative ways to break resistance . While the development of new treatments for Gram-negative bacteria is considered critical by the WHO ( Tacconelli et al . , 2018 ) , identifying novel approaches to target these organisms is particularly challenging due to their unique double-membrane permeability barrier and the vast range of AMR determinants they produce . For this reason , rather than targeting cytoplasmic processes , antimicrobial strategies that inhibit cell-envelope components or impair the activity of resistance determinants are being increasingly pursued ( Hart et al . , 2019; Laws et al . , 2019; Luther et al . , 2019; Nicolas et al . , 2019; Srinivas et al . , 2010 ) . The Gram-negative cell envelope is home to many different AMR determinants , with β-lactamase enzymes currently posing a seemingly insurmountable problem . More than 6500 unique enzymes capable of degrading β-lactam compounds have been identified to date ( Supplementary file 1 ) . Despite the development of more advanced β-lactam antibiotics , for example the carbapenems and monobactams , resistance has continued to emerge through the evolution of many broad-acting β-lactamases ( Bush , 2018 ) . This constant emergence of resistance not only threatens β-lactams , the most commonly prescribed antibiotics worldwide ( Meletis , 2016; Versporten et al . , 2018 ) , but also increases the use of last-resort agents , like the polymyxin antibiotic colistin , for the treatment of multidrug-resistant infections ( Li et al . , 2006 ) . As a result , resistance to colistin is on the rise , due in part to the alarming spread of novel cell-envelope colistin resistance determinants . These proteins , called mobile colistin resistance ( MCR ) enzymes , represent the only mobilizable mechanism of polymyxin resistance reported to date ( Poirel et al . , 2017 ) . Since their discovery in 2015 , 10 families of MCR proteins have been identified and these enzymes are quickly becoming a major threat to the longevity of colistin ( Sun et al . , 2018 ) . Alongside β-lactamases and MCR enzymes , Resistance-Nodulation-Division ( RND ) efflux pumps further enrich the repertoire of AMR determinants in the cell envelope . These multi-protein assemblies span the periplasm and remove many antibiotics ( Blair et al . , 2014; Cox and Wright , 2013 ) , rendering Gram-negative bacteria inherently resistant to important antimicrobials . Inhibition of AMR determinants has traditionally been achieved through the development of antibiotic adjuvants . These molecules impair the function of resistance proteins and are used in combination with existing antibiotics to eliminate challenging infections ( Laws et al . , 2019 ) . Whilst this approach has proven successful and has led to the deployment of several β-lactamase inhibitors that are used clinically ( Laws et al . , 2019 ) , it has so far not been able to simultaneously incapacitate different classes of AMR determinants . This is because traditional antibiotic adjuvants bind to the active site of a resistance enzyme and thus are only effective against specific protein families . To disrupt AMR more broadly , new strategies have to be developed that target the biogenesis or stability , rather than the activity , of resistance determinants . In this way , the formation of multiple resistance proteins can be inhibited at once , instead of developing specific compounds that inactivate individual AMR enzymes after they are already in place . In extracytoplasmic environments protein stability often relies on the formation of disulfide bonds between cysteine residues ( Goemans et al . , 2014; Heras et al . , 2007 ) . Notably , in the cell envelope of Gram-negative bacteria this process is performed by a single pathway , the DSB system , and more specifically by a single protein , the thiol oxidase DsbA ( Bardwell et al . , 1991; Denoncin and Collet , 2013; Hiniker and Bardwell , 2004; Kadokura et al . , 2004; Martin et al . , 1993 ) . DsbA has been shown to assist the folding of hundreds of proteins in the periplasm ( Kadokura et al . , 2004; Dutton et al . , 2008; Vertommen et al . , 2008; Figure 1A ) , including a vast range of virulence factors ( Heras et al . , 2009; Landeta et al . , 2018 ) . As such , inhibition of DSB proteins has been proposed as a promising broad-acting strategy to target bacterial pathogenesis without impairing bacterial viability ( Denoncin and Collet , 2013; Heras et al . , 2009; Landeta et al . , 2018; Heras et al . , 2015 ) . Nonetheless , the contribution of oxidative protein folding to AMR has never been examined . Since several cell envelope AMR determinants contain multiple cysteines ( Bardwell et al . , 1991; Piek et al . , 2014 ) , we hypothesized that interfering with the function of DsbA would not only compromise bacterial virulence ( Heras et al . , 2015 ) , but might also offer a broad approach to break resistance across different mechanisms by affecting the stability of resistance proteins . Here , we test this hypothesis by investigating the contribution of disulfide bond formation to three of the most important resistance mechanisms in the cell envelope of Enterobacteria: the breakdown of β-lactam antibiotics by β-lactamases , polymyxin resistance arising from the production of MCR enzymes and intrinsic resistance to multiple antibiotic classes due to RND efflux pumps . We find that some of these resistance mechanisms depend on DsbA and we demonstrate that when DsbA activity is chemically inhibited , resistance can be abrogated for several clinically important enzymes . Our findings prove that targeting protein homeostasis in the cell envelope allows the impairment of diverse AMR proteins and therefore could be a promising avenue for the development of next-generation therapeutic approaches .
DsbA has been shown to assist the folding of numerous periplasmic and surface-exposed proteins in Gram-negative bacteria ( Figure 1A; Heras et al . , 2009; Landeta et al . , 2018; Heras et al . , 2015 ) . As many AMR determinants also transit through the periplasm , we postulated that inactivation of the DSB system may affect their folding , and therefore impair their function . To test this , we first focused on resistance proteins that are present in the cell envelope and contain two or more cysteine residues , since they may depend on the formation of disulfide bonds for their stability and folding ( Bardwell et al . , 1991; Piek et al . , 2014 ) . We selected a panel of 12 clinically important β-lactamases from different Ambler classes ( classes A , B , and D ) , most of which are encoded on plasmids ( Table 1 ) . The chosen enzymes represent different protein structures , belong to discrete phylogenetic families ( Supplementary file 1 ) and have distinct hydrolytic activities ranging from the degradation of penicillins and first , second and third generation cephalosporins ( extended spectrum β-lactamases , ESBLs ) to the inactivation of last-resort β-lactams ( carbapenemases ) . In addition to β-lactamases , we selected five representative phosphoethanolamine transferases from throughout the MCR phylogeny ( Figure 1—figure supplement 1 ) to gain a comprehensive overview of the contribution of DsbA to the activity of these colistin-resistance determinants . We expressed our panel of 17 discrete resistance enzymes in an Escherichia coli K-12 strain ( E . coli MC1000 ) and its isogenic dsbA mutant ( E . coli MC1000 dsbA ) and recorded minimum inhibitory concentration ( MIC ) values for β-lactam or polymyxin antibiotics , as appropriate . We found that the absence of DsbA resulted in a substantial decrease in MIC values ( > 2 fold cutoff ) for all but one of the tested β-lactamases ( Figure 1B , Figure 1—figure supplement 2 , Supplementary file 2a ) . For the β-lactamase that seemed unaffected by the absence of DsbA , SHV-27 , we performed the same experiment under temperature stress conditions ( at 43 °C rather than 37 °C ) . Under these conditions , the lack of DsbA also resulted in a noticeable drop in the cefuroxime MIC value ( Figure 1—figure supplement 3 ) . A similar effect has been described for TEM-1 , whereby its disulfide bond becomes important for enzyme function under stress conditions ( temperature or pH stress ) ( Schultz et al . , 1987 ) . As SHV-27 has the narrowest hydrolytic spectrum out of all the enzymes tested , this result suggests that there could be a correlation between the hydrolytic spectrum of the β-lactamase and its dependence on DsbA for conferring resistance . In the case of colistin MICs , we did not implement a > 2 fold cutoff for observed decreases in MIC values as we did for strains expressing β-lactamases . Polymyxin antibiotics have a very narrow therapeutic window , and there is significant overlap between therapeutic and toxic plasma concentrations of colistin ( Nation et al . , 2016; Plachouras et al . , 2009 ) . Since patients that depend on colistin treatment are often severely ill , have multiple co-morbidities and are at high risk of acute kidney injury due to the toxicity of colistin , any reduction in the dose of colistin needed to achieve therapeutic activity is considered to be of value ( Nation et al . , 2019 ) . Expression of MCR enzymes in our wild-type E . coli K-12 strain resulted in colistin resistance ( MIC of 3 μg/mL or higher ) , while the strain harboring the empty vector was sensitive to colistin ( MIC of 1 μg/mL ) . In almost all tested cases , the absence of DsbA caused re-sensitization of the strains , as defined by the EUCAST breakpoint ( E . coli strains with an MIC of 2 μg/mL or below are classified as susceptible; Figure 1C ) , indicating that DsbA is important for MCR function . Taking into consideration the challenges when using colistin therapeutically ( Nation et al . , 2016; Plachouras et al . , 2009; Nation et al . , 2019 ) , we conclude that deletion of dsbA leads to clinically meaningful decreases in colistin MIC values for the tested MCR enzymes ( Figure 1C ) and that the role of DsbA in MCR function should be further investigated . Wild-type MIC values could be restored for all tested cysteine-containing enzymes by complementation of dsbA ( Figure 1—figure supplements 4 and 5 ) . Moreover , since DsbA acts on its substrates post-translationally , we performed a series of control experiments designed to assess whether the recorded effects were specific to the interaction of the resistance proteins with DsbA , and not a result of a general inability of the dsbA mutant strain to resist antibiotic stress . We observed no decreases in MIC values for the aminoglycoside antibiotic gentamicin , which is not affected by the activity of the tested enzymes ( Figure 1B , Figure 1—figure supplement 6 ) . Furthermore , the β-lactam MIC values of strains harboring the empty-vector alone , or a plasmid encoding L2-1 ( Figure 1B ) , a β-lactamase containing three cysteine residues , but no disulfide bond ( PDB ID: 1O7E ) , remained unchanged . Finally , to rule out the possibility that deletion of dsbA caused changes in cell envelope integrity that might confound our results , we measured the permeability of the outer and inner membrane of the dsbA mutant . To assess the permeability of the outer membrane , we used the fluorescent dye 1-N-phenylnaphthylamine ( NPN ) and complemented our results with vancomycin MIC assays ( Figure 1—figure supplement 7A ) . To test the integrity of the entire cell envelope , we used the fluorescent dye propidium iodide ( PI ) , as well as the β-galactosidase substrate chlorophenyl red-β-D-galactopyranoside ( CPRG ) ( Figure 1—figure supplement 7B ) . All four assays confirmed that the cell envelope integrity of the dsbA mutant is comparable to the parental strain ( Figure 1—figure supplement 7 ) . Together , these results indicate that many cell envelope AMR determinants that contain more than one cysteine residue are substrates of DsbA and that the process of disulfide bond formation is important for their activity . Unlike β-lactamases and MCR enzymes , none of the components of the six E . coli RND efflux pumps contain periplasmic cysteine residues ( Wang et al . , 2017 ) , and thus they are not substrates of the DSB system . Nonetheless , as DsbA assists the folding of approximately 300 extracytoplasmic proteins , and plays a central role in maintaining the homeostasis of the cell envelope proteome ( Kadokura et al . , 2004; Dutton et al . , 2008; Vertommen et al . , 2008 ) , we wanted to assess whether changes in periplasmic proteostasis that occur in its absence could indirectly influence efflux pump function . To do this , we determined the MIC values of three antibiotics that are RND efflux pump substrates using E . coli MG1655 , a model strain for efflux studies , its dsbA mutant , and a mutant lacking acrA , an essential component of the major E . coli RND pump AcrAB-TolC . MIC values for the dsbA mutant were lower than for the parental strain for all tested substrate antibiotics , but remained unchanged for the non-substrate gentamicin ( Figure 1D ) . This indicates that the MG1655 dsbA strain is generally able to resist antibiotic stress as efficiently as its parent , and that the recorded decreases in MIC values are specific to efflux pump function in the absence of DsbA . As expected for a gene deletion of a pump component , the acrA mutant had substantially lower MIC values for effluxed antibiotics ( Figure 1D ) . At the same time , even though gentamicin is not effluxed by AcrAB-TolC ( Nikaido , 1996 ) , the gentamicin MIC of the acrA mutant was twofold lower than that of E . coli MG1655 , in agreement with the fact that one of the minor RND pumps in E . coli , the aminoglycoside pump AcrD , is entirely reliant on AcrA for its function ( Aires and Nikaido , 2005; Rosenberg et al . , 2000; Yamasaki et al . , 2011 ) . As before , the observed phenotype could be reversed by complementation of dsbA ( Figure 1—figure supplement 8 ) and the recorded effects were not due to changes in membrane permeability ( Figure 1—figure supplement 9 ) . Chloramphenicol is the only antibiotic from the tested efflux pump substrates that has a EUCAST breakpoint for Gram-negative bacteria ( E . coli strains with an MIC of 8 μg/mL or below are classified as sensitive ) . It is notable that the MIC drop for this pump substrate , caused by deletion of dsbA , sensitized the E . coli MG1655 dsbA strain to chloramphenicol ( Figure 1D ) . Overall , the effect of DsbA absence on efflux pump efficiency is modest and much less substantial than that measured for a mutant lacking acrA ( 2–3-fold decrease in MIC versus 5–16-fold decrease , respectively ) ( Figure 1D ) . Nonetheless , the recorded decreases in MIC values are robust ( Figure 1D ) and in agreement with previous studies reporting that deletion of dsbA increases the sensitivity of E . coli to dyes like acridine orange and pyronin Y ( Bardwell et al . , 1991 ) , which are known substrates of AcrAB-TolC . While it is unlikely that the decreases in MIC values for effluxed antibiotics in the absence of DsbA are of clinical significance , it is interesting to explore the mechanistic relationship between DsbA and efflux pumps further , because there are very few examples of DsbA being important for the function of extra-cytoplasmic proteins independent from its disulfide bond forming capacity ( Alonso-Caballero et al . , 2018; Zheng et al . , 1997 ) . To understand the underlying mechanisms that result in the decreased MIC values observed for the dsbA mutant strains , we assessed the protein levels of a representative subset of β-lactamases ( GES-1 , L1-1 , KPC-3 , FRI-1 , OXA-4 , OXA-10 , OXA-198 ) and all tested MCR enzymes by immunoblotting . When expressed in the dsbA mutant , all Ambler class A and B β-lactamases ( Table 1 ) , except GES-1 which we were not able to visualize by immunoblotting , exhibited drastically reduced protein levels whilst the amount of the control enzyme L2-1 remained unaffected ( Figure 2A ) . This suggests that when these enzymes lack their disulfide bond , they are ultimately degraded . We did not detect any decrease in protein amounts for Ambler class D enzymes ( Table 1 , Figure 2B ) . However , the hydrolytic activity of these β-lactamases was significantly lower in the dsbA mutant ( Figure 2C ) , suggesting a folding defect that leads to loss of function . Like with class A and B β-lactamases , MCR enzymes were undetectable when expressed in a dsbA mutant ( Figure 3A ) suggesting that their stability or folding is severely compromised when they lack their disulfide bonds . We further confirmed this by directly monitoring the lipid A profile of all MCR-expressing strains where deletion of dsbA resulted in colistin MIC values of 2 µg/mL or lower ( i . e . strains expressing MCR-3 , –4 , –5 , and –8 , Figure 1C ) using MALDI-TOF mass spectrometry ( Figure 3BC ) . MCR activity leads to the addition of phosphoethanolamine to the lipid A portion of bacterial lipopolysaccharide ( LPS ) , resulting in reduced binding of colistin to LPS and , thus , resistance . In E . coli , the major lipid A peak detected by mass spectrometry is present at m/z 1796 . 2 ( Figure 3B , first spectrum ) and it corresponds to hexa-acyl diphosphoryl lipid A ( native lipid A ) . The lipid A profile of E . coli MC1000 dsbA was identical to that of the parental strain ( Figure 3B , second spectrum ) . In the presence of MCR enzymes two additional peaks were observed , at m/z 1821 . 2 and 1919 . 2 ( Figure 3B , third spectrum ) . The peak at m/z 1919 . 2 corresponds to the addition of a phosphoethanolamine moiety to the phosphate group at position 1 of native lipid A , and the peak at m/z 1821 . 2 corresponds to the addition of a phosphoethanolamine moiety to the 4ˊ phosphate of native lipid A and the concomitant loss of the phosphate group at position 1 ( Dortet et al . , 2018 ) . For dsbA mutants expressing MCR-3 , –5 , and –8 ( Figure 3C ) , the peaks at m/z 1821 . 2 and m/z 1919 . 2 could no longer be detected , whilst the native lipid A peak at m/z 1796 . 2 remained unchanged ( Figure 3B , fourth spectrum ) ; dsbA mutants expressing MCR-4 retain some basal lipid A-modifying activity , nonetheless this is not sufficient for this strain to efficiently evade colistin treatment ( Figure 1C ) . Together these data suggest that in the absence of DsbA , MCR enzymes are unstable ( Figure 3A ) and therefore no longer able to efficiently catalyze the addition of phosphoethanolamine to native lipid A ( Figure 3BC ) ; as a result , they cannot confer resistance to colistin ( Figure 1C ) . As RND efflux pump proteins do not contain any disulfide bonds , the decreases in MIC values for pump substrates in the absence of dsbA ( Figure 1D ) are likely mediated by additional cell-envelope components . The protease DegP , previously found to be a DsbA substrate ( Hiniker and Bardwell , 2004 ) , seemed a promising candidate for linking DsbA to efflux pump function . DegP degrades a range of misfolded extracytoplasmic proteins including , but not limited to , subunits of higher order protein complexes and proteins lacking their native disulfide bonds ( Clausen et al . , 2002 ) . We hypothesized that in a dsbA mutant the substrate burden on DegP would be dramatically increased , whilst DegP itself would not function optimally due to absence of its disulfide bond ( Skórko-Glonek et al . , 2003 ) . Consequently , protein turn over in the cell envelope would not occur efficiently . Since the essential RND efflux pump component AcrA needs to be cleared by DegP when it becomes misfolded or nonfunctional ( Gerken and Misra , 2004 ) , we expected that the reduced DegP efficiency in a dsbA mutant would result in accumulation of nonfunctional AcrA in the periplasm , which would then interfere with pump function . In agreement with our hypothesis , we found that in the absence of DsbA degradation of DegP occurred ( Figure 4A ) , reducing the pool of active enzyme ( Skórko-Glonek et al . , 2003 ) . In addition , AcrA accumulated to the same extent in a dsbA and a degP mutant ( Figure 4B ) , suggesting that in both these strains AcrA was not efficiently cleared . Finally , no accumulation was detected for the outer-membrane protein TolC ( Figure 4C ) , which is not a DegP substrate ( Werner et al . , 2003 ) . Thus , in the absence of DsbA , inefficient DegP-mediated periplasmic proteostasis affects RND efflux pumps ( Figure 1D ) through the accumulation of AcrA that should have been degraded and removed from the cell envelope . The data presented above validate our initial hypothesis . The absence of DsbA affects the stability and folding of cysteine-containing resistance proteins and in most cases leads to drastically reduced protein levels for the tested enzymes . As a result , and in agreement with the recorded decreases in MIC values ( Figure 1BC ) , these folding defects impede the ability of AMR determinants that are substrates of DsbA to confer resistance ( Figure 4D ) . In addition , changes in cell envelope protein homeostasis due to the lack of DSB activity can result in a generalized , albeit much more modest , effect on protein function in this compartment . This is suggested by the fact that prevention of disulfide bond formation seems to indirectly affect the AcrAB-TolC efflux pump ( Figure 1D ) , because of insufficient turnover of its AcrA component ( Figure 4D ) . DsbA is essential for the folding of many virulence factors . As such , inhibition of the DSB system has been proposed as a promising anti-virulence strategy ( Heras et al . , 2009; Landeta et al . , 2018; Heras et al . , 2015 ) and efforts have been made to develop inhibitors for DsbA ( Duprez et al . , 2015; Totsika et al . , 2018 ) , its redox partner DsbB ( Figure 1A; Landeta et al . , 2015 ) or both ( Halili et al . , 2015 ) . These studies have made the first steps toward the production of chemical compounds that inhibit the function of the DSB proteins , providing us with a laboratory tool to test our approach against AMR . 4 , 5-Dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one , termed ‘compound 12’ in Landeta et al . ( Landeta et al . , 2015 ) is a potent laboratory inhibitor of E . coli DsbB and its analogues from closely related organisms . Using this molecule , we could chemically inhibit the function of the DSB system . We first tested the motility of E . coli MC1000 in the presence of the inhibitor and found that cells were significantly less motile ( Figure 5AB ) , consistent with the fact that impairing DSB function prevents the formation of the flagellar P-ring component FlgI ( Dailey and Berg , 1993; Hizukuri et al . , 2006 ) . Furthermore , we directly assessed the redox state of DsbA in the presence of ‘compound 12’ to probe whether it was being re-oxidized by DsbB , a necessary step that occurs after each round of oxidative protein folding and allows DsbA to remain active ( Figure 1A ) . Under normal growth conditions , DsbA was in its active oxidized form in the bacterial periplasm ( i . e . C30 and C33 form a disulfide bond ) , showing that it was efficiently regenerated by DsbB ( Kishigami et al . , 1995; Figure 5C ) . By contrast , addition of the inhibitor to growing E . coli MC1000 cells resulted in accumulation of inactive reduced DsbA , thus confirming that DsbB function was impeded ( Figure 5C ) . After testing the efficacy of the DsbB inhibitor , we proceeded to examine whether chemical inhibition of the DSB system could be used to broadly impair the function of AMR determinants . We determined MIC values for the latest generation β-lactam that each β-lactamase can hydrolyze , or colistin , for our panel of E . coli MC1000 strains and found that addition of the compound during MIC testing phenocopied the effects of a dsbA deletion on β-lactamase and MCR activity ( Figure 5DE , Figure 5—figure supplement 1 , Supplementary file 2b ) . The observed effects are not a result of altered cell growth , as addition of the compound does not affect the growth profile of the bacteria ( Figure 5—figure supplement 2A ) , in agreement with the fact that deletion of dsbA does not affect cell viability ( Figure 5—figure supplement 2B ) . Furthermore , the changes in the recorded MIC values are due solely to inhibition of the DSB system as no additive effects on MIC values were observed when the dsbA mutant harboring a β-lactamase or mcr gene was exposed to the compound ( Figure 5—figure supplement 3 ) . Having shown that the DSB system is a tractable target in the context of AMR , we examined the effect of chemical inhibition on several species of β-lactamase-expressing Enterobacteria ( Suppementary Table 1 in Supplementary file 3 ) . We chose to test organisms that pose significant clinical or societal challenges , such as the ESKAPE pathogens Klebsiella pneumoniae and Enterobacter cloacae ( Mulani et al . , 2019 ) , or drug-resistant E . coli strains , which account for 50% of the economic burden of resistant infections ( O’Neill , 2014 ) . DSB system inhibition in a clinical isolate of K . pneumoniae expressing KPC-2 sensitized the strain to imipenem as defined by EUCAST breakpoints ( Figure 6A ) . The efficiency of this double treatment is evident from scanning electron micrographs of the tested strains ( Figure 6B ) . Addition of either the DSB system inhibitor or imipenem alone does not cause any changes in the morphology of K . pneumoniae cells , which remain healthy and dividing ( Figure 6B , top row ) . By contrast , the combination of the inhibitor with imipenem ( added at a sub-MIC final concentration of 6 µg/mL ) , led to dramatic changes in the appearance of the cells , whose integrity was entirely compromised ( Figure 6B , bottom row ) . Similarly , E . coli and Citrobacter freundii isolates expressing KPC-2 , including multidrug-resistant strains , also showed clinically relevant decreases in their MIC values for imipenem that resulted in sensitization when their DSB system was chemically inhibited ( Figure 6C ) . For an E . cloacae isolate expressing FRI-1 , chemical inhibition of DsbA caused reduction in its aztreonam MIC value by over 180 µg/mL , resulting in intermediate resistance as defined by EUCAST breakpoints ( Figure 6D ) . Along with β-lactamase-expressing strains , we also tested the effect of DsbA inhibition on MCR-producing clinical isolates . We found that combination of the DSB system inhibitor with colistin led to reduction of the colistin MIC and sensitization of MCR-1-expressing multidrug-resistant E . coli ( Figure 7A ) . In agreement with this , SEM images of this strain after combination treatment using sub-MIC amounts of colistin ( final concentration of 2 µg/mL ) revealed drastic changes in morphology , whereby cells blebbed intensely or their contents leaked out ( Figure 7B ) . We tested eight additional clinical E . coli isolates that encode diverse MCR enzymes ( most of which are multidrug resistant ) and have colistin MICs ranging from 3 to 16 µg/mL; DSB system inhibition also allowed sensitization to colistin ( Figure 7C ) for tested strains . At the same time , we were able to show that DSB system inhibition in E . coli CNR1790 ( i . e . the clinical isolate expressing both MCR-1 and the ESBL TEM-15 that was sensitized to colistin in Figure 7A ) , led to a decrease in its ceftazidime MIC , resulting in intermediate resistance ( Figure 7D ) . While we did not test the dependence of TEM enzymes on DsbA in our panel of E . coli K-12 strains , we chose to test the effects of DSB system inhibition on E . coli CNR1790 because we posited that the disulfide bond in TEM-15 may be important for its function , based on the fact that the narrow-spectrum TEM-1 enzyme has been shown to be reliant on its disulfide under stress conditions ( Schultz et al . , 1987 ) . Validation of our hypothesis provides evidence that DsbA inhibition can improve the resistance profile of the same isolate both for β-lactam ( Figure 7D ) and polymyxin ( Figure 7A ) antibiotics . Together these results , obtained using multiple clinical strains from several bacterial species , provide further proof of the significance of our data from heterologously expressed β-lactamase and MCR enzymes in E . coli K-12 strains ( Figure 1BC ) , and showcase the potential of this approach for clinical applications . To determine if our approach for Enterobacteria would be appropriate for other multidrug-resistant Gram-negative bacteria , we tested it on another major ESKAPE pathogen , Pseudomonas aeruginosa ( Mulani et al . , 2019 ) . This bacterium has two DsbB analogues which are functionally redundant ( Arts et al . , 2013 ) . The chemical inhibitor used in this study has been shown to be effective against DsbB1 , but much less effective against DsbB2 of P . aeruginosa PA14 ( Landeta et al . , 2015 ) , making it unsuitable for MIC assays on P . aeruginosa clinical isolates . Nonetheless , deletion of dsbA1 in a multidrug-resistant P . aeruginosa clinical isolate expressing OXA-198 ( PA43417 ) , led to sensitization of this strain to the antipseudomonal β-lactam piperacillin ( Figure 8A ) . In addition , we deleted dsbA1 in the multidrug-resistant P . aeruginosa PAe191 strain that produces OXA-19 , a member of the OXA-10 phylogenetic family ( Supplementary file 1 ) and the most disseminated OXA enzyme in clinical strains ( Mugnier et al . , 1998 ) . In this case , absence of DsbA caused a drastic reduction in the ceftazidime MIC value by over 220 µg/mL , and sensitized the strain to aztreonam ( Figure 8B ) . These results suggest that targeting disulfide bond formation could be useful for the sensitization of many more clinically important Gram-negative species . Finally , to test our approach in an infection context we performed in vivo survival assays using the wax moth model Galleria mellonella ( Figure 8C ) . G . mellonella has proven to be an invaluable non-vertebrate model for the study of P . aeruginosa pathogenesis as well as for testing antibiotic treatments against this organism ( Hill et al . , 2014; Miyata et al . , 2003 ) , making it an appropriate tool for assessing the in vivo efficacy of our approach on a multidrug-resistant strain of this pathogen . Larvae were infected with the P . aeruginosa PAe191 strain producing OXA-19 , and its dsbA1 mutant , and infections were treated once with ceftazidime at a final concentration below the EUCAST breakpoint . No larvae survived beyond 18 hr post infection with P . aeruginosa PAe191 , even when treatment with ceftazidime was performed ( Figure 8C , blue and red survival curves ) . Deletion of dsbA1 resulted in 80% mortality of the larvae at 50 hr post infection ( Figure 8C , light blue survival curve ) ; this increase in survival compared to larvae infected with P . aeruginosa PAe191 is due to the fact that absence of the principal DsbA protein likely affects the virulence of the pathogen ( Landeta et al . , 2019 ) . Nonetheless , treatment of the dsbA1 mutant with ceftazidime resulted in a significant increase in survival ( 17% mortality ) compared to the untreated condition , 50 hr post infection ( Figure 8C , compare the light blue and pink survival curves ) . This improvement in survival is even more noticeable if one compares the survival of larvae treated with ceftazidime after infection with P . aeruginosa PAe191 versus infection with P . aeruginosa PAe191 dsbA1 ( Figure 8C , compare the red and pink survival curves ) . Since OXA-19 , in this case produced by a multi-drug resistant clinical strain ( Supplementaty Table 1 in Supplementary file 3 , Figure 8B ) , is a broad-spectrum β-lactamase that cannot be neutralized by classical β-lactamase inhibitors ( Table 1 ) , these results further highlight the promise of our approach for future clinical applications .
This work is one of the first reports of a strategy capable of simultaneously impairing multiple types of AMR determinants by compromising the function of a single target . By inhibiting DsbA , a non-essential cell envelope protein which is unique to bacteria , we can inactivate diverse resistance enzymes and sensitize critically important pathogens to several existing antibiotics . This proof of principle will hopefully further incentivize the development of DsbA inhibitors and open new avenues toward the inception of novel adjuvants that will help reverse AMR in Gram-negative organisms . We have shown that targeting DsbA incapacitates broad-spectrum β-lactamases from three of the four Ambler classes ( class A , B and D , Figure 1B ) . This includes enzymes that are not susceptible to classical β-lactamase inhibitors ( Table 1 ) , such as members of the KPC and OXA families , as well as metallo-β-lactamases like L1-1 from the often pan-resistant organism Stenotrophomonas maltophilia . The function of these proteins is impaired without a small molecule binding to their active site , unlike most of the currently-used β-lactamase inhibitors which often generate resistance ( Laws et al . , 2019 ) . As DsbA dependence is conserved within phylogenetic groups ( Figure 1—figure supplement 2 ) , based on the number of enzymes belonging to the same phylogenetic family as the β-lactamases tested in this study ( Supplementary file 1 ) , we anticipate that a total of 195 discrete enzymes rely on DsbA for their stability and function , 84 of which cannot be inhibited by classical adjuvant approaches . DsbA is widely conserved ( Heras et al . , 2009 ) , thus targeting the DSB system should not only compromise β-lactamases in Enterobacteria but , as demonstrated by our experiments using P . aeruginosa clinical isolates ( Figure 8 ) , could also be a promising avenue for impairing the function of AMR determinants expressed by other highly-resistant Gram-negative organisms . As such , together with the fact that approximately 56% of the β-lactamase phylogenetic families found in pathogens and organisms capable of causing opportunistic infections contain enzymes with two or more cysteines ( Supplementary file 1 ) , we expect many more clinically relevant β-lactamases , beyond those already tested in this study , to depend on DsbA . MCR enzymes are rapidly becoming a grave threat to the use of colistin ( Sun et al . , 2018 ) , a drug of last resort often needed for the treatment of multidrug-resistant infections ( Li et al . , 2006 ) . Currently , experimental inhibitors of these proteins are sparse and poorly characterized ( Zhou et al . , 2019 ) , and only one existing compound , the antirheumatic drug auranofin , seems to successfully impair MCR enzymes , through displacement of their zinc cofactor ( Sun et al . , 2020 ) . As all MCR members contain multiple disulfide bonds , inhibition of the DSB system provides a broadly applicable solution for reversing MCR-mediated colistin resistance ( Figures 1C , 5E and 7ABC ) that would likely extend to novel MCR proteins that may emerge in the future . Since the decrease in colistin MIC values upon dsbA deletion ( Figure 1C ) or DsbB inhibition ( Figures 5E and 7ABC ) is modest , this phenotype cannot be used in future screens aiming to identify DsbA inhibitors , because such applications require a larger than 4-fold decrease in recorded MIC values to reliably identify promising lead compounds . Nonetheless , our findings in this study clearly demonstrate that absence of DsbA results in degradation of MCR enzymes and abrogation of their function ( Figure 3 ) , which , in turn , leads to sensitization of all tested E . coli clinical isolates to colistin ( Figure 7 ) . This adds to other efforts aiming to reduce the colistin MIC of polymyxin resistant strains ( Minrovic et al . , 2019; Zimmerman et al . , 2020 ) . As such , if a clinically useful DsbA inhibitor were to become available , it would be valuable to test its efficacy against large panels of MCR-expressing clinical strains , as it might offer a new way to bypass MCR-mediated colistin resistance . No clinically applicable efflux pump inhibitors have been identified to date ( Sharma et al . , 2019 ) despite many efforts to target these macromolecular assemblies as a way to overcome intrinsic resistance . While deletion of dsbA sensitizes the tested E . coli strain to chloramphenicol , the overall effects of DsbA absence on efflux function are modest at best ( Figure 1D ) . That said , our investigation of the relationship between DsbA-mediated proteostasis and pump function ( Figure 4A–C ) highlights the importance of other cell envelope proteins responsible for protein homeostasis , such as DegP , for bacterial efflux . Since the cell envelope contains multiple protein folding catalysts ( Goemans et al . , 2014 ) , it would be worth testing if other redox proteins , chaperones , or proteases could be targeted to indirectly compromise efflux pumps . More generally , our findings demonstrate that cell envelope proteostasis pathways have significant , yet untapped , potential for the development of novel antibacterial strategies . The example of the DSB system presented here is particularly telling . This pathway , initially considered merely a housekeeping system ( Kadokura et al . , 2003 ) , plays a major role in clinically relevant bacterial niche adaptation . In addition to assisting the folding of 40% of the cell-envelope proteome ( Dutton et al . , 2008; Vertommen et al . , 2008 ) , the DSB system is essential for virulence ( Heras et al . , 2009; Landeta et al . , 2018 ) , has a key role in the formation and awakening of bacterial persister cells ( Wilmaerts et al . , 2019 ) and , as seen in this work , is required for bacterial survival in the presence of widely used antibiotic compounds . As shown in our in vivo experiments ( Figure 8C ) , targeting such a system in Gram-negative pathogens could lead to adjuvant approaches that inactivate AMR determinants whilst simultaneously incapacitating an arsenal of virulence factors . Therefore , this study not only lays the groundwork for future clinical applications , such as the development of broad-acting antibiotic adjuvants , but also serves as a paradigm for exploiting other accessible cell envelope proteostasis processes for the design of next-generation therapeutics .
