pmid stringlengths 8 8 | pmcid stringlengths 8 11 ⌀ | source stringclasses 2
values | rank int64 1 9.78k | sections unknown | tokens int64 3 46.7k |
|---|---|---|---|---|---|
31393956 | PMC6687160 | pmc | 1,971 | {
"abstract": "We proposed a theoretical framework predicting mutualistic outcomes for the arbuscular mycorrhizal (AM) symbiosis based on host provenance (crop versus wild). To test the framework, we grew two isolates of Rhizoglomus irregulare (commercial versus an isolate locally isolated), with five crop plants and five wild plants endemic to the region that co-occur with the locally sourced fungus. While inoculation with either isolate had no effect on plant biomass, it decreased leaf P content, particularly for wild plants. All plants associating with the commercial fungus had lower leaf P. Overall, our data shows that wild plants may be more sensitive to differences in mutualistic quality among fungal isolates.",
"conclusion": "Conclusions Wild plants had highly variable response to inoculation by AM fungi compared to crop plants. This raises concerns about how inoculation practices may affect wild plant/soil communities. Our study provides evidence that the commercial isolate used in this study may be less mutualistic under some conditions. The commercial isolate invested in spore production at the expense of intraradical structures, suggesting a more “selfish” strategy. Correspondingly, plants experienced decreased P with the commercial isolate. It is important for future studies to consider fitness consequences associated with inoculation studies, as poor mutualists may not be apparent over one generation. Considering the number of propagules produced by the commercial isolate, there is a high likelihood of spread beyond the agricultural fields displacing native AM fungi. Future studies should focus on the viability and establishment of these propagules beyond agricultural systems.",
"introduction": "Introduction Arbuscular mycorrhizal (AM) fungi are obligate root symbionts that provide a wide spectrum of benefits to their hosts, such as improved nutrient uptake and stress tolerance [ 1 ]. These benefits have led to their use as bio-fertilizers in agriculture and horticulture over the past 30+ years [ 2 ]. Consumer demand for AM fungal biofertilizers is growing; the number of companies producing inoculum have more than doubled in the past decade [ 3 , 4 ]. Despite early promise [ 5 ], inoculation by AM fungi does not always lead to improved plant performance. Even under controlled greenhouse conditions, failure to colonize is common [ 6 , 7 ] and in cases of successful colonization, effects range from negative [ 8 – 11 ], negligible [ 12 , 13 ] to significant yield increase [ 14 ]. Inoculation with AM fungi in the field is likewise inconsistent, ranging from yield increases [ 15 – 18 ] to no significant effect [ 19 , 20 ]. Most of our knowledge about host responses to inoculation by AM fungi is based on domesticated cultivars [ 17 , 18 , 21 , 22 ], meaning we have a poor understanding of how inoculants may affect local plant populations and communities if they disperse beyond the target plant community [ 23 ]. AM fungal inoculants may pose little threat to natural plant communities because most commercial inoculants comprise cosmopolitan species with worldwide distribution [ 24 ]. But there exists large intraspecific variation among conspecifics, [ 25 – 28 ], including mutualistic quality [ 29 ] and genetics [ 27 , 30 , 31 ]. Such variation may make it difficult to predict the functioning of these inoculants, even if conspecifics naturally co-occur. Symbiotic outcomes for wild plants may differ from domesticated cultivars, leading to differential responses to inoculation. Because wild plants generally depend more on AM fungi compared to cultivars [ 32 – 35 ], symbiotic outcomes may be more pronounced (positive or negative) for wild plants compared to cultivars. This effect may be exacerbated by local adaptation between wild plants and soil biota as native plants respond more positively to local, versus exotic AM fungi [ 36 , 37 ]. Not only are local fungi adapted to local conditions [ 38 , 39 ], mutualisms are more beneficial when partners share evolutionary history [ 40 ]. Taken together, such differential responses to commercial fungal inoculum may result in less beneficial mutualisms for local plants if commercial inoculants become naturalized. We designed a theoretical framework to predict how wild plants and crops can respond differently to AM fungal inoculation ( Fig 1 ). To test the framework and to determine potential positive or negative impacts of commercial inoculants, we evaluated the mycorrhizal response of five wild plants and five crop plant species growing with a commercial AM fungal isolate and a locally sourced conspecific to test the questions: 10.1371/journal.pone.0221037.g001 Fig 1 Experimental hypotheses and theoretical framework of differential response of wild and crop plants to AM fungi. The thickness of arrows corresponds to the magnitude of host-plant response. Inoculation effects on crop and wild plants (up arrows): We hypothesized inoculation with AM fungi would have a weak effect, either positive or negative, on crops due to lack of coadaptation [ 36 , 37 ] and reduced mycorrhizal responsiveness of domesticated plants [ 32 – 35 ]. For wild plants, we hypothesized effects of inoculation will be magnified, having a strong, either positive or negative effect, due to strong mycorrhizal responsiveness of wild plants [ 32 – 35 ]. Effects of plant provenance on AM fungi (down arrows): We hypothesized wild plants would have a strong positive effect on AM fungal growth due to increased dependency while crops would have a weak positive effect on AM fungi due to lack of coadaptation and reduced mycorrhizal responsiveness [ 36 , 37 ]. Does plant provenance affect mycorrhizal response ? Does plant provenance affect fungal response ?",
"discussion": "Discussion Our experiment provides preliminary evidence that introduced fungi may negatively affect AM mutualistic outcomes. Wild plants had pronounced variation in their response to fungi compared to crops. Responses ranged from positive to strongly negative revealing the sensitivity of wild plants to fungal identity, even to isolates within the same fungal species. Future studies assessing the risk of fungal inoculants should examine this relationship on a variety of hosts, multiple functional groups and fungal isolates. Plant responses While wild plants have been reported to be more responsive to AM fungi [ 32 – 35 ], we did not find support for this in terms of biomass when looking at plants as either ‘crop’ or ‘wild’ plants. Because our wild plants were perennials, it is possible that our study did not allow enough time for full biomass differences to manifest, as the study ended when crops, but not wild plants, had senesced. Thus, our inability to detect a difference among crops may have been due to time constraints [ 13 ]. When we looked at responses of individual plant species, wild plants varied more in their response to inoculation with AM fungi, ranging from highly negative to highly positive. Variation in biomass has been documented for wild plants in the literature, particularly, for perennials versus annuals [ 51 ], and natives versus exotics [ 40 ] revealing the strong mycorrhizal responsiveness of wild plants. Inoculation with the commercial isolate led to increased shoot: root for crops but not for wild plants. It is not uncommon to observe alteration in root: shoot ratio with inoculation by AM fungi [ 52 – 54 ]. The “functional equilibrium” theory suggests that plants allocate biomass preferentially to maximize resource acquisition, a plant should favor above ground growth when carbon is limited. Because carbon allocation from plant to fungus can lead to carbon limitation [ 55 ], our results indicate that the commercial isolate may have posed more of a carbon demand than the local isolate, leading to increased shoot allocation. Such changes may lead to reduced nutrient acquisition for plants associated with the commercial fungus in some conditions. We found support for our hypothesis that plant provenance affects mycorrhizal response in terms of % leaf P . Wild plants, surprisingly, experienced a decrease in percent % leaf P when inoculated with AM fungi. While there is evidence for wild plants as more mycorrhizal dependent compared to crops [ 32 – 35 ], their increased sensitivity to AM fungi can lead to magnified negative effects as well particularly when fungi and plants are competing for limited resources [ 40 ]. In our study, the commercial isolate was less mutualistic in terms of % leaf P and this effect was magnified in wild plants. Other studies have shown of AM fungal inoculation leading to reduced host P [ 8 – 11 ]. While such reductions may be related to greenhouse growing conditions, reduced P following inoculation may also indicate a less mutualistic AM association in some cases [ 10 , 56 , 57 ]. In our study, plants inoculated with the commercial isolate had lower P compared to non-mycorrhizal controls which could indicate either direct competition between plant and fungus for P, or P hoarding by the fungus [ 58 ]. It may also mean that the commercial isolate does not have enhanced P uptake ability over plant-direct uptake routes, perhaps through loss of traits during domestication. Further studies comparing more isolates with isotope labelling and genomic studies could elucidate the mechanism involved. Nevertheless, in order to further examine wild and cultivar plant responses to inoculation, that will more accurately represent natural growing conditions, it is very important for future studies to consider soil physicochemical properties, since it has been shown that sympatric combinations of plants, fungi and soil, can lead to increased MR [ 37 ]. Fungal responses We did not find support for our hypothesis that wild plants were more beneficial to AM fungi compared to crops. Rather, large differences in growth strategies among the two fungal isolates may explain differences plant response. The commercial isolate in our study had few arbuscules at time of harvest. This is unusual as arbuscules (or coils) are considered fundamental to the mutualism under natural conditions [ 59 ]. Reduction of arbuscules has been reported for a variety of AM fungal species (including Rhizophagus sp .) under stressful environments [ 60 , 61 ], and due to differences in harvest time and level of fertilization [ 62 , 63 ]. Specifically, suppression of arbuscules can occur with increasing P or N [ 62 ] and changes in arbuscule formation due to time of harvest can be regulated by the species identity [ 63 ]. In our experiment, differences are likely do to fungal strategies since there was no suppression of arbuscules in the locally sourced isolate. Low levels of arbuscules in the commercial isolate may be explained by considering the conditions under which the commercial isolate is propagated. Large-scale inoculum production occurs mostly on transformed roots, which are able to directly uptake most of their resources from the nutrient medium [ 64 ] and have very low nutrient requirements [ 65 ]. Such a luxurious in nutrients environment, may require fewer arbuscules or enhance the resource sink abilities of the isolate, but this remains to be seen. Given that there is still considerable debate over the function of arbuscules [ 66 , 67 ], it is difficult to identify factors that promote or suppress their production. Alternatively, propagation using transformed roots for inoculum production may favor ruderal behavior since ruderal traits, such as rapid growth and early production of abundant propagules, are of interest to inoculum production industry [ 68 ]. If the commercial isolate is more ruderal, the extremely low number of arbuscules observed for the commercial isolate at harvest might be due to a faster`completion of its life cycle compared to the locally sourced isolate. Future studies examining the progress of the symbiosis over time would be able to reveal such significant differences in life history strategies. Overall, we observed low values of root colonization for both isolates across all plants. Our experiment was conducted during the winter. Thus, it is possible our plants reduced the amount of carbon allocated to the roots and subsequently to AM fungi [ 69 ]. While there was no difference in the extent of ERM among fungal isolates, the commercial isolate invested heavily in spore production compared to the local isolate. Large differences in spore production among isolates is not unusual, as there have been many reports of inter and intraspecific variation in fungal traits, over several orders of magnitude in some cases [ 28 , 54 , 70 – 72 ]. Nevertheless, the difference in sporulation rate observed in this study, is unusually large (50x) and represents a significant carbon drain for hosts associating with this fungus. Low levels of root colonization during winter growing conditions and especially due to light limitations is typical ([ 73 – 76 ]) but in our study it had the unintended benefit of potentially exacerbating differences in LHS among our isolates. Under the carbon limiting conditions of our study, the commercial isolate allocated resources into non-mutualistic structures (i.e. spores)–thereby competing with its host for nutrients. In contrast, the wild isolate allocated resources to mutualistic structures (i.e. arbuscules). It would be interesting to test whether such strategies are a response to resource levels, or robust across gradients. Allometry (Intraradical: Extraradical investment) Root colonization is not a good predictor of symbiosis quality [ 77 , 78 ]. On the contrary, examining specific traits can be more meaningful [ 79 , 80 ]. The commercial isolate had a significantly different growth pattern compared to the locally sourced isolate that was consistent among hosts and plant provenance, revealing important LHS variations between the two isolates. The commercial isolate had a high soil biomass, which could enhance soil exploration potential and subsequently, host benefit [ 81 ]. Considering the differences in spore number between the two isolates, deriving from the same quantity of ERM, means that the commercial isolate represented a nutrient sink rather than a source (including C and P)."
} | 3,554 |
30867493 | PMC6416308 | pmc | 1,972 | {
"abstract": "We report on memsensors, a class of two terminal devices that combines features of memristive and sensor devices. Apart from a pinched hysteresis (memristive property) and stimulus dependent electrical resistance (sensing property) further properties like dynamic adaptation to an external stimulus emerge. We propose a three component equivalent circuit to model the memsensor electrical behaviour. In this model we find stimulus dependent hysteresis, a delayed response to the sensory signal and adaptation. Stimulus dependent IV hysteresis as a fingerprint of a memsensor device is experimentally shown for memristive ZnO microrods. Adaptation in memsensor devices as found in our simulations resembles striking similarities to the biology. Especially the stimulus dependency of the IV hysteresis and the adaptation to external stimuli are superior features for application of memsensors in neuromorphic engineering. Based on the simulations and experimental findings we propose design rules for memsensors that will facilitate further research on memsensitive systems.",
"conclusion": "Conclusion We report on the concept of memsensors, a class of two terminal devices combining properties of memristive devices and sensors. The junction of features of memristive devices (pinched hysteresis, memory) and sensors (stimulus dependent resistance) gives rise to stimulus dependent IV hysteresis and adaptation to the applied stimulus, which are the fingerprints of a memsensor. We propose a memsensor model based on three components. In the equivalent circuit, two memristive elements are connected in parallel respectively in series to the sensory element. The model is capable of resembling adaptation behaviour to external stimulus and stimulus dependent hysteresis and thus showcases all features of a memsensor device. The experimental investigation on ZnO microrods replicates stimulus dependent hysteresis. Based on a review on the state of art in memristive devices with sensing capabilities and our investigations, we proposed three design guidelines to facilitate further research on memsensor devices.",
"introduction": "Introduction Since the postulation of the experimental realization of a memristor device in 2008 1 , there has been an increased effort in understanding and modelling memristive (resistive) switching for a broad variety of device classes 2 – 4 . Up to now (2018) an increasing group of concepts for memristive devices with sensing behaviour has been reported, typically based on metal oxide or semiconductor nanostructures with sensing of temperature 5 , (UV)-light 6 – 9 , gases 10 , 11 , magnetic field 7 , mechanical response 12 , biomolecules 13 – 15 or pH 16 , 17 . However, a systematic treatment of the junction between memristive and sensitive devices is still missing. In this work we report on the concept of memsensors, a class of two terminal electronic devices that combine the properties of memristive switching and sensing properties. By thoughtfully designing electronic devices with both, a sensing part (stimulus dependent resistance) and a memristive part (pinched hysteresis and memory) it is possible to make use of their enhanced capabilities. In classical electronics respective controlling theory, sensors that show a hysteresis are typically regarded as a poor sensing device. However, features like adaptation and forgetting are essential for efficient learning in biological systems such as neuronal networks 18 , 19 . In memsensors, the combination of sensing and memory resembles the dynamic response of biological systems to environmental stimuli. One example is adaptation: the response to a constant excitation decreases asymptotically over time, the system accommodates to the constant stimulus. After reducing the stimulus for a long period, the system recovers its initial sensitivity. Adaptation is key for efficient use of neuronal capabilities 18 . First indications for adaptation in memsensor devices, although not specifically mentioned, can be found in the work of Chiolerio et al . and Lupan et al . 6 , 11 Although over the past years there has been an increasing number of detailed reports on memristive switching devices combined with sensing capabilities, to the best of our knowledge a systematic description of the concept of memsensors as well as the link to adaptation is still missing. In the current work, we deduct design guidelines to facilitate further research on memsensors based on the simulations and experimental findings.",
"discussion": "Results and Discussion The concept of a memsensor A circuit element for a light controlled memsensor acting as a current source was, to the best of our knowledge, first proposed by Chiolerio et al . in 2015 (Fig. 2a ) 6 . Based on this, we propose a generalized circuit element for any memsensor that is sensitive to any kind of external stimulus (Fig. 2b ). Although not addressed specifically in their work, early indications of adaptation to external stimulus appeared in the electrical characterization of the devices of Chiolerio et al . and Lupan et al . 6 , 11 . Figure 2 ( a ) Circuit symbol for a UV memsensor as proposed by Chiolerio et al . 6 ; ( b ) generalized memsensor circuit symbol in agreement with the proposed ( c ) equivalent circuit of a memsensor device featuring memristive elements parallel (with R m,par ) and in series (with R m,ser ) to the sensing element (with R s ). The resistance R s is sensitive to an external stimulus α stim , e.g. gas concentration or UV-irradiation. Common to all memsensitive devices discussed in the previous section is the influence of the external stimulus on the electronic structure. Accordingly, the memristive device is expanded by a sensing capability for this external stimulus. In addition to the core feature pinched hysteresis and stimulus dependent response a memsensor may show further features. In this paper we report on adaptation as an emergent feature of a memsensor. Adaptation in case of a memsensor describes the asymptotically decrease of the memsensor response to a constant stimulus after an initial high response (Fig. 3a ). Figure 3 Adaptation to an external stimulus: ( a ) Amplitude adaptation: Amplitude of the response is proportional to the stimulus. ( b ) Spike frequency adaptation as proposed by E.D. Adrian in 1926: The frequency of spikes is proportional to the stimulus and fades over time. Spike frequency adaptation has been experimentally realized in a memristive spiking neuron circuit 23 – 25 , 37 . The system does not respond to subthreshold stimulus. The adaptation behaviour of a memsensor device shows parallels to adaptation in biology 18 , 23 – 25 . In the biological context adaptation is typically discussed with regard to firing frequency in pulsed operation (Fig. 3b ). In case of spike frequency adaptation, the number of spikes per unit time (spike frequency) is proportional to the strength of the stimulus (a certain threshold stimulus may be overcome in order to excite spiking). For a constant applied stimulus the frequency of spikes decreases with time – the system adapts to the stimulus. Adaptation is believed to be an essential process of a signalling system to be better suited to environmental changes. It can be found in nearly all biological systems. Adaptation enables the system to blank out background signals and offers the possibility to be more sensitive to abrupt changes. Macroscopic effects of this adaptation process are for example the habituation to certain smells that are only perceived by the individual for a short time after entering a room. In fact, such complex habituation processes are not realized by a single neuron but involve the mutual interplay between various neuron ensembles, so called neural networks 26 , 27 . In this paper we will show memsensors with adaptation in constant voltage mode with pulsed stimuli. Accordingly, the adaptation is in the amplitude of the memsensor response. This roughly relates to the integration over spike frequency modulated signal. However, this leads to an easier implementation to neuromorphic networks due to constant applied voltage. Accordingly, two terminal memsensor devices have high potential for neuromorphic engineering, e.g., as a feed forward network with input from external stimulus. For this purpose, sensory signals that can be applied locally are favourable, eliminating signals like temperature for the design of memsensors. Modelling of a memsensor device In order to model a memsensor device that is capable of adaptation to the sensory signal, we propose a simple two terminal equivalent circuit (Fig. 2c ) based on memristive elements in parallel (with R m,par ) and in series (with R m,ser ) to the sensing element (with R s ). The resistance of the sensory element R s depends on the signal (external stimulus) that is supplied to the sensor. For an ideal linear sensor the resistivity R s varies linearly between a high resistive state (HRS) and a low resistive state (LRS) Equation ( 1 ). 1 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${R}_{s}({\\alpha }_{stim})={R}_{s,LRS}+(\\frac{{\\alpha }_{stim}}{{\\alpha }_{stim,\\max }})\\cdot ({R}_{s,HRS}-{R}_{s,LRS})$$\\end{document} R s ( α s t i m ) = R s , L R S + ( α s t i m α s t i m , max ) ⋅ ( R s , H R S − R s , L R S ) For the memristive elements we approximate a simple linear dependence of the resistance on an internal state variable ( ω ) and the respective HRS and LRS of the memristive element (Equation ( 2 )). Conventionally the memristive element is in its high resistance state in case the internal state variable equals zero. 2 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${R}_{m}(\\omega )={R}_{m,LRS}+(1-\\omega )\\cdot ({R}_{m,HRS}-{R}_{m,LRS})$$\\end{document} R m ( ω ) = R m , L R S + ( 1 − ω ) ⋅ ( R m , H R S − R m , L R S ) Based on the equivalent circuit, for any applied voltage U in , the individual potential drops U m over the serial (Equation ( 3 )) and the parallel (Equation ( 4 )) memristive element are given. 3 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${U}_{m,ser}={U}_{in}\\cdot \\frac{{R}_{m,ser}}{{R}_{m,ser}+\\frac{{R}_{m,par}\\cdot {R}_{s}}{{R}_{m,par}+{R}_{s}}}$$\\end{document} U m , s e r = U i n ⋅ R m , s e r R m , s e r + R m , p a r ⋅ R s R m , p a r + R s 4 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${U}_{m,par}={U}_{in}\\cdot \\frac{\\frac{{R}_{m,par}\\cdot {R}_{s}}{{R}_{m,par}+{R}_{s}}}{{R}_{m,ser}+\\frac{{R}_{m,par}\\cdot {R}_{s}}{{R}_{m,par}+{R}_{s}}}$$\\end{document} U m , p a r = U i n ⋅ R m , p a r ⋅ R s R m , p a r + R s R m , s e r + R m , p a r ⋅ R s R m , p a r + R s In our simulation the internal state variable for each memristive element is updated in each time step and the respective resistances and voltages are calculated accordingly. We chose a model for the memristive device (Equation ( 5 )) in which the internal state variable ω is changed depending on the polarity of applied voltage, a threshold voltage U s and a constant internal back driving voltage U b , representing all internal forces driving the system back into its equilibrium state (which is in our model the HRS). The back driving voltage U b relates to the sum of all processes that relax the system into its equilibrium, comparable to the back driving (restoring) force in common Boltzmann-equation-like approaches 28 . These potentials translate to an electrical field across the device that results in forces on the electrically charged mobile species (e.g. cations, oxygen vacancies) in the memristive elements. The memristive device switches to its LRS if it is subjected to a voltage that is higher than the threshold voltage and the back driving voltage. This assumption of a voltage driven model with a threshold voltage and backdriving voltage is in close relation to memristive switching that relies on metal cation movement. In such electrochemical metallization cells 3 the ionization processes, which are necessary to oxidize the metal atoms from the active electrode, resemble a threshold. The applied voltage drives the transport of metal cations through the matrix. The back driving force and the time constant τ define the retention time of the state of the memristive device. 5 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\omega }_{i+1}=\\{\\begin{array}{ccc}{\\omega }_{i}+\\frac{|{U}_{m}|-{U}_{s}}{\\tau }\\cdot \\frac{{U}_{m}}{|{U}_{m}|}-\\frac{{U}_{b}}{\\tau } & {\\rm{for}} & |{U}_{m}|\\ge {U}_{s}\\\\ {\\omega }_{i}-\\frac{{U}_{b}}{\\tau } & {\\rm{for}} & |{U}_{m}| < {U}_{s}\\,{\\rm{and}}\\,0\\le {\\omega }_{i}-\\frac{{U}_{b}}{\\tau }\\,\\\\ 1 & {\\rm{for}} & 1 < {\\omega }_{i}+\\frac{1}{\\tau }\\cdot (|{U}_{m}|-{U}_{s})\\cdot \\frac{{U}_{m}}{|{U}_{m}|}-\\frac{{U}_{b}}{\\tau }\\\\ 0 & {\\rm{for}} & 0 > {\\omega }_{i}+\\frac{1}{\\tau }\\cdot (|{U}_{m}|-{U}_{s})\\cdot \\frac{{U}_{m}}{|{U}_{m}|}-\\frac{{U}_{b}}{\\tau }\\,{\\rm{or}}\\,0 > {\\omega }_{i}-\\frac{{U}_{b}}{\\tau }\\end{array}$$\\end{document} ω i + 1 = { ω i + | U m | − U s τ ⋅ U m | U m | − U b τ for | U m | ≥ U s ω i − U b τ for | U m | < U s and 0 ≤ ω i − U b τ 1 for 1 < ω i + 1 τ ⋅ ( | U m | − U s ) ⋅ U m | U m | − U b τ 0 for 0 > ω i + 1 τ ⋅ ( | U m | − U s ) ⋅ U m | U m | − U b τ or 0 > ω i − U b τ The built-in back driving potential and the threshold voltage both prevent the system from switching to LRS at low voltages. Equation ( 5 ) may be substantially simplified in case there is no threshold voltage (Equation ( 6 )). 6 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\omega }_{i+1}=\\{\\begin{array}{ccc}{\\omega }_{i}+\\frac{{U}_{m}-{U}_{b}}{\\tau } & {\\rm{for}} & 0\\le {\\omega }_{i}+\\frac{{U}_{m}-{U}_{b}}{\\tau }\\le 1\\\\ 1 & {\\rm{for}} & 1 < {\\omega }_{i}+\\frac{{U}_{m}-{U}_{b}}{\\tau }\\\\ 0 & {\\rm{for}} & 0 > {\\omega }_{i}+\\frac{{U}_{m}-{U}_{b}}{\\tau }\\end{array}\\,\\,$$\\end{document} ω i + 1 = { ω i + U m − U b τ for 0 ≤ ω i + U m − U b τ ≤ 1 1 for 1 < ω i + U m − U b τ 0 for 0 > ω i + U m − U b τ The corresponding switching behaviour for a single memristive element is visualized in Fig. 4 , once for the case without threshold voltage (Fig. 4a ) and once for the more complex description (Fig. 4b ). In both cases three distinct switching regimes are to be distinguished. SET for switching of the memristive device to low resistive state. Slow RESET, which is limiting the device retention time, as observed for a multitude of memristive systems. (Fast) RESET occurs at negative voltage over the threshold voltage for switching to HRS. Figure 4 Schematic overview over the switching regimes realized by the mathematical representation of the memristive device in Equation ( 6 ) ( a ) and Equation ( 5 ) ( b ). The SET regime (green) corresponds to a switching of the memristive device to LRS at voltages above the threshold voltage. The slow RESET (yellow) describes the finite retention of the LRS. The fast RESET (red) below the negative threshold voltage corresponds to the switching to HRS. In order to overcome the necessity of defining additional conditions for accounting for the boundary cases of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$\\omega =1$$\\end{document} ω = 1 and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$\\omega =0$$\\end{document} ω = 0 , the model may be extended by incorporation of a window function. Appropriate window functions for memristive models are discussed in the work of Prodromakis et al . and Zha et al . 29 – 31 . However, in this model we do not make use of any window function in order to keep a simple representation of all parameters by physical quantities. Hysteresis and adaptation in memsensor devices Based on the three component model the behaviour of a memsensor with regard to voltage ramps and stimulus pulses was investigated. All parameters for this simulation are given in the supporting material. Comparable results were obtained using the switching model with and without a threshold. For the investigation of the response of a memsensor to stimulus pulses, a constant unipolar voltage is applied to the two terminals. The successful adaptation to a pulsed external stimulus is demonstrated in Fig. 5 . The response of the memsensor reaches a maximum value in the first stimulus pulse and decreases asymptotically with each subsequent pulse. After a prolonged low stimulus pulse, the state of the memsensor is restored and its response to the next set of stimulus pulses is identical to the previous set. The step by step calculation in the memsensor model allows plotting the internal states of the memristive devices during the pulses (Fig. 5 ). At a low stimulus, the parallel memristive device is reset to its high resistive state due to its back driving force. The serial memristive element is at ON state, determined by the equilibrium between the voltage that drops over this element and its internal back driving force. At the high stimulus pulse, the resistance of the sensory element rises and there is a higher voltage drop over the parallel sub-circuit. Accordingly, the state of the parallel memristive element rises and the state of the serial memristive element drops. In the following pulse the memristive elements switch back. At this point one precondition for adaptation as we found in our simulations becomes clear: both memristive elements need to show different relaxation behaviour. As shown in the supplementary data, we chose different time constants and back driving forces to realize this. A similar behaviour can be observed for a single high stimulus pulse for a longer time (Fig. 5c ). First the faster serial memristive element drops to its HRS, then the parallel element saturates at a LRS. The overall memsensor resistance reaches its maximum shortly after the onset of the pulse and then gradually adapts to the signal. Especially this behaviour resembles striking similarities to earlier devices from Lupan et al . (c.f. first 100 ppm pulse Fig. 4c ) and Chiolerio et al . (photocurrent measurements, Fig. 2c,d ) 6 , 11 . Figure 5 Simulation of adaptation to an external stimulus: ( a ) The response of the memsensor (black line) decreases with each subsequent stimulus pulse (blue background). After a sufficient time at low stimulus, the memsensor recovers. In depth look at the impact of serial (green line) and parallel (red line) memristive element on the memsensor adaptation to external stimulus pulses (blue background). The adaptation is shown at the example of a short pulse ( b ) and a single long pulse ( c ). The core feature of a classical memristive device is its pinched current-voltage (IV) hysteresis curve 1 , 32 . Typically the shape of this hysteresis curve depends on the measurement time. The faster the IV measurement is recorded, the less time the memristive elements have to change their internal state. Accordingly, the hysteresis loop broadens with increasing cycling time 1 . This expected dependency is also observed in the IV characteristic of the memsensor model (Fig. 6a ). As the memsensor additionally incorporates a sensitive element, the shape and the extent of the hysteresis loop also depends on the applied external stimulus (Fig. 6b ). This dependency is easily explained by considering the equivalent circuit. At zero applied voltage, both memristive elements start in their high resistive state. In case R sens is low (and R ser is high), the main voltage drop is over the serial memristive element, which will be switched to its low resistive state until the applied voltages and the back driving voltage are equal. In case R sens is high (and R par is high), most voltage will drop over the parallel sub-circuit and accordingly the parallel memristive element will switch to its low resistive state. In addition, the memsensor also shows an offset in its HRS, depending on the applied stimulus. The origin of this offset is the change in resistance of the sensor element by applying a stimulus, which reflects on the overall resistance of the memsensor. Figure 6 Evaluation of simulated IV hysteresis based on the equivalent circuit. ( a ) Dependency of hysteresis on cycling time, the hysteresis loop narrows for faster cycling times; ( b ) dependency of hysteresis on external stimulus, the hysteresis loop widens for lower external stimulus. ZnO rods as a memsensor prototype A broad variety of materials and sensing signals has been proposed 5 – 17 . The memristive effect and UV-sensitivity in ZnO nanorods was recently described in detail by Russo et al . at the example of Ag/ZnO/FTO structures 8 . Accordingly, we chose Ag/ZnO/Ag structures, in which the ZnO rods are contacted on both sides by silver glue. Typically nanostructures are preferred for their capability of faster sensing. However, we chose bigger structures, namely ZnO microrods for their higher time constants in sensing UV-light, matching better to biological time constants of adaptation. In Fig. 7a a typical ZnO microrod device is depicted schematically, the insets show a SEM micrograph as well as a photographic image of a ZnO microrod sample. We prepared ten ZnO microrod samples and recorded IV curves with and without irradiation of UV-light (Fig. 7b ). The device yield for working memristive devices is approximately 20% and is mainly limited by handling of the ZnO microrods and contacting them macroscopically with silver glue. For completeness, a full overview over all prepared samples can be found in the supplementary material (Fig. S2 ). Figure 7 ( a ) Schematic drawing (top) and a photographic image (right) of a ZnO microrod device for comparison. The SEM micrograph is showing the diameter of the ZnO microrod is 13 µm (left); The IV hysteresis curves show strong dependence on the illumination by UV light ( b ) as shown in linear IV plot and logarithmic RV plot. Without UV illumination the ZnO device approached the limit of reliable detection (10 pA). The time constants for the sensing of UV light are determined ( c ) to be in the range of seconds. The memristive switching devices yield a good prototype to showcase stimulus dependent hysteresis as a memsensor feature. To investigate the sensing properties, namely relaxation times for excitation τ exc and relaxation τ rel and the sensitivity S (the ratio of the device resistance with and without applied stimulus), the time dependence of the ZnO samples overall resistance upon switching of the UV irradiation is recorded (Fig. 7c ). The Ag/ZnO/Ag microrod device shows a sensitivity of roughly 350 and the time constants are in the order of seconds to tens of seconds ( τ exc ≈ 11 s, τ rel ≈ 14 s), which offers good observability and is more close to biological time constants for spike frequency adaptation 25 . The typical µs to ms time constants obtained in classically well performing sensors are somewhat comparable to the time constants of individual neuron spikes 33 . The IV-characteristics for a memristive switching ZnO microrod show strong dependencies of pinched hysteresis curves on the applied UV light (Fig. 7b ). In the UV illuminated state, a single ZnO microrod is a memristive device with an ON/OFF ratio of up to 35. Without UV-stimulus, the rod shows about four orders of magnitude higher resistance, which is partially above the limit of reliable detection in the used setup. In general, the finding of stimulus dependent hysteresis in our model was also experimentally proven by the ZnO microrod devices. However, the details of the shape of pinched hysteresis curve in simulation and in experiment differ, as our model does not include the nonlinearity of a Schottky contact potentially formed at the contact of the microrod with the silver glue. The Schottky barrier to the silver paste seems to be highly relevant to achieve memristive behaviour. Switching by supplying charges to Schottky interface is well known for memristive devices 4 , 34 . Guidelines for memsensor design Based on the evaluation of state of art sensitive memristive devices and our investigations on ZnO microrods, we propose three design guidelines to facilitate further research on memsensor devices. The choice of material In any memsensor, memristive switching has to be joined with sensitivity towards a stimulus. A good starting point for possible memsensor materials are memristive device whose electronic structure of the bulk device or its interfaces with the contacts can be changed by an external stimulus. Metal oxide materials, such as CuO, ZnO, Fe 2 O 3 or TiO 2 , are promising candidates for memsensor devices, as they combine sensitivity (e.g. towards photons) and memristive properties. For an application in neuromorphic networks, e.g. as feed forward network with external sensitivity, a locally applicable signal is beneficial. Thus light, biomolecules or gases would resemble a better signal than delocalized signals like temperature. Matching properties The memristive device shows a pinched IV hysteresis as a characteristic property. In order to allow for adaptation, this switching has to be analog. Thus filamentary memristive devices with binary switching are inferior to, e.g., devices in which charges are shifted from and to a Schottky barrier. There is no need for ultrafast switching or sensing. In contrast, for applications as bio-inspired feed-forward networks, the time constants could match biological time constants. The configuration A prerequisite for any memsensor is its accessibility to the external stimulus. An overview on possible device schematics is shown in Fig. 1 . In case of horizontal structures like nanowires, the memsensor is open to the surrounding and thus also the sensory signal. However, large scale integration may render challenging with this device setup. For vertical structures, in which the active layer is sandwiched between two electrodes, integration into cross bar arrays is possible and was already demonstrated. However, a structuring of the top electrode is mandatory in order to make the memsensor accessible to external stimuli like light or gas molecules."
} | 6,932 |
25327887 | PMC4218951 | pmc | 1,973 | {
"abstract": "Species in ecological communities build complex webs of interaction. Although revealing the architecture of these networks is fundamental to understanding ecological and evolutionary dynamics in nature, it has been difficult to characterize the structure of most species-rich ecological systems. By overcoming this limitation through next-generation sequencing technology, we herein uncover the network architecture of below-ground plant–fungus symbioses, which are ubiquitous to terrestrial ecosystems. The examined symbiotic network of a temperate forest in Japan includes 33 plant species and 387 functionally and phylogenetically diverse fungal taxa, and the overall network architecture differs fundamentally from that of other ecological networks. In contrast to results for other ecological networks and theoretical predictions for symbiotic networks, the plant–fungus network shows moderate or relatively low levels of interaction specialization and modularity and an unusual pattern of ‘nested’ network architecture. These results suggest that species-rich ecological networks are more architecturally diverse than previously recognized.",
"discussion": "Discussion The characteristic network structure of below-ground plant–fungus networks ( Figs 2e,f and 3 ) may result from the unique biological features of these interactions. Unlike other symbiotic systems, a fungal symbiont individual can simultaneously interact with multiple host individuals in mycorrhizal interactions 21 22 . As reward levels provided by a host plant individual (for example, carbohydrates) change with the host’s physiological status or the soil nutrients available to the plants 13 21 , many mycorrhizal fungi are thought to have evolved wide rather than narrow ranges of host plants 23 . If fungi are phenotypically plastic and can abandon interactions with less profitable hosts depending on local biotic/abiotic environmental conditions 21 24 , natural selection would favour the ability to interact with a potentially broad range of hosts. The ability to use multiple plant species and the unique ability of fungi to interact simultaneously 9 13 may be partly responsible for the observed moderate modularity in below-ground plant–fungus symbiosis. Although many of the links in the network likely represent mutualistic interactions, especially those involving mycorrhizal fungi, the plant–fungus network may also include commensalistic and even antagonistic interactions. Diverse clades of root-endophytic and plant-pathogenic fungi are possibly present within the root-associated fungal community of the studied forest 12 16 . Non-mutualistic effects of partners are likely to occur even in interactions involving mycorrhizal fungi, as the benefit and cost of interacting with specific mycorrhizal hosts/symbionts depend on internal physiological status and/or abiotic/biotic environmental conditions 21 . Non-mutualistic links are likely part of all mutualistic networks. For example, the presence of cheaters (for example, nectar robbers) and the context-dependency of interaction type 25 are also expected in previously studied plant–pollinator interactions. Inclusion of these antagonistic links in the description of predominantly mutualistic ecological networks is meaningful, as the lifestyles of these antagonists rely on the existence of mutualistic networks and may affect the stability 26 and coevolutionary processes 9 of mutualisms. Thus, development of a comprehensive conceptual framework for understanding ecological and coevolutionary dynamics will require analysis of all types of possible interactions in a community 26 27 . While arbuscular mycorrhizal, ectomycorrhizal, endophytic and parasitic fungi are sampled and analysed separately in most mycological studies 13 , the present study indicates that compartmentalization by fungal functional or phylogenetic groups is incomplete in real ecological communities. This proposition is supported by recent studies showing that plant species can be simultaneously infected by both arbuscular and ectomycorrhizal fungi 28 and that fungal taxonomic clades can involve both mycorrhizal and endophytic species 29 . Of course, the present data set can include many links of weak, commensalistic or neutral interactions, and hence further technical advances that allow high-throughput evaluation of interaction type/strength are necessary. Ever since Darwin’s deliberation of an ‘entangled bank’ full of interacting species 30 , ecologists and evolutionary biologists have investigated how interspecific interactions are organized in biological communities. Although we have already had data sets of large predator–prey, plant–pollinator, and plant–seed disperser networks encompassing hundreds of species, those visible interactions represent only a tiny fraction of diverse interspecific interactions found in nature. By expanding the target of ecological network analysis to hyperspecies-rich symbiotic interactions by means of high-throughput sequencing, we have shown that the diversity of ecological network architecture has been underappreciated. The significantly low nestedness observed in the plant–fungus network is particularly important, as theoretical studies have argued that the commonly reported nested patterns in species networks could determine feasibility, resilience, persistence and structural stability of ecological communities 4 6 18 31 32 . A comprehensive understanding of the laws that organize the earth’s biosphere will require continued exploration of ecological network architecture in diverse symbiotic and non-symbiotic networks."
} | 1,402 |
36819038 | PMC9935839 | pmc | 1,974 | {
"abstract": "Coastal pollution, global warming, ocean acidification, and other reasons lead to the imbalance of the coral reef ecosystem, resulting in the increasingly serious problem of coral degradation. Coral bleaching is often accompanied by structural abnormalities of coral symbiotic microbiota, among which Vibrio is highly concerned. In this study, Vibrio fortis S10-1 (MCCC 1H00104), isolated from sea cucumber, was used for the bacterial infection on coral Seriatopora guttatus and Pocillopora damicornis . The infection of S10-1 led to coral bleaching and a significant reduction of photosynthetic function in coral holobiont, and the pathogenicity of V. fortis was regulated by quorum sensing. Meanwhile, Vibrio infection also caused a shift of coral symbiotic microbial community, with significantly increased abundant Proteobacteria and Actinobacteria and significantly reduced abundant Firmicutes; on genus level, the abundance of Bacillus decreased significantly and the abundance of Rhodococcus , Ralstonia , and Burkholderia–Caballeronia–Paraburkholderia increased significantly; S10-1 infection also significantly impacted the water quality in the micro-ecosystem. In contrast, S10-1 infection showed less effect on the microbial community of the live stone, which reflected that the microbes in the epiphytic environment of the live stone might have a stronger ability of self-regulation; the algal symbionts mainly consisted of Cladocopium sp. and showed no significant effect by the Vibrio infection. This study verified that V. fortis is the primary pathogenic bacterium causing coral bleaching, revealed changes in the microbial community caused by its infection, provided strong evidence for the “bacterial bleaching” hypothesis, and provided an experimental experience for the exploration of the interaction mechanism among microbial communities, especially coral-associated Vibrio in the coral ecosystem, and potential probiotic strategy or QS regulation on further coral disease control.",
"conclusion": "5. Conclusion To identify the relationship between V. fortis and coral health and to understand its role in bacterial infection, V. fortis S10-1 was designed as the only variable factor to infect the coral Seriatopora guttatus and Pocillopora damicornis . The results of color scale analysis of S10-1-infected coral branches in aquaria indicated that V. fortis was responsible for coral bleaching, which leads to the high probability of pathogenicity of V. fortis in coral. The significant reduction of photosynthetic function in coral holobiont and shift of coral symbiotic microbial community upon the infection of S10-1 provided strong evidence for the “bacterial bleaching” hypothesis. The positive effect of quorum quenching indicated the potential strategy of bacterial disease control. The infection of S10-1 led to the imbalance in the coral-associated bacterial community but had no significant effect on the algal symbionts, and this was accompanied by a significant decrease in the abundance of probiotic Bacillus and an increase in the abundance of Rhodococcus erythropolis and other opportunistic pathogens including Ralstonia and Burkholderia–Caballeronia–Paraburkholderia in the coral-associated community, as well as increased abundance of Shimia and other unclassified genus in family Rhodobacteraceae in the planktonic bacterial community in water. The study provided experimental experience on corals in aquaria for the exploration of the interaction among coral-associated microbial communities in coral relative micro-ecosystem and revealed the potential probiotic strategy or QS regulation on pathogen control for coral health.",
"introduction": "1. Introduction Coral reef ecosystems possess the highest productivity and biodiversity among marine ecosystems; however, due to changes in environmental conditions, such as climate change and ocean acidification, the ecological balance among coral, microorganisms, and the environment has been affected ( Stuart-Smith et al., 2018 ). In addition, the coastal pollution caused by human activities also led to ecological imbalance, especially the microorganism shift, resulting in the increasingly serious problems of coral degradation ( Putnam and Gates, 2015 ) and arousing the wide attention of researchers. The coral-associated microorganisms play important roles to maintain the health of coral holobiont ( Rosenberg et al., 2009 ; Kang et al., 2022 ), which assists the coral host to adapt to the environmental changes by a shift of the symbiotic microorganism community. However, the changes in environmental conditions might lead to the disorder of the microorganism community, including the dissociation of algal symbionts from the coral symbiotic micro-ecosystem, which normally causes coral bleaching, or a shift in the bacterial community, which results in unbalanced nutrition metabolism, which causes coral disease, or the increasing richness of opportunistic pathogens, which causes bacterial infection ( Mao-Jones et al., 2010 ). Although the proposed “bacterial bleaching” hypothesis ( Kushmaro et al., 1996 ) is still controversial ( Ainsworth et al., 2008 ), there have been continuous reports in recent years to prove the correlation between bacterial infection and coral disease ( Ziegler et al., 2019 ), which highlighted the hypothesis that coral disease might be caused by the synergistic effects from one or more pathogenic bacteria. Moreover, marine animals such as fish, shrimp, sea cucumbers, and even starfish in coral reefs play the role of intermediate hosts of pathogens, increasing the probability of bacterial infection during their movement. Coral bleaching is often accompanied by the abnormal structure of symbiotic microorganisms ( Mhuantong et al., 2019 ; Mohamed et al., 2022 ). Almost all bacterial pathogens causing coral diseases are opportunistic pathogens, which produce virulence factors as a response to the changes in environmental conditions such as temperature and pH, or the interaction among bacteria during colonization competition, resulting in coral diseases ( Kimes et al., 2012 ). Although Vibrios had been reported as the main kind of opportunistic pathogens to cause coral diseases, including V. coralliilyticus, V. natriegens, V. parahaemolyticus, and V. shilonii ( Ushijima et al., 2014 ; Li et al., 2017 ; Ahmed et al., 2022 ), the relationships between other Vibrio sp. and coral health remained unknown, and their infection pathway and the pathogenic mechanism remained unknown. In the previous study, a significantly increased abundance of V. fortis had been found in the microbial symbionts in Porites lutea with pigment abnormalities in Lembeh Strait, North Sulawesi, Indonesia, which indicated that V. fortis may be involved in the bacterial infection and caused the coral inflammatory reaction ( Ou et al., 2018 ). To verify this inference, and also to study the potential interaction of microorganism community caused by the V. fortis colonization, in this study, we carried out the Vibrio infection experiments on the laboratory-based model of Seriatopora guttatus and Pocillopora damicornis , both of which are fast-growing and easy-reproducing coral species, using a V. fortis strain from sea cucumber from coral reef area and a marine-source quorum quenching (QQ) enzyme, YtnP, with positive AHL degradation activity ( Sun et al., 2021 ) to inhibit the pathogenicity of V. fortis , to investigate the relationship between extrinsic V. fortis and coral health and to reveal its effects on coral-associated microorganism community by bacterial interaction.",
"discussion": "4. Discussion 4.1. Potential pathogenicity of Vibrio to coral health The role of Vibrios in coral diseases was still limitedly known ( Munn, 2015 ), though there are some Vibrio species had been identified as opportunistic pathogens to cause coral disease, there is disagreement about whether they should be regarded as primary causing agents, as some Vibrios appeared to contribute to nitrification ( Thurber et al., 2009 ) or involved in defense against pathogens ( Ritchie, 2006 ). Unlike the better-studied species, such as V. parahaemolyticus and V. coralliformis , V. fortis is still rarely reported. To the best of our knowledge, it was reported for its pathogenicity to marine animals such as shrimps ( Thompson et al., 2003 ), sea horses ( Wang et al., 2016 ), and sea urchins ( Ding et al., 2014 ), as well as one of the lists of abnormal abundant coral symbiotic microbes ( Ou et al., 2018 ). Predictably, the activity of sea cucumber carrying V. fortis in coral reefs area would increase the risk of bacterial infection to the coral holobiont. However, the disease development and pathogenicity mechanism of coral bleaching caused by V. fortis are not yet clear. It is currently known that the pathogenicity of Vibrio sp. is parallelly regulated by at least three QS pathways, LuxM/LuxN-related AHLs, CqsA/CqsS-related CAI-1, and LuxS/LuxP-related AI-2; all lead to the regulation of core regulatory protein LuxO ( Herzog et al., 2019 ) and then to the synergistic action of key regulatory proteins AphA ( Lu et al., 2018 ) and OpaR ( Zhang et al., 2016 ), which regulate downstream exopolysaccharide synthesis genes to affect biofilm formation and virulence relative genes to release virulence factors including hemolysin. The findings of QS signals produced by bacteria in the coral mucus layer ( Li et al., 2017 ), the signaling molecules and active hydrolytic enzymes released by V. shilonii AK1 in outer membrane vesicles ( Li et al., 2016 ), and the interference of QS regulation against the AHL-mediated opportunistic bacteria of resident microbes in bleaching’s initiation and progression ( Zhou et al., 2020 ), in recent years, all supported the hypothesis of bacterial infection under QS regulation, which is consistent with the positive inhibition of V. fortis pathogenicity under QQ enzyme treatment in this study ( Figure 4 ), and this phenomenon indicates the potential strategy of QQ on coral disease control. 4.2. Microbial shift and bacterial interaction behind Vibrio infection In this article, the infection of V. fortis was designed as the only variable factor in coral health. The symptom development of coral bleaching and tissue lysis verified and strongly supported the hypothesis of “bacterial bleaching.” The findings in this study demonstrated that its infection also caused significant changes in the coral-associated microsystem, especially the increased abundance of opportunistic pathogenic bacteria, which might synergistically affect coral health. Among the microbial community, the decreasing abundance of Bacillus belonging to phylum Firmicutes seems to be the most serious side effect caused by Vibrio infection ( Figure 5 ). Bacillus had been widely accepted as the main antagonistic bacteria and used for biological control in agriculture ( Jiang et al., 2018 ) and aquaculture ( Wang et al., 2019 ), and recently reported as marine probiotics to increase coral resistance to bleaching ( Rosado et al., 2019 ). Bacillus can produce secondary metabolites, including lipopeptides and polyketones with antibacterial, antiviral, or antitumor activities ( Wu et al., 2019 ), and the quorum quenching (QQ) enzymes, such as AiiA ( Dong et al., 2000 ) and YtnP ( Sun et al., 2021 ), to degrade AHLs, inhibiting the biofilm formation and toxin release of AHLs/Lux-mediated pathogens including V. fortis that was known to contain an AHLs intermediated LuxM/LuxN QS pathway ( Ding et al., 2014 ). The subsequent abnormal abundant Bacillus , B. circulans for instance, resulted from the simulated outbreaking of V. fortis when the inoculation of 10 5 CFU/ml was far beyond the threshold of QS; therefore, it might further weaken the resistance of coral holobiont to the stress from pathogens ( Satpute et al., 2010 ; Niu et al., 2014 ). The increased abundance of Rhodococcus in phylum Actinobacteria might also be related to the decreasing abundance of Bacillus , because Rhodococcus sp. was reported with antibacterial activity against Bacillus substilis ( Mahmoud and Kalendar, 2016 ). Among these, the most composition species, R. erythropolis , produces different types of QQ enzymes ( Ryu et al., 2020 ) to recognize and degrade AHLs and could be used for biofouling control ( Ergön-Can et al., 2017 ). The increasing abundance might be speculated to result from the self-regulation of the coral symbiotic microbiota after the sense of abnormally increased concentration of AHLs generated by the inoculated V. fortis . The increasing abundance of the genera Ralstonia and Burkholderia–Caballeronia–Paraburkholderia in phylum Proteobacteria was probably eligible for the causal relationship of the declined abundance of Bacillus because of the less inhibition from the Bacillus -produced surfactin ( Grady et al., 2019 ). However, the correlation between Ralstonia and Burkholderia–Caballeronia–Paraburkholderia of coral health is still unclear ( Bernasconi et al., 2019 ). Reports are claiming that Ralstonia sp. is a known gram-negative phytopathogenic bacteria ( Hayashi et al., 2019 ) and dental opportunistic pathogen ( Tuttlebee et al., 2002 ), while the Burkholderia–Caballeronia–Paraburkholderia , which is mostly constituted by Paraburkholderia fungorum , is reported as a plant probiotic bacteria ( Rahman et al., 2018 ) but still risky to human health ( Tan et al., 2020 ). In addition, considering that the planktonic bacteria in the water could flow into the gastrovascular cavity of coral along with the filter feeding, the significantly increased abundance of Thalassobius (also detected with higher abundance in bleaching coral but with no significant difference) and Tenacibaculum in water might also affect the coral health ( Figure 7 ). Thalassobius belonging to the family Rhodobacteraceae has been reported as a microbial bioindicator enriched in the Stony Coral Tissue Loss Disease accompanied by Vibrio ( Becker et al., 2021 ), and has been associated with invertebrate diseases ( Roder et al., 2014 ), while the increased abundance of Tenacibaculum was also consistent with the findings from microbial community shift in the White syndrome-affected Echinopora lamellosa in aquaria ( Smith et al., 2015 ). Another unclassified genus belonging to the family Rhodobacteraceae and the genus Shimia was also detected in the planktonic bacterial community in water, though there was no significant difference observed yet. The indicator species in the coral hosts, family Rhodobacteraceae and genus Shimia , were observed to increase their relative abundance when corals are under stress ( Casey et al., 2015 ; Pootakham et al., 2019 ) or with the emergence of Porites white patch syndrome accompanied with Vibrio ( Séré et al., 2013 ). This phenomenon was verified in the later coral bleaching in the blank group caused by the deterioration of water quality 2 days after the experiments stopped (data not shown). The virulence from the shift bacterial community of abundant opportunistic pathogens might not be the only reason for coral bleaching. The nutrients sources and waste products for coral holobiont are also a concern, which mainly come from the metabolism of symbiotic microbes such as carbon and nitrogen fixation ( Gibbin et al., 2019 ) and the metabolic integration from algal symbionts ( Sun et al., 2020 ), in such closed aquaria without extra exogenous nutrient sources. The shift in the bacterial community no doubt leads to an imbalance in the nutrient food chain in the coral holobiont ( Raina et al., 2009 ); however, in this study, there is no direct relevance between the infection of V. fortis and the status of algal symbionts. The dominant population in the tested S. guttatus was Cladocopium sp., though with few relative abundances of other Symbiodiniaceae according to the reported ITS2 type ( Yu et al., 2020 ), and no observation of Durusdinium sp., which was reported with stronger stress resistance ( Sun et al., 2020 ); therefore, the findings indicated that the bleaching in this study was caused by Vibrio infection and the following shift of bacterial community, but not by the dissociation of algal symbionts. To sum up, this study expands the cognition of the correlation between coral symbiotic Vibrio and coral health. According to the experience of other reported Vibrio pathogens, the outbreak and pathogenicity enhancement of Vibrio may be related to environmental changes ( Kimes et al., 2012 ). The symbiotic changes caused by the outbreak of V. fortis , especially the increased abundance of other pathogenic bacteria, may also cause more serious damage to the coral holobiont. In addition, the reduced abundance of potential probiotics Bacillus under the competition of microbiota might also provide an experimental experience for probiotic strategy ( Zhao et al., 2019 ) or QS regulation ( Zhou et al., 2020 ) on pathogen control or resistance enhancement ( Rosado et al., 2019 )."