Unless otherwise stated , chemicals and reagents were acquired from Sigma Aldrich , growth media were purchased from Oxoid and antibiotics were obtained from Melford Laboratories . Lysogeny broth ( LB ) ( 10 g/L NaCl ) and agar ( 1 . 5% w/v ) were used for routine growth of all organisms at 37 °C with shaking at 220 RPM , as appropriate . Unless otherwise stated , Mueller-Hinton ( MH ) broth and agar ( 1 . 5% w/v ) were used for Minimum Inhibitory Concentration ( MIC ) assays . Growth media were supplemented with the following , as required: 0 . 25 mM Isopropyl β-D-1-thiogalactopyranoside ( IPTG ) ( for strains harboring β-lactamase-encoding pDM1 plasmids ) , 0 . 5 mM IPTG ( for strains harboring MCR-encoding pDM1 plasmids ) , 12 . 5 μg/mL tetracycline , 100 μg/mL ampicillin , 50 μg/mL kanamycin , 10 μg/mL gentamicin , 33 μg/mL chloramphenicol , 50 μg/mL streptomycin ( for cloning purposes ) , and 2000–5000 μg/mL streptomycin ( for the construction of Pseudomonas aeruginosa mutants ) . Bacterial strains and plasmids used in this study are listed in the Key Resources Table and in Supplementary file 3 - Supplementary Tables 2 and 3 , respectively . Oligonucleotides used in this study are listed in Supplementary Table 4 . DNA manipulations were conducted using standard methods . KOD Hot Start DNA polymerase ( Merck ) was used for all PCR reactions according to the manufacturer’s instructions , oligonucleotides were synthesized by Sigma Aldrich and restriction enzymes were purchased from New England Biolabs . All DNA constructs were sequenced and confirmed to be correct before use . Genes for β-lactamase and MCR enzymes were amplified from genomic DNA extracted from clinical isolates ( Supplementary file 3 - Supplementary Table 5 ) with the exception of mcr-3 and mcr-8 , which were synthesized by GeneArt Gene Synthesis ( ThermoFisher Scientific ) . β-lactamase and MCR genes were cloned into the IPTG-inducible plasmid pDM1 using primers P1-P34 . pDM1 ( GenBank accession number MN128719 ) was constructed from the p15A-ori plasmid pACYC184 ( Chang and Cohen , 1978 ) to contain the Lac repressor , the Ptac promoter , an optimized ribosome binding site and a multiple cloning site ( NdeI , SacI , PstI , KpnI , XhoI , and XmaI ) inserted into the NcoI restriction site of pACYC184 . All StrepII-tag fusions of β-lactamase and MCR enzymes ( constructed using primers P1 , P3 , P9 , P11 , P13 , P15 , P17 , P21 , P23 , P25 , P27 , P29 , P35 , P36 , and P39-P48 ) have a C-terminal StrepII tag ( GSAWSHPQFEK ) except for OXA-4 , where an N-terminal StrepII tag was inserted between the periplasmic signal sequence and the body of the protein using the primer pairs P7/P38 , P9/P37 , and P7/P8 . Plasmids encoding ges-1 and kpc-3 were obtained by performing QuickChange mutagenesis on pDM1 constructs encoding ges-5 and kpc-2 , respectively ( primers P31-P34 ) . E . coli gene mutants were constructed using a modified lambda-Red recombination method , as previously described ( Kim et al . , 2014 ) ( primers P51-P58 ) . To complement the dsbA mutant , a DNA fragment consisting of dsbA preceded by the Ptac promoter was inserted into the NotI/XhoI sites of pGRG25 ( primers P49/P50 ) and was reintroduced into the E . coli chromosome at the attTn7 site , as previously described ( McKenzie and Craig , 2006 ) . The dsbA1 mutants of the P . aeruginosa PA43417 and P . aeruginosa PAe191 clinical isolates were constructed by allelic exchange , as previously described ( Vasseur et al . , 2005 ) . Briefly , the dsbA1 gene area of P . aeruginosa PA43417 and P . aeruginosa PAe191 ( including the dsbA1 gene and 600 bp on either side of this gene ) was amplified ( primers P59/P60 ) and the obtained DNA was sequenced to allow for accurate primer design for the ensuing cloning step . Subsequently , 500 bp DNA fragments upstream and downstream of the dsbA1 gene were amplified using P . aeruginosa PA43417 genomic DNA ( primers P61/P62 [upstream] and P63/P64 [downstream] ) . A fragment containing both regions was obtained by overlapping PCR ( primers P61/P64 ) and inserted into the XbaI/BamHI sites of pKNG101 . The suicide vector pKNG101 ( Kaniga et al . , 1991 ) is not replicative in P . aeruginosa; it was maintained in E . coli CC118λpir and mobilized into P . aeruginosa PA43417 and P . aeruginosa PAe191 by triparental conjugation . Unless otherwise stated , antibiotic MIC assays were carried out in accordance with the EUCAST recommendations using ETEST strips ( BioMérieux ) . Briefly , overnight cultures of each strain to be tested were standardized to OD600 0 . 063 in 0 . 85% NaCl ( equivalent to McFarland standard 0 . 5 ) and distributed evenly across the surface of MH agar plates . ETEST strips were placed on the surface of the plates , evenly spaced , and the plates were incubated for 18–24 hr at 37 °C . MICs were read according to the manufacturer’s instructions . β-lactam MICs were also determined using the Broth Microdilution ( BMD ) method , as required . Briefly , a series of antibiotic concentrations was prepared by twofold serial dilution in MH broth in a clear-bottomed 96-well microtiter plate ( Corning ) . When used , tazobactam was included at a fixed concentration of 4 μg/mL in every well , in accordance with the EUCAST guidelines . The strain to be tested was added to the wells at approximately 5 × 104 colony-forming units ( CFU ) per well and plates were incubated for 18–24 hr at 37 °C . The MIC was defined as the lowest antibiotic concentration with no visible bacterial growth in the wells . Vancomycin MICs were determined using the BMD method , as above . All colistin sulphate MIC assays were performed using the BMD method as described above except that instead of twofold serial dilutions , the following concentrations of colistin ( Acros Organics ) were prepared individually in MH broth: 32 μg/mL , 16 μg/mL , 12 μg/mL , 8 μg/mL , 7 μg/mL , 6 μg/mL , 5 . 5 μg/mL , 5 μg/mL , 4 . 5 μg/mL , 4 μg/mL , 3 . 5 μg/mL , 3 μg/mL , 2 . 5 μg/mL , 2 μg/mL , 1 . 5 μg/mL , 1 μg/mL , 0 . 5 μg/mL . The covalent DsbB inhibitor 4 , 5-dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one ( Landeta et al . , 2015 ) was used to chemically impair the function of the DSB system . Inactivation of DsbB results in abrogation of DsbA function ( Kishigami et al . , 1995 ) only in media free of small-molecule oxidants ( Dailey and Berg , 1993 ) . Therefore , MIC assays involving chemical inhibition of the DSB system were performed using M63 broth ( 15 . 1 mM ( NH4 ) 2SO4 , 100 mM KH2PO4 , 1 . 8 mM FeSO4 . 7H2O , adjusted to pH 7 . 2 with KOH ) and agar ( 1 . 5% w/v ) supplemented with 1 mM MgSO4 , 0 . 02% w/v glucose , 0 . 005% w/v thiamine , 31 µM FeCl3 . 6H2O , 6 . 2 μM ZnCl2 , 0 . 76 µM CuCl2 . 2H2O , 1 . 62 µM H3BO3 , 0 . 081 µM MnCl2 . 4H2O , 84 . 5 mg/L alanine , 19 . 5 mg/L arginine , 91 mg/L aspartic acid , 65 mg/L glutamic acid , 78 mg/L glycine , 6 . 5 mg/L histidine , 26 mg/L isoleucine , 52 mg/L leucine , 56 . 34 mg/L lysine , 19 . 5 mg/L methionine , 26 mg/L phenylalanine , 26 mg/L proline , 26 mg/L serine , 6 . 5 mg/L threonine , 19 . 5 mg/L tyrosine , 56 . 34 mg/L valine , 26 mg/L tryptophan , 26 mg/L asparagine and 26 mg/L glutamine . CaCl2 was also added at a final concentration of 0 . 223 mM for colistin sulfate MIC assays . Either DMSO ( vehicle control ) or the covalent DsbB inhibitor 4 , 5-dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one ( final concentration of 50 μM ) ( Enamine ) ( Landeta et al . , 2015 ) were added to the M63 medium , as required . The strain to be tested was added at an inoculum that recapitulated the MH medium MIC values obtained for that strain . Samples for immunoblotting were prepared as follows . Strains to be tested were grown on LB or MH agar plates as lawns in the same manner as for MIC assays described above . Bacteria were collected using an inoculating loop and resuspended in 0 . 85% NaCl or LB to OD600 2 . 0 ( except for strains expressing OXA-4 , where OD600 6 . 0 was used ) . For strains expressing β-lactamase enzymes , the cell suspensions were spun at 10 , 000 x g for 10 min and bacterial pellets were lysed by addition of BugBuster Master Mix ( Merck Millipore ) for 25 min at room temperature with gentle agitation . Subsequently , lysates were spun at 10 , 000 x g for 10 min at 4 °C and the supernatant was added to 4 x Laemmli buffer . For strains expressing MCR enzymes cell suspensions were directly added to 4 x Laemmli buffer , while for E . coli MG1655 and its mutants , cells were lysed as above and lysates were added to 4 x Laemmli buffer . All samples were boiled for 5 min before separation by SDS-PAGE . Unless otherwise stated , SDS-PAGE analysis was carried out using 10% BisTris NuPAGE gels ( ThermoFisher Scientific ) using MES/SDS running buffer prepared according to the manufacturer’s instructions and including pre-stained protein markers ( SeeBlue Plus 2 , ThermoFisher Scientific ) . Proteins were transferred to Amersham Protran nitrocellulose membranes ( 0 . 45 µm pore size , GE Life Sciences ) using a Trans-Blot Turbo transfer system ( Bio-Rad ) before blocking in 3% w/v Bovine Serum Albumin ( BSA ) /TBS-T ( 0 . 1 % v/v Tween 20 ) or 5% w/v skimmed milk/TBS-T and addition of primary and secondary antibodies . The following primary antibodies were used in this study: Strep-Tactin-HRP conjugate ( Iba Lifesciences ) ( dilution 1:3 , 000 in 3 w/v % BSA/TBS-T ) , Strep-Tactin-AP conjugate ( Iba Lifesciences ) ( dilution 1:3 , 000 in 3 w/v % BSA/TBS-T ) , rabbit anti-DsbA antibody ( dilution 1:1 , 000 in 5 w/v % skimmed milk/TBS-T ) , rabbit anti-AcrA antibody ( dilution 1:10 , 000 in 5 w/v % skimmed milk/TBS-T ) , rabbit anti-TolC antibody ( dilution 1:5 , 000 in 5 w/v % skimmed milk/TBS-T ) , rabbit anti-HtrA1 ( DegP ) antibody ( Abcam ) ( dilution 1:1 , 000 in 5 w/v % skimmed milk/TBS-T ) and mouse anti-DnaK 8E2/2 antibody ( Enzo Life Sciences ) ( dilution 1:10 , 000 in 5% w/v skimmed milk/TBS-T ) . The following secondary antibodies were used in this study: goat anti-rabbit IgG-AP conjugate ( Sigma Aldrich ) ( dilution 1:6 , 000 in 5% w/v skimmed milk/TBS-T ) , goat anti-rabbit IgG-HRP conjugate ( Sigma Aldrich ) ( dilution 1:6 , 000 in 5% w/v skimmed milk/TBS-T ) , goat anti-mouse IgG-AP conjugate ( Sigma Aldrich ) ( dilution 1:6 , 000 in 5% w/v skimmed milk/TBS-T ) and goat anti-mouse IgG-HRP conjugate ( Sigma Aldrich ) ( dilution 1:6000 in 5% w/v skimmed milk/TBS-T ) . Membranes were washed three times for 5 min with TBS-T prior to development . Development for AP conjugates was carried out using a SigmaFast BCIP/NBT tablet , while HRP conjugates were visualized with the Novex ECL HRP chemiluminescent substrate reagent kit ( ThermoFisher Scientific ) or the Immobilon Crescendo chemiluminescent reagent ( Merck ) using a Gel Doc XR + Imager ( Bio-Rad ) . β-lactam hydrolysis measurements were carried out using the chromogenic β-lactam nitrocefin ( Abcam ) . Briefly , overnight cultures of strains to be tested were centrifugated , pellets were weighed and resuspended in 150 μL of 100 mM sodium phosphate buffer ( pH 7 . 0 ) per 1 mg of wet-cell pellet , and cells were lysed by sonication . For strains harboring pDM1 , pDM1-blaL2-1 , pDM1-blaOXA-10 , and pDM1-blaGES-1 , lysates corresponding to 0 . 34 mg of bacterial pellet were transferred into clear-bottomed 96-well microtiter plates ( Corning ) . For strains harboring pDM1-blaOXA-4 and pDM1-blaOXA-198 , lysates corresponding to 0 . 2 mg and 0 . 014 mg of bacterial pellet were used , respectively . In all cases , nitrocefin was added at a final concentration of 400 μM and the final reaction volume was made up to 100 μL using 100 mM sodium phosphate buffer ( pH 7 . 0 ) . Nitrocefin hydrolysis was monitored at 25 °C by recording absorbance at 490 nm at 60-s intervals for 15 min using an Infinite M200 Pro microplate reader ( Tecan ) . The amount of nitrocefin hydrolyzed by each lysate in 15 min was calculated using a standard curve generated by acid hydrolysis of nitrocefin standards . 1-N-phenylnaphthylamine ( NPN ) ( Acros Organics ) uptake assays were performed performed as previously described ( Helander and Mattila-Sandholm , 2000 ) . Briefly , mid-log phase cultures of strains to be tested were diluted to OD600 0 . 5 in 5 mM HEPES ( pH 7 . 2 ) before transfer to clear-bottomed 96-well microtiter plates ( Corning ) and addition of NPN at a final concentration of 10 μM . Colistin sulphate ( Acros Organics ) was included at a final concentration of 0 . 5 μg/mL , as required . Immediately after the addition of NPN , fluorescence was measured at 60-s intervals for 10 min using an Infinite M200 Pro microplate reader ( Tecan ) ; the excitation wavelength was set to 355 nm and emission was recorded at 405 nm . Exponentially-growing ( OD600 0 . 4 ) E . coli strains harboring pUltraGFP-GM ( Mavridou et al . , 2016 ) were diluted to OD600 0 . 1 in phosphate buffered saline ( PBS ) ( pH 7 . 4 ) and cecropin A was added to a final concentration of 20 μM , as required . Cell suspensions were incubated at room temperature for 30 min before centrifugation and resuspension of the pellets in PBS . Propidium iodide ( PI ) was then added at a final concentration of 3 μM . Suspensions were incubated for 10 min at room temperature and analyzed on a two-laser , four-color BD FACSCalibur flow cytometer ( BD Biosciences ) . 50 , 000 events were collected for each sample and data were analyzed using FlowJo v . 10 . 0 . 6 ( Treestar ) . The cell envelope integrity of bacterial strains used in this study and of their dsbA mutants , was tested by measuring the hydrolysis of the β-galactosidase substrate chlorophenyl red-β-D-galactopyranoside ( CPRG ) by cytoplasmic LacZ , as previously described ( Paradis-Bleau et al . , 2014 ) . Briefly , exponentially growing ( OD600 0 . 4 ) E . coli MC1000 harboring pCB112 or MG1655 , as well as their dsbA mutants , were diluted to 1:105 in MH broth and plated on MH agar containing CPRG and IPTG at final concentrations of 20 μg/mL and 50 μM , respectively . Plates were incubated at 37 °C for 18 hr , were photographed , and images were analyzed using Adobe Photoshop CS4 extended v . 11 . 0 ( Adobe ) as follows . Images were converted to CMYK color space format , colonies were manually selected using consistent tolerance ( 26 , anti-alias , contiguous ) and edge refinement ( 32 px , 100% contrast ) , and the magenta color was quantified for each image and normalized for the area occupied by each colony . Lipid A profiles of strains to be tested were determined using intact bacteria , as previously described ( Larrouy-Maumus et al . , 2016 ) . The peak for E . coli native lipid A is detected at m/z 1796 . 2 , whereas the lipid A profiles of strains expressing functional MCR enzymes have two additional peaks , at m/z 1821 . 2 and 1919 . 2 . These peaks result from MCR-mediated modification of native lipid A through addition of phosphoethanolamine moieties ( Dortet et al . , 2018 ) . The ratio of modified to unmodified lipid A was calculated by summing the intensities of the peaks at m/z 1821 . 2 and 1919 . 2 and dividing this value by the intensity of the native lipid A peak at m/z 1796 . 2 . A total of 500 μL of overnight culture of each strain to be tested were centrifuged and the pellets were washed three times in M63 broth before resuspension in the same medium to achieve a final volume of 25 μL . Bacterial motility was assessed by growth in M63 medium containing 0 . 25% w/v agar supplemented as described above . DMSO ( vehicle control ) or the covalent DsbB inhibitor 4 , 5-dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one ( final concentration of 50 μM ) ( Enamine ) were added to the medium , as required . One μL of the washed cell suspension was inoculated into the center of a 90-mm diameter agar plate , just below the surface of the semi-solid medium . Plates were incubated at 37 °C in a humidified environment for 16–18 hr and growth halo diameters were measured . Bacterial strains to be tested were grown for 18 hr in M63 broth supplemented as described above . DMSO ( vehicle control ) or the covalent DsbB inhibitor 4 , 5-dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one ( final concentration of 50 μM ) ( Enamine ) were added to the medium , as required . Cultures were standardized to OD600 2 . 0 in M63 broth , spun at 10 , 000 x g for 10 min and bacterial pellets lysed by addition of BugBuster Master Mix ( Merck Millipore ) for 25 min at room temperature with gentle agitation . Subsequently , lysates were spun at 10 , 000 x g for 10 min at 4 °C prior to reaction with 4-acetamido-4ˊ-maleimidyl-stilbene-2 , 2ˊ-disulfonic acid ( AMS ) ( ThermoFisher Scientific ) . AMS alkylation was performed by vortexing the lysates in 15 mM AMS , 50 mM Tris-HCl , 3% w/v SDS and 3 mM EDTA ( pH 8 . 0 ) for 30 min at 25 °C , followed by incubation at 37 °C for 10 min . SDS-PAGE analysis and immunoblotting was carried out as described above , except that 12% BisTris NuPAGE gels ( ThermoFisher Scientific ) and MOPS/SDS running buffer were used . DsbA was detected using a rabbit anti-DsbA primary antibody and an AP-conjugated secondary antibody , as described above . To assess the effect of DSB system inhibition of the growth of E . coli , overnight cultures of the strains to be tested were centrifuged and the pellets were washed three times in M63 broth before transfer to clear-bottomed 96-well microtiter plates ( Corning ) at approximately 5 × 107 CFU/well ( starting OD600 ~0 . 03 ) . M63 broth supplemented as described above was used as a growth medium . DMSO ( vehicle control ) or the covalent DsbB inhibitor 4 , 5-dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one ( final concentration of 50 μM ) ( Enamine ) were added to the medium , as required . Plates were incubated at 37 °C with orbital shaking ( amplitude 3 mm , equivalent to ~220 RPM ) and OD600 was measured at 900-s intervals for 18 hr using an Infinite M200 Pro microplate reader ( Tecan ) . The same experimental setup was also used for recording growth curves of E . coli strains and their isogenic mutants , except that overnight cultures of the strains to be tested were diluted 1:100 into clear-bottomed 96-well microtiter plates ( Corning ) ( starting OD600 ~0 . 01 ) and that LB was used as the growth medium . The wax moth model Galleria mellonella was used for in vivo survival assays ( McCarthy et al . , 2017 ) . Individual G . mellonella larvae were randomly allocated to experimental groups; no masking was used . Overnight cultures of the strains to be tested were standardized to OD600 1 . 0 , suspensions were centrifuged and the pellets were washed three times in PBS and serially diluted . Ten μl of the 1:10 dilution of each bacterial suspension was injected into the last right abdominal proleg of 30 G . mellonella larvae per condition; an additional 10 larvae were injected with PBS as negative control . Immediately after infection , larvae were injected with 4 μl of ceftazidime to a final concentration of 7 . 5 μg/ml in the last left abdominal proleg . The larvae mortality was monitored for 50 hr . Death was scored when larvae turned black due to melanization , and did not respond to physical stimulation . Bacterial strains to be tested were grown for 18 hr in MH broth; the covalent DsbB inhibitor 4 , 5-dichloro-2- ( 2-chlorobenzyl ) pyridazin-3-one ( final concentration of 50 μM ) ( Enamine ) was added to the medium , as required . Cells were centrifuged , the pellets were washed three times in M63 broth , and cell suspensions were diluted 1:500 into the same medium supplemented as described above; the covalent DsbB inhibitor ( final concentration of 50 μM ) and/or antibiotics ( final concentrations of 6 μg/mL and 2 μg/mL of imipenem and colistin , respectively ) were added to the cultures , as required . After 1 hr of incubation as described above , 25 μl of each culture was spotted onto positively charged glass microscope slides and allowed to air-dry . Cells were then fixed with glutaraldehyde ( 2 . 5% v/v in PBS ) for 30 min at room temperature and the slide was washed five times in PBS . Subsequently , each sample was dehydrated using increasing concentrations of ethanol ( 5% v/v , 10% v/v , 20% v/v , 30% v/v , 50% v/v , 70% v/v , 90% v/v [applied three times] and 100% v/v ) , with each wash being carried out by application and immediate removal of the washing solution , before a 7-nm coat of platinum/palladium was applied using a Cressington 208 benchtop sputter coater . Images were obtained on a Zeiss Supra 40 V Scanning Electron Microscope at 5 . 00 kV and with ×26 , 000 magnification . The total numbers of performed biological experiments and technical repeats are mentioned in the figure legend of each display item . Biological replication refers to completely independent repetition of an experiment using different biological and chemical materials . Technical replication refers to independent data recordings using the same biological sample . For MIC assays , all recorded values are displayed in the relevant graphs; for MIC assays where three or more biological experiments were performed , the bars indicate the median value , while for assays where two biological experiments were performed the bars indicate the most conservative of the two values ( i . e . for increasing trends , the value representing the smallest increase and for decreasing trends , the value representing the smallest decrease ) . For all other assays , statistical analysis was performed in GraphPad Prism v8 . 0 . 2 using an unpaired T-test with Welch’s correction , a one-way ANOVA with correction for multiple comparisons , or a Mantel-Cox logrank test , as appropriate . Statistical significance was defined as p < 0 . 05 . Outliers were defined as any technical repeat >2 SD away from the average of the other technical repeats within the same biological experiment . Such data were excluded and all remaining data were included in the analysis . Detailed information for each figure is provided below: Figure 2C unpaired T-test with Welch’s correction; n = 3; 3 . 621 degrees of freedom , t-value = 0 . 302 , p = 0 . 7792 ( non-significance ) ( for pDM1 strains ) ; 3 . 735 degrees of freedom , t-value = 0 . 4677 , p = 0 . 666 ( non-significance ) ( for pDM1-blaL2-1 strains ) ; 2 . 273 degrees of freedom , t-value = 5 . 069 , p = 0 . 0281 ( significance ) ( for pDM1-blaGES-1 strains ) ; 2 . 011 degrees of freedom , t-value = 6 . 825 , p = 0 . 0205 ( significance ) ( for pDM1-blaOXA-4 strains ) ; 2 . 005 degrees of freedom , t-value = 6 . 811 , p = 0 . 0208 ( significance ) ( for pDM1-blaOXA-10 strains ) ; 2 . 025 degrees of freedom , t-value = 5 . 629 , p = 0 . 0293 ( significance ) ( for pDM1-blaOXA-198 strains ) . Figure 3C one-way ANOVA with Tukey’s multiple comparison test; n = 4; 24 degrees of freedom; F value = 21 . 00; p = 0 . 000000000066 ( for pDM1-mcr-3 strains ) , p = 0 . 0004 ( for pDM1-mcr-4 strains ) , p = 0 . 000000000066 ( for pDM1-mcr-5 strains ) , p = 0 . 00066 ( for pDM1-mcr-8 strains ) . Figure 5B one-way ANOVA with Bonferroni’s multiple comparison test; n = 3; 6 degrees of freedom; F value = 1878; p = 0 . 000000002 ( significance ) . Figure 8C Mantel-Cox test; n = 30; p =< 0 . 0001 ( significance ) ( P . aeruginosa versus P . aeruginosa dsbA1 ) , p > 0 . 9999 ( non-significance ) ( P . aeruginosa vs P . aeruginosa treated with ceftazidime ) , p =< 0 . 0001 ( significance ) ( P . aeruginosa treated with ceftazidime versus P . aeruginosa dsbA1 ) , p =< 0 . 0001 ( significance ) ( P . aeruginosa dsbA1 versus P . aeruginosa dsbA1 treated with ceftazidime ) . Figure 1—figure supplement 7A ( left graph ) : one-way ANOVA with Bonferroni’s multiple comparison test; n = 3; 6 degrees of freedom; F value = 39 . 22; p = 0 . 0007 ( significance ) , p = 0 . 99 ( non-significance ) . Figure 1—figure supplement 7B ( left graph ) : one-way ANOVA with Bonferroni’s multiple comparison test; n = 3; 6 degrees of freedom; F value = 61 . 84; p = 0 . 0002 ( significance ) , p = 0 . 99 ( non-significance ) . Figure 1—figure supplement 7B ( right graph ) : unpaired T-test with Welch’s correction , n = 3; 4 degrees of freedom; t-value = 0 . 1136 , p = 0 . 9150 ( non-significance ) . Figure 1—figure supplement 9A ( left graph ) : one-way ANOVA with Bonferroni’s multiple comparison test; n = 3; 6 degrees of freedom; F value = 261 . 4; p = 0 . 00000055 ( significance ) , p = 0 . 0639 ( non-significance ) . Figure 1—figure supplement 9B ( left graph ) : one-way ANOVA with Bonferroni’s multiple comparison test; n = 3; 6 degrees of freedom; F value = 77 . 49; p = 0 . 0001 ( significance ) , p = 0 . 9999 ( non-significance ) . Figure 1—figure supplement 9B ( right graph ) : unpaired T-test with Welch’s correction , n = 3; 4 degrees of freedom; t-value = 0 . 02647 , p = 0 . 9801 ( non-significance ) . The following bioinformatics analyses were performed in this study . Short scripts and pipelines were written in Perl ( version 5 . 18 . 2 ) and executed on macOS Sierra 10 . 12 . 5 . β-lactamase enzymes . All available protein sequences of β-lactamases were downloaded from http://www . bldb . eu ( Naas et al . , 2017 ) ( 5 August 2021 ) . Sequences were clustered using the ucluster software with a 90% identity threshold and the cluster_fast option ( USEARCH v . 7 . 0 ( 77 ) ) ; the centroid of each cluster was used as a cluster identifier for every sequence . All sequences were searched for the presence of cysteine residues using a Perl script . Proteins with two or more cysteines after the first 30 amino acids of their primary sequence were considered potential substrates of the DSB system for organisms where oxidative protein folding is carried out by DsbA and provided that translocation of the β-lactamase outside the cytoplasm is performed by the Sec system . The first 30 amino acids of each sequence were excluded to avoid considering cysteines that are part of the signal sequence mediating the translocation of these enzymes outside the cytoplasm . The results of the analysis can be found in Supplementary file 1 . MCR enzymes . E . coli MCR-1 ( AKF16168 . 1 ) was used as a query in a blastp 2 . 2 . 28+ ( Altschul et al . , 1990 ) search limited to Proteobacteria on the NCBI Reference Sequence ( RefSeq ) proteome database ( 21 April 2019 ) ( evalue <10e-5 ) . A total of 17 , 503 hit sequences were retrieved and clustered using the ucluster software with a 70% identity threshold and the cluster_fast option ( USEARCH v . 7 . 0 ( 77 ) ) . All centroid sequences were retrieved and clustered again with a 20% identity threshold and the cluster_fast option . Centroid sequences of all clusters comprising more than five sequences ( 809 sequences retrieved ) along with the sequences of the five MCR enzymes tested in this study were aligned using MUSCLE ( Edgar , 2004 ) . Sequences which were obviously divergent or truncated were manually eliminated and a phylogenetic tree was built from a final alignment comprising 781 sequences using FastTree 2 . 1 . 7 with the wag substitution matrix and default parameters ( Price et al . , 2010 ) . The assignment of each protein sequence to a specific group was done using hmmsearch ( HMMER v . 3 . 1b2 ) ( Finn et al . , 2015 ) with Hidden Markov Models built from confirmed sequences of MCR-like and EptA-like proteins . | Antibiotics , like penicillin , are the foundation of modern medicine , but bacteria are evolving to resist their effects . Some of the most harmful pathogens belong to a group called the 'Gram-negative bacteria' , which have an outer layer – called the cell envelope – that acts as a drug barrier . This envelope contains antibiotic resistance proteins that can deactivate or repel antibiotics or even pump them out of the cell once they get in . One way to tackle antibiotic resistance could be to stop these proteins from working . Proteins are long chains of building blocks called amino acids that fold into specific shapes . In order for a protein to perform its role correctly , it must fold in the right way . In bacteria , a protein called DsbA helps other proteins fold correctly by holding them in place and inserting links called disulfide bonds . It was unclear whether DsbA plays a role in the folding of antibiotic resistance proteins , but if it did , it might open up new ways to treat antibiotic resistant infections . To find out more , Furniss , Kaderabkova et al . collected the genes that code for several antibiotic resistance proteins and put them into Escherichia coli bacteria , which made the bacteria resistant to antibiotics . Furniss , Kaderabkova et al . then stopped the modified E . coli from making DsbA , which led to the antibiotic resistance proteins becoming unstable and breaking down because they could not fold correctly . Further experiments showed that blocking DsbA with a chemical inhibitor in other pathogenic species of Gram-negative bacteria made these bacteria more sensitive to antibiotics that they would normally resist . To demonstrate that using this approach could work to stop infections by these bacteria , Furniss , Kaderabkova et al . used Gram-negative bacteria that produced antibiotic resistance proteins but could not make DsbA to infect insect larvae . The larvae were then treated with antibiotics , which increased their survival rate , indicating that blocking DsbA may be a good approach to tackling antibiotic resistant bacteria . According to the World Health Organization , developing new treatments against Gram-negative bacteria is of critical importance , but the discovery of new drugs has ground to a halt . One way around this is to develop ways to make existing drugs work better . Making drugs that block DsbA could offer a way to treat resistant infections using existing antibiotics in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2022 | Breaking antimicrobial resistance by disrupting extracytoplasmic protein folding |
HIV-1 sequence diversity is affected by selection pressures arising from host genomic factors . Using paired human and viral data from 1071 individuals , we ran >3000 genome-wide scans , testing for associations between host DNA polymorphisms , HIV-1 sequence variation and plasma viral load ( VL ) , while considering human and viral population structure . We observed significant human SNP associations to a total of 48 HIV-1 amino acid variants ( p<2 . 4 × 10−12 ) . All associated SNPs mapped to the HLA class I region . Clinical relevance of host and pathogen variation was assessed using VL results . We identified two critical advantages to the use of viral variation for identifying host factors: ( 1 ) association signals are much stronger for HIV-1 sequence variants than VL , reflecting the ‘intermediate phenotype’ nature of viral variation; ( 2 ) association testing can be run without any clinical data . The proposed genome-to-genome approach highlights sites of genomic conflict and is a strategy generally applicable to studies of host–pathogen interaction .
Through multiple rounds of selection and escape , host and pathogen genomes are imprinted with signatures of co-evolution that are governed by Darwinian forces . On the host side , well-characterized anti-retroviral restriction factors , such as TRIM5α , APOBEC3G and BST2 , harbor strong signals of selection in primate genomes , clear examples of retroviral pressure ( Ortiz et al . , 2009 ) . On the virus side , obvious signs of selection are observable in the HIV-1 genome: escape mutations and reversions have been described in epitopes restricted by human leukocyte antigen ( HLA ) class I molecules and targeted by cytotoxic T lymphocyte ( CTL ) responses ( Goulder et al . , 2001; Kawashima et al . , 2009 ) . Sequence polymorphisms have also been reported recently in regions targeted by killer immunoglobulin-like receptors ( KIR ) , suggesting evasion from immune pressure by natural killer ( NK ) cells ( Alter et al . , 2011 ) . Evidence for the remodeling of retroviral genomes by host genetic pressure also comes from simian immunodeficiency virus ( SIV ) infection studies in rhesus macaques , where escape from restrictive TRIM5α alleles has been observed in the viral capsid upon cross-species transmission of SIVsm ( Kirmaier et al . , 2010 ) . In contrast , human alleles of TRIM5α do not result in escape mutations , likely because of adaptation of the pathogen to the host ( Rahm et al . , 2013 ) . Sequence adaptation is also a known feature of cross-species transmission . For example , a methionine in the matrix protein ( Gag-30 ) in SIVcpzPtt changed to arginine in lineages leading to HIV-1 and reverted to methionine when HIV-1 was passaged through chimpanzees ( Wain et al . , 2007 ) . To date , combined analyses of human and HIV-1 genetic data have addressed the association of HLA and KIR genes with variants in the retroviral genome ( Moore et al . , 2002; Brumme et al . , 2007; Bhattacharya et al . , 2007; Kawashima et al . , 2009; Alter et al . , 2011; Carlson et al . , 2012; Wright et al . , 2012 ) . Additionally , genome-wide association studies ( GWAS ) performed in the host have focused on various HIV-related clinical phenotypes ( Fellay et al . , 2007; Fellay et al . , 2009; Pereyra et al . , 2010 ) . In parallel , large amounts of HIV-1 sequence data have been generated for phylogenetic studies , which shed new light on viral transmission and evolution ( Kouyos et al . , 2010; Alizon et al . , 2010; Von Wyl et al . , 2011 ) , or allow clinically driven analyses of viral genes targeted by antiretroviral drugs ( resistance testing ) ( Von Wyl et al . , 2009 ) . Building on the unprecedented possibility to acquire and combine paired human and viral genomic information from the same infected individuals; we employ an innovative strategy for global genome-to-genome host–pathogen analysis . By simultaneously testing for associations between genome-wide human variation , HIV-1 sequence diversity , and plasma viral load ( VL ) , our approach allows the mapping of all sites of host–pathogen genomic interaction , the correction for both host and viral population stratification , and the assessment of the respective impact of human and HIV-1 variation on a clinical outcome ( Figure 1 ) . 10 . 7554/eLife . 01123 . 003Figure 1 . A triangle of association testing . The following association analyses were performed: [Study A] human SNPs vs plasma viral load ( 1 GWAS ) ; [Study B] human SNPs vs variable HIV-1 amino acids ( 3007 GWAS ) ; and [Study C] variable HIV-1 amino acids vs plasma viral load ( 1 proteome-wide association study ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01123 . 003
Full-length HIV-1 genome sequence and human genome-wide SNP data were obtained from seven studies or institutions on a total of 1071 antiretroviral naive patients of Western European ancestry , infected with HIV-1 subtype B . The homogeneity of the study population was confirmed by principal component analysis of the genotype matrix: together , the first five principal components explained 1% of total genotypic variation . After quality control of the human genotype data , imputation and filtering , ∼7 million SNPs were available for association testing . The full-length HIV-1 sequence is approximately 9 . 5 Kb long , corresponding to over 3000 encoded amino acids . Not all sequences were complete; on an average , viral residues were covered in 85% of the study population ( range: 75% in Tat to 95% in Gag ) . Due to its hypervariable nature , the portion of the HIV-1 envelope gene that encodes the gp120 protein was not sequenced in most study samples and was therefore excluded . Overall 1126 residues of the HIV-1 proteome were found to be variable in at least 10 samples , for a total of 3381 different viral amino acids that could be represented by 3007 distinct binary variables . We first performed a classical GWAS of host determinants of HIV-1 VL ( Figure 2A , Study A ) using data from 698 patients ( 65% of the study population ) for whom a VL phenotype could be reliably estimated . The top associations were observed in the HLA class I region on chromosome 6 and were highly consistent with results observed previously ( Fellay et al . , 2009; Pereyra et al . , 2010 ) . The strongest associated SNP , rs9267454 ( p = 1 . 5 × 10−8 ) , is in partial linkage disequilibrium ( LD ) with HLA-B*57:01 ( r2 = 0 . 47 , D′ = 0 . 92 ) , HLA-B*14:01 ( r2 = 0 . 12 , D′ = 1 . 0 ) , HLA-B*27:05 ( r2 = 0 . 01 , D′ = 0 . 99 ) , and the HLA-C -35 rs9264942 SNP ( r2 = 0 . 07 , D′ = 0 . 77 ) , and thus reflects these well-known associations with HIV-1 control . These results confirm the quality of the study population for the purpose of genome analysis of determinants of HIV-1-related outcomes . 10 . 7554/eLife . 01123 . 004Figure 2 . Results of the genome-wide association analyses . ( A ) Associations between human SNPs and HIV-1 plasma viral load . The dotted line shows the Bonferroni-corrected significance threshold ( p-value < 7 . 25 × 10−9 ) . ( B ) Associations between human SNPs and HIV-1 amino acid variants , with 3007 GWAS collapsed in a single Manhattan plot . The dotted line shows the Bonferroni-corrected significance threshold ( p-value < 2 . 4 × 10−12 ) . ( C ) Schematic representation of the HLA class I genes and of the SNPs associated with HIV-1 amino acid variants in the region . ( D ) Same association results as in panel B , projected on the HIV-1 proteome . Only the strongest association is shown for each amino acid . Significant associations are indicated by a blue dot . The gp120 part of the HIV-1 proteome was not tested . The colored bar below the plot area shows the positions of the optimally defined CD8+ T cell epitopes . An interactive version of this figure can be found at http://g2g . labtelenti . org ( which is also available to download from Zenodo , http://dx . doi . org/10 . 5281/zenodo . 7138 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01123 . 004 3007 genome-wide analyses of associations between human SNPs and HIV-1 amino acid variants were performed in the full sample of 1071 individuals ( Figure 2B , Study B ) using logistic regression corrected for viral phylogeny ( Carlson et al . , 2008; Carlson et al . , 2012 ) . Highly significant associations were observed between SNPs in the major histocompatibility complex ( MHC ) region and multiple amino acids throughout the HIV-1 proteome ( except in Vpu , Rev and the RNaseH subunit of RT ) ( Figure 2C ) , with Gag and Nef having a significantly higher density of associated variable sites than the rest of the proteome ( Gag: 6 . 8% vs 2 . 6% p=0 . 001; Nef: 11% vs 2 . 6% p = 1 . 2 × 10−5 , binomial tests ) . Using Bonferroni correction for multiple testing ( threshold p = 2 . 4 × 10−12 ) , significant human SNP associations were observed with 48 viral amino acids ( Figure 2 and Table 1 ) . None of these 48 amino acids mapped to known sites of major antiretroviral drug resistance mutations ( Hirsch et al . , 2008 ) . The strongest association found was between rs72845950 and Nef position 135 ( p = 2 . 7 × 10−66 ) . Associations were much stronger between human SNPs and HIV-1 amino acids than with VL . For example , the SNP rs2395029 , a proxy for HLA-B*57:01 ( r2 = 0 . 93 ) , has a p-value of 1 . 21 × 10−6 for association with VL , while it reaches a p-value of 4 × 10−59 for association with amino acid variation in Gag at position 242 ( a well known position of escape from HLA-B*57:01 ) . No significant signals were identified outside the MHC . A link to the complete set of association results can be found at http://g2g . labtelenti . org ( which is also available to download from Zenodo , http://dx . doi . org/10 . 5281/zenodo . 7139 ) . These results demonstrate the feasibility and improved power of performing association testing using viral genetic variation as outcome , independent of clinical phenotype . 10 . 7554/eLife . 01123 . 005Table 1 . Associations between HIV-1 amino acid variants and human polymorphismsDOI: http://dx . doi . org/10 . 7554/eLife . 01123 . 005HIV geneHIV positionSNPCTL epitope ( codons ) Tagging HLA ( D’/r2 ) SNP vs aa ( p ) SNP vs VL ( p ) aa vs VL ( p ) GAG12chr6:31285512–B*49:01 ( 1 . 00/1 . 00 ) 2 . 20E-136 . 70E-015 . 60E-01GAG26rs12524487–B*15:01 ( 1 . 00/0 . 82 ) 6 . 10E-192 . 10E-011 . 40E-01GAG28rs1655912RLRPGGKKK ( 20–28 ) A*03:01 ( 1 . 00/0 . 81 ) 2 . 70E-555 . 60E-012 . 00E-02GAG79chr6:31267544LYNTVATL ( 78-85 ) C*14:02 ( 1 . 00/0 . 96 ) 2 . 40E-123 . 50E-012 . 80E-01GAG147rs1055821–C*06:02 ( 0 . 95/0 . 71 ) 3 . 10E-173 . 30E-072 . 90E-05GAG242rs73392116TSTLQEQIGW ( 240–249 ) B*57:01 ( 1 . 00/0 . 98 ) 2 . 40E-621 . 90E-061 . 70E-05GAG248rs41557213TSTLQEQIGW ( 240-249 ) B*57:01 ( 1 . 00/0 . 97 ) 4 . 80E-152 . 00E-065 . 30E-03GAG264chr6:31376564KRWIILGLNK ( 263–272 ) B*27:05 ( 1 . 00/0 . 92 ) 2 . 30E-135 . 50E-023 . 50E-01GAG268rs2249935GEIYKRWIIL ( 259–268 ) B*08:01 ( 1 . 00/0 . 43 ) 2 . 20E-145 . 10E-011 . 90E-01GAG340rs11966319–B*15:01 ( 0 . 94/0 . 42 ) 6 . 70E-144 . 60E-017 . 70E-01–C*03:04 ( 0 . 99/0 . 59 ) GAG357rs2523612GPGHKARVL ( 355-363 ) B*07:02 ( 0 . 99/0 . 95 ) 2 . 70E-192 . 20E-011 . 20E-01–C*07:02 ( 0 . 99/0 . 84 ) GAG397rs61754472–A*31:01 ( 0 . 97/0 . 83 ) 8 . 80E-213 . 50E-018 . 30E-01GAG403rs28896571––8 . 90E-217 . 90E-018 . 60E-01GAG437rs34268928RQANFLGKI ( 429-437 ) B*13:02 ( 1 . 00/0 . 96 ) 8 . 70E-141 . 80E-026 . 80E-02GP41206rs17881210–B*15:01 ( 1 . 00/0 . 88 ) 6 . 10E-176 . 10E-013 . 00E-01GP41267rs9278477RLRDLLLIVTR ( 259–269 ) A*03:01 ( 1 . 00/0 . 01 ) 1 . 00E-127 . 80E-012 . 60E-01INT11rs2596477–B*44:02 ( 1 . 00/0 . 64 ) 5 . 10E-331 . 50E-011 . 80E-01INT32rs1050502–B*51:01 ( 0 . 97/0 . 92 ) 4 . 80E-187 . 20E-014 . 00E-01INT119rs9264954–C*05:01 ( 1 . 00/1 . 00 ) 1 . 30E-247 . 90E-011 . 10E-01INT122rs9264419–C*05:01 ( 1 . 00/0 . 95 ) 4 . 50E-228 . 30E-017 . 80E-01INT124chr6:31345421STTVKAACWW ( 123–132 ) B*57:01 ( 1 . 00/1 . 00 ) 3 . 00E-131 . 10E-069 . 70E-03NEF71rs2596488FPVTPQVPLR ( 68–77 ) B*07:02 ( 1 . 00/0 . 98 ) 3 . 80E-552 . 50E-018 . 10E-02–C*07:02 ( 0 . 95/0 . 83 ) NEF81rs9295987RPMTYKAAL ( 77–85 ) B*07:02 ( 1 . 00/0 . 01 ) 4 . 80E-362 . 50E-019 . 50E-02–C*04:01 ( 0 . 90/0 . 63 ) NEF83rs34768512–B*15:01 ( 1 . 00/0 . 47 ) 2 . 20E-172 . 80E-011 . 50E-02–C*03:04 ( 0 . 96/0 . 54 ) NEF85rs2395475RPMTYKAAL ( 77–85 ) B*07:02 ( 1 . 00/0 . 29 ) 1 . 90E-248 . 10E-011 . 30E-03–B*08:01 ( 1 . 00/0 . 22 ) –C*07:02 ( 0 . 97/0 . 30 ) NEF92rs16896166AVDLSHFLK ( 84–92 ) A*11:01 ( 1 . 00/0 . 99 ) 1 . 00E-275 . 30E-011 . 50E-01NEF94rs9265972FLKEKGGL ( 90–97 ) B*08:01 ( 1 . 00/0 . 97 ) 9 . 60E-359 . 80E-011 . 20E-01NEF102rs2524277–B*44:03 ( 0 . 98/0 . 96 ) 1 . 10E-134 . 40E-012 . 40E-01NEF105rs1049709–C*07:01 ( 1 . 00/0 . 98 ) 1 . 10E-359 . 00E-012 . 70E-01NEF116chr6:31402358HTQGYFPDW ( 116–124 ) B*57:01 ( 1 . 00/1 . 00 ) 3 . 00E-221 . 90E-063 . 30E-01NEF120chr6:31236168-C*14:02 ( 1 . 00/1 . 00 ) 4 . 40E-163 . 60E-011 . 20E-02NEF126chr6:31102273–B*51:01 ( 1 . 00/0 . 18 ) 1 . 10E-121 . 80E-014 . 90E-02NEF133chr6:31397689–B*35:01 ( 0 . 95/0 . 89 ) 2 . 80E-192 . 50E-013 . 40E-01NEF135rs72845950RYPLTFGW ( 134–141 ) A*24:02 ( 1 . 00/0 . 88 ) 2 . 70E-669 . 10E-025 . 50E-03PR35rs2523577EEMNLPGRW ( 34-42 ) B*44:02 ( 1 . 