} | 4,285 |
36134978 | PMC9496005 | pmc | 1,975 | {
"abstract": "This review article will discuss the ways in which various polymeric materials, such as polyethylene (PE), polypropylene (PP), polystyrene (PS), and poly(ethylene terephthalate) (PET) can potentially be used to produce bioplastics, such as polyhydroxyalkanoates (PHAs) through microbial cultivation. We will present up-to-date information regarding notable microbial strains that are actively used in the biodegradation of polyolefins. We will also review some of the metabolic pathways involved in the process of plastic depolymerization and discuss challenges relevant to the valorization of plastic waste. The aim of this review is also to showcase the importance of methods, including oxidative degradation and microbial-based processes, that are currently being used in the fields of microbiology and biotechnology to limit the environmental burden of waste plastics. It is our hope that this article will contribute to the concept of bio-upcycling plastic waste to value-added products via microbial routes for a more sustainable future.",
"conclusion": "5. Conclusions The bioconversion of plastics (petroleum-derived polymers) is a complicated process with multiple variables. The detection of micro- and nano-plastics in our waters means the pollution issue is much more personal than ever before. Due to biodegradation, thermo-oxidative degradation, photodegradation, thermal, and hydrolysis processes in the ecosystem, there is a major threat to sea-life and humans indirectly. The carbon backbone of plastics means they are very hydrophobic and inert in nature; however, by utilizing physical and chemical methods, micro-organisms have demonstrated various routes for degradation. Most importantly, non-conventional feedstocks are being applied and researched globally, and this will contribute to the development of circular economy systems [ 10 ]. In this article, several bacterial strains were highlighted that are capable of breaking down a variety of plastics, namely PS, PP, and PE, and some of the pretreatment techniques that could be applied. In some cases, commercially available polymers and films could be used as substrates (pre-treated or not). They may be composed of certain plasticizers, additives, or biodegradable impurities, which are more facile for microbial use than the actual carbon backbone itself, as waste fungal biomass from other processes could potentially act as a carbon source for bacterial growth for the purpose of plastic breakdown. This could lead to false positives being reported in some of the studies referenced. Therefore, the analysis of plastics undergoing fungal, enzymatic, or bacterial breakdowns requires further standardization. Further use of synthetic biology and mixed microbial cultures, in conjunction with metabolic engineering tools to generate microbes, will develop these value-added products for the post-consumer and contribute to a further improved model of production and consumption. Unfortunately, the slower functionality of biological systems is currently limiting this process. Long-term, coordinated clean-up operations are required to evaluate the detrimental ecosystem effects of plastics, and the use of locally sourced waste for bioconversion could be another way carbon footprints are reduced, i.e., by allowing the monomers and oligomers formed after breakdown to be utilized to create more suitable, biodegradable, and diverse commodities, particularly in developing countries.",
"introduction": "1. Introduction The synthesis of polyhydroxyalkanoates (PHAs), a bioplastic that can be used to replace traditional (petrol-based) plastics, is an important focus in today’s politically and environmentally conscious society. PHAs are part of a group of organic polymers containing 3-, 4-, 5-, and 6-hydroxyalkanoic acids that are biocompatible, 100% biodegradable, and nontoxic to the environment [ 1 , 2 ]. PHAs can be considered a greener alternative to synthetic plastic compounds, and they can be produced by plants and various strains of bacteria (as documented later). In addition, these bioplastics can be generated through microbial fermentation using waste materials which could offer more sustainability in a closed-cycle system of carbon materials [ 2 ]. Some of the waste materials that have been used to make PHAs have included used synthetic plastics, such as PE, PS, and PP [ 2 , 3 , 4 ]. From 2022, by changing the co-monomer type and distribution in PHAs, the properties can be considered to be comparable with seven of the most profitable crude-oil-based plastics, which is estimated to be 230 million tons of plastic per year [ 5 ]. There have also been global policies put in place and capacity expansions for the next 5 years for over 1.4 million tons, so there is a lot of encouragement for the industry to adopt biomaterials [ 5 , 6 ]. It is also predicted that 12,000 million tons of plastic waste will be added to landfills or the natural environment by 2050 [ 7 ]. Figure 1 shows a timeline of milestones related to bioplastic development over the last hundred years. Research projects around the world have focused on using microbes to break down some the most persistent types of plastics found in the environment. Currently, the most common way plastics are disposed of is by incineration, mechanical and chemical recycling, and the relatively cheap method of landfill sites [ 8 ]. However, all of these methods have their disadvantages; landfill occupies too much land space, and incineration creates secondary pollutants, such as dioxins and carbon monoxide. Even though mechanical recycling has become the main technique for refuge plastic, the chemical properties are usually compromised via processing, which results in reduced commercial value [ 9 ]. Chemical recycling is known to be able to recover monomers from plastic waste, but its success depends heavily on the efficiency of catalysts [ 9 ]. With up to 79% of waste plastic being discarded into landfills (or the surrounding environment), there is a huge requirement for novel recycling methodologies, and bioconversion is one possible answer [ 4 , 10 , 11 ]. The transfer of current feedstocks could be smoother if the true economical value (including the carbon footprint of products and practical benefits) were considered in detail. The efficiency of biotechnological methods can also be further improved using metabolic engineering, which could help achieve the aims of the internationally agreed Paris Agreement, a treaty on climate change [ 10 ]. Moreover, the recently estimated impact of the COVID-19 pandemic on plastic discharge indicates that around 8.4 million tons of pandemic-associated plastic waste was generated from 193 countries, as of 23 August 2021, and over 25.9 thousand tons were released into the global ocean [ 12 ]. With these issues in mind, the diagram in Figure 2 displays a possible system for generating biomaterials, such as PHAs from waste plastics via fermentation. Elements of this kind of system exist in parts, such as the optical scanning and separation and sorting of plastics; the innovation would be in having all of these sections interlinked. In some cases, processes such as milling, agglutination, or sonication would have to be selected on the basis of the target material’s properties. After thermal pre-treatments of PS or PE, where oxygenated groups were incorporated into the unsaturated carbon backbone, sonication was found to greatly increase the mixing of plastics with the growth media for fermentation [ 2 , 3 , 4 ]. Due to the nature of microbial cultivation, requiring time (normally 48hrs or more) and optimal growth conditions (usually ranging from 30 to 37 °C at 50 to 150 rpm), a lot of energy is needed [ 4 , 10 ]. This is perhaps the major bottleneck for such a system, which is why the choice of micro-organisms used is so important. Additionally, any biological PHA generation has the often-unreported issue of difficulty in controlling the precise purity and biopolymer composition. Moreover, there is evidence showing that the extraction processes can alter the PHA-polymer structure when conventional chemicals (such as chloroform) and Soxhlet extraction are used [ 13 ]. The extraction yields and the PHA properties can also be affected. Data revealed the two different extraction methods alter the crystallization degree and the chemical composition. When pure bioplastic is required, pre-treatments such as homogenization provided a 15% more extractive yield than the others, especially at high pressures, which also improved the visual appearance (transparency and clearness), thermal stability, and mechanical performances, which is ideal for medical grade PHAs [ 13 ]. For packaging (the major application of PHAs), these polymers have already been proposed to effectuate a significant shift in the industry, which currently utilizes almost 40% of plastics created [ 14 ]."
} | 2,222 |
32154089 | PMC7055565 | pmc | 1,976 | {
"abstract": "Abstract Rapid energy‐efficient movements are one of nature's greatest developments. Mechanisms like snap‐buckling allow plants like the Venus flytrap to close the terminal lobes of their leaves at barely perceptible speed. Here, a soft balloon actuator is presented, which is inspired by such mechanical instabilities and creates safe, giant, and fast deformations. The basic design comprises two inflated elastomer membranes pneumatically coupled by a pressurized chamber of suitable volume. The high‐speed actuation of a rubber balloon in a state close to the verge of mechanical instability is remotely triggered by a voltage‐controlled dielectric elastomer membrane. This method spatially separates electrically active and passive parts, and thereby averts electrical breakdown resulting from the drastic thinning of an electroactive membrane during large expansion. Bistable operation with small and large volumes of the rubber balloon is demonstrated, achieving large volume changes of 1398% and a high‐speed area change rate of 2600 cm 2 s −1 . The presented combination of fast response time with large deformation and safe handling are central aspects for a new generation of soft bio‐inspired robots and can help pave the way for applications ranging from haptic displays to soft grippers and high‐speed sorting machines.",
"conclusion": "3 Conclusion Inspired by nature, we introduced an actuator, which harnesses a snap‐through and snap‐back instability for giant high‐speed deformations. By using a system of coupled balloons, remote, high‐voltage triggered, bistable actuation is possible. Consisting of an electrically active and passive part, this actuator can be operated in a safe regime, far away from the EB voltage of the dielectric elastomer. Its combination of fast volume change rate and large maximum deformation makes this concept an attractive candidate for use in soft robotics.",
"introduction": "1 Introduction Nature inspires artificial systems, particularly regarding motion. Quick and large movements are keenly sought‐after for applications in rigid and soft robotics, industrial automation, and modern prosthetics. 1 , 2 , 3 , 4 , 5 In recent years, the development of artificial muscles that mimic the basic function of human and animal musculature has gained an importance due to its wide range of potential applications. 6 But there are examples in which direct muscle action alone cannot be responsible for the rapid movement. The jaw muscles of hummingbirds are not strong enough to close the beak in the observed short amount of time. However, hummingbirds are able to bend their lower jaw and use a controlled elastic instability to rapidly snap it from the open to the closed position. 7 Completely without muscle support, plants rely solely on mechanical instabilities for rapid movement. The inherent motility of plants is the consequence of selective swelling and shrinking caused by water flow driven by osmosis and evaporation phenomena. 8 These processes are rather slow and therefore limit the overall speed. Plants like the Venus flytrap can overcome this limit by suddenly releasing stored elastic energy resulting in one of the fastest movements (≈100 ms) in the plant world. 9 , 10 The rapid closure of the Venus flytrap is made possible by harnessing a snap‐buckling instability ( Figure \n \n 1 \n a). Due to geometric constraint and elastic properties of the doubly curved terminal lobes which form the trap at the tip of each leaf the plant can accumulate elastic energy and release it if triggered by an external stimulus. 9 , 11 Here, we took a lesson from plants and transferred the idea of how mechanical instabilities are key to the rapid actuation to technical applications. By exploiting an elastic snap‐through and snap‐back instability, we show that the speed of elastomer balloon actuators can be accelerated drastically. This possibility results from the nonmonotonic N‐shaped pressure–volume relation of an inflated rubber membrane. 12 , 13 , 14 As soon as the pressure reaches the critical values p \n 1 or p \n 2 , a further insignificant pressure increase or decrease rapidly changes the membrane volume within milliseconds (Figure 1 b). Pressure‐controlled deformation is required for the occurrence of an instability, 15 , 16 therefore the rubber membrane is mounted on a pressure vessel of suitable size. The pressure signal triggering the balloon instability can be provided by any kind of controllable pressure source like a pneumatic compressor, a loudspeaker, or a coupled dielectric elastomer actuator (DEA; Figure 1 b). We chose an electroactive acrylic elastomer (3M VHB 4910F) as actuator material to electrically manipulate the pressure of the system. 17 , 18 VHB has been widely employed due to its large stretchability and high electric breakdown strength. 16 With voltage‐triggered balloon instability there is no need for fast and complex pneumatic pressure equipment, after initially pressurizing the chamber. Actuators based on soft balloons are compliant, robust, light weight, simple in structure, and have low costs. Considering these desirable features, they are widely used, ranging from pneumatic applications in medical and health‐care robotics to wave‐handling systems to transport delicate objects in industry. 19 , 20 , 21 All the mentioned examples would benefit from the increased response speed. To demonstrate the underlying potential, we present prototypes utilizing the snap‐through instability of a balloon. In Figure 1 c, a fast sorting device is shown, which can be used to sort out defective parts on a conveyor belt. A gripper as depicted in Figure 1 d, is fast enough to reliably catch falling objects of various shapes. More controllable gripping can be provided by an appropriate geometric design. A simple example is shown in Figure 1 e, where a mounted rod blocks the balloon, resulting in better compliance to the object's shape and thus ensures safe handling. Figure 1 Harnessing mechanical instability to improve the speed of actuation. a) The Venus flytrap uses a stimulus‐triggered mechanical buckling instability. b) Mechanical balloon snap‐through instability enables high‐speed actuation. Characteristic pressure–volume behavior of a rubber membrane: As soon as the pressure reaches the critical pressures p \n 1 or p \n 2 , a further insignificant pressure increase or decrease rapidly changes the membrane volume in just a few milliseconds. Under pressure‐controlled conditions, snap‐through instability can be triggered by different kinds of pressure sources. c) Possible application as fast sorting device, e.g., for conveyor belts; d) fast and soft gripper catching a ping‐pong ball; e) handling of sensitive objects by improving compliance with additional constraints (Video S1, Supporting Information). Except for lab demonstrations, DEAs solely based on VHB acrylic adhesives—favored for achieving giant static strain—have very limited potential for commercial applications that require fast actuation due to viscoelasticity. 22 We overcome this problem by pneumatically coupling two balloon actuators of dissimilar materials. In our approach, a VHB DEA triggers the instability of an inflated natural rubber membrane, which serves as high‐speed balloon actuator. Therefore, we are able to report giant deformations within 20 ms. The properties of natural rubber enable high‐speed volume expansion rates up to 2300 cm 3 s −1 (2600 cm 2 s −1 area expansion rate) at forward and backward actuation at frequencies up to 8 Hz. We ensure safe operation, as there is no high voltage at the high‐speed balloon actuator. At the same time, the trigger actuator (TA) undergoes small deformations only, averting failure mechanisms such as electromechanical instability or electrical breakdown (EB). 23 The actuation is remote, potentially allowing significant spatial separation into a passive and an electrically active part, which can be arranged “behind the scenes.” Subsequent experiments and theory demonstrate how the rubber balloon achieves high‐speed and giant deformation by harnessing the snap‐through and snap‐back instability. The analysis is based on an electromechanical model using hyperelasticity of elastomer membranes.",
"discussion": "2 Results and Discussion 2.1 Experimental Setup and Operation Principle We build our system of coupled balloons by mounting a dielectric elastomer membrane (VHB) as TA and a rubber balloon featuring low viscoelasticity as high‐speed actuator (HSA) on a chamber of suitable volume. In a previous experiment with VHB elastomers, the snap‐through occurred on a time scale of 100 s, mainly due to the large viscoelasticity of the VHB elastomer. 15 Thus, VHB is not suitable for fast snap‐through. Natural rubber, on the other hand, has comparatively smaller viscoelasticity, enabling large volume changes over a short time. Figure \n \n 2 \n a compares the “creep‐through” of the VHB membrane with a fast snap‐through of the rubber membrane. The TA is constantly connected to a high voltage supply providing the voltage signal Φ TA (Figure 2 b). The schematic in Figure 2 b shows the cyclic process in the pressure–volume plane of the HSA ( p ‐ V \n HSA ). Each pressure–volume state of the HSA corresponds to a voltage Φ TA applied to the coupled TA. The operation of the HSA is separated into the following steps: The initial pressure of the system is set to p \n A of state A, slightly above the verge of instability of the rubber balloon in state E to enable electrically triggered, giant deformation. Then the voltage Φ TA is applied to the TA membrane. It becomes thinner, volume V \n TA increases, and the common pressure in the systems falls, consequently decreasing the volume V \n HSA of the HSA. As soon as the pressure in the system falls below p \n B , the HSA snaps back from state B to C. At state D, the voltage Φ TA and volume V \n TA reach their maximum and V \n HSA its minimum. Subsequent reduction of the voltage at the TA leads the system to the verge of instability in state E at pressure p \n E . The instability is triggered and the HSA snaps through to state F and finally returns to state A when the voltage at the TA reaches its minimum. The unstable states B and E are characterized by the pressure–volume curve of the HSA being tangent to the pressure–volume curve of the air in the whole system. 15 This is the first demonstration of harnessing an electrically triggered snap‐ back instability of a balloon actuator—a key element to achieve fast cyclic actuation. Figure 2 a) Comparison of the volume expansion rate when using elastomer membranes of low viscosity (natural rubber) and high viscosity (acrylic elastomer VHB 3M), while undergoing pressure‐controlled mechanical balloon snap‐through instability (Video S2, Supporting Information). b) Setup of a HSA remotely triggered by a coupled DEA balloon (TA) and schematic representation of a complete actuation cycle of the HSA in the pressure–volume plane. Giant volume changes occur rapidly as the HSA undergoes snap‐back (state B to C) or snap‐through (state E to F) instability. c) Sinusoidal voltage applied to the TA is used to trigger instability of a coupled balloon actuator. d) The volume change of an electrically driven DEA balloon actuator (TA) modulates e) the system pressure, f) enabling the coupled HSA to undergo instability with large volume changes. 2.2 Discussion of Measured Data In our experiment, we measured the voltage Φ TA and current I \n TA of the TA as well as the system overpressure p and determined the volumes V \n HSA and V \n TA of the TA and HSA for several cycles of operation. Despite the first two consecutive measurements being noticeably different, the behavior became reproducible after three to four cycles. Figure 2 c–f illustrates the measurement data for two consecutive cycles. By applying sinusoidal voltage (@ 0.5 Hz and 4 kV peak‐to‐peak) to the compliant electrodes, the TA deforms cyclically (Figure 2 c,d and Video S3, Supporting Information). Similar to a pumping piston, the alternating pressure caused by the TA membrane triggers the snap‐through and snap‐back instability of the HSA (Figure 2 e,f and Video S3, Supporting Information). A suitable prestretch of the TA is necessary to obtain sufficient actuation, 24 required to reach the unstable states B and E of the HSA. We observe that the sudden changes of volume and pressure during the instabilities from state B to C, or from state E to F result in distinct spikes in the current characteristics I \n TA (orange triangles with black connecting lines in Figure 2 c). They originate from the rapid alteration of the capacitance of the electroactive membrane due to the variation in geometry (area and thickness) during inflation or deflation. The observed small currents during operation permit the usage of cheap commercially available low power DC to HV converters for high voltage supply. The time t \n EF for the HSA to change from state E with volume V \n HSA,E = 13 cm 3 to state F with V \n HSA,F = 59 cm 3 is 20 ms. This results in a high volume expansion rate of 2300 cm 3 s −1 (2600 cm 2 s −1 area expansion rate in spherical membrane approximation). The total volume change of one half‐cycle between state A at volume V \n HSA,A = 88.4 cm 3 and state D with V \n HSA,D = 5.9 cm 3 is about 1398%. The snap‐through from state E to F results in a significant pressure drop and a rapid volume increase. During a snap‐back instability from state B to state C, the pressure rises rapidly and the volume drops quickly. The size of these jumps can be modified by changing the volume of the connecting chamber. 15 , 16 For subsequent theoretical analysis, the measured cyclic time‐dependent data sets are plotted parametrically in the pressure–volume plane for the HSA, as well as the voltage applied on the TA versus the volume of the HSA ( Figure \n \n 3 \n ). Figure 3 a) Measured pressure and volume data of the cyclic experiment. The system is pressurized to an initial pressure p \n A . As voltage is applied to the TA, the change of volume and pressure forces the coupled HSA to undergo the snap‐through and snap‐back instability (indicated by arrows). The characteristic states A to F of a full actuation cycle correspond to the points marked in Figure 2 b. The shapes of the balloon at state A and D of the TA and HSA are depicted for comparison. b) Measured voltage Φ TA applied on the TA plotted as a function of the volume V \n HSA of the HSA. 2.3 Theoretical Analysis The resulting blue pressure curve in Figure 3 a shows the typical hysteretic behavior of elastomer balloons. 25 This hysteresis is due to stretch‐induced crystallization (SIC) and viscoelastic effects. Its theoretical analysis is possible within the frame of microscopic theories 26 , 27 and quasi‐linear viscoelasticity, 28 but is not instructive for our purposes. The hysteresis in the pressure–volume plane of the HSA does not change the qualitative picture, though it does influence the exact numbers. We account for this effect by using different shear modulus μ HSA and J \n lim for the inflation and deflation stages (Table S1 and Figure S1a, Supporting Information). For theoretical analysis, the system is idealized by assuming the HSA and TA to be spherical with initial (unstretched) radii R \n HSA and R \n TA and thicknesses H \n HSA and H \n TA . The membranes of both balloons are taken to be incompressible and thin. When the pressure inside each balloon exceeds the atmospheric pressure p \n atm by p , the balloons deform to radii r \n HSA = λ HSA \n R \n HSA and r \n TA = λ TA \n R \n TA , where λ are the respective homogenous radial (lateral) stretches. In addition to pressure, the TA is subjected to the voltage Φ TA . We consider quasi‐static equilibrium with respect to pressure and voltage. We assume that air is an ideal gas obeying the ideal gas law\n (1) p + p atm 4 3 π R TA 3 λ TA 3 + 4 3 π R HSA 3 λ HSA 3 + V C = N k B T \nhere, V \n C is the chamber volume, N is the number of air molecules, and k \n B \n T is the temperature in energy units. The amount of air N enclosed by both balloons and chamber is fixed after the valve is closed. The deformation is assumed to be isothermal. In practice, the fast expansion and contraction of the considered prototype actuators may be closer to adiabatic processes, leading to additional effects, which will be discussed elsewhere. The adiabatic relative temperature changes, and the concomitant changes in the shear modulus Δμ/μ = Δ T / T can be estimated to be less than 1%. Such minor variations are below the accuracy of the current modeling. We account for the stiffening of the elastomer at large deformation by using a hyperelastic Gent material model 29 for the elasticity of both balloons, such that the volumetric strain energy density for equi‐biaxial deformation (configurational part of Helmholtz free energy) is of the form\n (2) W stretch λ = − μ J lim 2 ln 1 − 2 λ 2 + λ − 4 − 3 J lim \nhere, μ is the small‐strain shear modulus and J \n lim is a constant related to the stiffening of the elastomer at large deformation. 29 , 30 For the dielectric elastomer balloon, we adopt the model of ideal dielectric elastomers, 31 such that the free energy density is the sum of two parts: the elastic energy due to stretching in Equation (2) and the electrostatic energy density due to the polarization of the elastomer, W \n ele = D \n 2 / (2ε), where D is the electric displacement and ε is the absolute permittivity. For any variation of the system, the free energy of each balloon membrane varies by its respective 4 πR \n 2 \n H δW . When the charge on the electrodes of the TA varies by δQ , the applied voltage does work Φ TA \n δQ . When the radius of either balloon varies by δr , the pressure does work 4 πr \n 2 \n p δr . For the HSA, a state of equilibrium is reached when the variation of the free energy of the membrane equals to the work done by the pressure; in the case of the TA, the equilibrium state is reached, when the variation of the free energy of the membrane is equal to the combined work done by the pressure and the voltage. This results in standard N‐shaped single‐balloon pressure–volume dependences p \n HSA ( V \n HSA ) and p \n TA ( V \n TA , Φ TA ), as detailed in the Supporting Information (Equation (S2)). They correspond to the equilibrium conditions of the kinematic set of equations derived previously. 32 Fitting of the pressure–volume data for the rubber (HSA) and VHB (TA) membranes (Figure S1, Supporting Information) results in the values (Table S1, Supporting Information) used in the calculations. They comply with the data reported and used in literature. 16 The relative permittivity ε r of VHB is taken as 4.7. 33 \n The cyclic dynamic behavior shown in Figure 3 a,b is modeled in Figure \n \n 4 \n a,b, with the same notations for the key points. The solid blue curve and the dashed pink curve in Figure 4 a do not depend on the TA and visualize the p \n HSA ( V \n HSA ) dependences. They correspond to the inflation (solid blue) and deflation (dashed pink) stage—deflation has a smaller μ HSA and J \n lim value. The snapping hysteresis exists also with constant parameters, but the numbers differ significantly. The different black curves combine the p \n TA ( V \n TA , Φ TA ) dependence with Equation (1) , and do not depend on the HSA (see Equation (S5) in the Supporting Information for details). They correspond to 0 V, snap‐through, snap‐back, and the maximum value of the voltage. The parameters of the system are chosen such that the TA is always strongly stretched, and the applied voltage modulates its pressure. The equilibrium states are the intersections of the blue—or pink—and black curves. Snapping happens, when some of the common solutions disappear, i.e., when these curves become tangential. The snapping path roughly follows (dynamically modified) HSA single‐balloon curves. The p \n HSA ( V \n HSA ) dependence dictates a large pressure drop with a large volume expansion during snap‐through instability, and a smaller pressure rise with a smaller volume contraction during a snap‐back. The snapping magnitude is governed by the HSA properties, for snap‐through Δ p ≤ C μ H R HSA and V HSA , after V HSA , before ≥ λ lim 3 7 1 / 2 × 3 3 / 2 (see Equations (S8)–(S10) and S14, Supporting Information). Here, C is a constant C ≈ 12 × 7 − 7 / 6 − 3 3 / 2 λ lim − 1 ≈ 0.23 containing the reciprocal of λ lim which is the maximal possible (equal‐biaxial) stretch of the HSA according to the Gent model 2 λ lim 2 ≈ J lim . Without the effects of SIC, the hysteresis in the pressure–volume plane of the HSA vanishes and the blue and pink curves merge, but clearly, the voltage‐induced snapping hysteresis still persists, though the required voltage difference is significantly lower. Theoretical Figure 4 b explains the dynamic voltage behavior, observed in the experimental Figure 3 b; it is plotted according to the parametric procedure described in the Supporting Information. The voltages required for snapping depend on the HSA and TA properties and are of the order of Φ TA ≈ H R 2 ε λ lim TA 1 / 2 C μ H R HSA 1 / 2 (Equations (S14) and (S15), Supporting Information). The voltages only weakly depend on the shear modulus of the VHB, μ TA , because the change in the elastic pressure of the HSA between the snapping points is compensated largely by the changes in the electrostatic pressure of the strongly stretched TA. The theoretical predictions of Figure 4 are in a semi‐quantitative agreement with the experimental results from Figure 3 . The details of the theoretical analysis and the influence of various parameters are described in depth in the Supporting Information. Figure 4 Theoretical a) static pressure p and b) voltage Φ TA applied to the TA membrane as a function of the volume V \n HSA of the HSA with dimensional left‐bottom axes, and dimensionless right‐top scales (with inflation μ \n HSA value). The solid blue and the dashed pink curves represent the inflation and deflation stages and account for the material hysteresis observed in the experiment. Different dotted black curves combine p \n TA ( V \n TA ) dependences with air conservation (Equation (1) ) for different voltages Φ TA , applied to the coupled TA: 0 V, the snap‐through, the snap‐back, and the maximum value (see Equation (S5), Supporting Information). The equilibrium pressure–volume states of the HSA for these voltages are marked by red dots and correspond to states indicated in Figures 2 and 3 . Parameters are listed in Table S1 in the Supporting Information. With this good agreement between our model and our experiment, an optimized system that maximizes actuation performance can be designed in the future.\n\n2.2 Discussion of Measured Data In our experiment, we measured the voltage Φ TA and current I \n TA of the TA as well as the system overpressure p and determined the volumes V \n HSA and V \n TA of the TA and HSA for several cycles of operation. Despite the first two consecutive measurements being noticeably different, the behavior became reproducible after three to four cycles. Figure 2 c–f illustrates the measurement data for two consecutive cycles. By applying sinusoidal voltage (@ 0.5 Hz and 4 kV peak‐to‐peak) to the compliant electrodes, the TA deforms cyclically (Figure 2 c,d and Video S3, Supporting Information). Similar to a pumping piston, the alternating pressure caused by the TA membrane triggers the snap‐through and snap‐back instability of the HSA (Figure 2 e,f and Video S3, Supporting Information). A suitable prestretch of the TA is necessary to obtain sufficient actuation, 24 required to reach the unstable states B and E of the HSA. We observe that the sudden changes of volume and pressure during the instabilities from state B to C, or from state E to F result in distinct spikes in the current characteristics I \n TA (orange triangles with black connecting lines in Figure 2 c). They originate from the rapid alteration of the capacitance of the electroactive membrane due to the variation in geometry (area and thickness) during inflation or deflation. The observed small currents during operation permit the usage of cheap commercially available low power DC to HV converters for high voltage supply. The time t \n EF for the HSA to change from state E with volume V \n HSA,E = 13 cm 3 to state F with V \n HSA,F = 59 cm 3 is 20 ms. This results in a high volume expansion rate of 2300 cm 3 s −1 (2600 cm 2 s −1 area expansion rate in spherical membrane approximation). The total volume change of one half‐cycle between state A at volume V \n HSA,A = 88.4 cm 3 and state D with V \n HSA,D = 5.9 cm 3 is about 1398%. The snap‐through from state E to F results in a significant pressure drop and a rapid volume increase. During a snap‐back instability from state B to state C, the pressure rises rapidly and the volume drops quickly. The size of these jumps can be modified by changing the volume of the connecting chamber. 15 , 16 For subsequent theoretical analysis, the measured cyclic time‐dependent data sets are plotted parametrically in the pressure–volume plane for the HSA, as well as the voltage applied on the TA versus the volume of the HSA ( Figure \n \n 3 \n ). Figure 3 a) Measured pressure and volume data of the cyclic experiment. The system is pressurized to an initial pressure p \n A . As voltage is applied to the TA, the change of volume and pressure forces the coupled HSA to undergo the snap‐through and snap‐back instability (indicated by arrows). The characteristic states A to F of a full actuation cycle correspond to the points marked in Figure 2 b. The shapes of the balloon at state A and D of the TA and HSA are depicted for comparison. b) Measured voltage Φ TA applied on the TA plotted as a function of the volume V \n HSA of the HSA."
} | 6,517 |
32038137 | PMC6987377 | pmc | 1,978 | {
"abstract": "Dealing with big data, especially the videos and images, is the biggest challenge of existing Von-Neumann machines while the human brain, benefiting from its massive parallel structure, is capable of processing the images and videos in a fraction of second. The most promising solution, which has been recently researched widely, is brain-inspired computers, so-called neuromorphic computing systems (NCS). The NCS overcomes the limitation of the word-at-a-time thinking of conventional computers benefiting from massive parallelism for data processing, similar to the brain. Recently, spintronic-based NCSs have shown the potential of implementation of low-power high-density NCSs, where neurons are implemented using magnetic tunnel junctions (MTJs) or spin torque nano-oscillators (STNOs) and memristors are used to mimic synaptic functionality. Although using STNOs as neuron requires lower energy in comparison to the MTJs, still there is a huge gap between the power consumption of spintronic-based NCSs and the brain due to high bias current needed for starting the oscillation with a detectable output power. In this manuscript, we propose a spintronic-based NCS (196 × 10) proof-of-concept where the power consumption of the NCS is reduced by assisting the STNO oscillation through a microwatt nanosecond laser pulse. The experimental results show the power consumption of the STNOs in the designed NCS is reduced by 55.3% by heating up the STNOs to 100°C. Moreover, the average power consumption of spintronic layer (STNOs and memristor array) is decreased by 54.9% at 100°C compared with room temperature. The total power consumption of the proposed laser assisted STNO-based NCS (LAO-NCS) at 100°C is improved by 40% in comparison to a typical STNO-based NCS at room temperature. Finally, the energy consumption of the LAO-NCA at 100°C is expected to reduce by 86% compared with a typical STNO-based NCS at the room temperature.",
"conclusion": "Conclusion To reduce the power consumption of future STNO-based NCSs, a microwatt-nanosecond laser pulse is utilized for the first time to ease the magnetic oscillation of the STNO through heating. The power consumption of the spintronic layer and the total power consumption of the proposed LAO-NCS are improved by 54.9% and 40% at T = 100°C compared with operation at the room temperature. Moreover, 86% lower energy consumption can be expected for the LAO-NCA at 100°C compared with a typical NCS at the room temperature. It should be noted that scaling the technology and increasing the temperature above 100°C leads to further improvement of the power consumption.",
"introduction": "Introduction The grand challenge of exascale computing, 10 18 operations/second, calls for a dramatic change in hardware of the current petascale supercomputers. A paradigm shift is needed to tackle the issue of processing the explosively growing Big Data from different sources, which are mostly images and videos as the most time and power-consuming task for the existing Von-Neumann computing machines (VNCs). Filling the gap between the performance of the current computing systems and the brain requires development of a computing system with similar features as the brain; brain-inspired computing systems, so-called neuromorphic computing systems (NCSs). Such systems overcome the limitation of the word-at-a-time thinking of the VNCs by massive parallel data processing similar to the brain ( U.S. Department of energy, 2015 ; DeepMind, 2018 ; Hbp, 2018 ; Ibm, 2018 ; SpiNNaker, 2018 ). An NCS includes many parallel processors (neurons) communicating using simple messages (spikes) through programmable memory units (synapses). Although significant progress has been made in the CMOS implementation of NCSs, there are some fundamental limits to the simultaneous improvement of area and power in CMOS-based NCS ( Fong et al., 2016 ). Such limits have driven a significant effort to investigate beyond-CMOS NCSs. The spin-based devices integrated with electronics (i.e., spintronics) have opened a door for designers to implement low-power high-density NCSs. In spintronic-based NCSs, magnetic switching in magnetic tunnel junction (MTJ) ( Fong et al., 2016 ) or magnetic oscillation in spin-torque nano-oscillator (STNO) ( Yogendra et al., 2015 , 2016 ; Kurenkov et al., 2019 ) is used to mimic neuron firing. While using oscillation of magnetic moment decreases the power consumption by an order of magnitude compared with the magnetic moment switching [critical current: ∼10 6 Acm –2 ( Costa et al., 2017 ) vs. ∼10 –7 Acm –2 ( Fukami et al., 2016 )], still there is a huge gap between spintronic-based NCSs and the brain in terms of power consumption and speed. This is due to the fact that the traditional way of oscillating the magnetic moment through the bias current consumes high power and it is done at low speeds. Hence, there is a crucial need for eliminating or decreasing the bias current in spintronic-based NCSs. Magnetic tunnel junctions and STNOs can be used to perform Bayesian computation in networks inspired by cortical microcircuits of pyramidal stochastic neurons. This type of neurons spikes stochastically, observed in the cortex ( Sengupta et al., 2016 ). The membrane voltage of a cell can change from the rest potential to oscillatory mode as a result of bifurcation ( Bose, 2014 ). This is very similar to what happen inside STNOs, where the magnetization of FL starts to oscillate by increasing the current passing through the STNO to the currents higher than critical current (Hopf bifurcation). On the other hand, STNOs can show different precession modes based on their bias current (out-of-plane precession and in-plane precession with small or large angle), which are as the result of different bifurcation types, e.g., Hopf bifurcation causes in plane precession and heteroclinic bifurcation leads to out-of-plane precession ( Nakada and Miura, 2016 ). However, in this work, the STNOs with in-plane precession have been used and in order to mimic neuron firing the transition from the magnetization resting state (non-oscillating) to the magnetization oscillation is utilized. It should be noted that the STNOs cannot be used to mimic all bifurcations, for example STNOs unable to mimic SNIC (saddle node on an invariant circle) bifurcation where the f-I curve is continuous ( Bose, 2014 ). In neural networks inspired by biological behavior, the activation function represents the rate of action potential firing in the cell ( Hodgkin and Huxley, 1952 ). In this manuscript, STNOs are used to implement the binary activation function, which is widely used to implement the linear perceptrons in neural networks. The weakness of this type of activation function is that the number of neurons needed for achieving a certain amount of accuracy increases. However, the main goal of this manuscript is to investigate the proof-of-the-concept of improving the performance of the STNO-based systems by elevating the temperature of the STNOs using laser illumination. The STNO-based NCS is used as an application to explore the effectiveness of the proposed idea. In this manuscript, for the first time to our knowledge, we propose to design a laser-assisted STNO-based NCS (LAO-NCS) to improve power consumption of the state-of-the-art NCSs by at least 40%; narrowing the gap of power efficiency between the Brain and the current NCSs. Spin Torque Nano-Oscillators Basics The schematic of an STNO is shown in Figure 1A , which consists of a Pinned Layer (PL) with fixed magnetization and a free layer (FL) with changeable magnetization direction separated by a tunneling oxide layer e.g., MgO or Al 2 O 3 . Figure 1B shows the magnetization direction of the free layer (m) and different torques acting on it ( Yogendra et al., 2015 ). T P describes the precession torque that leads to the oscillation of m. T D is the damping torque that aligns m with H eff and T STT is the spin-transfer torque caused by a bias current ( Yogendra et al., 2016 ). The interaction of T STT and T D determines the oscillatory orbit of m. As T STT increases, m will be placed in an orbit farther than H eff , which will lead to a lower frequency of oscillation of m as shown in Figure 1B ( Csaba and Porod, 2013 ). It is shown experimentally and through simulation that the frequency of the STNO can be locked to the frequency of an RF current passing through it ( Rippard et al., 2005 , 2013 ) or an external oscillating RF field ( Slavin et al., 2010 ). Moreover, the frequency of two STNOs can lock if they are close to each other ( Kaka et al., 2005 ). In STNO-based NCSs, the frequency locking of the STNO and comparing its output power with a threshold power are two mechanisms used to implement neuron firing. However, in all cases, a very high DC current (i.e., bias current) is needed flowing through the STNO to generate the required torque (i.e., T STT ). FIGURE 1 (A) The schematic view of a MTJ as spin torque nano-oscillators (STNO) and (B) the magnetization direction of MTJ free layer (FL) and torques acting on it. Effect of Raising Temperature on Spin Torque Nano-Oscillators The dynamic behavior of the FL magnetic moment is modeled using Landau-Lifshitz-Gilbert-Slonczewski (LLGS) equation as follows ( Sengupta et al., 2016 ): (1) ( 1 + α 2 ) | γ | ∂ m ^ ∂ τ = - m ^ × H → E F F - α m ^ × m ^ × H → E F F + 1 q N s ( m ^ × I s × m ^ ) where, α, and m ^ are the gyromagnetic ratio, Gilbert damping factor and magnetization of the FL, respectively. H → E F F is the effective magnetic field acting on FL described by H → E F F = H → U A + H → e x + H → T F , where H → U A , H → e x , and H → T F are uniaxial anisotropy field, external magnetic field and thermal fluctuations field, respectively. N s = M S V μ B is the number of spins in the FL of volume V ( M S is the saturation magnetization and μ B is Bohr magneton) and I s is the input spin current. The first term in (1) represents the precession torque (T P ) that makes m ^ precess around the easy axis. The second term is the damping term (T D ) that tries to align m ^ with easy axis. The third term represents the transverse component of spin current being absorbed by m ^ (T STT ). In the absence of third term (no current passing through STNO), and in the equilibrium, m ^ is aligned with easy axis. By increasing the current, the third term starts to increase, and m starts to oscillate around the easy axis. Higher currents will make the m ^ to be placed on orbit farther than easy axis (i.e., higher output power). Increasing the temperature affects the dynamic behavior of the FL through decreasing the saturation magnetization (M S ) of it, decreasing the resistance of the STNO and increasing the dispersion of the initial deviation of the magnetic moment from easy axis due to higher thermal fluctuations. Saturation Magnetization It is shown theoretically ( Ashcroft and Mermin, 1976 ) and experimentally ( Alzate et al., 2014 ) that the dependency of M S can be well described by Bloch’s law as follows: (2) M S ( T ) = M S ( 0 ) ( 1 - ( T / T * ) 3 2 ) where T is the absolute temperature in Kelvin and M S (0) is the saturation magnetization at 0K, and T ∗ is a fitting factor. Equation (2) shows that increasing the temperature decreases the M S (T). This will lead to a degradation of the uniaxial anisotropy field, which decreases the minimum current required for the FL magnetic oscillation. Resistance Two tunneling mechanisms contribute to the STNO resistance including electron spin-polarized direct elastic tunneling and spin independent tunneling. The total conductance of the STNO can be described as ( Teixeira et al., 2010 ). (3) G ( θ ) = G T [ 1 + P 1 P 2 cos θ ] + G S T where θ is the angle between the magnetization of the FL and the PL. P 1 and P 2 are the effective tunneling spin polarization of the magnetic layers. G T is the pre-factor for direct elastic tunneling. All these parameters are temperature-dependent. Elevating the temperature will increase G T and reduces P 1 and P 2 ( Teixeira et al., 2010 ). As a result, R P is almost independent of temperature while R AP reduces approximately linearly with temperature. This has been experimentally shown in Teixeira et al., 2010 , Takeuchi et al., 2015 , and Hu et al., 2016 . Thermal Fluctuations The effect of temperature on random fluctuating field can be modeled by H → T F while its x , y , and z components have uncorrelated Gaussian distribution with zero mean and ( 2 α k B T ) / ( γ M S V △ t ) standard deviation ( Brown, 1963 ; Sankey et al., 2005 ; Yogendra et al., 2017 ). α, k B , γ, V, and Δt are the Gilbert damping parameter, the Boltzmann’s constant, the gyromagnetic ratio, the volume of the FL and the integration time step. Elevating the temperature will increase the dispersion of H → T F , which leads to an easier oscillation of the FL magnetic moment. In order to explore the mentioned effects on oscillation behavior of the STNO at elevated temperatures, different characteristics of the STNO (e.g., resistance, TMR, and output power of the oscillation) have been measured at different temperatures from 27°C up to 100°C in section “Memristor Behavior at Elevated Temperatures.” Memristor Behavior at Elevated Temperatures Tantalum-oxide (TaO x ) memristors are one of the best candidate in memory and NCS applications due to their unique characteristics such as CMOS compatibility ( Diokh et al., 2013 ), low power operation ( Strachan et al., 2011 ), high endurance ( Lee et al., 2011 ), and long retention of states ( Ninomiya et al., 2013 ). The conduction mechanism of the TaO x memristors can be modeled by two parallel conduction mechanisms including hopping conduction and Schottky thermionic emission ( Graves et al., 2017 ) as follows: (4) I t o t a l = 2 e l v p h N k B T e ( - W k B T ) e ( - 2 l ζ ) s i n h ( q l F k B T ) ⏟ I H o p + A T 2 e ( - ϕ B 0 - β F k B T ) ⏟ I S c h where, k B is the Boltzmann constant, I is the hopping distance, W is the hopping energy, ζ is the wave function localization, F is the applied field (converts from V), T is the temperature, v p h is the vibrational phonon frequency, A is the reduced effective Richardson constant multiplied by active device area, ϕ B o is the barrier height, and β is the barrier lowering factor. N is proportional to the density of electrons in the conduction path multiplied by the relevant conducting area. Based on this model, which is well fitted with experimental results, the temperature dependence of TaO x memristor resistance can be divided into two regions called cold and hot regions ( Graves et al., 2017 ). In the cold region ( T ≤ 350K), the state-dependent hopping conduction is dominant and the resistance of memristor is almost temperature insensitive. In the hot region, however, the Schottky emission of electrons determines the hot current, and the memristor’s resistance decreases with raising the temperature, rapidly. Note that, the amount of resistance change of memristor in hot region depends on the memristor’s initial resistance. Proposed Laser Assisted Neuromorphic Computing System Our novel envisioned LAO-NCS is shown in Figure 2 , which is a crossbar array of programmable TaO x memristors as synapses and the STNOs as neurons assisted thermally by a narrow laser-pulse. Considering the fact that in many applications, size of the memristor array is much larger than the area of the STNO-based neurons, there is no significant area improvement in stacking the memristor array on top of the STNOs. Hence, the memristor array and the STNOs are supposed to be next to each other in the proposed LAO-NCS. Moreover, this structure allows direct laser illumination on the STNOs’ top contacts. The resistance of the memristors can be tuned using an electric signal flowing through them. The NCS operation starts with a calibration phase in which the temperature of the STNOs will be elevated to 100°C and stabilized. Then, the NCS is ready for operation and the processing phase will start. The processing phase can be divided into two steps including stimulation and recovery, which will be repeated in sequence. In the stimulation step, the crossbar array sums the weighted input currents passing them to the STNOs, which are already set in AP-state (the magnetization direction of the FL and the PL are anti-parallel). In case, the weighted input currents are sufficiently large, the FL magnetic moment of the STNO starts to oscillate that will be detected by a sensing circuit immediately, and translated to neuron firing. The sensing circuit should use track and terminate method ( Farkhani et al., 2017 , 2018 ; Torrejon et al., 2017 ) in order to minimize the energy consumption of the NCS. Immediately after detecting the STNO oscillation, the recovery step begins. In the recovery step, the input corresponding to the fired neuron will be activated in the post-synaptic neuronal layer. Note that one of the advantages of using oscillation instead of magnetic moment switching is that there is no need for switching back the FL magnetization. Hence, the recovery step can be done in a very short time (∼600 ps) compared with the magnetic moment switching (∼2 ns) without extra energy consumption for switching back the magnetic moment. In our approach, the energy consumption needed for starting the STNO oscillation will be lowered significantly by increasing the temperature of the STNO using a nanosecond laser pulse. In fact, increasing the temperature of the STNO will decrease its energy barrier, which leads to a lower bias current needed for starting the oscillation in the STNO. On the other hand, in case, the temperature of TaO x memristor array increases to temperatures above 350K due to heat propagation, the resistance of memristors will decrease, as discussed in previous section. However, it seems unlikely that the memristor array temperature reaches above 350K due to limited laser power. Moreover, in order to keep the memristor temperature below 350K, a thermal insulator layer can be placed between the memristor array and the STNOs. Considering the fact that the STNO current passes through the memristor array, the total power consumption of the memristor array will be reduced, significantly. As a result, the power consumption of the LAO-NCS decreases compared with typical spintronic-based NCSs. Considering the fact that the control transistors (T ct ) act as switches, heating them up has no significant impact on the overall performance of the LAO-NCS. FIGURE 2 The schematic view of the novel LAO-NCS. The STNO and memristor act as neuron and synapse, respectively. The STNOs will be heated to 100°C by illuminating a laser pulse. Interaction Between Laser and the Spin Torque Nano-Oscillators The on-chip laser can be achieved through vertical cavity surface emitting laser (VCSEL) ( Chen et al., 2014 ; Zhou et al., 2015 ; Kozlov and Carusone, 2016 ). VCSEL’s unique specification is that, in contrast to the conventional edge-emitting semiconductor lasers, its laser beam is emitted perpendicular to its surface, which makes it a proper candidate for on-chip laser applications including the LAO-NCS. The output power of VCSEL can be tuned through changing the supply voltage of its driver ( Kozlov and Carusone, 2016 ). Hence, in order to control the output power of the laser, a CMOS interface circuit is designed, which is described below. CMOS Interfacing Circuit Figure 3A shows the block diagram of the proposed LAO-NCS. The spintronic layer includes a memristors array, STNOs, T c , and a sensing circuit. The CMOS interface circuit adjusts the output laser power by manipulating the supply voltage of the laser diode driver (LDD). In this way, the CMOS interfacing block can control the STNO temperature in the spintronic layer. Figures 3B,C show the circuit design of the CMOS interfacing block and its timing diagram, respectively. As mentioned before, the LAO-NCS operating time can be divided to calibration and processing phases. In the calibration phase, the temperature of the STNO is increased from 27°C to 100°C (first laser pulse with high power), and stabilized at this temperature (second laser pulse with low power). In the processing phase, the STNO temperature will be kept at 100°C with a sequence of low power laser illuminations as shown in Figure 3C . The operation of the CMOS interfacing circuit is as follows. The counter is clocked with a 500 MHz clock and generates the b0, b1, and b2 signals. Then, the logic circuit generates the V LDD signal from the output of the counter. During the first pulse of V LDD , the level shifter is enabled by a logic circuit and the voltage of V LDD will be set at V DDH that leads to a high output power laser pulse. During the next pulses, the transmission gate is enabled and the level shifter is disabled by the logic circuit. Hence, the voltage of the V LDD node is at V DD and the laser output power will be lower. FIGURE 3 (A) The block diagram of the proposed LAS-NCS including CMOS interfacing block, VCSEL array, LDD, and the spintronic layer. (B) CMOS interfacing circuit design. (C) Timing diagram of the CMOS interfacing block. Neurons’ Readout Approach The sensing circuit is to sense the magnetization oscillation of the STNOs (neurons) in order to find the fired neuron(s) and activate the corresponding input(s) in the post-synaptic neuronal layer. This can be done either by sensing the frequency or the output power of the oscillating signal across the STNOs, and comparing it with a threshold frequency or a threshold output power. Figures 4A,B shows the measured frequency and the output power of our STNO samples in response to different bias currents. At bias currents lower than 60 μA, the output power of oscillation is very low. As a result, the frequency of oscillation is not detectable. By increasing the bias current, the frequency of oscillation decreases. However, the frequency reduction rate is slow (just 10% frequency reduction at 600 μA). Hence, it is difficult to detect the fired neuron by comparing the frequency of oscillation with a reference frequency. In contrast, thanks to the advances in power detector (PD) circuits, signals with few nano-Watt output power are detectable within few nano-seconds and with micro-Watts power consumption ( Li et al., 2010 ; Qayyum and Negra, 2017 ). Hence, the output power of oscillation is used to detect the oscillating STNO. FIGURE 4 The measured (A) frequency and (B) output power of the STNO versus different bias current from 0 to 600 μA. The frequency of oscillation is unrecognizable from noise at I Bias <60 μA due to low output power of oscillation. The maximum frequency change is 10% @ I Bias =600 μA. Output powers higher than 10 nW are detectable by sensing circuits. (C) The schematic view of the neuron firing detection approach. The schematic view of the sensing approach is shown in Figure 4C . The current of memristor array passing through the STNO leads to its resistance oscillation. As a result, a weak signal with milli-Volt amplitude oscillating at GHz frequency will appear across the STNO. This weak AC signal, first, will be amplified by a low noise amplifier (LNA). Then, the output signal of LNA will be converted to a DC voltage by the PD. The output voltage of the PD will be compared with a threshold voltage by the comparator. In case, I Mem. passing through the STNO will be high enough, the DC output voltage of the PD becomes higher than the threshould voltage. Hence, the output voltage of the comparator switches from “0” to “1” and it will be considered as neuron firing."
} | 5,977 |
34776834 | PMC8586528 | pmc | 1,982 | {
"abstract": "Spiking neural networks (SNNs), with their inherent capability to learn sparse spike-based input representations over time, offer a promising solution for enabling the next generation of intelligent autonomous systems. Nevertheless, end-to-end training of deep SNNs is both compute- and memory-intensive because of the need to backpropagate error gradients through time. We propose BlocTrain, which is a scalable and complexity-aware incremental algorithm for memory-efficient training of deep SNNs. We divide a deep SNN into blocks, where each block consists of few convolutional layers followed by a classifier. We train the blocks sequentially using local errors from the classifier. Once a given block is trained, our algorithm dynamically figures out easy vs. hard classes using the class-wise accuracy, and trains the deeper block only on the hard class inputs. In addition, we also incorporate a hard class detector (HCD) per block that is used during inference to exit early for the easy class inputs and activate the deeper blocks only for the hard class inputs. We trained ResNet-9 SNN divided into three blocks, using BlocTrain, on CIFAR-10 and obtained 86.4% accuracy, which is achieved with up to 2.95× lower memory requirement during the course of training, and 1.89× compute efficiency per inference (due to early exit strategy) with 1.45× memory overhead (primarily due to classifier weights) compared to end-to-end network. We also trained ResNet-11, divided into four blocks, on CIFAR-100 and obtained 58.21% accuracy, which is one of the first reported accuracy for SNN trained entirely with spike-based backpropagation on CIFAR-100.",
"conclusion": "7. Conclusion End-to-end training of deep SNNs is memory-inefficient due to the need to perform error BPTT. In this work, we presented BlocTrain, which is a scalable block-wise training algorithm for deep SNNs with reduced memory requirements. During training, BlocTrain dynamically categorized the classes into easy and hard groups, and trained the deeper blocks only on the hard class inputs. In addition, we introduced a hard class detector per block to enable fast inference with early exit for the easy class inputs and conditional activation of deeper blocks only for the hard class inputs. Thus, BlocTrain provides a principled methodology to determine the optimal network size (in terms of number of layers) for a given task, depending on the accuracy requirements. We demonstrated BlocTrain for deep SNNs trained using spike-based BPTT, on the CIFAR-10 and the CIFAR-100 datasets, with higher accuracy than end-to-end training method. Future works could further improve the effectiveness of BlocTrain by using more complex methods for determining the hard classes, such as considering the false positives and negatives aside from the class-wise accuracy. Also, the local discriminative loss, which is used to separately train the individual blocks, could be augmented with other local losses as proposed in Nøkland and Eidnes ( 2019 ). Finally, well-established methods like neural architecture search could be used for selecting the BlocTrain hyperparameters such as the hardness threshold.",
"introduction": "1. Introduction Deep neural networks have achieved remarkable success and redefined the state-of-the-art performance for a variety of artificial intelligence tasks including image recognition (He et al., 2016 ), action recognition in videos (Simonyan and Zisserman, 2014a ), and natural language processing (Bahdanau et al., 2014 ; Sutskever et al., 2014 ), among other tasks. We refer to modern deep neural networks as analog neural networks (ANNs) since they use artificial neurons (sigmoid, ReLU, etc.) that produce real-valued activations. ANNs attain superhuman performance by expending significant computational effort, which is believed to be much higher compared to the human brain. The quest for improved computational efficiency has led to the emergence of a new class of networks known as spiking neural networks (SNNs) (Maass, 1997 ), which are motivated by the sparse spike-based computation and communication capability of the human brain. The salient aspect of SNN is its ability to learn sparse spike-based input representations over time, which can be used to obtain higher computational efficiency during inference in specialized event-driven neuromorphic hardware (Merolla et al., 2014 ; Davies et al., 2018 ; Blouw et al., 2019 ). Supervised training of SNNs is challenging and has attracted significant research interest in recent years (Lee et al., 2016 , 2020 ; Bellec et al., 2018 ; Jin et al., 2018 ; Shrestha and Orchard, 2018 ; Wu et al., 2018 ; Neftci et al., 2019 ; Thiele et al., 2020 ). Error backpropagation algorithms, which are the workhorse for training deep ANNs with millions of parameters, suffer from scalability limitations when adapted for SNNs. It is well known that end-to-end training of feed-forward ANNs, using backpropagation, requires the activations of all the layers to be stored in memory for computing the weight updates. SNNs, by virtue of receiving input patterns converted to spike trains over certain number of time-steps, require multiple forward passes per input. As a result, spike-based backpropagation algorithms need to integrate error gradients through time (Neftci et al., 2019 ). The ensuing weight update computation requires the spiking neuronal activation and state (also known as membrane potential) to be stored across time-steps for the entire network. SNNs are typically trained for hundreds of time-steps to obtain high enough accuracy for visual image recognition tasks (Lee et al., 2020 ). Hence, end-to-end training of SNN using backpropagation through time (BPTT) requires much higher memory footprint over that incurred for training similarly sized ANN on Graphics Processing Units (GPUs) (Gruslys et al., 2016 ). In this work, we propose input complexity driven block-wise training algorithm, referred to as BlocTrain , for incrementally training deep SNNs with reduced memory requirements compared to that incurred for end-to-end training. We divide a deep SNN into blocks, where each block consists of few convolutional layers followed by a local auxiliary classifier, as depicted in Figure 1 . We train the blocks sequentially using local losses from the respective auxiliary classifiers. For training a particular block, we freeze the weights of the previously trained blocks and update only the current block weights using local losses from the auxiliary classifier. The proposed algorithm precludes the need for end-to-end backpropagation, thereby considerably reducing the memory requirements during training, albeit with overhead incurred due to the addition of a classifier per block. Next, we present a systematic methodology to determine the optimal SNN depth for a given application based on the target accuracy requirements. New blocks are added only if the accuracy of prior blocks (obtained on the validation set) is lower than the desired accuracy. Further, the newly appended blocks are trained only on the “hard” classes as summarized below. Once a particular block is trained, we subdivide the classes into “easy” and “hard” groups based on the class-wise accuracy on the validation set. We incorporate and train a HCD in the following block to perform binary classification between the “easy” and the “hard” class inputs. The next deeper block is now trained only on the hard class instances, as illustrated in Figure 1 . Previous works on class complexity aware training built hierarchical classifier models, where the initial layers classify the inputs into coarse super-categories while the deeper layers predict the finer classes, which require end-to-end training and inference (Srivastava and Salakhutdinov, 2013 ; Yan et al., 2015 ; Panda et al., 2017a ). On the other hand, BlocTrain significantly minimizes the training effort with increasing block depth due to gradual reduction in the number of output classes. During inference, we obtain improved computational efficiency by using the HCD per block to terminate early for easy class inputs and conditionally activate deeper blocks only for the hard class inputs. The higher inference efficiency is achieved with increased memory requirement owing to the use of nonlinear auxiliary classifiers. We demonstrate the capability of BlocTrain to provide improved accuracy as well as higher training (compute and memory) and inference (compute) efficiency relative to end-to-end approaches for deep SNNs on the CIFAR-10 and the CIFAR-100 datasets. Note that BlocTrain , although demonstrated in this work for SNNs, can be directly applied for ANNs to achieve efficient conditional training and inference. Overall, the key contributions of our work are as follows: We propose a scalable training algorithm for deep SNNs, where the block-wise training strategy can help alleviate the larger memory requirement, which is bound by hardware limitations, and gradient propagation issues incurred by end-to-end training. We present a systematic methodology to determine the optimal network size (in terms of number of layers) for a given dataset based on the accuracy requirements, since new layers are added and trained sequentially until the desired accuracy is achieved. We improve the latency and compute efficiency during inference, which is achieved by using the HCD to exit early for the easy class instances and activate the deeper blocks only for the hard class instances. Figure 1 Illustration of BlocTrain methodology for block-wise input complexity aware training of deep SNNs. The blocks are trained sequentially using local losses from the respective classifiers. Each block B i has a hard class detector that is trained to perform binary classification between the easy ( E i −1 ) and the hard class instances ( H i −1 ), as determined from the preceding block. The next block B i +1 is trained only on the hard class instances ( H i −1 ). This process is repeated for every block, leading to fast learning with increasing block depth.",
"discussion": "6. Discussion 6.1. BlocTrain Hyperparameters Heuristics In this section, we present the heuristics for setting the BlocTrain hyperparameters, namely, the hard-class accuracy threshold, also referred to as the class hardness threshold ( Acc hard − thresh in Algorithm 1) and the softmax classifier confidence threshold (θ conf in Algorithm 3). The choice of these hyperparameters directly impacts the trade-off among memory overhead, compute efficiency, and test accuracy, as illustrated in Figures 6 , 7 . Our experiments using ResNet-9 on CIFAR-10 ( Figure 6 ) and ResNet-11 on CIFAR-100 ( Figure 7 ) establishes the following key heuristics and trends on the hardness threshold. First, the hardness threshold is experimentally found to be bounded within the range [μ acc −σ acc , μ acc +σ acc ], where μ acc is the mean and σ acc is the standard deviation of the class-wise accuracies on the validation set to obtain favorable trade-off among memory overhead, compute efficiency, and test accuracy. Second, higher the hardness threshold, larger is the memory overhead, lower is the compute efficiency, and better is the test accuracy. For ResNet-9 on CIFAR-10, we fixed the hardness threshold to 95.5%, which is roughly equal to the experimental lower bound of μ acc −σ acc , where μ acc and σ acc are 96.79 and 1.43%, respectively, calculated using the class-wise accuracies reported in Figure 4 . For CIFAR-10, using the lower bound on the hardness threshold provided favorable memory overhead-test accuracy trade-off since there were only 10 classes with clear separation between the easy and the hard classes, as illustrated in Figure 4 . On the other hand, for ResNet-11 on CIFAR-100, we experimented with hardness thresholds of 90.5–93%, which is roughly in the range of μ acc to μ acc +σ acc . Setting the hardness threshold closer to μ acc categorized roughly 50 classes as hard (refer to Figure 7A ) based on the validation accuracy of the first trained block in ResNet-11. Lowering the hardness threshold any further would provide <50% of the total number of classes for the deeper block. Hence, we did not investigate hardness thresholds much lower than μ acc . On the contrary, setting the hardness threshold to 93% (~μ acc +σ acc ) categorized close to 80 classes as hard, leading to higher memory overhead and lower compute efficiency relative to that achieved with hardness threshold of 92% (~μ acc +0.5*σ acc ). Hence, for any network to be trained on a complex dataset such as CIFAR-100 with a mix of easy and hard classes, setting the hardness threshold closer to μ acc +0.5*σ acc should yield favorable trade-offs among memory overhead, compute efficiency, and accuracy. However, if all the class probabilities are similar and the class-wise validation accuracies are high, it implies that the dataset has mostly “easy” classes, and hence, the hardness threshold can be set to the lower bound. On the other hand, if the class probabilities are similar and the class-wise validation accuracies are low, then the dataset has predominantly “hard” classes, and hence, the hardness threshold could be set closer to the upper bound. Thus, the hardness threshold, per se , does not introduce additional complexity during the training process. As far as the softmax classifier confidence threshold (θ conf ) is concerned, we investigated values ranging from ln(10 −2 ) to ln(10 −6 ) in logarithmic scale. Our experimental results across the CIFAR-10 and the CIFAR-100 datasets indicate that θ conf of ln(10 −3 ) or ln(10 −4 ) yields favorable compute efficiency-accuracy trade-off. Hence, the choice of θ conf should not require extensive experimentation to identify the optimal threshold. 6.2. Blocking Strategy for Deeper SNNs For the SNNs analyzed in this work, namely, ResNet-9 and ResNet-11, we divided the network at the granularity of a residual block and, consequently, inserted an auxiliary classifier for every residual block. Much deeper networks such as VGG-19 (Simonyan and Zisserman, 2014b ) and ResNet-34 (He et al., 2016 ) could be divided at the granularity of few VGG and residual blocks, respectively, to minimize the overhead stemming from the extra softmax layer while limiting the gradient flow to a few layers for stable training using spike-based BPTT. A more principled approach could be to take into account the memory and computational cost of adding a classifier after a certain block and the fraction of instances reaching the block (obtained from the HCD of the prior classifier block) for guiding the placement process as proposed in Panda et al. ( 2017b ). Such a principled methodology will help avoid inserting too many classifiers, and at the same time help determine the optimal network size for a given dataset based on the accuracy requirements. 6.3. Comparison With Early Inference The proposed BlocTrain method categorizes the classes as hard or easy, and trains deeper blocks only on the hard class instances. Inference is terminated at the earlier blocks for easy class instances while the deeper blocks are activated only when hard class instances are detected. It is important to note that BlocTrain attributes uniform hardness (or significance) to all instances of any given class. In practice, the hardness might not be uniform across all instances of a class, as noted in prior works (Panda et al., 2016 ; Teerapittayanon et al., 2016 ), which categorized individual instance as hard or easy irrespective of the general difficulty of the corresponding class. Therefore, we set forth to compare the efficacy of BlocTrain with respect to baseline method, designated as BlocTrain-base, wherein every block is trained on all the classes. Inference is terminated at a particular block based on the classifier confidence, that is, if the classifier prediction probability is higher than a specified confidence threshold (θ conf ). The BlocTrain-base method effectively classifies easy instances, belonging to any class, at the earlier blocks and activates the deeper blocks only for hard instances. For the proposed BlocTrain method, the original CIFAR-10 or CIFAR-100 dataset, containing 50,000 images, is split into training set of 40,000 images and validation set of 10,000 images. The validation set is used to subdivide the classes into easy and hard groups, as noted in section 5.1. On the contrary, the entire dataset is used for BlocTrain-base since each of the blocks is trained on all the classes. The classifier confidence threshold is set to unity for all the blocks, which causes inference to be terminated at a given block only if the prediction is obtained with 100% confidence. Setting the confidence threshold to unity yields the best test accuracy since it encourages more instances to be classified at the deeper blocks. We first present the training efficiency results followed by inference accuracy-efficiency trade-off provided by BlocTrain compared to the BlocTrain-base method. BlocTrain offers reduced or comparable training time (or effort) with increasing block depth. On the contrary, the training time increases steadily with block depth for BlocTrain-base, as shown in Figures 10A,B for ResNet-9 (on CIFAR-10) and ResNet-11 (on CIFAR-100), respectively. BlocTrain-base incurs higher training effort compared to BlocTrain due to the following couple of reasons. First, BlocTrain-base uses the entire training dataset while BlocTrain divides the original dataset into separate training and validation sets. Second, BlocTrain-base trains every block on all the class instances while BlocTrain uses only the hard class instances for deeper blocks. Despite the higher training effort, BlocTrain-base offers 88.31% test accuracy for ResNet-9 SNN on CIFAR-10, which is higher than an accuracy of 86.4% provided by BlocTrain. For ResNet-11 SNN on CIFAR-100, BlocTrain-base offers 62.03% accuracy, which is even higher compared to an accuracy of 58.33% provided by BlocTrain. The higher accuracy provided by BlocTrain-base can be attributed to the following factors. First, BlocTrain-base uses the entire original dataset for training all the blocks. Second, BlocTrain-base enables the harder instances in every class to be executed at the deeper blocks, resulting in higher accuracy. On the contrary, BlocTrain classifies both the easy and the hard instances of an “easy” class in the earlier blocks, leading to relatively inferior accuracy. The superior accuracy offered by BlocTrain-base is obtained with 8.5% and 7.8% higher computational effort (in terms of number of synaptic operations per inference) for ResNet-9 (on CIFAR-10) and ResNet-11 (on CIFAR-100), respectively. This is because BlocTrain-base classifies a larger fraction of hard instances at the ultimate block, as shown in Figures 10C,D . In summary, BlocTrain-base offers higher accuracy compared to BlocTrain, albeit, with longer training time and higher computational effort during inference. Figure 10 (A) Training time per epoch incurred by successive blocks of ResNet-9 spiking neural network (SNN), trained on CIFAR-10 using BlocTrain, wherein deeper blocks are trained on hard classes, and BlocTrain-base, wherein deeper blocks are trained on all classes. (B) Training time per epoch incurred by successive blocks of ResNet-11 SNN, trained using BlocTrain and BlocTrain-base methods, on the CIFAR-100 dataset. (C) Percentage of exiting inputs per block for ResNet-9 SNN, trained using BlocTrain and BlocTrain-base methods, on the CIFAR-10 dataset. (D) Percentage of exiting inputs per block for ResNet-11 SNN, trained using BlocTrain and BlocTrain-base methods, on the CIFAR-100 dataset. 6.4. Comparison With End-to-End Training 6.4.1. Accuracy Comparison Deep SNNs consisting of 7–11 layers, trained using end-to-end spike-based backpropagation approaches, have been shown to achieve >90% accuracy on CIFAR-10, as shown in Table 1 . These networks are trained end-to-end with different surrogate gradient approximations, for the discontinuous spiking nonlinearity, than the one used in this work. The various surrogate gradient-based backpropagation approaches can be readily integrated into BlocTrain to further improve its efficacy. In the ANN domain, Mostafa et al. ( 2018 ) performed layer-wise training of 10-layer deep ANN using only the local discriminative loss and reported best accuracy of ~83% on CIFAR-10. BlocTrain, on account of block-wise rather than layer-wise training, provides much higher accuracy on CIFAR-10. On the other hand, very few works have reported CIFAR-100 accuracy for SNN trained entirely with spike-based BPTT, as noted in Table 2 . Thiele et al. ( 2020 ) reported 64.69% accuracy for 8-layer deep SNN, wherein the training was performed on an equivalent ANN using the proposed SpikeGrad algorithm. Interestingly, Ledinauskas et al. ( 2020 ) trained ResNet-50 using end-to-end spike-based backpropagation and obtained 58.5% accuracy, which is comparable to that provided by ResNet-11 and lower than that obtained with VGG-16, trained using BlocTrain. Table 1 Accuracy of spiking neural network (SNN) trained using BlocTrain and end-to-end spike-based backpropagation through time (BPTT) methods, and SNN/analog neural network (ANN) trained using only the local losses, on the CIFAR-10 dataset. \n Model \n \n Training method \n \n Dataset size \n \n %Accuracy \n CIFARNet w/ 7 layers (Wu et al., 2019 ) End-to-end STBP (Wu et al., 2018 ) 50,000 90.53 ResNet-9 (Lee et al., 2020 ) End-to-end Spike BP 50,000 90.35 SNN w/ 8 layers (Thiele et al., 2020 ) End-to-end ANN-based SpikeGrad 50,000 89.72 ResNet-11 (Ledinauskas et al., 2020 ) End-to-end Spike BP 50,000 90.2 VGG-16 (Rathi et al., 2020 ) ANN-SNN and end-to-end STDB 50,000 91.13 VGG-16 (Zhou et al., 2020 ) Direct end-to-end BP 50,000 92.68 SNN w/ 4 layers (Panda and Roy, 2016 ) Local AutoEncoder 50,000 70.16 ANN w/ 10 layers (Mostafa et al., 2018 ) Local training 50,000 ~83 \n ResNet-9 (our work) \n \n BlocTrain \n \n 40,000 \n \n 86.4 \n \n ResNet-9 (our work) \n \n BlocTrain-base \n \n 50,000 \n \n 88.31 \n The bold values are used to highlight the results reported in this work over prior works . Table 2 Accuracy of spiking neural network (SNN) trained using BlocTrain and end-to-end spike-based backpropagation through time (BPTT) methods on the CIFAR-100 dataset. \n Model \n \n Training method \n \n Dataset size \n \n %Accuracy \n SNN w/ 8 layers (Thiele et al., 2020 ) End-to-end ANN-based SpikeGrad 50,000 64.69 VGG-11 (Rathi et al., 2020 ) ANN-SNN and end-to-end STDB 50,000 67.87 ResNet-50 (Ledinauskas et al., 2020 ) End-to-end Spike BP 50,000 58.5 \n ResNet-11 (our work) \n \n BlocTrain \n \n 40,000 \n \n 58.21 \n \n ResNet-11 (our work) \n \n BlocTrain-base \n \n 50,000 \n \n 62.03 \n \n VGG-16 (our work) \n \n BlocTrain \n \n 50,000 \n \n 61.65 \n The bold values are used to highlight the results reported in this work over prior works . Finally, we note that prior works have demonstrated much deeper SNNs, with competitive accuracy, for CIFAR-10, CIFAR-100, and ImageNet datasets, using either standalone ANN–SNN conversion (Rueckauer et al., 2017 ; Sengupta et al., 2019 ; Han and Roy, 2020 ; Han et al., 2020 ) or a combination of ANN–SNN conversion and spike-based BPTT methods (Rathi et al., 2020 ; Wu et al., 2020 ). The hybrid approach initializes the weights and firing thresholds of the SNN using the trained weights of the corresponding ANN, and then performs incremental spike-based BPTT to fine-tune the SNN weights. Such a hybrid SNN training methodology can be incorporated into BlocTrain to achieve further improvements in accuracy on standard vision datasets. However, the primary objective of our work is to improve the training and inference capability of deep SNN for event-driven spatiotemporal inputs, such as those produced by dynamic vision sensors (Lichtsteiner et al., 2008 ), which could potentially require exclusive spike-based training to precisely learn the input temporal statistics. We demonstrated higher accuracy using BlocTrain over end-to-end spike-based BPTT methods on CIFAR-10 and CIFAR-100 data, mapped to spike trains, which indicates the capability of BlocTrain to scale to deep SNNs for complex event-based inputs. 6.4.2. Training Time Comparison The training time incurred by BlocTrain, relative to end-to-end training, depends on the training hardware memory limitations. We evaluated the training time on two different GPU configurations, namely, Nvidia GeForce GTX and RTX GPUs. The GeForce GTX GPU, on account of higher memory capacity, could sustain the same batch size of 64 for both BlocTrain and end-to-end training methods. Figure 11A indicates that ResNet-9 SNN and ResNet-11 SNN, trained using BlocTrain on the GeForce GTX GPU, incurs 1.13× and 1.22× longer training time, respectively, compared to end-to-end training. The longer training time incurred by BlocTrain over end-to-end training, when the same batch size is used for both the methods, can be attributed to the following twofold reasons. BlocTrain requires multiple forward passes per block during training, as detailed below for ResNet-9 SNN, consisting of 3 blocks. Block 1 incurs 3 separate forward passes for individually training each of the blocks. The second block incurs 2 forward passes to train Block 2 and Block 3 . The third and final block entails a single forward pass to train Block 3 . On the other hand, end-to-end training incurs only a single forward pass for all the blocks. Each block in the original network has an additional nonlinear classifier that needs to be trained. Figure 11 (A) Normalized training time per epoch of ResNet-9 and ResNet-11, trained using BlocTrain, relative to end-to-end training on the (A) GeForce GTX GPU and (B) GeForce RTX GPU. Next, we evaluated the training times for BlocTrain and end-to-end training on the GeForce RTX GPU, which has relatively lower memory capacity. BlocTrain, by virtue of higher memory efficiency, could be used to train both ResNet-9 and ResNet-11 with a batch size of 64. End-to-end training, on account of hardware memory limitation, necessitated the batch size to be reduced to 60. Smaller batch size leads to higher number of batches (or iterations) per training epoch. As a result, BlocTrain incurs comparable training time for ResNet-9 and 0.85× shorter training time for ResNet-11 SNN over end-to-end training. For much deeper networks, the larger memory requirement needed for end-to-end training could either preclude SNN training or cause the batch size to be much smaller than that used for BlocTrain, depending on the hardware memory limitations. In the case that end-to-end training uses comparatively smaller batch size, BlocTrain would be both training time and memory efficient, as shown in Figure 11B ."
} | 6,719 |
31649246 | PMC6813331 | pmc | 1,983 | {
"abstract": "The soil microbiome is highly diverse and comprises up to one quarter of Earth’s diversity. Yet, how such a diverse and functionally complex microbiome influences ecosystem functioning remains unclear. Here we manipulated the soil microbiome in experimental grassland ecosystems and observed that microbiome diversity and microbial network complexity positively influenced multiple ecosystem functions related to nutrient cycling (e.g. multifunctionality). Grassland microcosms with poorly developed microbial networks and reduced microbial richness had the lowest multifunctionality due to fewer taxa present that support the same function (redundancy) and lower diversity of taxa that support different functions (reduced functional uniqueness). Moreover, different microbial taxa explained different ecosystem functions pointing to the significance of functional diversity in microbial communities. These findings indicate the importance of microbial interactions within and among fungal and bacterial communities for enhancing ecosystem performance and demonstrate that the extinction of complex ecological associations belowground can impair ecosystem functioning.",
"introduction": "Introduction Microbes are the unseen majority on Earth and comprise a large portion of life’s genetic diversity 1 – 3 . A multitude of microorganisms associate with humans, animals, insects, plants, and soils around the globe 4 – 8 . In each of these biomes, microbes usually form highly diverse and complex communities that collectively function as a microbiome. Earlier studies focused on the description of these microbial communities, but currently there is much interest to link microbiome composition and diversity to function 1 , 9 , 10 . This is not surprising because it is well known that microbes impact all living organisms and play a central role in many biogeochemical cycles on earth, driving global carbon and nutrient cycling with direct feedback effects on ecosystem functioning and productivity 1 – 3 . Experiments carried out in microcosms 11 – 17 and at global observational scales 18 , 19 revealed that microbial diversity is linked to ecosystem functioning, implying that communities with higher microbial richness perform better. The extremely high microbial diversity on small spatial scales has led to hypotheses that these highly diverse microbiomes are functionally redundant 20 . Yet, functional redundancy is an important feature of biodiversity as greater diversity provides a greater likelihood that some species are present that can perform a function under temporally and spatially varying conditions and buffers functioning against the loss taxa so that ecosystem functioning is maintained 21 – 23 . Furthermore, although such a vast soil microbial diversity may appear to be functionally redundant, microbes are involved in multiple functions simultaneously and thus functional redundancy is likely to fade as more functions are considered, as has been shown for plant richness–multifunctionaliy relationships 24 , 25 . To understand how changes in soil biodiversity affect ecosystem functioning it is therefore important to consider not only whether the total number of taxa present relates to a function, but how the reduction in the number of species that support a single function relates to the loss of multiple functions simultaneously. Importantly the influence of an individual species on an ecosystem function is not independent of other species present and is a result of a myriad of positive and negative, direct and indirect associations among the different species that as a whole drive ecosystem functioning. For instance, microbial communities are not only characterized by the number and composition of taxa, but also by the ecological associations among microbiome members. In recent years, microbial co-occurrence analyses have shed light on microbiome complexity and the interrelationships among community members 26 . Emerging studies have revealed that microbiomes are structured, and form complex interconnected microbial networks 26 – 31 , where microbes associate with each other directly or indirectly through processes, such as competition, facilitation, and inhibition. The complexity of these microbial networks and its relation to function is not necessarily determined by the number of taxa in the community, but rather by the number of associations that those taxa share amongst them 31 . A next frontier is now to empirically test whether changes in microbiome complexity, as indicated by both the diversity and interconnectivity among co-occurring microbes, is important for the way microbial communities affect ecosystem functioning. By fractionating soil organisms according to size, using filters of decreasing mesh size we have previously shown that the loss of soil biodiversity resulted in reduced plant diversity, productivity, nutrient retention, and belowground carbon allocation using self-contained grassland microcosms that restrict external contamination 15 . However, the role of microbiome diversity, functional redundancy, and network complexity within and among bacterial and fungal communities in regulating ecosystem performance has not been assessed along such a soil biodiversity gradient. Thus, we capitalize on this model system with a strong gradient in soil biodiversity here to further assess these different features of soil microbial diversity and their relationship with 10 soil functions that are known to be mediated by soil microbes 1 – 3 , and that reflect nutrient cycling efficiency, here termed soil multifunctionality 25 , 32 . We used soil collected from these microcosms and used next generation sequencing to characterize the fungal and bacterial soil microbiome (see the “Methods” section). Although next generation sequencing tools have allowed us to capture a vast diversity of soil microbes, many of the taxa detected may not play a significant role in the ecosystem functions of interest, thus resulting in ‘noise’ that may obscure the realized diversity–function relationship. This is in contrast to classic plant diversity–productivity relationship where each plant present inherently contributes biomass to the net ecosystem productivity. Thus, we used feature selection, a statistical tool, to identify taxa that contribute to predicting the performance of each ecosystem function considered (see the “Methods” section 33 ). This provided us with the identities of fungal and bacterial taxa that support a function (directly or indirectly), thereby removing such ‘noise’ in assessing diversity–function relationships. The association of microbial taxa to functions then allowed us to further assess the effects of greater microbiome diversity on increasing the redundancy of taxa that support a common function, where greater redundancy means that there are a greater number of taxa that support the same function. We also quantified the functional diversity within the microbial communities using the functional uniqueness index, which is the product of Raos quadradic entropy and the inverse Simpsons index and summarizes the diversity in the relative abundance among microbes that support different functions 34 . Here we hypothesize (1) that microbiome richness and microbial network complexity promotes ecosystem multifunctionality, ( 2 ) that if a given ecosystem function is not the result of the presence of a single taxon, then having more taxa present that positively contribute, directly or indirectly, to the underlying processes that drive a response in a function should lead to a positive redundancy–function relationship. At the same time, if greater microbial richness enhances ecosystem multifunctionality then it would be hypothesized that this is because (3) greater richness provides a greater diversity of taxa that support multiple different functions resulting in a positive functional uniqueness–ecosystem multifunctionality relationship. We assessed microbiome complexity by first generating a meta-association matrix including all fungal and bacterial taxa from all microcosms. We used a cross-validation and a graphical model inference framework to define the most parsimonious links among the taxa 35 (see the “Methods” section). From this, sub-networks based on taxa present in specific microcosms were used to generate indices of soil microbiome complexity (linkage density) among fungal and bacterial taxa. This work demonstrates that more complex microbial networks contribute more to improved ecosystem function multifunctionality than simpler or low-diversity networks. Moreover, different microbes support different functions pointing to the significance of functional diversity within microbial communities.",
"discussion": "Discussion Soils harbor a vast diversity of microbes 1 – 3 and recent studies have identified the drivers of microbiome composition, network association patterns, and complexity in a wide range of ecosystems 1 , 6 – 8 , 26 , 27 . However, there is a pressing need for moving beyond mere descriptions of microbial community composition and delving into the functional implication of compositional patterns and changes in microbial network structure as was highlighted recently 1 , 10 , 11 . In particular, while a large number of studies employing microbial network analysis have enriched our understanding of microbial co-occurrence patterns in various soil ecosystems 26 – 31 , very little is known of whether differences in the structure of microbial networks have consequences for microbiome functioning. Although earlier social network studies have linked network structure to functional complexity 36 , 37 , the task of relating microbial community structure to function is a non-trivial one largely due to the contentious nature of structure–function relationship that has perplexed microbial ecologists for the last two decades 38 – 41 . To our knowledge, this is one of the first studies to link microbial network complexity to ecosystem multifunctionality. Our results reveal that while taxonomic richness is an important feature that drives multifunctionality it does so because richness supports greater microbiome complexity and interkingdom associations (here by considering fungi and bacteria simultaneously). Thus, combining these microbiome characteristics can enhance our assessment of the attributes of soil microbiome diversity that explain an aggregate of process functions, i.e., soil multifunctionality. This adds a new dimension to earlier observations that soil biodiversity and microbial richness act as a driver of soil multifunctionality 18 , 19 . This study also shows that the impact of microbial communities and their taxonomic richness on ecosystem functioning may be better understood by considering various microbiome characteristics that may include: a) functional redundancy—here the increasing number of taxa that support a common function, b) the diversity of taxa that support different functions and c) the complexity of associations among taxa that support functioning. This is because here we show that greater microbiome richness is needed to support (1) greater functional redundancy to secure individual functions and (2) greater functional diversity to support multiple functions simultaneously. There are a number of experimental studies which manipulated the diversity of plant 24 , 25 , 42 – 44 and microbial communities 11 – 16 that have demonstrated the importance of biodiversity for ecosystem functioning, and for maintaining multiple ecosystem functions. Experiments in plant communities have provided empirical evidence for the widely regarded insurance and redundancy hypotheses of biodiversity for sustaining ecosystem functioning, where greater richness can provide a greater guarantee of the maintenance in functioning under various spatial–temporal environmental conditions 22 – 25 . For soil microbial communities confronted with a potentially large functional redundancy 20 , here we similarly show that greater soil microbiome diversity can also ensure the greater performance in multiple ecosystem functions. This was largely due to the effects of greater microbiome richness increasing (1) the redundancy effect of having more taxa present that support the same function and (2) an increase in the presence of microbes that were for the most part associated with different ecosystem functions. These results show the importance of maintaining a greater taxonomic richness because it supports greater functional redundancy and diversity that parallels observations in aboveground plant communities 21 – 25 , 42 – 44 . Importantly, functional redundancy and diversity are both key features of biodiversity that provide support for the ‘insurance’ 21 , 22 and ‘rivit-redundancy’ 21 , 45 hypotheses and the ‘portfolio’ effect 46 as to why greater biodiversity is needed to maintain greater functioning. Our findings further extend these concepts to show that greater microbial richness also provides greater association complexity within microbial communities and resultantly a greater association among taxa that support the multiple functions of interest. Further we found that by combining results from both fungal and bacterial communities on their functional associations, along with results from network analyses, we could achieve some of the strongest relationship with multifunctionality. This supports our hypothesis (3) that a more taxonomically rich soil microbiome can underpin soil multifunctionality because it also ensures greater association complexity among microbes that together are required to support multiple ecosystem functions simultaneously. Perhaps what is most intriguing about our findings is that often the consideration of both fungi and bacteria together improved our ability to predict soil multifunctionality. Until now, most microbiome studies have focused on particular groups (e.g. bacteria or fungi), while still relatively few assessed other key groups of soil organisms, such as protists, Archaea, and nematodes 29 , 47 , 48 in order to obtain a more complete picture of the soil biome. Here we found that considering both fungal and bacterial community characteristics simultaneously was often a better predictor of multifunctionality in nutrient cycling compared to considering these two microbial kingdoms separately. This is in line with earlier observations revealing that there is division of metabolic labor among microbes leading to complementarity among those with unique physiological properties, such as between fungi and bacteria 49 , 50 . For instance, litter decomposition may be performed by distinct groups of soil microbes that inhabit different parts of the soil horizon 51 . Moreover, it has been shown that different plant symbionts (arbuscular mycorrhizal fungi and nitrogen-fixing bacteria) can complement each other by providing different limiting nutrients to plants resulting in higher plant productivity 52 . This points to the importance of microbial interkingdom associations as a driver of ecosystem functioning and parallels recent observations that associations among guilds of microbes promote plant health in the model plant Arabidopsis thaliana 53 . Such unseen synergisms might be much more widespread and ecologically important for the soil microbiome functioning than previously thought. Although here our focus was solely on the soil microbiome encompassing soil fungi and bacteria, our results also allude that including information from other organismal groups beyond fungi and bacteria may further improve our ability to predict soil multifunctionality. Therefore, a next challenge is to link the composition of the soil biome, including multitrophic levels, to ecosystem functioning. It has been hypothesized that vertical diversity (among guilds of organisms) may be just as important, if not more, than the horizontal diversity within a single guild of organisms 54 . Considering this, it is important to note that the filtering approach that we used to manipulate microbial communities also altered the composition of other key groups of soil biota not assessed in our study. These unassessed groups, such as microbial and fungal feeders, may also have contributed to the observed effects directly or indirectly and this deserves further attention in future work (see Table 2 ). Although here we focus on the soil microbiome, there remains numerous facilitative, antagonistic, and multi-trophic interactions among the many individual members of the soil biome that are still poorly understood as to how such vertical diversity affects soil multifunctionality that needs further exploration. In nature, soil ecosystems are highly heterogeneous since soil microbial biodiversity hot spots can form spatial and temporally within soil aggregates 55 – 57 and microbial abundance and diversity declining with greater soil depth 58 , 59 . This spatial heterogeneity likely plays an important role for the interactions among microbes and the mechanisms by which more complex and diverse communities drive various nutrient cycling processes on small spatial scales. For instance, fungal hyphal networks can span air pores within the soil facilitating the movement of bacterial communities to new resource patches 49 . Furthermore, diverse microbial interactions within soil aggregates are likely not only microbial diversity hot spots but are likely also hot spots for key soil processes 57 . Considering these additional spatial complexities of the soil ecosystem in natural environments together with our results could indicate the further importance of various indices of microbial diversity and functional complexity across spatial and temporal scales as has been shown for aboveground plant communities 24 . Moreover, perturbations to the soil ecosystem through compaction and tilling that physically damage larger soil organisms 60 and restrict movement among soil pore space 49 , 61 , will likely not only result in the loss of soil microbial diversity and the structural spatial complexities in natural soil environments, but also its ability to function and cycle nutrients between above and belowground compartments effectively. Considering this, our approach of using a model system with a relatively homogenous soil environment may have underestimate the importance of the complexity of soil microbial diversity and its role in supporting ecosystem function that requires further investigation in situ. Recent studies have shown that microbial network complexity varies between different ecosystem types. For instance, network complexity was much higher in late successional fields compared to early successional fields 29 . Moreover, organically managed agricultural fields harbored much more complex fungal networks with many more keystone taxa, compared to conventional managed fields 31 . Extrapolation of the findings in this study, to these systems, implies that the microbial contribution to ecosystem functioning is higher in systems with higher association complexity, such as in late successional fields and the organically managed fields 29 – 31 . Our study emphasizes that both horizontal functional diversity (i.e. diversity within a guild of organisms such as fungi) and vertical or interkingdom functional diversity (e.g. diversity among functional guilds, such as among fungi and bacteria) are both important for maintaining ecosystem functioning. Our work further demonstrates that the extinction of complex ecological interactions belowground impairs important ecosystem services that soils provide us."
} | 4,902 |
39809803 | PMC11733112 | pmc | 1,988 | {
"abstract": "Lignin, as the abundant carbon polymer, is essential for carbon cycle and biorefinery. Microorganisms interact to form communities for lignin biodegradation, yet it is a challenge to understand such complex interactions. Here, we develop a coastal lignin-degrading bacterial consortium (LD), through “top-down” enrichment. Sequencing and physiological analyses reveal that LD is dominated by the lignin degrader Pluralibacter gergoviae (>98%), with additional rare non-degraders. Interestingly, LD, cultured in lignin-MB medium, significantly enhances cell growth and lignin degradation as compared to P. gergoviae alone, implying a role of additional outliers. Using genome-scale metabolic models, metabolic profiling and culture experiments, modeling of inter-species interactions between P. gergoviae , Vibrio alginolyticus , Aeromonas hydrophila and Shewanella putrefaciens , unravels cross-feeding of amino acids, organic acids and alcohols between the degrader and non-degraders. Furthermore, the sub-population ratio is essential to enforce the synergy. Our study highlights the unrecognized role of outliers in lignin degradation.",
"introduction": "Introduction Lignin is the most abundant aromatic carbon polymer on earth and possesses a recalcitrant and heterogenous nature 1 . Microbes have evolved various enzymes and pathways to decompose lignin and drive carbon and energy cycles. Moreover, in nature, various microorganisms co-exist to form consortia that display improved lignin conversion and resist environmental perturbations 2 – 4 , indicating that consortia are superior to single species in lignin conversion. Thus, investigation of microbial ecological roles and interaction modes in communities is essential to understand and optimize lignin bioconversion for bioenergy production and climate change alleviation. Recently advanced high-throughput sequencing technologies have revealed the diversities of in situ and cultivated lignin-degrading microbiomes from various ecosystems. For instance, North American forest soil harbored a lignin-degrading consortium composed of Sphingomonadaceae , Comamonadaceae and Caulobacteraceae families 5 . Another lignin-degrading consortium (LigMet) was enriched from sugarcane plantation soil and revealed to be composed of Alcaligenaceae , Micrococcaceae , Phyllobacteriaceae and Paenibacillaceaea families 6 . Recently, coastal wetlands were also reported to be enriched in lignocellulolytic degrading consortia 7 , 8 . Various coastal lignin-/lignocellulose degrading consortia were obtained 9 . Among them, the prevalence of Vibrio , Aeromonas , Desulfovibrio and Shewanella genera was observed 8 . Metagenome sequencing further revealed the multiple lignin oxidative enzymes (e.g., dye-decolorizing peroxidases (DyP) and laccases) and aerobic-/anaerobic guaiacyl (G), syringyl (S), and p -hydroxyphenyl (H) type lignin unit degradation pathways in these coastal communities 8 . Such studies described taxonomic and functional profiling for various lignin-degrading consortia. However, a comprehensive and mechanistic understanding of these biological systems remains elusive. Notwithstanding the performance of a consortium (e.g., lignin degradation and growth) would be largely attributed to the inter-species interactions within a community. There are six well-known classical bidirectional interaction modes in microbial communities, e.g., commensalism, mutualism, competition, neutralism, amensalism and predation 10 . These interactions are driven by the sharing, exchange and competition of metabolites. These metabolites, designated public goods, are commonly produced at high cost to individual members for the benefit of the greater community. Labor division and/or aggregation patterns are mostly used to explain the synergy between lignocellulose degraders 11 , 12 . It is worth noting that communities include not only degraders but also non-degraders that are seen as outliers 13 – 15 . What are the exact roles of the non-degraders? Do they contribute to lignin degradation, or are they just ‘cheaters’ within the community? Experimental verification of the microbial interactions is employed as the direct method 16 , 17 . It requires tedious and trial-and-error exploration and hence underestimates the interactions. Progress in sequencing technologies promotes the development of indirect methods. Molecular ecological networks are constructed based on the statistical correlations of taxa among numerous samples 18 , 19 . The co-occurrence networks are widely used to rapidly elucidate possible interactions in microbial communities, especially for uncultivated microbiomes 15 , 20 . However, correlation-based inference not only commonly overestimates the interaction but also hardly provides a mechanistic understanding of correlations. Alternatively, meta-genomic/transcriptomic and metabolomic analyses reveal the metabolic pathways of the community, demonstrating the metabolic potential of the community 8 , 21 , 22 . It is a challenge to definitely assign biochemical reactions to those different members in the community, restricting to uncover inter-species interactions 17 . Parallel computational efforts have enabled in silico simulation of microbial interactions through genome-scale metabolic models (GSMM) and constraint-based reconstruction and analysis (COBRA) models 23 – 26 . The biologically realistic metabolic models, coupled with the experimental verification by synthetic communities (the complementary reductionist communities), provide a promising approach to explicitly describe the microbial interactions at the genome scale. In this study, we “top-down” enriched the lignin-degrading community from the coastal intertidal wetlands of the East China Sea. Through high-throughput sequencing and physiological analyses of the community and isolates, we not only characterized the taxonomic and functional gene profiles of the community but also observed that the low abundant non-degraders supported the community’s growth and lignin degradation. To understand the observed advantage of community, an integrative system biology strategy was employed, which combined multiple biological information from metagenomics, GSMM, metabolic profiling, and mono-/co-culture experiments. Our results elucidate cross-feeding metabolites (e.g., alcohols, amino acids and organic acids) among the degrader and non-degraders, revealing the underestimated role of non-degraders in the complex polymer degradation. Furthermore, it provides a clue for a “bottom-up” design synthetic community based on defined inter-species interactions for lignin bioconversion.",
"discussion": "Discussion Low abundant non-degraders boost the performance of LD consortium In this study, we developed the coastal lignin-degrading LD consortium via “top-down” enrichment. The LD consortium was dominated by P. gergoviae , encoding 96 gene families involved in lignin degradation. In contrast, the additional ASVs not only just occupied ~ 2% (Fig. 1D ) but also did not harbor the gene families, participating in the complete lignin-degrading pathways (Fig. S3B ). Mono-culture investigation of P. gergoviae further confirmed that it can grow in the MB mineral medium with kraft lignin as the sole carbon source (Fig. 2A ). Consequently, P. gergoviae is the only identified species in LD community to perform the task of lignin degradation. P. gergoviae , a Gram-negative, facultatively anaerobic bacterium, is widely distributed in various environments (e.g., soil, plant leaf, aquatic sources and coastal wetlands). It can grow on chlorphenesin and ethylhexylglycerin, and resist parabens and other preservatives 39 . However, its capacity for lignin degradation had been previously unrecognized. Our study not only reveals that it can degrade kraft lignin but also develops the P. gergoviae -dominated lignin-degrading consortium, LD. Interestingly, the growth of LD consortium was superior to P. gergoviae alone. Lignin degradation of LD consortium was also higher, generating more abundant DOC than P. gergoviae mono-culture (Fig. S2C, D ). Furthermore, we noticed that some of the rare ASVs in LD consortium, e.g., Vibrio ASV5, Aeromonas ASV18 and Shewanella ASV20, also exist in the other coastal lignin/lignocellulose degrading consortia 8 , 9 , implying that they might be commensal with P. gergovia in LD consortium, instead of random partnerships (Fig. S3A ). Overall, the low abundant non-degraders in LD consortium, overshadowed by P. gergoviae , play unrecognized roles in lignin conversion. Cross-feeding of metabolites between degraders and non-degraders promote community growth and lignin degradation Microbial interactions in communities are key to perform their ecological function (e.g., polymer degradation). Numerous studies explored aggregation and division of labor, based on their distinct physical or metabolic features 40 , 41 . Taking the lignocellulose degradation community as an example, both Citrobacter freundii and Sphingobacterium multivorum are lignocellulosic degraders. Co-culturing of these organisms greatly enhances cellulose degradation, compared to the individual mono-cultures 42 . Alternatively, Sphingomonas sp. in a bisphenol A (BPA)-degrading microbial community is responsible for the degradation of BPA, whereas the additional members, e.g., Pseudomonas sp., degrades the downstream intermediates of 4-DM (4,4′-dihydroxyl-α-methylstilbene) to drive the complete mineralization of BPA 14 . In contrast, most species in a community, especially those with low abundance, are harder to disentangle, as they have not definitely roles in the degradation of complex polymers. Here, the lignin degrader P. gergoviae was identified in LD consortium, growth experiments supported its metabolic capabilities for lignin and its derivates (Fig. S4 ). In contrast, the other members, e.g., V . alginolyticus , A . hydrophila and S . putrefaciens , do not contain any of the lignin-degrading gene families and, thus, could not individually grow with these substrates (Figs. S3B , S4 ). These species are easy to be classified as “cheaters” and, thus, occupy a tiny niche in LD community. In fact, the low abundant non-degraders boost the performances of LD consortium and four-species combination (Fig. 4B ). An uncharacterized boosting strategy should be employed in the LD community. Mathematical models (e.g., GSMM and COBRA), hence, were employed to pinpoint the potential inter-species metabolic interactions, providing mechanistic interpretation. Our model, coupled with an array of metabolic analysis and mono-/co-culture experiments, suggested that the lignin degrader ( P. gergoviae ) degraded lignin and provided succinate and malate to non-degraders as labile carbon sources. The enhanced growth of non-degraders, in return, supplied glycerol, alanine and aspartate as nutrients to foster P. gergoviae growth and further stimulate lignin degradation. In addition, the cross-feedings also might have an influence on the whole metabolome, generating the various unique extracellular compounds in the four-species combination. Each species, by releasing and sensing these compounds, would harness interactions to stabilize cooperative behaviors 43 , 44 . Eventually, degraders and non-degraders, at the sub-population ratio (100:1), generated a positive feedback loop via mutualistic cross-feeding interactions. Interestingly, when the populations of non-degraders were artificially increased to levels similar to the degrader, the reconstructed communities immediately collapsed (Fig. S7 ). This could be interpreted with the “snow-drift game” 45 , in which degraders and non-degraders can co-exist in a system where only degraders can acquire enough net interest. Degraders commonly have a lower growth rate, as they produce public goods at high cost. When non-degraders invade and dominate the community, public goods cannot be produced and shared to the required degree, resulting in community collapse 46 . This study not only uncovers the positive roles of non-degraders in lignin degradation but also demonstrates that cross-feeding amino acids, organic acids and alcohols, rather than lignin metabolites, contribute to lignin degradation. Synthetic communities pave the road for more accurate modeling of natural communities With the rapid development of high-throughput sequencing technologies, numerous in situ microbiomes or “top-down” enriched communities have been well-studied, including their compositions, ecological functions and succession properties 8 , 9 , 47 , 48 . In contrast, far fewer “bottom-up” synthetic communities have been constructed, in view of defined microbial interactions 49 , 50 . Moreover, successfully constructed synthetic communities are mostly of bi-culture design, based on either aggregation or division of labor 49 , 51 . This is because each member in these strategies has desirable biological function that promotes the performances of the community. Co-culturing of cellulolytic Ruminiclostridium cellulolyticum and sulfate-reducing Desulfovibrio vulgaris was observed to contribute to cellulose degradation 49 . A bi-culture of Acinetobacter sp. AG3 and Bacillus sp. R45, both of which can degrade bromoxynil octanoate (BO), was shown to accelerate BO degradation 52 . Furthermore, R45 also can degrade bromoxynil, the downstream metabolite of BO. However, these observed interactions are pairwise between two organisms. Much more complex interactions, tertiary interactions, should occur in natural communities. Here, the model-based analysis guided us in exploring interactions of the four-species synthetic community (Fig. 3A ). A large number of variables (1421–1886 reactions and 1418–1771 metabolites) in each strain were first considered. Then, mathematical algorithms were constructed to explore all the possible strategies for metabolic exchanges in the four-species synthetic community (Fig. 3A, B ). Therefore, syntrophic exchanges of small metabolites among the four bacteria were clearly predicted and targeted (Fig. 3 ). Finally, the quad-culture not only showed a more than 2-fold higher biomass than either mono-culture of P. gergoviae or bi-culture of P. gergoviae with any one of the other 3 species but also exhibited higher lignin degradation (Fig. 4B ), laying foundations for the application of synthetic communities in lignin valorization. In summary, this study explores the outliers in LD consortium. Our work demonstrates that cross-feeding of metabolites between degraders and non-degraders boost community performance, highlighting the key roles of non-degraders in lignin degradation. Moreover, we provide new insights into the “bottom-up” construction of a synthetic community for lignin utilization, which holds promise to efficiently manipulate microbiomes. Limitations of the synthetic community, due to the current state of technology (e.g., microbial isolation and cultivation, and scale of metabolic models), should also be acknowledged 13 , 53 . The performance (e.g., growth and lignin degradation) of the enriched LD community still surpasses the synthetic community. It not only indicates that the additional ASVs in the LD consortium should also provide positive synergies but also implies that much complex cross-feedings, e.g., quorum sensing signals, exoenzymes, and nanowires, might support the cooperation 38 . In the near future, with the advancement of in silico mathematical models, the facilitation, reliability, and throughput of in silico simulations will increase, and provide more accurate and comprehensive predictions about the cooperative and competitive behaviors in natural communities."