00/0 . 64 ) 1 . 70E-181 . 60E-015 . 70E-01PR93rs2263323–B*15:01 ( 0 . 98/0 . 92 ) 5 . 60E-304 . 70E-019 . 50E-01RNASE28rs2428481–B*08:01 ( 1 . 00/1 . 00 ) 1 . 80E-128 . 10E-016 . 20E-01RT135rs1050502TAFTIPSI ( 128–135 ) B*51:01 ( 0 . 97/0 . 92 ) 6 . 70E-457 . 20E-013 . 00E-01RT245chr6:31411714IVLPEKDSW ( 244–252 ) B*57:01 ( 1 . 00/0 . 98 ) 2 . 90E-211 . 20E-065 . 40E-02RT277rs3128902QIYPGIKVR ( 269–277 ) A*03:01 ( 1 . 00/0 . 99 ) 1 . 20E-658 . 20E-012 . 70E-01RT369rs17190134–B*13:02 ( 0 . 93/0 . 86 ) 3 . 50E-206 . 40E-021 . 40E-01RT395rs17194293–-1 . 50E-121 . 20E-017 . 70E-02TAT29rs9260615–A*32:01 ( 0 . 98/0 . 95 ) 4 . 40E-143 . 90E-011 . 40E-01TAT32rs16899214CCFHCQVC ( 30–37 ) C*12:03 ( 0 . 98/0 . 96 ) 6 . 40E-213 . 40E-014 . 90E-01VIF33chr6:31430060ISKKAKGWF ( 31–39 ) B*57:01 ( 1 . 00/0 . 98 ) 1 . 50E-139 . 90E-079 . 30E-03VIF51rs7767850–B*49:01 ( 1 . 00/1 . 00 ) 1 . 40E-125 . 20E-012 . 10E-01VIF74rs2395029–B*57:01 ( 1 . 00/0 . 98 ) 5 . 40E-139 . 70E-072 . 80E-01VPR32chr6:31362941VRHFPRIWL ( 31–39 ) B*27:05 ( 1 . 00/0 . 94 ) 3 . 10E-135 . 40E-026 . 50E-01Significant associations ( p < 2 . 4 × 10-12 ) were observed for 48 HIV-1 amino acid variants . The table shows the major amino acid variants present at each specific HIV-1 position , the strongest associated SNP and its linked HLA class I allele ( s ) , if applicable . The column ‘CTL Epitope ( codons ) ’ lists published , optimally described CTL epitopes ( available at http://www . hiv . lanl . gov/content/immunology/tables/optimal_ctl_summary . html and in [Carlson et al . , 2012] ) restricted by the tagged HLA class I allele ( s ) specified , and their positions within the protein . Where multiple overlapping epitopes restricted by the same HLA class I allele have been described , only one is shown . Associations where no relevant CTL epitope has been described are indicated with a dash . The last three columns give association p-values for comparisons between human SNPs and viral amino acids , human SNPs and plasma VL and viral amino acids and plasma VL , respectively . For tests involving viral amino acids accommodating more than 1 alternate allele , the smallest association p-value observed at that position is reported . We next assessed whether the top SNPs associated with HIV-1 amino acids represent indirect markers of HLA class I alleles known to exert evolutionary pressure on HIV-1 ( Table 1 ) . We tested pairwise correlations between significant MHC SNPs and HLA class I alleles . The analysis confirmed the existence of high LD between SNPs and HLA alleles targeting corresponding epitopes . For example , the strongest association ( p = 2 . 7 × 10−66 ) was observed between residue 135 in Nef , located in an optimally defined A*24:02 epitope , and rs72845950 , which strongly tags HLA-A*24:02 ( r2 = 0 . 89 ) . Furthermore , we observed that a substantial fraction of the identified viral amino acids ( 24/48 , 50% ) were located within an optimally defined CTL epitope restricted by one or more HLA alleles tagged by the associated SNP ( http://www . hiv . lanl . gov/content/immunology/tables/optimal_ctl_summary . html , supplemented with a recently updated list of epitopes [Carlson et al . , 2012] ) . However , in seven cases , the classical HLA allele implicated through LD with a tagging SNP did not match previously reported restriction patterns ( Table 1 ) . These data demonstrate that this approach can reconstruct a map of targets of HLA pressure across the viral proteome and identify sites outside classical epitopes that could represent additional escape variants or compensatory mutations . That a substantial proportion of associated viral amino acids lay outside known CTL epitopes also highlights this approach as a tool to guide novel epitope discovery ( Bhattacharya et al . , 2007; Almeida et al . , 2011 ) . Analysis of polymorphic amino acids within the HLA genes has been shown to improve power for detection of association with clinical outcome and has demonstrated the biological relevance of key residues in the HLA-B binding groove ( Pereyra et al . , 2010 ) . Therefore , we used the genome-to-genome framework to characterize the evolutionary pressure of HLA class I amino acids on the viral genome . The top associations in all classical class I genes mapped to discrete residues in the binding grooves of the HLA molecule: HLA-A position 62 ( p = 3 . 3 × 10−76 with HIV Nef 135 ) , HLA-B position 70 ( p = 7 . 1 × 10−57 with HIV Gag 242 ) , HLA-C position 99 ( p = 5 . 4 × 10−63 with HIV Nef 70 ) . These data indicate that all class I HLA genes can exert strong pressure on the viral proteome through a shared mechanism . The association results for HLA amino acids can also be found at http://g2g . labtelenti . org ( which is also available to download from Zenodo ) . To address whether there was an observable impact of viral mutation on a clinical outcome in this sample , we tested for associations between all HIV-1 amino acid variant and VL ( Study C ) . After correction for multiple testing ( p threshold = 1 . 6 × 10−5 based on 3125 viral amino acids ) , we did not observe any significant associations . We then focused on estimating the changes in VL associated with the 48 HIV-1 amino acid variants that were identified as significantly associated with host SNPs ( Figure 2D ) . The effects of amino acid variation at these sites on VL ranged from –0 . 16 to +0 . 07 log10 copies/ml ( Figure 3A ) . We also explored the combined fitness effect of multiple HIV-1 viral amino acid variants targeted by a single host marker using the well-understood model of HLA-B*57:01 . We evaluated the effect on VL of 23 viral residues that associated with host variant rs2395029 ( r2 = 0 . 93 with HLA-B*57:01 ) in the genome-to-genome analysis ( selection cutoff: p<0 . 001 ) . The marker rs2395029 was associated with a 0 . 38 log decrease in viral RNA copies/ml . The univariate effect on VL for each of the 23 viral amino acids targeted by this allele ranged from –0 . 16 to +0 . 12 ( Figure 3B ) . These results suggest that the genome-to-genome approach can be linked to clinical/laboratory phenotypes , allowing for detailed understanding of the distribution and relative contribution of sites of host–pathogen interaction to disease outcome . 10 . 7554/eLife . 01123 . 006Figure 3 . Association of HIV-1 amino acid variants with plasma viral load . ( A ) Changes in VL ( slope coefficients from the univariate regression model and standard error , log10 copies/ml ) for the 48 HIV-1 amino acids that are associated with host SNPs in the genome-to-genome analysis . ( B ) rs2395029 , a marker of HLA-B*57:01 is associated with a 0 . 38 log10 copies/ml lower VL ( black bar ) in comparison to the population mean . Gray bars represent changes in VL for amino acid variants associated with rs2395029 ( p<0 . 001 ) . In case of multiallelic positions , the change in VL is shown for all minor amino acids combined vs the major amino acid ( e . g . , GAG147 not I ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01123 . 006
HIV-1 host genomic studies performed so far have focused on clinically defined outcomes ( resistance to infection , clinical presentation , disease progression or death ) or on pathogen-related laboratory results ( such as CD4+ T cell counts and VL set point ) . While useful , these phenotypes have significant drawbacks . First , consistency of phenotypic determination can be hard to achieve , and such inconsistency can adversely affect power in large-scale genetic studies performed across multiple centers ( Evangelou et al . , 2011 ) . Second , a relatively long follow-up in the absence of antiretroviral treatment is necessary to obtain informative data about the natural history of infection . However , international guidelines now propose an early start of antiretroviral therapy in most HIV-1 infected individuals ( Thompson et al . , 2012 ) , making the collection of large numbers of long-term untreated patients not only unrealistic but also ethically questionable . To overcome these limitations , we developed a novel approach for host genetic studies of infectious diseases , built on the unprecedented possibility to obtain paired genome-wide information from hosts and pathogens . We combined human polymorphism and HIV-1 sequence diversity in the same analytical framework to search for sites of human-virus genomic conflict , effectively using variation in HIV-1 amino acids as an ‘intermediate phenotype’ for association studies . Intermediate phenotypes have recently been shown to be useful in uncovering association signals that are not detectable using more complex clinical endpoints: illustrative examples include metabolomic biomarkers in cardiovascular research ( Suhre et al . , 2011 ) , serum IgE concentration in the study of asthma ( Moffatt et al . , 2010 ) , or neuroimaging-based phenotypes in psychiatry genetics ( Rasetti and Weinberger , 2011 ) . Variation in the pathogen sequence is an as-yet-untapped intermediate phenotype , specific by nature to genomic research in infectious diseases . Importantly , it depends on sequencing the pathogen , which could prove in many cases easier and more standardized than obtaining detailed clinical phenotypes . Our approach allowed the mapping of host genetic pressure on the HIV-1 genome . The strongest association signals genome-wide were observed between human SNPs tagging HLA class I alleles and viral mutations in their corresponding CTL epitopes . Additional association signals were observed outside of optimally defined CTL epitopes , which could indicate novel epitopes , or represent secondary ( compensatory ) mutations . In a single experiment , these results recapitulate extensive epidemiological and immunogenetic research and represent a proof-of-concept that biologically meaningful association signals are identifiable using a hypothesis-free strategy . Indeed , host factors leading to viral adaptation can be uncovered by searching for associated imprints in the viral genome . Of note , the International HIV Controllers Study demonstrated the importance of specific amino acid positions in the HLA-B binding groove on a clinical outcome ( elite control ) ( Pereyra et al . , 2010 ) . We here extend this observation to the HLA-A and C grooves , emphasizing the similarity in mechanism of host pressure on the viral proteome that is not necessarily translated into observable clinical outcomes . We found a higher density of amino acid positions under selection in Gag and Nef compared with the rest of the HIV proteome . This is consistent with earlier findings that indicate the importance of Gag p24-specific CTL responses in slower progression to AIDS ( Borghans et al . , 2007; Brennan et al . , 2012 ) or controller status ( Dyer et al . , 2008 ) . Moreover , this further demonstrates that mapping host pressure on the pathogen proteome can reveal biologically relevant effects . Analyses were performed using samples from clinically well-characterized patients , most of them with repeated and reliable HIV-1 VL measurements in the absence of antiretroviral therapy . We were thus able to compare the results of GWAS assessing human genetic determinants of mean VL , a standard clinical correlate of HIV-1 control , and genome-to-genome GWAS on amino acid variants in the viral proteome . The use of HIV-1 variation as outcome resulted in a considerable gain in power to detect host factors: the lowest p-values were observed for SNPs mapping to the HLA class I region in both approaches , but associations were much stronger with HIV-1 amino acid variation than for HIV-1 VL ( 2 . 7 × 10−66 vs 1 × 10−08 ) , even when accounting for the increased number of multiple tests . In addition to identifying sites of interaction between the host and the pathogen , the study design allowed the scoring of biological consequences of such interaction , by assessing associations between host-driven escape at viral sites and an in vivo phenotype ( VL ) . For example , we decomposed the effect of rs2395029 ( a marker of HLA-B*57:01 ) on VL to the effects of the multiple viral amino acid variants that are associated with that SNP . While some HIV-1 amino acid changes individually associate with decrease in VL , the compound image that emerges is one of a multiplicity of modest effects distributed across many residues . Correlations between host-associated variants and VL are difficult to interpret , because they may reflect fitness costs or compensation , the existence of strong ( Iversen et al . , 2006; Carlson et al . , , 2012 ) or novel ( Almeida et al . , 2011 ) immune responses , or the indirect impact of specific HLA class I alleles . Nevertheless , the observation that the majority of host-associated HIV-1 mutations do not correlate with any detectable change in VL confirms HIV’s remarkable capacity to adapt and compensate to immune pressure , often without measurable fitness cost . A significant confounder in both human and viral genomic analyses is the existence of population stratification , where shared ancestry between infected individuals , stratification by ethnic groups , non-random distribution of HIV-1 subtypes , or clusters of viral transmission can all have an influence on the population frequencies of specific mutations , and thus create spurious associations if not carefully controlled for . Previous studies usually controlled for viral population substructure but were limited in the control of human population stratification ( Moore et al . , 2002; Bhattacharya et al . , 2007 ) . Our approach offers the opportunity to correct for both factors , thanks to the availability of extensive host and viral genomic information . The present sample size provided approximately 80% power to detect a common human variant ( minor allele frequency of 10% ) with an odds ratio of 4 . 2 in the genome-to-genome analysis ( Study B ) and a viral amino acid explaining approximately 4% of the variation in plasma viral load ( Study C ) at the respective significance thresholds ( Purcell et al . , 2003 ) . Consistent with most studies performed in HIV-1 host genetics over the past few years ( reviewed in Telenti and Johnson ( 2012 ) ) , we did not identify previously unknown host genetic loci involved in host-viral interaction and HIV-1 restriction . The proposed approach can only detect polymorphic host factors that leave an imprint on the virus , which may exclude mediators of immunopathogenesis or genes involved in the establishment of tolerance ( Medzhitov et al . , 2012 ) . An additional limitation is the incomplete nature of genomic information available both on the host side ( common genotypes from GWAS ) and on the viral side ( near full-length consensus sequence; gp120 was not included in the analyses ) . Finally , the multiple hypothesis burden of a genome-to-genome scan is extremely high . It is conceivable that larger studies , or studies that focus on a subgroup of predefined host genes , would have power to detect novel associations . A comprehensive , but computationally challenging description of host–pathogen genomic interactions would require human genome sequencing , coupled with deep sequencing of intra-host retroviral subpopulations . In summary , we used a genome-to-genome , hypothesis-free approach to identify associations between host polymorphisms and HIV-1 genomic variation . This strategy allows a global assessment of host–pathogen interactions at the genome level and reveals sites of genomic conflict . Comparable approaches are immediately applicable to explore other important infectious diseases , as long as polymorphic host factors exert sufficient selective pressure to trigger escape mutations in the pathogen . The observation that pathogen sequence variation , used as an intermediate phenotype , is more powerful than clinical and laboratory outcomes to identify some host factors allows smaller-scale studies and encourages analyses of less prevalent infectious diseases . Researchers involved in pathogen genome studies and host genetic studies should strongly consider the gathering of paired host–pathogen data .
Participating centers provided local Institutional Review Board approval for genetic analysis . Study participants provided informed consent for genetic testing , with the exception of a subset where a procedure approved by the relevant Research Ethics Board allowed the use of anonymized historical specimens in the absence of a specific informed consent . Study participants are treatment-naïve individuals followed in one of the following cohorts or institutions: the Swiss HIV Cohort Study ( SHCS , www . shcs . ch , [Schoeni-Affolter et al . , 2010] ) ; the HAART Observational Medical Evaluation and Research ( HOMER ) study in Vancouver , Canada ( www . cfenet . ubc . ca/our-work/initiatives/homer ) ; the AIDS Clinical Trials Group ( ACTG ) Network in the USA ( actgnetwork . org ) ; the International HIV Controllers Study in Boston , USA ( IHCS , www . hivcontrollers . org ) ; Western Australian HIV Cohort Study , Perth , Australia; the AIDS Research Institute IrsiCaixa in Badalona , Spain; and the Instituto de Salud Carlos III in Madrid , Spain . To reduce noise due to host and viral diversity , we only included individuals of recent Western European ancestry ( confirmed by clustering with HapMap CEU individuals in principal component analysis of the genotype data [Price et al . , 2006] ) , and infected with HIV-1 subtype B ( as assessed by the REGA Subtyping Tool [De Oliveira et al . , 2005] ) . Plasma VL determinations in the absence of antiretroviral therapy were available from patients from the SHCS and the HOMER study . The VL phenotype was defined as the average of the log10-transformed numbers of HIV-1 RNA copies per ml of plasma , excluding measurements obtained in the first 6 months after seroconversion and during advanced immunosuppression ( i . e . , with <100 CD4+ T cells per ml of blood ) . Consequently , 698 study participants were eligible for VL analysis . DNA samples were genotyped in the context of previous GWAS ( Fellay et al . , 2009; Pereyra et al . , 2010 ) or for the current study on various platforms , including the HumanHap550 , Human 660W-Quad , Human1M and HumanOmniExpress BeadChips ( Illumina Inc . , San Diego , CA , USA ) , as well as the Genome-Wide Human SNP Array 6 . 0 ( Affymetrix Inc . , Santa Clara , CA , USA ) ( Table 2 ) . Study participants were filtered on the basis of genotyping quality , a sex check , and cryptic relatedness . SNP quality control was performed separately for each dataset: SNPs were filtered on the basis of missingness ( excluded if called in <99% of participants ) , minor allele frequency ( excluded if <0 . 01 ) , and marked deviation from Hardy-Weinberg equilibrium ( excluded if p<0 . 00005 ) . Missing genotype imputation was performed with the Mach software per genotyping platform ( in separate batches for Illumina 1M , OmniExpress , 550K and Affymetrix data ) using 1000 Genomes Phase I CEU population data as reference haplotypes . Imputed markers were filtered on minor allele frequency ( excluded if <0 . 01 ) and imputation quality using Mach’s reported r-squared measure ( excluded if <0 . 3 ) . SNPs with a deviation in the allele frequencies between platforms were excluded . High-resolution HLA class I typing ( 4 digits; HLA-A , HLA-B , and HLA-C ) was obtained using sequence-based methods , or imputed from the SNP genotyping data as described elsewhere ( Jia et al . , 2013 ) . 10 . 7554/eLife . 01123 . 007Table 2 . Distribution of samples across genotyping platforms and cohortsDOI: http://dx . doi . org/10 . 7554/eLife . 01123 . 007NGenotyping platformCohort140Illumina 1MACTG6Illumina OmniExpress 12v1HCARLOS III518Affymetrix 6 . 0HOMER136Illumina OmniExpress12v1HHOMER47Illumina 650kIHCS6Illumina 660W-QuadIRSICAIXA2Illumina 1MSHCS79Illumina 550kSHCS122Illumina OmniExpress12v1HSHCS15Illumina 550kWAHCSACTG = AIDS Clinical Trials Group Network; CARLOS III = Instituto de Salud Carlos III; HOMER = HAART Observational Medical Evaluation and Research Study; IHCS = International HIV Controllers Study; IRSICAIXA = AIDS Research Institute IrsiCaixa; SHCS = Swiss HIV Cohort Study; WAHCS = Western Australian HIV Cohort Study . Near full-length retroviral sequence data were obtained by bulk sequencing of viral RNA present in pretreatment-stored plasma , and in 11 cases , of proviral DNA isolated from peripheral blood mononuclear cells , as previously described ( Sandonís et al . , 2009; John et al . , 2010 ) . We defined an amino acid residue as variable if at least 10 study samples presented an alternative allele . Per position , separate binary variables were generated for each alternate amino acid , indicating the presence or absence of that allele in a given sample . To globally assess the association between human genomic variation ( SNPs ) , HIV-1 proteomic variation ( amino acids ) and clinical outcome ( VL ) , we performed three series of analyses ( Figure 1 ) : [A] human SNPs vs VL; [B] human SNPs vs HIV-1 amino acids; and [C] HIV-1 amino acids vs VL . To test for association between human SNPs and HIV-1 amino acids , we used phylogenetically corrected logistic regression ( Carlson et al . , 2008; Carlson et al . , 2012 ) . For association testing between polymorphic amino acids in human HLA genes and HIV sequence variation , we used standard logistic regression ( for a binary HLA amino acid ) or a multivariate omnibus test ( when more than one alternate allele was present ) including sex , cohort , and the coordinates of the first two principal component axes as covariates . We used linear regression models in PLINK to test for association between human SNPs and VL , and between HIV-1 amino acids and VL ( Purcell et al . , 2007 ) , including sex , cohort , and the coordinates of the first two principal component axes as covariates ( Price et al . , 2006 ) . An additive genetic model was used for all analyses involving human SNPs . Significance was assessed using Bonferroni correction ( significance thresholds of 7 . 25 × 10−9 , 2 . 4 × 10−12 , and 1 . 6 × 10−5 for analyses A , B , and C , respectively , Figure 1 ) . | Developing treatments or vaccines for HIV is challenging because the genetic makeup of the virus is constantly changing in an effort to outwit the human immune system . Moreover , the immune system is highly variable as a result of the long-standing co-evolution of humans and microbes . Each individual will try to oppose the invading virus in a unique way , forcing the virus to acquire specific mutations that can be interpreted as the genetic signature of this one-against-one battle . To explore the influence of co-evolution on HIV , Bartha et al . took samples of both human and viral genomes from 1071 individuals infected with HIV , the AIDS virus , and used genotyping and sequencing technology to obtain a comprehensive description of the genetic variation in both . Computational techniques were then used to search for links between variants in the human DNA sequences and variants in the viral sequences . The most common type of genetic variation found in the human genome is a single nucleotide polymorphism , or SNP for short: a SNP is produced when a single nucleotide – an A , C , G or T – is replaced by a different nucleotide . Bartha et al . found that SNPs within the human DNA sequences in their study were linked to variations in 48 amino acids in HIV . Moreover , all these SNPs were found within a group of genes known as the HLA ( human leukocyte antigen ) system , which encodes for proteins that play a vital role in the immune response . This work identified the areas of the human genome that put pressure on the AIDS virus , and the regions of the virus that serve to escape human control . The approach developed by Bartha et al . allows the interactions between a microbe and a human host to be studied by looking at the genome of the microbe and the genome of the infected person . It also differentiates host-induced mutations that limit the capacity of the virus to do harm from those that are tolerated by the pathogen . A similar strategy could be used to study other infectious diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"genetics",
"and",
"genomics"
] | 2013 | A genome-to-genome analysis of associations between human genetic variation, HIV-1 sequence diversity, and viral control |
T cells engineered to express a tumor-specific αβ T cell receptor ( TCR ) mediate anti-tumor immunity . However , mispairing of the therapeutic αβ chains with endogenous αβ chains reduces therapeutic TCR surface expression and generates self-reactive TCRs . We report a general strategy to prevent TCR mispairing: swapping constant domains between the α and β chains of a therapeutic TCR . When paired , domain-swapped ( ds ) TCRs assemble with CD3 , express on the cell surface , and mediate antigen-specific T cell responses . By contrast , dsTCR chains mispaired with endogenous chains cannot properly assemble with CD3 or signal , preventing autoimmunity . We validate this approach in cell-based assays and in a mouse model of TCR gene transfer-induced graft-versus-host disease . We also validate a related approach whereby replacement of αβ TCR domains with corresponding γδ TCR domains yields a functional TCR that does not mispair . This work enables the design of safer TCR gene therapies for cancer immunotherapy .
T lymphocytes recognize cellular targets with extraordinary sensitivity and specificity . The T cell receptor ( TCR ) on the T cell surface binds to its cognate peptide antigen presented on major histocompatibility complex ( MHC ) molecules on the target cell surface ( Reinherz , 2015 ) . The TCR - a heterodimer that is uniquely rearranged in each T cell during its development - is the sole determinant of T cell specificity ( DembiĆ et al . , 1986 ) . Accordingly , transfer of the α and β genes encoding a tumor-reactive αβ TCR can impart anti-tumor reactivity to a cancer patient’s T cells ( Clay et al . , 1999 ) . Upon reinfusion into the patient , these engineered T cells mediate potent anti-tumor cytotoxicity ( Park et al . , 2011 ) . This approach – termed TCR gene therapy – has demonstrated clinical responses for several malignancies ( Johnson et al . , 2009; Morgan et al . , 2006; Davis , 2010; Parkhurst , 2011; Robbins et al . , 2015 , 2011 ) , which may be bolstered in the future through combination with complementary immunotherapies like dendritic cell-based vaccination and checkpoint blockade ( Abate-Daga et al . , 2013; Brahmer and Hammers , 2015; Mellman et al . , 2011; Sharma and Allison , 2015; Topalian et al . , 2015; Chodon , 2014 ) . Despite the promise of TCR gene therapy , there remain significant limitations to its safety and efficacy . Among these is that TCR-transduced αβ T cells express two α and two β chains , a non-physiological situation in which the introduced TCR α and β chains can mispair with the endogenous TCR β and α chains , respectively . TCR chain mispairing reduces the level of correctly-paired , tumor-reactive TCR heterodimers expressed on the T cell surface , possibly reducing therapeutic efficacy ( Jorritsma et al . , 2007 ) . More ominously , mispairing can generate self-reactive TCR heterodimers that do not undergo thymic negative selection . These mispaired TCRs engender autoreactivity in TCR-transduced human T cells in vitro ( van Loenen et al . , 2010 ) and mediate lethal graft-versus-host disease in mice administered TCR-transduced T cells following a protocol mimicking human clinical trials ( Bendle et al . , 2010; Bunse , 2014 ) . Although no serious adverse events have been attributed to TCR mispairing in engineered T cell trials ( Rosenberg , 2010 ) , autoreactive off-target and off-tumor engineered T cell responses have caused deaths ( Linette et al . , 2013; Morgan et al . , 2013 , 2010 ) . These underscore the need to safeguard against TCR mispairing-related autoreactivity , particularly as more potent immunotherapy regimes are employed . Efforts to prevent TCR mispairing can be broadly categorized as either engineering the transduced TCR ( adding interchain disulfide bonds , murinizing portions of the TCR , expressing TCR as a single chain ) ( Uckert and Schumacher , 2009 ) or reducing expression of the endogenous TCR ( shRNA knockdown ( Bunse , 2014; Okamoto et al . , 2009 ) or genomic knockout [Provasi et al . , 2012] ) . Although several engineering strategies improve pairing between the transduced chains , complete prevention of mispairing has not been achieved ( Thomas et al . , 2007 ) and murine TCRs are immunogenic ( Davis , 2010 ) . Endogenous TCR knockout prevents mispairing , but the extensive processing currently required to generate these cells is incompatible with clinical protocols . The ideal solution will prevent mispairing entirely , eliminating the risk of autoimmunity . Additionally , modifications made to the introduced TCR chains should avoid foreign sequences to minimize immunogenicity . Finally , these modifications must be restricted to the constant TCR domains , such that they can be applied without further optimization to any TCR of therapeutic interest . We describe a novel approach for preventing TCR mispairing that meets these criteria . We show that this approach is further improved by combining it with the complementary strategy of endogenous TCR knockdown .
Our approach to prevent TCR mispairing exploits the molecular requirements for TCR biogenesis and function . The TCR α and β chains each contain a membrane-distal variable immunoglobulin domain ( V ) , which imparts specificity , and several constant domains including a membrane-proximal constant immunoglobulin domain ( C ) , a connecting peptide ( cp ) , a transmembrane helix ( TM ) , and a short cytoplasmic tail ( cyto ) ( Figure 1A ) . To achieve functional form , the TCR αβ heterodimer must assemble with six additional chains ( CD3 dimers γε , δε , and ζ2 ) , which facilitate export of the TCR complex to the cell surface and mediate signal transduction upon antigen binding ( Call and Wucherpfennig , 2005 ) . If the TCR/CD3 complex is not assembled properly prior to export , it is degraded ( Bonifacino , 1989 ) . Assembly with CD3 requires contacts within the constant domains of both the TCR α and β chains ( Call et al . , 2002; Kuhns and Davis , 2007; Xu and Call , 2006 ) , most critically the basic residues within the transmembrane domains ( Call et al . , 2002 ) ( Figure 1B ) . We designed interchain domain-swapped ( ds ) TCRs in which select constant domains of the TCR α and β chains are exchanged in a reciprocal manner ( Figure 1C ) . Correctly paired α/β dsTCRs retain all domains necessary to assemble with CD3 and to enact tumor-targeted immunity . By contrast , mispaired heterodimers comprising one dsTCR chain and one wild-type ( wt ) TCR chain lack domains necessary to assemble with CD3 or to enact autoimmune responses ( Figure 1d ) . 10 . 7554/eLife . 19095 . 003Figure 1 . Schematic outlining the domain-swapped TCR strategy . ( A ) The TCR/CD3 complex comprises the antigen-specific variable ( V ) Ig domain and constant domains ( constant Ig , C; connecting peptide , cp; transmembrane helix , TM; and cytoplasmic tail , cyto ) , which assemble with CD3 chains . CD3 chains are required for export of the TCR/CD3 complex to the cell surface and for signaling . Parallel horizontal lines represent the cell membrane . ( B ) Schematic showing key interactions between basic residues in the TCR TM domain and acidic residues in the CD3 TM domains . ( C ) Domain-swapped TCRs retain all domains of the wild-type TCR with altered covalent connectivity . ( D ) Mispairing between therapeutic and endogenous TCR chains can result in autoreactivity ( middle ) . Mispairing between domain-swapped therapeutic and endogenous TCR chains will result in heterodimers that lack constant domains needed to assemble with CD3 , preventing surface expression and function of potentially autoreactive TCRs ( right ) . Domain-swapped TCRs are thus expected to be functional but incapable of mediating mispairing-related autoreactivity . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 003 We designed three dsTCR architectures in which α and β constant domains were swapped at the V-C junctions ( dsTCRV ) , at the C-cp junctions ( dsTCRC ) , or at the cp-TM junctions ( dsTCRcp ) ( Figure 2A ) . Initially , we generated these dsTCR variants for the HLA-A2-restricted , MART1-specific F5 TCR ( Johnson et al . , 2006 ) ( Figure 2—figure supplement 1 ) . We transfected CD3+ 293T cells with each F5 TCR variant and evaluated A2/MART1 tetramer binding by flow cytometry . The dsTCRV construct did not bind tetramer ( Figure 2B ) , and variation of the precise position at which the V domains were exchanged did not rectify this ( Figure 2—figure supplement 2A ) . However , both F5 dsTCRC and dsTCRcp bound tetramer , indicating these architectures express on the cell surface and retain specificity for their cognate ligand ( Figure 2B ) . Cells transfected with F5 dsTCRC bound tetramer with comparable avidity to wtTCR , whereas tetramer binding for dsTCRcp was lower . We varied the position at which the domains were exchanged in the dsTCRcp architecture to identify the optimal design for this construct ( α1β5 , Figure 2—figure supplement 2B ) . The dsTCRC and optimized dsTCRcp constructs were selected for further investigation . 10 . 7554/eLife . 19095 . 004Figure 2 . Domain-swapped TCRs assemble with CD3 and retain antigenic specificity . ( A ) Schematic of dsTCR architectures . Domains are swapped following the V domains ( dsTCRV ) , C domains ( dsTCRC ) , or connecting peptides ( dsTCRcp ) . ( B ) Flow cytometric analysis of peptide-MHC multimer binding by CD3+ 293T transfected with F5 wtTCR or dsTCRs . ( C ) Flow cytometric analysis comparing peptide-MHC multimer binding by CD3+ 293T transfected with F5 wtTCR , dsTCRC , or simulated mispaired constructs . Schematics of constructs are above corresponding data . ( D–E ) Flow cytometric analysis comparing anti-Vβ antibody binding by CD3 + 293T transfected with wtTCR , dsTCRC , or simulated mispaired constructs for ( D ) two additional human TCRs and ( E ) mouse OTI TCR . Means ± SD for two replicate transfections are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 00410 . 7554/eLife . 19095 . 005Figure 2—figure supplement 1 . Amino acid sequences of initial domain-swapped TCR construct chimeric junctions . Sequence derived from TCRα is in red and sequence derived from TCRβ is in blue . The junction sequence shown for dsTCRcp’ was only used for experiment in Figure 2 . Final optimized junction for dsTCRcp is shown in Supplemental Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 00510 . 7554/eLife . 19095 . 006Figure 2—figure supplement 2 . Optimization of domain-swapped junctions for dsTCRV and dsTCRcp . ( A ) Amino acid sequences of F5 TCR Vα-Cα and Vβ-Cβ domain boundaries are shown . Positions used in iterating the position at which domains were swapped are indicated with arrows , and an illustrative example sequence for one dsTCRV iteration ( α3β2 ) is provided . To test for optimal surface expression , 293T were cotransfected with CD3 and dsTCRV constructs and percent of transfected cells that bound tetramer was measured by flow cytometry . Means ± SD for triplicate assays from a single experiment shown . ( B ) Amino acid sequences of F5 TCR cpα-TMα and cpβ-TMβ domain boundaries are shown . Positions used in iterating the position at which domains were swapped are indicated with arrows . The optimized dsTCRcp sequence ( α1β5 ) is provided . To identify construct with optimal function , IL-2 secretion was measured by ELISA following 48 hr coincubation of dsTCRcp–transduced Jurkat T cells with antigen-expressing K562 target cells . Means ± SD for 3 technical replicates from a single experiment shown . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 006 We posit that dsTCRs are safer than wtTCR because mispaired TCRs comprising one ds and one wtTCR chain should have impaired capacity to assemble with CD3 and express on the cell surface . To determine whether dsTCR chains properly assemble CD3 when mispaired with a wtTCR chain , we prepared hybrid constructs in which only the α or β chain of F5 TCR was domain-swapped at the C-cp junction , and tested for expression on the surface of CD3+ 293T ( Figure 2C ) . As expected , compared to fully wtTCR or dsTCRC , surface expression of these simulated mispaired constructs was significantly reduced , indicating impaired assembly of CD3 . To have broad therapeutic potential , our approach must be applicable to any human TCR of clinical interest without further optimization . Ideally , this approach should also be applicable to murine TCRs , enabling preclinical studies of safety and efficacy . We constructed dsTCRC and hybrid wt/ds variants of two additional clinically relevant human TCRs – HLA-A2/MART1-specific M1 TCR ( Chhabra et al . , 2008 ) and HLA-A2/NY-ESO-1-specific 1 G4 TCR ( Chen et al . , 2000 ) –as well as the H-2Kb/ovalbumin-specific OTI murine TCR . For both additional human TCRs as well as the murine TCR , dsTCR derivatives express on the surface of CD3+ 293T and mitigate mispairing with wtTCR chains ( Figure 2D , E ) , suggesting this approach can be applied generally to human and mouse TCRs . Altering the spatial relationship between constant TCR domains could potentially alter the orientation of CD3 chains in the TCR/CD3 complex , impairing its function . To compare the orientation of CD3 chains around wt- and dsTCRs , we used a dimerization reporter assay previously employed to determine the orientation of CD3 chains around the αβ wtTCR ( Kuhns et al . , 2010 ) . The murine pro-B cell line , BaF3 , will grow in the absence of IL-3 only when provided Jak-Stat signaling via dimerized signaling domains of the erythropoietin receptor ( EpoR ) . A fusion of CD3ε to EpoR ( εE ) was previously shown to drive proliferation of BaF3 when incorporated in the TCR/CD3 complex , indicating that the CD3 δε and γε dimers cluster on one face of the wtTCR , juxtaposing the CD3ε chains ( Kuhns et al . , 2010 ) ( Figure 3A ) . We transduced BaF3 cells with all CD3 chains , including εE , and either human F5 wtTCR , dsTCRC , or dsTCRcp and then measured TCR/CD3 complex on the cell surface . All three TCRs formed complexes with CD3/εE and bound cognate peptide-MHC multimer ( Figure 3—figure supplement 1 ) . Following removal of IL-3 from the cell media , εE drove proliferation of BaF3 cells when complexed with wtTCR , dsTCRC , or dsTCRcp , indicating that dsTCRs assemble with CD3 chains in the same orientation as wtTCR ( Figure 3B ) . Results were similar for the murine OTI TCR ( Figure 3C and Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 19095 . 007Figure 3 . Domain-swapped TCRs are functional in lymphocytes . ( A ) Schematic of the BaF3 proliferation assay for determining CD3 orientation within the TCR/CD3 complex . CD3ε-EpoR fusion proteins promote IL-3-independent proliferation of BaF3 only if assembly in the TCR/CD3 complex orients them in juxtaposition to one another . ( B-C ) CD3ε-EpoR-driven proliferation of BaF3 cells expressing wild-type and domain-swapped ( B ) human F5 TCR and ( C ) mouse OTI TCR . Cell number was measured 72 hr after IL-3 deprivation . Means ± SD for 3 independent experiments are shown . ( D-E ) Representative flow cytometry histograms comparing peptide-MHC multimer binding by Jurkat T cells transduced with ( D ) F5 wtTCR or dsTCR constructs or ( E ) simulated mispaired constructs . Multimer binding is reliant on cell surface expression following assembly with endogenous CD3 chains . ( F ) Flow cytometric measurement of TCR and Myc271 expressed on the surface of transduced Jurkats . Myc271 is an independent transduction marker expressed from the same vector as TCR . ( G ) ELISA measuring secretion of IL-2 from TCR-transduced Jurkats following 48 hr coincubation with cognate or control antigen-expressing K562 target cells . ( H ) Flow cytometric measurement of dextramer-binding TCR and LNGFR expressed on the surface of transduced primary human T cells . LNGFR is an independent transduction marker expressed from the same vector as TCR . ( I ) ELISA measurement of secreted IFN-γ and ( J ) flow cytometric measurement of CD25 expressed on the T cell surface from TCR-transduced primary human T cells following 48 hr coincubation with cognate or control antigen-expressing K562 target cells . For panels ( f-j ) , means ± SD for 3 technical replicates are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 00710 . 7554/eLife . 19095 . 008Figure 3—figure supplement 1 . Domain-swapped TCRs are expressed on the surface of CD3+ BaF3 cells . Representative flow cytometry plots showing surface expression of wt and dsTCR constructs in transduced CD3+ BaF3 cells . Events from untransduced ( black ) and transduced ( red ) cells are overlaid . Three independent experiments were performed . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 008 In both the 293T and BaF3 assays , CD3 chains are expressed at high , non-physiological levels . We have observed that some TCR is exported to the surface of transfected 293Ts even when CD3 assembly is incomplete ( i . e . 26% , 0% , 52% , and 100% residual export respectively when omitting CD3δ , CD3ε , CD3γ , or CD3ζ from transfection ) . We suspected that this accounted for the low level of apparent pairing between dsTCR and wtTCR chains observed on 293T cells ( Figure 2C–E ) . To determine whether dsTCRs form a complex with endogenous CD3 in T cells , we transduced the Jurkat human T cell line with wt- and dsTCRs and used flow cytometry to assess surface expression . Both dsTCRC and dsTCRcp assemble endogenous CD3 , express on the cell surface , and bind cognate A2/MART1 multimers to an extent similar to the wtTCR ( Figure 3D ) . By contrast , hybrid constructs comprising one dsTCR chain and one wtTCR chain were not detectable on the surface of transduced Jurkat cells , indicating that , as expected , these mispaired constructs failed to form a mature complex ( Figure 3E ) . Virtually all transduced ( Myc271+ ) cells were positive for A2/MART1 staining ( Figure 3F ) , indicating the dsTCRs competed favorably for CD3 recruitment in the presence of the endogenous Jurkat TCR . To determine whether dsTCR/CD3 complexes are functional , dsTCR-transduced Jurkat T cells were coincubated with K562 target cells expressing either cognate A2/MART1 or control A2/NYESO1 single-chain trimer ( Hansen et al . , 2009 ) . Consistent with multimer staining results , dsTCRs mediated antigen-specific release of IL-2 , whereas hybrid wt/dsTCRs simulating mispairing did not ( Figure 3G ) . Unlike the Jurkat T cell line , primary T cells express a diverse repertoire of endogenous T cell receptors , which may mispair or compete for surface expression with the transduced TCR to differing extents across cells . We therefore transduced primary T cells with wt- and dsTCRs to assess their function in a more physiologically relevant context . As expected , wtTCR was detected in only a subset ( 74% ) of transduced ( LNGFR+ ) cells , indicating ~25% of T cells express an endogenous TCR that out-competes F5 TCR for CD3 recruitment and cell surface expression ( Figure 3H ) . Dextramer staining was lower for dsTCRs than wtTCRs in primary T cells ( 25% and 11% of LNGFR+ cells , for dsTCRC and dsTCRcp , respectively; MFI = 21% and 13% that of wtTCR ) . This indicates a larger subset of endogenous TCRs outcompete the dsTCRs for CD3 recruitment . Nonetheless , T cells transduced with either wtTCR or dsTCRs were functional , secreting IFN-γ and upregulating cell surface CD25 in an antigen-specific manner when coincubated with target cells ( Figure 3I , J ) . As with Jurkat cells , simulated mispaired constructs elicited no such activation ( Figure 3I , J ) . Thus , dsTCRs assemble with endogenous CD3 to mediate antigen-specific immunity in primary lymphocytes and are rendered nonfunctional when mispaired with wtTCR chains . Transduced α and β chains can each mispair with endogenous chains , and autoreactivity is an unpredictable result of the particular transduced and endogenous chains involved ( van Loenen et al . , 2010 ) . To investigate whether dsTCR chains can mispair with any of the diverse array of endogenous TCR chains expressed across primary T cells , we devised an assay for measuring mispairing without regard to antigenic specificity . We tagged the N-termini of the α and β chains of F5 TCR with cMyc and V5 epitope tags , respectively , and confirmed that the addition of these tags did not alter TCR surface expression on Jurkat T cells ( Figure 4—figure supplement 1 ) . We then delivered each chain individually ( i . e . Myc-α only or V5-β only ) , such that the export of the tagged TCR chain is dependent on mispairing with endogenous TCR chains . The wtα and wtβ chains of F5 TCR mispaired with endogenous TCR chains in 88% and 58% of transduced T cells , respectively . By contrast , neither dsα nor dsβ chains exhibited detectable mispairing with endogenous TCR chains ( Figure 4A ) . Similar results were obtained with the murine OTI TCR: wtα and wtβ chains were mispaired in 63% and 84% of transduced murine T cells , whereas mispairing was not observed with dsα or dsβ chains ( Figure 4B ) . Thus , even when mispairing propensity is maximized by transduction of a single dsTCR chain in the absence of stoichiometric expression of its correct partner , dsTCR chains exhibit no detectable mispairing with endogenous TCR chains . 10 . 7554/eLife . 19095 . 009Figure 4 . Domain-swapped TCRs do not mispair with endogenous TCRs and do not cause graft-vs-host disease . ( A-B ) Flow cytometric measurement of mispairing of ( A ) human F5 TCR chains in human peripheral blood T cells and ( B ) murine OTI TCR chains in murine splenic T cells . Cells were transduced with only TCRα ( left panel ) or only TCRβ ( right panel ) such that expression of the introduced chain on the cell surface required mispairing with endogenous TCR chains . Representative histograms are displayed above bar graphs quantifying means ± SD for 3 technical replicates . ( C-D ) Mouse model of TCR gene transfer-induced graft-vs-host disease ( TI-GvHD ) . Mice were irradiated on day -1 , administered TCR-transduced T cells on day 0 , and injected twice daily with IL-2 from day 10–12 . Weight was monitored daily and mice were euthanized at 85% initial body weight . Results are aggregated from 3 independent experiments . Respective group sizes , n = 9 , 15 , 15 , and 25 . *p<0 . 05 , ****p<0 . 0001 , ns , not significant . ( C ) Percent initial body weight of surviving mice at 16 days after T cell administration . ( D ) Kaplan-Meier survival curve . ( E ) Tumor size in mice administered TCR-transduced T cells and then injected with ovalbumin-expressing E . G7 thymoma tumor cells . Respective group sizes , n = 9 , 6 , 10 , and 10 . Mean ± SD for each group is overlaid on traces for individual animals . ( F ) Tumor size in mice ( n = 10 ) administered OTI dsTCR-transduced T cells and then injected on opposing flanks with EL4 ( control ) or E . G7 ( target ) thymoma tumor cells . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 00910 . 7554/eLife . 19095 . 010Figure 4—figure supplement 1 . N-terminal epitope tags do not impair surface expression of TCR chains . ( A ) Schematic of untagged and epitope-tagged TCRs delivered to Jurkat T cells via retroviral transduction . ( B-E ) Flow cytometry histograms measuring ( B ) background staining with anti-mouse IgG ( secondary antibody ) only , ( C ) cMyc-TCRα , ( D ) V5-TCRβ , or ( E ) TCR expressed on the surface of untransduced and transduced Jurkat cells . JOVI . 1 antibody detects a TCRβ extracellular constant domain epitope on the surface of ~10% of untransduced Jurkat T cells . This is increased to the same extent by transduction with untagged or tagged TCR . Histograms are representative of 3 similar experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 01010 . 7554/eLife . 19095 . 011Figure 4—figure supplement 2 . Wild-type TCR- and dsTCR-transduced T cells secrete a similar complement of cytokines upon stimulation . Mouse splenocytes were activated with 2 μg/mL concanavalin A and 1 ng/mL IL-7 for 48 hr , transduced with OTI wtTCR , OTI dsTCR , or no TCR overnight , washed , and then incubated for 16 hr in either media alone or media containing H-2Kb/Ova tetramer and anti-CD28 antibody . Supernatants collected after 48 hr ConA/IL-7 activation and after 16 hr tetramer stimulation were tested for ensemble cytokine analysis based on a DNA-encoded Antibody Library ( DEAL ) assay52 . Results are from a single experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 01110 . 7554/eLife . 19095 . 012Figure 4—figure supplement 3 . Domain-swapped TCR-transduced T cells are functional in vivo . ( A-C ) Functional analysis of splenic T cells from mice 45 days after administration of OTI ( Vα2Vβ5 ) TCR-transduced T cells . Splenocytes were incubated for 72 hr with or without Ova257-264 peptide and analyzed by flow cytometry and ELISA . Means ± SD for 3 technical replicates are shown . ( A ) Flow cytometric measurement of percent of Vα2+Vβ5+ ( left panel ) and H-2Kb/Ova257-264 tetramer+ ( right panel ) splenic T cells . ( B ) ELISA measurement of secreted IFN-γ . ( C ) Flow cytometric measurement of CD25 ( left panel ) , CD44 ( middle panel ) , and CD62L ( right panel ) expressed on the T cell surface . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 01210 . 7554/eLife . 19095 . 013Figure 4—figure supplement 4 . Wild-type and domain-swapped TCR-transduced T cells similarly protect against tumor growthStacked daily tumor cross-sectional area measurements in mice administered a variable number ( 104 , 105 , or 3 . 3 x 105 ) of ovalbumin-specific TCR-transduced T cells ( unsorted ) and then injected with 5 x 106 ovalbumin-expressing E . G7 thymoma tumor cells . Group size is n = 8 for control T cells ( no transduced TCR ) and n = 4 for each input level for each TCR-transduced T cell group ( wild-type , domain-swapped , or codon-optimized wild-type with endogenous TCR knockdown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 013 To determine whether dsTCRs prevent deleterious autoimmunity resulting from TCR mispairing , we used a previously reported mouse model for TCR gene transfer-induced graft-versus-host disease ( TI-GvHD ) ( Bendle et al . , 2010 ) . Mice that develop TI-GvHD due to TCR chain mispairing exhibit destruction of the hematopoietic compartment and rapid weight loss , generally necessitating sacrifice within 1–2 days of onset of cachexia . We transduced T cells with wt- and dsTCR derivatives of OTI TCR , as this TCR resulted in the highest incidence of TI-GvHD of those TCRs tested in the initial report of the model . Based on our finding that dsTCR chains do not mispair with endogenous TCR chains , we expected to recapitulate TI-GvHD in wtTCR-transduced T cell recipients , but not in dsTCR-transduced T cell recipients . Indeed , mice that received wtTCR ( IRES ) -transduced T cells - in which the transduced wtTCR chains were linked by an internal ribosome entry site ( IRES ) sequence – exhibited significant cachexia following IL-2 administration ( Figure 4C ) . Although some of these mice regained weight and survived , 47% ( 7/15 ) developed TI-GvHD to an extent that necessitated euthanasia ( Figure 4D ) . Consistent with previous results ( Bendle et al . , 2010 ) , the incidence of TI-GvHD could be reduced but not eliminated by replacing the IRES with a viral 2A sequence to ensure stoichiometric expression of the wtTCR chains . In this wtTCR ( 2A ) group , less weight loss was observed and only 1/15 mice developed fatal TI-GvHD . By contrast , cachexia was not observed for mice that received dsTCRC ( 2A ) -transduced T cells , and none ( 0/25 ) of these mice developed TI-GvHD . Thus , dsTCRs prevent mispairing and thereby prevent mispairing-related autoimmune complications in a model of TCR gene therapy . The lack of TI-GvHD in mice receiving dsTCR-transduced T cells was not due to lack of function in vivo . Mouse T cells transduced with OTI wtTCR or dsTCRC produced a similar complement of cytokines upon stimulation with cognate peptide-MHC multimer ( Figure 4—figure supplement 2 ) . For at least 6 weeks following T cell injection , Vα2+/Vβ5+ T cells were present in the spleen and retained specificity for H-2Kb/Ova tetramer ( Figure 4—figure supplement 3A ) . These splenic T cells exhibited antigen-specific IFN-γ release and modulated the surface expression of T cell activation markers upon stimulation with ovalbumin peptide ex vivo ( Figure 4—figure supplement 3B , C ) . The ovalbumin-expressing murine thymoma line , E . G7 , grew in all control mice injected with LNGFR-transduced T cells , but did not grow detectably in any mice injected with either wtTCR- or dsTCR-transduced T cells ( Figure 4E ) . Tumor cell killing was antigen-specific , because EL4 – the ovalbumin-negative parental line from which E . G7 was derived – grew readily in mice injected with dsTCR-transduced T cells ( Figure 4F ) . The relative efficacies of wtTCR- and dsTCR-transduced T cells were compared by titrating down the number of T cells injected prior to tumor cell injection . Both wtTCR- and dsTCR-transduced T cells slowed tumor growth when mice were injected with 1 T cell per 500 tumor cells , and prevented tumor growth entirely at higher tested ratios ( Figure 4—figure supplement 4 ) . Thus , dsTCRs mediate antigen-specific immunity but not deleterious autoimmunity in vivo . Swapping TCR domains prevents mispairing with endogenous TCR chains , but also reduces the amount of transduced TCR on the cell surface in the presence of competing endogenous TCR chains . Accordingly , employing dsTCRs in adjunct with approaches to reduce endogenous TCR expression may increase dsTCR surface expression while also improving the safety of these approaches in a complementary manner . We explored the compatibility of dsTCRs with two such approaches: knockdown of endogenous TCR in mature T cells , and hematopoietic stem cell TCR gene therapy . A recent report demonstrated that RNAi-mediated silencing of endogenous TCR genes improves surface expression and reduces mispairing of a co-delivered codon-optimized TCR ( Bunse , 2014 ) . However , TI-GvHD was not eliminated entirely , indicating incomplete knockdown of endogenous TCR . We hypothesized that combining dsTCRs with endogenous TCR knockdown would improve cell surface expression of dsTCRs while also eliminating TI-GvHD entirely . To investigate this combination approach , we transduced mouse T cells with vectors that co-delivered 1 ) shRNA targeting endogenous TCR α and β genes and 2 ) codon-optimized wt- and dsTCR versions of the lymphocytic choriomeningitis virus ( LCMV ) -specific P14 murine TCR . As previously shown ( Bunse , 2014 ) , the β chain ( Vβ8 ) of P14 wtTCR was disproportionally detected on the cell surface relative to the α chain ( Vα2 ) - indicative of mispairing - even when translational stoichiometry was ensured through the use of a viral 2A sequence ( Figure 5A ) . Co-delivery of shRNA resulted in knockdown of endogenous TCR surface expression to ~40% of normal levels ( Figure 5—figure supplement 1 ) , improving the balance of P14 wtTCR chains expressed on the surface and increasing tetramer staining ( Figure 5A , B ) . Reflecting the lack of mispairing exhibited by dsTCRs , the domain-swapped α and β chains were detected in equal amounts on the cell surface with or without endogenous TCR knockdown ( Figure 5A ) . However , knockdown greatly improved surface expression and tetramer binding of the P14 dsTCR by reducing competition for assembly with CD3 chains ( Figure 5B ) . 10 . 7554/eLife . 19095 . 014Figure 5 . Complementarity of TCR domain-swapping and endogenous TCR knockdown . ( A-B ) Representative flow cytometry plots showing percent of CD8+ T cells that are ( A ) Vα2+Vβ8+ and ( B ) gp33 tetramer+ following transduction with a vector delivering codon-optimized gp33-specific P14 ( Vα2Vβ8 ) TCR ± shRNA targeted to endogenous TCR Cα/Cβ . ( C-E ) Mouse model of TCR gene transfer-induced graft-vs-host disease ( TI-GvHD ) , as in Figure 4 , except mice were administered T cells transduced with codon-optimized P14 TCR ± shRNA targeted to endogenous TCR Cα/Cβ . Respective group sizes , n = 8 , 8 , 13 , 12 , and 11 . ***p<0 . 001 , ****p<0 . 0001 ns , not significant . ( C ) Mean percent initial body weight ± SD for each group over time . ( D ) Percent initial body weight of surviving mice at 22 days after T cell administration . ( E ) Kaplan-Meier survival curve . ( F-G ) Flow cytometric measurement of mispairing of ( F ) P14 TCRα and ( G ) P14 TCRβ chains in murine splenic T cells transduced with either chain ± shRNA targeted to endogenous TCR Cα/Cβ . Histograms are representative of triplicate results from 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 01410 . 7554/eLife . 19095 . 015Figure 5—figure supplement 1 . shRNA-mediated knockdown reduced endogenous TCR expression on surface of mouse T cells . Flow cytometric measurement of cell surface TCR Cβ and CD3ε on mouse splenic T cells transduced with LNGFR ± shRNA targeting endogenous TCR Cα/Cβ . Mean ± SD from triplicate measurements of MFI for TCR Cβ , MFI for CD3ε , and % double positive for TCR Cβ/CD3ε among CD45 . 1+CD8a+LNGFR+ cells , calculated as 100%* ( value for LNGFR+shRNA/value for LNGFR ) . Data are representative of 2 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 015 Mice injected with P14 wtTCR-transduced T cells exhibited weight loss and developed TI-GvHD , which was prevented by co-delivery of shRNA ( Figure 5C–E ) . By contrast , no TI-GvHD was observed in mice receiving P14 dsTCR-transduced T cells , whether or not shRNAs were co-delivered . As with OTI dsTCR ( Figure 4B ) , P14 dsTCR did not induce TI-GvHD because the P14 domain-swapped α and β chains did not mispair with endogenous chains ( Figure 5F , G ) . By contrast , individually delivered P14 wild-type α and β chains mispaired with endogenous chains even when both endogenous chains were knocked down . In fact , knockdown of both endogenous chains increased mispairing of P14 wtα , suggesting the reduction of competing endogenous α enabled more mispairing of the weak P14 wtα with the remaining endogenous β chain . Prevention of TCR mispairing via endogenous TCR knockdown therefore relies on both the reduction of endogenous TCR and the equal ( over ) expression of the transduced TCR chains . Taken together , these results suggest that a combination approach of endogenous TCR knockdown and dsTCR transduction may optimize safety and efficacy over either approach alone . Hematopoietic stem cells ( HSCs ) transduced with pre-rearranged TCR genes differentiate into antigen-specific T cells in the periphery ( Vatakis et al . , 2011; Yang and Baltimore , 2005 ) . This is an attractive therapeutic approach as it provides a continuous supply of antigen-specific naïve T cells with optimal replicative potential ( Hinrichs et al . , 2011 ) . Additionally , expression of a pre-rearranged TCR in HSCs suppresses rearrangement of the endogenous TCR in these cells via allelic exclusion ( Hinrichs et al . , 2013; Vatakis , 2013 ) . As allelic exclusion in TCR-transduced HSCs is incomplete , however , there remains a risk that TCR mispairing can precipitate autoimmunity . HSCs transduced with dsTCRs would retain the advantages of wtTCR-engineered HSCs while eliminating the risk of autoimmunity . We injected HLA-A*0201+ immunodeficient ( NSG ) mice with F5 wtTCR- or dsTCR-transduced human CD34+ HSCs . Three months after injection , A2/MART1-specific CD3ε+ T cells constituted up to 87 . 7% of human CD45+ thymocytes and up to 66 . 7% of human CD45+ CD19- CD3ε+ splenocytes in mice engrafted with dsTCR-transduced stem cells ( Figure 6 ) . Therefore , dsTCRs are compatible with TCR-engineered HSC adoptive transfer . 10 . 7554/eLife . 19095 . 016Figure 6 . dsTCR-transduced human HSCs develop into antigen-specific T cells in humanized mice . Flow cytometry plots showing dissociated thymocytes or splenocytes from humanized mice stained with A2/MART1 tetramer and anti-CD3ε . Mice were injected with TCR-transduced HSCs three months prior to analysis . Mouse 3 and 4 are replicates , mouse 5 and 6 are replicates . Mouse 2 has no replicate because wtTCR-transduced HSCs did not engraft in Mouse 1 ( not shown ) . Results are from a single experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 016 A subset of T cells encode γδ TCRs , which share a broadly similar architecture to αβ TCRs , are somatically rearranged , associate with CD3 chains , and mediate antigen-specific cellular immunity , but which are functionally distinct from αβ TCRs ( Attaf et al . , 2015 ) . Interestingly , αβ TCR chains transduced into γδ T cells do not mispair with endogenous γδ TCR chains ( Saito , 1988; van der Veken et al . , 2006 ) . These αβ TCR-transduced γδ T cells mediate antigen-specific immunity in vitro and in mouse models ( van der Veken et al . , 2006 , 2009 ) . However , the relative paucity of γδ T cells in the blood , the requirement of co-delivery of CD4 and CD8 co-receptors , and the lower proliferative capacity and in vivo effector function of these cells relative to αβ T cells are challenges of using αβ TCR-transduced γδ T cells for adoptive cell therapies ( van der Veken et al . , 2006 , 2009 ) . Ideally , the lack of mispairing between αβ and γδ TCRs could be exploited in the context of the more abundant and co-receptor-expressing αβ T cells . We explored swapping corresponding constant domains of αβ and γδ TCRs as an approach to prevent mispairing that is conceptually related to but distinct from our αβ interchain domain-swapping approach . We designed three chimeric constructs in which the F5 TCR domains Vα and Vβ were fused to Cδ and Cγ ( dsTCRCδγ ) ; VαCα and VβCβ were fused to cpδ and cpγ ( dsTCRδγ ) ; or VαCα and VβCβ were fused to cpγ and cpδ ( dsTCRγδ ) , ( Figure 7A ) . Mispairing between these constructs and wild-type endogenous αβ TCR chains is expected to be disfavored due to juxtaposition of αβ and γδ constant domains ( Figure 7B ) . 10 . 7554/eLife . 19095 . 017Figure 7 . Substitution of αβ TCR constant domains with corresponding domains from γδ TCRs prevents TCR mispairing . ( A ) Schematic of αβ/γδ chimeric TCR architectures . Domains are swapped following the V domains ( dsTCRCδγ ) or C domains ( dsTCRδγ and dsTCRγδ ) . ( B ) Schematic of simulated mispaired constructs , in which αβ and γδ constant domains are juxtaposed . Such mispairing is expected to be unproductive because αβ and γδ constant domains do not interact . ( C-E ) Flow cytometry histograms comparing peptide-MHC multimer binding by Jurkat T cells transduced with ( C ) F5 wtTCR or αβ/γδ chimeric dsTCR constructs , ( D ) F5 wtTCR , F5 dsTCRδγ and constructs simulating mispairing between these , or ( E ) F5 wtTCR , F5 dsTCRγδ and constructs simulating mispairing between these . Histograms are representative of triplicate results from 2 independent experiments . ( F ) Flow cytometric measurement of percent of dextramer-binding TCR and LNGFR expressed on the surface of transduced primary human T cells . LNGFR is an independent transduction marker expressed from the same vector as TCR . ( G ) Flow cytometric measurement of CD25 expressed on the T cell surface and ( H ) ELISA measurement of secreted IFN-γ from TCR-transduced primary human T cells following 48 hr coincubation with cognate or control antigen-expressing K562 target cells . For panels ( F-H ) , means ± SD for 3 technical replicates are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 01710 . 7554/eLife . 19095 . 018Figure 7—figure supplement 1 . Chimeric αβ/γδ dsTCRs function similarly with cytoplasmic domains derived from αβ or γδ TCRs . ( A ) Flow cytometric measurement of dextramer-binding TCR and LNGFR expressed on the surface of transduced primary human T cells . LNGFR is an independent transduction marker expressed from the same vector as TCR . Schematics of constructs are above corresponding data . Colors used in schematic correspond to those in Figure 6 . ( B ) Flow cytometric measurement of CD25 expressed on the T cell surface from TCR-transduced primary human T cells following 48 hr coincubation with cognate or control antigen-expressing K562 target cells . For both panels , means ± SD for 3 technical replicates are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 19095 . 018 Constructs were initially tested for surface expression and dextramer binding in transduced Jurkat T cells . As with the interchain domain-swapped constructs ( Figure 2B ) , chimeric domain-swapping at the V-C junction resulted in loss of TCR surface expression ( Figure 7C ) . By contrast , both dsTCRδγ and dsTCRγδ expressed on the surface and bound cognate dextramer to a similar extent as wtTCR . As expected , 3 of the 4 simulated mispaired constructs comprising one chimeric chain and one wtTCR chain did not bind dextramer ( Figure 7D , E ) . Only the wtα/dsβγ mispaired construct was detected at a low level on the cell surface , binding dextramer at 20% the level of wtTCR ( Figure 7D ) . Dextramer binding results were similar in primary T cells , indicating that chimeric dsTCRδγ and dsTCRγδ assemble with CD3 chains with similar efficiency as wtTCR in the presence of competing endogenous TCR ( Figure 7F; MFI for dsTCRδγ and dsTCRγδ = 157% and 174% that of wtTCR , respectively ) . T cells expressing chimeric dsTCRs were activated and secreted IFN-γ in response to coincubation with cognate antigen-expressing target cells ( Figure 7G , H ) . In contrast to dextramer binding , these functional outputs were reduced for chimeric dsTCRs relative to wtTCR , indicating chimeric domain-swapping partially impaired signaling capacity . Replacing the chimeric TCR cytoplasmic signaling domains with those from αβ TCR did not substantively alter the function of chimeric dsTCRs ( Figure 7—figure supplement 1 ) . As observed in Jurkats , chimeric dsαδ , dsαγ , and dsβδ chains did not mispair productively with wild-type αβ TCR chains ( Figure 7F–H ) . Only the wtα/dsβγ mispaired construct exhibited low levels of dextramer binding ( 7% of wt ) and secreted IFN-γ ( 20% of wt ) , indicating some allowance for pairing between constant TCRα and TCRγ domains . Thus , substitution of constant domains of tumor-specific αβ TCRs with corresponding γδ TCR domains produces a chimeric receptor that mediates functional , antigen-specific T cell immunity with minimized mispairing risk .
The promise of TCR gene transfer – to enable us to initiate or amplify beneficial immune responses – is already coming to fruition in the field of cancer immunotherapy ( Johnson et al . , 2009; Morgan , 2006; Davis , 2010; Parkhurst , 2011; Robbins , 2015 , 2011 ) . However , mispairing between introduced and endogenous TCR chains produces autoreactive human T cells in vitro ( van Loenen et al . , 2010 ) and causes graft-vs-host disease in mice ( Bendle et al . , 2010; Bunse , 2014 ) , indicating that TCR gene transfer poses a non-theoretical risk of autoimmunity . As the potency of TCR gene therapy is augmented through combination with other immunotherapies ( e . g . checkpoint blockade ) , there may be a concomitant rise in mispairing-related serious adverse events . This work describes a domain-swapping strategy to prevent TCR mispairing , thereby improving the safety of this promising approach . By swapping constant domains between α and β chains of a clinically-relevant αβ TCR , we generated a receptor that retains antigenic specificity and function but does not mispair with endogenous TCRs . Interchain domain-swapped TCRs prevent mispairing-related autoreactivity through three potential mechanisms: ( 1 ) dsTCR chains mispair with wtTCR chains inefficiently due to juxtaposition of non-interacting domains ( i . e . neither α nor β self-associate ) ; ( 2 ) mispairing between wt and ds chains results in a receptor that incompletely assembles with CD3 chains and is degraded; ( 3 ) mispaired wt/dsTCRs that escape degradation and express on the surface with an incomplete complement of CD3 chains will have impaired signaling capacity . Thus , proper dsTCR/CD3 complex assembly is central to the differential function of correctly paired versus mispaired dsTCRs . The detailed mechanism by which dsTCRs assemble with CD3 chains remains open . The transmembrane domains and connecting peptides of wtTCRs participate in complex assembly while the ectodomains contribute to complex stability and signaling ( Call et al . , 2002; Kuhns and Davis , 2007; Xu and Call , 2006 ) . The dsTCRs reorient these domains with respect to one another . Despite this , dsTCRs assemble with CD3 chains with proper juxtaposition of CD3ε chains , express on the cell surface , and mediate functional , antigen-specific responses , including tumor cell killing . This indicates that dsTCR/CD3 complexes preserve the transmembrane contacts between the TCR and CD3 chains that are essential for complex assembly ( Call et al . , 2002 ) . If transmembrane orientation is preserved , it is unlikely that the short , rigid CD3 connecting peptides can contact the reoriented TCR ectodomains , resulting in presumed ( though undemonstrated ) disruption of these stabilizing contacts . This could account for the generally lower surface expression observed for dsTCRs relative to the wtTCR . The dsTCRcp construct additionally reorients the connecting peptides with respect to the transmembrane domains , consistent with its generally lower surface expression relative to the dsTCRC construct . That both of these constructs transduce extracellular ligand binding into intracellular signaling suggests that the function of the TCR/CD3 complex is remarkably robust toward structural perturbation . The functional avidity of dsTCR-transduced T cells can be optimized for clinical application in several ways . First , wtTCR- and dsTCR-transduced Jurkat T cells performed equivalently in functional assays , indicating the importance of high transduction efficiency . Second , assembly of dsTCRs with CD3 may be improved by reducing the number of endogenous TCRs competing for these chains . In this work , we showed that dsTCRs are compatible with TCR-transduced HSC transfer , as well as with shRNA-mediated knockdown of endogenous TCR genes . Combining dsTCRs with these adjunct approaches improved cell surface expression of dsTCRs and safeguarded against mispairing with residual endogenous TCR . Third , swapping domains closer to the V-C junction would avoid reorienting constant domains with respect to each other , thereby preserving all contacts with CD3 . We attempted to optimize such a construct in this work ( dsTCRV and dsTCRCδγ ) , but found that all constructs that swapped domains at the V-C junction abrogated TCR surface expression completely . Interestingly , a conceptually related approach has been used to ensure correct pairing of heavy and light chains in bispecific antibodies ( Schaefer et al . , 2011 ) , and the lead architecture ( CrossMabCH1-CL ) from that application swapped domains at the V-C junction ( i . e . the elbow ) . It is unclear why the dsTCRV variant was entirely nonfunctional whereas analogously manipulated CrossMab antibodies bind antigen with affinity comparable to wild-type Fab . This may be attributable to a more flexible antibody hinge or to the specifics of how the domain-swap was achieved . Notably , the CrossMab approach differed from ours in that domains were swapped asymmetrically between chains ( i . e . heavy chain residues were duplicated in the hinge/elbow region of both the heavy and light chains ) , so a similar modification to our approach may identify a functional dsTCR design that is swapped near the V-C junction . Finally , although we focus on TCRs relevant to cancer immunotherapy , domain-swapping was applied successfully to all TCRs attempted , including four human TCRs and two mouse TCRs . Thus , the general dsTCR approach can be applied broadly to TCRs of therapeutic interest , enabling development of safer immunotherapies for a variety of cancers , infections , and autoimmune diseases .
Primers were purchased from Integrated DNA Technologies ( Coralville , IA ) . KOD polymerase master mix and polybrene were purchased from EMD Millipore ( Darmstadt , Germany ) . Sequencing was performed by Retrogen Inc ( San Diego , CA ) . Fluorescently labeled antibodies and 7-AAD for flow cytometry were purchased from Biolegend ( San Diego , CA ) , eBioscience ( San Diego , CA ) , or Beckman Coulter ( Brea , CA ) . Peptides were purchased from Anaspec Inc ( Fremont , CA ) or Thermo Fisher Scientific ( Waltham , MA ) . Anti-CD3 ( OKT3 ) and anti-CD28 ( CD28 . 2 ) activating antibodies were purchased from eBioscience ( San Diego , CA ) . Short-dated Proleukin was provided by Prometheus Laboratories Inc ( San Diego , CA ) through an investigator-initiated trial program . All other cytokines were purchased from Peprotech , Inc . ( Rocky Hill , NJ ) . Concanavalin A was purchased from Sigma-Aldrich ( St . Louis , MO ) . RetroNectin was purchased from Clontech Laboratories ( Mountain View , CA ) . BioT transfection reagent was purchased from Bioland Scientific ( Paramount , CA ) . Cell culture media , antibiotics , and fetal bovine serum were purchased from Corning ( Corning , NY ) . Human AB serum was purchased from Omega Scientific ( Tarzana , CA ) . Peptides were purchased from Anaspec ( San Jose , CA ) . 293T/17 , Jurkat E6-1 , and K562 cells were purchased from the American Type Culture Collection ( Manassas , VA ) . Their identities were authenticated by STR analysis and they were confirmed to be mycoplasma-free ( DDC Medical , Fairfield , OH ) . 293T cells were grown in D10 media ( Dulbecco’s Modified Eagle Medium ( DMEM ) supplemented with antibiotics ( penicillin/streptomycin ) and 10% ( v/v ) fetal bovine serum ( FBS ) ) . Jurkat and K562 cells were grown in C10 media ( ( RPMI 1640 ) medium supplemented with antibiotics , 10% ( v/v ) FBS , 10 mM HEPES , 50 μM β-mercaptoethanol , 1x MEM NEAA , and 1 mM sodium pyruvate ) . EL-4 and E . G7 mouse thymoma cell lines were provided by Lili Yang ( UCLA ) and grown in C10 media . The cells were split every 2–3 days . All cells were grown and assayed at 37°C with 5% atmospheric CO2 . Female 6 wk old C57BL/6 J ( B6 ) mice and B6 . SJL-PtprcaPepcb/BoyJ ( CD45 . 1 B6 ) mice were purchased from Jackson Laboratory ( Bar Harbor , ME ) and housed in the California Institute of Technology animal facility . Animals were granted access to food ( PicoLab Rodent Diet 5053 , PMI Nutrition International , St . Louis , MO ) and water ad libitum throughout the study . The room was maintained on a 13:11 hr light:dark cycle . The ambient temperature remained between 71–75 °F with a relative humidity of 30-70% . All injections ( retroorbital , intraperitoneal , and subcutaneous ) were performed under anesthesia ( isoflurane ) . Local anesthetic ( 0 . 5% proparacaine hydrochloride ) was administered following retro-orbital injections . Animals were sacrificed by controlled CO2 inhalation . Experiments conducted at Caltech were approved by the Institutional Animal Care and Use Committee of the California Institute of technology ( IACUC protocol 1611 ) . Animals for TCR/shRNA combination experiments at MDC Berlin were purchased as 6 wk old female C57BL/6 ( Charles River ( Sulzfeld , Germany ) and housed at the MDC SPF facility . These experiments were conducted in accordance with institutional and national guidelines , after approval by a state animal ethics committee ( Landesamt für Gesundheit und Soziales protocol 0233/11 ) . Animals for HSC transduction experiment were bred , housed , and monitored according to UCLA Department of Laboratory Animal Medicine standards ( UCLA Animal Research Committee protocol 2008–167 ) . HLA-A2 and β2-microglobulin were expressed in E . coli , refolded in the presence of MART126-35 ( A27L ) ( ELAGIGILTV ) or NY-ESO-1157-165 ( C165V ) ( SLLMWITQV ) heteroclitic peptides , purified , and biotinylated as previously described ( Toebes et al . , 2006 ) . Biotinylated H-2Kb/Ova257-264 monomers were provided by the NIH Tetramer Core Facility ( Atlanta , GA ) . Peptide-MHC monomers were stored at −20°C or −80°C until use . Peptide-MHC tetramers were prepared from monomers using fluorescently-labeled streptavidin from Molecular Probes ( Eugene , OR ) according to the NIH Tetramer Core Facility protocol . Peptide-MHC dextramers were prepared by doping the tetramer assembly reaction with biotinylated dextran ( manuscript in preparation ) . Briefly , peptide-MHC and fluorescently-labeled streptavidin were added at a molar ratio of 3:1 ( 1 . 5 μM:0 . 5 μM ) and incubated 10 min . Biotinlyated dextran , MW 500 kDa ( Molecular Probes ) was then added to the incubation at 25 nM ( 1:20 with respect to streptavidin ) and incubated an additional 10 min . Peptide-MHC multimers were prepared fresh or stored briefly at 4°C . OTI TCR genes were obtained from Gavin Bendle ( Univ . of Birmingham , UK ) . F5 TCR genes were obtained from Steve Rosenberg ( NCI , Bethesda , MD ) . 1 G4 TCR genes were obtained from Richard Koya ( RPCI , Buffalo , NY ) . M1 TCR genes and mouse and human CD3 genes were obtained from Lili Yang ( UCLA , Los Angeles , CA ) . Novel TCR architectures were prepared by assembly PCR . Amino acid sequences of hybrid joints are provided within relevant figures . For epitope-tagged TCR constructs , Myc or V5 epitope tags were inserted between the leader sequence and the N-terminus of the TCR chain . Peptide-MHC single-chain trimers of A2/NY-ESO-1 ( SLLMWITQV ) and A2/MART1 ( ELAGIGILTV ) were prepared with a disulfide trap modification as described ( Hansen et al . , 2009 ) . 293T were plated ( 25 K/well on 96-well plate ) a day prior to transfection . Cells in each well were transfected with 0 . 2 μg each of plasmids encoding TCR , CD3δ , CD3ε , CD3γ , and CD3ζ , using BioT transfection reagent according to manufacturer’s instructions . Media was refreshed after 24 hr . Cells were de-adhered from the plate using 2 mM EDTA in PBS , stained with fluorescent antibodies and pMHC multimers in FACS buffer ( 2% FBS in PBS ) , and analyzed by flow cytometry 48 hr after transfection . TCR constructs were subcloned into a pCCLc-MND-based lentiviral vector in the format TCRα-F2A-TCRβ-P2A-Myc271 . Myc271 is a novel chimeric transduction marker comprising the transmembrane and truncated extracellular domains of CD271 ( LNGFR ) fused to an extracellular cMyc epitope tag . Lentiviral vectors were prepared from 293T producer cells . One day prior to transfection , 1 x 106 293T cells/well were plated on 6-well plates . In 100 μL PBS , added 1 μg TCR lentivector , 1 μg pCMV-R8 . 2 ( gag-pol helper plasmid ) , 0 . 2 μg pCAGGS-VSVG ( pseudotyping helper plasmid ) , and 6 . 6 μL polyethyleneimine transfection reagent ( Sigma ) and incubated for 10 min . Fresh 2 mL media was replaced on plated 293T , and then transfection mix was added to cells . 48 hr following transfection , Jurkat T cells ( 10 ( Johnson et al . , 2009 ) cells/well in 250 μLC10 ) were added on a 24-well plate and 250 μL filtered ( 0 . 45 μm ) viral supernatant was added to cells to initiate transduction . After 60–72 hr of transduction , Jurkats were assayed for TCR surface expression and function . To test for surface expression , cells were rinsed and stained with fluorescent antibodies and pMHC multimers in FACS buffer , and analyzed by flow cytometry . Dead cells were excluded from analysis using 7-AAD . To test for function , unselected Jurkat T cells were co-incubated in a 96-well plate at a 1:1 ratio with K562 cells stably expressing control or cognate peptide-MHC single chain trimers . After 48 hr of coincubation , the supernatant was tested for secreted interleukin-2 with a commercial ELISA kit according to the manufacturer’s protocol ( BD , Franklin Lakes , NJ ) . TCR constructs were subcloned into an MSCV-based retroviral vector in the format LNGFRΔ-P2A-TCRα-F2A-TCRβ . LNGFRΔ is a transduction marker comprising the low-affinity nerve growth factor receptor with intracellular domain truncated . Retroviral vectors were prepared from 293T producer cells . One day prior to transfection , 6-well plates were coated with poly-L-lysine and 1 x 106 293T cells/well were plated . On day of transfection , 2 μg TCR lentivector , 2 μg pHIT60 ( gag-pol helper plasmid ) , and 1 . 3 μg pRD114 ( pseudotyping helper plasmid ) were mixed in 100 μL 250 mM CaCl2 . DNA/CaCl2 was added while vortexing to 100 μL 2x HEPES-buffered saline ( HBS ) , incubated 5 min , and then added to cells in fresh D10 media . The day after transfection , media was replaced with D10 containing 20 mM HEPES and 10 mM sodium butyrate , incubated 8 hr , and then media was replaced with D10 containing 20 mM HEPES . Viral supernatant was harvested 40–48 hr after transfection , filtered ( 0 . 45 μm ) , and used directly to transduce activated T cells . Primary human peripheral blood mononuclear cells ( PBMCs ) were purchased from the CFAR Virology Core Lab at the UCLA AIDS Institute . Two days prior to transduction , a 24-well plate was coated with 1 μg/mL anti-CD3 ( OKT3 ) and peripheral blood mononuclear cells were activated at 2 x 106 cells/mL on the anti-CD3 coated plate in the presence of 1 μg/mL anti-CD28 and 300 U/mL IL-2 in T cell media ( AIM-V media supplemented with 5% ( v/v ) human AB serum and antibiotics ( pen/strep ) ) . On day of transduction , activated T cells were centrifuged with virus and 10 μg/mL polybrene at 1350xg for 90 min at 30°C . Following spinfection , virus was replaced with fresh T cell media supplemented with 1 μg/mL anti-CD28 and 300 U/mL IL-2 . 48 hr following transduction , transduced primary T cells were assayed for TCR surface expression . Functional assays were performed with unselected transduced T cells as described for Jurkat T cells , except that after coincubation with antigen-expressing K562 derivatives , surface expression of CD25 was measured for primary T cells and the supernatant was tested with a commercial ELISA kit for secreted interferon-γ ( BD , Franklin Lakes , NJ ) . BaF3 cell lines were generated and analyzed by flow cytometry as previously described ( Kuhns et al . , 2010; Lee et al . , 2015 ) . Briefly , Phoenix E packaging cells were used to generate retrovirus for each cell line . In 6 cm plates , 3 µg of the desired retroviral construct along with 0 . 75 µg gag polymerase and 0 . 75 µg eco envelope were mixed in DMEM with 10% FCS and then transfected into 1 x 106 Phoenix E packing cells using Turbofect ( Fermentas ) according to the manufacturer’s instructions . The next day , transfected Phoenix E cells were shifted to 32°C after media change . Viral supernatants were harvested at 48 and 72 hr post-transfection . The supernatants for all constructs used to generate a cell line were pooled and concentrated to 250 µl using an Amicon Ultra 15 100 kDa ( Millipore ) . BaF3 cells were co-transduced simultaneously with a polycistronic vector encoding all CD3 chains ( cytoplasmic domains truncated and CD3ε fused to EpoR cytoplasmic domain ) and a vector encoding a TCR construct . Parental BaF3 cells ( 1 x 106 ) were plated in 2 ml of complete RPMI in one well of a 12 well plate in 8 µg/ml polybrene , 1 ng/µL IL-3 ( R and D Systems ) , and the viral supernatant . Cells were then spun for 2 hr at 32°C at 2700 rpm in a Legend XTR centrifuge ( ThermoFisher ) , after which the media was exchanged with fresh complete RPMI containing 1 ng/µL IL-3 . Transduced cells were cultured overnight at 37°C prior to selection for εE expression with 10 µg/mL puromycin ( InvivoGen ) . Cells transduced with the 2B4 control TCR were selected with 10 µg/mL puromycin and 100 µg/mL zeocin ( LifeTech ) . BaF3 cells were split as necessary to keep them sparse and under heavy drug selection . All cell lines were maintained at a density below 1 x 106/mL and split the night before experiments to be in mid growth phase the day of the experiment . Surface expression was assessed by flow cytometry with the following antibodies: anti-CD3ε-PE-Cy7 ( clone 2 C11 ) , anti-Vα2-PE ( clone B20 . 1 ) , anti-Vβ5-FITC ( clone MR9-4 ) , and HLA-A2/MART1 ( ELAGIGILTV ) and H-2Kb/ovalbumin ( SIINFEKL ) dextramers . BaF3 cells ( 2 . 5 X 104/well ) were setup in 96 well plates in triplicate under drug selection for a 3-day proliferation assay in the absence of IL-3 . Cells were enumerated by flow cytometry with count beads , which allowed the total cell number per well to be extrapolated . Humanized mice were generated as previously described ( Gschweng et al . , 2014 ) . Briefly , HLA-A*0201+ , CD34-enriched human peripheral blood stem cells were stimulated for 24 hr before transduction with spin concentrated virus encoding the indicated constructs at a vector concentration of 1 x 108 transduction units/mL . Twenty four hours later , cells were collected , washed , and resuspended at a concentration of 2 x 107 cells/mL in X-VIVO-15 medium ( Lonza ) . Neonatal ( d3-5 ) NSG-A0201 mice ( Jax 014570 ) received 1 Gy from a 137Cs source immediately before injection . A dose of 1 x 106 cells ( 50 μl ) was delivered to each mouse via intrahepatic injection . Three months post-transplant , mice were harvested for analysis . Thymus and spleen from each mouse was surgically excised and dissociated over 70 μm strainers in RPMI + 10% FBS . Flow cytometry analysis was performed on 1 x 106 cells per sample , and stained with anti-mouse CD45 , HLA-A2/MART1 ( ELAGIGILTV ) multimer , and the following anti-human antibodies: CD45 , CD19 , CD3 , CD4 , and CD8 . Data were acquired on an LSRFortessa ( BD ) . Murine OTI TCR constructs were subcloned into the MSCV-based retroviral vector pMX in the format TCRβ- ( F2A or IRES ) -TCRα . Retroviral vectors were prepared from 293T producer cells . Two days prior to transfection , 5 x 106 293T cells were seeded per 15 cm plate . On day of transfection , 11 . 25 μg retroviral vector ( encoding OTI wtTCR , OTI dsTCR , or LNGFRΔ control ) and 11 . 