} | 3,941 |
38900914 | PMC11304949 | pmc | 1,991 | {
"abstract": "Abstract Stony coral tissue loss disease (SCTLD) has devastated coral reefs off the coast of Florida and continues to spread throughout the Caribbean. Although a number of bacterial taxa have consistently been associated with SCTLD, no pathogen has been definitively implicated in the etiology of SCTLD. Previous studies have predominantly focused on the prokaryotic community through 16S rRNA sequencing of healthy and affected tissues. Here, we provide a different analytical approach by applying a bioinformatics pipeline to publicly available metagenomic sequencing samples of SCTLD lesions and healthy tissues from 4 stony coral species. To compensate for the lack of coral reference genomes, we used data from apparently healthy coral samples to approximate a host genome and healthy microbiome reference. These reads were then used as a reference to which we matched and removed reads from diseased lesion tissue samples, and the remaining reads associated only with disease lesions were taxonomically classified at the DNA and protein levels. For DNA classifications, we used a pathogen identification protocol originally designed to identify pathogens in human tissue samples, and for protein classifications, we used a fast protein sequence aligner. To assess the utility of our pipeline, a species-level analysis of a candidate genus, Vibrio , was used to demonstrate the pipeline's effectiveness. Our approach revealed both complementary and unique coral microbiome members compared with a prior metagenome analysis of the same dataset.",
"conclusion": "Conclusions In all, we provide a novel pipeline to understand coral disease, and further investigate the role of bacteria and DNA viruses in SCTLD. Our new pipeline found both incongruities and parallels with the original analysis of these data. For instance, both studies revealed a prevalence of Rhodobacteraceae and Flavobacteriaceae in lesion samples. One of the major differences between the studies was the dominance of Synechococcus and Vibrio in this study, which was likely driven by the use of k -mers to quantify taxonomic abundance (as the protein analysis with read counts aligned more with previous results). In addition to providing novel insights, our pipeline also reduced the computational load for downstream analysis, making large metagenomic analysis more accessible to those with less computational resources. This method is useful not only for coral scientists, but also for fields that study nonmodel organisms and lack comprehensive genomic resources. In corals, the lack of such resources can impede the progress of metagenomic analysis and the identification of potential pathogens. Specifically, host-associated metagenomes without a reference genome can dominate and confound metagenomic results ( Rosales et al. 2022 ). As genome assemblies are unlikely to be available anytime soon for the over 30 species of coral affected by SCTLD, this pipeline offers a useful tool for the simultaneous examination of viruses, bacteria, and eukaryotic species living on or within the host tissue that may be associated with infection for any coral species or nonmodel organism.",
"introduction": "Introduction Stony coral tissue loss disease (SCTLD) was discovered off the coast of Miami, FL, in 2014 and has since had negative consequences on the function of coral reefs across Florida and the Caribbean ( Walton et al. 2018 ; Alvarez-Filip et al. 2022 ). To date, despite many efforts, no pathogen has been definitively identified as the causative agent of SCTLD. The stony coral (order Scleractinia) microbiome is a complex system of interactions between the host, bacteria, viruses, fungi, archaea, and algal symbionts ( Bourne et al. 2009 ); thus, a disturbance in any number of these symbiotic relationships could be involved in SCTLD progression. Multiple studies have explored viruses that may infect stony coral symbionts, notably Symbiodiniaceae , but no causative relationships have been detected ( Work et al. 2021 ; Veglia et al. 2022 ; Beavers et al. 2023 ; Howe-Kerr et al. 2023 ). Bacterial species are particularly under scrutiny for their potential involvement in SCTLD, due to the effectiveness of antibiotics in halting lesion progression in multiple affected coral species ( Aeby et al . 2019 ; Neely et al. 2020 ; Shilling et al. 2021 ; Studivan et al. 2023 ). Consequently, SCTLD studies have predominantly focused on understanding changes in the bacterial community between apparently healthy and SCTLD-affected corals. Studies to identify bacterial pathogens have relied primarily on small subunit 16S ribosomal RNA (rRNA) sequencing, followed by computational analysis ( Callahan et al. 2016 ) typically using the Silva database ( Quast et al. 2013 ) to assign and classify amplicon sequence variants (ASVs) into taxa. ASVs found in diseased lesion samples are then compared with samples from apparently healthy colonies to determine which ASVs are associated with the tissue loss lesions ( Meyer et al. 2019 ; Rosales et al. 2020 ; Clark et al. 2021 ). These methods have characterized many notable shifts in coral bacterial communities due to SCTLD and several bacterial taxa associated with SCTLD lesions, including Rhizobiales , Clostridiales , Peptostreptococcales-Tissierellales , Rhodobacteraceae , Flavobacteriaceae , and Vibrionaceae ( Rosales et al. 2023 ). However, because of the difficulty in determining whether an associated bacterial taxon is a harmless commensal, an opportunistic secondary infection, or the primary pathogen, none of the bacterial taxa associated with SCTLD have been identified as the causative agent. An alternative approach to understanding disease dynamics is the use of metagenomic sequencing, in which all of the DNA from a source is sequenced, including not only the host, but also viruses, bacteria, and eukaryotic species living on or within the host tissue. For example, by analyzing metagenomic sequencing data of human tissue samples taken from the site of infection, researchers have identified pathogenic agents in brain infections ( Salzberg et al. 2016 ; Wilson et al. 2019 ), corneal infections (Li et al. 2018) , and other diseases ( Kostic et al. 2012 ). The sensitivity of this approach relies on first, sequencing the source DNA deeply enough to capture the pathogen of interest, and second, the existence of genome assemblies closely related to the pathogen in public databases. While the number of complete genomes has grown enormously over the past 2 decades, databases still contain few or no genomes for nonmodel organisms, including scleractinian corals. Currently, data from only 1 SCTLD metagenome study are publicly available. While the authors of that study ( Rosales et al. 2022 ) were able to assemble and annotate genomes for SCTLD-associated bacterial taxa such as Rhodobacterales , Rhizobiales , and Flavobacteriales , the results were focused on only 5 of the 20 diseased lesion tissue samples, all from the same coral species ( Stephanocoenia intersepta ), because the majority of samples were dominated by host sequences. In metagenomic studies, host sequences can confound results, so they are typically removed by aligning all reads to a host reference genome ( Lu et al. 2022 ; Gihawi et al. 2023 ). Currently, the GenBank database has 53 genome assemblies from scleractinian corals, of which only 7 are at the chromosome level ( NCBI 2023 ). Of these 53 genomes, none are from the species of corals previously investigated for SCTLD ( Rosales et al. 2022 ), emphasizing the additional challenges associated with using metagenomics in nonmodel organisms. Additionally, given the complex symbiotic microbiome (i.e. algal symbiont, viruses, and prokaryotic community) of stony corals ( Bourne et al. 2009 ), the host DNA is only one of the hurdles. In this study, we applied new classification methods to understand this devastating coral disease. We used a method to filter host reads from metagenome data by using data collected from apparently healthy corals of the same species to approximate a species-specific healthy host coral genome and microbiome. We then applied the Kraken software suite for pathogen identification ( Lu et al. 2022 ) using KrakenUniq ( Breitwieser et al. 2018 ) to identify putative pathogens present in diseased samples and not present in healthy ones. Using these methods, we identified a number of taxa that have previously been associated with SCTLD, providing further support for their involvement in SCTLD pathogenesis. Finally, from the pool of bacterial taxa we identified as associated with SCTLD lesions, we selected a candidate genus, Vibrio , with which to explore the utility of our pipeline at a finer taxonomic level.",
"discussion": "Discussion In this study, we used previously published sequencing data from coral affected by SCTLD and developed a novel metagenomic analysis pipeline to explore the microbial communities present in those data. The data consisted of samples from 4 coral species collected from Florida's coral reefs during an SCTLD outbreak. To investigate the microbial taxonomy of these samples, the previous study used small subunit rRNA gene assemblies and metagenome-assembled genomes (MAGs). Our investigation differed by using the Kraken software suite and focusing on unique k -mer count data to understand abundances. By predominantly working with k -mers instead of MAGs, we maximized the utility of the read data, as our approach allowed us to capture reads that may have been discarded during MAG assembly and binning. Samples with high intrapopulation diversity can provide challenges in MAG assembly and binning ( Ramos-Barbero et al . 2019 ; Meziti et al . 2021 ), which may have hindered the generation of a MAG bin in the previous analysis. The previous work also did not filter out host sequences because reference genomes do not exist for the sampled coral species. However, here we applied a novel technique to filter these reads by using data derived from apparently healthy (AH) samples as a surrogate for reference genomes. This allowed us to examine unique sequences from DL samples by approximating a species-specific host coral genome and reduce computational load. In addition, in this study, we investigated the SCTLD DNA virome, which has not been previously reported. The families Rhodobacteraceae and Flavobacteriaceae were found to be associated with SCTLD in our protein analysis, consistent with other SCTLD studies (reviewed in Papke et al. 2024 ), including previous examinations of the same sample set ( Rosales et al. 2020 , 2022 ). Rhodobacteraceae is one of the most common bacterial families associated with coral diseases ( Gignoux-Wolfsohn et al. 2017 ) in diverse geographic locations, but no member has been identified as a causative coral disease agent ( Mouchka et al. 2010 ), and members of this group are broadly found across ocean habitats and fulfill diverse ecological functions ( Brinkhoff et al. 2008 ). Flavobacteriaceae has been enriched in White Band Disease in the Scleractinia staghorn coral Acropora cervicornis ( Gignoux-Wolfsohn and Vollmer 2015 ), but has never been identified as a causative agent in coral tissue loss. In SCTLD, Flavobacteriaceae has previously been found enriched in unaffected (nonlesion) tissue from diseased colonies, potentially indicating colony stress and initial dysbiosis due to disease ( Rosales et al. 2023 ). In contrast to the protein analysis, which used read count data, our DNA-level analysis using k -mers did not detect a singular genus belonging to Rhodobacteraceae or Flavobacteriaceae in high proportions among the unique diseased reads, providing support for the idea that members of these 2 families are likely a diverse set of bacteria species that associate with SCTLD opportunistically. In our DNA-level analysis, Vibrionaceae and Synechococcaceae were among the most abundant families within the unique DL reads and these high abundances showcase how the k -mer-based abundance approach can provide a different perspective of highly abundant taxa in particular samples. These families were not observed as highly abundant in the MMSeqs2 protein analysis; however, we were primarily interested in whether new candidates emerged from the MMSeqs2 analysis, not the relative proportions of the candidates. Interestingly, the genera Synechococcus and Vibrio were not detected in the previous analysis of these data ( Rosales et al. 2022 ). In the 16S rRNA metabarcoding study of these same samples, Synechococcus and Vibrio were differentially abundant in M. meandrites samples ( Rosales et al. 2020 ). These results did not show a clear trend as Synechococcus was present across multiple AH samples and Vibrio was not prevalent across DL coral samples. In this study, Synechococcus was the most or second most proportionally abundant genus across all 4 coral species. Synechococcus belong to phylum Cyanobacteria , which are photosynthetic picoplankton ( Kim et al. 2018 ), and typically not involved in pathogenesis. Even so, Synechococcus have been enriched in other SCTLD studies that compared healthy colonies and healthy tissue on diseased colonies, and it was hypothesized that their increase in abundance is a response to disease stress ( Rosales et al. 2023 ). The high proportion of Synechococcus in this study supports the suggestion that Synechococcus may have some role in microbial community interactions during SCTLD. \n Vibrio also have been associated with SCTLD (reviewed in Papke et al. 2024 ), including in these samples ( Rosales et al. 2020 ), but in contrast to Synechococcus , Vibrio have been associated with other coral tissue loss diseases, such as white pox disease ( Kemp et al. 2018 ) and yellow band disease ( Cervino et al. 2008 ), and diseases in other marine organisms, such as sponges ( Dinçtürk et al . 2023 ) and crustaceans ( de Souza Valente and Wan 2021 ). In 3 coral species analyzed here, Vibrio was one of the 2 most abundant genera, suggesting it may be associated with SCTLD tissue loss. In the fourth species, M. meandrites , Vibrio was only the sixth most abundant genus. However, M. meandrites had the fewest AH reads to create the database used for filtering the DL reads ( Table 1 ), which may reduce the effectiveness of our novel pipeline that relies on reads from control groups. This was also a surprising result, since as mentioned, Vibrio were abundant in the 16S rRNA M. meandrites analysis of these samples ( Rosales et al. 2020 ). When using our pipeline to explore a species-level analysis within the candidate genus Vibrio , we found read matches to V. mediterranei/shilonii ( Tarazona et al . 2014 ), V. coralliilyticus , V. harveyi , and V. owensii —all known coral pathogens associated with bleaching and tissue loss ( Kushmaro et al. 1997 ; Ben-Haim et al. 2003 ; Luna et al. 2007 ; Ushijima et al. 2012 ; Munn 2015 ). Vibrio mediterranei/shilonii , which comprised a majority of the classified Vibrio species in D. labyrinthiformis and D. stokesii , was previously found to be responsible for the annual bleaching of the scleractinian coral Oculina patagonica off the Israeli coast from 1993 ( Kushmaro et al. 1996 , 1997 ) to 2003 ( Reshef et al. 2006 ). Additionally, when V. mediterranei / shilonii is experimentally introduced together with V. coralliilyticus to healthy corals, they appear to have synergistic pathogenic effects ( Rubio-Portillo et al. 2014 ). Given these associations, V. mediterranei / shilonii may be of particular interest in future SCTLD studies as potentially increasing the virulence of SCTLD, as does V. coralliilyticus ( Ushijima et al. 2020 ), which also appeared in our samples. However, the very low similarity between the cultured V. coralliilyticus sequences in Ushijima et al. (2020) and our Vibrio sequences leads us to believe that multiple Vibrio species may be involved in SCTLD lesion development. It is important to note that our picture of the SCTLD microbiome is restricted by the genomes in the databases used. The Vibrio genus has been found to have a large degree of genetic plasticity between species ( Gu et al . 2009 ), so while the classifications to different Vibrio species may truly represent the presence of an array of Vibrio species, it may instead be the result of various reads from a novel Vibrio species matching different Vibrio species based on closest genomic similarity. Therefore, while matches to different Vibrio species and their potential role in SCTLD may offer some insights, a more robust interpretation is to consider the implications of disease association by Vibrio at the genus level. The etiology of SCTLD has been hypothesized to be viral ( Work et al. 2021 ), and gene expression data show that there is an increase in coral viral immune response in corals with SCTLD ( Beavers et al. 2023 ). Researchers have explored the potential role of RNA viruses in SCTLD, but no RNA viruses have been found exclusively in corals with SCTLD ( Veglia et al. 2022 ) and these viruses are likely ubiquitous in corals without any potential relationship to SCTLD ( Howe-Kerr et al. 2023 ). However, the involvement of DNA viruses has not previously been explored. Our data show the majority of DNA viruses in diseased samples represent phages. Not surprisingly, phage sequences correspond with some of the most abundant bacteria identified in this study, such as Rhodobacteraceae , Vibrionaceae , and Synechococcaceae . The Paracoccus phage, which infects Rhodobacteraceae , and the Vibrio phage would be interesting to further explore as potential avenues for disease mitigation. In addition to phages, sequences were found with similarities to the C. ericina virus. However, with only 2 contigs and little coverage of its genome, we do not believe this virus plays a role in SCTLD. Thus, we did not find any DNA viruses with a definitive association with SCTLD. Future studies may consider viral enrichment protocols prior to sequencing to help better characterize the SCTLD DNA virome. In addition to differences in the bacterial and viral communities, members of Symbiodiniaceae were found to be abundant in DL samples, particularly in M. meandrites . SCTLD disrupts the relationship between the host coral and its Symbiodiniaceae through symbiont necrosis and peripheral nuclear chromatin condensation, among other physiological changes ( Landsberg et al. 2020 ). This may result from an increase in rab7 signaling among the Symbiodiniaceae to degrade dead and dysfunctional cells through endocytic phagosomes ( Beavers et al. 2023 ). The Symbiodiniaceae DNA identified in diseased samples in this study may be a byproduct of this necrosis and degradation of the symbiont. This was especially notable in M. meandrites , which was the coral in this study most susceptible to acute tissue loss and mortality from SCTLD ( Precht et al. 2016 ); this accelerated tissue loss may lead to higher levels of dead and dysfunctional symbionts being produced during visual lesion progression in M. meandrites than in other coral species. Although, this pipeline was initially developed under the one-pathogen-one-disease assumption for humans, our results show that we can identify a consortium of putative pathogens with this method (i.e. Rhodobacteraceae , Flavobacteriaceae , and Vibrio ). However, this pipeline is limited by the amount of prior information about pathogens in the species investigated; for example, human pathogens are well represented in metagenomic classification databases, whereas nonmodel organism pathogens may not be, which can lead to a higher rate of unclassified or misclassified reads. In addition, disease states are more clearly defined in humans than in nonmodel organisms, making the assumptions about sample health more reliable in human analyses. Conclusions In all, we provide a novel pipeline to understand coral disease, and further investigate the role of bacteria and DNA viruses in SCTLD. Our new pipeline found both incongruities and parallels with the original analysis of these data. For instance, both studies revealed a prevalence of Rhodobacteraceae and Flavobacteriaceae in lesion samples. One of the major differences between the studies was the dominance of Synechococcus and Vibrio in this study, which was likely driven by the use of k -mers to quantify taxonomic abundance (as the protein analysis with read counts aligned more with previous results). In addition to providing novel insights, our pipeline also reduced the computational load for downstream analysis, making large metagenomic analysis more accessible to those with less computational resources. This method is useful not only for coral scientists, but also for fields that study nonmodel organisms and lack comprehensive genomic resources. In corals, the lack of such resources can impede the progress of metagenomic analysis and the identification of potential pathogens. Specifically, host-associated metagenomes without a reference genome can dominate and confound metagenomic results ( Rosales et al. 2022 ). As genome assemblies are unlikely to be available anytime soon for the over 30 species of coral affected by SCTLD, this pipeline offers a useful tool for the simultaneous examination of viruses, bacteria, and eukaryotic species living on or within the host tissue that may be associated with infection for any coral species or nonmodel organism."
} | 5,427 |
26981620 | PMC4794232 | pmc | 1,992 | {
"abstract": "Many living organisms transform inorganic atoms into highly ordered crystalline materials. An elegant example of such biomineralization processes is the production of nano-scale magnetic crystals in magnetotactic bacteria. Previous studies implicated the involvement of two putative serine proteases, MamE and MamO, during the early stages of magnetite formation in Magnetospirillum magneticum AMB-1. Here, using genetic analysis and X-ray crystallography, we show that MamO has a degenerate active site, rendering it incapable of protease activity. Instead, MamO promotes magnetosome formation through two genetically distinct, noncatalytic activities: activation of MamE-dependent proteolysis of biomineralization factors and direct binding to transition metal ions. By solving the structure of the protease domain bound to a metal ion, we identify a surface-exposed di-histidine motif in MamO that contributes to metal binding and show that it is required to initiate biomineralization in vivo. Finally, we find that pseudoproteases are widespread in magnetotactic bacteria and that they have evolved independently in three separate taxa. Our results highlight the versatility of protein scaffolds in accommodating new biochemical activities and provide unprecedented insight into the earliest stages of biomineralization.",
"introduction": "Introduction Biomineralization is the widespread phenomenon by which living organisms transform inorganic atoms into highly ordered, crystalline structures. Controlling the size and shape of such materials requires specialized protein machinery that can define the nano-scale trajectory of crystal growth [ 1 ]. Incorporating biochemical principles uncovered from studying biomineralization has the potential to revolutionize the design and synthesis of nanomaterials in vitro [ 2 ]. In addition to the well-known examples of tooth, bone, and shell production by multicellular eukaryotes, a number of bacteria have the ability to biomineralize small magnetic crystals within subcellular compartments called magnetosomes [ 3 , 4 ]. These particles allow the cells to passively align in the earth’s magnetic field, facilitating the search for their preferred oxygen environments [ 5 ]. Although these magnetotactic bacteria have drawn longstanding interest due to their ability to manipulate transition metals, the biochemical details of how they transform iron into magnetite (Fe 3 O 4 ) remain poorly understood. Magnetotactic organisms are phylogentically diverse. Nearly all isolates come from the α-, δ-, or γ- classes of Proteobacteria , but representatives from the Nitrospirae and Omnitrophica phyla have recently been identified [ 6 ]. The genes responsible for making magnetosomes are often contained in a genomic region called the magnetosome island (MAI) [ 7 – 11 ]. Comparative genomic and phylogenetic studies have identified a set of core genes that appears to have been assembled a single time and inherited vertically, indicating that magnetosome formation likely predates the divergence of the Proteobacteria [ 12 , 13 ]. The MAI seems to have formed by incorporating elements from other cellular processes, as the majority of the core factors have homology to ancient and widespread protein domains [ 14 , 15 ]. Uncovering the biochemical functions encoded in the MAI in relation to its evolutionary history provides a unique opportunity to understand how new cellular processes evolve. Due to the availability of genetic systems, α -Proteobacteria such as Magnetospirillum magneticum AMB-1 are used as models for studying the molecular biology of magnetosome formation [ 16 ]. AMB-1 contains 15–20 magnetite crystals, each formed within a cytoplasmic membrane invagination and organized in a chain spanning the length of the cell [ 17 , 18 ]. By making deletions within the MAI and characterizing the ultrastructure of the mutant cells, specific genes have been assigned roles in various stages of magnetosome formation [ 10 , 11 , 19 , 20 ]. Genes whose deletions produced empty magnetosome compartments or compartments with abnormally small magnetite crystals were termed biomineralization factors. It is the proteins encoded by these genes that hold the secrets of how magnetotactic organisms interface with solid magnetite. Two genes in the MAI, mamE and mamO , are homologous to the HtrA proteases, a ubiquitous family of trypsin-like enzymes that functions using His-Asp-Ser catalytic triads [ 21 ]. An additional pair of genes, called limE and limO , with homology to the protease domains of mamE and mamO , exists in a secondary genomic region termed R9 [ 10 ]. Disrupting mamE or mamO causes cells to produce empty magnetosome membranes, but removing R9 has no effect, showing that mamE and mamO are required for biomineralization and that limE and limO are not ( Fig 1 ) [ 10 , 22 ]. Adding variants of mamE or mamO with all three predicted active site residues mutated to alanine could not restore normal biomineralization in the Δ mamO Δ R9 or Δ mamE Δ R9 strains but complemented single Δ mamO or Δ mamE deletions [ 23 ]. These genetic analyses show that mamE and mamO are required for the initiation of magnetite biomineralization. Furthermore, limE and limO are partially redundant in that they can cross-complement the active site-dependent crystal maturation defects of their respective orthologs. 10.1371/journal.pbio.1002402.g001 Fig 1 MamO promotes the nucleation of magnetosome crystals. ( A ) TEM micrograph of a wild-type AMB-1 cell. The electron-dense particles make up a magnetosome chain. ( B ) Cellular organization of the magnetosome compartment. MamO promotes the nucleation of magnetite within inner membrane invaginations. ( C ) Domain structure of the three biomineralization factors discussed in the text. “ c ” represents a CXXCH c -type cytochrome motif. Here, we use a combination of in vivo and in vitro approaches to reveal an unexpected dual role for MamO. It promotes MamE-dependent proteolysis of three biomineralization factors through the use of its C-terminal transporter domain. Separately, the protease domain has lost the ability to carry out catalysis and has instead been repurposed to bind transition metal ions. Two surface-exposed histidine residues that contribute to this metal-binding function are required for initiating magnetite biosynthesis in vivo. Bioinformatic analysis shows that similar pseudoproteases evolved independently in the three major taxa of magnetotactic organisms, highlighting a unique evolutionary mechanism behind microbial nanoparticle synthesis.",
"discussion": "Discussion Magnetotactic bacteria control the growth of their associated magnetite crystals with a level of precision that cannot be replicated in vitro. The molecular details of how they perform this task can reveal novel bioinorganic interfaces and be exploited for improved synthesis of nanomaterials. Genetic analysis has shown that magnetite biomineralization is surprisingly complex. It requires over 15 factors in AMB-1, nearly all of which are predicted integral membrane proteins [ 2 , 19 , 34 ]. A subset of these is required for the initial crystallization of iron within the magnetosome compartment [ 10 ]. While the key players for this step are known, their biochemical functions have only been inferred from sequence homology, leaving the mechanism of how magnetite biosynthesis begins a mystery. Here, we examined two magnetite nucleation factors: the putative HtrA proteases MamE and MamO. We find that the presence of both MamE and MamO is required for the proteolysis of three biomineralization factors, MamE, MamO, and MamP. These events depend on an intact catalytic triad from MamE but not MamO, indicating that MamO activates MamE in a noncatalytic manner. Thus far, we have not been able to detect a physical interaction between MamE and MamO, but we have found that the C-terminal TauE-like transporter domain of MamO is required for activation. The putative ion transport activity of this domain could be the feature that promotes proteolysis. This model is attractive because it does not require a direct interaction between the two proteins. Indirect evidence suggests that TauE family proteins transport sulfite or sulfur containing organic ions, leading us to speculate that the concentrations of specific solutes in the magnetosome might control MamE’s activity [ 35 , 36 ]. Separately, we discovered that a metal-binding function in the protease domain of MamO is required for the initiation of magnetite biomineralization. In our structure, H148 and H263 directly coordinate a single metal ion. Disrupting these residues alters the binding behavior in a modified tmFRET assay, but the effect is unusual in that it lowers the FRET efficiency while slightly increasing the overall affinity for metal. Using the Förster equation and reported radius for Ni 2+ , we calculated that the fluorophore to metal distance changes from 12.8 Å in wild-type to 15.2 Å in the H148A/H263A mutant, suggesting that the binding geometry changes in a way that allows metal binding in the same vicinity [ 37 ]. We favor the explanation that the dihistidine motif identified here is part of a more complex metal coordinating network. Our soaking strategy may have missed additional sites that are inaccessible in our crystal form, and separate attempts at characterizing a fully metal-bound state using co-crystallization have been unsuccessful. Nevertheless, we identified dihistidine motif using the structure that contributes to an unexpected transition metal-binding activity in MamO. Despite the presence of other binding features, metal binding through H148 and H263 is absolutely required in vivo as disruption of either residue completely abolishes magnetite formation. Our structural analysis shows that MamO has lost its ability to perform proteolysis altogether, supporting the idea that metal binding is now the central function of the protease domain. Consistent with this, T225, the predicted catalytic nucleophile, is completely dispensable for biomineralization. Though disrupting H116 and D149 in the predicted catalytic triad causes conditional crystal maturation defects, magnetite nucleation is not affected. Interestingly, H116 and D149 participate in a hydrogen bond on the opposite face of loop LC from the H148/H263 metal binding motif, suggesting that the conditional phenotypes could be due to temperature-dependent flexibility near the metal binding site ( S5 Fig ). A potential link between the two motifs is consistent with the reported inhibition of protease activity through a highly analogous metal binding site in the kallikrein family that rearranges the H-D catalytic pair ( S5 Fig ). While templating of magnetite growth via an interaction between biomineralization factors and the mineral surface has been proposed, our findings with MamO emphasize that direct interactions with individual solute ions also play a role [ 34 , 38 , 39 ]. One of the most fascinating aspects of MamO’s metal ion interaction is that the H148A and H263A forms of MamO maintain the ability to bind metals but cannot support any magnetite biosynthesis in vivo. It appears that binding is insufficient and that the precise coordination geometry must be maintained, leading us to speculate that MamO directly promotes nucleation by guiding individual iron atoms into the magnetite lattice. This model is consistent with the phenotypes observed in vivo, the modest binding affinity and the surface exposed nature of the simple dihistidine motif. Additionally, it agrees with topological predictions for MamO placing the protease domain in periplasm, which is continuous with the magnetosome lumen in AMB-1 [ 18 ]. More broadly, our results define an unexpected mechanism for MamO in biomineralization. It appears to have lost the ability to perform serine protease activity and instead performs two noncatalytic functions: direct metal binding to promote magnetite nucleation and activation of MamE’s proteolytic activity ( Fig 8A ). 10.1371/journal.pbio.1002402.g008 Fig 8 MamO in magnetosome formation and evolution. (A) A dual role in biomineralization. Distinct regions of the protein contribute to each activity separately. The protease domain promotes nucleation by binding iron and the TauE domain manipulates solute conditions that regulate MamE’s activity. (B) Specialization of the trypsin-like protease family in magnetotactic bacteria through gene duplication and subsequent neofunctionalization. In addition to the surprising mechanism for MamO in AMB-1, we uncovered a fascinating evolutionary expansion of the trypsin family within magnetotactic bacteria. Our analysis suggests that the ancestral MAI contained a single trypsin-like protease homologous to MamE. The δ- and γ- Proteobacteria experienced independent duplications of this ancestral enzyme, while the α- Proteobacteria appear to have acquired a second, distantly related trypsin-like protease. Despite these different origins, having two redundant proteases seems to have allowed one copy to lose its catalytic ability in all three clades ( Fig 8B ). In α- Proteobacteria , MamO specialized to promote biomineralization through the two noncatalytic activities identified here. While the pseudoproteases in the other clades remain uninvestigated, the fact that inactive copies are retained strongly suggests that they also play important noncatalytic roles. The pathway that led to convergent evolution of pseudoproteases in magnetotactic organisms highlights the critical role duplication and redundancy play in facilitating diversification of protein function [ 40 ]. Perhaps more intriguing is the fact that MamO’s metal binding motif is placed at the same site on the chymotrypsin fold as the highly analogous zinc-binding site seen in the distantly related kallikreins [ 31 , 32 ]. This hints toward the possibility that the ability to bind metals may be a latent biochemical function carried within the fold. Such activities are absent in specific evolutionary states of a protein but can quickly surface under selective pressure [ 41 , 42 ]. Consistent with this, trypsin-like proteases utilize catalytic residues on loops that are well separated from the core, a property termed fold polarity that correlates with the capacity for functional diversification [ 43 , 44 ]. Perhaps neofunctionalization of the trypsin scaffold within magnetotactic organisms is due to an inherent stability and adaptability in the fold that makes it a useful building block for biochemical innovation."
} | 3,651 |
25425235 | PMC4251993 | pmc | 1,993 | {
"abstract": "ABSTRACT Dissimilatory metal-reducing bacteria, such as Geobacter sulfurreducens , transfer electrons beyond their outer membranes to Fe(III) and Mn(IV) oxides, heavy metals, and electrodes in electrochemical devices. In the environment, metal acceptors exist in multiple chelated and insoluble forms that span a range of redox potentials and offer different amounts of available energy. Despite this, metal-reducing bacteria have not been shown to alter their electron transfer strategies to take advantage of these energy differences. Disruption of imcH , encoding an inner membrane c- type cytochrome, eliminated the ability of G. sulfurreducens to reduce Fe(III) citrate, Fe(III)-EDTA, and insoluble Mn(IV) oxides, electron acceptors with potentials greater than 0.1 V versus the standard hydrogen electrode (SHE), but the imcH mutant retained the ability to reduce Fe(III) oxides with potentials of ≤−0.1 V versus SHE. The imcH mutant failed to grow on electrodes poised at +0.24 V versus SHE, but switching electrodes to −0.1 V versus SHE triggered exponential growth. At potentials of ≤−0.1 V versus SHE, both the wild type and the imcH mutant doubled 60% slower than at higher potentials. Electrodes poised even 100 mV higher (0.0 V versus SHE) could not trigger imcH mutant growth. These results demonstrate that G. sulfurreducens possesses multiple respiratory pathways, that some of these pathways are in operation only after exposure to low redox potentials, and that electron flow can be coupled to generation of different amounts of energy for growth. The redox potentials that trigger these behaviors mirror those of metal acceptors common in subsurface environments where Geobacter is found.",
"introduction": "INTRODUCTION Geobacter sulfurreducens is a model dissimilatory metal-reducing anaerobe able to completely oxidize organic compounds inside the cell and transfer the resulting electrons to terminal acceptors beyond the outer membrane ( 1 , 2 ). Extracellular electron acceptors utilized by G. sulfurreducens include chelated transition metals, particulate Fe(III) and Mn(IV) oxides ( 2 ), and electrodes poised at oxidizing redox potentials ( 3 ). Geobacter representatives are abundant in anoxic metal-reducing habitats, including aquatic sediments ( 2 ), Fe(III)-rich petroleum-contaminated sites ( 4 ), zones where U(VI) reduction is stimulated by organic acid addition ( 5 ), subsurface aquifers where Fe(III) reduction releases arsenic into drinking water ( 6 , 7 ), and on electrodes used to produce electrical energy ( 8 , 9 ). Despite their contribution to global biogeochemical processes and emerging biotechnological applications, the molecular mechanism for electron transfer across the inner membrane of Geobacter is not known, and there is no respiratory protein-based marker for monitoring the activity of these ubiquitous metal-reducing bacteria in their natural environment. Part of the difficulty in studying Geobacter stems from the diversity of redox proteins potentially utilized by these organisms for respiration. Geobacter genomes typically encode 60 to 90 multiheme c -type cytochromes, few of which are conserved between all species ( 10 ). This is in stark contrast to another well-studied metal-reducing family, the Shewanellaceae . This group of facultative anaerobes encodes a single inner membrane NapC/NirT family tetraheme cytochrome (CymA) ( 11 ) that passes electrons to outer membrane conduits comprised of two decaheme c -type cytochromes ( 12 – 14 ). Synthesis of all proteins involved in the Shewanella pathway is simply induced by a shift to anaerobic conditions, rather than the presence of metals, and deletion of the CymA inner membrane cytochrome eliminates growth with all extracellular electron acceptors ( 11 , 15 ). In contrast, G. sulfurreducens exhibits a complex transcriptional response to different extracellular electron acceptors ( 16 – 19 ), and no single deletion eliminates electron transfer to all electron acceptors ( 20 , 21 ). Despite this evidence for complexity, published models of the Geobacter electron transport chain invoke a single Shewanella -like route from the quinone pool to an array of outer surface proteins able to interact with soluble compounds, which then are suggested to pass electrons further to extracellular proteins interacting with larger acceptors ( 22 – 24 ). If such models are true, mutants defective in reduction of soluble metals, such as Fe(III) citrate, should also be defective in utilization of all insoluble acceptors, such as Fe(III) oxides. Such a hypothesis is challenged by directed mutant ( 20 , 25 ) and transposon mutagenesis ( 21 ) studies that continue to find G. sulfurreducens mutants defective in electron transfer to only a subset of extracellular acceptors. The array of Geobacter mutant phenotypes argues against a simple single pathway, as well as models where different proteins are required based on solubility (chelates versus oxides) or metal content (Fe versus Mn) of the acceptor. In this report, we describe how redox potential explains many of these discrepancies, through discovery of ImcH, an inner membrane c- type cytochrome in G. sulfurreducens . The imcH mutant reduced insoluble Fe(III) oxides yet could not reduce insoluble Mn(IV) oxides or chelated Fe(III). Because of the low electron-accepting potential of Fe(III) oxides relative to these other compounds, poised electrodes were used to investigate the role of redox potential. A switch from high to low potential induced respiration of the imcH mutant but only when electrodes were poised at a sufficiently low redox potential of −0.1 V (versus the standard hydrogen electrode [SHE]). These experiments are consistent with different pathways being used by G. sulfurreducens for transfer of electrons out of the quinone pool to low- versus high-potential acceptors and suggests a mechanism for sensing the redox potential of extracellular objects. These findings also provide a molecular explanation for recent electrochemical evidence supporting at least two separate electron transfer pathways out of Geobacter electrode-grown cells ( 26 ). Since Fe(III) naturally occurs in up to 15 different oxide or oxyhydroxide forms ( 27 ), spanning more than half a volt of redox potential ( 28 , 29 ), multiple electron transfer pathways could be utilized by Geobacteraceae in response to the energy available in environmentally relevant metals and provide an explanation for cytochrome diversity in these organisms.",
"discussion": "DISCUSSION These results show that G. sulfurreducens cannot transfer electrons to high redox potential electron acceptors without the inner membrane multiheme c- type cytochrome ImcH. However, after exposure to low redox potentials, either in the form of certain Fe(III) oxides or poised electrodes, imcH mutants are able to utilize an alternative pathway that enables respiration and growth. The growth rate of wild-type and mutant cultures at lower redox potentials is slower, supporting the hypothesis that these strategies are distinct and that cells generate less ATP per electron when using lower-energy strategies. Based on these data, G. sulfurreducens has electron transfer pathways which can be differentially utilized in response to the redox potential of an extracellular electron acceptor, such as a metal oxide or an electrode. Since the potentials of many environmentally relevant metals lie on opposite sides of the threshold able to trigger this choice, these findings help explain conflicting Geobacter mutant phenotypes, suggest that genetic markers may exist for monitoring subsurface redox conditions, and imply that bacteria can sense the redox potential of terminal electron acceptors. With its transmembrane helices, N-terminal NapC/NirT homology, and c -type hemes, ImcH represents a new family of bacterial redox proteins implicated in extracellular electron transfer. Homologs of ImcH are present in all Geobacteraceae isolated for a metal-reducing phenotype, as well as related metal-reducing Anaeromyxobacter spp . , but are notably absent in fermentative ( Pelobacter spp . ) and chlororespiratory ( Geobacter lovleyi ) members of this cluster ( Fig. 5 ). Homologs of imcH are found in the Acidobacteria and Planctomyces - Verrumicrobia-Chlorobium phyla, which contain Fe(III)-reducing genera ( Geothrix and Melioribacter , respectively) ( 38 , 39 ), suggesting that other relatives should be reinvestigated for their respiratory abilities ( Fig. 5 ). With the increasing availability of single-cell and metagenomic sequences, imcH may serve as an aid for prediction of respiratory strategies, but discovery of what proteins interact with ImcH will greatly improve sequence-based predictions. FIG 5 ImcH homologues are widely distributed. Only proteins with at least 75% total of the sequence length of Geobacter sulfurreducens ’ ImcH were included. Alignment was performed using the default settings in the ClustalO program, with FigTree v 1.4.0 used to generate the graphical representation of the alignment. Clusters with 50% or greater identity were collapsed. Clusters in which pure culture-based evidence for extracellular electron transfer has been found are shown in red, with the representative species/strain names also shown in red. In the laboratory, electrochemical systems are routinely used to study electron transfer kinetics ( 3 , 36 ), biofilm development ( 21 , 37 , 40 ), and long range electron flow between metal-reducing bacteria ( 40 – 42 ), with the justification that these insights help explain reduction of environmentally relevant metal oxides. These data show that most electrodes are not poised at the proper potential to mimic electron transfer to poorly crystalline Fe(III) oxides, such as ferrihydrite ( Fig. 6 ), typically the largest reservoir of reducible Fe(III) in sediments and aquifers ( 27 ). Based on recent measurements, the redox potential of laboratory goethite is −0.17 V versus SHE ( 43 ), and ferrihydrite/Fe(II) mixtures begin at ~0.0 V versus SHE but decrease rapidly to −0.1 V or lower as Fe(II) accumulates in the medium ( 28 , 32 , 44 ). The difference between high-potential laboratory electrode experiments and environmental Fe(III) oxide electron acceptors may explain how outer membrane cytochromes, such as OmcS or OmcB, can appear to be important to reduction of Fe(III) oxides but not electrodes and lead to confusion when proposing electron transfer models ( 33 , 45 , 46 ). FIG 6 Summary of reduction potentials tested and general model for inner membrane electron transfer pathways. Reduction potentials versus standard hydrogen electrode (SHE) for all metal oxides, chelates, and electrodes tested as electron acceptors for wild-type and Δ imcH:: Kan r strains in this study. Due to the heterogeneity of metal oxides, a range of potentials from previously reported values are shown. Reduction of the electron acceptor by a given strain is indicated as “+,” whereas lack of reduction is indicated as “−.” According to this model, cells typically route electron flux out of the quinone pool via an ImcH-dependent pathway for reduction of high-potential acceptors, while an as yet undiscovered pathway is used for low-potential acceptors. Looking forward, the observation that growth at low redox potential acceptors requires different proteins and induces a characteristically lower growth rate produces a number of hypotheses which can be tested in Geobacter . First, multiple inner membrane quinone oxidoreductases must exist in G. sulfurreducens , with at least one that is essential for low-potential respiration. These low-potential pathways should demonstrate a lower midpoint potential for electron transfer and result in a lower H + /e − stoichiometry than the ImcH-dependent pathway. At least five candidates for alternative quinone oxidoreductases exist in the G. sulfurreducens genome, most containing b -type cytochromes next to periplasmic c -type cytochromes. These “ cbc family” proteins are among the most conserved within the metal-reducing Geobacteraceae ( 10 ). Unlike the case of ImcH, which proteomic studies show to be abundant under all conditions ( 17 , 18 ), cbc family protein levels are often altered in the presence of different metals ( 20 ), and mutants lacking these proteins yield intermediate phenotypes depending on the metal acceptor ( 21 ). It is possible that studies attempting to uncover the role of cbc family proteins were confounded by the presence of other electron transfer pathways, such as the ImcH-dependent pathway, and the need to control redox potentials more precisely to observe phenotypes. A second hypothesis is that each inner membrane cytochrome interacts with separate periplasmic and/or outer membrane redox proteins to create independent conduits out of the cell, based on external redox potential. This could explain why different acceptors induce differential expression of the five periplasmic triheme c- type cytochromes (PpcA to -E), why there are many outer membrane cytochromes of G. sulfurreducens , and why strains evolved to overcome some cytochrome deletions regain only 60% of the wild-type growth rate after mutations that lead to expression of repressed or cryptic cytochromes ( 46 ). With the identification of ImcH, it may be possible to construct strains lacking one or more routes out of the Geobacter cell and study electron transfer pathways independently of each other. In nearly every microbial respiration, separate strategies are utilized in response to thermodynamic constraints. Oxygen is reduced by multiple terminal oxidases, depending on available concentrations ( 47 ). Nitrate reductases exist in the periplasm or cytoplasm, leading to differences in energy conservation ( 48 ). Different methanogens compete at distinct thermodynamic thresholds of hydrogen and acetate ( 49 ). Since their discovery, metal-reducing bacteria have been outliers, despite the fact that for every 0.5-V difference in redox potential [such as with Mn(IV) oxides versus Fe(III) oxides], enough energy is available to make at least one extra ATP/2e − . These new findings show that Geobacter also has a response to this thermodynamic challenge. By sensing and responding to the energy available in external surfaces, Geobacter ’s respiratory strategy may provide a competitive advantage in subsurface and syntrophic habitats where redox potential, rather than substrate concentration, can be the dominant environmental variable."