25 μg pCL-Eco ecotropic packaging vector were mixed in 1000 μL serum-free DMEM . Transfection reagent ( 33 . 75 μL BioT ) was added and transfection mix was immediately vortexed , incubated for 5 min , and then added to cells in fresh D10 media . The following day , media was replaced on 293T and a Retronectin-coated plate was prepared ( 24-well plate coated overnight with 30 μg/mL Retronectin overnight at 4°C and then blocked with 2% ( w/v ) BSA for 30 min at room temperature ) . Viral supernatant was harvested at 48 and 72 hr after transfection , filtered ( 0 . 45 μm ) , and added directly to blocked Retronectin-coated plates . Virus was bound to plate by spinning at 3000 xg for 120 min at 4°C . Supernatant was removed immediately prior to adding activated T cells to virus-coated plate . Primary mouse T cells were harvested from mouse spleen . Mouse spleens were dissociated and treated with RBC lysis buffer ( Biolegend ) to remove erythrocytes . Mouse splenocytes ( 2 x 106 cells/well on a 24-well plate ) were activated in C10 containing 2 μg/mL concanavalin A and 1 ng/mL IL-7 for 24 hr . Activated T cells were pooled , supplemented with 4 μg/mL protamine sulfate , spun on the virus-coated plate at 800 xg for 30 min at 32°C , and then incubated in the presence of virus overnight at 37°C . The next day , cells were washed with PBS , and then transduced a second time on fresh , virus-coated plates . The day after the second transduction , cells were counted and analyzed for transgene expression by flow cytometry . TCR-transduced cells were evaluated directly for activity or transferred to recipient mice for in vivo assays as described below . Mouse splenocytes ( 2 x 106 cells/well ) were activated in 1 mL C10 containing 2 μg/mL concanavalin A and 1 ng/mL IL-7 for 48 hr , then transduced with retroviral supernatant encoding OTI wtTCR , OTI dsTCR , or no TCR as described . After overnight transduction , T cells were washed and incubated for 16 hr in either C10 media alone or media containing 1:100 H-2Kb/Ova tetramer ( final 39 nM ) and 1 μg/mL anti-CD28 ( clone 37 . 51 ) . Supernatants were collected after the 48 hr ConA/IL-7 activation and after the 16 hr tetramer stimulation for ensemble proteomic analysis based on a DNA-encoded Antibody Library ( DEAL ) assay ( Bailey et al . , 2007 ) . Briefly , a DNA barcoded slide was sectioned into wells using molded elastomers . The surface of the slide was blocked for 1 hr with 3% bovine serum albumin ( BSA , Sigma ) in PBS buffer ( Irvine Scientific ) . A cocktail of antibody-DNA conjugates that capture the measured proteins was added and incubated for an hour in order to hybridize them to the surface ( Bailey et al . , 2007; Ma et al . , 2011 ) . After washing the wells 3 times with 3% BSA , the supernatant containing the secreted proteins was added to individual wells and incubated for an hour . Afterwards , wells were washed again 3 times with 3% BSA . The assay was completed by applying biotinylated antibodies and streptavidin-Cy5 and a final wash with 3% BSA to remove excess dye . Finally , the slide was washed with PBS before spin drying and scanning on a GenePix 4400A fluorescent scanner ( Molecular Devices ) . The capacity of OTI wtTCR- and dsTCR-transduced T cells to induce graft-vs-host disease was evaluated by an assay simulating TCR gene therapy in mice ( Bendle et al . , 2010 ) . Donor mouse ( CD45 . 1 B6 ) splenocytes were activated and transduced with retroviral supernatant encoding OTI wtTCR , OTI dsTCR , or no TCR as described above . Recipient mice received 5 Gy from a 137Cs source to render them lymphopenic one day prior to T cell injection ( d-1 ) . The next day ( d0 ) , a dose of 1 x 106 transduced cells ( 100 μl ) was delivered to each mouse via retro-orbital injection . Initial body weight was recorded on d10 , after which mice received twice daily intraperitoneal injections of 7 . 2 x 105 U IL-2 ( Prometheus Labs ) for three days ( d10-d12 ) . Animals were monitored thereafter for signs of TI-GvHD ( i . e . cachexia ) , and mice were euthanized if body weight fell to 85% initial weight . Donor mouse splenocytes were activated with ConA/IL-7 , transduced with OTI wtTCR , dsTCR , or control vector , and then injected retro-orbitally into lymphopenic recipient mice as described above . Forty-four days following T cell injection , recipient mice were sacrificed and splenocytes were dissociated , RBC-lysed , and seeded in a 48-well plate at 106 cells/mL in C10 alone or C10 containing 1 μg/mL ovalbumin257-264 ( SIINFEKL ) peptide . After 72 hr incubation , the supernatant was tested with a commercial ELISA kit for secreted mouse interferon-γ ( BD ) and the T cells were tested by flow cytometry for surface expression of Vα2 TCRα , Vβ5 TCRβ , H-2Kb/Ova257-264 tetramer-specific TCR , and activation markers ( anti-CD25 ( clone PC61 ) , anti-CD44 ( clone IM7 ) , and CD62L [clone MEL-14] ) . A subset of mice evaluated for TCR gene transfer-induced GvHD were further tested for capacity of transferred T cells to prevent tumor growth of tumor xenografts . Recipient mice were rendered lymphopenic by γ-irradiation , injected with TCR-transduced T cells retro-orbitally , and monitored for TI-GvHD as described above . Forty days following T cell injection , mice were injected on their right flank subcutaneously with 5 x 106 ovalbumin-expressing E . G7 thymoma tumor cells . For tumor specificity experiment , mice were injected with OTI dsTCR-transduced T cells as above and 25 days thereafter injected subcutaneously with 5 x 106 E . G7 ( target ) thymoma tumor cells on the right flank and 5 x 106EL4 ( control ) thymoma tumor cells on the left flank . For T cell titration experiment , mice were injected as above with unsorted splenocytes containing the specified number of transduced ( Vα2+Vβ5+ ) CD8+ T cells and 25 days thereafter injected with 5 x 106 E . G7 thymoma tumor cells . The two longest perpendicular dimensions of each injected tumor were measured with calipers daily and used to calculate the tumor cross-sectional area . Mice were euthanized if the longest dimension measured 15 mm . Retroviral vectors encoding codon-optimized TCR and shRNA targeting endogenous TCR α and β genes were designed and produced as described ( Bunse , 2014 ) . Briefly , ecotropic virus particles were produced using Platinum-E packaging cells and γ-retroviral MP71 vector plasmids . Supernatants were harvested 48 hr after transfection , filtered ( 0 . 45 μm ) and frozen ( −80°C ) . Mouse splenocytes were activated using concanavalin A ( 2 μg/ml , Sigma-Aldrich ) and IL-7 ( 1 ng/ml , Peprotech ) for 24 hr and then the first round of transduction was performed . The cells were transferred into virus-coated plates , protamine sulfate ( 4 µg/mL , Sigma-Aldrich ) was added and the plates were centrifuged ( 800 g , 32°C ) for 30 min . The virus-coated plates were prepared by centrifugation of Retronectin-coated 24-well culture plates with 0 . 5 ml/well viral supernatant for 2 hr ( 3000 g , 4°C ) . The next day a second round of transduction was performed . After overnight incubation , the transduction rate was determined and a population of cells containing 1 x 106 P14 TCR-expressing CD8 T cells was transferred into mice ( C57BL/6 ) that were treated 1 d before with total body irradiation ( TBI , 5 Gy , X-Ray irradiator RS2000 , Rad Source ) . Ten days after T cell injection , the mice received 7 . 2 × 105 ( Johnson et al . , 2009 ) IU Proleukin ( Novartis ) intraperitoneally twice a day for 3 d and were monitored thoroughly ( weight , appearance , behavior ) . Group assignments were blinded and mice were sacrificed according to predefined human endpoints . Human PBMCs were activated with anti-CD3/anti-CD28 and transduced with RD114-pseudotyped virus as described above . To determine whether epitope tags affect TCR surface expression , T cells transduced with F5 wtTCRα ( ± cMyc tag ) and wtTCRβ ( ± V5 tag ) were analyzed by flow cytometry using anti-cMyc and anti-V5 ( Genscript , Piscataway , NJ ) and anti-TCRβ ( clone JOVI . 1 ) primary antibodies , followed by anti-mouse IgG-PE secondary antibody . To measure mispairing of a single introduced F5 TCR chain with endogenous human TCR chains , T cells transduced with cMyc-wtTCRα , cMyc-dsTCRα , V5-wtTCRβ , or V5-dsTCRβ individually were analyzed for surface expression of introduced chains with anti-tag antibodies . Similar analysis was performed with mouse OTI TCR in mouse T cells , but tags were not required due to the availability of antibodies against the variable domains of OTI TCR ( anti-mouse Vα2 ( clone B20 . 1 , Biolegend ) and anti-Vβ5 . 1/5 . 2 ( clone MR9-4 , Biolegend ) . Fresh mouse splenocytes were depleted of T cells expressing endogenous TCRs with these variable domains by binding them with anti-Vα2-PE and anti-Vβ5 . 1/5 . 2 PE and then removing bound cells using anti-PE microbeads and MACS LD Separation Columns ( Miltenyi Biotec , San Diego , CA ) according to the manufacturer’s instructions . Remaining T cells were activated with ConA/IL-7 and transduced as described above with ecotropic virus encoding LNGFR ( transduction marker ) and either OTI wtTCRα , dsTCRα , wtTCRβ , or dsTCRβ . Transduced T cells ( gated on CD45 . 1+CD8a+LNGFR+ ) were analyzed for surface expression of Vα2 and Vβ5 . 1/5 . 2 to measure mispairing of these single introduced OTI chains with endogenous murine TCR chains . Analysis of mispairing of P14 TCR chains ( Vα2 and Vβ8 ) ± codelivered shRNA was conducted identically except that 1 ) recipient cells were depleted of endogenous Vα2+ and Vβ8+ cells prior to transduction , and transduced cells were analyzed with anti-Vα2 and anti-Vβ8 ( clone MR5-2 , Biolegend ) and 2 ) transduced cells were cultured in media with 10 ng/mL IL-15 for 72 hr following transduction and prior to flow cytometry analysis to enable endogenous gene knockdown by shRNA . Extent of endogenous TCR knockdown was determined by flow cytometry , comparing the amount of TCR complex ( anti-TCRβ ( clone H57-597 ) and anti-CD3ε ( clone 145-2C11 ) ) expressed on the surface of T cells transduced with LNGFR ± shRNA Significance analysis of mouse survival ( Kaplan-Meier ) curves was conducted with the log-rank ( Mantel-Cox ) test . Significance analysis of mouse weight loss was conducted with one-way analysis of variance and Tukey’s post-test for multiple comparisons . A statistical probability of p<0 . 05 was considered significant . No power analysis was conducted for animal experiments; the number of animals tested was based on budget and feasibility . No data were excluded as outliers . Descriptive statistics are presented as mean ± standard deviation of replicate assays . Technical replicates are defined as replicate testing of tissue from the same mouse ( only Figure 4—figure supplement 3 ) or identical assays conducted on independent manipulations of the same cell line or primary T cell sample ( e . g . activity assays employ T cell replicates transduced separately and coincubated separately with target cells ) . Biological replicates are defined as identical experiments conducted in multiple mice . | T cells enable the immune system to recognize invading microbes and diseased cells while ignoring healthy cells . The ability of a T cell to recognize a specific microbe or diseased cell is determined by two proteins that pair to form its “T cell receptor . ” The paired receptors are exported to the surface of the T cell , where they bind to infected or cancerous cells . Those T cells that produce receptors that bind healthy cells are eliminated during development . T cells can generally distinguish between the body’s own cells and the cells of invading bacteria or other microbes . However , cancer cells are more difficult to identify because they are similar to healthy cells . Efforts to develop therapies that enhance the immune system’s ability to recognize cancer cells have had only limited success . One successful approach – known as T cell receptor gene therapy – modifies T cells to destroy cancer cells by arming them with a cancer-specific T cell receptor . This technique produces T cells possessing two T cell receptors – the cancer-specific receptor and the one it had originally – so it is possible for proteins from the two receptors to mispair . This impedes the correct pairing of the cancer-specific T cell receptor , reducing the effectiveness of the therapy . More importantly , mispaired T cell receptors may cause the immune cells to attack healthy cells in the body , leading to autoimmune disease . To make T cell receptor gene therapy safe , the cancer-specific receptor must not mispair with the resident receptor . Here , Bethune et al . describe a new strategy to prevent T cell receptors from mispairing . The researchers altered the arrangement of particular regions in a cancer-specific T cell receptor to make a new receptor called a domain-swapped T cell receptor ( dsTCR ) . Like normal T cell receptors , the dsTCRs were exported to the T cell surface and were able to interact with other proteins involved in immune responses . Furthermore , T cells armed with dsTCRs were able to kill cancer cells and prevent tumor growth in mice . Unlike other cancer-specific receptors , dsTCRs did not mispair with the resident T cell receptors in mouse or human cells , and did not cause autoimmune disease in mice . The findings of Bethune et al . show that the structure of the T cell receptor is unexpectedly robust , in that it still works even if it is modified . The next step is to study dsTCRs in more detail with the aim of optimizing them so that they might be used in human clinical trials in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation",
"cancer",
"biology"
] | 2016 | Domain-swapped T cell receptors improve the safety of TCR gene therapy |
Periodontal disease is an age-associated disorder clinically defined by periodontal bone loss , inflammation of the specialized tissues that surround and support the tooth , and microbiome dysbiosis . Currently , there is no therapy for reversing periodontal disease , and treatment is generally restricted to preventive measures or tooth extraction . The FDA-approved drug rapamycin slows aging and extends lifespan in multiple organisms , including mice . Here , we demonstrate that short-term treatment with rapamycin rejuvenates the aged oral cavity of elderly mice , including regeneration of periodontal bone , attenuation of gingival and periodontal bone inflammation , and revertive shift of the oral microbiome toward a more youthful composition . This provides a geroscience strategy to potentially rejuvenate oral health and reverse periodontal disease in the elderly .
Old age is associated with failure to maintain homeostasis resulting in degradation of cellular maintenance and repair processes ( López-Otín et al . , 2013 ) and is the single greatest risk factor for many human diseases including cardiovascular disorders , dementias , diabetes , and most cancers ( Kennedy et al . , 2014 ) . Interventions that target specific aging hallmarks have been shown to delay or prevent age-related disorders and extend lifespan in model organisms ( Kaeberlein et al . , 2015 ) . Rapamycin , an FDA-approved drug , which directly inhibits the mechanistic target of rapamycin complex I ( mTORC1 ) , is one such intervention that extends lifespan and ameliorates a variety of age-related phenotypes ( Johnson et al . , 2013 ) . In mice , rapamycin extends lifespan when administered beginning at 9 or 20 months of age ( Harrison et al . , 2009 ) , and short-term treatments ranging from 6 to 12 weeks during adulthood have been shown to increase lifespan ( Bitto et al . , 2016 ) , improve cardiac function ( Flynn et al . , 2013; Dai et al . , 2014 ) and restore immune function as measured by vaccine response ( Chen et al . , 2009 ) . Initial indications suggest that mTORC1 inhibition may also reverse declines in age-related heart function in companion dogs ( Urfer et al . , 2017a; Urfer et al . , 2017b ) , and age-related immune function ( Mannick et al . , 2014; Mannick et al . , 2018 ) and skin aging ( Chung et al . , 2019 ) in humans . Periodontal disease is clinically defined by inflammation of the periodontium , the specialized tissue surrounding and supporting the tooth structure , resulting in clinical attachment loss , alveolar ( periodontal ) bone loss and periodontal pocketing , and pathogenic changes in the oral microbiome ( Könönen et al . , 2019; Lang and Bartold , 2018 ) . Most recent epidemiologic data in the U . S . population suggests that more than 60% of adults aged 65 years and older have periodontitis ( Eke et al . , 2012; Eke et al . , 2015 ) , and diagnosis with periodontal disease is associated with increased risk for other age-related conditions including heart disease , diabetes , and Alzheimer’s disease ( Gil-Montoya et al . , 2015; Kim and Amar , 2006; Razak et al . , 2014 ) . Given that periodontal disease shows a similar age-related risk profile as other age-associated diseases ( An et al . , 2018 ) , we predicted that interventions which target biological aging could be effective at treating periodontal disease . Consistent with that hypothesis , aged mice treated with rapamycin have greater levels of periodontal bone than control animals ( An et al . , 2017 ) . In order to further test this idea and to understand potential mechanisms by which mTOR activity influences oral health during aging , we carried out a longitudinal study in which we asked whether transient rapamycin treatment during middle age can impact three clinically defining features of periodontal disease: loss of periodontal bone , inflammation of periodontal tissues , and pathogenic changes to the microbiome . Here we report that 8 weeks of treatment with rapamycin in aged mice is sufficient to regrow periodontal bone , reduce inflammation in both gingival tissue and periodontal bone , and revert the composition of the oral microbiome back toward a more youthful state .
In order to understand potential mechanisms by which aging and mTOR activity influence oral health , we carried out two parallel longitudinal studies at two sites in which aged mice were treated with either vehicle control or rapamycin for 8 weeks . NIA-UW mice were housed at the University of Washington and JAX mice were housed at The Jackson Laboratory ( see Materials and methods ) . We used microCT ( μCT ) imaging to measure the amount of periodontal bone present in the maxilla and mandible of young ( 6 month ) , adult ( 13 month ) , and old ( 20 month ) mice from the NIA-UW cohort ( Figure 1A ) . The amount of periodontal bone for the maxilla and mandible of each animal was calculated as the distance from the cementoenamel junction ( CEJ ) to the alveolar bone crest ( ABC ) for 16 landmarked sites each on the buccal aspect of the maxillary and mandibular periodontium ( Figure 1B ) . Thus , larger values represent greater bone loss . As expected , there was a significant loss of periodontal bone with age in the NIA-UW cohort ( Figure 2 , A to C ) . Mice treated with rapamycin for 8 weeks had significantly more bone at the end of the treatment period compared to mice that received the control diet ( eudragit ) ( Figure 2C ) . To determine whether the increase in periodontal bone upon rapamycin treatment reflects attenuation of bone loss or growth of new bone , we performed μCT imaging on mice before and after treatment in the JAX cohort ( Figure 1A ) . Old mice randomized into either the eudragit control or rapamycin treatment groups had significantly less periodontal bone than young mice prior to the treatment period ( Figure 2F ) . After 8 weeks , the rapamycin treated mice had significantly more periodontal bone compared to eudragit controls and also compared to the pre-treatment levels for the same animals ( Figure 2 , D to F ) . The presence of new bone following rapamycin treatment can be observed by comparison of μCT images from the same animals before and after treatment ( Figure 2 , D and E ) . Normal bone homeostasis results from a balance between new bone growth and bone resorption , which is reflected by the ratio of RANKL ( receptor-activator of nuclear factor-κB ligand ) to OPG ( osteoprotegerin ) , and dysregulation of this balance contributes to bone loss in periodontitis ( Darveau , 2010 ) . Consistent with bone loss during aging , we detected significantly greater levels of RANKL in old animals of both cohorts compared to young animals ( Figure 3A and B ) . OPG levels remained relatively stable , resulting in an increase in the RANKL:OPG ratio indicative of bone resorption exceeding bone formation ( Figure 3C ) . These age-associated defects in bone homeostasis were suppressed by eight weeks of rapamycin treatment ( Figure 3 ) . In addition to increased RANKL:OPG ratio , a significant increase in TRAP+ cells was also observed in periodontal bone with age ( Figure 3D and E ) . TRAP ( tartrate-resistant acid phosphatase ) is a histochemical marker of bone resorbing osteoclasts ( Hayman , 2008; Ballanti et al . , 1997 ) . Rapamycin treatment for eight weeks also decreased TRAP+ cells . Together , our data indicate that rapamycin reverses periodontal bone loss in the aging murine oral cavity at least in part through inhibition of bone resorption . Along with bone loss , gingival inflammation is a defining feature of periodontal disease . Aging is also associated with chronic accumulation of pro-inflammatory factors , a collective term referred to as inflammaging ( Chung et al . , 2009; Franceschi and Ottaviani , 1997; Franceschi et al . , 2000; De Martinis et al . , 2005 ) . The nuclear factor-κB ( NF-κB ) is a hub of immune and inflammatory response activated both during normal aging and as a consequence of periodontal disease ( Arabaci et al . , 2010; Ambili and Janam , 2017; Abu-Amer , 2013; Liu et al . , 2017 ) . We first evaluated the NF-κB hub through NF-κB p65 and IκBα expressions levels . The NF-κB heterodimer consists of RelA ( or p65 ) and p50 . IκBα functions as a negative regulator of NF-κB by sequestering it in the cytoplasm . Degradation of IκBα or phosphorylated-IκBα leads to nuclear localization of NF-κB subunits which induce expression of target inflammatory genes , such as TNF-α and IL-1β ( Liu et al . , 2017 ) . In both the gingival tissue and periodontal bone , there was an increase in p65 expression with corresponding decrease of IκBα levels , indicating an age-associated increase in NF-κB inflammatory signaling in the periodontium ( Figure 4 , A and B ) . Eight weeks of rapamycin treatment was sufficient to reverse these changes . We also examined the levels of inflammatory cytokines in the oral cavity associated with normative aging and rapamycin treatment in mice . Consistent with the increase in NF-κB signaling , we found elevated expression of several cytokines in both the gingival tissue and the periodontal bone ( Figure 4 , E and F ) . Eight weeks of rapamycin treatment reversed most age-associated chemokine and cytokine changes in both the gingival tissue and periodontal bone . Thus , transient treatment with rapamycin during middle-age can largely restore a youthful inflammatory state in both the gingiva and periodontal bone of mice ( Wikham , 2016 ) . Dysbiotic shifts in the oral microbiome are thought to play a significant role in the progression of periodontal disease in humans . We and others have previously shown that rapamycin treatment can remodel the gut microbiome in mice ( Bitto et al . , 2016; Jung et al . , 2016; Hurez et al . , 2015 ) ; however , the effect of rapamycin on the oral microbiome has not been explored . Therefore , we sought to evaluate effects of rapamycin on the aged oral microbiome using 16S rRNA gene sequencing and Amplicon Sequence Variant ( ASV ) analysis approach . Examination of the alpha diversity of the oral cavity illustrated a significant increase in species richness during aging that rapamycin attenuated ( Figure 5A , Figure 5—figure supplement 1 ) . Among the most notable alterations in taxonomic abundance between groups was the reduction of Bacteroidetes phylum in the rapamycin-treated old animals ( Figure 5B ) . When pooled across sites , no significant difference was observed between levels of Bacteroidetes in untreated young animals and old animals treated with rapamycin . Old animals treated with rapamycin in the JAX cohort had even lower levels of Bacteroidetes than young untreated mice ( p<0 . 05 ) , whereas in the UW cohort rapamycin treatment lowered the levels of Bacteroidetes to the level of the young mice ( Figure 5—figure supplement 2 ) . The Bacteroidetes phylum consists of over 7000 different species ( Thomas et al . , 2011 ) and includes bacteria associated with human periodontal disease such as Porphyromonas gingivalis , Treponema denticola , and Bacteroides forsythus ( van Winkelhoff et al . , 2002; Torres et al . , 2019; Socransky et al . , 1998 ) . Further , both the Firmicutes and Proteobacteria phyla also showed a significant difference that was age dependent ( Figure 5B ) but was not significantly altered by rapamycin treatment . In order to assess whether rapamycin is shifting the composition back towards a youthful state , we evaluated the beta diversity using weighted UniFrac distances . We discovered a significant separation of the oral microbiome between old control and old rapamycin-treated animals , while no significant differences were observed between young mice and old rapamycin-treated mice ( Figure 5C ) . Overall , we observed no significant differences in alpha diversity , beta diversity , nor relative taxonomic abundance between young untreated mice and old mice treated with rapamycin , suggesting an eight-week treatment with rapamycin reverted the old oral microbiome to a more youthful state . This observation is further supported when analysis of the samples is performed independently by facility ( UW-NIA or JAX ) ( Figure 5—figure supplement 3 ) . Despite differences in animal facility and diet composition , no batch effect was detected when comparing the JAX and NIA-UW cohorts ( PERMANOVA , nperm = 999 , p=0 . 34 ) ( Figure 5—figure supplement 4 ) .
Taken together , our data demonstrate that a short-term treatment with rapamycin in aged mice is sufficient to reverse three clinically defining features of periodontal disease: periodontal bone loss , periodontal inflammation , and pathogenic changes to the oral microbiome . This adds further support for the Geroscience Hypothesis , which posits that any intervention which targets the biological aging process will simultaneously delay multiple age-related diseases and functional declines ( Kaeberlein , 2017; Sierra and Kohanski , 2017 ) . To the best of our knowledge , this is the first report of rejuvenation in the aged oral cavity . This work suggests several interesting questions that it will be important to evaluate in future studies . One such question is whether the effects of rapamycin on the aged periodontium will persist after the treatment period or will rapidly revert back to the aged state . Improvements in age-related cardiac function associated with a similar rapamycin treatment regimen have been found to persist for at least eight weeks following cessation of treatment ( Quarles et al . , 2020 ) , and it will be of interest to determine whether similar outcomes are observed for improvements in oral health . It will also be important in future studies to determine whether these effects are mediated through local inhibition of mTORC1 in the gingiva and periodontal bone or through systemic effects on immune function or other tissues . Likewise , it will be of interest to understand whether additional features of oral health that are known to decline with age , such as salivary function , are improved by rapamycin treatment . Finally , these results suggest the intriguing likelihood that additional geroscience interventions , such clearance of senescent cells , may phenocopy the effects of rapamycin in this context . Such interventions could pave the way for the first effective treatments to reverse periodontal disease and improve oral health in the elderly .
To enhance rigor and reproducibility , experiments were performed on two different cohorts housed at two sites: the University of Washington in Seattle , WA and the Jackson Laboratory in Bar Harbor , ME . To examine the impact of rapamycin on the periodontium during normative aging , we designed a cross institutional study between the University of Washington ( UW ) and the Jackson Laboratory ( JAX ) ( Figure 1A ) . The UW cohorts of C57BL/6Nia ( hereafter termed NIA-UW Colony ) were received directly from the National Institute on Aging ( NIA ) Aged Rodent Colony and acclimated within the UW facilities . The JAX cohorts of C57BL/6J ( hereafter termed JAX Colony ) were born and raised within the JAX facilities . We then treated mice at both sites with encapsulated rapamycin ( eRAPA ) in the diet at 42ppm , which has been shown to significantly increase lifespan of UMHET3 and C57BL6/J mice ( Miller et al . , 2014; Zhang et al . , 2014 ) , or control food ( eudragit ) . All data are from female mice , which have previously been found to have greater increases in lifespan and some health metrics compared to male mice at this dose of rapamycin ( Miller et al . , 2014; Zhang et al . , 2014 ) . For the NIA-UW colonies , five young , five adult , and 20 old ( 10 eudragit and 10 rapamycin ) animals were utilized . While for the JAX colonies a total of 13 young and 26 old ( 13 eudragit and 13 rapamycin ) animals were used . For this study , young , adult , and old mice were 6 , 13 , and 20 months of age , respectively . Twenty NIA-UW mice ( 10 on eudragit , 10 on rapamycin ) received assigned diet treatments at 20 months of age , lasting for 8 weeks , along with five young and five adult mice as normative aging controls . Animals were housed individually in Allentown NexGen Caging ( Allentown , Allentown , NJ ) containing corncob bedding and nestlets . Mice were fed irradiated Picolab Rodent Diet 20 #5053 ( Lab Diet , St . Louis , MO ) . Animals were maintained in a specific pathogen free facility within a Helicobacter spp . -free room . Mice were housed in groups and inspected daily . National Guidelines for the Care and Use of Animals and the IACUC guidelines were followed . All methods are in accordance with The Jackson Laboratory Institutional Animal Care and Use Committee ( IACUC ) -approved protocols . Animals were fed standard Lab Diet 6% 5K52 with eRapa at 42 mg/kg/day or control . Animals had ad libitum access to food and water throughout the study . Animals were checked daily , and once per week the food was topped off . Animals were housed at 3–5 animals per cage . A cohort of mice were transferred into the JAX Center for Biometric Analysis and brought into the imaging suite in groups of 10 mice per scan group . Prior to scanning , the weight of each mouse was recorded and anesthesia induced with 2–3% isoflurane . The mice were then placed in a prone position in the CT scanner and kept anesthetized for the duration of the scan with an isoflurane level of 1 . 2–1 . 5% . A whole head scan was performed with bone mineral density phantoms included on the specimen positioning bed . After the CT scan , the mouse was placed in a warmed isolation cage and allowed to fully recover from the anesthesia . At the end of the imaging session , the cohort was returned to animal housing facility . Animal experimentation was performed in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health . All animals were handled according to approved institutional animal care and use committee ( IACUC ) protocols ( #4359–01 ) of the University of Washington and ( #06005-A24 ) of the Jackson Laboratory . Encapsulated rapamycin ( eRAPA ) was obtained from Rapamycin Holdings , Inc . Food pellets were ground and mixed with encapsulated rapamycin at 42ppm . 300 ml of 1% agar melted in sterile water , and 200 ml of sterile distilled water were added per kilogram of powdered chow in order to make pellets . Pellets were stored at −20°C until use . Control food contained the same concentration of agar and encapsulated material ( eudragit ) without rapamycin at the concentration that matched the rapamycin chow . Eudragit is the encapsulation material used in eRAPA and is a copolymer derived from esters of acrylic and methacrylic acids . Eudragit without rapamycin was thus added to the regular chow at 42 ppm as a vehicle control . For protein analysis by western blot , gingival tissue and alveolar bone was dissected . Total cellular proteins were extracted in RIPA Lysis and Extraction Buffer ( Thermo Scientific , MA , USA ) and EDTA-free Halt protease and phosphatase inhibitor cocktail included to prevent protein degradation during extraction process . Gingival tissue was pooled from co-housed animals and bone samples were from single specimens . Protein concentration was determined by Pierce BCA Protein Assay Kit ( Thermo Scientific ) . 10–20 μg of total protein was separated by SDS-PAGE on 10% or 12% ( w/v ) polyacrylamide gel , then transferred to PVDF membrane using Trans-Blot Turbo Transfer System ( Bio-Rad , CA , USA ) . Antibodies to NF-κB p65 ( D14E12 ) XP ( 8242 , Cell Signaling Technology ) , phospho-IκBα ( B-9 , Santa Cruz ) , IκBα ( 32518 , Abcam ) , GAPDH ( D16H11 ) XP ( 5174 , Cell Signaling Technology ) , RANKL ( G-1 , sc377079 , Santa Cruz ) , and Mouse OPG ( R and D Systems , AF459 ) were used to probe the membrane . Dependent upon the strength of the antibody-dependent signal , either the membranes were stripped with Restore Plus Western Blot Stripping Buffer and reprobed for total antibody , or duplicate gels were run and separate blots probed . Analysis of the cytokine proteome was completed using a Mouse XL Cytokine Array Kit ( R and D Systems , Bio-Techne Corporation , MN , USA ) . Gingiva and alveolar bone samples were individually pooled , protein concentration determined by Pierce BCA Assay Kit and 200 μg of protein lysate loaded . Detection and imaging were performed using ChemiDoc XRS+ ( Biorad , USA ) and Image Lab Software ( Biorad , USA ) . Data analysis was completed per the manufacture’s protocol . Tissues were fixed in Bouin’s solution , and demineralized in AFS ( acetic acid , formaldehyde , sodium chloride ) . Mandibles were processed and embedded in paraffin . Serial sections of 5 μm thickness were collected in the coronal ( buccal-lingual ) plane . Sections were stained for tartrate-resistant acid phosphatase ( TRAP ) to examine osteoclast activity and numbers ( Sigma-Aldrich Kit , St . Louis , MO , USA ) , and Fast Green counterstaining and examined with a Nikon Eclipse 90i Advanced Research Scope . Representative images ( 40x ) were taken of the alveolar bone furcation . The V3-V4 variable region of the 16 s ribosomal RNA gene was amplified using gene-specific primers with Illumina adapter overhang sequences ( 5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3’ and 5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3’ ) . Each reaction mixture contained 2 . 5 µl of genomic DNA , 5 µl of each 1 µM primer , and 12 . 5 µl of KAPA HiFi HotStart ReadyMix . Amplicon PCR was carried out as follows: denaturation at 95°C for 3 min , 35–40 cycles at 95°C for 30 s , 55°C for 30 s , 72°C for 30 s , followed by a final extension step at 72°C for 5 min . PCR products were verified using gel electrophoresis ( 1% agarose gel ) and cleaned with AMPure XP beads ( Agencourt , Beckman Coulter Inc , Pasadena , CA , USA ) . Amplicons were then indexed using the Nextera XT Index Kit V2 set A and set D ( Illumina ) and purified again with AMPure XP beads to remove low molecular weight primers and primer-dimer sequences . DNA concentrations were concentration of 1–2 nM using the SequalPrep Normalization Kit ( Invitrogen ) . Samples were pooled into a single library which was analyzed using the TapeStation 4200 High Sensitivity D1000 assay ( Agilent Technologies , Waldbronn , Germany ) and Qubit High Sensitivity dsDNA assay ( Thermo Fischer Scientific ) to assess DNA quality and quantity . The final pooled library was then loaded on to an Illumina MiSeq sequencer with 10% PhiX spike , which served as an internal control to balance for possible low diversity and base bias present in the 16S amplicon samples , and was run for 478 cycles and generated a total of 5 . 68 million paired-end reads ( 2 × 239 bp ) . Raw paired-end sequences were imported in to Qiime2 ( v . 2019 . 1 ) and were trimmed by 15 nt from the 5’ end and truncated to 239 nt for the 3’ end for both the forward and reverse reads respectively . The trimmed reads were then demultiplexed and denoised using the DADA2 package ( Callahan et al . , 2016 ) . Forward reads were only used in our analysis . Taxonomy was then assigned using the feature-classifier suite trained on the Human Oral Microbiome Database ( HOMD v . 15 . 1 ) ( Escapa et al . , 2018 ) . Samples were then filtered for taxonomic contaminants excluded samples with less than 10 , 000 reads . Alpha and Beta diversity as well as other analysis were done in R-Studio using the Phyloseq ( McMurdie and Holmes , 2013 ) Clustvis ( Metsalu and Vilo , 2015 ) , ggplot2 ( Wikham , 2016 ) , ampvis2 ( Andersen KS et al . , 2018 ) , vegan ( Oksanen J et al . , 2019 ) , ade4 ( Bougeard and Dray , 2018 ) packages as part of the R suite . Taxonomy filtered from samples was determined by analysis of kit controls with no template and zymo sequencing controls of known diversity and abundance in the QIAamp DNA Microbiome Kit ( Qiagen , Hilden , Germany ) and the DNA Clean and Concentrator Kit ( Zymo Research ) . The following taxonomic assignments were removed as part of the dada2 workflow ( Callahan et al . , 2016 ) : Unassigned , Cyanobacteria , acidovorans , pestis , coli , flavescens , sakazakii , durans , diminuta , anthropi , monocytogenes , parasanquinis_clade_411 , otitidis , subtilis , aeruginosa , fermentum . Results for μCT analysis , including measurements , quantitative histology , proteome analysis are expressed as mean ± standard error of mean ( SEM ) . Data were analyzed where appropriate using Student’s t-test or paired t-test ( comparing two groups only ) , or one-way analysis of variance ( ANOVA ) with post-hoc Tukey test for multiple comparisons , where p-values<0 . 05 were considered statistically significant . Statistical analysis was completed GraphPad Prism 8 . 00 ( Graphpad , Software , La Jolla , CA , USA ) . For the 16 s rRNA sequencing , to identify statistically significant differences among agglomerated and normalized amplicon sequence variants ( ASV ) between samples as well as differences in alpha and beta diversity measures , we applied both the unpaired Wilcoxon rank-sum test as well as the two-tailed paired t-test – both with a 95% confidence interval ( α = 0 . 05 ) . Alpha diversity was assessed measuring Shannon , Chao1 , Observed ( ASV ) , and Fisher diversity measures . Beta diversity was measured using weighted Unifrac distances . Statistical analysis for the microbiome analysis was completed in R ( v . 3 . 5 . 3 ) . All data used in the development of this manuscript and the supplemental material are available in the manuscript or the supplemental materials or upon request . Bioinformatic scripts and microbiome data used in the analysis and generation of figures for this manuscript are available on the McLean Lab GitHub repository: https://github . com/kkerns85/Rapamycin_rejuvenates_oral_health_in_aging_mice . git ( Kerns , 2020; copy archived at https://github . com/elifesciences-publications/Rapamycin_rejuvenates_oral_health_in_aging_mice ) . In addition , a web version of the R Markdown is available on Rpubs: https://rpubs . com/kkerns85/Rapamycin_Rmrkdown . The V4-16S rDNA sequences in raw format , prior to post-processing and data analysis , have been deposited at the European Nucleotide Archive ( ENA ) under study accession no . PRJEB35672 . | Age is the single greatest risk factor for many human diseases , including cancer , heart disease , and dementia . This is because , as the body ages , it becomes less able to repair itself . One way to prevent age-related disease and extend lifespan , at least in laboratory animals , is to use a drug called rapamycin . Mice treated with rapamycin live longer , have stronger hearts , and respond better to vaccination . But , despite these promising observations , the use of rapamycin as an anti-aging treatment is still under investigation . One open question is what age-related diseases rapamycin can help to prevent or treat . In the United States , more than 60% of adults over the age of 65 have gum disease . These people are also more likely to have other age-related diseases , like heart disease or Alzheimer's . This association between gum problems and other age-related diseases prompted An et al . to ask whether it might be possible to treat gum disease by targeting aging . To find out whether rapamycin could improve gum health , An et al . performed three-dimensional CT scans on mice as they aged to measure the bone around the teeth . Some of mice were treated with rapamycin , while the rest received a placebo . The mice that received the placebo started to show signs of gum disease as they aged , including inflammation and loss of bone around the teeth . The types of bacteria in their mouths also changed as they aged . Treating mice with rapamycin not only delayed the onset of these symptoms , but actually reversed them . After eight-weeks of the drug , the older mice had lost less bone and showed fewer signs of inflammation . There was also a shift in their mouth bacteria , restoring the balance of species back to those found in younger mice . Rapamycin is already approved for use in people , so a clinical trial could reveal whether it has the same effects on gum health in humans as it does in mice . But there are still unanswered questions about how rapamycin affects the mouth as it ages . These include how the drug works at a molecular level , and how long the changes to gum health persist after treatment stops . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2020 | Rapamycin rejuvenates oral health in aging mice |
From an herbivore's first bite , plants release herbivory-induced plant volatiles ( HIPVs ) which can attract enemies of herbivores . However , other animals and competing plants can intercept HIPVs for their own use , and it remains unclear whether HIPVs serve as an indirect defense by increasing fitness for the emitting plant . In a 2-year field study , HIPV-emitting N . attenuata plants produced twice as many buds and flowers as HIPV-silenced plants , but only when native Geocoris spp . predators reduced herbivore loads ( by 50% ) on HIPV-emitters . In concert with HIPVs , plants also employ antidigestive trypsin protease inhibitors ( TPIs ) , but TPI-producing plants were not fitter than TPI-silenced plants . TPIs weakened a specialist herbivore's behavioral evasive responses to simulated Geocoris spp . attack , indicating that TPIs function against specialists by enhancing indirect defense .