} | 3,651 |
35514730 | PMC9065573 | pmc | 1,994 | {
"abstract": "Surfaces that have unique wettabilities and are simultaneously superhydrophobic with water contact angles > 150°, and superoleophilic with oil contact angles < 5°, are of critical importance in the oil/solvent–water separation field. This work details the facile preparation of highly efficient oil–water separation devices that successfully combine hierarchical surface roughening particles and low surface energy components with porous substrates. Coatings were generated using TiO 2 and hydrophobic-SiO 2 micro/nanoparticle loadings which were then embedded within polydimethylsiloxane, commercially known as Sylgard® 184, and 1 H ,1 H ,2 H ,2 H -perfluorooctyltriethoxysilane (FAS) polymer mixtures. The resulting slurries were dip coated onto copper meshes with varying pore diameters (30, 60 and 100 meshes had 595, 250 and 149 μm pore dimensions respectively). Functional testing proved that mesh substrates coated in the lowest Sylgard® 184 : FAS polymer ratio formulations displayed heightened water repellency and retained their superoleophilic properties upon repeat testing. The largest average water contact angle of 145 ± 1°, was recorded on a copper 30 mesh substrate with a coating comprising H-SiO 2 microparticles and TiO 2 nanoparticles in a 1 : 9 polymer mixture of Sylgard® and FAS. The coating's extreme oil affinity was supported by high solvent–water separation efficiencies (≥99%) which withstood numerous testing/washing cycles.",
"conclusion": "4. Conclusion A facile route to fabricate a superior (super)hydrophobic and oleophilic separation surface was devised throughout this research. Hierarchical surface roughening in combination with silicon-based polymers generated water repellent coatings with oil-loving properties. Numerous different combinations and concentrations of dual scale TiO 2 and/or hydrophobic-SiO 2 (H-SiO 2 ) particles in varying Sylgard®184 : FAS polymer ratio mixtures were explored in order to enhance surface wettabilities. Average water contact angles were linked to improved solvent separation efficiencies, which, in turn, were most favourable on samples with higher quantities of the fluorine-rich fluoroalkylsilane, FAS, polymer. For example, coating F's dual scale TiO 2 particle system embedded in a 1 : 4 Sylgard® 184 : FAS polymer mixture was applied to copper 60 mesh substrates (pore diameter of 250 μm). The resulting device was highly functional with 100 ± 0%, 85 ± 0% and 97 ± 0% separation efficiencies for toluene–, hexane– and dichloromethane–water solutions respectively. A trade-off between the highly adhesive and durable Sylgard® 184 polymer and the low surface energy FAS, in a 1 : 4 ratio, afforded robust separation coatings that showed no deviation in functionality after 3 wash–separation–wash cycles nor after the Scotch tape test.",
"introduction": "1. Introduction The increasing volume of oil and/or solvent contamination in industrial wastewater is a lasting contribution to the deterioration of our marine and terrestrial ecosystems. The food, leather, textile and petrochemical industries all generate significant amounts of unwanted oily water during daily production. 1 When these continuous waste outputs couple with freak accidents, such as oil spills, the natural environment suffers severe implications for up to one decade after each isolated event. 2 During the period between 2010 and 2017 there was more than 47 000 tonnes of petrochemical oil spilt in aquatic environments, 80% of this amount was lost in just ten incidents. 3 It is therefore of great importance to engineer efficient oil–water separation techniques capable of rapidly cleaning the environment in the event of future spills and to decontaminate oily industrial waste. Interactions between liquids and surfaces are central to this field of research. Superhydrophobic materials often make use of nano and/or microscale protrusions combined with very low surface energy modifications to achieve static water contact angles ≥ 150°. 4,5 Water droplets upon these materials can exist in a Wenzel state, whereby the droplets ‘stick’ to all parts of the surface, or Cassie–Baxter regime where droplets ‘slip’ on a lubricating air layer trapped within surface asperities. 6,7 The literature details various ways of fabricating novel water repellent surfaces inspired by nature's waterproof materials such as lotus leaves butterfly wings and geckos' feet. 5 Plasma enhanced chemical vapor deposition (PECVD) was carried out by Bico et al. to generate carbon flake-like multiscale surface structures. 8 Aerosol-assisted chemical vapour deposition (AACVD) of Sylgard® 184 or polydimethylsiloxane (PDMS) and tetraethyl orthosilicate (TEOS) afforded appropriately rough and transparent coatings on glass substrates. 9–11 Alternative superhydrophobic generating methods such as chemical etching, 12 electrospinning 13,14 and lithographic imprinting/templatation 15 have enhanced the functionality of oil–water separation, self-cleaning, anti-icing and anti-corrosion materials. 12,16–18 Whilst water repellency is crucial to certain oil–water separation devices, materials with two wettability regimes have been known to improve separation efficiency. A surface displaying both superhydrophobic ‘water-hating’ and superoleophilic ‘oil-loving’ properties can be tuned for selective separation providing its surface energy resides between that of the solvent/oil and water it aims to separate (approximately between 30 mN m −1 and 70 mN m −1 ). 19 Filtration and/or absorption are successful types of oil-removing techniques that encompass both size exclusion and wettability selectivity mechanisms. Pore sizes are commonly designed to either let oil run straight through the mesh/membrane or for the material to absorb the hydrocarbon whilst water is repelled. 20 Copper meshes have been etched with nitric acid and alkali (NaOH/K 2 S 2 O 8 ) solutions before modification with self-assembled monolayers of hexadecanethiol and dodecanethiol respectively. High water contact angles, ∼150°, and low oil contact angles, <5°, produced materials that separated at least 97% of water from an oily solution. 21,22 Vertically-aligned multi-walled carbon nanotubes were synthesised on a stainless steel mesh via CVD. Needle-likes tubes created a high surface area and aided capillary action; the oil contact angle was recorded at 0° and oil penetrated the mesh in 0.4 s. 23 Li et al. fabricated candle soot and silica coated meshes that repelled hot water and corrosive liquids. These coated meshes worked as effective gravity separation devices with ∼99% of organic solvents permeating through the device. 24 Nano protrusions have also been introduced by electrochemical etching, a copper mesh was anodized in a NaOH solution (1 M) to create a needle-like Cu(OH) 2 film. The roughened substrate was then coated in a self-assembled monolayer of 1 H ,1 H ,2 H ,2 H -perfluorooctyltriethoxysilane (FAS) to form a durable solvent/water separator. 25 Spin coating metal mesh substrates with ZnO nanorods has been yet another effective fabrication method. 26 Swapping a porous substrate for porous structures (on special materials) enhances robustness by removing the possibility for coating stripping. Tu et al. designed a micro-bead and nanofiber film created in one step by spraying onto any given substrate. The water contact angle on these surfaces was not as high as other but the cost and fabrication time were drastically reduced. 27 A thermoplastic polyurethane mat, created via electrospinning, immersed in a hexadecyltrimethoxysilane (HDTMS) modified nanosilica solution also generated an efficient structured separation material. 28 Several other surface modification methods such as electrostatic deposition, grafting polymerisation, spray drying and photo-initiated polymerisation have gained traction by utilising various polymers, gels and biomaterials. 29,30 Herein we present an extremely facile one-pot method to fabricate highly efficient oil–water separation devices, >99% efficiency. Combinations of hierarchical surface roughening particles and low surface energy components upon porous substrates lead to enhanced robustness and functionality when tested with a variety of common industrial solvents; this elevated device effectiveness across a substrate and solvent range has substantially advanced work in this area. Here, TiO 2 and hydrophobic-SiO 2 micro/nanoparticles were embedded within polydimethylsiloxane (Sylgard® 184) and 1 H ,1 H ,2 H ,2 H -perfluorooctyltriethoxysilane (FAS) polymer mixtures. Resulting slurries were dip coated onto copper meshes with varying pore diameters (30, 60 and 100 meshes had 595, 250 and 149 μm pore dimensions respectively). Water contact angles, solvent separation efficiencies and mechanical stability were found to outperform many of the existing optimum superhydrophobic/superoleophilic filtration surface previously documented.",
"discussion": "3. Results and discussion A one-pot synthesis was used to generate a range of highly functional ‘water hating’ and ‘oil loving’ coatings for application in oil–water separation. Various iterations of TiO 2 and hydrophobic-SiO 2 (H-SiO 2 ) micro/nanoparticles were dispersed in Sylgard® 184 and 1 H ,1 H ,2 H ,2 H -perfluorooctyltriethoxysilane (FAS) mixtures, as detailed in Tables S1, † 1 and 2. The resulting coatings were dip coated onto 3 types of copper meshes (30, 60 and 100 meshes had 595, 250 and 149 μm pore dimensions respectively). 3.1. Characterisation X-ray photoelectron spectroscopy (XPS) was carried out on the functional coatings to identify the oxidation state of Ti and chemical environments surrounding Si atoms, Fig. 3 . Elemental scans of the dried optimised TiO 2 particle containing coatings (D, E, F and G (formulations listed in Table 2 )) confirmed that Ti was constantly in the +4 oxidation state; a 2p doublet was characteristic of the 459.00 eV Ti 2p 3/2 and the 464.70 eV Ti 2p 1/2 peaks. Additionally, optimised coating formulations D, E and G also contained SiO 2 particles. For sample D, consisting of SiO 2 5–15 μm particles (0.6 g) and TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS, the Si 2p peak was deconvoluted into a 104.00 eV Si–O 2 and a 101.82 eV Si–OR environment. Values agreed with the literature and were consistent with other TiO 2 and SiO 2 particle containing samples. 31–34 Fig. 3 Deconvoluted X-ray photoelectron spectroscopy (XPS) Ti 2p and Si 2p scans of coating D comprising SiO 2 5–15 μm particles (0.6 g) and TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS. The X-ray diffraction pattern of coating D displays a high intensity TiO 2 anatase peak at 25.4° and subsequently less intense peaks at 36.7°, 37.8°, 38.8°, 47.9°, 54.6°, 55.1° and 62.5°. A medium intensity rutile peak was identified at 27.4° and SiO 2 particles were attributed to the quartz peak at 20.8°. As coating F was absent of SiO 2 particles the only detectable peaks were characteristic of the anatase and rutile phases of TiO 2 . All coating patterns presented in Fig. S1 † agreed with particle standards and the Inorganic Crystal Structure Database. Fourier transform infrared (FT-IR) analysis, consistent with the NIST Standard Reference Database, resulted in the identification of several bending and stretching modes characteristic of the Sylgard® 184 and 1 H ,1 H ,2 H ,2 H -perfluorooctyltriethoxysilane (FAS) polymer mixtures, Fig. 4 . Samples containing hydrophobic-SiO 2 (H-SiO 2 ) and/or TiO 2 particles embedded in 1 : 0, 0 : 1, 1 : 1, 1 : 4 and 1 : 9 Sylgard® 184 : FAS ratios all had similar polymer transmittance bands. A sharp peak at 3120 cm −1 was seen in the spectra of coatings D, E, F and G which depicted the C–H alkane stretch present in both polymers. Varying intensities of the following peaks were observed: 1257 cm −1 (sh, w) intense out of phase vinyl ether stretch in Sylgard® 184, 35 1010 cm −1 (s) Si–OR Sylgard® 184 and FAS stretch, 790 cm −1 (m) alkene out of plane bending in Sylgard® 184, 622 cm −1 (w) C–F FAS stretch and 457 cm −1 (w) antisymmetric Si–(CH 3 ) Sylgard® 184 stretch. 35 The presence of these C–F FAS stretches were synonymous with functional effectiveness; fluorine's extreme electronegativity made it only weakly susceptible to fleeting dipoles that form the basis of van der Waals forces. Consequently, the samples' fluorinated carbon chains had small intermolecular forces and therefore the low surface energy hydrophobic requirement. Fig. 4 Fourier transform infrared (FT-IR) spectrum of coating D; SiO 2 5–15 μm particles (0.6 g) and TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS. Scanning electron microscopy (SEM) images provided visual information about the topography of separation materials, particle size/distribution and the extent of copper mesh substrate pore blockage. This qualitative evidence was attributed to water contact angle measurements as well as separation efficiency results. Initial trial coatings containing TiO 2 particles in high ratios of Sylgard® 184 to FAS (1 : 1 or 1 : 0) resulted in almost complete pore blockage, Fig. S2. † As expected, pore obstruction physically impeded the passage of any solvent through the membrane and resulted in poor separation efficiencies < 25%. Coatings with higher proportions of the more viscous Sylgard® 184 polymer were also seen to reduce the prominence of particle roughening structures thus had reduced average water contact angle values, <110°. \n Fig. 5a presents SEM images of the unobstructed pores of a 30 mesh (595 μm pore dimension) copper substrate functionalised with coating D (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS). Dual scale surface structures were accentuated in this coating due to the reduced quantity of Sylgard® 184 and were reflected in the high average water contact angle, 145 ± 1°. However, solvent–water separation testing indicated that the pores were too large to selectively filter solvents from water and so the use of copper 30 mesh substrates were discontinued. Fig. 5b displays a section of coating D adhered to a copper 60 mesh substrate (250 μm pore dimension). Once again, imaging proved that pore openings were clear. The average water contact angle remained relatively high, ∼135°, and the toluene–water separation efficiency in equal parts was recorded as 100%. Fig. 5c shows images of coating E (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 0 : 1 polymer mixture of Sylgard® 184 and FAS) on copper 100 mesh substrates (149 μm pore dimensions). Open pore structures were conserved, due to the absence of Sylgard® 184, and there was a clear presence of protruding micro and nanoscale particles. The associated average water contact angle and toluene–water separation efficiency in equal parts were 147 ± 1° and 99 ± 1% respectively. Lastly, Fig. 5d displays coating F on a copper 100 mesh substrate (TiO 2 60–200 nm particles (1.5 g) with TiO 2 21 nm particles (1.5 g) in a 1 : 4 polymer mixture of Sylgard® 184 and FAS). The small pores were ∼50% blocked by this polymer mixture making this an unsuitable solvent–water separation device despite preserving dual scale surface contours, average water contact angle of 146 ± 1°. SEM analysis lead to the assumption that coatings with high proportions of the non-viscous fluorinated FAS polymer mixture created highly structured surface topographies and maximised the appropriate pore area of copper 60 mesh (and in some cases copper 100 mesh) substrates. This in turn elevated the average contact angle values and enhanced efficiency of solvent separation; low Sylgard® 184 levels were tolerated to preserve device coating robustness. Fig. 5 SEM images of (a) coating D on a copper 30 mesh substrate (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS), (b) coating D on a copper 60 mesh substrate (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS), (c) coating E on a copper 100 mesh substrate (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 0 : 1 polymer mixture of Sylgard® 184 and FAS) and (d) coating F on a copper 100 mesh substrate (TiO 2 60–200 nm particles (1.5 g) with TiO 2 21 nm particles (1.5 g) in a 1 : 4 polymer mixture of Sylgard® 184 and FAS). 3.2. Functional testing Average water contact angle results are plotted in Fig. 6 to demonstrate the impact of changing particle loading/combination and polymer mixture on hydrophobicity. The highly durable, viscous and adhesive Sylgard® 184 polymer had to be delicately balanced with the easily abraded, non-viscous and highly water repellent fluoroalkylsilane (FAS) polymer to achieve a formulation that wouldn't compromise membrane pore size (for effective solvent–water separation) but maintain functionality and durability. Coatings were prepared with 1 : 0, 0 : 1, 1 : 1, 1 : 4 and 1 : 9 Sylgard® 184 : FAS polymer ratios with embedded hydrophobic-SiO 2 (H-SiO 2 ) and/or TiO 2 loading combinations and functionally contrasted. With the majority of polymer combinations dual scale H-SiO 2 particles produced the lowest average water contact angle values, for example 125 ± 10° was obtained with the 0 : 1 Sylgard® 184 : FAS polymer ratio mixture. Dual scale TiO 2 particles consistently generated the highest average water contact angle results for each coating composition (with the exception of the Sylgard® 184 : FAS ratio of 1 : 0), a maximum of 151 ± 7° was achieved with the 1 : 9 Sylgard® 184 : FAS polymer mixture. The coatings that contained the highest concentration of Sylgard® 184 often performed comparatively worse than mesh coatings with a larger FAS content. These findings were expected, the abundance of –C–F polymer bonds in FAS is known to effectively reduce a material's surface energy and consequently increase the average water contact angle value. All samples were in the Cassie–Baxter regime, assumed from low tilting angle results, and improved upon the hydrophobicity of separation meshes detailed in the literature. Fore example. Xue's group generated a superhydrophilic mesh whilst Zhang's coated mesh afforded average water contact angles < 130°. 36,37 Fig. 6 Average water contact angle data on coatings containing various particle combinations and loadings embedded in a Sylgard® 184 and FAS polymer mixture. Abbreviated particle combination and loadings are as follows; A – TiO 2 60–200 nm particles (0.6 g) with TiO 2 21 nm particles (0.6 g), B – SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) and C – TiO 2 60–200 nm particles (0.6 g) with TiO 2 21 nm particles (0.6 g). Each particle combination was embedded in 0 : 1, 1 : 0, 1 : 1, 1 : 4 and 1 : 9 Sylgard 184 : FAS ratios. Coatings were applied to copper 60 mesh substrates prior to oven drying. Error bars show the maximum and minimum values obtained after three repeat readings. Average percentage of toluene separated from 25%, 50% and 75% toluene solutions (75%, 50% and 25% water respectively) were contrasted on coated 60 mesh copper substrates, via the setup in Fig. S3. † The H-SiO 2 and TiO 2 particle combination, B, embedded in a 1 to 9 Sylgard® 184 : FAS mixture was the most effective at separating 25% toluene–50% water solutions on copper 60 mesh substrates (95 ± 6% separation efficiency). Fig. 7 illustrates that all H-SiO 2 /TiO 2 particle combinations, A, B and C, reached 100% separation efficiencies for 50% toluene–50% water solutions when coupled with their optimised Sylgard® 184 : FAS polymer mixture. Subsequently, high separation efficiencies of 99 ± 1% were achieved on dual scale H-SiO 2 as well as H-SiO 2 /TiO 2 containing particle systems embedded in a 1 to 9 Sylgard® 184 : FAS mixture when tested with 75% toluene–25% water solutions. All particle and polymer mixture combinations had slightly reduced separation efficiencies for 25% toluene–75% water solutions but remained >90%. Fig. 7 Average percentage of toluene separated from a 50% toluene–water solution on coatings containing various particle combinations and loadings embedded in a Sylgard® 184 and FAS polymer mixture. Abbreviated particle combination and loadings are as follows; A – TiO 2 60–200 nm particles (0.6 g) with TiO 2 21 nm particles (0.6 g), B – SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) and C – TiO 2 60–200 nm particles (0.6 g) with TiO 2 21 nm particles (0.6 g). Each particle combination was embedded in 0 : 1, 1 : 0, 1 : 1, 1 : 4 and 1 : 9 Sylgard 184 : FAS ratios. Coatings were applied to copper 60 mesh substrates prior to oven drying. Error bars show the maximum and minimum values obtained after three repeat readings. Final separation efficiencies on optimised coatings on copper 60 and 100 separation meshes have been recorded in Table S2. † Coating D (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS) coating E (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 0 : 1 polymer mixture of Sylgard® 184 and FAS), coating F (TiO 2 60–200 nm particles (1.5 g) with TiO 2 21 nm particles (1.5 g) in a 1 : 4 polymer mixture of Sylgard® 184 and FAS) and coating G (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 1 : 4 polymer mixture of Sylgard® 184 and FAS) were functionally tested using various solvent systems. Table S2 † documents solvent separation efficiencies for toluene–, hexane– and dichloromethane–water solutions along with associated errors. The results indicated that coated copper 60 mesh substrates again were most likely to generate a favourable separation efficiency when compared to coated copper 100 mesh substrates. The larger pore diameter remained unblocked after coating with even the most viscous of polymer mixtures and therefore allowed toluene, hexane and dichloromethane to filter through the mesh; small pore 100 mesh substrate blockages most strongly inhibited the densest solvent, dichloromethane, from filtering through the device and consequently has not been featured in Table S2. † Despite the exceptional performance of all optimised coatings D–G, formulation F on 60 mesh substrates was an extremely effective separation coating with 100 ± 0%, 85 ± 0% and 97 ± 0% efficiencies for toluene–, hexane– and dichloromethane–water solutions respectively. The innately hydrophobic dual scale TiO 2 particles embedded coating F's 1 : 4 Sylgard® 184 : FAS mixture helped elevate the average water contact angle, 146 ± 1°. This favourable particle combination coupled with the relatively high proportion of FAS in the coating elevated average contact angles and improved separation potential. Various other separation materials that used longer more elaborate fabrication procedures have been well documented over recent years. Separation efficiencies ∼ 95% were achieved but the vast number of readily available oils/solvents and the absence of a standardized separation method makes it difficult to directly contrast like results from across the field. 36–41 Due to the adhesive properties of Sylgard® 184, optimised coatings D, F and G remained comparably functional, within the original error limits represented in Fig. 6 , after 3 wash–separation–wash cycles and the Scotch tape test. Fig. 8 displays the water contact angle image post separation testing of coating D; unaltered water repellency further supports coating robustness. The 1 : 9 and 1 : 4 Sylgard® 184 : FAS ratios provided the ideal particle embedded polymer systems to preserve surface morphology and pore structures (copper 60 mesh substrates), establish preferred wettabilities and achieve surface durability. Fig. 8 Water droplet on a copper 60 mesh substrate with a coating D surface (SiO 2 5–15 μm particles (0.6 g) with TiO 2 21 nm particles (0.6 g) in a 1 : 9 polymer mixture of Sylgard® 184 and FAS). Image was recorded post separation functional testing. The most favorable separation meshes were identified as a result of ensuring surface morphologies and open pore structures were preserved whilst tailoring the surface energy to reside between that of oil and water (approximately between 30 mN m −1 and 70 mN m −1 ). 42 Trends were identified that indicated solvent density had a significant part to play. For example, copper 100 mesh separation rates were substantially higher using toluene–water solutions as opposed to hexane for identical surface coatings. This was attributed to toluene's increased density of 865 kg m −3 , 43 as opposed to 672 kg m −3 for hexane. 44 A denser solvent was able to more quickly permeate a mesh substrate when originally mixed with water."
} | 6,240 |
27386556 | PMC4928881 | pmc | 1,996 | {
"abstract": "Researchers report an organic nanowire artificial synapse that emulates working principles and energy consumption of biological synapses.",
"conclusion": "CONCLUSION In conclusion, we used ONW-based electronic devices to emulate the morphology, working principles, and energy consumption of synaptic junctions of nerve fibers. Important working principles, such as short-term plasticity, LTP, and LTD, were emulated in a single electronic device. These properties are essential for pattern recognition and associative learning. Extremely low energy consumption of the artificial synapses was realized by using core-sheath–structured ONWs across a short nanochannel. The femtojoule-level energy consumption (~1.23 fJ per synaptic event) rivals that of biological synapses (~10 fJ per synaptic event), and this achievement is a significant step toward fabrication of brain-inspired electronic devices. This work presents important progress in the development of high-density and soft brain-inspired computational systems that consume ultralow amounts of energy.",
"introduction": "INTRODUCTION Synapses are among the most important functional units for learning and memory ( 1 , 2 ). They consume an extremely small amount of energy (10 fJ per synaptic event), so a human brain consumes only as much energy as a domestic lightbulb but can outperform a supercomputer in many aspects. Recently, great efforts have been made to develop synapse-emulating circuits ( 3 ) and electronic devices ( 4 – 6 ), but they still consume orders of magnitude more energy (generally above picojoule level) than do natural synapses ( 7 – 11 ). Solving these problems requires appropriate choice of materials and design of devices; these remain difficult challenges. Here, we focus on design and fabrication of an artificial synapse. Attaining this goal presents several challenges: the first is to select a material that mimics the morphology of nerve fibers that form synaptic junction. We selected organic nanowires (ONWs) because they mimic nerve fiber morphology in terms of elongated shape, flexibility, and good scalability to large areas, which enable formation of a three-dimensional grid of crisscrossing alignment of nerve fibers in a human brain ( 12 ). This is the first morphological emulation of nerve fibers for use in an artificial synapse, which has great potential in various soft bio-inspired and bio-integrated electronics ( 13 – 18 ). The second challenge is to design a material structure that can emulate, in a single device, the important working principles of a synapse. Considering the distinct reversibility of trapping and detrapping ions in different polymer matrices, ONWs with the structure of a polyethylene oxide (PEO) sheath wrapped around a poly(3-hexylthiophene-2,5-diyl) (P 3 HT) core were fabricated to facilitate the development of short-term and long-term plasticity. Furthermore, we aimed to achieve an artificial synapse that has energy consumption comparable to that of a biological synapse. This low energy consumption is essential to constructing highly integrated, very large-scale neuromorphic computing systems. With a nanowire lithographic 300-nm channel length, ~1.23 fJ per synaptic event was attained, which is the lowest attained so far and even rivals that of biological synapses.",
"discussion": "RESULTS AND DISCUSSION We designed ONW synaptic transistor (ST) architecture to emulate the functions of a biological synapse ( Fig. 1 ). A typical ONW ST was composed of a conducting line probe (A′), an ion gel, an ONW (B′), and two metal contact pads on a substrate (fig. S1). The conducting line probe (A′), which mimicked a biological axon and a presynaptic membrane, was in contact with the ion gel to supply signals analogous to presynaptic spikes (fig. S2). The ion gel provided mobile ions that migrate in response to the presynaptic spikes. The ion gel between the probe and ONW emulates the synaptic cleft, which is ionically conductive to allow chemical transmission across it but is electrically insulating to separate electrical input and output and thereby to ensure that synaptic processes can occur independently at the presynaptic membrane and postsynaptic membrane. The core-sheath ONW (B′) has a P 3 HT inner core in a PEO sheath (fig. S3). The ONW combined with a drain electrode mimics a biological dendrite (B). The core-sheath structure of ONW underlies the mechanism of short-term and long-term plasticity in the artificial synapse. The presynaptic spikes provide an electrical field that helps ions penetrate the PEO sheath or even the P 3 HT core. The process in which ions penetrate the PEO sheath under presynaptic spikes and later rapidly diffuse back to the ion gel induces short-term plasticity. A number of consecutive presynaptic spikes increase the depth to which ions penetrate into the P 3 HT core, in which ions have low solubility and cannot easily diffuse back to the ion gel spontaneously; this restricted mobility induces long-term potentiation (LTP) (fig. S4). Here, gold electrodes were used for electrical connections, and a SiO 2 -coated Si substrate was used to provide physical support for the device. Our homebuilt electrohydrodynamic NW (e-NW) printer rapidly and inexpensively fabricated ONWs on a large scale while simultaneously controlling the location and alignment of individual ONWs ( 19 ). The spacing of ONWs was computer-controlled (fig. S5). The diameter of ONWs could be controlled by adjusting the solution concentration (fig. S6). The printing technique can print ONWs of other types of polymer semiconductor materials (fig. S7). Fig. 1 Schematic of biological neuronal network and an ONW ST that emulates a biological synapse. The conductive lines and probe (A′) mimic an axon (A) that delivers presynaptic spikes from a preneuron to the presynaptic membrane. The mobile ions in the ion gel move in the electrical field analogous to the neuron transmitters in the synaptic cleft that later induces an excitatory postsynaptic current (EPSC) in the dendrite through the contact of postsynaptic membrane. An ONW (B′) combined with a drain electrode mimics a biological dendrite (B). EPSC is generated in the ONW in response to presynaptic spikes and is delivered to a postneuron through connections to the drain electrode. A biological synapse permits a neuron to pass an electrical or chemical signal to another cell ( 20 ). Synapses are widely believed to contribute to the formation of memory, and the synaptic strength is thought to result in the storage of information by changing the amplitude of a postsynaptic current. A brief application (presynaptic spike) of voltage (−1 V, 50 ms) on the ion gel triggered an excitatory postsynaptic current (EPSC), which reached a peak value of 3.82 nA, and then gradually decayed to a resting current of ~1 nA ( Fig. 2 , A to C). This trend emulates a biological process, in which a neuron generates an action potential (spike) that propagates along the axon and is transmitted across a synapse to the next neuron. This behavior in the device originates from ion migration in the ion gel (fig. S4, A to C). Before the spike, anions and cations are distributed randomly. A negative presynaptic spike causes anions to accumulate at the ion gel–ONW interface during the spike; the high density of anions surrounding the NW attracts a certain number of holes in the P 3 HT channel, which move in response to a driving voltage to form an EPSC. After the spike, the distribution of anions returns to random, and the EPSC gradually decays. Analogously, inhibitory postsynaptic current (IPSC) was triggered by a positive presynaptic spike ( Fig. 2D ). EPSC and IPSC are complementary processes that underlie the basic features of neuronal transmission. Fig. 2 Short-term synaptic plasticity. ( A ) Schematic of application of presynaptic spikes to a postneuron through a synaptic joint. ( B ) Schematic of application of electrical pulses to an ONW ST analogous to presynaptic spikes that induces current responses through the ONW active channel. ( C ) EPSC triggered by a negative presynaptic spike (−1 V, 50 ms). ( D ) IPSC triggered by a positive presynaptic spike (1 V, 50 ms). ( E ) EPSCs triggered by two spikes, with an interspike interval of 780 ms. A1 and A2, amplitudes of the first and second EPSCs, respectively. ( F ) Schematic of applying spatiotemporally correlated presynaptic spikes onto a postneuron through two presynaptic connections to one single strand of dendrite. ( G ) Schematic of EPSC triggered by a pair of spatiotemporally correlated spikes applied to an ONW ST through two laterally coupled gates. ( H ) EPSC triggered by a single or pair of spatiotemporally correlated presynaptic spikes versus time. ( I ) Amplitude of the EPSC at t = 0 (when the spike is applied to synapse 1) versus interval Δ t pre2 − pre1 between the spikes. The change in the strength of a synapse’s response over time in response to external stimuli is referred to as synaptic plasticity and is regarded as an important foundation of learning and memory. Short-term plasticity can either strengthen or weaken a synapse for a short time on scales from tens of milliseconds to a few minutes. Short-term synaptic enhancement can be attained when two spikes arrive in rapid succession ( Fig. 2E ) ( 21 ). It is involved in encoding temporal information in auditory and visual signals ( 22 ) and has an important function in associative learning, information processing, pattern recognition, and sound source localization ( 2 , 23 – 25 ). Although many biological mechanisms remain unclear, for example, how exactly short-term plasticity is involved in pattern recognition and associative learning, scientists continue to apply existing knowledge to achieve some of these functions in artificial systems ( 4 , 25 , 26 ). Short-term synaptic enhancement was successfully mimicked in our ONW ST by applying two consecutive negative presynaptic spikes with an interspike interval (780 ≤ Δ t pre ≤ 7800 ms). The amplitude of the second EPSC peak (A2) was 1.62 times the amplitude triggered by the first spike (A1) ( Fig. 2E ). The increase in A2 was caused by residual ions that had accumulated during the first spike. Some of these ions accumulated near the ONW–ion gel interface and added to the total amount of ions accumulated during the second-spike period. The quantity of residual ions increased as Δ t pre decreased, thereby increasing the summation of EPSCs. The A2/A1 ratio decreased as Δ t pre increased (fig. S8). This spontaneous decay of current retention can be regarded as analogous to loss of memory by a human brain because the EPSC trends fit very well with a forgetting curve ( y = b × t −m ) (fig. S9). Comparably, further decreased IPSC was also obtained by applying additional positive pulses (fig. S10). Excitatory and inhibitory synapses exhibit enhancement and depression, the combined use of which underlies short-term memory. Spatiotemporally correlated stimuli from different presynapses could be used to trigger a postsynaptic current to establish dynamic logic in a neural network. The basic spatiotemporal dynamic logic was demonstrated in a simple structure that uses an ONW ST with two laterally coupled gates ( Fig. 2 , F and G). EPSC responses were recorded under a constant source-drain voltage of 0.75 V. A single spike on presynapse 1 (−1 V, 50 ms) triggered EPSC 1 with an amplitude of ~4 nA; a single spike on presynapse 2 (−2 V, 50 ms) triggered EPSC 2 with an amplitude of ~6 nA ( Fig. 2H ). When presynaptic spike 1 and presynaptic spike 2 were applied sequentially with an interspike interval (Δ t pre2 − pre1 ), the accumulation of migrated ions caused the EPSC 2 induced by the second spike to be superimposed onto the EPSC 1 induced by the first spike; as a result, the total amplitude of EPSC increased. When Δ t pre2 − pre1 = 0, the EPSC 1 and EPSC 2 were triggered simultaneously and resulted in a maximum EPSC of ~22 nA in the postsynapse. The EPSC amplitude at the end of the presynaptic spike 1 (set as t = 0, where EPSC summation was recorded) was plotted as a function of Δ t pre2 − pre1 to understand the influence of a spike on a correlated presynapse ( Fig. 2I ). The amplitude of EPSC at t = 0 is consistent with the peak amplitude of the EPSC 1 if the EPSC 2 were triggered afterward (Δ t pre2 − pre1 > 0). However, when EPSC 2 was triggered before EPSC 1 (Δ t pre2 − pre1 < 0), EPSC at t = 0 became the superimposition of EPSC 1 and the remaining EPSC 2. This is similar to the neural response to spatiotemporally correlated stimuli from different presynapses. Long-term plasticity that usually occurs at excitatory synapses includes LTP and long-term depression (LTD) ( 27 ), which are respectively a persistent increase and a persistent decrease in synaptic strength, following a number of consecutive stimulations of a synapse. Because memories are thought to be encoded by modification of synaptic strength ( 28 ), LTP is widely regarded as a mechanism that underlies learning and memory in the brain ( 22 ). LTP in the ONW ST was demonstrated by applying an increased number of presynaptic spikes ( Fig. 3A ); after this treatment, the EPSC increased by as much as ~15 times and was ~1.5 times the resting current 5 min after the spikes. LTP tends to greatly increase electrical responses of neurons to stimuli, and this increase (“potentiation”) in synaptic strength lasts much longer than do other processes that affect synaptic strength. In contrast, LTD selectively weakens specific synapses, may prevent synaptic strength from reaching some level of maximum efficacy that results from simply continued LTP, and inhibits encoding of new information ( 29 ). LTD was obtained by applying consecutive positive synaptic spikes ( Fig. 3B ). The mimicking of short-term and long-term plasticity in a single ONW ST has potential in constructing spiking neural network, which adapts to real-time functions, for example, real-time pattern recognition ( 4 , 25 ). Fig. 3 Long-term synaptic plasticity. ( A ) Postsynaptic current versus time stimulated by 30 presynaptic spikes (−1 V, 50 ms) of the ONW ST; V D (driving voltage) = 0.75 V. ( B ) Postsynaptic current versus time stimulated by 30 presynaptic spikes (1 V, 50 ms) of the ONW ST, V D = 0.75 V. ( C ) Postsynaptic current triggered by 60 negative and 60 positive pulses. ( D ) Postsynaptic current as a function of nonlinear presynaptic spikes. Application of approximately 60 cycles of negative spikes increased the EPSC amplitude by 22 times; to reduce the ONW conductance back to the initial value, 60 positive pulses were required ( Fig. 3C ). The results suggest that the present configuration of the ONW device can simulate the analog storage capacity of synapses. When a series of negative (or positive) spikes were applied, the negative ions continuously accumulated toward (or away from) the conductive channel, resulting in significant potentiation (or depression). The change rate in the synaptic strength increased with the number of spikes. However, when 60 negative (or positive) presynaptic spikes were applied, the postsynaptic current tended to saturate without further significant change due to a newly established equilibrium of ion movement. The negative presynaptic spikes drove the mobile ions to near the channel region, but the residual accumulated ions tended to drift back during the interval between the presynaptic spikes. These two trends coexisted because of the fixed number of mobile ions in the ion gel matrix. We hypothesize that the rates of the two opposite trends of ion movement became equal when sufficient ions were urged to approach the channel region by a certain number of presynaptic spikes and that at this point, equilibrium was finally established so that no further significant increase in EPSC was observed. Among simulations of various synaptic behaviors, symmetry characteristics of potentiation and depression were found to be the most important requirement. To improve the symmetry in the curve, we applied nonidentical presynaptic spikes and observed significant improvement in symmetry ( Fig. 3D ). Spike timing–dependent plasticity (STDP), which describes the change of synaptic weight depending on relative timing of the spikes of pre- and postneurons, is an important synaptic adaption rule of Hebbian learning. A typical asymmetric form of STDP (fig. S11, A and B) was obtained using an ONW ST, which could be useful in the construction of a network ( 30 – 32 ). Ultralow energy consumption is one of the most important superiorities of a neural system; to achieve this, different types of electronic devices have been investigated. Although some of those have shown picojoule-level energy consumptions, further reducing synaptic energy to the femtojoule level remains a challenging task, but is an important requirement for the construction of very large systems. To achieve significantly reduced energy consumption, we used polyvinyl carbazole (PVK) ONWs as lithographic tools to produce a narrow gap of ~300 nm between electrodes to provide a short channel length (fig. S12, A to C). Synaptic enhancement and depression of a nanogap ONW ST were also demonstrated from consecutive small presynaptic spikes (−1 mV) ( Fig. 4A ). Energy consumption was estimated as E = AIW , where A = 20 mV is the drain voltage, I = 5.52 pA is the current flowing across the device, and W = 111.2 ms is the width of the programming pulse. The minimum E for the 10-ONW artificial synapses with nanogap (fig. S12D) was ~12.3 fJ per spike, so the average energy consumption was ~1.23 fJ per spike for each NW. This value is orders of magnitude lower than those of currently available synapse-inspired electronic devices ( Fig. 4B ) and rivals that of biological synapses. This result shows that these ONW STs have the potential to mimic the low energy consumption of a biological synapse in a human brain, in which each spike has an average frequency of ~10 Hz with an amplitude of ~100 mV, and each synaptic event has a duration of ~100 ms, and consumes energy on a scale of ~10 fJ ( 33 ). The low energy consumption of the devices is a benefit of (i) the large surface-to-volume ratio of ONWs that improves the contact area between ONWs and surrounding ion gel and (ii) the short channel length that significantly shortens the path along which energy can dissipate (fig. S13 and related discussion) ( 34 ). The ion gel is ionically conductive and electrically insulating, with leakage current orders of magnitude smaller than that of the semiconducting path in similar transistor geometries ( 19 ), so energy dissipation through the ionic gating is negligible ( 19 , 35 ). Spike rate was correlated with EPSC in the short-channel ST (fig. S14). EPSC could also be increased by applying large-amplitude presynaptic spikes (fig. S15) or longer-time stimulation (fig. S16), but this approach requires increase in energy consumption. Narrow pulse width that is in the reasonable range of dopaminergic neurons in the central brain also triggered EPSC in ONW STs (fig. S17). Furthermore, large-scale 144 STs were fabricated on a 4-inch wafer using the highly aligned ONWs ( Fig. 4C ). To study the transparency of ONWs, we compared the transmittance of bare polyethylene terephthalate (PET) (red) and PET covered with 50-μm–pitched ONWs (blue). The 50-μm–pitched ONW coverage reduced transparency by <1% ( Fig. 4D ). The photograph (inset of Fig. 4D ) showed no obvious reduction in the transparency of PET by the ONW coverage. Our e-NW printing technique that precisely aligns ONWs on an extended area ensures individually addressable synaptic device arrays and low-cost rapid fabrication of large-area electronics. This type of device has a transistor structure, and transfer and output characteristics can be recorded (fig. S19). Fig. 4 Special features: Low energy consumption, scalability and transparency. ( A ) EPSC triggered by an even lower-intensity negative presynaptic spike and short-term synaptic enhancement by two negative pulses. ( B ) Energy consumption per synaptic event of current available synaptic devices. NG, nanogap; PCM, phase change memory; RRAM, resistive switching random access memory. ( C ) Array of 144 ONW STs fabricated on a 4-inch silicon wafer. Inset: Scanning electron microscopy (SEM) image of a typical ONW with a diameter of 200 nm. ( D ) Ultraviolet–visible light spectroscopy as a measure of the transparency of bare and ONW array–loaded PET sheets. Inset: Photograph of a bare PET sheet and a 50-μm–pitched ONW array–loaded PET sheet."
} | 5,161 |
35431782 | PMC9012531 | pmc | 1,997 | {
"abstract": "Frequency-modulated continuous wave radar sensors play an essential role for assisted and autonomous driving as they are robust under all weather and light conditions. However, the rising number of transmitters and receivers for obtaining a higher angular resolution increases the cost for digital signal processing. One promising approach for energy-efficient signal processing is the usage of brain-inspired spiking neural networks (SNNs) implemented on neuromorphic hardware. In this article we perform a step-by-step analysis of automotive radar processing and argue how spiking neural networks could replace or complement the conventional processing. We provide SNN examples for two processing steps and evaluate their accuracy and computational efficiency. For radar target detection, an SNN with temporal coding is competitive to the conventional approach at a low compute overhead. Instead, our SNN for target classification achieves an accuracy close to a reference artificial neural network while requiring 200 times less operations. Finally, we discuss the specific requirements and challenges for SNN-based radar processing on neuromorphic hardware. This study proves the general applicability of SNNs for automotive radar processing and sustains the prospect of energy-efficient realizations in automated vehicles.",
"introduction": "1. Introduction Automated driving is currently a very appealing area of research continuously drawing attention from academic and industrial research groups alike. One key aspect of this development is the success of modern machine learning approaches over the past decade, particularly deep learning by achieving remarkable results on several tasks necessary for fully automated driving, such as traffic sign recognition (Ciresan et al., 2012 ), semantic segmentation (Badrinarayanan et al., 2015 ), 2D and 3D object detection (Zhou et al., 2019 ; Yin et al., 2020 ), and behavior prediction of other traffic participants (Deo and Trivedi, 2018 ). Therefore, the use of such powerful learning approaches in automated vehicle functions and components is likely to increase in the near future. On the other hand, automated vehicle prototypes are typically equipped with a rich setup of various sensor units (Aeberhard et al., 2015 , see also Figure 1A ) to ensure a sufficient coverage of the vehicle's surroundings as well as safety through sensor redundancy. This combination of increasing in-vehicle deployment of modern and power-hungry machine learning approaches; rich and redundant sensor setups; and limited on-board energy resources poses significant challenges on the realization of automated vehicles: Already today, a significant amount of energy in automated vehicle prototypes is dedicated to computing (Gawron et al., 2018 , see also Figure 1B ). Furthermore, in electric vehicles high processing demands can significantly reduce the travel range. While the energy per operation in CPUs and GPUs decreases for smaller semiconductor manufacturing processes, researchers see an asymptotic efficiency wall that is slowly approached in the next years (Marr et al., 2013 ): Therefore, alternative approaches regarding hardware and algorithms are demanded that fulfill both the efficiency and safety requirements for autonomous vehicles. Figure 1 (A) Exemplary sensor setup of an automated vehicle prototype. Image source: BMW. (B) Sources of added energy consumption on a medium automated vehicle system on an electric vehicle prototype. Reprinted with permission from Gawron et al. ( 2018 ) Copyright 2018 American Chemical Society. The neuromorphic computing field (Roy et al., 2019 ) presents an attractive alternative to overcome the previously described challenges. It takes inspiration from the brain by means of a highly-parallel and local processing of information in neural networks, where the memory—the synaptic weights—is physically close to the computing units (neurons). Spiking neural networks (SNNs) employ event-based communication of information, which is fast, efficient and sparse, as information flows when something significant changes or happens. In turn, neuromorphic engineering (Mead, 1990 ; Indiveri et al., 2011 ) integrates neuro-inspired building blocks into electronic circuits for an energy-efficient sensing and information processing suitable for low-power edge applications or large-scale brain simulation. There exist several large-scale neuromorphic hardware systems for SNNs using either purely digital (Merolla et al., 2014 ; Davies et al., 2018 ), multi-processor based (Furber et al., 2014 ) or mixed-signal approaches (Qiao et al., 2015 ; Wunderlich et al., 2019 ) (see Furber, 2016 ; Thakur et al., 2018 for reviews). This is complemented with a new generation of sensors, such as dynamic vision sensors (Lichtsteiner et al., 2008 ; Brandli et al., 2014 ) or dynamic audio sensors (Liu et al., 2014 ), which enable a neuro-inspired pre-processing to directly output events, allowing a seamless integration to neuromorphic compute platforms. Still, those sensors and hardware platforms are mainly used in academic research and are just gradually making their way to commercial products, particularly in the automotive context. In this article, as one step toward energy-efficient neuro-inspired processing for automated driving, we investigate the use of spiking neural networks for automotive radar signal processing . Automotive radars complement LIDAR sensors and cameras for the perception of the street scene and other road users. The used frequency modulated continuous wave (FMCW) radar sensors operate in the 77 GHz band and provide accurate range and relative velocity measurements for distances up to 250 m. In contrast to LIDAR and camera, automotive radar works reliably under all weather conditions and in scenarios with poor lighting, and it also achieves fast reaction times for automatic emergency breaking systems (Patole et al., 2017 ). However, traditional radars lack fine angular resolution to recognize and separate close targets in complex automotive scenarios, and to fully exploit their capabilities in the new artificial intelligent (AI) era. Recent research efforts (Khalid et al., 2018 ; Arkind et al., 2020 ; Rao et al., 2020 ) are tackling this problem by significantly increasing the number of transmit and receive antennas in a multiple input multiple output (MIMO) configuration, which enables a very high angular resolution (down to 1°). This new imaging radar trend has the potential to address the perception challenges in traditional automotive radar sensors, extend the detection to occluded situations in which a pedestrian is not yet exposed to the visual sensors, and provide an accurate radar-based classification of targets in all scenarios, which are all key aspects to enable fully automated driving. Motivated by the successful application of SNNs for a wide range of signal processing and pattern recognition tasks (Zhou et al., 2020 ; Davies et al., 2021 ; Göltz et al., 2021 ; Yin et al., 2021 ), we want to explore whether the signal processing steps of automotive radars can be implemented with SNNs and how well those SNNs perform compared to conventional algorithms. To this end, we first collect and discuss SNN concepts for all steps of the radar processing chain. Next, in order to provide concrete examples, we implement and evaluate SNNs for two processing steps in software. Furthermore, as we plan a future implementation on digital neuromorphic hardware, such as Loihi (Davies et al., 2018 ) or SpiNNaker2 (Mayr et al., 2019 ), we derive the specific requirements and challenges of neuromorphic radar processing. Our main contributions in this article are:\n We perform a comprehensive analysis of the state-of-the-art digital signal processing (DSP) steps for automotive radars and discuss SNN-based approaches for all stages of the processing chain. For the radar target detection step, we implement SNNs for two variants of the constant false alarm rate (CFAR) algorithm and compare their object detection performance and computational cost to classic approaches. For the first time, we apply an SNN to automotive radar object classification achieving an accuracy close to a reference artificial neural network (ANN) at significantly reduced computational cost. We derive the requirements for realizing SNN-based radar processing in neuromorphic hardware systems and discuss the encountered challenges. The remainder of this article is organized as follows: Section 2 describes the operating principle of automotive radars and the digital signal processing chain. It further introduces spiking neural networks and the CARRADA automotive radar dataset used in this article. Section 3 presents a detailed assessment of SNN concepts with the potential to enhance or extend the previously described DSP chain. Section 4 implements and evaluates spiking neural networks for two radar processing steps. Finally, Section 5 discusses the challenges and future outlook in this direction.",
"discussion": "5. Discussion 5.1. Summary, Related Work and Limitations In this article, we reviewed the state-of-the-art digital signal processing steps for automotive radars and discussed for each step various SNN approaches as replacements (Section 3). To the best of our knowledge, such comprehensive analysis of concepts for radar processing with SNNs has not been done before. Yet, we consider this collection of approaches preliminary, and we are sure that more and enhanced approaches will be adopted or developed in the future. Furthermore, for two processing steps we have provided concrete SNN examples and compared them to classical approaches: For the CFAR object detection we developed two temporally coded SNNs and analyzed their accuracy depending on time steps. Starting from 100 time steps, the spiking version is competitive with the reference approach. For object classification, we trained a deep recurrent SNN with BPTT and surrogate gradients on ROI sequences of range-Doppler maps from the CARRADA dataset. The accuracy of the SNN with real-valued inputs of 92.6% is close to the 94.7% achieved by a reference ANN while requiring only 18% of the operations. Instead, the pure SNN with spike input achieves 90.0% with less than 0.5% of the operation of the ANN. Further improvement is expected by increasing the size of the dataset and performing a systematic hyperparameter search. Regarding related work in the context of FMCW radar, so far, SNNs have only been used for gesture recognition, cf. Section 3.4. Very recently, Stuijt et al. ( 2021 ) have demonstrated radar gesture recognition using an ultra-low-power SNN chip and a 8 GHz FMCW radar. They turn the micro-Doppler map into a small binary image and classify it with a rate-based feed-forward SNN on the chip. In López-Randulfe et al. ( 2022 ), the time-coded spiking Fourier transform introduced in Section 3.1.2 was implemented and validated on Loihi to compute the range and Doppler-FFT on recorded radar data. Compared to dedicated hardware FFT accelerators, the neuromorphic solution lags behind by one to three orders of magnitudes in terms of energy and latency. Brown et al. ( 2021 ) have developed an SNN hardware accelerator for compressed sensing with pulse-Doppler radars. A spiking locally competitive algorithm (LCA) solves the sparse optimization to achieve highly accurate and efficient target and velocity estimation. This compressed sensing approach is not directly applicable to the FMCW automotive radar processing chain discussed in this article. Further, Barnell et al. ( 2020 ) use spiking DNNs on Loihi for classification of synthetic aperture radar images. While this demonstrates the efficiency of neuromorphic hardware for image classification, new network models will have to be developed for automotive FMCW radar data. The SNN concepts presented in this work apply to single steps of the radar processing chain. How to combine several SNNs or how to build a radar processing chain completely with spiking neurons was not the objective of this paper and remains an open research subject. 5.2. Toward Neuromorphic Radar Sensors Whether or not spiking neural networks can outperform conventional radar processing depends on how efficiently they can be realized in neuromorphic hardware. In the following, we summarize the requirements of a neuromorphic application-specific integrated circuit (ASIC) to process the radar data in real time. For this, we assume that a neuromorphic processor replaces or complements a DSP (cf. Figure 2A ) and receives raw ADC data or preprocessed data that has to be converted to spikes on the chip. Our analysis includes the required memory for buffering input data, the required input bandwidth, the number of neurons and synapses, and the processing speed of those neuromorphic components. As a radar sensor setup we take the one from the CARRADA dataset with 2 transmitters and 4 receivers (cf. Table 3 ), yet we note that the requirements for high-resolution radars will strongly increase. Reviewing the radar processing steps from Figure 3 , the hardware requirements vary significantly for each processing step, e.g., the amount of input data per frame that needs to be processed varies a lot, as shown in Table 3 . Especially processing the full raw data or high-resolution range-angle maps requires more than 100 kB of memory for buffering the input. This amount does not pose a problem for typical embedded micro-processors, yet it might become challenging for high-resolution radars with more than 10 times as much data or when fed into edge neuromorphic processors. Similarly, for the communication between a radar sensor and neuromophic hardware at least a bandwidth of 10–100 MBit/s is needed. Table 3 Requirements for a neuromorphic ASIC for radar processing. \n Frame parameter \n \n Value \n \n N \n TX \n 2 \n N \n RX \n 4 \n N \n chirps \n 64 N samples (complex) 256 Frames per second 10 \n Input type \n \n Memory [kB] \n \n Bandwidth [MBit/s] \n Raw data cube (256x64x4 á 2x16b) 197 15.7 Range-Doppler map (256x64 á 16b) 32.8 2.62 Range-angle map (256x256 á 16b) 131 10.5 \n Processing step \n \n Inputs \n \n Neurons \n \n Synapses \n \n Time steps/ frame \n \n Repetitions/frame \n Range-FFT (López-Randulfe et al., 2022 ) 512 4608 37 888 550 256 OS-CFAR on RD map (this work) 16 384 16 384 2 899 968 100 1 Object classification (this work, pure SNN) 520 3747 311 328 1 1-20 At the bottom of Table 3 , we review SNN requirements for some of the radar processing steps: The range S-FT with time coding from López-Randulfe et al. ( 2022 ) can be realized with sparse connectivity and one spike per synaptic connection for 550 time steps. While the S-FT network itself is rather small, the challenge is to run the model 256 times (64 chirps × 4 receivers) per frame on a neuromorphic processor (e.g., within 20 ms assuming that 20% of the 100 ms frame time are budgeted for the range-FFT). This seems possible, according to the results obtained in López-Randulfe et al. ( 2022 ), where a 1024-point spiking FFT can be calculated every 105 μs on the Loihi neuromorphic chip. For the spiking OS-CFAR from Section 4.1, a network of 16 k input and output neurons with nearly 3 million synapses is required to process an entire range-Doppler map. Compared to the range-FFT, this SNN is run only once per frame and thus has lower neuromorphic compute demands. Finally, the SNN-based radar object classification (Section 4.2) has the least requirements for implementation on neuromorphic hardware as the network is smaller and there is only one time step per frame (cf. Supplementary Section 2.3 ). Note, however, that the network needs to be newly instantiated for each detected object and we expect in the order of up to 20 radar objects in simple street scenes. Looking at the neuromorphic requirements for the different automotive radar processing steps, we expect that SNN-based object classification has the highest potential for energy-efficient realization in neuromorphic hardware. SNN-based object tracking should also be evaluated further in the future. For the earlier processing steps like the FT and CFAR object detection, further work shall determine if neuromorphic hardware tailored at these operations can implement these operations more efficiently than digital signal processors and close the current gap in terms of energy and time performance (López-Randulfe et al., 2022 ). At the system level, one could alternatively combine a DSP with a neuromorphic processor to achieve maximum efficiency. When split onto different chips, the data bandwidth requirements from Table 3 need to be fulfilled. An even more radical approach for radar processing with neuromorphic hardware is to use analog spiking neurons in hardware with the radar IF signal as input. Resonate-and-fire neurons are the perfect candidates for that, but this might be limited to radar systems that don't need phase information, e.g., using a single transmitter and receiver. Yet, further research may clarify whether a full SNN pipeline on dedicated neuromorphic hardware can outperform classical DSP or hybrid DSP/ANN approaches. 5.3. Toward Neuromorphic Automated Driving As motivated in the introduction, the use of neuromorphic hardware has a high potential to significantly reduce the energy demands for highly-automated driving. Besides radar signals, also camera and LIDAR data need to be processed in order to get a complete understanding of the automotive scene. For image processing there already exist first attempts to solve complex tasks with SNNs, e.g., for object detection (Kim et al., 2020 ) or semantic segmentation (Kim et al., 2021 ). Also recently, Viale et al. ( 2021 ) realized an SNN on Loihi for car detection using a dynamic vision sensor. Using LIDAR data, which is naturally sparse and thus predestined for SNNs, Zhou et al. ( 2020 ) showed a spiking convolutional network for real-time 3D object detection. Shalumov et al. ( 2021 ) use LIDAR data for SNN-based collision avoidance with a control network based on the neural engineering framework. All these examples show that SNN-based sensor processing for autonomous driving is a trending topic. Besides the development of SNNs and their implementation on neuromorphic hardware, also the combined processing, i.e., sensor fusion using SNNs, will become an important topic. When it comes to AI-based autonomous driving, ensuring functional safety of both software and hardware is a critical issue. The principles that are currently developed to support machine learning models (Henriksson et al., 2018 ; Mohseni et al., 2019 ) will also apply to SNNs. Similarly, neuromorphic hardware will have to fulfill the same standards as any automotive electronic system: adhere to temperature ranges, be resistant to vibrations, be deterministic and redundant, or contain self-monitoring. For that reason, only digital neuromorphic systems are candidates for integration in cars, while the use of analog or mixed-signal neuromorphic hardware seems out of scope at the moment due to their intrinsic variability. Hence, we suggest to focus on advanced digital systems such as SpiNNaker2 (Yan et al., 2021 ) or Loihi2 (Orchard et al., 2021 ) to further explore neuromorphic hardware for automotive radar processing and automated driving in general."