Plant indirect defenses are traits that disable or remove herbivores by manipulating tri-trophic interactions to the advantage of the plant ( Price et al . , 1980 ) . They attract and inform the third trophic level , predators or parasitoids , resulting in increased attacks on herbivores ( Turlings and Wäckers , 2004 ) . Indirect defenses are widespread and include domatia , extrafloral nectar , and food bodies which provide shelter and nutrition for predators and parasitoids , as well as herbivory-induced plant volatiles ( HIPVs ) which convey information about feeding herbivores ( Heil , 2008 ) . Field studies with the native tobacco Nicotiana attenuata , a desert annual , and with maize have shown that HIPVs can reduce herbivore loads by 24% to more than 90% , by increasing both predation and parasitization of herbivores ( Kessler and Baldwin , 2001; Rasmann et al . , 2005; Halitschke et al . , 2008; Degenhardt et al . , 2009; Allmann and Baldwin , 2010 ) and deterring herbivore oviposition ( Kessler and Baldwin , 2001 ) . If HIPVs really function as defenses , they should increase Darwinian fitness , defined as successful reproduction , for plants under herbivore attack ( Karban and Baldwin , 1997 ) . But because HIPVs can be perceived by many other members of the ecological community—from herbivores , pollinators , predators and parasitoids to competing or parasitic plants—it is not clear whether HIPVs increase plant fitness in nature ( Dicke and Baldwin , 2010; Kessler and Heil , 2011 ) . The field studies described above have either spanned too short a time to reveal Darwinian fitness benefits , or have not reported fitness data at all ( Kessler and Baldwin , 2001; Rasmann et al . , 2005; Halitschke et al . , 2008; Degenhardt et al . , 2009; Allmann and Baldwin , 2010 ) . Two laboratory studies showed that parasitization of herbivores can increase plant reproduction ( van Loon et al . , 2000; Hoballah and Turlings , 2001 ) , but the parasitization in these studies was not mediated by HIPVs . Hence three decades after their description , it remains unclear whether HIPVs are really indirect defenses . Long-term field studies comparing HIPV-emitting vs -deficient plants are required in order to demonstrate a defensive function for HIPVs . Experimental additions of pure volatiles or mixes to plants growing in nature has worked well to test short-term effects of specific compounds ( Kessler and Baldwin , 2001; Allmann and Baldwin , 2010 ) , but only endogenously produced HIPV emissions can ensure specific , lasting and consistent differences under field conditions . Most evidence for the utility of HIPVs comes from studies in which predators and parasitoids learn to associate HIPVs with prey; naïve predators and parasitoids are just as likely to respond to HIPVs as not to respond ( Allison and Hare , 2009 ) . Thus the inducibility of HIPV emission , which ensures association with herbivore feeding , is likely essential for HIPV function , but it is difficult to engineer ( Kos et al . , 2009 ) . Engineered constitutive HIPV emissions have been used , either on predators and parasitoids trained to associate target volatiles with prey in short-term laboratory experiments ( Kappers et al . , 2005; Schnee et al . , 2006 ) , or in set-ups in which target volatiles are always associated with prey ( Rasmann et al . , 2005; Degenhardt et al . , 2009 ) . When plants are engineered constitutively to emit HIPVs , they no longer provide accurate information about the location of feeding herbivores , and predators will not associate these signals with prey in nature . Genetically silencing the biosynthesis of HIPVs , however , permits naturally inducible wild-type ( WT ) plants to serve as HIPV emitters , for comparison with transformed lines lacking specific volatile components ( Halitschke et al . , 2008; Skibbe et al . , 2008 ) . Furthermore , field experiments that manipulate the production of HIPVs which not only attract the third trophic level , but also influence the second trophic level ( e . g . , as feeding stimulants and host location cues ) , require additional experimental manipulations to preserve the plant-herbivore part of the tritrophic interaction . When HIPVs do attract the third trophic level , how can herbivores adapt ? Many herbivores can outgrow their vulnerability to predators and parasitoids , but plant direct defenses can slow herbivore growth and prolong vulnerability as postulated by the slow growth-high mortality hypothesis ( Benrey and Denno , 1997; Williams , 1999; Kessler and Baldwin , 2001 , 2004; Kaplan and Thaler , 2011 ) . The solanaceous specialists Manduca sexta and M . quinquemaculata ( Lepidoptera , Sphingidae ) are resistant to the potent alkaloid toxin nicotine ( Wink and Theile , 2002 ) , but sensitive to the nutritional value of plant tissue ( Zavala and Baldwin , 2004 ) . Non-toxic protease inhibitor ( PI ) proteins , which inhibit protein digestion and thus decrease the availability of organic nitrogen in the form of amino acids ( Zavala et al . , 2008 ) , are widespread in flowering plants ( Hartl et al . , 2011 ) , and trypsin protease inhibitors ( TPIs ) slow the growth of M . sexta on N . attenuata ( Zavala et al . , 2008 ) . However , herbivores can overcome PIs by producing insensitive or desensitized proteases , inactivating or degrading PIs , eating more plant tissue , and eating more nutritious young tissue ( Winterer and Bergelson , 2001; Steppuhn and Baldwin , 2007; Zavala et al . , 2008 ) . In the latter two cases , PIs could reduce plant fitness . Although TPI-producing N . attenuata plants produce more seeds than TPI-deficient plants when attacked by M . sexta under controlled glasshouse conditions ( Zavala and Baldwin , 2004 ) , whether TPIs function as a direct defense in nature is unknown . We tested the hypotheses that HIPVs and TPIs defend plants in nature by increasing herbivore predation and thereby plant Darwinian fitness . To do so , we monitored the performance , predation and mortality of Manduca spp . ( M . sexta and M . quinquemaculata ) on wild-type N . attenuata plants and RNAi transformed lines silenced for the production either of a specific group of HIPVs , or of TPIs , and compared the resulting plant reproductive output in terms of bud and flower production ( we are not permitted to allow transgenic plants to disperse ripe seed ) . Because N . attenuata is an annual , opportunistic out-crosser , seeds are produced within one growing season , mostly from fertilization via self-pollen ( Sime and Baldwin , 2003 ) , and we can relate bud and flower production to lifetime seed production , which is commonly accepted as a measure of Darwinian fitness ( Baldwin , 1998; van Loon et al . , 2000; Hoballah and Turlings , 2001 ) . We hypothesized that HIPVs would increase plant reproduction by increasing predation of herbivores , and that TPIs alone would not increase reproduction under herbivore attack , but would either increase predation or increase herbivores' susceptibility to predators . We then assembled a toolbox of wild-type and transgenic lines chosen to test these hypotheses . We chose a genotype of N . attenuata native to the Great Basin Desert of southwestern Utah . In many years , Manduca spp . larvae cause the most defoliation of N . attenuata plants in this area ( Kessler and Baldwin , 2001 ) and thus the N . attenuata 'UT' genotype is likely adapted to defend against Manduca spp . Eggs and young larvae of Manduca spp . are predated by Geocoris spp . ( big-eyed bugs ) which occur naturally in the Utah habitat and are attracted to components of N . attenuata's HIPV bouquet ( Kessler and Baldwin , 2001; Halitschke et al . , 2008; Skibbe et al . , 2008 ) . Specifically , Utah Geocoris spp . predators are attracted to the sesquiterpene ( E ) -α-bergamotene as well as green leaf volatiles ( fatty acid-derived C6 aldehydes , alcohols and esters ) ( Kessler and Baldwin , 2001; Halitschke et al . , 2008; Schuman et al . , 2009 ) . Green leaf volatiles , or GLVs , can be silenced via a single upstream 13-lipoxygenase , NaLOX2 , which specifically supplies lipid hydroperoxides for their production ( Allmann et al . , 2010 ) . Although GLVs are released upon mechanical damage , the oral secretions ( OS ) of M . sexta convert 3- ( Z ) -GLVs to the 2- ( E ) -structures , resulting in greater Geocoris spp . predation than the damage-induced ( Z ) : ( E ) ratio ( Allmann and Baldwin , 2010 ) . GLVs are released immediately upon damage ( Allmann and Baldwin , 2010 ) and may therefore be a 'first line of defense' . Like GLVs , many other HIPVs are also released after mechanical damage , but change in amount or ratio upon herbivory , and thus GLVs mirror the functional complexity of the total HIPV blend . Furthermore , GLVs prime or directly regulate responses in neighboring plants ( Kessler et al . , 2006; Paschold et al . , 2006 ) , attract herbivores as well as predators ( Halitschke et al . , 2008 ) , and are important cues for pollinating and ovipositing moths ( Kessler and Baldwin , 2001 , 2006; De Moraes et al . , 2001; Fraser et al . , 2003 ) , thus performing several roles which may harm or benefit plant fitness in addition to their role in attracting predators . Perhaps most significantly , GLVs also stimulate Manduca spp . feeding , and silencing plant GLV production results in reduced herbivore damage ( Halitschke et al . , 2004; Meldau et al . , 2009 ) . All these qualities made the manipulation of GLV emissions an ideal means to test rigorously the fitness consequences of HIPV emissions and to evaluate whether these emissions can truly be considered defensive .
We chose a line of irPI plants with no detectable TPI activity ( Steppuhn and Baldwin , 2007 ) , and a line of irLOX2 plants with GLV emissions <20% of WT ( Allmann et al . , 2010 ) ; non-target defense metabolites are not affected in either line ( Steppuhn and Baldwin , 2007; Allmann and Baldwin , 2010 ) , including emission of ( E ) -α-bergamotene measured in a glasshouse characterization of all lines prior to field release ( see 'Non-target metabolites are not affected in irLOX2 , hemi-irLOX2 or irPI plants' ) . Because of the importance of GLVs for the plant-herbivore interaction , we used both homozygous ( Allmann et al . , 2010 ) and hemizygous irLOX2 plants to provide different levels of GLV silencing . Hemizygous ( hemi- ) irLOX2 plants were created by crossing homozygous irLOX2 and irPI plants , but the irPI construct was not active in this cross ( Figure 1 see 'Discussion' ) . 10 . 7554/eLife . 00007 . 003Figure 1 . Trypsin protease inhibitor ( TPI ) activity and transcripts in transformed lines; graphs show means+SEM . ( A ) TPI activity measured in systemic leaves of field-grown ( top two panels , 2011 , N=11–14 for panel 1 and N=21 for panel 2 ) or glasshouse-grown plants ( bottom panel , N=10 ) attacked by Manduca spp . larvae . Only WT , irPI and hemi-irLOX2 plants were used in M4 . For a timeline of Manduca spp . infestations M1–M4 see Figure 4A . For raw data , see F2A_SchumanBarthelBaldwin2012TPIactivity . xlsx ( Dryad: Schuman et al . , 2012 ) . ( B ) Transcripts of PI in unelicited leaf tissue ( control ) , and at the point of maximum accumulation in W+OS-elicited leaf tissue in glasshouse-grown plants ( N=5 ) . For raw data , see F2B_SchumanBarthelBaldwin2012PItranscripts . xlsx ( Dryad: Schuman et al . , 2012 ) . *W+OS treatment had a significant effect on PI ( p<0 . 001 ) transcript accumulation . a , b Different letters indicate significant differences between genotypes ( p<0 . 001 ) in Scheffe post hoc tests following a two-way ANOVA on log2-transformed data with factors treatment and genotype ( genotype F3 , 29=174 . 077 , p<0 . 001; treatment F1 , 29=75 . 909 , p<0 . 001 ) . L . O . D . : below limit of detection for measurement . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 003 The irPI plants ( Steppuhn and Baldwin , 2007 ) had no detectable TPI activity in the glasshouse or throughout the field experiment in 2011 , and PI transcripts accumulated to only 0 . 3% of WT levels in irPI ( transcripts , N=5 , p<0 . 001 in Scheffe post hoc tests following two-way ANOVAs on log2-transformed data with factors W+OS treatment and genotype: treatment F1 , 29=75 . 909 , p<0 . 001; genotype F3 , 29=174 . 077 , p<0 . 001 ) ; in contrast , TPI activity and PI transcripts were similar to WT plants in irLOX2 and hemi-irLOX2 ( transcripts , N=5 , p>0 . 2 in Scheffe post hoc tests following two-way ANOVAs on log2-transformed data with factors W+OS treatment and genotype; activity , N=10–17 , p>0 . 05 in one-way ANOVAs with factor genotype ) ( Figure 1 ) . We assessed GLV production by hexane extraction of GLVs from frozen leaf tissue , and GLV emission by GC analyses of leaf headspaces . GLVs in hemi-irLOX2 plants were reduced to levels similar to those in irLOX2 plants , but hemi-irLOX2 plants still produced detectable amounts of ( Z ) -3-hexenol ( Figures 2 and 3 ) . The dominant GLV in hexane tissue extracts was ( E ) -hex-2-enal , and ( Z ) -hex-3-en-1-ol was additionally quantifiable as a minor component . Only ( E ) -hex-2-enal was quantifiable in extracts from field-grown plants on May 28 , 2011 , and was below quantifiable levels in irLOX2 and hemi-irLOX2 plants , but detectable in pooled samples from hemi-irLOX2 ( Figures 2 and 3 ) . Extracts from later in the season also contained quantifiable amounts of ( Z ) -hex-3-en-1-ol and hemi-irLOX2 extracts contained up to 50% as much of this alcohol as WT and irPI extracts ( N=10 , p<0 . 05 in Scheffe post hoc tests following one-way ANOVAs with factor genotype: June 14 , 2011 , F2 , 26=9 . 556 , p=0 . 001; June 22 , 2011 , F2 , 26=12 . 196 , p<0 . 001; p>0 . 6 for irPI vs WT in a t-test for May 28 and in Scheffe post hoc tests for June 14 and 22 ) ( Figure 3 ) . Headspace measurements from field- and glasshouse-grown plants detected a similar 80–100% reduction in GLV emissions from irLOX2 and hemi-irLOX2 plants compared to WT and irPI ( field , N=3 , p=0 . 024 for hemi-irLOX2 v irPI , p>0 . 05 for hemi-irLOX2 vs WT and irLOX2 vs irPI and WT , but p=0 . 939 for irPI vs WT in Scheffe post hoc tests following one-way ANOVA: F3 , 8=7 . 346 , p=0 . 011; glasshouse , N=4: irLOX2 and hemi-irLOX2 below limit of detection , p=0 . 834 for t-test irPI vs WT ) ( Figure 3 ) , and transcript accumulation of LOX2 was 2% of WT levels in irLOX2 and hemi-irLOX2 ( N=5 , p<0 . 001 in Scheffe post hoc tests following two-way ANOVAs on log2-transformed data with factors W+OS treatment and genotype: treatment F1 , 32=0 . 021 , p=0 . 887; genotype F3 , 32=635 . 477 , p<0 . 001 ) but unaffected in irPI ( p>0 . 9 in Scheffe post hoc test vs WT ) ( Figure 3 ) . 10 . 7554/eLife . 00007 . 004Figure 2 . Hexane extracts of leaves from field-grown plants . ( A ) Hexane extracts from pooled leaf samples of field-grown plants for a qualitative assessment of green leaf volatile ( GLV ) pools , analyzed by GC-MS with a split ratio of 1/100 onto a nonpolar column; only ( E ) -hex-2-enal was identified due to poor resolution of ( E ) -hex-2-enal and ( Z ) -hex-3-en-1-ol on the nonpolar column; no ester peaks were detected . For raw data , see F3A_SchumanBarthelBaldwin2012chromatograms . xlsx ( Dryad: Schuman et al . , 2012 ) . ( B ) Example chromatograms from hexane extracts of individual leaf samples from field-grown plants , analyzed by GC-FID on a wax column . The dominant compound was ( E ) -hex-2-enal; ( Z ) -hex-3-en-1-ol was also present in quantifiable amounts . ( Z ) -3-hexenyl acetate was chosen as an internal standard because no esters were detectable in the preliminary qualitative GC-MS analysis ( 1A ) , and because its chemical similarity to ( E ) -hex-2-enal and ( Z ) -hex-3-en-1-ol made it a good choice of internal standard for normalization and calculation of yield from extracts . For raw data , see F3B_SchumanBarthelBaldwin2012chromatograms . xlsx ( Dryad: Schuman et al . , 2012 ) . IS: internal standard . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 00410 . 7554/eLife . 00007 . 005Figure 3 . GLV production and emission in transformed lines; graphs show means+SEM . ( A ) GLVs extracted with hexane from leaf tissue of field-grown WT , irPI , and hemi-irLOX2 plants grouped in triplets for infestation M4 in 2011 ( Figure 4A ) . Leaves were harvested from every plant at the beginning ( June 14 ) and in the middle of M4 ( June 22 ) and leaves from plants in 10 randomly chosen triplets were analyzed . Only ( E ) -hex-2-enal and ( Z ) -hex-3-en-1-ol were quantifiable in leaf extracts . Different letters ( a and b ) indicate significant differences ( p≤0 . 05 ) in Scheffe post hoc tests following one-way ANOVAs for ( Z ) -hex-3-en-1-ol ( top panel , F2 , 26=9 . 556 , p=0 . 001; bottom panel , F2 , 26=12 . 196 , p<0 . 001 ) . For raw data , see F4A_SchumanBarthelBaldwin2012GLVpools . xlsx ( Dryad: Schuman et al . , 2012 ) . ( B ) GLVs measured in headspace samples of leaves from field-grown ( top panel , N=3 ) or glasshouse-grown plants ( bottom panel , N=4 ) . For field-grown plants , leaves were harvested and measured on May 21 ( just before M3 ) . Intact leaves were kept fresh by placing petioles in water . Immediately before each measurement , one leaf was treated with wounding and M . sexta oral secretions ( W+OS ) ; a 1-cm2 disc was stamped out and placed in a 4-mL GC vial . After 15 min the headspace in the vial was measured with a Z-Nose 4200 and total alcohols and aldehydes were quantified . Different letters ( a and b ) indicate significant differences ( p<0 . 05 ) in Scheffe post hoc tests following one-way ANOVA ( F3 , 8=7 . 346 , p=0 . 011 ) . For glasshouse-grown plants , leaves were left on plants , treated with W+OS , and enclosed in padded , 50 mL food-quality plastic containers for 3 hr while the headspace was pulled over a Poropak Q filter . Filter eluents were measured by GC-MS . Three-hour headspace samples contained ( Z ) -hex-3-en-1-ol , ( E ) -hex-2-en-1-ol ( forms from ( E ) -hex-2-enal on filters over trapping periods longer than 20 min ) , ( Z ) -hex-3-enyl acetate , ( Z ) -hex-3-enyl butanoate , ( Z ) -hex-3-enyl isobutyrate , and ( Z ) -hex-3-enyl propanoate , all of which showed the pattern shown for the total amount . For raw data , see F4B_SchumanBarthelBaldwin2012GLVheadspace . xlsx ( Dryad: Schuman et al . , 2012 ) . ( C ) Transcripts of LOX2 in unelicited leaf tissue ( control ) , and at the point of maximum accumulation in W+OS-elicited leaf tissue in glasshouse-grown plants ( N=5 ) . For raw data , see F4C_SchumanBarthelBaldwin2012LOX2transcripts . xlsx ( Dryad: Schuman et al . , 2012 ) . a , b Different letters indicate significant differences between genotypes ( p<0 . 001 ) in Scheffe post hoc tests following a two-way ANOVA on log2-transformed data with factors treatment and genotype ( genotype F3 , 32=635 . 477 , p<0 . 001 , treatment F1 , 32=0 . 021 , p=0 . 887 ) . L . O . D . : below limit of detection for measurement . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 005 For the ‘UT’ genotype of N . attenuata used in our experiments , the induction of all HIPVs except GLVs is mediated by jasmonate signaling ( Halitschke and Baldwin , 2003; Kessler et al . , 2004 ) . The irPI line A-04-186-1 ( Steppuhn and Baldwin , 2007 ) and irLOX2 line A-04-52-2 ( Allmann et al . , 2010 ) have been characterized previously , and neither is affected in jasmonate signaling . Particularly , the emission of ( E ) -α-bergamotene , the best-characterized HIPV in N . attenuata apart from GLVs ( Halitschke et al . , 2000; Kessler et al . , 2004; Halitschke et al . , 2008; Skibbe et al . , 2008 ) , does not differ significantly among the lines used ( N=4 measured 24–32 hr after W+OS treatment as according to Halitschke et al . ( 2000 ) and normalized as a percentage of the internal standard peak: WT , 67 . 9±17 . 1%; irPI , 30 . 2±13 . 2%; irLOX2 , 26 . 6±5 . 8%; hemi-irLOX2 , 42 . 7±23 . 3%; ANOVA: F3 , 12=1 . 338 , p=0 . 308 ) . The transformation process itself does not affect plant fitness or competitive ability ( Schwachtje et al . , 2008 ) , TPI production or volatile emission ( Figures 1–3 ) . We monitored the predation of Manduca spp . larvae and eggs daily , and counted Geocoris spp . individuals around plants every 2–3 days ( Geocoris spp . counts , Figure 4 ) . In 2010 , we planted into a first-year plot . Although plants were infested with laboratory strain M . sexta larvae ( N=51 ) and baited with M . sexta eggs ( N=50 ) over a 2-week period in 2010 ( Figure 4 ) , no Geocoris spp . individuals were observed on this plot through May . There were also no Geocoris spp . observed through May on a nearby , older plot: Geocoris spp . first arrived and began to predate Manduca spp . eggs on the older plot on June 9 . In 2011 , we planted into the older plot , where we observed Geocoris spp . in May prior to the first infestation ( M2 , Figure 4 ) . 10 . 7554/eLife . 00007 . 006Figure 4 . Experimental timeline and layout . ( A ) Timeline of field experiments in 2010 and 2011 . Different assays and measurements are represented by individual arrows , and rectangles span the time frame of each assay or measurement; narrow rectangles represent single days . Four experimental Manduca infestations ( M1–M4 ) structure the overall experimental design: M1–M3 , with laboratory Manduca , and M4 , with wild Manduca larvae . ( B and C ) Layouts of field plots in ( B ) 2010 and ( C ) 2011 . Thick lines denote the borders of the experiment , thin lines denote irrigation lines ( vertical borders of plot were also irrigation lines in [B] 2010 ) , and R# denotes row number ( used for identifying replicates during the experiment ) . The genotype key in ( B ) applies to both ( B ) and ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 006 During infestation M2 ( Figure 4 ) , we allowed Geocoris spp . to associate all four plant genotypes with the presence of prey: we infested half of all plants with equal numbers of first-instar M . sexta larvae from the laboratory strain and , because Geocoris spp . predate more from GLV-emitting or-perfumed plants ( Kessler and Baldwin , 2001; Halitschke et al . , 2008; Allmann and Baldwin , 2010 ) , we supplemented GLV emission from irLOX2 and hemi-irLOX2 plants by placing cotton swabs with lanolin paste containing GLVs representative of the M . sexta-fed N . attenuata headspace ( Table 1 Allmann and Baldwin , 2010 ) adjacent to M . sexta-infested leaves . Swabs containing lanolin with solvent were placed next to irPI and WT as a control . M . sexta larvae were predated at a rate of 12–37% over two 2- to 3-day trials . Geocoris spp . tended to predate more larvae from GLV-supplemented plants ( Fisher's exact tests , 35–37% vs 22–27% May 5–6 , N=59–60 larvae , p=0 . 066; 17–21% vs 12% May 13–15 , N=92–100 larvae , p=0 . 069; combined trials , Bonferroni-corrected p=0 . 0063 ) ( Figure 5 ) . 10 . 7554/eLife . 00007 . 007Table 1 . GLV mix used to externally supplement plant GLV emission in M2 ( see Figure 4 ) ( Allmann and Baldwin , 2010 ) DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 007ComponentNanogram/20 μL lanolin ( Z ) -hex-3-enal3530 ( E ) -hex-2-enal2690 ( Z ) -hex-3-en-1-ol1780 ( E ) -hex-2-en-1-ol2440 ( Z ) -hex-3-enyl acetate46 . 6 ( E ) -hex-2-enyl acetate35 . 5 ( Z ) -hex-3-enyl propanoate9 . 00 ( E ) -hex-2-enyl propanoate8 . 08 ( Z ) -hex-3-enyl butanoate97 . 0 ( E ) -hex-2-enyl butanoate35 . 6Pure GLVs were diluted in 1 mL of hexane and mixed into 14 mL of lanolin to yield the amount shown per 20 μL , representing the emission per g leaf material within the first 20 minutes of W+OS elicitation . Lanolin containing an equivalent amount of hexane was used as a control . 10 . 7554/eLife . 00007 . 008Figure 5 . Predation of M . sexta larvae and eggs by Geocoris spp . ( A ) Examples of predated M . sexta larva ( left panel ) and egg ( right panel ) . Left , the carcass of a predated first-instar M . sexta larva and typical feeding damage from early-instar Manduca spp . larvae . Right , an intact ( lower left ) and a predated ( upper right ) Manduca spp . egg . In this case , the predated egg collapsed during predation . ( B ) Total predation of M . sexta larvae per trial over two trials during infestation M2 . GLVs were supplemented externally by placing cotton swabs next to Manduca-infested leaves ( 1 per plant ) . Cotton swabs next to irLOX2 and hemi-irLOX2 plants received 20 μL of a GLV mixture in lanolin paste ( Table 1 ) ; those next to WT and irPI plants received lanolin with hexane as a control because hexane was used to dissolve GLVs before mixing with lanolin . N=59–60 larvae on May 5–6 and 92–100 larvae on May 13–15 . Geocoris spp . tended to predate more larvae from GLV-supplemented plants ( Fisher's exact tests , 35–37% vs 22–27% May 5–6 , p=0 . 066; 17–21% vs 12% May 13–15 , p=0 . 069; combined trials , Bonferroni-corrected p=0 . 0063 ) . ( C ) Total percentage of M . sexta larvae ( left panel , N=30 larvae ) and eggs ( right panel , N=88 eggs ) predated in two separate trials during infestation M3 in 2011 ( Figure 4 ) . There was no predation of larvae or eggs by Geocoris spp . in 2010 . Raw data for ( B ) and ( C ) is in F5BC_SchumanBarthelBaldwin2012predation . xlsx ( Dryad: Schuman et al . , 2012 ) . Pictures are of a G . pallens adult predating a first-instar Manduca spp . larva ( left ) and a fifth-instar G . pallens nymph predating a Manduca spp . egg ( right picture , S . Allmann ) . *p<0 . 05 , ***p<0 . 001 in Fisher's exact tests against WT ( irLOX2 ) or irPI ( hemi-irLOX2 , which also contains the irPI construct ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 008 We removed the cotton swabs and the remaining larvae . We then monitored predation of newly-infested larvae and eggs without GLV supplementation during infestation M3 ( Figure 4 ) . We staggered infestation to accommodate differences in plant growth: WT and irPI seedlings were initially larger and therefore were planted into the field on average 3 days earlier than irLOX2 and hemi-irLOX2 plants , so that all plants were planted at a similar size , which is important for even establishment . We therefore re-infested WT and irPI plants earlier after M2 , to allow irLOX2 and hemi-irLOX2 plants to catch up in their growth to WT and irPI before re-infestation , so as not to bias further assays . However , we left M . sexta larvae on irLOX2 and hemi-irLOX2 as long as on WT and irPI , and we made several control measurements to ensure that differences in Geocoris spp . predation were not due to our staggering of infestation: we counted Geocoris spp . populations around all genotypes over this period ( Table 2 ) and saw that they were not different ( p>0 . 05 in Fisher's exact tests ) , indicating that Geocoris spp . continued to explore irLOX2 and hemi-irLOX2 plants but not to predate from them over a week of infestation followed by 5 days of M . sexta egg predation assays ( during which M . sexta eggs were simultaneously applied to all genotypes ) ; and we followed predation of M . sexta larvae from all four genotypes in parallel over 1 week , during which irLOX2 and hemi-irLOX2 were infested with more larvae than WT and irPI due to sustained higher predation rates on WT and irPI; but predation remained higher on WT and irPI . 10 . 7554/eLife . 00007 . 009Table 2 . Numbers ( N ) of Geocoris spp . individuals ( nymphs and adults ) within 5 cm radii around plants used for predation experiments , counted within half an hour during the main period of Geocoris spp . activityDOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 009ExperimentGenotypeGeocoris spp . per day ( n ) Plants ( n ) Larval predationDatesMay2122May 21–23 , 2011WT3419irPI6624irLOX26420hemi-irLOX28220Total231683Egg predationDatesJune3457June 2–6 , 2011WT252118irPI371521irLOX2420321hemi-irLOX2112224Total101551184Numbers are shown as subtotals for each plant genotype and grand totals per day ( in bold ) . Predation of both larvae and eggs without GLV supplementation was two to four times as great on GLV-emitting WT and irPI plants: 43%/60% ( WT/irPI ) for larvae and 34%/39% for eggs , vs 17%/33% ( irLOX2/hemi-irLOX2 ) for larvae and 9%/20% for eggs ( Fisher's exact tests: N=30 larvae , p=0 . 047 for irLOX2 vs WT , p=0 . 069 for hemi-irLOX2 vs irPI; N=88 eggs , p<0 . 001 for irLOX2 vs WT , p=0 . 013 for hemi-irLOX2 vs irPI ) ( Figure 5 ) . Predation was associated with a steady Geocoris spp . population of 16–23 individuals per day within a 5 cm radius around plants ( Table 2 ) . However , there was no difference among plant genotypes in the number of Geocoris spp . individuals ( p>0 . 05 in Fisher's exact tests ) , indicating that Geocoris spp . regularly survey all plants and use GLVs as a short-distance cue to determine which plants harbor prey . Figure 5 shows larval predation rates at the beginning of the assay , when the M . sexta load was comparable across plant genotypes . Over the following week , Geocoris spp . predated a total of 80% of these larvae from WT and irPI vs 47% from irLOX2 ( Fisher's exact test , p=0 . 015 vs WT ) and 67% from hemi-irLOX2 ( p=0 . 382 vs irPI ) . In summary , Geocoris spp . had the same opportunity to locate M . sexta larvae and eggs on all genotypes , but consistently preferred to predate from GLV-supplemented or -emitting plants . We took the different number of ‘days in field’ for each plant into account in our comparison of growth and reproduction among genotypes and therefore the staggered planting did not affect this comparison ( Figure 6 , statistics Table 3 ) . The irLOX2 and hemi-irLOX2 plants suffered , in total , a similar amount of M . sexta damage to WT plants in trials M2 and M3 ( Figure 4 ) , and only irPI plants suffered significantly less M . sexta damage ( Figure 7 ) . 10 . 7554/eLife . 00007 . 010Figure 6 . Growth and reproduction of plants during the 2010 and 2011 field seasons; graphs show means±SEM . ( A ) Final growth measurements for M . sexta-infested and uninfested control plants of each genotype in 2011 ( left , 44–45 days after planting , N=11–17 ) or M . sexta-infested plants in 2010 ( right , June 6 , 2 days after the removal of fifth-instar M . sexta larvae ) . *p<0 . 05 for Wilks' Lambda test of the effect of M . sexta feeding on growth and reproduction in 2011 , day 44–45 , in a two-way MANOVA with factors genotype and treatment ( F6 , 52=2 . 287 , p=0 . 049 ) . *p-values above individual graphs denote the significance of M . sexta feeding over all genotypes in 2011 for the measurement shown in the MANOVA , or in a separate Mann-Whitney U-test for side branches ( stem F1 , 57=9 . 155; side branches , U = 270; buds F1 , 57=4 . 572 ) ; values for individual genotypes are in Table 3 . a , b , c Different letters denote significant ( p<0 . 05 ) differences between genotypes in 2011 for Scheffe post hoc tests ( rosette diameter F3 , 57=8 . 791 , p<0 . 001 , stem length F3 , 57=4 . 192 , p=0 . 009 , number of buds F3 , 57=9 . 876 , p<0 . 001 ) or Bonferroni-corrected p-values for Mann-Whitney U-tests following a Kruskal-Wallis test ( side branches χ2 = 10 . 958 ) . In 2010 , in the absence of Geocoris spp . activity , there were no significant differences between genotypes in the parameters shown with or without M . sexta infestation ( Table 3 ) . Bud numbers from 2010 are also shown in Figure 9 . ( B and C ) Flower production for M . sexta-infested and uninfested control plants from the beginning of flowering in ( B ) 2011 and ( C ) 2010 . Flowers were counted and removed at the time points shown: each time point represents new flower production . Insets in ( C ) show the first two time points for irPI and hemi-irLOX2 . *p<0 . 05 for the main effect of M . sexta infestation in a repeated-measures ANOVA with log2-transformed data ( Table 3 ) . Raw data for 2011 is in F6AB_SchumanBarthelBaldwin2012growth_reproduction2011 . xlsx and T4_SchumanBarthelBaldwin2012growth_reproduction2011 . xlsx , and data for 2010 is in F6AC_SchumanBarthelBaldwin2012growth_reproduction2010 . xlsx ( Dryad: Schuman et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 01010 . 7554/eLife . 00007 . 011Table 3 . Results of Mann–Whitney U-tests , Kruskal–Wallis tests , and ANOVAs for control vs M . sexta-infested plants of each genotype grown in the field in 2010 and 2011 ( Figures 6 and 9 ) DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 0112010BranchesStem , buds , flowersMann–Whitney , Kruskal–WallisMANOVA , Wilks' lambdaComparisonGenotypedfχ2p*dfFpTreatmentAll10 . 0221 . 0003 , 1480 . 4630 . 709GenotypeAll32 . 9090 . 8029 , 360 . 3441 . 1860 . 303Branches ( n ) Rosette diameter ( cm ) Stem length ( cm ) Buds ( n ) Flowers ( n ) 2011Student's t-testStudent's t-testMANOVA , Wilks' lambdaMANOVA , Wilks' lambdaMANOVA , Wilks' lambdaComparisonGenotypedftpdftpdfFpdfFpdfFpTreatment×timeWT261 . 6960 . 10226−0 . 8700 . 9325 , 223 . 8710 . 0115 , 223 . 1880 . 0263 , 241 . 2130 . 326irPI261 . 0240 . 31526−0 . 1610 . 8735 , 220 . 9910 . 4465 , 220 . 6560 . 6605 , 220 . 5250 . 755irLOX2251 . 1120 . 27725−0 . 0580 . 9545 , 210 . 6060 . 6965 , 210 . 5350 . 7485 , 210 . 5400 . 744hemi-irLOX2221 . 7530 . 094221 . 1400 . 2675 , 181 . 1180 . 3865 , 183 . 0010 . 0384 , 190 . 7230 . 5872010: Numbers of side branches ( Mann–Whitney , Kruskal–Wallis ) , stem length , and final numbers of buds and flowers ( MANOVA ) were recorded in a single measurement at the end of M1 ( Figure 4 ) . Numbers of newly produced flowers were counted repeatedly upon flower removal , and Wilks' Lambda F values for the main effect of M . sexta feeding are shown from repeated-measures ANOVAs across all measurements; Wilks' F values for the M . sexta-by-time interaction were not significant . * Bonferroni-corrected p-values . 2011: Because many plants had few or no side branches before the final measurement , and rosette diameters did not change over the period that plants were measured , t-tests are shown for the final measurement of these parameters in M3 ( Figure 4 ) . For stem lengths , numbers ( n ) of buds , and numbers of flowers , Wilks' lambda F values for the M . sexta-by-time interaction are shown from repeated-measures ANOVAs across all measurements . Significant p-values are given in bold . 10 . 7554/eLife . 00007 . 012Figure 7 . Herbivore damage to plants during the 2010 and 2011 field seasons ( means+SEM ) . For a timeline of Manduca infestations M1–M4 , see Figure 4A . ( A ) Total canopy damage due to naturally occurring herbivores before the start of infestation M1 in 2010 , N=17 . For raw data , see F7A_SchumanBarthelBaldwin2012herbivoreDamage2010 . xlsx ( Dryad: Schuman et al . , 2012 ) . ( B ) Total canopy damage due to naturally occurring herbivores before infestation M2 ( May 5 ) and near the end of M3 ( May 27 ) in 2011 , N=24–28 . a , b Different letters denote significant ( p<0 . 05 ) differences between genotypes in Scheffe post hoc tests following one-way ANOVAs for arcsine-transformed data at each timepoint ( mirids May 27 F3 , 103=5 . 291 , p=0 . 002; noctuids May 27 F3 , 103=3 . 503 , p=0 . 018 ) ; n . s . : not significantly different . #p<0 . 05 for the main effect of genotype on noctuid damage in a Bonferroni-corrected Kruskal-Wallis test , May 5 ( χ2=11 . 239 , p=0 . 027 ) . ( C ) Damage in 2011 from M . sexta larvae used in the predation assays in M2 ( left panel ) and M3 ( right panel ) . GLVs were externally supplemented to plants in infestation M2 and not in M3 . Total canopy damage was estimated , using the index , by an independent observer without knowledge of plant identity ( N=11–17 ) . *p<0 . 05 in a Mann-Whitney U-test between irPI and WT on May 28 ( U=54 , p=0 . 046 ) ; the difference on May 15 was not significant ( p>0 . 1 ) . Note that scales differ . Raw data for ( B ) and ( C ) is in F7BC_SchumanBarthelBaldwin2012herbivoreDamage2011 . xlsx ( Dryad: Schuman et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 012 Reduced predation of M . sexta from irLOX2 and hemi-irLOX2 in trials M2 and M3 in 2011 ( Figure 4 ) correlated with the reduced growth and reproduction of both genotypes , by 30–50% for irLOX2 and 20–30% for hemi-irLOX2 vs WT , although this reduction was also apparent in plants not infested with M . sexta . ( Figures 4 and 6 , statistics Table 3 ) . In 2010 however , in the absence of predation , there was no difference in stem growth , branching , or bud and flower production among genotypes irrespective of M . sexta infestation ( Figure 6 , statistics Table 3 ) . Although M . sexta feeding significantly affected growth and reproduction of plants overall , the effect was not significant for irLOX2 or irPI plants in either year ( Figure 6 , statistics Table 3 ) , possibly due to reduced feeding damage resulting from a lack of TPI-induced compensatory feeding in irPI ( Steppuhn and Baldwin , 2007 ) ( Mann-Whitney U-test between irPI and WT on May 28 , U=54 , p=0 . 046 , Figure 7 ) . Although GLVs are feeding stimulants ( Halitschke et al . , 2004 ) , we could not measure reduced M . sexta feeding damage in hemi-irLOX2 or irLOX2 ( Figure 7 ) . Yet hemi-irLOX2 plants , despite strongly reduced GLVs , still suffered reduced growth and reproduction from Manduca spp . feeding: M . sexta feeding reduced flower production rates by about 50% in WT and by about 30% in hemi-irLOX2 plants in 2010 , although the overall reduction was only significant in WT; and reduced bud production significantly for both WT and hemi-irLOX2 by 25–30% in 2011 ( Figure 6 , statistics Table 3 ) . We monitored herbivore attack to determine whether GLV-silenced plants suffered different amounts of damage from naturally occurring herbivores , which could also cause differences in their growth and reproduction . All genotypes were attacked by mirid ( Tupiocoris notatus ) and noctuid herbivores which caused similar amounts of damage across genotypes and years ( ca . 15% and 3% of total canopy area , respectively ) although irLOX2 plants suffered 60% less mirid and noctuid damage by the end of M3 in 2011 ( N=24–28; Bonferroni-corrected Kruskal-Wallis test , noctuids May 5 , p=0 . 027 , all pairwise tests Bonferroni-corrected p>0 . 05; one-way ANOVAs with factor genotype on arcsine-transformed data: mirids May 27 , F3 , 103=5 . 291 , p=0 . 002 , p<0 . 05 for irLOX2 vs hemi-irLOX2 and irPI in Scheffe post hoc tests , noctuids May 27 , F3 , 103=3 . 503 , p=0 . 018 , all post hoc tests p>0 . 05; all other comparisons p>0 . 05; Figures 4 and 7 ) . Reduced herbivore damage on irLOX2 in 2011 could have increased the growth and reproduction of irLOX2 plants relative to WT , but cannot explain why irLOX2 plants instead displayed reduced growth and reproduction . Plants in 2011 were also damaged by flea beetles and grasshoppers ( <3% of canopy area , Kruskal-Wallis tests , N=24–28 , all comparisons p>0 . 05 , Figure 7 ) . We cannot exclude the possibility that reduced growth and reproduction of uninfested irLOX2 and hemi-irLOX2 plants in 2011 ( Figure 6 , statistics Table 3 ) might have been due to non-herbivory-related factors ( e . g . , differences in root health corresponding to GLV antimicrobial properties ) which did not play a role in 2010 . Because of this uncertainty , we conducted assay M4 ( Figure 4 ) in which plants were carefully matched for size and prior reproduction ( Figure 8 ) , and this experiment is the more robust basis for our argument that GLV-mediated indirect defense increases plant reproduction . 10 . 7554/eLife . 00007 . 013Figure 8 . Comparison of plants used in triplets for infestation M4 in 2011 ( see Figure 4A ) ; graphs show means+SEM ( N=21 plants ) . ( A ) Parameters used to match plants in triplets . Measurements and assessments are from the first day of M4 . ( B ) Final measurement of prior growth and reproduction for plants used for triplets; data are from the final two measurements during infestation M3 ( see Figure 4A ) . a , bDifferent letters denote significant differences ( p<0 . 001 ) for flower number in Scheffe post hoc tests following a MANOVA with all measurements and genotype as the factor ( F2 , 60=8 . 668 , p<0 . 001 ) . ( C ) Health index used in ( A ) . For raw data , see F8_SchumanBarthelBaldwin2012triplets . xlsx ( Dryad: Schuman et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 013 To ensure that the correlated differences we observed in plant reproduction and M . sexta mortality were due to plant GLV emission and not to different timing and amounts of M . sexta damage , and to avoid the influence of non-herbivory-related factors , we conducted a Manduca spp . predation and plant performance assay during infestations M1 in 2010 , and M4 in 2011 ( Figure 4 ) for which all plants used were matched for size , as well as former damage and reproduction as necessary ( Figure 8 ) , and infested simultaneously with Manduca spp . neonates . We hypothesized that the 50% lower predation rates of Manduca spp . from GLV-deficient plants ( Figure 5 ) , combined with Manduca spp . 's negative effect on growth and reproduction ( Figure 6 ) , would result in reduced reproduction for GLV-deficient versus matched GLV-producing plants if Geocoris spp . were present . Homozygous irLOX2 plants were excluded from these 'matched' experiments because they did not suffer reduced growth or reproduction from M . sexta feeding , and because they were too small in comparison to other lines in 2011 ( Figure 6 ) . In both 2010 and 2011 , we selected triplets of WT , irPI and hemi-irLOX2 plants similar in size , reproductive output , apparent health , and prior damage; damage from naturally occurring herbivores did not differ among these genotypes ( Figure 7 ) . In 2010 , matched plants were part of infestation M1 ( Figure 4 ) and thus it was not necessary to control for prior reproduction or M . sexta damage . Plants in 2010 received three lab strain M . sexta larvae per plant to a lower stem leaf . We recorded the mortality of M . sexta larvae and the reproductive output of plants until they began to set unripe seed . No reproductive meristems were removed , but flowers were removed and counted periodically over the first 10 days , as was done during infestation M3 in 2011 ( Figures 4 and 6 , statistics Table 3 ) , to track plant reproduction while avoiding ripe seed capsules: the distribution of ripe seed is not permitted for genetically modified plants . In the absence of Geocoris spp . in 2010 , genotypes did not differ in M . sexta mortality ( N=51 larvae ) —which in every observed case was due to a failure of the larva to feed—or plant reproduction ( N=17 plants ) ( Figure 9 ) . This , and the fact that flower production did not differ among genotypes in 2011 through infestation M3 despite flower removal ( Figure 6 ) , indicates that flower removal itself does not cause a difference among genotypes , and suggests that the other differences among genotypes in growth and reproduction seen in 2011 ( Figure 6 , statistics Table 3 ) are real . In 2011 , hemi-irLOX2 , irPI and WT plants were matched prior to infestation M4 to exclude differences in growth , reproduction and Manduca spp . damage arising during M . sexta infestations M2 and M3 and from caged Manduca spp . during the egg predation assay ( Figures 4 , 7 , and 8 ) . Instead of regularly removing flowers , we removed all reproductive meristems from matched plants in 2011 by cutting inflorescences at their base . This allowed us to follow a new set of reproductive meristems through to seed set without incurring ripe seed . Because plants were matched prior to the assay , a similar number of reproductive meristems were cut from all plants , and thus all plants were similarly affected by this cutting ( Figure 8 , see 'Discussion' ) . Because oviposition by native Manduca spp . moths provided sufficient eggs prior to the beginning of M4 ( Figure 4 ) , we decided to conduct this infestation with wild larvae and thereby demonstrate that native larvae , like larvae of the lab strain , are susceptible to GLV-mediated predation . To make M4 a realistic test , we placed one wild Manduca spp . neonate per plant on a lower stem leaf to mimic natural oviposition rates ( Kessler and Baldwin , 2001 ) . We again recorded the mortality of Manduca spp . larvae and the new reproductive output of plants until they began to set unripe seed . During the first to third larval instars in which larvae are vulnerable to Geocoris spp . predation ( Kessler and Baldwin , 2001 ) , wild Manduca spp . mortality was 38% on hemi-irLOX2 plants vs 62–76% on matched WT and irPI plants; the overall mortality of larvae on all three lines was significantly different ( N=21 larvae , Bonferroni-corrected pairwise comparisons by Friedman tests , p<0 . 01 ) ( Figure 9 ) . Although Manduca spp . mortality on hemi-irLOX2 jumped to 70% in the fourth and semi-final larval instar , this was likely due to predation by whiptail lizards ( Cnemidophorus spp . ) which were present on the field plot: these lizards predate late-instar Manduca and are attracted to short-chain fatty-acid volatiles produced by the larvae due to ingestion of acyl sugars in plant trichomes ( Stork et al . , 2011; Weinhold and Baldwin , 2011 ) . 10 . 7554/eLife . 00007 . 014Figure 9 . Cumulative mortality of Manduca spp . larvae and numbers of reproductive units produced by infested plants in 2010 , in the absence of Geocoris spp . predation , and in 2011 , when Geocoris spp . were active predators of Manduca spp . ( A ) In 2010 , flowering plants matched for size ( N=17 ) were each infested with three M . sexta neonates from a laboratory culture ( N=51 larvae ) , which were allowed to reach the final instar on plants . The upper panel shows larva mortality over time , which reached a maximum of 40% by the fifth instar , after 12 days . Flower production ( lower panel ) did not differ , nor did any other parameters of plant size and reproduction ( Figure 6 , Table 3 ) including number of buds produced by June 6 , which was day 19 after infestation and day 49 after planting in the field . For raw data , see F9A_SchumanBarthelBaldwin2012data2010 . xlsx ( Dryad: Schuman et al . , 2012 ) . ( B ) In 2011 , plants ( N=21 ) were matched for size , prior reproduction , health , and previous damage by Manduca spp . and other herbivores ( Figures 7 and 8 ) following the end of infestation M3 ( Figure 4 ) , and reproductive meristems were removed . Matched plants were infested with one wild Manduca spp . neonate each ( M4 in Figure 4 ) , and Manduca spp . larvae were allowed to reach the fourth ( penultimate ) instar . Larval mortality ( upper panel ) reached a maximum of 76% after larvae transitioned from the second to third instar ( days 9 and 10 ) , at which time larval mortality on hemi-irLOX2 was only half as great as on WT or irPI; larvae beyond this stage are not susceptible to Geocoris spp . ( Kessler and Baldwin , 2001 ) . Flower and bud production ( lower panel ) was twice as great in WT and irPI as in hemi-irLOX2 , and numbers of flowers and buds correspond to numbers of seed capsules: hemi-irLOX2 plants also produced fewer unripe seed capsules than WT or irPI plants . For raw data , see F9B_SchumanBarthelBaldwin2012data2011 . xlsx ( Dryad: Schuman et al . , 2012 ) . a , b , c Different letters indicate significant differences ( p<0 . 01 ) in Bonferroni-corrected pairwise Friedman tests ( Manduca spp . mortality ) , or Scheffe post hoc tests of hemi-irLOX2 versus WT and irPI flowers and buds following a repeated-measures MANOVA over all flower and bud counts shown ( results of Greenhouse-Geisser-corrected univariate tests for the interaction of line and day: buds , F4 . 988 , 149 . 653=5 . 297 , p<0 . 001; flowers , F3 . 722 , 111 . 657=4 . 403 , p=0 . 003 ) , or significant differences ( p<0 . 05 ) in Scheffe post hoc tests following an ANOVA for unripe seed capsules at day 15 with genotype as the factor ( F2 , 60=4 . 142 , P=0 . 021 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 014 The plants used in M4 had not previously differed in their reproduction except that hemi-irLOX2 plants had produced more flowers than WT , but not irPI plants ( Figure 8 ) . By the end of the assay , the hemi-irLOX2 plants had produced 40–50% fewer buds and flowers than matched WT and irPI plants ( N=21 plants , p<0 . 05 in Scheffe post hoc tests for hemi-irLOX2 vs WT and irPI flowers and buds following a repeated-measures MANOVA over all flower and bud counts , Wilks' Lambda for the interaction of line and day: F12 , 110=2 . 835 , p=0 . 002 ) ( Figure 9 ) . This reduced bud and flower production was not due to accelerated seed set: unripe seed capsules on hemi-irLOX2 plants were also reduced by 50% ( N=21 plants , p=0 . 021 for hemi-irLOX2 vs irPI in a Scheffe post hoc test following an ANOVA with genotype as the factor , F2 , 60=4 . 142 , p=0 . 021 ) ( Figure 9 ) . These data demonstrate that herbivore-induced GLV emissions function as indirect defenses by increasing predation of Manduca spp . larvae twofold , resulting in a twofold increase in bud and flower production for N . attenuata in its native habitat . To ensure that our results were not biased by the use of wild Manduca spp . larvae , which comprised both M . sexta and M . quinquemaculata , we analyzed the growth ( length over time ) and instar change of larvae on plants in M4 by larval species . M . sexta and M . quinquemaculata did not differ in their growth or instar progression ( N=11–13 , repeated measures ANOVA for days 4–11 , Wilks' Lambda for the interaction of day and species: F11 , 12=1 . 356 , p=0 . 311 ) . Because larvae of the two species cannot be distinguished before the third instar , we could not test whether mortality was equal for both species in the first three instars; however , because other collections of wild eggs around the same time as the collection for our experiment yielded a 1:1 ratio of species , and because our ratio of the species remained 1:1 after larvae reached the third instar , it is likely that mortality of the two species was equal prior to the third instar . TPIs had a less consistent and , contrary to our expectations , negative effect on Manduca spp . predation ( Figure 5 ) ; furthermore , there was no positive effect of TPIs on plant growth and reproduction ( Figure 6 , statistics Table 3 ) and only a marginal effect of TPIs on Manduca spp . growth under natural conditions ( N=13–26 second instar larvae during M2 , one-way ANOVAs with genotype as the factor , F3 , 77=2 . 792 , p=0 . 046 , all post hoc tests p>0 . 05 , Figure 10; N=8 second instar larvae during M4 , paired t-test between matched WT and irPI , p=0 . 052 ) . We hypothesized that the reduced access to protein for larvae feeding on TPI-producing plants might nevertheless affect Manduca spp . behavior independently of larval size . Indeed , wild Manduca spp . larvae feeding on WT plants ( infestation M4 , Figure 4 ) reacted more sluggishly to experimental provocation than size-matched larvae on irPI plants: they were 75% less likely to attack when lifted off of the leaf ( N=5 second-instar larvae matched for size , p=0 . 035 in paired t-test ) ( Figure 10 , Videos 1 and 2 ) . 10 . 7554/eLife . 00007 . 015Figure 10 . Mock Geocoris spp . predation assays with Manduca spp . larvae fed on WT or irPI plants . ( A ) Response of wild Manduca spp . ( Figure 4A ) on plants in the field to poking with a toothpick and lifting with a featherweight forceps ( N=5 second-instar larvae matched for size ) . We first poked larvae below the horn three times , 3 s apart , with the end of a toothpick and counted how often they attacked the toothpick , defined as the larva whipping its head around toward the toothpick and making contact . We then lifted larvae from the plant using the forceps and counted how often they attempted to attack , or succeeded in attacking the forceps over 15 s . In an attempted attack , the larvae moved from hanging at a 180° angle below the forceps vertically toward the forceps; and in a successful attack , the front end of the larva made contact with the forceps , before returning to its original position . All individuals were recorded and responses were counted from videos ( see Videos 1 and 2 ) . *p<0 . 05 in a paired t-test . ( B ) Left , response of M . sexta from a laboratory strain raised for 48 hr in boxes on either WT or irPI leaf tissue ( N=20 first-instar larvae matched for size ) to being poked , pierced and lifted with an insect pin . Right , growth of larvae in the following 24 hr . The procedure was identical to that for the on-plant assay described above , except that larvae were poked with an insect pin rather than a toothpick , and then pierced in the rear flank and lifted with the same insect pin ( see Videos 3 and 4 ) . *p<0 . 05 in a paired t-test . The length of each larva was measured prior to poking and lifting . Afterward , larvae were placed in individual cups , each with a moist paper towel round and fresh WT or irPI leaf tissue , and length of the larvae in millimeters was again measured after 24 hr; mortality did not differ between WT- and irPI-fed larvae . *p<0 . 05 in a Student's t-test . ( C ) Upper panel , length of first instar larvae fed for 2 days on WT or irPI tissue and size-matched for use in the off-plant behavioral assay mimicking Geocoris attack ( B ) ; lower panel , mortality of first instar larvae 24 hr after mock Geocoris attack as described in ( B ) . Mortality was not significantly different in a Fisher's exact test . ( D ) Larval length in the first instar after 2 days on plants in the field: larvae on irPI were not significantly larger . Length of surviving larvae was measured in a predation assay during infestation M3 ( Figures 4 and 5C ) , N=13–26 larvae . Length was not significantly different for larvae feeding on irPI in a one-way ANOVA with genotype as the factor ( F3 , 77=2 . 792 , p=0 . 046 , all post-hoc tests p>0 . 05 ) . For raw data , see F10_SchumanBarthelBaldwin2012Manduca . xlsx ( Dryad: Schuman et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 01510 . 7554/eLife . 00007 . 016Video 1 . On-plant assay , plant 7u , WT , June 18 , 2011 . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 01610 . 7554/eLife . 00007 . 017Video 2 . On-plant assay , plant 2o , irPI , June 18 , 2011 . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 017 We were careful not to harm wild larvae so that we could monitor their natural mortality and consequences for plant reproduction ( Figure 9 ) . To more accurately imitate Geocoris spp . attack , we developed an off-plant assay with larvae from the laboratory M . sexta strain feeding on detached leaves from field-grown plants , in which size-matched larvae were poked , pierced and lifted using an insect pin to mimic the Geocoris spp . beak ( Figure 10 , Videos 3 and 4 ) . Similarly to the on-plant assay , larvae fed on WT leaves were 50% less likely to successfully attack the insect pin , either when initially poked , or poked and lifted with the pin ( N=20 first-instar larvae matched for size , p<0 . 05 in paired t-tests ) ( Figure 10 ) . We also monitored recovery post-trial and found that WT-fed larvae ceased to grow for at least 24 hr after simulated attack , while irPI-fed larvae continued to grow ( p=0 . 016 in Student's t-test ) ; mortality did not differ ( p=0 . 527 in Fisher's exact test ) ( Figure 10 ) . 10 . 7554/eLife . 00007 . 018Video 3 . Off-plant assay , replicate 3 , WT , June 24 , 2011 . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 01810 . 7554/eLife . 00007 . 019Video 4 . Off-plant assay , replicate 3 , irPI , June 24 , 2011 . DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 019 Thus TPIs did not increase plant reproduction under attack from Manduca spp . in nature , but may support indirect defense by weakening the response of larvae to predator attack . The contradictory higher predation rates of Manduca spp . larvae from irPI than from WT plants ( Figure 5 ) might reflect Geocoris spp . 's feeding preference , if irPI-fed larvae are more nutritious than WT-fed larvae ( Kaplan and Thaler , 2011 ) .