} | 4,848 |
36340378 | PMC9632346 | pmc | 1,998 | {
"abstract": "The plant microbiome profoundly affects many aspects of host performance; however, the ecological processes by which plant hosts govern microbiome assembly, function, and dispersal remain largely unknown. Here, we investigated the bacterial and fungal communities in multiple compartment niches (bulk soil, rhizosphere soil, root endosphere, phylloplane, and leaf endosphere) of Casuarina equisetifolia L. at three developmental stages in Hainan Province, China. We found that microbiome assemblages along the soil–plant continuum were shaped by the compartment niches. Bacterial diversity and richness decreased from the soils to roots to leaves, with the highest network complexity found in the roots and the lowest found in the phylloplane. However, fungal diversity gradually increased from the soils to roots to phyllosphere, whereas fungal richness decreased from the soils to roots but increased from the roots to phyllosphere; the greatest network complexity was found in bulk soils and the lowest was found in the roots. Different biomarker taxa occurred in the different ecological niches. Bacterial and fungal communities exhibited distinct ecological functions; the former played important roles in maintaining plant growth and providing nutrients, whereas the latter predominantly decomposed organic matter. The bacterial community of C. equisetifolia mostly originated from bulk soil, whereas the fungal community was mainly derived from rhizosphere soil and air. Leaf endophytes were positively correlated with organic carbon, and root and soil microorganisms were positively correlated with total nitrogen, total phosphorus, and total potassium. Our findings provide empirical evidence for plant–microbiome interactions and contribute to future research on non-crop management and the manipulation of non-crop microbiomes.",
"conclusion": "Conclusions In this study, we provide comprehensive and empirical evidence for the relative contribution of compartment niches to microbiome assembly in C . equisetifolia . Our results suggest that microbiome assemblages along the soil–plant continuum were primarily shaped by the compartment niches rather than the developmental stage. Moreover, bacterial diversity and richness decreased from the soils to roots to leaves, with the highest network complexity found in the roots and the lowest in the phylloplane. However, fungal diversity gradually increased from the soils to roots to phyllosphere, whereas richness decreased from the soils to roots and increased from the roots to phyllosphere; the greatest network complexity was found in bulk soils, and the lowest was in the roots. Furthermore, different biomarker taxa were present in different ecological niches, and significant differences in ecological function were found between bacterial and fungal communities, with bacteria playing an important role in maintaining plant growth and providing nutrients, whereas fungi played a dominant role in the decomposition of organic matter. In addition, the bacterial community of C . equisetifolia was mainly derived from bulk soil, whereas the fungal community primarily originated from the rhizosphere soil and air, with leaf microorganisms positively correlated with organic carbon, and root and soil microorganisms positively correlated with total nitrogen, total phosphorus, and total potassium. These results suggest strong selective and regulatory effects of plant hosts on the composition and potential function of plant microbial communities. The results of this study have implications for future non-crop management by providing baseline data to inform translational research into harnessing the plant microbiome.",
"introduction": "Introduction Microorganisms are ubiquitous in plant tissues such as roots, stems, leaves, flowers, and fruits ( Lindow and Brandl, 2003 ; Hacquard et al., 2015 ). Plants and their inhabiting microbiome together constitute a “holobiont,” with the plant microbiome acting as a secondary genome and a link to host fitness ( Vandenkoornhuyse et al., 2015 ). Therefore, comprehending the mechanisms underlying plant microbiome assembly, function, and dispersal is important for our fundamental understanding of the development of microbiome-based strategies required to maximize plant survival and increase their tolerance to low soil fertility, alkalinity, and salinity. Assembly of the plant microbiome begins shortly after sowing and develops with plant growth under the influence of stochastic (e.g., random dispersal) and deterministic (e.g., selection mediated by biotic and abiotic factors) processes ( Bulgarelli et al., 2013 ; Müller et al., 2016 ). In addition to vertical transmission ( Müller et al., 2016 ; Abdelfattah et al., 2021 ), microorganisms from the soil and air can migrate to and colonize different compartments of plants ( Bulgarelli et al., 2013 ; Compant et al., 2021 ). A study has highlighted the important contribution of plant developmental stage to plant microbiome assembly such as the physiological needs of plants and the composition of plant secretions ( Zhang et al., 2018 ). However, the mechanisms by which the host and environment shape microbiome assemblages and symbiotic patterns across the soil, root endosphere, and phyllosphere remain largely unknown. Our understanding of the microbiome assembly mechanism in plants is still in its infancy, with several key questions yet to be answered. The root microbiome is also assembled from soil microorganisms, and based on the composition of microbial communities in the root endosphere and rhizosphere, this assembly may be achieved in two steps: (1) recruitment of rhizosphere microorganisms by plants; (2) entry of microorganisms into the plant ( Bulgarelli et al., 2013 ). Furthermore, the phylloplane is an important interface among plants, microorganisms, and the environment, and microorganisms from the surrounding environment can colonize the leaf endosphere ( Lindow and Brandl, 2003 ). Although the aforementioned hypotheses are reasonable, the microbial composition among the bulk soil, rhizosphere, root endosphere, phylloplane, and leaf endosphere, as well as the relationship between the rhizosphere and phyllosphere, remain unclear. In addition, the roles of microorganisms in different ecological niches have yet to be clarified. In the late 1950s, Casuarina equisetifolia L. was first introduced to China in Guangdong Province ( Li et al., 2015 ), where it serves as an important shelter tree species on the southeastern coast of China ( Liu et al., 2020 ). C. equisetifolia is characterized by wind speed reduction and sand fixation, as well as salt and alkali tolerance. Moreover, it is of great importance for the restoration of ecological functions and protection against natural disasters in coastal areas, and it contains abundant endophytic microorganisms that perform important functions related to saline–alkali tolerance, low soil fertility tolerance, and allelopathy ( Lin et al., 2008 ). Previous studies have shown that vertical transmission of endophytes in C. equisetifolia affects the dispersal of endophytic fungi from seeds to the leaf endosphere and the transmission of endophytic bacteria from seeds to the root endosphere (Lin, unpublished). However, there has been little investigation of the impact of the soil–plant continuum on the plant microbiome, such as epiphytes and endophytes. Here, we use C. equisetifolia as a model system to fill these key knowledge gaps. Specifically, this study has two objectives: (1) evaluation of the microbiome assembly at different developmental stages (young, mature, and aged forests) and in different compartment niches (bulk soil, rhizosphere soil, root endosphere, phylloplane, and leaf endosphere) of C. equisetifolia ; (2) identification of the potential sources, dominant taxa, and ecological functions of plant microbial communities, as well as the effects of environmental factors on microorganisms. To achieve these objectives, we use high-throughput sequencing technology and chemical analysis to evaluate samples associated with C. equisetifolia at different developmental stages. The following hypotheses are proposed: (1) the microbial community migrates from the soil (a reservoir of microorganisms) to plant roots and then to plant leaves, with a gradual decrease in microbial diversity; (2) bacteria and fungi perform distinct functions in different compartment niches.",
"discussion": "Discussion While the diversity of plant-associated microbiomes is increasingly recognized, we still need to further clarify the relationship between microbiomes and their host plant. In this study, we investigated the microbial communities along the soil–plant continuum of C . equisetifolia at different forest ages. We found that the structure and diversity of microbial communities were shaped by compartment niches. Our results further showed that bacterial diversity and richness decreased from the soils to roots to leaves, with the highest network complexity found in the roots and the lowest in the phylloplane. However, fungal diversity gradually increased from the soils to roots to phyllosphere, whereas richness decreased from the soils to roots but increased from the roots to phyllosphere; the greatest network complexity was found in bulk soils and the lowest was found in the roots. These findings provide comprehensive and empirical evidence for theories regarding non-crop host selection and niche occupation involved in C . equisetifolia microbiome assembly. In addition, we identified the dominant taxa, ecological functions, and the potential sources as well as the effects of environmental factors on non-crop microbiomes. Our results provide key information for non-crop microbiome manipulation. Assembly and diversity of microbial communities in compartment niches In our study, we found clear and separate clustering among different compartment niches rather than among developmental stages ( \n Figures 1A \n , \n 2A \n ). Thus, we inferred that the C . equisetifolia microbiome assemblage was predominantly shaped by the compartment niches, as suggested by previous studies ( Beckers et al., 2017 ; Cregger et al., 2018 ). We also found that the bacterial Shannon diversity decreased from the soil to roots and then to the leaves, whereas the fungal diversity showed the opposite trend ( \n Figures 3A, B \n ), which might be attributed to the external environment, host plant, and microbial properties ( Edwards et al., 2015 ; Hamonts et al., 2018 ). These results suggest significant differences in the assembly of bacteria and fungi across compartment niches. We further found that the lowest Chao1 richness, Shannon diversity, and network complexity of bacteria occurred in the phyllosphere ( \n Figures 3A, C \n ). A leaf environment with low bacterial diversity is considered unfavorable for bacterial colonization because the leaf surface is exposed to rapidly changing temperature and humidity, as well as the alternating presence and absence of rainwater, which can reduce the bacterial diversity ( Guttman et al., 2014 ). Moreover, leaves provide limited nutrients for bacteria, and you could find less competition. As a result, most bacteria that migrate to leaves may find themselves in a nutrient-poor environment that limits their growth and metabolism ( Kembel et al., 2014 ; Hacquard et al., 2015 ). Interestingly, we found that leaves harbored a diverse range of fungal taxa ( \n Figure 3B \n ), probably because fungi are typically more tolerant to drought and can proliferate in harsh environments, enabling better survival of the fungal community in above-ground plant tissues ( Whipps et al., 2008 ; Rodriguez et al., 2009 ). Hainan Province has a large diurnal temperature range and distinct seasons with abundant and deficient rainfall, with drought typically occurring in the season with low rainfall ( Cai et al., 2010 ). Moreover, C . equisetifolia bears coriaceous scale leaves, and photosynthesis mainly occurs in needle-like branchlets, which leads to low water retention. These factors may contribute to the harsh microenvironment of C . equisetifolia leaves. We also found that bacterial diversity in the rhizosphere soil was consistently higher than that in the root endosphere ( \n Figure 3A \n ). This is not surprising because the root exudates produced by C . equisetifolia in the rhizosphere (e.g., 2,4-di-tert-butylphenol, methyl stearate, and arginine) enhance the bacterial chemoattraction and colonization of the rhizosphere soil and rhizoplane. These factors lead to the formation of a unique, highly abundant, and diverse microbial community in the rhizosphere (Lin, unpublished data). After colonization at the rhizoplane, soil bacteria compatible with the plant lifestyle will reach the xylem vessels by active or passive transport through the endodermis and pericycle. The rhizoplane acts as a selective barrier responsible for a considerable loss of microbial diversity ( Hardoim et al., 2008 ). Via a systematic evaluation of the relative contribution of various niches to the microbial community, our work enhances the understanding of plant microbiome assembly. Dominant taxa and biological functions of microbial communities in compartment niches Dominant taxa can be key microbes with important ecological implications in microbiome assembly and ecosystem functioning ( Banerjee et al., 2018 ; Delgado-Baquerizo et al., 2018 ). Our results indicated that the dominant taxa in the root endosphere were Pseudomonas (Gammaproteobacteria) and Bacillus (Bacilli) ( \n Figure 4 \n ). Members of Gammaproteobacteria can colonize the rhizosphere and a wide range of niches, thereby playing a key role in regulating host fitness, pathogen inhibition, and plant tolerance ( Mendes et al., 2011 ; Álvarez-Pérez et al., 2017 ). We further found that Pseudomonas was more abundant in the roots than in the soils ( \n Figure 4 \n ) because, although found in the rhizosphere where it greatly promotes plant growth, Pseudomonas has poor environmental adaptability and low competitiveness ( Berg, 2009 ). Therefore, we speculated that Pseudomonas could be more adapted to the root endosphere than to the soil, which explains their dominance in this environment. Pseudomonas is also growth-promoting bacteria capable of nitrogen-fixing and phosphorus dissolution ( Berg, 2009 ). Bacilli are frequently reported as antagonistic bacteria against soil-borne diseases. Moreover, members of Bacillus are repeatedly shown to be growth-promotors living in the rhizosphere and capable of phosphorus solubilization and nitrogen-fixing ( Gong et al., 2014 ). Our results indicated that the nitrogen content was higher than the potassium content in the roots ( \n Table 1 \n ), which might be related to nitrogen-fixing by Pseudomonas and Bacillus ; however, the fact that the phosphorus content was relatively lower requires further study. Our results also revealed that Dothideomycetes were present in the phyllosphere ( \n Figure 4 \n ), which agrees with the findings of ( Adams et al., 2013 ) and suggests that Dothideomycetes may be airborne. In addition, many members of Dothideomycetes are saprophytic fungi associated with litter decomposition and nutrient cycling ( Adams et al., 2013 ; Hyde et al., 2013 ). In the present study, most fungi were saprophytes enriched in the leaves ( \n Figure 6 \n ), suggesting that the occupation of aged host plants by saprophytic fungi resulted from reduced immunity and the increasingly important ecological functions of saprophytic fungi as decomposers. As a result, we speculated that the adaptability of C . equisetifolia was improved by the growth and proliferation of probiotic microorganisms with various functions and living in different niches, which also supported the low soil fertility tolerance of this plant. Together, these results suggest that plants can recruit microbial taxa with specific functions and adaptability in various compartment niches ( Foster et al., 2017 ; Cordovez et al., 2019 ). Identification of these dominant taxa provides essential information for the development of strategies to manipulate the microbial community in C . equisetifolia . Sources of microbial communities in different ecological niches Identifying the potential sources and enrichment processes of microbial communities in C . equisetifolia is essential for understanding the interactions among plants, soil, and microorganisms. Although previous studies have reported that the above-ground and below-ground compartments of plants share a large proportion of microbial taxa ( Bai et al., 2015 ), little is known about enrichment of the microbiome in C . equisetifolia . Our results showed that the bacterial community in rhizosphere soil was mainly derived from bulk soil (unknown source values < 17%) and was sequentially filtered by the plant niches ( \n Figures 7A, B \n ). This was expected because the rhizoplane acts as a selective barrier. Meanwhile, limited bacterial species can colonize the root endosphere, such as those that express chemotaxis-related genes, present the formation of flagella and pellets, produce cell wall-degrading enzymes, and have complex interactions with the host plant immune system. Thus, the diversity of endophytic bacteria was lower than that of soil bacteria ( Hardoim et al., 2008 ; Bulgarelli et al., 2012 ). Moreover, although a small proportion of fungal OTUs were shared between the phylloplane and rhizosphere soil, more than 96% of fungi were of unknown sources ( \n Figures7C, D \n ). Therefore, we inferred that phylloplane fungi were mainly derived from the surrounding environment, suggesting air, dust, and rainwater as the primary sources of the phylloplane fungal community. We also found that leaf endosphere fungi originated from the phylloplane and soil ( \n Figures 7C, D \n ), and the roots may serve as an important transition boundary ( Hacquard et al., 2015 ; Vandenkoornhuyse et al., 2015 ), allowing rhizosphere microbes to enter plant tissues and migrate to the above-ground plant compartments. In the phyllosphere, the unknown source values of bacteria were lower than those of fungi ( \n Figure 7 \n ), indicating that a greater proportion of bacterial communities in the above-ground plant tissues was derived from rhizosphere soil. In addition, the unknown source values of leaf endophytes were lower than those of leaf epiphytes ( \n Figure 7 \n ), further highlighting the selection of endophytes by hosts. These findings have identified potential sources and driving forces of microbial communities in the phyllosphere. Furthermore, they further confirm the phylloplane and rhizoplane as important interfaces among hosts, microorganisms, and the environment ( Lindow and Brandl, 2003 ; Vorholt, 2012 ; Vacher et al., 2016 ; Remus-Emsermann and Schlechter, 2018 ). In summary, our research showed that plants can recruit microbial taxa with specific function and niche adaptability, which provides a theoretical basis for the analysis of plant niches and transmission routes of microorganisms. However, the molecular mechanisms by which hosts regulate plant–microbiome interactions and microbial community dispersal are not fully understood; thus, further research is required."
} | 4,828 |
23251679 | PMC3522635 | pmc | 1,999 | {
"abstract": "Non-Hebbian learning is often encountered in different bio-organisms. In these processes, the strength of a synapse connecting two neurons is controlled not only by the signals exchanged between the neurons, but also by an additional factor external to the synaptic structure. Here we show the implementation of non-Hebbian learning in a single solid-state resistive memory device. The output of our device is controlled not only by the applied voltages, but also by the illumination conditions under which it operates. We demonstrate that our metal/oxide/semiconductor device learns more efficiently at higher applied voltages but also when light, an external parameter, is present during the information writing steps. Conversely, memory erasing is more efficiently at higher applied voltages and in the dark. Translating neuronal activity into simple solid-state devices could provide a deeper understanding of complex brain processes and give insight into non-binary computing possibilities.",
"conclusion": "Conclusions We have analyzed the behavior of light-controlled resistive memory devices [22] as synaptic-mimics for inhibitory learning processes. We have demonstrated that the learning processes are not only controlled by the voltage pulses applied to the device (that are equivalent to the neuronal action potentials) but also by an external third factor, in our case the presence of light. This is, to the best of our knowledge, the first implementation of a non-Hebbian learning process in a single solid-state device. We believe that light-controlled non-volatile resistive memory devices offer a new perspective for the investigation of neuromodulated learning processes. More complex structures, bringing together a number of devices that would act as neuronal networks, could be the next step into implementation of biological processes into silicon-compatible architectures.",
"introduction": "Introduction Ramón y Cajal postulated that the nervous system is formed of individual fundamental units called neurons linked to each other by small contacts [1] , that were later called synapses [2] . In spite of the astonishing continuous progress made in neuroscience for more than a century, there is still a lack of understanding of many neuronal mechanisms, mainly due to their complexity and versatility. For example, for the specific case of neuronal processes, Hebb proposed that its basis stands on the synaptic strength (weight) increase caused by the simultaneous activity of both presynaptic and postsynaptic neurons [3] . The learning process proposed by Hebb is inherently unstable because of the so-called autocorrelation aspect. In simple terms, autocorrelation represents the trend of the synaptic weight for self-amplification, that is, the more a synapse drives a postsynaptic cell the more the synaptic weight will grow. Likewise, once depressed, the synaptic weight decreases invariably to zero. One realistic way of stabilizing the synaptic weight is to introduce an extra, third factor capable of modulating the learning process so as to control the self-amplification. Such third factors are typically neuromodulators and are usually inputs external to the analyzed synaptic system. An example of an external neuromodulator is dopamine in certain brain areas manifested by mechanisms of learning and forgetting processes, for example in classical or operant conditioning [4] – [7] . In addition to the experimental influence of diverse neuromodulators, mathematical models in neuronal networks have demonstrated their role in the plasticity of memory processes [8] – [9] . Resistive memories are solid-state devices in which a resistance state can be set by an appropriate sequence of voltage pulses of well-determined durations [10] – [14] . This behavior resembles some key aspects of synapses in the brain, since the voltage pulses act very much like the neuronal action potentials (or spikes) in Hebbian processes [15] , [16] . In biological synapses the learning process is strengthened by its repetition, and a similar behavior has also been observed in solid-state resistive memory devices mimicking Hebbian learning [17] – [21] . In this paper we move a step further and present the first experimental implementation of a three-factor non-Hebbian learning in a single memory device. In this specific example, we employ a light-controlled resistive memory device [22] . In our system the electrical resistance (which is equivalent to the synaptic weight) is modulated not only by the voltage (equivalent to neuronal action potentials) as in conventional resistive memories, but also by a third, external factor, which in our case is the presence of light. Moving closer to translating complex neuronal operations into simple solid-state devices can provide both a deeper understanding of neuromodulated brain processes and give insight into non-binary computing possibilities.",
"discussion": "Results and Discussion In Figure 1a we show the design of a light-controlled resistive memory. It consists of a 20-nm-thick Al 2 O 3 film deposited on a p-doped Si substrate covered with a thin layer (1.9 nm) of native SiO 2 . The Si substrate works as bottom electrode in a metal-oxide-semiconductor (MIS) configuration while Pd electrodes, patterned by photolithography, act as top electrodes. Light can reach the optically active silicon substrate through the spaces uncovered by palladium and after crossing the transparent aluminum oxide and silicon oxide layers. The behavior of the light-controlled memory devices is based on the photogeneration of charge carriers in silicon under illumination. With suitable applied voltage-pulses, the photogenerated electrons from Si are injected in the Al 2 O 3 layer. A fraction of these electrons is then trapped in the Al 2 O 3 layer, changing quasi-permanently its resistance state [22] . The electrical characterization of the devices was performed by means of remnant resistance hysteresis switching loops (HSL; Figure 1b ). Each step on the HSL represents the remnant resistance (Rrem) measured at 7 V either in the dark (curve labeled ‘dark’ in Figure 1b ) or under illumination (the curve labeled ‘light’ in Figure 1b ), after sweeping the voltage between −Voperate and +Voperate (each pulse of the sweep has a length of 100 ms), again either with light or in the dark, respectively. Each Voperate pulse is followed by a waiting time of 100 ms at 0 V to discard capacitive effects before finally reading the device state with the 7 V voltage pulse. In the dark, the absence of photogenerated electrons in Si results in a non-hysteretic remnant resistance HSL curve. Under illumination, the resistance decreases strongly due to the presence of photogenerated charge carriers in the system, and a hysteretic non-volatile memory behavior is obtained. The HSL proves that a non-volatile memory state can be intrinsically defined by a chosen voltage pulse under illumination, although it also depends on any previous memory state. 10.1371/journal.pone.0052042.g001 Figure 1 Device and measurements protocols. (a) Configuration of the light-controlled resistive memory devices. 20 nm of Al 2 O 3 are deposited on p-Si/SiO 2 . The bottom electrode is the Si substrate, whereas the top electrode is Pd patterned by photolithography. Light can reach the optically active Si substrate through the spaces uncovered by Pd and after crossing the transparent Al 2 O 3 and SiO 2 layers. Inside the Si, under illumination, charge carriers are photogenerated. (b) Typical remnant resistance hysteresis switching loops measured at 7 V after applying a voltage pulse between -Voperate and Voperate either in the dark or under illumination, followed by a 100 ms waiting time at 0 V to discharge capacitive effects. (c-e) Steps of voltage-pulses applied to the memory in different learning processes (summarized in Table 1 and 2 ); the sequence is for example repeated 1000 times in the case of the ‘usual activity’ measurements (process A1 in Table 1 ). S 0 is the typical voltage input and is applied in the usual activity or in the reading steps, while S E is the extra-external voltage input, applied in addition to S 0 during the inhibitory learning (S E >0) or the Memory erasing (S E <0 and |S E |>|S 0 |) steps. Processes A2–A6 ( Table 1 ) contain ‘inhibitory learning’ steps (figure d), meaning that positive S voltages are applied a number of times. Processes B1–B4 ( Table 2 ) comprise ‘Memory erasing’ steps (figure e), meaning that negative S voltages are applied a number of times. The operating voltage pulse (S 0 or S = S 0 +S E ) is applied either in the dark or with light, as a third extra-control parameter, followed by a waiting time in short-circuit conditions (at 0 V) and then followed by the measurement of the induced memory state by a reading voltage of S 0 = 7 V under illumination. The step time t 0 is 100 ms in all cases. 10.1371/journal.pone.0052042.t001 Table 1 Summary of the inhibitory learning processes tested. Process ‘Operate’ voltage pulses \n Usual activity, S 0 \n Inhibitory learning,S = S 0 +S E (S E >0) \n Usual activity, S 0 \n Inhibitory learning,S = S 0 +S E (S E >0) \n Usual activity, S 0 \n A1 \n 1000x (S 0 , light) \n A2 \n 100x (S 0 , light) \n 500x (S = 10 V, light) \n 400x (S 0 , light) \n A3 \n 100x (S 0 , light) \n 500x (S = 10 V, dark) \n 400x (S 0 , light) \n A4 \n 100x (S 0 , light) \n 500x (S = 8 V, light) \n 400x (S 0 , light) \n A5 \n 100x (S 0 , light) \n 500x (S = 8 V, dark) \n 400x (S 0 , light) \n A6 \n 100x (S 0 , light) \n 500x (S = 10 V, light) \n 100x (S 0 , light) \n 50x (S = 10 V, light) \n 250x (S 0 , light) \n The outcome of the different processes is displayed in Figures 2 and 3 . For processes A2–A6 the first 100 voltage pulses sets are the same, followed by different inhibitory learning V-sets. S 0 = 7 V in all cases and represents ‘usual activity’. After this brief initial presentation and characterization of basic properties of the light-controlled resistive switching devices, we move now to the testing of different learning processes, thus showing the influence of the external non-Hebbian third factor in the results. In Figures 1c to 1e we present the individual different voltage-pulses steps that constitute the core of the different processes implemented. We have denoted them as ‘Usual activity’ ( Fig. 1c ), ‘Inhibitory learning’ ( Fig. 1d ) and ‘Memory erasing’ ( Fig. 1e ). The measuring sequence for each point of these processes is: I. the application of a 100 ms Operate voltage pulse (of different amplitudes, S 0 or S, in dark or under illumination), II. the application a 100 ms waiting step in short-circuit conditions for discharging any capacitive effects, and finally III. the measurement of the state of the memory system under illumination (please refer to the Experimental Section for details) with a voltage pulse S 0 = 7 V of 100 ms duration. For segment I. of the measurement sequence presented above, the Operate voltage pulse is applied in the dark for processes A3, A5 (only during the Inhibitory learning part, as presented in Table 1 ) and B2, B4 (only during the Memory erasing part, see Table 2 ) and under illumination for all other processes. In all the different protocols tested we keep the same conditions for reading the remnant resistance (Rrem) state of the memory device, namely S 0 = 7 V under illumination, in order to sense the changes created by the different learning activities performed in the ‘Operate’ step. These different core steps are repeated a certain number of times during the learning processes presented in this paper (summarized in Tables 1 and 2 ). Before the beginning of each process (presented in Tables 1 and 2 ) a cleaning protocol is performed with the aim of removing any trapped charges in the Al 2 O 3 layer. This cleaning consists in 400 pulses of −10 V, 100 ms each, in the dark, and we observed that this protocol is equivalent to bringing the memory to a pristine state, as all past events are fully erased. 10.1371/journal.pone.0052042.t002 Table 2 Summary of the different ‘inhibitory learning’ and ‘Memory erasing’ processes tested. Process ‘Operate’ voltage pulses \n Usual activity, S 0 \n Inhibitory learning,S = S 0 +S E (S E >0) Memory erasing,S = S 0 +S E (S E <0) \n Usual activity, S 0 \n B1 \n 100x (S 0 , light) \n 500x (S = 10 V, light) 100x (S = −10 V, light) \n 300x (S 0 , light) \n B2 \n 100x (S 0 , light) \n 500x (S = 10 V, light) 100x (S = −10 V, dark) \n 300x (S 0 , light) \n B3 \n 100x (S 0 , light) \n 500x (S = 10 V, light) 100x (S = −7 V, light) \n 300x (S 0 , light) \n B4 \n 100x (S 0 , light) \n 500x (S = 10 V, light) 100x (S = −7 V, dark) \n 300x (S 0 , light) \n For processes B1 to B4 the first 600 sets of voltage pulses are the same, followed by different memory erasing voltage pulses. S 0 = 7 V in all cases. The outcomes are presented in Figure 4 . 10.1371/journal.pone.0052042.g002 Figure 2 Inhibitory learning of light controlled resistive memory devices. (a) Output of the different processes used for analyzing the efficiency of the different ‘inhibitory learning’ processes. The remnant resistance was measured at S 0 = 7 V with light after the inhibitory learning voltage pulses (S 0 +S E ) were applied either in the dark or with light. The curves are labeled following the protocols described in Table 1 . (b) Efficiency of the different learning processes described in Table 1 . Processes A2, A4 are inhibitory learning with light, whereas processes A3, A5 are learning in the dark. Approximately double efficiency is obtained when light is present during learning as compared to learning in the dark. A1 is the ‘usual activity’ curve, which serves as a base line. For positive applied voltages S>0, the remnant resistance increases (see Figure 1b ), while at negative applied voltages S<0, the remnant resistance decreases. Within this paper we identify the former step as an inhibitory learning process, while we associate the latter step with erasing the learned information. The inhibitory learning in our device correlates with a neuronal plasticity-rule where the synaptic strength is reduced through the learning process. In this context, the opposite process, excitatory learning, would be a learning process involving the increase of synaptic strength (i.e. decrease of resistance). 18 The step we refer to as erasing step resets the synaptic weight to a reproducible initial condition with low resistance. 10.1371/journal.pone.0052042.g003 Figure 3 Output remnant resistance measured in processes A2 and A6, as described in \n Table 1 \n . After a learning process followed by a forgetting time, the system only needs a small reminder to reach back to the ‘well-learned’ state. This behavior is similar to processes encountered in living organisms. In Table 1 we summarize the different ‘inhibitory learning’ processes tested (the outcomes of the processes are presented in Figures 2 and 3 ). Process A1 ( Figure 2a ) is equivalent to the experience achieved during a usual day-to-day activity (the applied voltage pulses are shown in Figure 1c ) as we observe that the system learns slowly over time (the resistance increases monotonically). We can further study the inhibitory learning by applying different sets of pulses, distinct from process A1 (the usual activity). Process A2 departs from A1 after the application of the 100 initial voltage pulses sets. In this particular case we continue by applying 500 voltage pulses of S = 10 V under illumination (in the ‘Operate’ step, see Figure 1d ), followed by 400 steps of S 0 = 7 V that represent again a usual activity protocol ( Figure 1c ). Other protocols follow similar routines although we change the amount and amplitude of the inhibitory learning part of the protocol. 10.1371/journal.pone.0052042.g004 Figure 4 Output of the different processes used to analyze the efficiency of memory erasing through positive reinforcement voltage pulses. The remnant resistance is read at S 0 = 7 V under illumination. Process A1 represents usual activity (see Table 1 ). A2 is a process without any erasing step and information is slowly forgotten over time (starting from step 600). Processes B1, B3 contain negative-voltages Memory erasing steps performed under illumination, while processes B2, B4 contain Memory erasing steps in the dark. Additionally, a more efficient memory erasing is obtained with higher-amplitude negative voltages and in the dark. In Figure 2 we show the behavior of our system after different learning sequences (A1 to A5, Table 1 ). As described above, the process A1 is equivalent to the experience achieved during a usual day-to-day activity, the system learns slowly over time. Distinctly, processes A2–A5 contain steps that are equivalent to applying inhibitory stimuli on a bio-system (S E >0). The learning is faster than in case A1, and this is represented by a more pronounced permanent increase of the remnant resistance of the device (in agreement with Figure 1b , with light). A stronger stimulus or a higher number of events lead to a more efficient inhibitory learning, as it also happens in Hebbian learning ( Figure 2b ). In addition, the presence of light as an extra, external (to the two interconnected neurons) control parameter during learning increases its efficiency. After the inhibitory learning steps (N >600 in processes A2–A5), the system returns to the usual activity protocol and starts forgetting the information. That is, learning is only maintained if the stimulus is present and forgetting occurs as an exponential decay. The key point of our work is that in addition to the normal inhibitory learning, we obtain higher learning efficiency when the system is illuminated. This external light input is the equivalent of a third factor that controls learning in bio-organisms. This is more evident in Figure 2b , where we summarize the efficiency of the different learning processes. We define the efficiency of each inhibitory learning process as the change in the electrical resistance induced in the inhibitory learning part of the overall protocol and in the following form: abs(‘Rrem at step 100’ - ‘Rrem after inhibitory learning pulses’)×100/‘Rrem at step 100’. From Figure 2b it is clear that the remnant resistance change (that is, the efficiency of the inhibitory learning process) is higher in the processes performed under illumination (A2, A4) as compared to the corresponding equivalent processes in the dark (A3, A5). It is especially noticeable that the light irradiation approximately doubles the efficiency of the inhibitory learning process as compared to similar processes performed in the dark (see Figure 2b ). In Figure 3 we present the comparison between protocols A2 and A6. Protocol A6 is equivalent to protocol A2 for the 700 initial sets of V-pulses, after which we have introduced a second inhibitory learning step (50 pulses, 10 V, with light) and a final ‘usual activity’ routine. The comparison between both protocols presents another example of similitude between the learning processes in bio-systems and in light-controlled resistive memories. After a process in which information was efficiently learned (in our case in steps 200 to 600), followed by a time of normal forgetting, the system only needs a small reminder to reach back to the ‘well-learned’ state. In Table 2 we summarize the different processes in which we investigate the efficiency of erasing the information previously learned by the system ( Figure 4 shows the outcome of these processes). The first 600 voltage-pulses sets in processes B1 to B4 are the same as in processes A2 to A6, a usual activity part followed by an inhibitory learning part. These 600 pulses are followed by different information-erasing voltage sets. Forgetting of inhibitory-information in bio-systems is accelerated by positive reinforcements, in our case by applying negative voltage pulses to the device (S<0, voltage pulses presented in Figure 1e ). \n Figure 4a shows the behavior of the memory system after the different Memory erasing processes. The last 400 steps in process A2 ( Table 1 ) are equivalent to a slow, continuous, day-to-day forgetting in bio-systems and are here exemplified by a progressive decrease in the electrical resistance after the inhibitory learning factors have been removed. Processes B1 to B4 (in which we also include erasing steps with negative ‘Operate’ voltage pulses; see Table 2 ) are the equivalent of applying a positive-reinforcement on a bio-system after the inhibitory learning took place. Information is forgotten more efficiently in the dark and with higher-amplitude stimuli, meaning higher-amplitude negative applied voltage pulses (−10V, processes B1, B2, Rrem is more sharply decreasing). This effect can be observed comparing processes B1, B2 at −10 V and B3, B4 at −7 V (in agreement with Figure 1b ). Note that most of the learned information is forgotten within the first two steps of the applied erasing voltages, nevertheless a higher number of erasing pulses lead to better memory cleaning ( Figure 4b ). The Rrem change in this case is defined as: abs((‘Rrem at step 600’ - ‘Rrem after Memory erasing pulses’) *100/‘Rrem at step 600’). As it can be observed, information can be efficiently erased with an adequate set of voltage pulses. Again in this case, the processes studied ( Table 2 , information removal) have a different efficiency in the dark or under illumination. Dark conditions lead to a slightly higher efficiency (B2, B4) than under illumination (B1, B3), highlighting again the role of the external third factor in the memory process."
} | 5,421 |
36923605 | PMC10009452 | pmc | 2,000 | {
"abstract": "Microbial fuel cells (MFCs) incorporating air-breathing cathodes have emerged as a promising eco-friendly wastewater treatment technology capable of operating on an energy-free basis. However, the inevitable biofouling of these devices rapidly decreases cathodic catalytic activity and also reduces the stability of MFCs during long-term operation. The present work developed a novel microbial separator for use in air-breathing MFCs that protects cathodic catalytic activity. In these modified devices, microbes preferentially grow on the microbial separator rather than the cathodic surface such that biofouling is prevented. Trials showed that this concept provided low charge transfer and mass diffusion resistance values during the cathodic oxygen reduction reaction of 4.6 ± 1.3 and 17.3 ± 6.8 Ω, respectively, after prolonged operation. The maximum power density was found to be stable at 1.06 ± 0.07 W m −2 throughout a long-term test and the chemical oxygen demand removal efficiency was increased to 92% compared with a value of 83% for MFCs exhibiting serious biofouling. In addition, a cathode combined with a microbial separator demonstrated less cross-cathode diffusion of oxygen to the anolyte. This effect indirectly induced the growth of electroactive bacteria and produced higher currents in air-breathing MFCs. Most importantly, the present microbial separator concept enhances both the lifespan and economics of air-breathing MFCs by removing the need to replace or regenerate the cathode during long-term operation. These results indicate that the installation of a microbial separator is an effective means of stabilizing power generation and ensuring the cost-effective performance of air-breathing MFCs intended for future industrial applications.",
"conclusion": "4 Conclusion An innovative design for air-breathing MFCs based on incorporating a microbial separator was demonstrated. During long-term operation, microbes were found to preferentially grow on the microbial separator rather than the cathodic surface as a consequence of niche-selectivity. Microbial separators with mature biofilms effectively blocked the cross-cathode diffusion of oxygen to the anolyte and limited the loss of cathodic catalytic activity. These effects provided a maximum power density as high as 1.02 ± 0.03 W m −2 . In addition, these microbial separators showed no negative effects on internal ionic transfer and were found not to be ion selective. Assessments of bacterial biodiversity suggested that the presence of a microbial separator indirectly promoted the growth of an electroactive anodic biofilm while Bugbase phenotypic prediction results demonstrated that the separator biofilm exhibited improved oxygen tolerance. The installation of a microbial separator can greatly improve the sustainability and economics of an air-breathing MFC by removing the need for cathodic replacement or regeneration during prolonged use. The present innovative design employing a microbial separator is evidently an efficient approach to obtaining air-breathing MFCs exhibiting stable power generation and improved removal of pollutants together with lower operational costs and therefore may have future industrial applications.",
"introduction": "1 Introduction Microbial fuel cells (MFCs) represent a new generation of eco-friendly wastewater treatment devices providing simultaneous wastewater treatment and power generation [ 1 , 2 ]. MFCs incorporating air-breathing cathodes were first used to treat wastewater in 2004 and since then have received significant attention in various fields [ 3 , 4 ] with numerous breakthroughs. Zhang et al. [ 5 ] proposed the concept of a separator-assembled cathode configuration that was shown to greatly improve performance [ 5 ]. Feng et al. [ 6 ] demonstrated that a so-called f factor was highly correlated with power generation in scaled-up MFCs [ 6 ]. Dong et al. [ 7 ] suggested a modular design comprising dense stacks that avoided a size effect during MFC scale-up [ 7 ]. Research based on the use of microbial electrochemical processes to address environmental issues has also been reported. The devices that have been studied include the microbial electrolysis cell [ 8 ], microbial desalination cell [ 9 ], microbial carbon capture cell [ 10 ] and microbial reverse electrodialysis cell [ 11 , 12 ]. Over the past decade, many landmark studies have also assessed the extracellular electron transport mechanism [ [13] , [14] , [15] , [16] , [17] , [18] ], as well as device construction and operational optimization [ [19] , [20] , [21] ], innovative material designs [ [22] , [23] , [24] ] and functional expansion [ [25] , [26] , [27] ]. Several pilot-scale MFC devices having the potential for practical applications have already been reported [ [28] , [29] , [30] , [31] , [32] ]. Thus, MFCs are presently undergoing a transition from laboratory research to practical use, although several remaining challenges still limit further engineering applications. Air-breathing MFCs have received much attention because they allow energy-free operation [ 33 , 34 ]. However, because the cathodes in these devices are in direct contact with the anolyte, biofouling of the electrode surfaces inevitably occurs during long-term operation [ 35 ]. This fouling, in turn, increases the cathodic overpotential and decreases the oxygen reduction reaction (ORR) kinetics at the cathode, thus significantly reducing the performance of the MFC [ 36 ]. The cathodic biofouling mechanism is now well understood and many mitigation strategies have been developed [ 36 , 37 ]. The most direct approach of physical cleaning is effective in the case of short operational durations but is not helpful during prolonged use because the deposits tend to bond strongly with the cathode and so cannot be fully removed [ 38 ]. The in situ magnetic cleaning method has been reported to rapidly remove part of the cathodic biofilm [ 39 ]. Chemical lysis systems using acids (such as hydrochloric acid or nitric acid) or alkali (such as sodium hydroxide) [ 40 , 41 ] have been found to regenerate cathodes but requires the use of additional chemicals. The use of ultraviolet radiation for cleaning also results in high energy consumption [ 41 ]. Specially designed antifouling electrodes, such as those impregnated with antibacterial Ag particles, can alleviate cathodic biofouling but tend to shed Ag particles and present toxicity concerns [ 42 ]. The strategies discussed above are primarily meant to control or mitigate biofouling but neglect the positive aspects of the presence of a biofilm layer. As an example, it has been reported that a biofouling layer can decrease the oxygen transfer coefficient at the cathode and limit the diffusion of oxygen into the anolyte [ 43 ]. The permeation characteristics of this film can also promote ion transfer inside the MFC, meaning that the biofilm layer acts similarly to a separator membrane [ 44 ]. Recently, the concept of a microbial separator was proposed and demonstrated in trials using dual chamber biocathode MFCs. These units were able to restrict the direct diffusion of chemical oxygen demand (COD) in the anode and dissolved oxygen (DO) in the cathode to opposite chambers inside MFCs while exhibiting improved ionic cross-separator transfer performance in the electrolyte [ 45 , 46 ]. On this basis, a microbial separator can be expected to perform similar functions in air-breathing MFCs. Furthermore, a microbial separator may also promote the degradation of pollutants [ 47 ]. Thus, it may be important to examine the positive functions of biofouling to allow the further development and application of air-breathing MFCs. In the present study, the novel concept of a microbial separator in an air-breathing MFC was demonstrated. The current density and power density of such devices were investigated as a means of monitoring stability during prolonged operation. In addition, the migration of oxygen across the cathode with or without the separator (as reflected in transfer coefficients) was determined to quantify the ability of the microbial separator to inhibit oxygen diffusion. Electrochemical impedance spectroscopy (EIS) was also employed to ascertain the effect of the microbial separator on the cathodic ORR while linear sweep voltammetry (LSV) data were acquired to demonstrate stable cathodic catalytic activity. The composition and biodiversity of the biofouling biofilm, the microbial separator and the electroactive biofilm were evaluated based on 16S rDNA high-throughput sequencing. A correlative network analysis, BugBase species phenotype contribution analysis and FAPROTAX function prediction were all employed to establish detailed relationships between microbial characteristics and MFC performance.",
"discussion": "3 Results and discussions 3.1 Effect of a microbial separator on long-term stability A separator is typically employed in a dual-chamber rather than a single-chamber MFC. In the case of air-breathing MFCs, the cathode works to close the MFC circuit and can be considered as a compressed cathode chamber. This configuration reduces the apparent internal resistance ( R in ) of the dual-chamber MFC by 50%. However, cathodic biofouling is inevitable when the electrode is in direct contact with the anolyte. In the work reported herein, the air-breathing cathode was covered with a thick biofilm after six months of operation, such that the maximum current density ( I max ) obtained from the MFC was decreased to 0.61 ± 0.02 A m −2 ( Fig. 2 a–c). The installation of a fresh cathode ( Fig. 2 d) immediately increased I max to 0.73 ± 0.01 A m −2 ( Fig. 2 b), indicating the significant effect of cathodic biofouling on MFC performance. The maximum power density was 0.69 ± 0.02 W m −2 after biofouling but 1.02 ± 0.04 W m −2 with a fresh cathode ( Fig. 2 i, j). The apparent internal resistance value ( R in ) for the cathode was approximately 15.1 Ω after fouling but no obvious changes were observed in the value for the anode ( Fig. 2 k, l). In subsequent trials, a microbial separator matrix was installed inside a new cathode ( Fig. 2 e) and there were no significant changes in output current density or apparent internal resistance ( Fig. 2 b, l). After approximately two weeks, a biofilm had been grown on this separator ( Fig. 2 f) and the highest current density ( I max ) was maintained at approximately 0.76 ± 0.01 A m −2 ( Fig. 2 a, b). More importantly, the mature biofilm was found to have grown selectively on the separator skeleton rather than on the cathode ( Fig. 2 g, h) after six months’ operation, even though the microbial separator and cathode were in close proximity to one another. This phenomenon completely eliminated the biofouling of the air-breathing MFCs. During long-term operation for approximately six months, the P max values of MFCs with microbial separators and air-breathing cathodes remained stable at 1.02 ± 0.03 W m −2 ( Fig. 2 j) while the cathodic R in remained low at 50.5 Ω ( Fig. 2 l). It is of note that LSV analyses also indicated that the cathodic catalytic activity was maintained ( Fig. S4 ). The response current obtained from the microbial separator reached a value of approximately 40 mA at a potential of −1.0 V after six months of operation, equal to twice the value observed after biofouling. The coulombic efficiency (CE) and coulombic recovery (CR) of the MFCs with microbial separators were determined to be 24.5% and 26.7%, respectively, and so were much higher than the values of 16.8% and 20.4% for the devices showing serious cathodic biofouling. Therefore, the microbial separator appears to have completely eliminated rather than simply reduced cathode biofouling in these devices to ensure the long-term stability of air-breathing MFCs. 3.2 Restricted cross-cathode transfer of oxygen The stability of MFCs incorporating air-breathing cathodes is greatly affected by the use of an anaerobic anolyte and an efficient cathodic ORR rate. The ideal separator for an MFC should restriction oxygen and COD diffusion while maintaining the cross-separator transfer of ions [ 52 ]. In the present study, the k [COD] for the microbial separator was determined to be approximately 7.5 × 10 −6 cm s −1 ( Fig. S5 ). Although this value was much higher than those obtained using ion-selective membranes ( Table 1 ), this result suggests that the cross-separator transfer of COD could be inhibited [ 45 ]. In addition, the k [DO] of a fresh cathode without biofouling was found to be (1.000 ± 0.003) × 10 −3 cm s −1 ( Fig. 3 a) such that the anodic DO concentration was 0.86 ± 0.21 mg L −1 . Oxygen diffusing into the anolyte has been found to compete with exoelectrogenic bacteria for electrons to limit the generation of current in MFCs [ 36 ]. Prior research has attempted to optimize the electrodes in these devices to balance the amount of oxygen migrating into the anolyte with that reacting during the cathodic ORR [ 50 ]. In the work reported herein, the presence of a thick biofilm on the cathodic electrode caused the k [DO] value to rapidly drop to (7.1 ± 0.1) × 10 −5 cm s −1 ( Fig. 3 a) while the DO concentration in the anolyte was also low at 0.64 ± 0.17 mg L −1 ( Fig. 3 b). Under these conditions, the maximum power density was reduced by approximately 32% because the biofouling significantly lowered the cathodic catalytic activity ( Fig. 3 b). However, a microbial separator covered with a mature biofilm was determined to also inhibit oxygen diffusion. Specifically, an oxygen transfer coefficient of (1.63 ± 0.02) × 10 −4 cm s −1 was obtained ( Fig. 3 a). This value was equal to those observed using ion-selective membranes ( Table 1 ). The microbial separators also maintained the DO concentration in the anolyte below 0.66 ± 0.08 mg L −1 ( Fig. 3 b). In sharp contrast, the biofouling that caused cathodic deactivation was completely avoided in the case that bacteria were directionally enriched at the oxic/anoxic interface of the microbial separator between the air-breathing cathode and anolyte ( Fig. 2 f–h). In fact, this interface was a favorable environment for the growth of mixed microbes [ 62 ]. In addition, the microbes in the microbial separator had preferential access to the substrate, thus allowing niche-selective growth ( Fig. 1 b) such that the output P max was comparable to those of MFCs with fresh cathodes. Table 1 The comparison of key parameters between various separators and microbial separator. Table 1 Separator Oxygen transfer coefficient ( × 10 −4 cm s −1 ) COD transfer coefficient ( × 10 −4 cm s −1 ) Internal resistance (Ω) Cost (USD m −2 ) Reference PEM 6.7 2.2 93 ± 2 1400 [ 53 , 54 ] J-cloth 29 - - 400 [ 52 ] Nafion 1.3 0.00043 84 ± 2 400 a [ 53 , 54 ] UFM 0.19 0.000089 4779 350 [ 54 , 55 ] CEM 0.94 0.00014 84 ± 2 200 [ 54 ] AEM 0.94 0.00055 88 ± 4 80 [ 52 , 54 ] Zirfon 19 - 2727 45 a [ 56 , 57 ] Earthenware - - - 5 a [ 58 ] Non-woven cloth - - 37–51 2–4 [ 53 , 59 ] EPS 2.7 - 500 1.3 a [ 60 ] Glass fiber 0.5 - 2.2 0.31 [ 55 , 61 ] Microbial separator 1.63 ± 0.02 0.075 ∼0.3 0.2 This study Abbreviations: PEM, Proton exchange membrane; UFM, Ultrafiltration membrane; CEM, Cation exchange membrane; AEM, Anion exchange membrane; EPS, Expanded polystyrene. a Calculated or extrapolated from the original research. Fig. 3 a , The dissolved oxygen transfer coefficients of the cathode with biofouling (BF), fresh cathode (FC), and cathode assembled with microbial separator (MS). b , The response of maximum power density to anodic DO concentrations under various conditions. c , The electrochemical impedance spectroscopy (EIS) analysis for the cathodic reduction reaction (The inserted image is a partial enlargement of the original image). d – e , The fitting results of the charge transfer resistances ( R ct , d ), the solution resistances ( R s , e ). f , The mass diffusion resistances ( R d ) in EIS analysis (The inserted image is a partial enlargement of the original image). g , The ionic conductivity testing results (inserted image) of the microbial separator represented by ohmic resistances. h , The schematic diagram of microbial separator model in air-breathing MFCs. (FC: fresh cathode, BF: cathode with biofouling, MS: microbial separator with mature biofilm, FMS: fresh microbial separator in development, SK: the skeleton of microbial separator, RX: reactor X). Fig. 3 Analyses based on EIS were employed to characterize ionic transfer inside the MFCs as well as the extent of charge transfer and mass diffusion during the cathodic ORR ( Fig. 3 c). The anolyte ionic transfer value ( R s ) was not impacted by the installation of a microbial separator ( Fig. 3 g, Table 1 ) and the separator was found to be beneficial in terms of providing improved permeability and a lack of ion selectivity [ 45 ]. The charge transfer resistance ( R ct ) and mass diffusion resistance ( R d ) values were also evaluated and demonstrated the positive effects of a microbial separator on the cathodic ORR. In the case of an MFC incorporating a new cathode, the R ct and R d for the cathodic ORR were 3.5 ± 0.6 and 11 ± 2 Ω, respectively ( Fig. 3 d, f) but these values rapidly increased to 4.9 ± 0.7 and 2369 ± 424 Ω when the cathode was subjected to a high level of biofouling ( Fig. 3 d, f). Biofouling also increased the solution resistance ( R s ) from 14.9 ± 0.2 Ω with a fresh cathode to 18.5 ± 1.2 Ω after fouling ( Fig. 3 e). In contrast, during a prolonged trial incorporating a mature microbial separator, the R ct and R d associated with the ORR were both low and relatively constant at approximately 4.6 ± 1.3 and 17.3 ± 6.8 Ω ( Fig. 3 d, f). These results indicate that the separator worked to maintain cathodic catalytic activity. In summary, a mature microbial separator allowed each air-breathing MFC to generate high currents by maintaining catalytic activity during the cathodic ORR ( Fig. 3 h). 3.3 Enhanced electroactive anodic biofilm growth The biodiversity and microbial compositions of the microbial separators and electroactive anodic biofilms were closely related to the performance of each MFC. Thus, the 16S rDNA sequencing method and associated function prediction analyses were used to establish the correlations between biofilm characteristics and current generation ( Fig. 4 ). The Shannon index values obtained from rarefaction curves showed that the resulting data accurately reflected microbial compositions ( Fig. S6a ). In addition, a co-occurrence network analysis demonstrated obvious otherness among microbial communities ( Fig. S6b ). A partial least squares discriminant analysis also confirmed significant classification differences, especially in the case of the separator biofilm and biofouling ( Fig. 4 a). On the phylum level, Proteobacteria and Actinobacteriota were found to be the dominant species in the biofouling biofilm and the microbial separator communities while Bacteroidota were also common in the biofouling biofilm ( Fig. S6c ). Desulfobacterota, Bacteroidota, and Proteobacteria were the preponderant phyla in the electroactive anodic biofilms along with numerous bacterial genera undergoing extracellular electron transfer, such as the typical genus Geobacter (Desulfobacterota) [ 14 ]. Fig. 4 a , The partial least squares discriminant analysis (PLS-DA) of microbial communities. b , The Bugbase Phenotypic Prediction and the species-phenotypic contribution analysis in (oxidation) stress tolerant (ST) and forms of biofilm (FB). c , The single factor co-occurrence network relationship of top 30 genera. d , The response of maximum power density to the relative abundance of Geobacter species. e , The relative abundance of the functions predicted by FAPROTAX. f , The circus map of species to samples on genus level. g – h , The Wilcoxon rank-sum test bar plots of the biofilm between the microbial separator and cathodic biofouling ( g ) and the anodic samples on genus level ( h ), while the blue and pink circles represent the algebraic difference between a group's data. (FC: fresh cathode, BF: cathode with biofouling, MS: microbial separator with mature biofilm, An: anode, Ca: cathode). Fig. 4 A Bugbase phenotypic analysis was employed to assess the phenotypic properties of the various bacterial communities and to establish the primary species contributing to phenotype expression ( Fig. 4 b). The relative abundance of functional species contributing to biofilm formation was approximately 50% in the biofilm on the microbial separator ( Fig. 4 b) and this biofilm included Rhodococcus , Brucella and others ( Fig. 4 b). The composition of the biofouling biofilm exhibited obvious otherness, and Leucobacter and Aquamicrobium were the two dominant genera ( Fig. 4 b). Although there were large differences in the compositions of the various biofilms, these major phylotypes had similar effects in terms of promoting biofilm development. The microbial separator also promoted directed evolution of anodic electroactive bacteria. In the case of those MFCs exhibiting serious cathodic biofouling, the relative abundance of these bacteria in the anodic biofilm was on the order of 48.1% ( Fig. 4 f), but increased to 54.6% in MFCs with microbial separators. The particular microbes included Geobacter , Aquamicrobium , and Cloacibacillus among others ( Fig. 4 f) [ 14 ]. In particular, the relative abundance of Geobacter rapidly increased to 50.2% in anodic electroactive biofilm in MS stage (An-MS), which was higher than the proportion of 30.2% in the electroactive biofilm in BF stage (An-BF) ( Fig. 4 d). As reported previously, high current densities in MFCs tend to favor the selective enrichment of electroactive bacteria [ 23 ]. In this study, the installation of a microbial separator protected the cathode from biofouling and so maintained the cathodic catalytic activity such that efficient current generation was observed and the quantity of electroactive bacteria was increased. A correlative network analysis demonstrated a close synergistic relationship between the dominant species in the microbial separator, with the exception of Geobacter ( Fig. 4 c). As an example, the correlation coefficients for the relationships between the most common genera were all greater than 0.5 ( Fig. 4 c). The FAPROTAX function prediction suggested that the 21 ± 4% of operational taxonomic units (OTUs) in the microbial separators were related to chemoheterotrophy ( Fig. 4 e), especially for aerobic chemoheterotrophy. This resulted in a favorable anodic habitat with a DO concentration of about 0.66 ± 0.08 mg L −1 that was similar to the value observed in devices with serious biofouling (0.64 ± 0.21 mg L −1 ) ( Fig. 3 b). A species-phenotypic contribution analysis suggested that Rhodococcus and Romboutsia , both of which are oxygen-tolerant, were the two dominant genera in the separator biofilms. In addition, Leucobacter was the most important genus in terms of contributing to the oxidative stress tolerance of the biofouling biofilms ( Fig. 4 b). Although there were differences in bacterial abundances, the Wilcoxon rank-sum test on the genus level demonstrated that there was no significant otherness with regard to bacterial biodiversity ( Fig. 4 g, h). Thus, each of these major phylotypes had a similar effect in terms of restricting cross-biofilm oxygen transfer. Overall, the in situ formation of a biofilm on the microbial separator limited the COD and DO cross-separator diffusion and, more importantly, the COD and DO were simultaneously available to the microbial separator rather than the cathode ( Fig. 3 h). These findings suggest that a microbial separator is superior in terms of realizing the niche-selective enrichment of bacteria and other microbes. 3.4 Improved economics The economic aspects of next generation devices that are still in the development stage must be carefully evaluated. Thus, to date, several pilot-scale studies have been performed to examine the potential applications of air-breathing MFCs by assessing efficiency and optimization of operation [ 6 , 30 , 32 , 63 , 64 ]. It has been reported that severe cathodic biofouling greatly reduces power density (by approximately 91%) within 77 days [ 64 ], such that the replacement or regeneration of cathodes must be conducted within two months. In addition, previous work found that the expense associated with cathodes accounted for 51% of the operating cost of a pilot-scale MFC without expensive cation exchange membrane (CEM) (based on an initial investment of 10,000 USD m −3 ) [ 6 ]. These results demonstrate that the stable and sustainable operation of cathodes would reduce the cost of MFCs by at least half during long-term operation. In this study, very inexpensive nylon textile fabric (approximately 0.2 USD m −2 ) was employed to fabricate microbial separators ( Table 1 ). Thus, compared with other material reported previously, the present microbial separators are highly economical ( Table 1 ). Consequently, cathodic biofouling was completely avoided and cathodic sustainability was achieved at negligible cost. This practice could therefore eliminate the cost of cathodic regeneration in future pilot-scale equipment during long-term operation. Stable power generation also improves the economics of air-breathing MFCs. In the present work, a maximum power density of 1.06 ± 0.07 W m −2 was obtained from MFCs incorporating microbial separators, which was 35% higher than the value of 0.69 ± 0.02 W m −2 for MFCs with serious biofouling. MFCs having higher current densities have been found to promote the growth of electroactive bacteria [ 65 ]. The proportional COD removal was also increased from 83% in conjunction with biofouling to 92% when using a microbial separator. It has been suggested that efficient pollutant degradation and lower sludge yields are more important than high coulombic efficiency or power density values during the operation of MFCs [ 66 , 67 ]. In this regard, the biomass formed on the microbial separators was also superior in terms of the extent to which pollutants were removed [ 66 ]. As an example, a microbial separator installed in a dual chamber MFC improved COD removal by approximately 8.3% [ 66 ] and provided an important pathway for the simultaneous nitrification and denitrification reactions [ 47 ]. Although the precise contribution of the microbial separator in an air-breathing cathode MFC to COD removal was not evaluated in this study, it is conceivable that the separator biofilm facilitated the degradation of organic compounds. 3.5 Facile installation of air-cathode MFCs with microbial separators Energy-free air-breathing MFCs have received spread attention in the wastewater treatment field as more focus has been placed on the efficient use of resources and energy. Over the past decade, several trials of pilot-scale air-breathing MFCs have been reported [ 6 , 7 , 28 , 30 , 32 ], although concerns related to cathodic stability and sustainability continue to limit practical applications. The data acquired using lab-scale MFCs in this work demonstrate the excellent sustainability of cathodes in these devices during long-term operation. On this basis, a potential configuration for a microbial separator unit was proposed for pilot-scale trials, based on multi-module plug-in stacks ( Fig. 5 ). In each module, a single extremely thin porous plate (approximately 0.1 mm) was installed between the microbial separator and cathode to avoid direct contact, while the microbial separator matrix was supported by anolyte pressure ( Fig. 5 c). Recently, a novel air-breathing cathode capable of withstanding a high pressure (water depth was 13 ± 0.7 m) was reported [ 49 ] and this concept was used as the basis for the construction of this novel pilot-scale configuration. This design allowed the fabrication of modular air-breathing cathode MFCs having a plug-in structure that were assembled to build pilot-scale equipment ( Fig. 5 a, b). As reported, the energy consumed by wastewater treatment plants accounts for close to 3% of the total electrical energy requirement in many municipalities [ 68 ]. Thus, a new generation of low-energy microbial electrochemical technology devices would be highly beneficial. Fig. 5 The schematic structure of a pilot-scale air-breathing microbial fuel cell (MFC) installed with microbial separators. a , The schematic of multi-module MFC stacks operated at parallel flow mode. b , The schematic of a single module of pilot-scale MFC constructed with multi-panel air-breathing cathode mode (one anode array paired with two cathode arrays), in which the air-breathing cathode is assembled with microbial separator. c , The assembly details of 2D side view and 3D section view of separator-cathode design (cathodic electrode, porous plate, and skeleton of microbial separator) installed in the pilot-scale MFC. Fig. 5"
} | 7,250 |
20559621 | null | s2 | 2,001 | {
"abstract": "Biofilms are communities of bacteria whose formation on surfaces requires a large portion of the bacteria's transcriptional network. To identify environmental conditions and transcriptional regulators that contribute to sensing these conditions, we used a high-throughput approach to monitor biofilm biomass produced by an isogenic set of Escherichia coli K-12 strains grown under combinations of environmental conditions. Of the environmental combinations, growth in tryptic soy broth at 37 degrees C supported the most biofilm production. To analyze the complex relationships between the diverse cell-surface organelles, transcriptional regulators, and metabolic enzymes represented by the tested mutant set, we used a novel vector-item pattern-mining algorithm. The algorithm related biofilm amounts to the functional annotations of each mutated protein. The pattern with the best statistical significance was the gene ontology 'pyruvate catabolic process,' which is associated with enzymes of acetate metabolism. Phenotype microarray experiments illustrated that carbon sources that are metabolized to acetyl-coenzyme A, acetyl phosphate, and acetate are particularly supportive of biofilm formation. Scanning electron microscopy revealed structural differences between mutants that lack acetate metabolism enzymes and their parent and confirmed the quantitative differences. We conclude that acetate metabolism functions as a metabolic sensor, transmitting changes in environmental conditions to biofilm biomass and structure."
} | 382 |
24965827 | PMC4262009 | pmc | 2,003 | {
"abstract": "Ferrous iron has been known to function as an electron source for iron-oxidizing microorganisms in both anoxic and oxic environments. A diversity of bacteria has been known to oxidize both soluble and solid-phase Fe(II) forms coupled to the reduction of nitrate. Here, we show for the first time Fe(II) oxidation by Sphaerotilus natans strain DSM 6575 T under mixotrophic condition. Sphaerotilus natans has been known to form a sheath structure enclosing long chains of rod-shaped cells, resulting in a thick biofilm formation under oxic conditions. Here, we also demonstrate that strain DSM 6575 T grows mixotrophically with pyruvate, Fe(II) as electron donors and nitrate as an electron acceptor and single cells of strain DSM 6575 T are dominant under anoxic conditions. Furthermore, strain DSM 6575 T forms nanoball-shaped amorphous Fe(III) oxide minerals encrusting on the cell surfaces through the mixotrophic iron oxidation reaction under anoxic conditions. We propose that cell encrustation results from the indirect Fe(II) oxidation by biogenic nitrite during nitrate reduction and that causes the bacterial morphological change to individual rod-shaped single cells from filamentous sheath structures. This study extends the group of existing microorganisms capable of mixotrophic Fe(II) oxidation by a new strain, S. natans strain DSM 6575 T , and could contribute to biogeochemical cycles of Fe and N in the environment.",
"introduction": "Introduction Iron (Fe) exists in divalent or trivalent states within the biosphere depending on the environmental conditions ( Cornell & Schwertmann, 2004 ). Although the abiotic redox changes between Fe(II) and Fe(III) play an important role in redox processes in the environment, microorganisms also significantly contribute to iron biogeochemical cycling in both oxic and anoxic environments on Earth ( Kappler & Straub, 2005 ; Weber et al. , 2006a , b , c ), because they are able to utilize both Fe(II) and Fe(III) as electron donor and acceptor, respectively. However, the exact mechanisms of Fe mineral formation during microbial Fe(II) oxidation is barely understood ( Benzerara et al. , 2011 ). Previous studies suggested that the anaerobic microbiological Fe(II) oxidation occurs either chemotrophically with nitrate ( ) as the electron acceptor ( Nealson & Saffarini, 1994 ; Hafenbradl et al. , 1996 ; Straub et al. , 1996 ; Lack et al. , 2002 ) or phototrophically ( Widdel et al. , 1993 ; Ehrenreich & Widdel, 1994 ; Hegler et al. , 2008 ; Poulain & Newman, 2009 ), which results in the formation of Fe(III) precipitates under anoxic environments. Hafenbradl et al . (1996) reported that the microbiological Fe(II) oxidation coupled with nitrate reduction was achieved using enrichment cultures and pure cultures in the absence of oxygen as a light-independent, chemotropic microbial process. The appearance of microbiological nitrate-dependent Fe(II) oxidation under anoxic natural conditions may play significant roles in coupling the redox cycling of N and Fe in sedimentary environments ( Weber et al. , 2001 ). Furthermore, the anaerobic nitrate-dependent Fe(II) oxidation has important implications for soil and sediment mineralogy and geochemistry through the formation of Fe(III) oxides, including a variety of environmentally relevant Fe(III)-bearing minerals such as ferric oxyhydroxide, goethite, haematite, green rust, and magnetite ( Straub & Buchholz-Cleven, 1998 ; Chaudhuri et al. , 2001 ; Lack et al. , 2002 ; Weber et al. , 2006a , b , c ; Pantke et al. , 2012 ). In addition to the geochemical importance of the anaerobic microbiological Fe(II) oxidation, it has been widely recognized that the aerobic microbiological Fe(II) oxidation under acidic or neutral pH conditions successfully competes with the chemical iron oxidation ( Blake et al. , 1993 ; Emerson & Revsbech, 1994 ; Blake & Johnson, 2000 ; Emerson, 2000 ). Indeed, aerobic Fe(II)-oxidizing bacteria mostly belonging to Betaproteobacteria such as Sphaerotilus/Leptothrix, Gallionella , and Rhodocyclus spp. or to some genera in the Alpha - and Gammaproteobacteria ( Emerson et al. , 2010 ) have been found in a broad range of environments. Among Fe(II)-oxidizing bacteria, Sphaerotilus natans has been characterized by a sheath-forming bacterium enclosing long chains of rod-shaped cells ( Hoeniger et al. , 1973 ). It has been suggested that S. natans is the dominant filamentous bacterium causing bulky biofilm and pipe clogging in waste water treatments due to the formation of sheaths which allow a means of attachment to solid surfaces ( Hoeniger et al. , 1973 ). Despite the environmental issues due to this filamentous bacterium, less information is available regarding iron biomineralization by S. natans and potential importance for the scavenging of inorganic pollutants ( Seder-Colomina et al. , 2013 ). This has led many studies with S. natans to solve bulking problems caused by the filamentous growth of the cells in activated sludge ( Gaudy & Wolfe, 1961 ; Takeda et al. , 1998 ; Suzuki et al. , 2002 ). In addition, it has been known that S. natans strains harbored nitrate reductase activity ( Pellegrin et al. , 1999 ). Although genetic information on nitrate reductase of S. natans was not available to date, the presence of nitrate reductase activity motivated us to study the capacity of Fe(II) oxidation by S. natans at neutral pH under nitrate-reducing conditions. In this study, we tried to determine anaerobic Fe(II) oxidation by S . natans causing cell encrustation and to identity the formed Fe (III) oxide minerals at the cell surfaces. Interestingly, under nitrate-reducing conditions, sheath-forming filamentous S. natans displayed a morphological change to individual rod-shaped single cells encrusted by nanoball-shaped Fe(III) oxide minerals formed from the oxidation of Fe(II). Demonstration of the anaerobic nitrate-dependent Fe(II) oxidation process with formation of Fe(III) oxides encrusting single cells of S. natans could provide information about the extended bacteria group for mixotrophic iron oxidation and also essentially contributes to anaerobic Fe cycling with N.",
"discussion": "Results and discussion Mixotrophic Fe(II) oxidation by S. natans strain DSM 6575 T In the previous study, Gaudy & Wolfe (1961) reported that iron-oxidizing bacteria, S. natans strain DSM 6575 T , grew as a single cell in the presence of both 0.5% of glucose and peptone, and S. natans also showed filamentous growth with sheath formation in the presence of both 0.1% of glucose and peptone (Supporting Information, Fig. S1a and b). However, no additional factors that regulate filamentous growth with sheath formation have been described ( Pellegrin et al. , 1999 ). In our result, it was observed that single rod cells of S. natans strain DSM 6575 T were dominant rather than filamentous long-chained cells in the basal medium with pyruvate and nitrate under anoxic conditions (Fig. S1c). It is suggested that bacterial culture condition such as either the presence of nitrate or anoxic condition can affect the bacterial morphology of S. natans . Furthermore, Pellegrin et al . (1999) confirmed that microaerobic Fe(II)-oxidizing and sheathed filamentous S. natans strain DSM 6575 T harbored nitrate reductase. The amino acid sequence of nitrate reductase from strain DSM 6575 T showed 75.7%, 62.3%, 21% and 20.2% identities to those from Leptothrix cholodnii SP-6 (accession number YP001791710), Acidovorax delafieldil (accession number WP005798151), Paracoccus denitrificans PD1222 (accession number YP918478), and Pseudomonas sp. G-179 (accession number AAC79443) (Fig. S2), respectively. The sequence comparison suggests that nitrate reductase from nitrate-dependent Fe(II) oxidizing bacteria could be quite different structurally from denitrifying bacteria, P. denitrificans and Pseudomonas sp. In this study, strain DSM 6575 T almost completely consumed the provided pyruvate with decreasing from 2 mM to c . 0.02–0.04 mM by 10 days of incubation in bacterial medium containing either nitrate or nitrate and Fe(II) during 10 days of incubation with a similar metabolizing trend for pyruvate (Fig. 1 ). However, the amount of consumed pyruvate in the bacterial medium containing both nitrate and Fe(II) as an electron acceptor and donor, respectively, was slightly higher than that in the bacterial medium containing nitrate alone as an electron acceptor (Fig. 1 ), and the rate of pyruvate consumption for the first 2 days in the presence of both Fe(II) and nitrate was higher than that in the presence of nitrate only, suggesting physiological effect of Fe(II) on the initial anaerobic metabolism of pyruvate. Based on the results, strain DSM 6575 T was able to grow heterotrophically with pyruvate and also mixotrophically with pyruvate and Fe(II) by utilizing nitrate as an electron acceptor under anoxic conditions. In addition, strain DSM 6575 T hardly oxidized Fe(II) in the absence of pyruvate (Fig. 2 a), similar to other bacteria previously described ( Straub et al. , 1996 , 2004 ; Benz et al. , 1998 ; Lack et al. , 2002 ). Strain DSM 6575 T was not able to utilize ferrous iron as a sole electron donor, that is, they need an organic cosubstrate such as pyruvate for both growth and Fe(II) oxidation. Muehe et al . (2009) demonstrated that mixotrophic oxidation of ferrous iron with cosubstrate, acetate, enhanced growth yields with acetate alone (12.5 g dry mass mol −1 acetate) by about 1.4 g dry mass mol −1 Fe(II) and contributes to the energy metabolism of bacteria. Considering the fact that lithoautotrophical growth of the pure culture with both Fe(II) and nitrate was reported to be weak, with hardly two doublings in one cultivation period ( Weber et al. , 2006a , b , c 2006b ), mixotrophic Fe(II) oxidation with pyruvate is the preferred process for nitrate-dependent Fe(II) oxidation in most environments ( Straub & Buchholz-Cleven, 1998 ; Hauck et al. , 2001 ). Indeed, when both nitrate (4 mM) and pyruvate (2 mM) were present in the bacterial culture as an electron acceptor and as an organic substrate, respectively, strain DSM 6575 T almost oxidized the dissolved Fe(II, 4 mM) within 10 days of incubation while consuming only 0.5 mM for the first 2 days (Fig. a ). This two-phase Fe(II) oxidation pattern was also previously reported with Acidovorax strain BoFeN1 ( Klueglein & Kappler, 2013 ; Klueglein et al. , 2014 ), Pseudogulbenkiania strain 2002 ( Weber et al. , 2006a , b , c 2006b ), P. denitrificans ATCC 19367, and P. denitrificans Pd 1222 ( Klueglein et al. , 2014 ). The decrease of Fe(II) in bacterial medium without either nitrate, pyruvate (Fig. 2 a), or bacterial inoculation (Fig. 2 b) was not observed. As mentioned before, strain DSM 6575 T barely oxidized Fe(II) in the absence of pyruvate (Fig. 2 a), indicating that strain DSM 6575 T was likely to depend on the presence of pyruvate for the oxidation of Fe(II). As strain DSM 6575 T oxidized Fe(II) in the bacterial culture containing nitrate and pyruvate during incubation, the provided nitrate (4 mM) was also decreased to c . 0. 3 mM with a small amount of nitrite production in the range of 0.9–1.0 mM at day 10 (Fig. 3 ). In addition, strain DSM 6575 T consumed more nitrate in the bacterial culture containing pyruvate in the presence of Fe(II) than in the absence of Fe(II) during 10 days of incubation (Fig. 3 a). It is assumed that addition of Fe(II) in the presence of pyruvate is likely to provide another electron sources to strain DSM 6575 T for the assimilatory or dissimilatory nitrate utilization with more consumption of pyruvate under anoxic conditions. It should be noted that nitrite accumulation in bacterial culture started at day 1 and reached the maximum concentration, c . 1.0 mM after 6 days (Fig. 3 b). Neither decrease of nitrate nor increase of nitrite was observed in the control experiment without bacterial inoculation (Fig. S3). Nitrite was not formed at bacterial culture in the absence of either Fe(II) or pyruvate (Fig. 3 b). The accumulation of nitrite in bacterial culture of nitrate-dependent Fe(II) oxidation reaction has been also observed in previous studies ( Kappler et al. , 2005 ; Larese-Casanova et al. , 2010 ; Chakraborty et al. , 2011 ). The exact reason that nitrite is accumulated in the mixotrophic Fe(II) oxidation culture has been remained an enigma ( Picardal, 2012 ). However, there is possibility that precipitation of Fe(III) mineral on the nitrite reductase through abiotic Fe(II) oxidation by biogenic nitrite causes inhibition of the nitrite reductase by mineral deposition and leads to nitrite accumulation ( Miot et al. , 2011 ; Picardal, 2012 ). However, further reduced forms of nitrogen species, N 2 O and NO, in bacterial cultures amended with dissolved Fe(II), pyruvate, and nitrate, were not detected in the culture headspace (data not shown). Fig 1 Concentration of consumed pyruvate by Sphaerotilus natans strain DSM 6575 T with alone as electron acceptor (▾), and with nitrate and Fe(II) (▵) in anaerobic medium, and no bacterial inoculation with nitrate (●), and with nitrate and Fe(II) (○) as control experiments. The error bars indicate standard deviation calculated from three independent parallels. The absence of error bars indicates that the error was smaller than the symbol size. Fig 2 Concentration of remaining Fe(II) in the bacterial culture of Sphaerotilus natans DSM 6575 T (a) and control experiment without bacterial inoculation (b). The bacterial culture and control experiment were incubated with Fe(II) (●), with Fe(II) and nitrate (○), with Fe(II) and pyruvate (▾), and with Fe(II), nitrate, and pyruvate (▵), respectively. The error bars indicate standard deviation calculated from three independent parallels. The absence of error bars indicates that the error was smaller than the symbol size. Fig 3 Concentration of remaining nitrate (a) and present nitrite (b) in the bacterial culture of Sphaerotilus natans strain DSM 6575 T . The bacterial culture was incubated with nitrate alone (●), with nitrate and Fe(II) (▵), with nitrate and pyruvate (■), and with nitrate, Fe(II), and pyruvate (Δ), respectively. The error bars indicate standard deviation calculated from three independent parallels. The absence of error bars indicates that the error was smaller than the symbol size. It has been recognized that abiotic oxidation of Fe(II) by nitrate at neutral pH occurs at high temperature (75 °C) ( Buresh & Moraghan, 1976 ) or in the presence of green rust ( Hansen et al. , 1996 ). Therefore, the abiotic oxidation of Fe(II) by nitrate in this study is negligible. Indeed, it has been known that Fe(II) oxidation by nitrite is more rapid than by nitrate ( Picardal, 2012 ). The question is whether the observed Fe(II) oxidation was an enzymatic reaction or chemical reaction by nitrite produced from the nitrate reduction. Klueglein et al . (2014) reported recently that Fe(II) oxidation is a nitrite-driven, indirect mechanism during heterotrophic denitrification of Acidovorax strain BoFeN1, Pseudogulbenkiania strain 2002, P. denitrificans ATCC 19367, and P. denitrificans Pd 1222. As shown in Figs 2 and 3 , fast decrease of Fe(II) after 2 days incubation of strain DSM 6575 T was correlated with nitrite accumulation found only in bacterial culture with pyruvate and Fe(II), suggesting abiotic Fe(II) oxidation by nitrite produced during mixotrophic denitrification. However, it does not rule out the possibility of inducible enzymatic reactions by different bacteria. No concrete evidences have been revealed for the enzymatic or abiotic reactions for Fe(II) oxidation under denitrifying conditions ( Glasauer et al. , 2013 ; Klueglein & Kappler, 2013 ; Klueglein et al. , 2014 ). Cell encrustation and Fe(III) mineral formation SEM image analyses also showed that single cells of strain DSM 6575 T encrusted with Fe(III) mineral crusts at their cell surface in the presence of pyruvate and Fe(II) as the electron donor and nitrate as the electron acceptor (Fig. 4 ). Interestingly, some of the enlarged Fe(III) oxides seemed to show holes or hollowness within the ball-shape structure (see arrows, Fig. 4 c). Fe(III) mineral crusts around the cell surfaces were formed as soon as Fe(II) oxidation occurred (Fig. 4 a and b), and bacterial cells were completely encrusted with Fe(III) mineral formed via iron oxidation (Figs 4 c, d and 5 a, c). TEM analyses revealed the images of bacterial cell cross-section showing the cell interior as well as cell–mineral interfaces (Fig. 5 c). It is observed that S. natans cells contained iron mineral in the cellular membrane, and the thickness of the iron mineral layer is c . 30–40 nm (Fig. 5 c). The thickness of the mineral layer varies among the cells. In addition, the presence of ball-shaped on the surface of the cells as observed in SEM analyses was confirmed (Fig. 5 a and c). However, cell encrustation was not observed in the bacterial culture containing both ferrous iron and nitrite instead of nitrate (Fig. 4 b). Several studies have been also demonstrated that Fe(III) mineral precipitation by nitrate-reducing Fe(II) oxidizer starts in the periplasm, continuous on the cell surface, and then terminates in the cytoplasm ( Miot et al. , 2009 ; Klueglein et al. , 2014 ). Fig 4 SEM images of Sphaerotilus natans strain DSM 6575 T grown in the presence of Fe(II), , and pyruvate under anaerobic conditions. Samples of S. natans strain DSM 6575 T were taken after 1 day (a–b) and 3 days (c–d). The insert (d) shows a close-up image of S. natans strain DSM 6575 T with Fe(III) oxide mineral deposition at the cell surface during Fe(II) oxidation under anaerobic conditions. Fig 5 TEM image of Sphaerotilus natans DSM 6575 T with inserted SAED pattern (a), XRD pattern of precipitates collected from bacterial cultures after 5 days incubation (b), TEM image of cross-sectioned S. natans strain DSM 6575 T (c), and EDS spectrum (d), showing encrustation of the cell surface with amorphous Fe(III) oxide mineral deposition at the cellular membrane structure. The bacterial culture were incubated with pyruvate, Fe(II), and nitrate. It has been reported that metabolically diverse Fe(II)-oxidizing bacteria were encrusted mainly by crystallized Fe(III) goethite mineral ( Emerson & Moyer, 1997 ; Kappler & Straub, 2005 ; Schaedler et al. , 2009 ). Partly or fully encrusted bacterial cells have been also observed in natural environments, such as Fe(II)-rich rivers and springs ( Benzerara et al. , 2008 ; Preston et al. , 2011 ). Moreover, many studies suggest that the encrustation of bacterial cells by Fe(III) oxide minerals has been considered biosignatures or microfossils resulted from the presence of microbial activity in modern and ancient environments ( Banfield et al. , 2001 ; Posth et al. , 2008 ; Cosmidis et al. , 2013 ; Glasauer et al. , 2013 ). However, it is unclear whether the Fe(III) oxide mineral formation encrusting the cell surface is beneficial to the bacterial cells for uptaking or diffusing substrate ( Hallberg & Ferris, 2004 ). In our study, the result of consumed pyruvate by strain DSM 6575 T indicates that encrusted cells could still utilize pyruvate (Fig. 1 ), and therefore, substrate may be transported to the encrusted cells. Previously, the results were reported that encrusted cells were metabolically active ( Miot et al. , 2009 ) and even divided ( Schaedler et al. , 2009 ). As some of researchers suggested, there are possibilities that cell encrustation can protect cells from UV radiation, predation, and dehydration ( Pierson et al. , 1993 ; Phoenix & Konhauser, 2008 ). Furthermore, Fe(III) oxide mineral crust at the cell surface may play a role in electrons transfer from Fe(II) to the cells via the conductive iron mineral crust, as it was shown in abiotic mineral environment ( Schaefer et al. , 2011 ). In addition, XRD analysis for the Fe(III) oxide minerals formed from the nitrate-dependent Fe(II) oxidation by strain DSM 6575 T did not reveal significant signals of crystalline phases (Fig. 5 b and d), indicating formation of amorphous or less crystalline Fe(III) oxide minerals. It has been also known that biologically produced amorphous Fe(III) oxide minerals could be an excellent substrate for a diverse Fe(III)-reducing bacteria in environments ( Straub & Buchholz-Cleven, 1998 ; Lovley et al. , 2004 ) and highly reactive to other metal(loids)s such as arsenic, uranium, and organic pollutants ( Weber et al. , 2001 ; Borch et al. , 2009 ; Hohmann et al. , 2009 , 2011 ; Vaughan & Lloyd, 2011 ; Hitchcock et al. , 2012 ). Thus, this iron species transformation between Fe(II) and Fe(III) by coupling Fe(II)-oxidizing bacteria with Fe(III)-reducing bacteria in the environments could provide important ecological and environmental implications for understanding biogeochemical cycling of Fe. On one hand, the abiotic reaction of ferrous iron with nitrite in the basal medium without bacterial culture at neutral pH leaded to nano-sized yellowish precipitates (Fig. S4a) and identified as goethite [XRD Power Diffraction File (PDF) number 81-0462] (Fig. S4b). The recent study demonstrated that nitrite reacts abiotically with aqueous Fe(II), resulting in the formation of green-rust-like minerals, which are further oxidized to form goethite as the final product ( Kampschreur et al. , 2011 ). In the recent studies, Pantke et al . (2012) reported the formation of green rust during Fe(II) oxidation by the nitrate-reducing Acidovorax sp. strain BoFeNa, and Klueglein et al . (2014) also identified goethite mineral formed by heterotrophic denitrification of Acidovorax strain BoFeN1, Pseudogulbenkiania strain 2002, P. denitrificans ATCC 19367, and P. denitrificans Pd 1222. However, as previously mentioned, the bacterial culture of strain DSM 6575 T in the basal medium containing ferrous iron and nitrate showed reddish precipitates of amorphous or less crystalline Fe(III) oxide minerals (Fig. 5 b and d). Considering Fe(II) oxidation under denitrifying conditions and formation of amorphous or less crystalline Fe(III) oxide minerals on the bacterial surface with changing bacterial morphology, we propose that S. natans strain DSM 6575 T mediates abiotic Fe(II) oxidation using nitrate as an electron acceptor in the presence of pyruvate and provides inducible enzymatic reaction mechanism for the cell encrustation with amorphous iron oxide minerals on the cell membrane. Further experiments are necessary to identify the corresponding enzymes involved in the iron oxidation reactions and their location in the cells. In summary, Fe(II)-oxidizing and sheath-forming S. natans strain DSM 6575 T was able to oxidize Fe(II) using nitrate as an electron acceptor in the presence of pyruvate and appeared as single cells encrusted by the nanoball-shaped amorphous Fe(III) oxide minerals, which contributes to biogeochemical cycles of Fe and N in anaerobic environment. Nucleotide sequence accession numbers This Whole-Genome Shotgun project has been deposited at DDBJ/EMBL/GenBank under the accession AZRA00000000. The version described in this paper is version AZRA01000000."
} | 5,858 |
27748755 | null | s2 | 2,008 | {
"abstract": "Legumes are essential components of agricultural systems because they enrich the soil in nitrogen and require little environmentally deleterious fertilizers. A complex symbiotic association between legumes and nitrogen-fixing soil bacteria called rhizobia culminates in the development of root nodules, where rhizobia fix atmospheric nitrogen and transfer it to their plant host. Here we describe a quantitative proteomic atlas of the model legume Medicago truncatula and its rhizobial symbiont Sinorhizobium meliloti, which includes more than 23,000 proteins, 20,000 phosphorylation sites, and 700 lysine acetylation sites. Our analysis provides insight into mechanisms regulating symbiosis. We identify a calmodulin-binding protein as a key regulator in the host and assign putative roles and targets to host factors (bioactive peptides) that control gene expression in the symbiont. Further mining of this proteomic resource may enable engineering of crops and their microbial partners to increase agricultural productivity and sustainability."
} | 261 |
25641987 | null | s2 | 2,009 | {
"abstract": "Nanotextured surfaces are widely used throughout nature for adhesion, wetting, and transport. Chemistry, geometry, and morphology are important factors for creating tunable textured surfaces, in which directionality of droplets can be controlled. Here, we fabricated nano textured polymeric surfaces, and studied the effect of tilting on the mobility of frequency modulated water droplet transported on asymmetric nano-PPX tracks. Plastically-deformed tracks guided water droplets for sorting, gating, and merging them as a function on their volume. Polymeric ratchets open up new avenues for the fields of digital fluidics and flexible device fabrication."
} | 164 |
38573964 | PMC11009646 | pmc | 2,011 | {
"abstract": "Significance Microbial life is essential for the functioning of ecosystems, yet we often do not know which microbes are responsible for which processes and the speed at which they cycle elements. We developed a proxy to estimate element turnover rates for two critical microbial processes—sulfate reduction and biogenic methane production—that also links these physiologies to specific microbial cells. When applied to a deep fractured rock ecosystem, where current methods for detecting microbial activity would be difficult, we showed this approach to be a powerful tool to measure microbial activity and to link activity to individual cells. Moreover, we resolved single-cell rates of sulfate reduction in a deep subsurface environment, which has not been possible before.",
"conclusion": "Concluding Remarks This study demonstrates the ability to make single-cell specific measurements of anaerobic respiration in ultralow biomass and low-energy ecosystems using flow cytometry integrated with single-cell genomics. Bulk rates of sulfate reduction derived from this single-cell method agreed well with a traditional radiotracer technique. The benefits of the reported approach include superior sensitivity, small sample volume requirements (1 mL is sufficient), short incubation times (30 min), single-cell resolution, and the direct linking of respiration measurements to the genome sequences of analyzed cells. We found that abundant Ca. Desulforudis audaxviator cells respire sulfate in situ, with both acetate and H 2 as electron donors, and likely contribute to a large proportion of elemental cycling and could be responsible for a majority of primary production. We also showed that less abundant members of the Desulfobacterota also had high specific sulfate reduction rates ( Fig. 5 ), decoupling respiration rates from abundance of species. Unexpectedly, the study failed to find conclusive evidence for the presence and activity of methanogenic archaea, which differs from samples collected at Inyo-BLM 1 a decade earlier and other deep subsurface environments where sulfate reducers and hydrogenotrophic methanogens coexist. While this study focused on a deep subsurface habitat, this cell-specific method may be applicable for constraining anaerobic respiration rates of microbial cells, and in resolving the role of essential microbial taxa in a variety of anoxic environments. Fig. 5. Schematic summarizing the respiration activity, general relative abundance, and contributions of the dominant species in the subsurface Inyo BLM-1 ecosystem which was sampled via a well that was cased and packed to prevent localized fluid intrusion. Ca. D. audaxviator cells are the most abundant cells and one of the most active species, and exhibit evidence for active sulfate reduction activity, with highly active Desulfobacterota and active Thermodesulfovibrionales also capable of sulfate reduction. However, these other lineages are present in smaller numbers and exhibit lower levels of respiration. This subsurface system also hosts largely inactive Ca . Hadarchaeota cells, which are likely either not suited for the particular location in the aquifer that was sampled, or are dormant, waiting for substrate to become available. Image not to scale. Surface-land interface taken from Funeral Mountains Geologic Map (978-1-4113-3313-0) and geologic units from Mullin et al. ( 44 ). Fluid flow hypothesis from Merino et al. ( 26 ). Schematic not to scale.",
"discussion": "Results and Discussion Anaerobic Respiration Rates Can Be Determined with an RSG-Based Proxy. RSG has been used in previous studies as a qualitative indicator of cell viability or respiratory activity ( 24 , 28 , 29 ) and with seawater samples to quantify aerobic respiration rates of marine prokaryoplankton ( 17 ). However, to our knowledge, cell-specific rates of anaerobic respiration rates have not been reported. To determine whether the RSG-based proxy could be extended to study anaerobic respiration, we analyzed the relationship between RSG fluorescence measured with a ZE5 flow cytometer and the rates of sulfate reduction ( Fig. 1 A ) and methanogenesis ( Fig. 1 B ) in a variety of stationary-phase pure cultures following a similar, recently described approach that was used to study aerobic respiration ( 17 ) ( Materials and Methods and Dataset S1 ). These particular metabolic processes were targeted because of the dominance of sulfate reducers and methanogens in many terrestrial deep-subsurface ecosystems ( 21 , 30 ). Our culture experiments yielded positive relationships between RSG fluorescence and the rate of production of sulfide by phylogenetically distinct sulfate-reducing microbial cultures (R 2 = 0.95), and the rate of methane produced from both acetoclastic and hydrogenotrophic methanogen cultures (R 2 = 0.66) according to the equations fully described in the Materials and Methods . Given uncertainty about the electron-transfer mechanisms by which RSG can be reduced ( 28 ), potential quantitative relationships between RSG fluorescence and metabolic rates could be further explored with other anaerobic metabolisms. Further calibrations and benchmarking could allow for the exploration of other anoxic ecosystems. Fig. 1. Calibration of RSG-based proxy for sulfate reduction and methanogenesis. ( A ) Comparison of normalized RSG fluorescence intensity against cell-specific sulfide production/sulfate reduction rates in laboratory cultures. Each data point represents one experiment with a different sulfate reducer under defined culture conditions, including the species A. fulgidus at 85 °C and 65 °C, Desulfovibrio vulgaris at 23 °C & 10 mM lactate and 10 °C & 20 mM lactate, and D. putei at 65 °C (isolated from Inyo-BLM 1; more information on cultures and experimental conditions in Dataset S1 ). A comparison of this figure with one high rate value outlier point removed is depicted in SI Appendix , Fig. S1 . ( B ) Comparison of normalized RSG fluorescence intensity against cell-specific methanogenesis rates in laboratory cultures. Each data point represents one experiment with a different methanogenic archaeon, including a Methanobacterium enriched from Inyo-BLM 1 at 65 °C and 55 °C, M. barkeri at 25 °C and 35 °C, and Methanocaldococcus jannaschii at 80 °C. For both comparisons, all fluorescence values were measured on a ZE5 Cell Analyzer flow cytometer but were normalized to fluorescence bead standards used to calibrate the flow cytometers in order to display values comparable to fluorescence measurements from the BD InFlux Mariner flow cytometer (see normalization procedure in Materials and Methods ). Growth curve proxies based on sulfide concentrations (for sulfate reducers) or methane concentrations (for methanogens) and cell counts are provided in SI Appendix , Figs. S2 and S3 , respectively. Representative flow cytometry gates for each culture species are depicted in SI Appendix , Figs. S11 and S12 . Within the cultures used for the RSG calibration, sulfide production ranged from 0.063 to 688 fmol sulfide cell −1 h −1 , (while one experiment exhibited an extremely high production value of 688 fmol sulfide cell −1 h −1 , the next highest value was 76 fmol sulfide cell −1 h −1 ), and methane production ranged from 1.97 to 258 amol methane cell −1 h −1 , with the lower value for each process being the lower limit of quantification for our culture calibration RSG rate calculations. There was one outlier sulfate reduction rate datum which minimally affected results depicted in Fig. 1 (culture calibration), and therefore calculated rates of cell-specific sulfate reduction. For comparison purposes, rate results calculated without this one outlier data point are presented in SI Appendix , Fig. S1 . Additional experiments with the cultures of the sulfate reducer Desulforamulus (basonym: Desulfotomaculum ) putei that was originally obtained from Inyo-BLM 1 subsurface fluids (described in Materials and Methods ) were done in both exponential and stationary growth phases to assess effects of growth phase on RSG fluorescence. These cultures grew to a relatively low density compared with the other cultures ( SI Appendix , Fig. S2 ) but exhibited some of the highest per-cell rates of sulfate reduction measured in cultures compared with previously published rates of psychrophilic or mesophilic cultures ( Fig. 1 A ) ( 31 ). These high rates could be because these are environmental thermophiles that were isolated from a nutrient-poor environment and given an abundance of carbon sources ( Materials and Methods ), which would allow them to potentially metabolize a larger amount of substrate in a shorter period of time. We detected a stronger relationship between RSG fluorescence and activity for stationary-phase cultures ( SI Appendix , Fig. S4 ), which was also observed for aerobic respiration ( 17 ). Thus, we used stationary-phase cultures for calibrating RSG fluorescence and respiration activity rates ( Fig. 1 ). Together, these calibrations confirmed the use of RSG fluorescence intensity as a proxy for cell-specific respiration rates of these anaerobic metabolisms. Phylogenetically Diverse Sulfate-Reducing Bacteria Are Abundant in a Deep Fractured Carbonate Aquifer. To negate potential drilling- and casing-associated artifacts, all samples were collected from a gastight flowing manifold supplied by a submersible pump after flushing with >50 hole volumes. Cell counts determined by staining with SYTO-9 and flow cytometry were 770 cells mL −1 in a pristine manifold sample. Cell concentrations of samples stained with RSG were 465 cells mL −1 in one of the unamended control samples that was incubated for 24 h, and 933, 332, and 748 cells mL −1 in parallel samples amended with acetate (14.9 µM), CO (9 µM), and H 2 (8 µM), respectively. Following FACS and single-amplified genome (SAG) sequencing, 138 SYTO-9- and 119 RSG-stained cells could be grouped into 17 species clusters of two or more cells at 95% average nucleotide identity (ANI) ( SI Appendix , Fig. S5 ). The most abundant of these species clusters were identified by the Genome Taxonomy Database Toolkit (GTDB-Tk) as Ca. D. audaxviator (cluster 1; 47 SAGs), an unidentified species in the order Candidatus Hadarchaeales (cluster 2; 24 SAGs), an unidentified species in the order Thermodesulfovibrionales (cluster 3; 11 SAGs), and an unidentified species in the phylum Desulfobacterota (cluster 4; 11 SAGs). The dominance of Ca. D. audaxviator in the Inyo-BLM 1 subsurface ecosystem was consistent with previous studies at Inyo-BLM 1 and other ultradeep subsurface habitats ( 10 , 18 , 21 ). Overall, Ca. D. audaxviator, which encompasses SAGs belonging to several clusters (clusters 1, 14, and 16, and several unclustered SAGs) represented 34% of all SYTO-9-stained cells (47 SAGs) and 30% of all RSG-stained cells (36 SAGs) and was abundant in both the unamended manifold sample and in amended samples ( Fig. 2 and Dataset S1 ). Ca. Hadarchaeales, Thermodesulfovibrionales , and Desulfobacterota are also often abundant in other terrestrial deep subsurface ecosystems globally ( 21 , 32 ). Generally, the community composition of the manifold sample determined with single-cell sequencing agreed with 16S rRNA gene amplicon data ( SI Appendix , Fig. S6 ), including samples collected from Inyo-BLM 1 pumped water collected at various timepoints since the hole was first drilled in 2007 ( 26 , 30 ). However, there were minor differences in community composition between these two approaches, likely due to the limited number of SAGs that could be analyzed (209 total cells). Fig. 2. Fluorescence intensity and taxonomic identity of cells from Inyo-BLM1 subsurface fluids as observed by flow cytometry. (A) Green fluorescence (y-axis) and red fluorescence (x-axis) of SYTO-9-stained particles recovered directly from the Inyo-BLM 1 manifold samples, from one FACS sort and one 384-well plate of sequencing (microplate AM-294), where green is fluorescence detected with 40 nm bandpass centered by 531 nm wavelength when excited with 488 nm light, and red is fluorescence detected with 40 nm bandpass centered by 692 nm wavelength when excited with 488 nm light. (B) Green fluorescence and red fluorescence from RSG-stained particles recovered directly from Inyo-BLM 1 manifold samples with similar detector settings as described for (A). This figure was constructed with data from three separate plates sorted from one sample (microplates AM-295, AM-296, and AM-297). (C) Green fluorescence and red fluorescence from RSG-stained particles recovered from Inyo-BLM 1 manifold fluids incubated at 65 °C under anoxic conditions with added acetate at 14.9 µM (“amended sample”). This figure was constructed with data from two separate plates sorted from one sample (microplates AM-300 and AM-301). For all three samples depicted in (A–C), particles with sequenced SAGs are highlighted in colored larger dots. The top nine phyla are colored with large dots, with all other phyla pooled and depicted as “Other.” The smaller gray points indicate particles that also passed through the detectors but were not sorted into the 384-well gate for sequencing and analysis, including “background” noise and other unsorted cell-like particles. Other samples not included in the flow cytograms shown here (incubated with no amendment, CO, and H2) are plotted in SI Appendix , Fig. S8 and flow cytograms plotting green fluorescence (y-axis) versus side scatter (x-axis) are depicted in SI Appendix , Fig. S7 . (D) Total RSG fluorescence of gated particles that represent “cell-like” particles in manifold fluids and incubated fluids with different additions incubated at 65 °C for 24 h. Total fluorescence was calculated by summing the green fluorescence of each cell in a sample, normalized to volume. Number of samples analyzed per treatment is indicated by “n.” For logistical reasons, we were unable to analyze enough replicates to determine statistical significance of this stimulation, as the number of samples analyzed per treatment varied between 1 and 3. Three of the four most abundant species clusters, Ca. D. audaxviator (cluster 1), Thermodesulfovibrionales (cluster 3), and Desulfobacterota (cluster 4) encoded key genes involved in sulfate reduction, including sulfate adenylyltransferase ( sat ), adenylylsulfate reductase ( aprAB ), and dissimilatory sulfite reductase ( dsrABC ) ( Fig. 3 ). Descriptions of these genes are detailed in Dataset S3 , and the taxonomy of each SAG is detailed in Dataset S1 . Notably, Ca. D. audaxviator encodes acetyl-CoA synthetase ( acs ) and was enriched among RSG-stained cells incubated with acetate, suggesting the cells were able to utilize acetate as a carbon source and electron donor, likely coupled to sulfate reduction ( Figs. 2 B and D and 4 D ). A putative H 2 uptake hydrogenase belonging to the [NiFe]-hydrogenase group 1a ( 33 ) was also found in Ca. D. audaxviator genomes. Both acetate and hydrogen supported growth of a pure culture of Ca. D. audaxviator in the laboratory ( 20 ). Fig. 3. Putative metabolic capabilities of the four most abundant SAG species clusters in the Inyo-BLM 1 samples—Firmicutes (cluster 1 which can be further classified to Ca. Desulforudis audaxviator) , Ca. Hadarchaeota (cluster 2), Nitrospirota (cluster 3 which can be further classified to Thermodesulfovibrionales ), and Desulfobacterota (cluster 4)—focusing on carbon cycling and energy sources. Genes in green text were present in transcriptome data ( Dataset S3 ), whereas genes in black text were only present in the genomes. An * denotes genes that were classified from transcripts and mapped to SAG reads but were not annotated from SAG reads with DRAM. The number of copies of each gene out of the total number of SAGs for each species cluster (47, 24, 11, and 11 for Ca. D. audaxviator, Ca. Hadarchaeota, Thermodesulfovibrionales , and Desulfobacterota respectively) are shown below the gene names. All four species may use glycolysis and the TCA cycle for central carbon metabolism, while three species clusters ( Ca. D. audaxviator, Thermodesulfovibrionales , and Desulfobacterota) encode complete Wood–Ljungdahl pathways, and some of the Ca. Hadarchaeota genomes only encode the carbonyl branch of the Wood–Ljungdahl pathway. Energy could be conserved in Ca. Desulforudis audaxviator and Thermodesulfovibrionales cells through dissimilatory sulfate reduction, as genomes in both species encode for key enzymes in the sulfate-reduction pathway (key genes include sat : sulfate adenylyltransferase, dsrAB : dissimilatory sulfite reductase, and aprAB : adenylylsulfate reductase). Desulfobacterota cells also encode dsrAB . Energy could be conserved in Ca . Hadarchaeota from hydrogenic CO oxidation, alongside proton reduction using carbon monoxide dehydrogenase ( cooF ), carbon monoxide dehydrogenase ( codh ), ferredoxin oxidoreductase complex ( rnf ), and a [NiFe]-hydrogenase group 4G, which is capable of forming a respiratory complex that couples ferredoxin oxidation with proton reduction ( 33 ). A full list of genes outlined in this figure with more detailed descriptions are in Dataset S4 . Metabolic maps showing presence and absence of genes detailed in this figure for additional clusters can be found in SI Appendix , Fig. S13 ( Ca. Desulforudis audaxviator), SI Appendix , Fig. S14 ( Ca. Hadarchaeota), SI Appendix , Fig. S15 (Desulfobacterota), and SI Appendix , Fig. S16 ( Thermodesulfovibrionales ). Application of the RSG-Based Proxy to Measure Single-Cell Sulfate Reduction Rates. As expected for an oligotrophic environment, the average RSG fluorescence of cells from Inyo-BLM 1 groundwater above detection limit of the flow cytometer (described in SI Appendix , SI Text ) was low ( Fig. 2 and SI Appendix , Figs. S7 and S8 ). Electron donor stimulation experiments with Inyo-BLM 1 fluids raised concentrations of acetate from ambient (below detection limit, <5 µM) to 14.9 (±5.1) µM, of CO from below detection limit (<5 µM) to 9 µM, and of H 2 from 0.206 µM to 8 µM (geochemistry and cell abundance for all samples are described in Dataset S2 ). There was an increase in cumulative RSG fluorescence of the whole community following acetate amendment, indicating metabolism of this substrate ( Fig. 2 C and D ). This comparison was done by adding bulk fluorescence measurements of all cells in each sample normalized to volume. RSG fluorescence of Inyo-BLM 1 cells with added CO or H 2 did not increase, indicating that the amendments did not stimulate additional respiration during the 24-h incubation on a bulk community scale. We observed an overall decrease in RSG-derived sulfate reduction rates in the CO-amended incubation ( Figs. 2 D and 4 D ). This could be caused by the switch of some organisms from respiration to fermentative metabolisms that do not reduce RSG, or the toxicity of CO and/or formic acid, which decomposes into CO. Exceptions to this overall trend were Thermoproteota cells (clusters 8, 9, and 12) that each exhibited increased RSG fluorescence when CO was added ( Fig. 4 D ), indicating that CO stimulated the respiration of these three species, which belong to three different taxonomic orders. This result points to an advantage of this single cell-resolved analysis: while the bulk measurements showed a potential inhibition by CO, we could still identify three taxa whose respiration was stimulated by CO addition. The addition of H 2 did not measurably inhibit or stimulate respiration in any species, which was surprising, given genomic predictions ( 18 ) and the ability of pure cultures of Ca. D. audaxviator to oxidize H 2 and the prevalence of hydrogenotrophy among known sulfate-reducing bacteria. Alternatively, the lack of H 2 stimulation could have been due to the adaptions of these lineages to low concentrations of H 2 , or we may not have added a high enough concentration of H 2 to stimulate respiration. Fig. 4. Evidence for metabolic activity, active cells, and sulfate reducers in low-biomass Inyo-BLM 1 subsurface fluids. ( A ) Inferred rates of sulfate reduction by individual single cells of known taxonomy in Inyo-BLM 1 fluids calculated from the RSG-fluorescence calibration ( Fig. 1 A ) versus estimated cell diameter inferred from forward scatter versus size-calibrated beads (methods for calculation described in SI Appendix , SI Text from ref. 17 ). Each cell either encoded for key genes involved in sulfate reduction in the SAG ( dsrAB and aprAB ) or was closely related (≥95% ANI) to another SAG within this study that encoded the key genes. SAG source indicated with symbol shapes and taxonomy indicated with symbol colors per legend. Per-cell sulfate reduction rates ranged from 0.1 to 27 fmol cell −1 h −1 , with highest per-cell rates observed for Desulfobacterota , Thermoproteota , and Firmicutes cells and lowest rates for Nitrospirota ( Thermodesulfovibrionales ) cells. Estimated cell diameters ranged from 0.4 to 1.8 µm. ( B ) Comparison of bulk sulfate reduction rates in Inyo-BLM 1 samples measured with two different approaches in this study—RSG-based [Translated Bulk Rates (BLM 1)] and 35 SO 4 2− radiotracer incubations [Radiotracer Rates (BLM 1)]—compared to radiotracer incubation-based measurements from other subsurface environments including the CROMO serpentinizing subsurface fluids ( 14 ); and subsurface oceanic crustal fluids from the Juan de Fuca Ridge flank ( 15 ). Translated Bulk Rate (RSG-based) estimates were calculated from cell-specific rates of sulfate reduction (from A ) multiplied by the number of cells identified as sulfate-reducing bacteria per mL and normalized to volume run through the flow cytometer used to sort cells for sequencing. Values listed in Dataset S6 . ( C ) Average 16S rRNA transcript quantities per cell and average relative abundance of cells within different 95% ANI species clusters from the manifold sample stained with RSG. The colors of the dots, in addition to being labeled according to the cluster #, follow the same color pattern as Figs. 2 and 3 . ( D ) Average rates of sulfide production per cell for different sulfate reduction-capable SAG 95% ANI clusters (1, 3, 4, 5, 8, 9, 10, 12, 14, 15) across RSG-stained samples. Previous modeling studies of subsurface environments that take into account total numbers of cells when normalizing bulk rates have assumed that all cells of sulfate-reducing bacteria are equally active ( 34 , 35 ). Our work shows that rates of sulfate reduction varied widely from cell to cell and species to species ( Fig. 4 A and D ). Although we did not have a sufficient number of SAGs to test the statistical significance of these differences, the variability in cell-specific rates implies that bulk rates cannot be accurately extrapolated to individual cells. This further underscores the importance of resolving species-specific activity rates when evaluating the physiological state of cells, especially considering that less abundant microbes can have a disproportionately large ecological impact ( 16 , 17 , 36 ). We also calculated the average sulfate reduction rate for cells from different taxa in the incubations, and this revealed the complexity in the response of different groups to different conditions ( Fig. 4 D ). Overall, Ca. D. audaxviator (cluster 1) cells exhibited the highest cell-specific respiration rates in samples amended with acetate, while cells from the phylum Desulfobacterota (cluster 4 and cluster 15) exhibited high cell-specific respiration rates in the manifold sample (cluster 4 only) and in samples incubated with no amendment, acetate amendment, and H 2 amendment (cluster 4 and cluster 15) ( Fig. 4 D ). Thermodesulfovibrionales cells (cluster 2 and cluster 5) either had low cell-specific respiration rates in all samples except the CO amendment (cluster 2) or were highly active in the manifold sample but were not recovered from electron donor amendments (cluster 5). Overall, these results indicate that the numerical abundance of various microbial taxa in the subsurface environment accessed through Inyo-BLM 1 does not always correlate with respiration activity, which was also shown in a prior study in coastal ocean water using similar tools ( 17 ). Across all cells putatively capable of sulfate reduction based on our flow cytometry sorting, genome annotations, and ANI analysis of the SAGs ( SI Appendix , SI Text and Materials and Methods ), respiration rates varied between 0.14 and 26.9 fmol sulfate cell −1 h −1 , with no correlation between this process and estimated cell diameter ( Fig. 4 A and Dataset S1 ). This large range of sulfate reduction rates assigned to single cells indicates that cells contribute to elemental cycling at different orders of magnitude, and our RSG-based analyses offer a more detailed and potentially more accurate assessment of respiration rate differences between taxa and within taxa across different conditions ( Fig. 1 A ). We then combined these single-cell rate data with counts of cells with the genetic potential for sulfate reduction ( SI Appendix , SI Text and Materials and Methods ). This resulted in volume-specific sulfate reduction rate estimates of 1,377 (±254), 1,789 (±793), 1,068 (±165), 314 (±78), and 763 (±175) pmol sulfate L −1 d −1 in the manifold sample, the incubated unamended sample, the acetate-amended, the CO-amended, and the H 2 -amended samples, respectively ( Fig. 4 B ). In comparison, rates of bulk sulfate reduction determined by the 35 SO 4 2− radiotracer ranged from 25 to 383 pmol sulfate L −1 d 1 ( Fig. 4 B ), i.e., were slightly lower than RSG-based estimates. The discrepancy between radiotracer- and RSG-based rate estimates could be due to the radiotracer approach being a net rate measurement influenced by multiple 35 S cycling processes ( 37 ), while the RSG approach is a gross measurement of oxidoreductase activity. Additionally, radiotracer-based rates were at the limit of detection for the method and required 4 wk of incubation, during which electron donors or other resources may have become limiting, and the composition of microbial communities may have shifted. By contrast, the RSG incubations took only 30 min, substantially reducing the risk of major changes in chemical and biological properties of the samples. However, it is important to keep in mind that the RSG approach may overestimate sulfate reduction because the reduction of RSG may not be fully coupled to sulfate reduction in the analyzed cells. This is due to the uncertainties in the specific electron donors used, or potentially other alternative pathways used by each species that may trigger electron transfer to RSG. For example, the use of different electron donors including hydrogen or formate in Ca. D. audaxviator cells ( 18 , 20 ) may have an unknown effect on RSG fluorescence based on the different fates of electrons. Interestingly, the rates measured with our RSG approach are higher than previously measured sulfate reduction rates that were normalized to cell number ( 31 ). These higher cell-specific rates over bulk rates make sense in light of the fact that bulk rates take into account both active and inactive cells, which could underestimate the activity of the most active cells in situ. This further highlights the diverse cell-specific rates made possible by this RSG approach and the importance of single-cell rate measurements. Nevertheless, both the RSG-based and radiotracer-based estimates indicate that despite relatively high sulfate concentrations (1.75 mM, Dataset S2 ), rates of sulfate reduction in the Inyo-BLM 1 system (25 to 383 pmol sulfate L −1 d 1 ) are lower than in other, previously studied, low-biomass deep biosphere systems, such as the Coast Range Ophiolite Microbial Observatory (600 to 2,047,700 pmol sulfate L −1 d 1 ) ( 14 ) and subsurface crustal fluids of the Juan de Fuca ridge flank (8,000 to 109,000 pmol sulfate L −1 d 1 ) ( 15 ). Species-Specific Gene Expression by Subsurface Microbiota. To further tie RSG fluorescence to specific species and metabolisms, we used all Inyo-BLM 1 SAGs as a reference database for the taxonomic and functional annotation of a metatranscriptome obtained from the Inyo-BLM 1 manifold sample ( Dataset S4 ). Remarkably, metatranscriptome reads were recruited only to 196 out of a total of 297,564 protein-encoding genes. Of these 196 genes, 107 belonged to Ca. D. audaxviator. Ca. D. audaxviator also recruited the greatest number of rRNA reads (1,022 reads normalized to total cell number). These results provide further evidence that this species is abundant and active in this low-energy, deep subsurface environment ( Fig. 4 C ). The detected Ca. D. audaxviator transcripts included sat, aprA, and dsrABC, which are subunits of key enzyme complexes involved in sulfate reduction, and acsC and hyaB, which are involved in the use of acetate and hydrogen as electron donors ( Fig. 3 and Dataset S4 ). We also detected Ca. D. audaxviator transcripts of ATP synthase, which is consistent with chemiosmotic ATP production ( Figs. 3 and 4 C and Dataset S3 ). Cells of Desulfobacterota , particularly cluster 15, also exhibited high transcript counts (903 transcripts per cell), consistent with the relatively high respiration activity of this order as shown by high RSG fluorescence. However, none of these transcripts mapped to genes directly involved in sulfate reduction or any other catabolic pathway ( Fig. 4 C ). It is notable that 16S rRNAs, which is sometimes used as a general metabolic activity proxy ( 38 ), varied between 1 and 1,000 among genera ( Fig. 4 C ). This range was similar to prior reports from seawater ( 17 , 39 , 40 ), although the average value was lower. We speculate that this overall low level of gene expression, and especially the very narrow spectrum of expressed genes, may represent an adaptation to living in an extremely oligotrophic environment, where cells minimize their transcription to conserve energy. Consistent with this hypothesis, as has been shown in other continental deep subsurface sites ( 41 ), most transcripts were related to energy metabolism, such as dsrAB or aprAB . Unexpected Absence of Subsurface Methanogenesis. Intriguingly, no evidence for methanogenesis, methanotrophy, or alkanotrophy [oxidation of short-chain alkanes; ( 42 )] was detected in samples from Inyo-BLM 1 either by taxonomic assignment of SAGs to canonical methanogens, annotation of methanogenic, methanotrophic, or alkanotrophic pathways in SAGs, or through bulk radiotracer methanogenesis measurements ( Fig. 3 ), despite other studies of subsurface environments demonstrating the co-occurrence of sulfate reducers and methanogens ( 21 , 43 ) and the presence of methane in these subsurface fluids measured at 151.5 µM in this study ( Dataset S2 ). Methane was also previously observed at concentrations of up to 100 μM in previous years at this site ( 44 ). Additionally, attempts to measure bulk rates of hydrogenotrophic methanogenesis with 14 C-labeled dissolved inorganic carbon yielded values that were indistinguishable from formaldehyde-killed controls. This result may be time-dependent, however, since earlier pumped sampling intervals at Inyo-BLM 1 (in 2007 and 2011) did reveal much higher proportions of Methanobacterium and Methanothermobacter spp. in 16S rRNA gene libraries [28.6 and 52.7% of prokaryotes], than noted here ( 30 ). In these 2021 samples, the most abundant archaea, cluster 2 and cluster 6, belonged to the phylum Ca. Hadarchaeota and the order Ca. Hadarchaeales ( Dataset S1 and SI Appendix , Fig. S5 ). Members of this phylum were first reported from the same ultradeep mine in South Africa, Driefontein, where Ca. Desulforudis audaxviator was initially detected ( 21 , 45 ) and have been shown to encode one or two methyl-coenzyme M reductase ( mcr ) gene clusters that may function in alkanotrophy ( 46 ). However, in the present study the SAGs from cluster 2 and cluster 6 only encoded distant homologs of a single subunit of mcr , mcrB ( SI Appendix , Fig. S9 ), clustered separately from known methanogenic lineages ( SI Appendix , Fig. S10 ), and lacked other mcr genes and other key genes for alkanotrophy. While the estimated genome completeness for Ca. Hadarchaeota SAGs ranged from 4 to 78%, our confidence for the absence of these pathways is high, due to the large number of Ca. Hadarchaeota SAGs sequenced. Future studies could examine the source of the methane that is observed in these well waters [ Dataset S2 , ( 44 )], or directly assess anaerobic methanotrophy. The abundant Ca. Hadarchaeota cells in Inyo-BLM 1 could potentially use carbon monoxide to fuel anaerobic CO oxidation via carbon monoxide dehydrogenase ( cdhABCDE ) using the “ferredoxin-like” electron carrier ( cooF ) ( 47 ) and an ech complex ([NiFe]-hydrogenase group 4G) that couples ferredoxin oxidation with proton reduction ( 33 ). The Ca. Hadarchaeota cells could fix carbon using the acetyl-CoA synthase/carbon monoxide dehydrogenase ( acs/codh ), and pyruvate ferredoxin oxidoreductase ( porABDG ). This basic metabolism has been previously described as a possible metabolism for other Ca. Hadarchaeota ( 32 ). CO was below method detection limits in Inyo-BLM 1 aquifer fluids (<5 µM, Dataset S2 ) but is common in environments influenced by hydrothermal activity and has been detected in other subsurface habitats where Ca. Hadarchaeota have been detected ( 21 ). CO could be produced by the thermal decomposition of organic matter, or as a side product of some anaerobes ( 48 , 49 ). After sulfate reduction, the CO oxidation reaction has the greatest free energy flux among modeled anaerobic metabolisms in similar deep continental subsurface environments ( 9 , 21 ). However, there was little evidence of respiration activity of Ca. Hadarchaeota in Inyo-BLM 1, aside from two cells present in the hydrogen-amended sample that were relatively low in RSG fluorescence ( SI Appendix , Fig. S7 I and J ). Ca. Hadarchaeota cells were not recovered from the CO-amended samples and very few mRNA or rRNA reads mapped to Ca. Hadarchaeota SAGs ( Fig. 4 C and Dataset S4 )."