The hemi-irLOX2 plants used in this study were created by crossing homozygous irPI and irLOX2 plants , but the irPI construct was not active in the hemizygous state ( Figure 1 ) . We continued to use this cross for its less severely reduced GLV production in comparison to homozygous irLOX2 plants ( Figures 2 and 3 ) , which likely permitted growth and reproduction comparable to irPI and WT in 2011 that was essential for plant matching prior to the final assays of Manduca spp . mortality and plant reproduction ( M4 , Figures 4 and 9 ) . It is common molecular biology knowledge that functional RNAi constructs may be rendered ineffective as a result of insufficient gene dosage , for example , Travella et al . ( 2006 ) and references therein , which may occur when an RNAi construct is present in the hemizygous state ( García-Pérez et al . , 2004 ) . The 35S promoter which drives the transcription of the RNAi construct may also be methylated: an epigenetic effect which can reduce the dose of RNAi in individual plants within a single transformed line ( A Weinhold , unpublished data ) . This may have occurred in the irPI parent used for the creation of the hemi-irLOX2 line , although loss of activity of the irPI construct was not observed in the homozygous irPI line over the lifetime of plants in the field ( Figure 1 ) . Although the production of viable offspring is the accepted definition of Darwinian fitness , we are not permitted to allow transgenic plants to disperse ripe seed in the field , and measures to prevent seed dispersal , such as bagging meristems , strongly affect production of buds and flowers and can also affect seed viability by increasing temperature , and decreasing respiration and photosynthesis of reproductive tissue and associated green tissue . For field-grown N . attenuata plants , fewer than 5% of buds and flowers ( in total ) are aborted by healthy ( not diseased ) plants , and abortion seems always to be due to damage by insects ( M C Schuman and I T Baldwin , personal observation , June 2010 ) . Plants are self-compatible and more than 70% of seed set from plants in native populations results from fertilization via self-pollen ( Sime and Baldwin , 2003 ) . Thus numbers of buds and flowers correlate to lifetime seed capsule production , which in turn correlates to lifetime seed production , which has been used as a proxy measure of Darwinian fitness ( Baldwin , 1998; van Loon et al . , 2000; Hoballah and Turlings , 2001 ) . The transgenic lines used do not vary in seed mass or their seedling viability under laboratory conditions . During the 2011 experiment , we saw a large and reproducible difference in predation from GLV-emitting versus GLV-deficient plant genotypes , which was not observed in 2010 due to an absence of Geocoris spp . predators in that year . This difference in predation rate correlated to a difference in plant growth and reproduction which was also not observed in 2010 . To rigorously test the consequences of GLV-mediated predation of Manduca spp . on plant reproduction , we selected triplets of WT , irPI and hemi-irLOX2 plants similar in size , previous reproductive output , apparent health , and prior damage to carry out Manduca spp . mortality and plant reproduction assays ( M1 in 2010 , M4 in 2011 , Figure 4 ) . We removed all reproductive meristems from matched plants in 2011 to allow us to follow plant reproduction over full Manduca spp . larval development without incurring ripe transgenic seed capsules . In 2010 ( and during infestation M3 in 2011 , Figure 4 ) , we had removed and counted flowers regularly to track reproduction while avoiding ripe seed; this did not cause a difference in reproduction among genotypes ( Figure 6 ) , but we elected to avoid flower removal during infestation M4 by removing reproductive meristems prior to the beginning of the assay . The hemi-irLOX2 plants chosen in 2011 had produced more flowers than WT – but not more than irPI—prior to the start of infestation M4 ( Figures 4 and 8 ) . This did not correspond to more cuts on average for hemi-irLOX2 when removing reproductive meristems: meristems were cut at the bases of inflorescences which contained mostly buds , and the number of these did not differ for the plants chosen , nor did the number of side branches ( Figure 8 ) which bore most reproductive meristems . Therefore , in the absence of additional effects during infestation M4 , the reproduction of the matched hemi-irLOX2 plants should have been similar to that of WT and irPI . By indicating the long-sought indirect defensive function of HIPVs , these data set the stage for the use of HIPVs as part of integrated pest management strategies ( IPM ) , which rely in part on recruiting biological control agents to reduce pesticide use ( Horne and Page , 2008 ) . These agents are usually naturally occurring generalist parasitoids and predators , such as Geocoris spp . ( Eubanks and Denno , 1999 , 2000; Allison and Hare , 2009; Allmann and Baldwin , 2010 ) . HIPVs are produced by genotypes of most , if not all crop plants and IPM would benefit from selective breeding or engineering of HIPV emission ( Kos et al . , 2009 ) rather than relying on alternatives such as controlled release dispensers , which have mixed success and require large amounts of synthetic HIPVs ( Kaplan , 2012 ) . PIs may be employed to enhance the efficiency of indirect defense , especially combined with toxins like Bt that directly target herbivores and are safe for biological control agents . With growing concerns about field-evolved Bt resistance ( Liu et al . , 2010 ) , indirect defenses promise an effective 'first line of defense' against agricultural pests , to which not even specialist herbivores are likely to rapidly evolve resistance .
Seed germination , glasshouse growth conditions , and the Agrobacterium tumefaciens ( strain LBA 4404 ) –mediated transformation procedure have been described previously ( Krügel et al . , 2002 ) . Seeds of the 31st generation of the inbred 'UT' line of Nicotiana attenuata ( Torr . ex S . Wats . ) were used as the wild-type plant in all experiments . For the field experiment , seedlings were transferred to 50 mm peat pellets ( Jiffy ) 15 days after germination and gradually hardened to the environmental conditions of high sunlight and low relative humidity over 10 days . Small , adapted , size-matched rosette-stage plants were transplanted into a field plot in a native habitat in Utah and watered thoroughly once at planting and as needed over the first 2 weeks until roots were established; all plants received the same watering regime in each year . WT , irPI , irLOX2 and hemi-irLOX2 plants were arranged in quadruplets ( N=40–50 ) of one plant per genotype , with individuals 0 . 5 m apart , a distance sufficient to allow predators and herbivores to distinguish volatiles from neighboring plants ( Kessler and Baldwin , 2001 ) . Quadruplets were arranged so that no two adjacent plants were of the same genotype ( Figure 4 ) . In 2010 , the field plot was a first-year plot located at latitude 37 . 141 , longitude 114 . 027; in 2011 , plants were planted at a second , older field site across a river from the first , located at latitude 37 . 146 , longitude 114 . 020 . Field plantations were conducted under APHIS permission numbers 06-242-3r-a3 ( 2010 and 2011 ) and 10-349-102r ( 2011 ) . We used previously characterized , homozygous , inverted-repeat ( ir ) RNAi transformed lines of the second transformed generation ( T2 ) to silence GLV biosynthesis: irLOX2 line number A-04-52-2 ( Allmann et al . , 2010 ) , and TPI activity: irPI line number A-04-186-1 ( Steppuhn and Baldwin , 2007 ) . Vector construction and the pSOL3 plasmid have been described previously ( Bubner et al . , 2006 ) . A cross was created between irLOX2 and irPI homozygous lines; however , the hemizygous irPI construct did not silence TPI activity or transcripts , and these plants therefor served as vector controls for comparison with irPI and had slightly greater residual GLV production than irLOX2 ( see 'Results' ) . They are thus referred to as hemizygous ( hemi- ) irLOX2 plants . Wild Manduca spp . eggs were collected for field assays when available from natural ovipositions . M . sexta and M . quinquemaculata ( hereafter Manduca spp . ) were both ovipositing at the time experiments were conducted; the species of larvae was identified at the third instar and recorded ( earlier instars of these two species cannot be distinguished morphologically ) . M . sexta and M . quinquemaculata oral secretions ( OS ) are highly similar in their composition ( Halitschke et al . , 2001 ) and elicit similar volatiles ( Halitschke et al . , 2001; Kessler and Baldwin , 2001 ) and defense genes ( Schittko et al . , 2001 ) in N . attenuata . Eggs from laboratory-reared M . sexta , kindly provided by Dr . Carol Miles at SUNY Binghampton , were used in the field when wild Manduca spp . eggs were not sufficiently abundant . Eggs were allowed to hatch in well-aerated boxes on fresh N . attenuata leaf tissue over a moistened paper towel . M . sexta larvae used to elicit glasshouse-grown plants , or to collect oral secretions ( OS ) for plant treatments , were taken from an in-house colony at the Max Planck Institute for Chemical Ecology in Jena , Germany . Because GLVs influence Manduca spp . oviposition ( De Moraes et al . , 2001; Kessler and Baldwin , 2001; Fraser et al . , 2003 ) , and the timing and extent of Manduca spp . oviposition varies from year to year , we created even , synchronous oviposition events by infesting plants with Manduca spp . larvae , either from a lab-reared culture or from wild collections ( see Manduca spp . eggs and larvae ) . Larvae used for plant infestations were placed as neonates on a rosette or lower stem leaf at a standardized position for each assay , and monitored mornings and evenings , during times outside of the main period of Geocoris spp . activity that occurs at midday . Plants in field experiments were either infested with Manduca spp . larvae as described above , or left uninfested ( control ) . There were four infestations over both years of the experiment , denoted M1-M4 in Figure 4 . For measuring headspace GLVs in the field and for glasshouse assays , plants were treated with wounding and M . sexta OS ( W+OS ) as a standardized method to mimic Manduca spp . feeding . Pure OS collected from fourth to fifth instar M . sexta larvae from the Jena colony fed on WT plants was diluted 1:5 with distilled water before use; even 1000-fold diluted OS is still sufficient to cause most OS-elicited responses ( Schittko et al . , 2000 ) . For field-grown plants , a similar , mature , non-senescent leaf was chosen from each plant; for glasshouse-grown plants , the two adjacent older leaves ( nodes +1 , +2 ) to the leaf undergoing a source-sink transition ( node 0 ) on rosette-stage plants were used for PI and LOX2 transcript quantification , and the +2 node of a separate set of bolting plants was used for measuring headspace volatiles . The leaf chosen for treatment was wounded by using a fabric pattern wheel run over the adaxial surface to make six rows of holes in the lamina , three rows on either side of the midvein . 20 µL of 1:5 diluted OS were deposited on the adaxial surface and gently rubbed over the holes with a gloved finger . Control plants were left untreated . For field-grown plants and glasshouse-grown M . sexta-fed plants , a similar , mature , non-senescent systemic ( undamaged ) leaf was chosen from each plant; for glasshouse-grown plants used to measure PI and LOX2 transcripts , the leaves at nodes +1 and +2 ( treated leaf positions ) were harvested . Leaves were cut at the petiole and wrapped in a double layer of aluminum foil . In the field , harvested leaves were immediately frozen on dry ice insulated with ice packs frozen at −20°C; samples were stored at −20°C until transport to Jena on dry ice , where they were kept at −80°C until analysis . Leaves harvested from glasshouse-grown plants were flash-frozen in liquid nitrogen and kept at −80°C until analysis . All sample processing was carried out over liquid nitrogen until the addition of the extraction solvents . Prior to analysis , entire leaves were ground with a mortar and pestle and transferred to a 2 mL microcentrifuge tube for storage . For specific measurements , aliquots were weighed into microcentrifuge tubes containing two steel balls and finely ground in a GenoGrinder ( SPEX Certi Prep ) prior to extraction . TPI activity was quantified in 100 mg of tissue from systemic leaves on Manduca spp . -infested plants using a radial diffusion assay as previously described ( van Dam et al . , 2001 ) . Leaf samples were from control plants or plants treated with W+OS . Treated leaf positions were harvested at the peak of transcript accumulation for PI , 12 hr ( Wu et al . , 2006 ) and LOX2 , 14 hr ( Allmann et al . , 2010 ) . Total RNA was extracted from leaves using the TRIzol reagent ( Invitrogen ) , and a 0 . 5 μg aliquot of total RNA of each sample was reverse-transcribed using oligo ( dT ) 18 and RevertAid H Minus reverse transcriptase ( Fermentas ) following the manufacturer's instructions . Quantitative real-time PCR ( qPCR ) was performed with a Mx3005P Multiplex qPCR system ( Stratagene ) and the qPCR Core kit for SYBR Green I ( Eurogentec ) . Transcripts were quantified using external standard curves for each gene . Elongation factor 1A ( EF1A ) transcript abundance in each sample was used to normalize total cDNA concentration variations . Samples of RNA used to make cDNA were pooled to the same dilution as in cDNA samples and run alongside cDNA in all qPCRs to control for gDNA contamination; no contamination was detected . The sequences of primers used for qPCR ( Kallenbach et al . , 2010; Fragoso et al . , 2011 ) are provided in Table 4 . 10 . 7554/eLife . 00007 . 020Table 4 . Primers used for quantitative PCR ( SYBR Green ) DOI: http://dx . doi . org/10 . 7554/eLife . 00007 . 020GeneForward primer sequence ( 5′-3′ ) Reverse primer sequence ( 5′-3′ ) CitationPITCAGGAGATAGTAAATATGGATCTGCATGTTCCACATTGCFragoso et al . ( 2011 ) LOX2TTGCACTTGGTGTTTGAGATGGTTTAGTAGAAAATGAGCACCACAAKallenbach et al . ( 2010 ) To assess qualitatively the GLV pools in leaf tissue from field-grown plants , and to determine appropriate amounts of leaf tissue and internal standard ( IS ) for GLV extraction , we extracted pooled samples from leaves collected June 6 , 2011 , from M . sexta-infested plants during infestation M3 ( Figure 4 ) . Each sample was pooled from all leaves collected from one genotype . Hexane ( 300 µL ) was added to 100 mg tissue spiked with 3 µg tetralin as an internal standard ( IS ) and incubated by rotating at RT overnight . Samples were allowed to settle and 100 µL of water- and tissue-free hexane was transferred to a GC vial containing a 250 µL microinsert . Individual analytes were analyzed by a Varian CP-3800 GC-Saturn 4000 ion trap MS connected to a ZB5 column ( 30 m×0 . 25 mm i . d . , 0 . 25 μm film thickness; Phenomenex ) . 1 μL of samples was injected by a CP-8400 autoinjector ( Varian ) onto the column with a 1:10 split ratio; the injector was returned to a 1:70 split ratio from 2 min after injection through the end of each run . The GC was programmed as follows: injector held at 250°C , initial column temperature at 40°C held for 5 min , then ramped at 5°C/min to 185°C and finally at 30°C/min to 300°C , held for 0 . 17 min . Helium carrier gas was used and the column flow set to 1 mL/min . Compounds eluted from the GC column were transferred to the MS for analysis . The MS was programmed as follows: transfer line at 250°C , trap temperature 110°C , manifold temperature 50°C , source heater 200°C and scan range from 40 to 399 m/z at 1 . 33 spectra per second as previously described ( Schuman et al . , 2009 ) . The identification of compounds was conducted by GC retention time compared to pure standards and mass spectra compared to standards and mass spectra databases , Wiley version 6 ( Wiley ) and NIST ( National Institute of Standards and Technology ) spectral libraries . For the quantification of GLV pools in leaf tissue from field-grown plants , the hexane extraction protocol was adjusted based on GC-MS results from pooled samples ( described above ) , and a GC-FID with a wax column was used for the quantitative analysis of extracts . Hexane ( 300 µL ) were added to 50 mg tissue ( N=10 ) spiked with 15 µg ( Z ) -hex-3-enyl acetate , a GLV not found in GC-MS samples , as an IS . The extraction proceeded as described above . Analytes were separated by Varian CP-3800 GC-FID connected to a ZB-Wax column ( 30 m×0 . 25 mm i . d . , 0 . 25 μm film thickness; Phenomenex ) . 1 μL of samples was injected by a CP-8400 autoinjector ( Varian ) onto the column in a splitless mode; the injector was returned to a 1:70 split ratio 2 min after injection through the end of each run . The GC was programmed as follows: injector held at 230°C , initial column temperature at 40°C held for 7 min , then ramped at 5°C/min to 115°C and finally at 30°C/min to 250°C , held for 0 . 5 min . Helium carrier gas was used and the column flow set to 1 mL/min . Compounds eluted from the GC column were transferred to a Varian FID set at 250°C for analysis ( airflow 300 mL/min , hydrogen 30 mL/min , nitrogen make-up gas 5 mL/min ) . Individual volatile compound peaks were quantified by peak areas using MS Work Station Method Builder and Batch Report software ( Varian ) and normalized to the peak area of the IS ( Z ) -hex-3-enyl acetate in each sample . Peak identification and quantification was done by comparison to standard curves of pure compounds in hexane . Compounds present in quantifiable amounts were ( Z ) -hex-3-en-1-ol , ( E ) -hex-2-enal and the IS ( Z ) -hex-3-enyl acetate . For measurement of GLVs in the headspace of field-grown plants , intact leaves were harvested ( N=3 ) and kept fresh by placing the petioles in microcentrifuge tubes filled with water . Immediately before each measurement , one leaf was treated with W+OS , and a 1 cm2 disc was stamped out and placed in a 4 mL GC vial . After 15 min , the headspace in the vial was measured with a ZNose 4200 portable gas chromatograph with a 1 m DB5 column ( Electronic Sensor Technology , Newbury Park , CA , USA ) by inserting the ZNose inlet needle through the septum of the GC vial into the headspace . The program was as follows: valve set at 165°C , inlet at 200°C , trap at 250°C; 30 s sampling time , column ramped from 30°C to 190°C at 4°C/s , data collection for 20 s . Genotypes were analyzed in an alternating order within each replicate: first replicate 1 of all genotypes , then replicate 2 , then replicate 3 . Retention times of GLV aldehydes and alcohols , the most abundant GLV headspace components , were determined using pure standards . For the analysis of GLVs in the headspace of glasshouse-grown plants , the +2 leaf was enclosed immediately after W+OS elicitation in a food-quality 50 mL plastic container ( Huhtamaki ) with an activated charcoal filter attached to one side for incoming air , and connected to self-packed Poropak Q filters containing 20 mg of Poropak ( Sigma-Aldrich ) packed with silanized glass wool and Teflon tubing in the column bodies ( ARS , Inc . ) as previously described ( Halitschke et al . , 2000; Schuman et al . , 2009 ) . Ambient air was pulled by vacuum pump for 3 hr through an activated charcoal filter , over the leaf in the trapping container , and through a Poropak Q filter connected by PVC tubing ( Rotabilo ) to a custom-made valve manifold , as previously described ( Schuman et al . , 2009 ) ; the manifold was adjusted such that flow rates through traps were ca . 300 mL/min . After trapping , sampled leaves were excised at the base of the petiole , scanned , and the leaf area was measured in comparison to a 1 cm2 standard ( SigmaScan 5 . 0; Systat Software Inc . ) for normalization of volatile emission to cm2 leaf area . Poropak Q filters were wrapped in aluminum foil and stored at −20°C until elution of volatiles with 250 μL dichloromethane ( Sigma-Aldrich ) . Immediately prior to elution , each filter was spiked with 320 ng of tetralin internal standard ( IS ) in hexane ( Sigma-Aldrich ) . Filters were eluted into a GC vial containing a 250 μL glass insert . Samples were analyzed by a CP-3800 GC Varian Saturn 2000 ion trap MS ( Varian ) connected to a polar ZB-wax column ( 30 m×0 . 25 mm i . d . , 0 . 25 μm film thickness; Phenomenex ) . 1 μL of samples was injected by a CP-8200 autoinjector ( Varian ) onto the column in a splitless mode; the injector was returned to a 1:70 split ratio from 2 min after injection through the end of each run . The GC was programmed as follows: injector held at 230°C , initial column temperature at 40°C held for 3 min , then ramped at 5°C/min to 180°C and finally at 10°C/min to 240°C , held for 1 min . Helium carrier gas was used and the column flow set to 1 mL/min . Eluted compounds from the GC column were transferred to the MS for analysis . The MS was programmed as follows: transfer line at 230°C , trap temperature 150°C , manifold temperature 80°C and scan range from 40 to 399 m/z at 1 . 33 spectra per second as previously described ( Schuman et al . , 2009 ) . Individual volatile compound peaks were quantified by peak areas of two specific and abundant ion traces per compound using MS Work Station Data Analysis software ( Varian ) and normalized by the 104+132 ion trace peak area of the IS ( tetralin ) in each sample . The identification of compounds was conducted by GC retention time compared to pure standards and mass spectra compared to standards and mass spectra databases , Wiley version 6 ( Wiley ) and NIST ( National Institute of Standards and Technology ) spectral libraries . In 3 hr headspace samples we detected ( Z ) -hex-3-en-1-ol , ( Z ) -hex-3-en-1-ol , ( E ) -hex-2-en-1-ol ( forms from ( E ) -hex-2-enal on filters over trapping periods longer than 20 min ) , ( Z ) -hex-3-enyl acetate , ( Z ) -hex-3-enyl butanoate , ( Z ) -hex-3-enyl isobutyrate , and ( Z ) -hex-3-enyl propanoate . The collection of ( E ) -α-bergamotene from the headspace of glasshouse-grown plants and its extraction from Poropak Q filters was carried out as for GLVs , except that ( E ) -α-bergamotene was collected 24–32 hr after W+OS treatment of the leaf . Eluted samples were analyzed by an HP 6890 GC-5973 quadropole MS ( Hewlett-Packard ) connected to a nonpolar DB-5ms column ( 30 m×0 . 25 mm i . d . , 0 . 25 μm film thickness; Agilent ) . 1 μL of samples was injected by a HP 7683 autoinjector ( Hewlett-Packard ) onto column in a splitless mode; the injector was purged at 50 mL/min 1 . 5 min after injection and switched to gas saver mode ( 20 mL/min ) from 10 min through the end of each run . The GC was programmed as follows: injector held at 230°C , initial column temperature at 40°C held for 2 min , then ramped at 5°C/min to 165°C and finally at 60°C/min to 300°C , held for 2 min . Helium carrier gas was used and the column flow set to 2 mL/min . Eluted compounds from GC column were transferred to the MS for analysis . The MS was programmed as follows: source at 230°C , quad temperature 150°C , and scan range from 33 to 350 m/z at 4 . 49 spectra per second . ( E ) -α-Bergamotene was quantified by peak area using the ion trace 119 m/z in Chemstation software ( Agilent ) and normalized by the 104 ion trace peak area of the IS ( tetralin ) in each sample . The identification of ( E ) -α-Bergamotene and tetralin IS was conducted by GC retention time and mass spectra compared to mass spectra of known standards as previously described ( Schuman et al . , 2009 ) . One or two larvae were placed on plants at a time for each assay , depending on the number available , and were equally distributed among plants as described under ‘Manduca spp . infestation and W+OS treatment of plants' . However , for infestation M3 ( Figure 4 ) , we staggered the infestation of different plant genotypes to accommodate differences in plant growth: WT and irPI plants were initially larger and therefore went into the field on average earlier than irLOX2 and hemi-irLOX2 plants , so that all plants were planted at a similar size , which is important for even establishment . We therefore re-infested WT and irPI plants earlier after M1 , to allow irLOX2 and hemi-irLOX2 plants to catch up in their growth to WT and irPI before re-infestation . However , we then left M . sexta larvae on irLOX2 and hemi-irLOX2 as long as on WT and irPI , and we used a combination of Geocoris spp . counts and additional predation assays to make sure that differences in Geocoris spp . predation were not due to this staggering of infestation ( see 'Results: Geocoris spp . preferentially predate from GLV-perfumed or -emitting plants' ) . Manduca spp . behavior , predation , and growth assays were conducted with first- and second-instar larvae , except infestations M1 2010 and M4 in 2011 , in which larvae were reared from the first through fifth instars on plants; and egg predation assays , in which M . sexta eggs were used . Larvae used in the off-plant mock predation assays were hatched on the appropriate N . attenuata genotype ( WT or irPI ) and hatching was monitored three times per day ( morning , noon , evening ) so that the mock predation assay could be timed to 48 hr after larvae hatched . A protocol of the mock predation assays is given in Figure 10 , and Videos 1–4 depict on-plant ( 1 and 2 ) and off-plant ( 3 and 4 ) behavioral assays . Larvae for off-plant mock predation assays were kept in aerated plastic boxes on cut leaves over moist paper towels . Leaves were refreshed twice daily and were kept fresh by placing the petioles in water in 1 . 5 mL microcentrifuge tubes which were closed around the petiole with Parafilm ( Pechiney Plastic Packaging Company ) . Larval growth was measured as increases in body length ( in millimeters ) using calipers or a small , flexible , transparent plastic ruler . Predation rates were recorded for larvae placed on plants as described above , or for two eggs per plant fixed with droplets of α-cellulose glue ( Kessler and Baldwin , 2001 ) to the underside of a rosette or lower stem leaf at a standardized position . For egg predation assays , a wild Manduca spp . larva was enclosed on a nearby leaf to ensure continual GLV emission: a clip cage was closed around the larva to make it inaccessible to Geocoris spp . predators . Predation was monitored mornings and evenings . Larvae were considered to be predated when either the larva was missing over multiple days , but clear Manduca spp . feeding damage was present , or when the predated larval carcass was found ( Figure 5 ) . Mortality was defined as the total number of missing larvae . Eggs were considered predated when the eggshell was empty but intact except for a small hole which characterizes the typical damage caused by Geocoris spp . feeding; eggs occasionally collapse during Geocoris spp . predation , but collapsed eggs were not counted unless the eggs were mostly or fully empty and with a visible hole ( Figure 5 ) . During infestation M2 , GLVs were added back to irLOX2 and hemi-irLOX2 headspaces by placing a cotton swab adjacent to the M . sexta-infested leaf and adding ca . 20 µL of lanolin paste , measured with a seed spoon , containing a mix of pure GLVs representative of the M . sexta-fed headspace and dissolved in hexane ( Table 1 ) ( Allmann and Baldwin , 2010 ) to the cotton swab . Cotton swabs bearing ca . 20 µL of lanolin paste with hexane as a control were placed next to M . sexta-infested leaves of WT and irPI plants . Lanolin pastes were regularly refreshed by adding 20 µL in the early afternoon and in the morning . Placing GLVs next to , rather than on the leaf ensured that the supplemented headspace would not be altered by plant metabolism , and that we could terminate the supplementation by removing the cotton swabs . Field plots were monitored daily for Geocoris spp . presence during the experiments in 2010 and 2011; both G . pallens and G . punctipes were present in 2011 , but most individuals observed on and around plants were G . pallens . Soon after the first Geocoris spp . sightings in May 2011 ( before the first infestation , M2 ) , Geocoris spp . populations in the immediate vicinity of experimental plants were monitored every 2–3 days by counting individuals . Counts were conducted during the main period of Geocoris spp . activity in the early afternoon , by at least two observers in parallel , in order to complete the count around all Manduca spp . -infested and control plants within 20–30 min . Each observer proceeded by looking at a focal plant and its immediate vicinity for 15 s and then quickly inspecting the rosette leaves; all Geocoris spp . adults and nymphs seen on , under , or within 5 cm around the rosette of the plant during this time were counted . Observers moved in synchrony with each other from one end of the field plot to the other , in this way counting predators around plants which had not yet been disturbed . Plant size ( rosette diameter , stem length and branching ) was monitored at the end of infestation M1 in 2010 and from the beginning of infestation M2 in 2011 ( Figure 4 ) : rosette diameter was measured as the maximum diameter found by gently laying a ruler over the rosette; stem length was measured from the base of the stem to the tip of the apical inflorescence by placing a ruler beside the stem; and all side branches 5 cm or longer were counted . Reproductive output was monitored by counting the number of closed flowers removed every 2–3 days ( before they opened ) from the beginning of flowering , and by counting numbers of closed flowers and buds 2 mm or larger at the end of infestation M1 in 2010 , and during all infestations in 2011 . All growth and reproduction data were analyzed for differences in control versus M . sexta-infested plants within each genotype ( statistics Table 3 ) . Because size-matched plants had been planted over one week in 2011 , growth and reproduction data from plants in 2011 were organized by the number of days since planting for comparison between genotypes ( Figure 6 ) . Flowers were removed and counted periodically over the first 10 days of infestation M1 and during infestation M3 ( Figures 4 and 6 , statistics Table 3 ) , in order to track plant reproduction while avoiding ripe seed capsules: the distribution of ripe seed is not permitted for genetically modified plants . For infestation M4 in 2011 ( Figure 4 ) , instead of regularly removing flowers , we removed all reproductive meristems by cutting inflorescences , which contained mostly buds , at their base , so that we would be able to follow a new set of reproductive meristems through to seed set without incurring ripe seed . Because plants were matched prior to M4 and had the same number of buds and of side branches ( which usually terminate in inflorescences ) ( Figure 8 , see 'Discussion' ) , all plants were similarly affected by the removal of reproductive meristems . Photographs were taken of entire plants and M . sexta-damaged leaves during infestations M2 and M3 in 2011 . Damage caused by M . sexta larvae was rated from photographs by an independent observer with no knowledge of plant identity . Total percent canopy damage due to M . sexta was rated as 1 , 2 , 3 , or 4 using the damage index in Figure 7 . We monitored herbivore attack to determine whether GLV-silenced plants suffered different amounts of herbivore damage , which could influence fitness measurements . The naturally occurring herbivore community on plants in 2010 and 2011 comprised mirids ( Tupiocoris notatus ) and noctuid larvae; in 2011 grasshoppers ( Trimerotropis spp . ) and flea beetles ( Epitrix spp . ) were also present . Total canopy damage due to herbivores occurring naturally on the field plot was quantified prior to the infestations in 2010 and 2011 and again during infestation M3 in 2011 . Damage was calculated by identifying damage from specific herbivores according to their characteristic feeding patterns , counting the number of leaves per plant ( small leaves were counted as 1/5 to 1/2 of a leaf based on leaf area and large leaves were counted as 1 leaf ) , estimating the total percentage of leaf area damage due to each herbivore , and dividing the total leaf area damage from each herbivore by the total number of leaves . Leaf area damage was estimated in categories of 1% , 5% , 10% , 15% , and so on , in steps of 5% . All such damage estimates were made by MCS or KB , who first practiced quantifying damage together until they consistently arrived at the same numbers . As part of matching plants prior to infestation M4 in 2011 , plant health was rated on a scale of 1 ( dead ) to 5 ( healthy ) using the index in Figure 7 . Fisher's exact tests were conducted using a macro ( J H Macdonald , http://udel . edu/~mcdonald/statfishers . html ) for Excel ( Microsoft ) . All other statistical analyses were conducted with SPSS 17 . 0 ( IBM ) . Count data were analyzed either by Fisher's exact tests ( independent values ) or by Friedman tests ( repeated measures ) . Levene's test for homogeneity of variance was performed prior to all t-tests and ANOVAs and when necessary , data were log2 transformed ( volatile and transcript data ) , square root transformed ( count data ) or arcsin transformed ( herbivore damage data ) to meet requirements for homogeneity of variance . Parametric data were compared using ANOVAs , MANOVAs , or repeated-measure ANOVAs followed by Scheffe post hoc tests . If variance was not homogeneous following transformation , data were compared using Kruskal-Wallis tests ( for multiple comparisons ) or Mann-Whitney U-tests ( for two-way comparisons ) and Bonferroni p-value corrections were used to correct for nonparametric multiple comparisons . For Kruskal-Wallis tests and Mann-Whitney U-tests , a Monte Carlo algorithm was used with 10 , 000 permutations and a 95% confidence level . | As the population of the world continues to increase beyond 7 billion , and agricultural pests continue to rapidly evolve resistance to pesticides , it is becoming ever more important to cultivate arable land in a way that is sustainable for both humans and the environment . A better understanding of the different mechanisms used by wild plants to deter herbivores will help to increase crop production without harming the environment . Plants use both direct and indirect methods to fend off herbivores . Direct defense methods include the production of chemicals that are toxic to herbivores or give them indigestion , and the growth of sticky prickles and spines that can injure or kill the herbivore . Indirect defense methods , on the other hand , generally rely on the plant attracting organisms that are either predators or parasites of the herbivore . Plants produce odors known as herbivory-induced plant volatiles ( HIPVs ) that are thought to offer indirect defense against herbivores by betraying their location to predators and parasites . However , HIPVs also influence other members of the ecological community , sometimes in ways that are detrimental to plants . Moreover , despite 30 years of research , no study has demonstrated that HIPVs increase the fitness of a plant , so it is unclear what they have evolved to do . Now , a 2-year field study by Schuman et al . has shown plants that emit green leaf volatiles ( which are a type of HIPV ) produce twice as many buds and flowers—a measure of fitness—as plants that have been genetically engineered not to emit green leaf volatiles . This study was conducted with Nicotiana attenuata , which is a wild tobacco plant that is often targeted by Manduca sexta , a type of moth that is also known as the tobacco hornworm . Green leaf volatiles only increased plants' fitness when various species of Geocoris—a bug that preys on Manduca sexta—reduced the number of herbivores by a factor of two . This is the first evidence that HIPVs offer indirect defense against herbivores . Schuman et al . also studied the effects of molecules called protease inhibitors that are thought to function as direct defenses by making it difficult for herbivores to digest plants . They found that the ability to produce protease inhibitors did not increase the fitness of plants under herbivore attack; however , tobacco hornworms that had been fed plants containing protease inhibitors were found to be more sluggish in response to attack , which suggests that protease inhibitors can enhance the indirect defenses of plants . The results suggest that employing both direct and indirect defenses—such as a combination of biological pesticides and genetic engineering to produce both HIPVs and protease inhibitors—is the best approach for defending agricultural plants against pests . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"evolutionary",
"biology"
] | 2012 | Herbivory-induced volatiles function as defenses increasing fitness of the native plant Nicotiana attenuata in nature |
This study defined the genetic epidemiology of dengue viruses ( DENV ) in two pivotal phase III trials of the tetravalent dengue vaccine , CYD-TDV , and thereby enabled virus genotype-specific estimates of vaccine efficacy ( VE ) . Envelope gene sequences ( n = 661 ) from 11 DENV genotypes in 10 endemic countries provided a contemporaneous global snapshot of DENV population genetics and revealed high amino acid identity between the E genes of vaccine strains and wild-type viruses from trial participants , including at epitope sites targeted by virus neutralising human monoclonal antibodies . Post-hoc analysis of all CYD14/15 trial participants revealed a statistically significant genotype-level VE association within DENV-4 , where efficacy was lowest against genotype I . In subgroup analysis of trial participants age 9–16 years , VE estimates appeared more balanced within each serotype , suggesting that genotype-level heterogeneity may be limited in older children . Post-licensure surveillance is needed to monitor vaccine performance against the backdrop of DENV sequence diversity and evolution .
Dengue is the commonest arboviral disease of humans and has been a major public health problem in tropical Asia and Latin America for decades ( Stanaway et al . , 2016 ) . Reducing the population of competent mosquito vectors of dengue viruses has been the central aim of disease control efforts , but these have had little success in eliminating or stopping the spread of dengue globally . Effective dengue vaccines will be essential tools to achieving dengue control . Accordingly , the licensure of the first tetravalent dengue vaccine ( chimeric yellow fever–dengue virus tetravalent dengue vaccine ( CYD-TDV ) , Sanofi Pasteur ) together with recommendations from The World Health Organisation’s Strategic Advisory Group of Experts ( SAGE ) on Immunization on its use in highly endemic countries , has provided the first prospects of an integrated public health approach to disease control ( WHO , 2016 ) . Dengue vaccine development has been challenging , in part because dengue viruses ( DENV ) exist as four phylogenetically and antigenically distinct serotypes ( DENV-1 to −4 ) . Within each virus serotype exists considerable genetic diversity at local , national and continental scales ( Holmes , 2008 ) . Subtle antigenic differences can also be measured amongst members of the same virus serotype and are speculated to be of epidemiological and clinical importance ( Katzelnick et al . , 2015 ) . The virus population dynamics of DENV in hyperendemic areas are complex , often involving the emergence and extinction of viral lineages against a backdrop of multiple virus types co-circulating and oscillating in their relative prevalence . Human population immunity and intrinsic virus fitness in mosquitoes and humans are potential drivers of DENV evolution in these settings ( Holmes and Burch , 2000 ) . Acting to balance high mutation rates of DENV within individual hosts , the vector-human transmission cycle subjects viral populations to strong purifying selection , whereby emergent virus variants that are less fit for disseminated infection of both humans and mosquitoes are lost from the viral population ( Holmes , 2003 ) . CYD-TDV was found to be safe and efficacious for use in children 9 years of age and older , with efficacy varying according to age , baseline serostatus and virus serotype ( Capeding et al . , 2014; Villar et al . , 2015 ) . Furthermore , a trend toward reduced efficacy against DENV-2 was observed in the Asian phase III trial compared to the Latin American trial ( Hadinegoro et al . , 2015 ) . This finding suggested that the efficacy of CYD-TDV might be affected by sub-serotype ( i . e . genotype ) level diversity in DENV populations , often associated with geographical boundaries . Beyond the epidemiological factors identified in previous studies of CYD-TDV efficacy , the performance of dengue vaccines could also be influenced by the evolving nature of DENV populations in endemic settings . For example , the possibility that circulating DENV populations could ‘escape’ vaccine-elicited immune responses was nominated as one of several possible explanations for the relatively low efficacy of CYD-TDV against DENV-2 in a phase IIb trial in Thailand ( Sabchareon et al . , 2012 ) . Two phase III efficacy trials of CYD-TDV , involving more than 31 , 000 children between the ages of 2–14 years in the Asia–Pacific region ( CYD14 trial ) and between the ages of 9–16 years in Latin America ( CYD15 trial ) ( Hadinegoro et al . , 2015 ) enable , for the first time , a post hoc investigation of vaccine efficacy versus DENV population diversity . Thus , the aims of the present study were threefold . First , to document the genetic distance between the components of the CYD-TDV formulation and the DENV strains detected amongst cases in the CYD14 and CYD15 trials . Second , to perform focused analysis of the level of sequence conservation between CYD-TDV vaccine strains and wild-type DENV at epitope locations targeted by potent virus neutralising human monoclonal antibodies ( mAbs ) . Lastly , we aimed to explore if a more complex genotype-specific efficacy pattern existed in the CYD14 and CYD15 trials , notwithstanding the limitations inherent to post hoc analysis . Collectively , these data provide insights into the characteristics of the CYD-TDV product relative to contemporary DENV populations and provide preliminary insight into genotype-level vaccine efficacy that can serve as a baseline for future research .
433 acute serum samples from 595 virologically-confirmed dengue ( VCD ) cases in CYD14 and 512 samples from 662 VCD cases in CYD15 were selected for investigation on the basis of subject consent , viremia level and sample volume considerations ( Figure 1A and B , respectively ) . From CYD14 , 314 complete DENV envelope ( E ) gene nucleotide sequences ( 1485 nt for DENV-1 , –2 , −4; 1479 nt for DENV-3 ) were acquired directly from 433 serum samples ( 72 . 5% , including three mixed infections ) , with a subset of 299/433 ( 69 . 1% ) samples also having a complete premembrane ( prM ) nucleotide sequence . From CYD15 , 333 complete DENV E gene nucleotide sequences were acquired directly from 512 serum samples ( 65 . 0% , including eight mixed infections ) , with a subset of 313/512 ( 61 . 1% ) samples also having a complete prM nucleotide sequence . The proportion of serum samples that yielded an E gene sequence was similar between control and dengue vaccine recipients within each study ( Supplementary file 1a ) . The probability of acquiring an E gene sequence from serum samples was positively associated with the DENV viremia level ( Figure 1—figure supplement 1 ) . Full and partial E gene sequences determined directly from serum samples collected in CYD14 and CYD15 trials ( 253 DENV-1 , 191 DENV-2 , 107 DENV-3 and 110 DENV-4 ) were aligned with E gene sequences corresponding to the CYD-TDV vaccine strains and sequences from GenBank for which the year and country of sampling were known . Maximum likelihood trees representing subsampled E gene sequence datasets allowed the classification of CYD14/15 viruses to the major intra-serotype lineages ( genotypes ) previously described for DENV ( Figure 2—figure supplements 1–4 ) . At the country level , CYD14/15 viruses were closely related to publicly available DENV sequences acquired from the same country , an indicator of ongoing local evolution . Figure 2 shows the genotypes detected in the CYD14/15 virus populations according to their country of sampling . Collectively , these data define the population genetics of viruses responsible for dengue cases in the CYD14/15 trials and provide a unique contemporaneous snapshot of DENV diversity in ten endemic countries . We quantified the differences between the E gene amino acid sequences in the components of the tetravalent CYD-TDV formulation and viruses from VCD cases in the CYD14 and CYD15 trials . The mean level of E gene amino acid sequence difference between vaccine strains and viruses from VCD cases in CYD14 and CYD15 was <3% for all serotypes ( Figure 3 and Supplementary file 1b ) . To define the nature of these sequence differences , the amino acid positions that varied between CYD-TDV vaccine strains and the E gene sequences sampled in CYD14/15 trials and in the subsampled GenBank sequences were annotated adjacent to the subsampled maximum likelihood phylogenetic trees for each serotype . The DENV-2 E gene phylogeny ( incorporating the vaccine strain ) of relevance to the CYD14 trial is shown in Figure 4A and for CYD15 in Figure 4B . The equivalent annotated phylogenies for DENV-1 , –3 and −4 are shown in Figure 4—figure supplements 1–6 . These data reveal that positions of amino acid non-identity between CYD-TDV vaccine strains and wild-type viruses were dispersed across the E protein and do not cluster to any particular structural domain . We examined amino acid sequence identity between vaccine strains and wild-type CYD14/15 virus sequences at twelve B cell epitopes . The twelve epitopes represent some of the best structurally defined epitopes in DENV that are targeted by potent virus neutralising human mAbs and are thus of particular interest in vaccine development and immune correlate assays ( Fibriansah et al . , 2014; Cockburn et al . , 2012a; Smith et al . , 2013; Fibriansah et al . , 2015a , 2015b; Teoh et al . , 2012; Rouvinski et al . , 2015; Costin et al . , 2013; Cockburn et al . , 2012b ) . Sequence analyses indicated limited variation at these epitope regions in the CYD14/15 sequences , as well as in a global database of wild-type virus sequences ( Figure 5 and Figure 5—figure supplement 1 ) . The conservation of these epitope sequences between the decades-old ‘donor’ viruses from which the CYD-TDV product was derived and contemporary virus populations suggests that these amino acid sites are not highly prone to evolutionary drift . Given the high degree of overall amino acid sequence identity , including at key epitope positions , between the E protein found in CYD-TDV vaccine strains and contemporary wild-type CYD14/15 viruses , we postulated that vaccine efficacy would be largely independent of virus genotype . We report two levels of intention to treat genotype-level efficacy from the CYD14 and CYD15 trials: the observed estimates and the observed+imputed estimates . The observed estimate refers to vaccine efficacy in the population of VCD cases who had serum samples yielding an E gene sequence that was empirically assigned a genotype . The observed+imputed estimates used the observed genotype data plus imputation to give genotype assignments to VCD cases where the serotype was known but genotype information was absent . Imputation was likely to be accurate because data from this study ( Figure 2 ) indicated eight out of the ten study countries had only a single genotype of each serotype in circulation during the study period . Publicly available sequence data largely mirror the genotype distributions observed in this study; greater diversity is found in some Asian countries relative to those detected in this study , likely because the publicly available sequences are collated at the country level , whereas the CYD14/15 sequences represent those circulating only within the geographically limited trial populations ( Supplementary file 1c ) . The count of observed and imputed genotypes is summarised in Supplementary file 1d . Estimates of genotype-level vaccine efficacy amongst the observed and observed+imputed case populations are described in Table 1 ( all ages ) and Table 2 ( participants 9–16 years of age ) . For completeness , we also show the observed genotype-level vaccine efficacy for participants < 9 years of age in Supplementary file 2 but do not consider it in the main analyses because this age-class was only present in the CYD14 trial and is below the age for which the licensed vaccine is now indicated ( i . e . ≥ 9 years ) ( WHO , 2016 ) . For each serotype , a Cox proportional hazards regression model ( expressing the hazard function ) was used to estimate vaccine efficacy ( derived as 100* [1- Hazard Ratio] ) with vaccine group , genotype and the interaction between vaccine group and genotype included as covariates . The parameter estimates and the 95% confidence intervals of the interactions are given in Table 3 ( all ages ) and Table 4 ( participants 9–16 years of age ) . For DENV-1 , vaccine efficacy estimates against the three different genotypes were highly similar in the all ages group and in participants 9–16 years of age ( Tables 1 and 2 ) . Additionally , the genotype interaction parameter estimates in the all ages group ( Table 3 ) were close to zero and had reasonably tight 95% confidence bounds . This suggests it is unlikely that an interaction exists between genotype and vaccine efficacy , but if such an interaction does exist , it is small . Amongst participants 9–16 years of age , the interaction parameter estimates had 95% confidence intervals that bounded zero and were wider than the all ages group , making conclusions relatively difficult to draw . For DENV-2 , vaccine efficacy estimates against the American-Asian genotype ( 50 . 2%; 95% CI: 32 . 6–63 . 2% ) and the Cosmopolitan genotype ( 43 . 8%; 95% CI: 16 . 1–62 . 2% ) were similar , and both were higher than against the Asian I genotype ( 19 . 8%; 95% CI: −30 . 0–49 . 6% ) in the all ages group ( Table 1 ) . However , the genotype interaction estimates had 95% confidence intervals that , although reasonably tight , included zero in the all ages group ( Table 3 ) and in participants ≥ 9 years ( Table 4 ) . We note however that the upper bound of the confidence interval was very close to zero for the Asian/American versus Asian I genotype interaction ( all ages , Table 3 ) , leaving open the possibility that true heterogeneity may exist . Within DENV-3 , the confidence intervals for the interaction estimates were very wide when comparing genotype II versus genotype I in the all ages population ( Table 3 ) and in participants ≥ 9 years ( Table 4 ) , and thus no conclusions could be drawn from these data . For DENV-3 genotype III versus genotype I , the 95% confidence intervals around the interaction estimates ( Tables 3 and 4 ) were much tighter but nonetheless passed through zero . This suggested an interaction between genotype and vaccine efficacy remained possible but unlikely , and that more data would be needed to address this question . Against DENV-4 , vaccine efficacy was significantly lower against genotype I ( 58 . 3% , 95% CI: 29 . 9–75 . 2% ) , which circulates endemically only in Asia , compared to the globally distributed genotype II ( 81 . 8% , 95% CI: 74 . 3–87 . 1% , across CYD14/CYD15 ) in the all ages population ( p=0 . 009 ) ( Table 1 ) . Confidence intervals around estimates of the interaction between genotype I and genotype II and vaccine group exclude zero , consistent with a lower efficacy against genotype I relative to genotype II ( Table 3 ) . However , when efficacy against DENV-4 genotype I versus genotype II was considered only in participants ≥ 9 years , efficacy was similar between genotypes ( Table 2 ) and confidence intervals for interaction estimates included zero ( Table 4 ) . An important caveat is that relatively wide 95% confidence intervals for all interaction estimates amongst participants ≥ 9 years suggests limited power to detect differences in vaccine efficacy ( Table 4 ) , i . e . this study was generally underpowered to assess heterogeneity in this age subgroup analysis . A visualization of genotype-specific vaccine efficacy versus amino acid identity of trial viruses to CYD-TDV components is shown in Figure 3—figure supplement 1 . These data illustrate the absence of a direct relationship between vaccine efficacy and genetic similarity between wild-type and vaccine strains of DENV .
The pivotal Phase III efficacy trials of CYD-TDV and longer term follow-up have revealed the complex efficacy profile of this vaccine ( Capeding et al . , 2014; Hadinegoro et al . , 2015; L'Azou et al . , 2016 ) . These trials also highlighted generic challenges in dengue vaccine development , e . g . the goal of balanced immunity to four DENV types , achieving efficacy in naïve and partially immune populations , and the need for long-term safety evaluation . A potential additional layer of complexity stems from the ongoing evolution of DENV populations in endemic countries and whether vaccines derived from viruses that circulated decades ago are ‘fit for purpose’ as immunogens against contemporary virus populations . Here , we demonstrate that the DENV E protein components in the CYD-TDV formulation shared high-level amino acid sequence identity , including at prominent B cell epitopes targeted by virus neutralising human mAbs , with viruses sampled in the CYD14 and CYD15 trials . Additionally , within the constraints of the available sample size and confounding factors discussed later , we found limited statistical evidence of genotype-specific differences in the efficacy profile . With a total of 253 DENV-1 , 191 DENV-2 , 107 DENV-3 and 110 DENV-4 E gene sequences generated from CYD14 and CYD15 , this study provides a contemporary characterization of DENV population genetics in ten highly endemic countries . Across all four serotypes , the E gene phylogenies positioned the CYD14 and CYD15 sequences together with those from geographically identical locations , consistent with long-term endemic circulation of a single virus genotype within each viral serotype in nearly all locations . Additionally , the phylogeographical profile demonstrates within-region sharing of virus genotypes but little inter-regional mixing ( distinct viral population profiles can be distinguished within three major geographical categories: mainland Southeast Asia , maritime Southeast Asia , the Americas ) ; only DENV-4 genotype II was found in multiple countries in both regions . The number of genotypes , and hence genetic diversity , detected for any given serotype was greater in Southeast Asia than in Latin America , as expected given the long history of hyperendemicity , within-serotype diversity , and high forces of infection in Southeast Asia ( Halstead , 2006; Holmes , 2009; Rodríguez-Barraquer et al . , 2014; Imai et al . , 2015 ) . Collectively , these data are informative for vaccine and drug development and design of molecular diagnostics . They also serve as virus population baseline profiles from which to monitor DENV evolution in countries where a selective pressure such as CYD-TDV might be widely introduced . Indonesia , with two DENV-1 genotypes , and the Philippines , with two DENV-4 genotypes , were the only countries with genotype co-circulation detected within the trial . As each of these viral populations appears closely related to previously sequenced viruses from their respective country , these likely represent true local circulation in the population rather than recent importations of novel viruses . In general , however , long-term co-circulation of multiple genotypes within a single serotype in one location is rare and these viral populations may instead represent a cross-section of the viral populations during a process of genotype replacement , in which an endemic viral population is rapidly replaced ( often completely ) by a novel population imported from another geographical region ( Zhang et al . , 2005; Lambrechts et al . , 2012; Loroño-Pino et al . , 2004 ) . This process may also be responsible for the presence of DENV-3 genotype III in Thailand , while neighboring Vietnam harbors only genotype II viruses . Genotype II was the dominant DENV-3 lineage in Thailand and mainland Southeast Asia from the early 1980s until at least 2010 ( Rabaa et al . , 2013 ) , but was not detected in Thailand in this study . Recent reports of DENV-3 genotype III infections elsewhere in mainland Asia suggest that this lineage may be moving through the region , potentially replacing genotype II ( Lao et al . , 2014; Jiang et al . , 2012 ) . Ongoing virological surveillance in the CYD14 and CYD15 trial populations , as well as populations vaccinated post-licensure , will be used to further investigate the relationships between the vaccine , virus evolution and local DENV genotype variation . Antigenic differences between viruses of the same serotype have been observed with neutralizing polyclonal and monoclonal antibodies in laboratory assays , and these differences have been postulated to be relevant to clinical epidemiology and vaccine development ( Katzelnick et al . , 2015; Kochel et al . , 2002; OhAinle et al . , 2011 ) . Arguing against a critical role for within-serotype sequence diversity is the common acceptance that natural infection with one serotype elicits life-long clinical immunity to that serotype in the vast majority of instances ( Waggoner et al . , 2016 ) . With respect to CYD-TDV , immunization of non-human primates elicited antibodies that neutralised geographically , phylogenetically and clinically diverse DENV serotypes and genotypes ( Barban et al . , 2012 ) . That CYD-TDV immunisation induced similar measured levels of efficacy against all genotypes of DENV-1 is supportive of the concept that clinical immunity elicited by CYD-TDV to this serotype was pan-genotype in nature ( Tables 1 and 2 ) . Analyses also suggest potential pan-genotype immunity in the case of DENV-3 , although the relatively low numbers of DENV-3 cases detected within the CYD14 trials in Southeast Asia result in this study being underpowered to assess potential heterogeneity . Further research is warranted to understand if smaller differences exist in genotype-level vaccine efficacy than could be measured in these trials and that might be relevant to programmatic use of vaccine . In DENV-4 , efficacy against genotype I ( found in Asia ) in the all ages population was significantly lower than that against genotype II in both Asia and the Americas . Subgroup analyses of genotype and age-stratified vaccine efficacy are inevitably speculative because of diminishing sample sizes and wide confidence intervals around the interaction estimates . Nonetheless we observed that in participants 9–16 years of age , the age group eligible for the licensed vaccine , the vaccine efficacy point estimates were similarly high ( >80% ) between DENV-4 genotypes . DENV-2 is of particular interest because CYD-TDV efficacy is lowest against this serotype . In a previous , single-centre phase IIb trial ( CYD23 ) in Thai children ( Sabchareon et al . , 2012 ) , where all the circulating DENV-2 viruses belonged to the Asian I genotype , efficacy was just 3 . 5% ( -59 . 8; 40 . 5 ) against this serotype/genotype . In CYD14/15 , with a larger sample size , efficacy against the DENV-2 Asian I genotype in the all ages population ( 19 . 8% ( -30 . 0; 49 . 6 ) ) was lower , but not significantly so , than that against other DENV-2 genotypes . Although heterogeneity could not be confirmed , further analysis of the interaction between genotype and vaccine group suggested a potentially decreased efficacy profile of the DENV-2 Asian I genotype in the all ages population compared to the American/Asian genotype , which currently circulates only in the Americas . As above for DENV-4 , it is speculative to examine subgroups , but for the age group 9–16 years of age , the age group eligible for the now licensed vaccine , the efficacy against DENV-2 Asian I genotype ( 34 . 6% ( -27 . 4; 65 . 7 ) was comparable to that seen against the other DENV-2 genotypes ( Asian/American and Cosmopolitan ) , albeit with inevitably wide confidence intervals around the point estimates . The basis for reduced efficacy against DENV-2 Asian I genotype and DENV-4 genotype I in the all ages population could be complex and linked to uncharacterised differences in how vaccine-elicited immunity acts on these virus genotypes . We note that DENV-2 Asian I genotype and DENV-4 genotype I viruses were only detected in Asia ( CYD14 ) , where younger trial participants were included compared to Latin America ( CYD15 ) . While a sole impact of age in vaccine-elicited immunity may account for a proportion of this difference , high DENV diversity within Asia resulted in the detection of additional DENV-2 and DENV-4 genotypes within the CYD14 study ( DENV-2 Cosmopolitan genotype and DENV-4 genotype II ) , against which no evidence of decreased vaccine efficacy was shown . It will be of interest to monitor efficacy against DENV-2 and DENV-4 genotypes in post-marketing effectiveness studies of CYD-TDV . Interestingly , while the parental strain of the CYD DENV-2 component is in fact based on an historical Asian I strain , contemporary DENV-2 Asian I populations diverge from the parental CYD strain at multiple amino acid residues , some of which are postulated to increase transmission fitness in some circumstances ( Vu et al . , 2010 ) . Additional investigations , which might include animal model studies coupled with virological surveillance in the post-vaccine licensure period , could assist further understanding and precision of estimates of vaccine efficacy against different genotypes of DENV-2 . Examination of amino acid sequences among circulating viruses , vaccine components , and epitope sequences targeted by potent , virus neutralising human mAbs provides a framework to predict and possibly understand genotype-specific vaccine performance . These analyses generally underscore the similarity between vaccine components and circulating viruses . Where there were differences at epitope sequence locations , we did not observe a measurable effect in the genotype-level vaccine efficacy . For example , mismatches between circulating DENV-1 and the vaccine component at IF4 epitope sites ( Fibriansah et al . , 2014 ) ( sites E155 , E161 , and E171; Figure 5—figure supplement 1 ) are present across individual genotypes , yet the point estimates of genotype-specific vaccine efficacy were not measurably different . In DENV-4 , there was evidence of lower vaccine efficacy against genotype I viruses and also amino acid mismatches between the vaccine component and genotype I virus sequences at known 5H2 epitope positions ( E155 and E160; Figure 5—figure supplement 1 ) ( Cockburn et al . , 2012a ) . Further research will be needed to understand the significance of these differences for clinical immunity . Several mismatches between circulating DENV-2 viruses and the vaccine component were observed at important epitope sites ( Figure 5 ) , but these were shared by two circulating genotypes in most cases ( sites E71 , E149 and E226 ) . Comparing wild-type DENV-2 sequences sampled in CYD14/CYD15 to the CYD-TDV vaccine component , only site E83 showed a mismatch in a high proportion of the Asian I population alone . Sequence data obtained from GenBank confirm that , while there is some variability at this site among all contemporaneous DENV-2 lineages , the Asian I lineage is defined by this amino acid difference at E83 . An important caveat to these sequence comparisons is that amino acid differences between vaccine strains and wild-type viruses at known B cell epitopes does not necessarily imply that antigenicity ( or immunogenicity ) is altered – functional assays of antibody binding will be needed for this . Our study had several limitations . These were post hoc analyses and , inevitably , sample sizes became small for genotype-level vaccine efficacy estimates and in particular the age-class ( 9–16 year olds ) subgroup analysis . This manifests as wide 95% confidence intervals around the genotype-level point estimates of vaccine efficacy . We generally did not obtain E gene sequences from VCD cases with low viremia and , hence , whether some rare genotypes are not represented in the population of E gene sequences is unknown . In countries where only a single genotype was detected , it was assumed that no undetected genotypes were circulating concurrently in the same location , while publicly available data indicate greater diversity in some Asian countries during this period than detected in this study ( Serotypes 1 and 3; Supplementary file 1c ) . This assumption may have affected the imputed estimates of vaccine efficacy within CYD14 , but because vaccine efficacy was ultimately estimated across the entire study population rather than at the country level , the impact of this assumption is expected to be limited . Baseline serostatus is an important determinant of the efficacy profile of CYD-TDV ( Hadinegoro et al . , 2015 ) . It is possible that some of the genotype-level efficacy results are confounded by the baseline serostatus of vaccine recipients in particular countries , but because only 10% of all participants were characterized immunologically at baseline it is not possible to explore this further ( Capeding et al . , 2014 ) . Such questions may be addressed in post-licensure research . Finally , our results and conclusions may be specific to this particular vaccine given its unique composition . Other vaccines in development employ different donor viruses and have different compositions and , hence , deserve their own evaluations; in any large-scale trial of a candidate DENV vaccine , continual monitoring will be key to understanding the landscape and evolution of circulating DENV populations and further elucidating the potential relationships between virological factors , vaccine efficacy and post-immunization transmission dynamics . Despite these limitations , the results described here improve the understanding of CYD-TDV vaccine performance . Post-licensure research is needed to further understand the complex profile of this vaccine , and to monitor the impact of vaccination programs on the evolution of DENV populations .
Briefly , viremic serum samples from VCD cases detected during the active phase of surveillance in CYD14 ( Indonesia , Malaysia , the Philippines , Thailand , Vietnam; collected between June 2011 and December 2013 ) and CYD15 ( Brazil , Colombia , Honduras , Mexico , Puerto Rico; collected between June 2011 and April 2014 ) were eligible for inclusion in this study ( Figure 1 ) . Samples with very low viremia or low sample volume that were highly unlikely to be fit for purpose with respect to nucleic acid isolation , amplification and sequencing of prM/E genes were excluded from further investigation , as were samples for which consent was not obtained . A total of 433 and 512 viremic serum samples were available and subjected to sequencing from CYD14 and CYD15 , respectively . Viral RNA was extracted from serum using the MagNA Pure 96 DNA and Viral NA Small Volume Kit ( Roche , Mannheim , Germany ) . PrM and E gene regions were amplified by PCR using 16 different primer pairs , with universal tails at the 5’ end to allow the addition of 454 sequencing-specific nucleotides and isolate-specific multiplex identifiers ( MIDs ) in a second PCR round , ‘barcode incorporation PCR’ . The first PCR round was performed in 20 µl reaction volumes using the FastStart High Fidelity Reaction Kit ( Roche ) with the addition of 0 . 25 µM of each PCR primer . The target genes were amplified by PCR in 96 well plates , with the following cycling conditions: denaturation at 94°C for 2 min followed by 40 cycles of PCR , with cycling conditions of 30 s at 94°C , 1 min at 57°C for DENV-1; 60°C for DENV-2; 56°C for DENV-3; 55°C for DENV-4 , 60 s at 72°C and 72°C in 5 min for final extension . After PCR , the amplicons were purified using magnetic AMPure XP beads ( Agencourt , Woerden , The Netherlands ) . The purified first round PCR amplicons were re-amplified to incorporate 454 sequencing-specific nucleotides and isolate-specific MIDs . For this , we used fusion primers that are composed of three parts: 454 sequencing-specific adapter nucleotides , MID sequences and the sequence target of interest on the DNA sample . The second PCR reactions were performed in 10 µl reaction volumes using the FastStart High Fidelity Reaction Kit ( Roche ) with the addition of 0 . 1 µM of each fusion primer . The thermo cycling conditions were: denaturation at 94°C for 2 min followed by 30 cycles of PCR , with cycling conditions of 30 s at 94°C , 30 s at 57°C , 60 s at 72°C and 72°C in 5 min for final extension . Amplicons spanning the same genomic coordinates , but from different viruses , were pooled . Amplicon pools were measured using Quant-iT PicoGreen dsDNA Assay Kit ( Invitrogen , Carlsbad , California ) after purification by magnetic AMPure XP beads . In preparation for 454 sequencing , the concentration of the pooled amplicons was adjusted to 106 copies/mL . The purified amplicons were the pooled into one library tube at a concentration of 5 . 105copies/mL . An emulsion-based clonal amplification ( emPCR ) was performed according to the manufacturer’s instructions as described in the emPCR Amplification Method Manual - Lib-A , revision June 2010 ( Roche ) . DNA sequencing was performed using the GS Junior Titanium Sequencing Kit and the GS Junior Titanium PicoTiterPlate using the Sequencing Method Manual , revision June 2010 ( Roche ) . GS Mapping software ( Roche ) was used for primer trimming and alignment of reads against reference sequences . Briefly , each read per amplicon was mapped to a reference sequence ( DENV-1/VN/BID-V2732/2007 , GenBank accession number GQ199773 . 1; DENV-2/VN/BID-V1873/2007 , GenBank accession number FJ461321 . 1; DENV-3/VN/BID-V1933/2008 , GenBank accession number KF955460 . 1; DENV-4/KH/BID-V2055/2002 ( GenBank accession number KF955510 . 1 ) . Sequence quality was high; the Phred scores for E gene sequences are provided in Supplementary file 3 . Technical controls were included in all sequencing runs , and showed 100% sequence concordance across the prM and E gene in all cases . CYD-TDV prME sequences from CYD1-CYD4 vaccine components are deposited in GenBank under accession numbers KX239894-KX239897 , respectively . All sequences obtained from study subjects are deposited in GenBank under accession numbers KY818060-KY818289 , KY851378-KY851758 , and KY882502-KY882554 . A total of 664 DENV prM and/or E gene sequences were obtained using the above protocol ( DENV-1 , 253; DENV-2 , 191; DENV-3 , 108; DENV-4 , 112 ) ( Figure 1 , Figure 1—source datas 1–4 ) . Sequences were manually aligned using Geneious ( v7 . 1 . 7; RRID:SCR_010519 ) and validated on both nucleotide and amino acid levels . Due to the availability of a larger , less geographically biased public database of E gene sequences compared to prM/E , we focused analysis on the E gene only ( 1485 nucleotide/495 amino acids for DENV-1 , –2 , −4; 1479 nucleotide/493 amino acids for DENV-3 ) , excluding three sequences for which only the prM sequence could be obtained . Thus for each serotype , all full and partial E gene sequences ( DENV-1 , 253; DENV-2 , 191; DENV-3 , 107; DENV-4 , 110 ) were aligned to large datasets of publicly available E gene sequences from GenBank for which the country and year of sampling are known . Maximum likelihood ( ML ) phylogenies were inferred for nucleotide sequences using RAxML ( v8 . 0 , http://www . exelixis-lab . org/; RRID:SCR_006086 ) under the GTRGAMMAI model and were visually assessed to determine the genotype of each virus obtained from CYD14 and CYD15 . Visual inspection further indicated that all viruses sequenced in this study fell into expected lineages corresponding to previously sampled sequences from the countries from which they were isolated . To investigate DENV sequences from the CYD14 and CYD15 studies in the context of the viruses circulating in their respective countries and make datasets more tractable , datasets were subsampled to include all CYD14 and CYD15 E gene sequences and up to three randomly selected , publicly available sequences per country per year from the countries involved in this study , along with up to five representative sequences of each known genotype ( Goncalvez et al . , 2002; Twiddy et al . , 2002; Wittke et al . , 2002; Patil et al . , 2012 ) , regardless of the country from which they were isolated ( Total number of taxa used for phylogenetic reconstruction , CYD14: DENV-1 , 317; DENV-2 , 279; DENV-3 , 222; DENV-4 , 207 . Total number of taxa used for phylogenetic reconstruction , CYD15: DENV-1 , 236; DENV-2 , 252; DENV-3 , 159; DENV-4 , 119 ) . Evolutionary models for each dataset were determined using jModeltest ( v2 . 0; RRID:SCR_015244 ) ( Posada , 2009 ) . ML trees were then inferred from these nucleotide sequences using RAxML ( v8 . 0 ) under the GTRGAMMAI model with 500 bootstrap replications , and p-uncorrected sequence identity ( pairwise comparison of genetic differences across the nucleotide and amino acid alignments ) was determined using Geneious ( v7 . 1 . 7 ) . To investigate potentially novel amino acid residues or changes suggestive of selection , all amino acid sites showing a difference between two or more CYD14/CYD15 sequences relative to the vaccine or circulating viruses of the same genotype were mapped to the aforementioned phylogenies using Phandango ( https://jameshadfield . github . io/phandango; RRID:SCR_015243 ) . All phylogenies were visualized and annotated using FigTree ( v1 . 4 . 2; RRID:SCR_008515 ) and Phandango . Gene annotations were done using the GR7 sequence viewer . To assess the diversity of DENV amino acid sequences and vaccine strains at sites targeted by virus neutralising human mAbs , the serotype-specific E gene alignments used for genotype determination were trimmed to include only sites at which a relevant epitope has previously been identified ( Fibriansah et al . , 2014; Cockburn et al . , 2012a; Smith et al . , 2013; Fibriansah et al . , 2015a; 2015b; Teoh et al . , 2012; Rouvinski et al . , 2015; Costin et al . , 2013; Cockburn et al . , 2012b ) . For each site , the sequences were compared to vaccine components , strains isolated in CYD14 and CYD15 , and publicly available sequences to determine the frequency at which viral sequences matched mAbs targets across all known human DENV lineages . Vaccine efficacy against symptomatic VCD cases according to each genotype during the active phase ( i . e . from D0 to Month 25 ) was calculated using the number of cases ( i . e . , children/adolescents with one or more episodes of VCD ) and the person-time at risk in all participants who received at least one injection according to intention to treat . The incidence density was derived as the number of cases per 100 person-years at risk in each group . A Cox regression model was used to estimate vaccine efficacy ( derived as 100* [1- Hazard Ratio] ) with vaccine group included as a covariate and 95% CI . To further investigate the interaction between vaccine efficacy and genotype , an additional Cox proportional hazards regression model ( expressing the hazard function ) was used to estimate vaccine efficacy with vaccine group , genotype and the interaction between vaccine group and genotype included as covariates . VCD cases with missing genotype were imputed using multiple imputation techniques ( logistic regression ) by serotype with the country included in the model . The twenty imputed ( completed ) datasets were then analyzed separately and the resultant estimates combined using Rubin’s variance rules and their multivariate generalizations ( Rubin , 1987 ) . Analyses were run based on the available data ( i . e . no imputed values ) and on the imputed data ( raw imputation data are shown in Supplementary file 1d ) . A Chi² test ( or Fisher’s exact test ) was used to test the heterogeneity of genotype distribution between vaccine groups . The alternative procedure for pooling chi-square distributed statistics that was proposed by Rubin ( Rubin , 1987 ) and further investigated by Li et al . was used on imputed data ( Li et al . , 1991 ) . A p-value of less than 0 . 10 was considered to indicate statistical significance . All statistical analyses were performed using SAS ( Version 9 . 3; RRID:SCR_008567 ) . | Each year , about 100 million people—mostly children in tropical parts of Asia and Latin America—are infected with the dengue virus . It has been difficult to produce a vaccine against the virus , because there are four different types of the virus , and people respond to infections with different types in an unusual way . Once a person is infected with one type of dengue , they are protected from future infections with that type . However , if that person later becomes infected with a different type , they are more likely to experience severe illness . As a result , a dengue vaccine must simultaneously protect against all four types of the virus to be safe and effective . The first dengue vaccine has recently become available . Clinical studies of the vaccine show that it can protect against all four virus types , but that the protection against certain types and in some age groups varies . Complicating matters , the four types of the dengue virus have continued to evolve since scientists first began developing the vaccine . Therefore , scientists are concerned that the vaccine may not be as effective against the newly evolved subtypes . To find out , scientists would have to carefully compare the genetics of the strains used to develop the vaccine with the strains currently circulating . They would also have to see how well the vaccine protects against current strains . Now , Rabaa et al . show that there is a high level of genetic similarity between the viruses used to create the vaccine , and dengue viruses that caused infections in people participating in clinical studies of the vaccines . The analyses also showed that in children between the ages of 2 and 16 , the vaccine is more effective against one subtype of the dengue type-4 , compared to the other circulating subtype . In children between the ages of 9 and 16 , who are eligible to receive the vaccine in some countries , the vaccine was largely equally effective across the various subtypes . In addition to providing reassurance that the vaccine is working against currently circulating types , Rabaa et al . provide a valuable snapshot of the genetic diversity of dengue viruses . This snapshot will help scientists develop more effective dengue vaccines and treatments . More studies following vaccinated people are needed to ensure that the current vaccine remains effective as circulating strains of the virus evolve . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2017 | Genetic epidemiology of dengue viruses in phase III trials of the CYD tetravalent dengue vaccine and implications for efficacy |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.