} | 8,613 |
36862663 | PMC9980807 | pmc | 2,012 | {
"abstract": "It is possible to generate small amounts of electrical power directly from photosynthetic microorganisms—arguably the greenest of green energy. But will it have useful applications, and what are the hurdles if so?"
} | 53 |
38982629 | PMC11233273 | pmc | 2,013 | {
"abstract": "Abstract Coral microbiomes differ in the mucus, soft tissue and skeleton of a coral colony, but whether variations exist in different tissues of a single polyp is unknown. In the stony coral, Fimbriaphyllia ancora , we identified 8,994 amplicon sequencing variants (ASVs) in functionally differentiated polyp tissues, i.e., tentacles, body wall, mouth and pharynx, mesenterial filaments, and gonads (testes and ovaries), with a large proportion of ASVs specific to individual tissues. However, shared ASVs comprised the majority of microbiomes from all tissues in terms of relative abundance. No tissue‐specific ASVs were found, except in testes, for which there were only two samples. At the generic level, Endozoicomonas was significantly less abundant in the body wall, where calicoblastic cells reside. On the other hand, several bacterial taxa presented significantly higher abundances in the mouth. Interestingly, although without statistical confirmation, gonadal tissues showed lower ASV richness and relatively high abundances of Endozoicomonas (in ovaries) and Pseudomonas (in testes). These findings provide evidence for microbiome heterogeneity between tissues within coral polyps, suggesting a promising field for future studies of functional interactions between corals and their bacterial symbionts.",
"conclusion": "CONCLUSIONS Overall, this study constituted the first survey of bacterial communities in coral polyp tissues using the stony coral, F. ancora, and identified microbial heterogeneity between tissues. A significantly lower abundance of Endozoicomonas was identified in body walls and several bacterial genera were more abundant in the mouth than in other somatic tissues. At the same time, gonadal tissues showed lower ASV richness and higher abundances of Endozoicomonas and Pseudomonas compared to other somatic tissues. These findings pave the way for future studies on functional interaction between corals and their bacterial symbionts. It should be mentioned that our samples comprise multiple polyps from each of the F. ancora colonies subjected in this study. Although polyps from the same colony showed great variation, their belonging to the same genet may introduce some bias. Further investigation with more samples and techniques with higher taxonomic resolution is therefore needed to provide more detailed insights.",
"introduction": "INTRODUCTION Corals form symbioses with a great diversity of microorganisms, including photosynthetic dinoflagellates, bacteria, archaea, viruses, and fungi, collectively termed the coral holobiont. Bacterial symbionts, sometimes simply referred as the coral microbiome, serve various critical functions in the coral holobiont, such as nutrient recycling and synthesis of vitamins and antimicrobial chemicals (Bourne et al., 2016 ; Nissimov et al., 2009 ; Pogoreutz et al., 2022 ). Shifts in the bacterial community in bleached or diseased corals suggest links between microbiomes and coral physiology. In addition, biochemical and genomic studies have demonstrated DMSP (dimethylsulfoniopropionate) degradation or biosynthetic capacity in several coral‐associated bacteria (Chiou et al., 2023 ; Doering et al., 2023 ; Kuek et al., 2022 ; Raina et al., 2009 ; Tandon et al., 2020 ). Based on metagenomic and nanoSIMS (nanoscale secondary ion mass spectrometry) data, Wada et al. ( 2022 ) also revealed polyphosphate accumulation in coral‐associated microbial aggregates (CAMAs) formed predominantly by Endozoicomonas bacteria. These discoveries suggest that in addition to their roles in coral physiology, coral microbiomes may also participate in the biogeochemical cycles of some elements in oligotrophic reef waters. Coral microbiomes are generally dominated by Proteobacteria , Actinobacteria , Bacteroidetes , Firmicutes, and Cyanobacteria , with thousands of operational taxonomic units (OTUs) often found in a single coral species (Blackall et al., 2015 ; Chen et al., 2011 ; Hong et al., 2009 ; Huggett & Apprill, 2019 ; Lee et al., 2016 ; Sweet et al., 2021 ). Coral microbiomes are specific to host species and can vary according to numerous factors, such as the physiological conditions of coral hosts, geographic locations, and environmental fluctuations (Ainsworth et al., 2010 ; Glasl et al., 2019 ; Sunagawa et al., 2010 ; Ziegler et al., 2019 ). In addition, coral microbiomes are heterogeneous within a colony. Distinct microbial communities reside in different compartments of a coral colony, i.e., the coral surface mucus layer, soft tissue, and skeleton (Lee et al., 2016 ; Pollock et al., 2018 ; Sweet et al., 2011 ; Tandon et al., 2023 ). Moreover, factors such as light intensity, water movement, dissolved oxygen level, and even chemical components, can vary greatly between sublocations of a coral colony, e.g., the top and bottom sides of a colony. This variation may create different microhabitats for bacteria, increasing the complexity of coral microbiomes (Hernandez‐Agreda et al., 2017 ). As cnidarians, corals are diploblastic. They lack true organs and consist of two tissue layers (epidermis and gastrodermis) with a layer of extracellular matrix between them (mesoglea). Despite this simple organization, corals exhibit several locally differentiated tissues that are distributed in radial symmetry around the oral‐aboral axis, such as tentacles, mouth (actinopharynx), body wall, mesenteries, and gonads (Galloway et al., 2007 ; Nielsen, 2012 ; Ruppert et al., 2004 ). The coral body wall is the only tissue that is directly in contact with the calcareous skeleton, whereas tentacles contain nematocysts that are used for defence and prey capture. Once a prey item is transported through the mouth into the gastrovascular cavity, mesenterial filaments, which contain abundant glandular cells that produce digestive enzymes, degrade prey and absorb nutrients. Gonads, on the other hand, are specially differentiated tissues located in mesenteries, composed of germ cells, gonadal somatic cells and several types of neurons (Galloway et al., 2007 ). Given that these tissues can encounter very different extracellular conditions, such as concentrations of nutrients, oxygen, and biochemical compounds, they represent fine‐scale microhabitats in individual coral polyps. However, due to the small polyp size (1–3 mm diameter) of most scleractinian corals, it is technically challenging to investigate microbiomes in specific tissues. The anchor coral, Fimbriaphyllia ancora (Cnidaria, Anthozoa, Scleractinia, Euphylliidae), formerly called Euphyllia ancora (Luzon et al., 2017 ), is a common stony coral in the Indo‐Pacific Ocean, including reefs around Taiwan. Its relatively large polyp size (3–5 cm diameter) allows the microscopic isolation of specific tissues for histological and molecular studies (Chiu et al., 2019 ; Shikina et al., 2013 , 2016 ; Shikina, Chiu, et al., 2015 ). In the past two decades, the reproductive biology of F. ancora has been intensively studied in Taiwan, from the possible role of sex steroids in coral spawning (Twan et al., 2003 , 2006 ) to gametogenesis (Chiu et al., 2020 ; Shikina et al., 2012 , 2020 ; Shikina, Chung, et al., 2015 ). In the present study, we employed F. ancora for a survey of bacterial communities in different coral tissues. Tissue‐specific microbiomes have been reported in some other marine invertebrates, which were attributed to differences in biological functions and/or cellular components (Dubé et al., 2019 ; Høj et al., 2018 ; Meisterhans et al., 2016 ). Given that F. ancora exhibits tissue‐specific gene and protein expression (Chiu et al., 2019 ; Shikina et al., 2016 ; Shikina, Chung, et al., 2015 ), we hypothesized that preferential bacterial colonization occurs in different tissues in F. ancora . The results of this study are expected to illuminate microbial heterogeneity in single coral polyps and provide valuable insights into functional interactions between corals and associated microbiomes.",
"discussion": "DISCUSSION Symbiotic associations between corals and bacteria have received great attention due to their contribution to coral health and disease. Coral microbiomes may vary by environmental conditions, coral species, and even among sublocations or compartments of coral colonies (Li et al., 2014 ; Pollock et al., 2018 ; Zhang et al., 2015 ; Ziegler et al., 2019 ). However, microbiome heterogeneity between coral tissues is largely underexplored due to difficulties in isolating tissues from individual polyps. In this study, we examined bacterial communities in different polyp tissues of F. ancora , which has relatively large polyps. Sequencing results showed that F. ancora microbiomes are diverse, with about 80% of identified ASVs specific to individual tissue types and fewer than 2% being common to all tissue types (Figure 3 ). However, these shared ASVs comprised the majority of the bacterial community in terms of abundance (all shared ASVs: 79.3%; conserved shared ASVs: 70.1%), whereas tissue‐specific bacteria constituted only 4.7% of F. ancora microbiomes. Furthermore, no tissue‐specific ASV was conserved in samples of the corresponding tissue except for one ASV specific to testes (ASV0080; affiliated with Endozoicomonas ), for which there were only two samples in our dataset. The identified tissue‐specific ASVs in this study may represent mostly opportunistic growth rather than functional symbiosis with specific coral tissues. Nevertheless, as our rarefied dataset cannot cover the full microbiomes of F. ancora tissues (Figure S1 ) and the sample preparation in this study, i.e., tissue fixing, can introduce some bias, more shared and tissue‐specific bacteria in F. ancora may be discovered when more data become available. Along with the predominance of shared ASVs, beta diversity analysis showed no significant differences among tissue types examined in this study (Figure 5 ). Despite that, the top two components in PCoA explained only ~20% of the variation among samples in this study, suggesting the complexity of microbiomes in F. ancora tissues. Furthermore, several bacterial genera exhibited significantly different abundances between tissue types, including Endozoicomonas and several other bacterial genera (Figure 6 ). Endozoicomonas bacteria are commonly found in stony corals and their abundance has been linked to coral health and stress (Bourne et al., 2008 ; Chuang et al., 2024 ; Neave et al., 2016 ). Earlier studies have shown that Endozoicomonas bacteria form dense cellular aggregates (CAMAs) and are predominantly located in specific polyp tissues such as tentacles (in S. pistillata ) and gastrodermal tissues (in P. damicornis ) (Bayer et al., 2013 ; Neave et al., 2017 ; Wada et al., 2022 ). Our findings add further evidence for the complexity of Endozoicomonas spatial distribution in coral polyps. Furthermore, when compared to other tissues, body walls showed a significantly lower abundance of Endozoicomonas . One of the unique characteristics of body walls is the highly differentiated calicodermis, which synthesizes the calcareous exoskeletons of corals (Allemand et al., 2011 ; Galloway et al., 2007 ). Using whole transcriptomic sequencing techniques, we recently identified several highly expressed genes related to coral skeletogenesis in F. ancora body wall (Shikina et al., 2023 ). Biomineralization in corals involves the secretion of a complicated organic extracellular matrix by calicoblastic cells and delicate control of ion exchange with the extracellular calcifying medium, such as protons, Ca 2+ and HCO 3 \n − (Tambutté et al., 2011 ). Whether these molecules affect the colonization of Endozoicomonas is an open question for future studies. In the mouth tissue of F. ancora , we found significantly higher ASV evenness and relative abundances of several bacterial genera compared to other somatic tissues. The mouth of coral polyps is where food particles enter the gastrovascular cavity and where digestive wastes and gametes are expelled to the surrounding seawater. In the mouth tissue of F. ancora , our recent work identified several highly expressed neuropeptides and neurotransmitters (Shikina et al., 2023 ). These molecules may contribute to the recruitment of specific bacteria to the mouth of F. ancora . Furthermore, the mouth opening constitutes a pathway for bacterial pathogens to enter and infect coral polyps, such as Vibrio coralliilyticus in the stony coral P. damicornis (Gavish et al., 2021 ). Preferential colonization of the mouth of F. ancora by specific bacteria may be linked to coral immunity. However, unlike Endozoicomonas , these mouth‐associated bacteria comprised only a small fraction of the microbiomes of individual tissues (Figure 6 ). Rare bacterial symbionts are often overlooked in microbiome studies but have been proposed to contribute significantly to coral physiology (D Ainsworth et al., 2015 ). The exact functional roles of these mouth‐associated bacteria in F. ancora warrant further investigation. Interestingly, we observed lower Chao1 richness in F. ancora gonads compared to other somatic tissues (Figure 4 ). These findings suggest that gonadal tissues may exhibit more stringent environments than somatic tissues in corals. Steroid hormones have been reported to prime the innate immune system in eukaryotes (García‐Gómez et al., 2013 ; Pace & Watnick, 2021 ; Vom Steeg & Klein, 2017 ). Early studies on F. ancora showed increased levels of steroid hormones during spawning (Twan et al., 2003 , 2006 ). In another stony coral, Mussismilia harttii , Vilela et al. ( 2021 ) showed that the estrogen, ethinylestradiol (EE2), induced significant microbial changes. Considering that our sampling was conducted at the early stages of F. ancora gametogenesis and only included a few gonadal samples, future studies with larger sample sizes among different stages of F. ancora gametogenesis should provide better insights into causation between hormonal differences and microbiome variation between gonadal and somatic polyp tissues. Notably, F. ancora gonads also showed relatively high abundances of Endozoicomonas and Pseudomonas among tissues examined in this study (Figure 7 ). The ability to degrade steroid hormones has been proposed to mediate interactions between bacteria and symbiotic hosts (Chiang et al., 2020 ; Vom Steeg & Klein, 2017 ). Based on both genomic and experimental data, several strains of Endozoicomonas and Pseudomonas are capable of degrading testosterone (Chiang et al., 2020 ; Ding et al., 2016 ; Shintani et al., 2013 ; Yin et al., 1991 ). A gene encoding steroid delta isomerase, which converts pregnenolone to progesterone, was also found in the genome of E. montiporae CL‐33 (Ding et al., 2016 ). In fact, the 4 dominant Endozoicomonas ‐affiliated ASVs identified in this study all matched E. montiporae CL‐33 when BLAST searched against the NCBI rRNA/ITS database (98.28%–100% sequence identity). These findings suggest that Endozoicomonas is not just preferentially associated with gonadal tissues, but may also facilitate gametogenesis in F. ancora . On the other hand, the dominant Pseudomonas ‐affiliated ASV in our dataset was identical to multiple Pseudomonas bacteria ( P. boanensis , P. oleovorans and P. indoloxydans ). Unfortunately, none of these Pseudomonas taxa was originally identified in corals or seawater and no genomic data is available. Thus, no conclusions can be drawn regarding the function of Pseudomonas bacteria in F. ancora , particularly their involvement in gametogenesis. It should also be mentioned that in this study we only sequenced the V6‐V8 region of the bacterial 16S rRNA gene. Therefore, the species‐ or strain‐level taxonomy of our ASVs may be less robust and requires further confirmation."
} | 3,982 |
28559808 | PMC5432669 | pmc | 2,016 | {
"abstract": "Recent advances in the development of data structures to represent spiking neuron network models enable us to exploit the complete memory of petascale computers for a single brain-scale network simulation. In this work, we investigate how well we can exploit the computing power of such supercomputers for the creation of neuronal networks. Using an established benchmark, we divide the runtime of simulation code into the phase of network construction and the phase during which the dynamical state is advanced in time. We find that on multi-core compute nodes network creation scales well with process-parallel code but exhibits a prohibitively large memory consumption. Thread-parallel network creation, in contrast, exhibits speedup only up to a small number of threads but has little overhead in terms of memory. We further observe that the algorithms creating instances of model neurons and their connections scale well for networks of ten thousand neurons, but do not show the same speedup for networks of millions of neurons. Our work uncovers that the lack of scaling of thread-parallel network creation is due to inadequate memory allocation strategies and demonstrates that thread-optimized memory allocators recover excellent scaling. An analysis of the loop order used for network construction reveals that more complex tests on the locality of operations significantly improve scaling and reduce runtime by allowing construction algorithms to step through large networks more efficiently than in existing code. The combination of these techniques increases performance by an order of magnitude and harnesses the increasingly parallel compute power of the compute nodes in high-performance clusters and supercomputers.",
"introduction": "1. Introduction Simulation has become an essential part of the scientific method. In neuroscience, it is employed to investigate the relationship between anatomical and physiological data, to explore dynamical systems not accessible by analytical methods, and to validate approximations made in theoretical derivations. This was made possible by progress in computer hardware as well as in simulation technology for models ranging from the molecular dynamics of ion channels via detailed compartmental models of individual nerve cells (neurons) to brain-scale networks of simple neuron models and field models. Today, simulation codes exist for all of these levels, but the degree of usage by the community varies (Carnevale and Hines, 2006 ; de Kamps et al., 2008 ; Helias et al., 2012 ; Hepburn et al., 2012 ; Ritter et al., 2013 ). Nerve cells interact primarily through stereotyped point events called action potentials or spikes, which are transmitted unidirectionally with delay from sending to receiving cell through contacts known as chemical synapses. After the sending (presynaptic) neuron emits a spike, the receiving (postsynaptic) neuron experiences an excursion of the electric potential difference between the inside and the outside of the cell, the postsynaptic potential (PSP). Typically on the order of ten to one hundred PSPs need to arrive within a few milliseconds in order to elicit an action potential in the postsynaptic neuron (see Abeles, 1991 ; Sterratt et al., 2011 , for textbooks). The essential challenge for a simulation code aiming to implement brain models based on simplified neuron and synapse models is the number of network elements that need to be represented in the computer. In the mammalian cortex, a neuron receives some 10, 000 local inputs. Since two neurons form a connection with a probability of 0.1, the smallest network in which both constraints are simultaneously fulfilled already has 100, 000 neurons. The corresponding volume of roughly one cubic millimeter of tissue can be considered as an elementary unit of cortex. However, the local connections constitute only about half of the input to a neuron, while the other half originates from more distant locations (see Potjans and Diesmann, 2014 , and references therein). A substantial fraction of these long-range connections directly links neurons from different areas in the brain. The human brain is divided into some two hundred areas per hemisphere, but an individual area is only connected to a fraction of them (Glasser et al., 2016 ). The brain thus forms recurrent networks at multiple levels of organization, and due to this intricate coupling between the local and the global level neuroscientists need to study brain-scale networks in order to arrive at self-consistent descriptions of brain activity. In the past decade, research on simulation technology for spiking neuronal networks focused on data structures capable of representing networks of increasing size. Morrison et al. ( 2005 ) presented the first code capable of full-scale simulation of local cortical networks, representing the 100, 000 neurons with their one billion synapses, using distributed computing to aggregate the memory from some ten compute nodes (see also Migliore et al., 2006 , for work carried out at about the same time). Whether downscaling or dilution of neuronal networks preserves the dynamical state of a neuronal network model has been a matter of debate. Recently van Albada et al. ( 2015 ) found that the first-order statistics (e.g., spike rates) can excellently be maintained, but distortions occur already for second-order statistics (e.g., correlations). This observation is relevant not only because correlations of spiking activity are an important measure for the experimentalist and impact spike-timing dependent plasticity (STDP), but also because correlations in neuronal activity drive fluctuations on the population level and thus determine meso- and macroscopic measures such as the local field potential (LFP) and the EEG (Lindén et al., 2011 ; Tetzlaff et al., 2012 ). In light of the limited explanatory power of downscaled network models, the technology of Morrison et al. ( 2005 ) represents a breakthrough, because at the scale of 100, 000 neurons each neuron is supplied with the number of synapses found in nature. Larger networks are necessarily less densely connected, and therefore from this threshold on memory consumption grows linearly with network size, instead of quadratically as is typical for down-scaled networks (see Lansner and Diesmann, 2012 , for details). Morrison and colleagues already point out that the time required to construct a neuronal network model in the main memory of the computer may take up a considerable fraction of the total simulation time and therefore one should make use of all the compute power available. They show furthermore that network construction is ideally parallelizable for a class of network structures. Their technology enables new findings on the dynamics and function of local cortical networks to the present day (Potjans and Diesmann, 2014 ). Due to the progress in computer hardware, networks of 10, 000 neurons are today comfortably studied on a laptop and networks of 100, 000 neurons just require one node of an HPC cluster. Since 2005 (Morrison et al., 2005 ), improvements in neuron (Kunkel et al., 2012 ) and connectivity (Kunkel et al., 2014 ) representation in simulations of networks of spiking neurons have expanded the range of brain models that can be simulated on available computing hardware. The focus of those studies is to minimize memory requirements without sacrificing performance in the propagation of the network state for a given span of biological time. Using supercomputers, networks with more than one billion neurons and the corresponding number of synapses can now be simulated. This already exceeds the number of neurons in the brain of a mouse (100 million), but is still two orders of magnitude away from the number of neurons of the human brain (100 billion). These advances enable the construction of multi-area models addressing the lack of self-consistency in local cortical network models mentioned above and making the link to meso- and macroscopic observables. First neuroscientific results are emerging (Schmidt et al., 2016 ). With the problem of network representation being solved for the range of systems from laptops to petascale supercomputers, increasing the speed of simulations becomes an urgent issue. Simulation times for brain-scale networks are orders of magnitude slower than real time, ruling out the investigation of plasticity and learning which span minutes and hours of biological time. As a first step toward faster simulation, we focus on the time required to create instances of neuronal network models in the main memory of modern computing hardware. We distinguish between two different simulation use cases that we address in this paper, both of which depend on fast network instantiation. One use case is the rapid exploration of the parameter space defining network model properties, including systematic parameter optimization (Martínez-Cañada et al., in press ). This is useful to identify parameter ranges for which a network model shows stable behavior, and typically combines network instantiation with a short simulation run covering on the order of a second of biological time. This use case requires that instantiation is not significantly slower than simulation. The other use case are models at the scale of entire brains, filling petascale supercomputers with on the order of 100, 000 CPU cores. Network construction time on these systems is presently in the range of 15 min (Kunkel et al., 2014 ). Compute time on such systems is a limited commodity and should not be wasted on sub-optimal network instantiation. Furthermore, entire compute nodes or even racks are often allocated for single jobs, implying that usage is only efficient if a simulation employs all processor hardware available in these units. As we focus on network instantiation in this work, especially on connection instantiation, we limit ourselves to simplified neuron models, describing the dynamics of a neuron by a small number of—often linear—ordinary differential equations combined with a threshold and reset mechanism representing the generation of action potentials. Our results apply also to more complex neuron models, including multicompartment models, provided that significantly less memory is required to represent neurons than synapses, and that the number of synapses is much larger than the number of neurons. We further confine our investigation to the creation of connections representing chemical synapses. While neurons interact also by other biophysical mechanisms, some of which are supported by current simulation technology (see Hahne et al., 2015 for electrical synapses (gap junctions) and Potjans et al., 2010 for neuromodulatory control of synaptic plasticity), these do not pose any new challenges for network instantiation from a simulation technology perspective. In practice, neuroscientists formulate their simulation setups using a high-level programming language designed for expressiveness in the problem domain. Research on suitable languages is ongoing. In common use are variants embedded into the Python language (Davison et al., 2008 ; Eppler et al., 2008 ) which has become a de facto standard in computational neuroscience (Muller et al., 2015 ). Network models commonly use combinations of deterministic and probabilistic rules to specify the connectivity among subpopulations of neurons (Crook et al., 2012 ). The script specifying a simulation essentially consists of a sequence of collective Create() commands for the instantiation of populations of different cell types and collective Connect() commands establishing and parameterizing the corresponding synapses. The specifications of the use cases in the present work follow this approach to perform all analyses under realistic conditions. Modern computing hardware beyond the desktop computer typically consists of a number of compute nodes connected by a fast interconnect such as Infiniband. Each compute node contains a number of CPUs, which in turn contain a number of cores that execute instructions. Since all cores within a single compute node share a common main memory and are managed by a single instance of the operating system, it is possible to parallelize simulations within a compute node using threads. A particular specification of a programming model for multi-threading in widespread use is OpenMP (OpenMP Architecture Review Board, 2008 ). Parallelization across multiple compute nodes, on the other hand, requires communication over a physical network. In common use is the message passing interface MPI (Message Passing Interface Forum, 2009 ). As MPI-based parallelization commonly incurs a memory and communication overhead compared to thread-based parallelization, a combination of both technologies is desirable. Initial work concentrating on the phase were the state of the network is advanced in time was already carried out a decade ago (Plesser et al., 2007 ). Here, each MPI process is split into a number of threads and each such thread is called a virtual process (VP). The present work builds upon these early explorations. We consider first the time required to simulate a neuronal network model of a size typically used in computational neuroscience today. The computer is a single multi-core system commonly used in theoretical laboratories. We call this network model small, because it represents only 25% of the neurons within the reach of the local connectivity in the mammalian cortex and only 6.25% of the one billion synapses in a cubic millimeter of cortex. Figure 1 compares MPI- and OpenMP-based parallelization and separates the total time for a simulation run into the time required to construct the network (Figure 1B ) and the time it takes to simulate the network, i.e., to advance the dynamical state of the network over the desired span of biological time (Figure 1C ). Simulation time declines with increasing number of processes for both MPI (blue) and OpenMP (red) until the simulation exhausts the number of computational cores (24). In spite of hardware support for two parallel processes per core (hyperthreading), simulation times increase at first when using more than 24 processes. Even with 48 processes, simulation times are only about 25% shorter than with 24 processes. Still, simulation time is reduced from over five minutes for a single process to slightly more than ten seconds for 48 processes. Figure 1 Performance of a small neuronal network model on a single shared-memory compute node . A balanced random network model (Brunel, 2000 ) representing 25,000 neurons and 62.5 million synapses is simulated for one second of biological time (small benchmark). The compute node houses two CPUs with 12 cores each and up to two hardware threads per core. Table 1 summarizes the configuration. For detailed system specifications see Sections 2.2.1, and 2.2.2 for model specifications. (A) Memory consumption and (B) runtime of network construction as a function of the degree of parallelization. Red indicates parallelization using OpenMP threads and blue using MPI processes. Virtual processes first bind to cores on one CPU (up to 12 VPs), then on the second CPU (up to 24 VPs), and finally to the second hardware thread on each core (up to 48 VPs). The data are averages over five simulations with identical seeds of random number generators. Error bars in (B) show one standard deviation of measurements. (C) Runtime of the simulation of network dynamics, excluding the network construction phase shown in (B) . Same notation as in (B) . The dashed vertical line indicates the total number of physical cores of the compute node. Network construction (Figure 1B ) shows very different scaling for MPI and OpenMP. When using OpenMP, construction times decline markedly only for up to four threads, followed by a complex non-monotonic course and saturate at a network construction time about five times as long as with MPI. Parallel processes do not communicate during network construction and therefore there is a priori no reason why the two parallelization schemes should exhibit different runtime performance. It is instructive to inspect memory consumption shown as a bar diagram in Figure 1A . While memory consumption is independent of parallelization when using OpenMP, it increases when using MPI, exceeding OpenMP by more than 60% for high degrees of parallelization. The scenario may just be indicative of the common runtime vs. memory consumption dilemma. Thus, we need to find out whether the more compact representation enabled by the shared memory access of OpenMP incurs runtime costs due to the need to coordinate access to joint data structures. Figure 2 compares the breakdown of memory consumption for the two parallelization schemes at the highest degree of parallelization studied in Figure 1 . The major part of memory is occupied by synapses and neurons and is of equal absolute size in both schemes. The discrepancy is explained by the overhead for running multiple processes and the data structures of MPI. While in the OpenMP scheme there is only one process, the overhead is multiplied by a factor of 48 for MPI. As the analysis shows that OpenMP does in fact not reduce the memory required for the representation of the network, the runtime vs. memory dilemma does not explain the inferior runtime of OpenMP. Figure 2 Map of memory consumption at the end of the network construction described in Figure 1 . For parallelization with OpenMP (top) 48 threads are used and, correspondingly, for MPI (bottom) 48 processes (rightmost data points in Figure 1 ). The absolute contributions to the total memory consumption are distinguished by color from left to right. In the case of OpenMP these are: synapses (blue, 2 GiB ), neurons (red, 72 MiB ), overhead at start of program (yellow, 6.4 MiB ), initialization of MPI with MPI_Init() (green, 9.4 MiB ), and other NEST data structures (black, 7.9 MiB ). MPI_Init() is not required for OpenMP parallelization and shown for comparison with MPI parallelization only. The enlargement (indicated by gray lines) shows the latter three contributions (total 23.7 MiB ) multiplied by 48 (1138 MiB ). In the case of the MPI simulation the data are sums over all processes, the last three components occupying a total of 1184 MiB . Memory consumption is measured by the resident set size (see Section 2.5). This is the puzzle we need to resolve. Why does network construction in the OpenMP scheme scale so much worse while it does not use a more compact representation of the network in memory and exhibits no disadvantage in the propagation of the dynamical state? This question is relevant because on future computer systems with hundreds of computational cores but limited amounts of memory per core, we cannot afford to spend the major part of memory on the overhead. A lightweight parallelization scheme is required. At the same time, however, without a corresponding scaling of construction time, systems with many cores will not be of use either. As we can demonstrate excellent scaling using the same representation of the network on identical computer hardware, there seems to be no room for a fundamental obstacle in simultaneously achieving efficient usage of memory and scaling. In this study we explain the observed poor scaling for OpenMP and show how to eliminate this bottleneck. The remainder of this paper is organized as follows: In the methods section we first specify the simulation software our quantitative data refer to. Next, we describe in detail the neuronal network models and the computer systems used throughout the present work. In order to validate the generality of our conclusions we extend the benchmark scenario indicated in Figure 1 to a large neuronal network model executed on a supercomputer with thousands of compute nodes, where each compute node harbors only a minor fraction of the neurons. Subsequently, we introduce the state-of-the-art high-level data structures representing a network instance in main memory. When these data structures are created, the simulation engine needs to acquire suitable pieces of memory from the operating system. Section 2.4 explains commonly used allocation strategies and how an application can select between them. Finally, the methods section describes our protocol of obtaining quantitative data and the tools for assessing runtime and memory consumption. In the results section we present an analysis of contributions of different components of the software to the total time required for network construction. For small networks, the creation of connections between neurons dominates, but in the large network setting other components consume the major fraction of runtime. We then show that the standard method of memory allocation serializes the creation of connections when using OpenMP and explain the strategy of advanced allocation algorithms to overcome this without loss of performance for the simulation of the network dynamics. Subsequently we turn to large networks distributed over thousands of compute nodes and exhibit the loops over target neurons that only rarely find local targets as the bottleneck. The reorganization of critical loops eliminates this bottleneck, and combined with thread-aware allocators ensures excellent scaling also for the large benchmark. In Section 4 we discuss our findings in light of the upcoming massively-parallel compute nodes and exascale computers. The technology described in the present article is available in the open-source simulation software NEST version 2.12 (Kunkel et al., 2017 ). The conceptual and algorithmic work described here is a module in our long-term collaborative project to provide the technology for neural systems simulations (Gewaltig and Diesmann, 2007 ).",
"discussion": "4. Discussion Since Dennard scaling enabling the steady increase in processor clock speed came to an end a decade ago (Chang et al., 2010 ), progress in compute power comes from multi-core architectures. Computational neuroscience has taken up the challenge to develop simulation code for spiking neuronal networks coping with the increasing parallelism of the hardware. Today, the dynamics of even rather small networks of the size of a cortical microcolumn ( O ( 1 0 5 ) neurons) can be simulated with highly parallel code reducing the time required for the simulation of one second of biological time from several minutes to a few seconds. For brain-scale networks, petascale parallelization is essential to aggregate the main memory required to represent trillions of synapses. The success of the community's efforts in developing technology for the parallel execution of the dynamics rests on the natural microscopic parallelism of the dynamics of neuronal networks: at a common level of description in computational neuroscience, nerve cells are independent dynamical systems interacting with each other only by delayed point-like events. However, before a particular neuronal network can be simulated it needs to be instantiated in the memory of the computer system. If this phase of the simulation does not parallelize to the same degree as the phase concerned with the dynamics of the network, network construction ultimately limits the scaling of the application. Just at the time when Dennard scaling ended, researchers pointed out the importance of parallel network construction (Morrison et al., 2005 ). The Message Passing Interface (MPI) provided a mechanism to distribute a neuronal network simulation over the compute nodes of high-performance clusters. For the small number of processors per compute node and the small network sizes considered, the resulting code indeed showed excellent scaling of network construction on clusters and shared memory machines. Consequently this software architecture became the backbone of spiking neuronal networks simulators. In the present study we uncover fundamental limitations in the parallelization of network construction on cluster nodes with many compute cores and for networks orders of magnitude larger than previously considered. For small networks (we remind the reader that small in this context means 25, 000 neurons) simulated on a modern compute node, we find excellent strong scaling up to the limits of the multi-core architecture when using MPI for parallelization, although at the price of significant memory overhead (almost 50%). Using OpenMP threads for parallelization instead avoids the memory overhead, scales perfectly in the simulation of network dynamics, but does not scale beyond four parallel processes in network construction. This is disappointing as CPUs already have dozens of compute cores each equipped with hardware supporting multiple threads while fast memory remains limited. It is therefore essential to find technologies exploiting multi-core architectures without the growing memory overhead which parallelization by MPI entails. Our study traces the lack of scaling of OpenMP in the construction phase of small networks to memory allocation: Constructing the adjacency tables representing network connectivity requires a large number of small object allocations and deallocations. When more than four threads perform such allocations and deallocations simultaneously, the ptmalloc memory allocator used by default in current Unix-based systems, significantly slows parallel construction as all threads need to synchronize every time a single thread obtains or returns memory. Using modern allocators optimized for multi-threaded memory operations, such as tcmalloc, jemalloc, and tbb, practically eliminates the locking between threads and restores scaling. An alternative approach to reducing thread contention due to memory allocation is to reduce the number of allocations and frees by creating each connector object as a dynamically-sized container with sufficient memory capacity as soon as the first synapse is registered for any source neuron. Tests with a pre-allocation of 64 elements showed better performance for intermediate thread numbers but no advantages for large numbers of threads and signficantly worse performance than when using modern allocators without pre-allocation (data not shown). The absolute performance of thread-based parallelization is still 60% worse than MPI-based parallelization for more than twelve processes, see Figure 3B for 24 and 48 virtual processes. The origin of this effect is not yet fully understood, but may be related to the non-uniform memory access (NUMA) architecture of modern computers (Hager and Wellein, 2011 , Ch. 4): Each processor has local main memory but can also access the local memory of other processors, albeit with higher latency. For an MPI-program, the operating system usually attempts to hold all data for each process in the memory local to the CPU running the process. When a multi-threaded program runs on several CPUs, though, memory may be allocated by different threads in different parts of the main memory, requiring threads to access data in non-local memory, thus incurring delays. For large, brain-scale networks simulated on supercomputers, thread-based parallelization is vital: these computers typically provide significantly less memory per compute core than HPC clusters and are thus de facto limited to a single MPI process per compute node. We find no or negative scaling beyond 16 threads per compute node. Here, however, the effect found for small networks is overshadowed by an entirely different problem. The connection algorithm spends most of its time iterating over potential target neurons it is ultimately not responsible for, thereby effectively serializing the processing. Reorganization of the loop order for the price of a slightly more complex test removes the serialization and excellent scaling is achieved up to the full number of threads supported by the hardware. Combining the new algorithm with allocators designed for multi-threading further improves the performance. Overall, we have reduced network construction times for microcolumn-sized networks on 48 threads by a factor 3.4, from about 5.3 times to just 1.6 times slower than MPI, while reducing memory consumption by about 25% compared to MPI. At this level of parallelization, network construction now contributes only 1.6 s to the total runtime while it takes 11 s to simulate 1 s of biological activity. Further improvements will thus have less impact. Reducing the network construction times also leads to, albeit minimally, improved times for network simulation, as indicated by Figure 5C . The improvements for the large (brain-scale) benchmark are more significant: The original code requires about 247 s to construct the brain-scale network in the best case (32 threads). After optimization, the code scales much further, constructing the network in just 16 s using 64 threads, an improvement by more than a factor of 17. This difference matters: It reduces the time required to construct a brain-scale network on the entire supercomputer by nearly 4 min, or about 30, 000 core hours, almost 1% of a typical 5 million core hour allocation for a supercomputing project. Brain-scale simulations on supercomputers are not feasible without the improvements in network construction presented here. The quantitative data shown here are obtained using a concrete implementation of network construction and simulation code in the NEST simulator. The conclusions obtained are however generic. The instantiation of model neurons takes up negligible time and models are only distinguished by the number of state variables. Therefore, our results also hold for more complex neuron models than the integrate-and-fire model used throughout the study. The creation of synapses contributes considerably to the time required for network creation. With respect to network creation, synapse models supporting long or short-term plasticity differ from static synapses only in the number of state variables per synapse. Consequently, there is no effect on the scaling of network creation. More complex network structures typically require a larger number of calls of high-level connectivity routines issued from the serial simulation script. This does not reduce the improvement of allocation but an additional serial component reduces the total gain in performance. The advantage of MPI parallelization over multi-threading observed previously is not due to a typical runtime vs. memory consumption dilemma. Multi-threaded code achieves excellent scaling when used with modern memory allocators at considerably lower memory consumption than MPI. In brain-scale network simulations, the number of neurons represented on a single compute node is orders of magnitude smaller than the total number of neurons in the entire network. Connection-generating algorithms need to be designed such that loops extend over local elements only, independent of whether parallelization is realized using MPI or multithreading."
} | 7,764 |
39873757 | PMC11774983 | pmc | 2,017 | {
"abstract": "Abstract Microbial fuel cell (MFC) technology has received increased interest as a suitable approach for treating wastewater while producing electricity. However, there remains a lack of studies investigating the impact of inoculum type and hydraulic retention time (HRT) on the efficiency of MFCs in treating industrial saline wastewater. The effect of three different inocula (activated sludge from a fish-canning industry and two domestic wastewater treatment plants, WWTPs) on electrochemical and physicochemical parameters and the anodic microbiome of a two-chambered continuous-flow MFC was studied. For each inoculum, three different HRTs were tested (1 day, 3 days, and 6 days). The inoculum from the fish canning industry significantly increased voltage production (with a maximum value of 802 mV), power density (with a maximum value of 78 mW m −2 ), coulombic efficiency (with a maximum value of 19.3%), and organic removal rate (ORR) compared to the inocula from domestic WWTPs. This effect was linked to greater absolute and relative abundances of electroactive microorganisms (e.g., Geobacter , Desulfovibrio , and Rhodobacter ) and predicted electron transfer genes in the anode microbiome likely due to better adaption to salinity conditions. The ORR and current production were also enhanced at shorter HRTs (1 day vs. 3 and 6 days) across all inocula. This effect was related to a greater abundance and diversity of bacterial communities at HRT of 1 day compared to longer HRTs. Our findings have important bioengineering implications and can help improve the performance of MFCs treating saline effluents such as those from the seafood industry. Key points \n • Inoculum type and HRT impact organic matter removal and current production. \n \n • Changes in bioenergy generation were linked to the electroactive anodic microbiome. \n \n • Shorter HRT favored increases in the performance of the MFC. \n Graphical Abstract \n Supplementary Information The online version contains supplementary material available at 10.1007/s00253-024-13377-y.",
"introduction": "Introduction Microbial fuel cells (MFCs) have received increased interest as a promising and sustainable approach for treating domestic and industrial wastewater (Boas et al. 2022 ; Tsekouras et al. 2022 ). This technology not only facilitates the removal of organic compounds but also produces electricity from liquid waste, making it a significant advancement in wastewater treatment (Kumar et al. 2022 ; Saravanan et al. 2023 ). However, uncertainties regarding economic viability and energy production have hindered the scale-up and commercialization of MFCs (Jadhav et al. 2021 ; Tsekouras et al. 2022 ). While previous research has shown the efficacy of novel bioelectrochemical systems in desalinating saline wastewater (Odunlami et al. 2023 ), less attention has been directed toward investigating the potential of MFCs for simultaneously removing organic contaminants and generating electricity from saline wastewater (Kim and Logan 2013 ; Kumar et al. 2022 ; Saravanan et al. 2023 ; Castellano-Hinojosa et al. 2024 ). The seafood processing industry generates significant volumes of wastewater during various production stages, including cleaning, chilling, blanching, filleting, cooking, and marination. Estimates suggest that the treatment of 1 ton of seafood requires 10 to 40 cubic meters of water (Arvanitoyannis and Kassaveti 2008 ; Venugopal 2021 ). This wastewater is characterized by moderate levels of organic compounds (ranging from 0.4 to 2 g L −1 of total organic carbon, TOC) and salinity (ranging from 1.2% to 10%), posing unique challenges for treatment and distinguishing it from other industrial effluents (Lefebvre and Moletta 2006 ; Chowdhury et al. 2010 ; Venugopal and Sasidharan 2021 ; Correa-Galeote et al. 2021a ). Previous research has demonstrated the potential of saline industrial and domestic wastewater to support bioenergy generation in MFCs due to their high conductivity (Lefebvre et al. 2012 ; Li et al. 2013 ; Guo et al. 2021 ). A significant limitation in the biological remediation of saline wastewater persists because microorganisms have a restricted capacity to adjust their metabolism to elevated salinity levels (Lefebvre and Moletta 2006 ). Recent investigations have indicated that inoculation of continuous-flow MFCs with halophilic bacteria can achieve chemical oxygen demand (COD) removal rates exceeding 90% with a maximal power density of 420 mW m −2 using industrial saline wastewater (Jamal and Pugazhendi 2021 ). However, other studies have reported similar performance in MFCs operated in batch mode and inoculated with activated sludge from domestic wastewater treatment plants (WWTPs) where microbial communities were not acclimated to elevated saline conditions (Li et al. 2013 ; Guo et al. 2021 ; Xin et al. 2022 ; Sibi et al. 2023 ). These contrasting results could be due to differences in salinity levels and MFC design and operation between studies, which may impact anode colonization and subsequent energy production and organic matter (OM) removal efficiency. Continuous-flow MFCs are better suited for practical applications than batch-mode MFCs because they can handle larger volumes and operate for extended durations (Cai et al. 2014 ; Cabrera et al. 2022 ). Moreover, much of the research on saline wastewater in MFCs has been conducted using fed-batch mode for short periods, typically between 6 to 24 h (Cai et al. 2014 ; Cabrera et al. 2022 ). Yet, additional information is needed on the effect of the inoculum source on electrochemical, physicochemical, and microbial parameters in MFCs fed with saline wastewater and operated in continuous mode. The hydraulic retention time (HRT) plays a critical role in the performance of MFCs (Ye et al. 2020 ; Tanaka et al. 2024 ; Castellano-Hinojosa et al. 2024 ). In general, previous research indicates that higher HRTs tend to enhance OM removal and increase electricity production in MFCs (Kim et al. 2015 , 2016 ; Sobieszuk et al. 2017 ; Fazli et al. 2018 ; Ye et al. 2020 ; Castellano-Hinojosa et al. 2024 ). However, increases in current generation with lower HRT have been reported (Juang et al. 2012 ; Wei et al. 2012 ; Akman et al. 2013 ; Ge et al. 2013 ). Additionally, HRT impacts the level of shear stress experienced by the electroactive biofilm formed on the anode surface, which directly influences its formation (Lecuyer et al. 2011 ; Sorgato et al. 2023 ). Further research is imperative to comprehensively understand the impact of varying HRTs and inocula on MFC performance when treating saline wastewater, potentially aiding in selecting optimal operational conditions to maximize energy production and OM removal. Exoelectrogenic microorganisms play a pivotal role in MFCs (Logan et al. 2019 ; Lovley and Holmes 2021 ). They directly generate electrical currents by oxidizing OM within the anode chamber under oxygen-limiting conditions, transferring electrons to an electrode via various mechanisms (Koch and Harnisch 2016 ; Logan et al. 2019 ; Castellano-Hinojosa et al. 2022a , b ). In continuous-flow MFC systems operating at short HRTs, direct conductive species become particularly crucial. This is because soluble mediators and planktonic cells are swiftly carried away, contrasting with batch culture MFCs (Pannell et al. 2016 ; Logan et al. 2019 ). Previous research has shown that the study of the anodic microbiome can help to understand treatment efficiency and current generation in MFCs operated in continuous mode (Lovley and Holmes 2021 ; Castellano-Hinojosa et al. 2022a , b ; Jalili et al. 2024 ). However, studies on the impact of activated sludge selection and HRT on bioenergy generation and the abundance, diversity, and community composition of electroactive microorganisms in MFCs treating saline wastewater are scarce. In this work, the impact of three different inocula on OM removal %, current production, and the anodic microbiome (abundance, diversity, and community composition) was explored in a continuous-flow MFC fed with saline wastewater. For each inoculum, the MFC was operated at three HRTs to further understand the combined effect of the inoculum source and HRT on the physicochemical, electrochemical, and microbial parameters. The impact of the inoculum type and the HRT on the relative abundance of predicted genes involved in electron transfer were also examined. This study contributes to elucidating the impact of the inoculum type and HRT on physicochemical and electrochemical parameters, and microbial communities of MFCs treating saline wastewater. It was hypothesized that inocula previously adapted to saline conditions would result in higher current production and OM removal efficiency due to a greater abundance and diversity of electroactive microorganisms. It was also hypothesized that higher HRTs would enhance the efficiency of OM removal and energy production regardless of the inoculum source with effects on the anode microbiome.",
"discussion": "Discussion This study demonstrates that the inoculum type and the HRT determine changes in the organic matter removal efficiency, bioelectricity generation, and the anodic microbiome of a continuous-flow MFC fed with saline wastewater. The inoculum from the fish canning industry significantly increased current generation and OM removal % compared to the inocula from domestic WWTPs. This effect was linked to a greater absolute and relative abundance of electroactive microorganisms (e.g., Geobacter , Desulfovibrio , and Rhodobacter ) and predicted electron transfer genes in the anodic microbiome, likely due to better adaption to previous salinity conditions by microbial communities of the FCI inoculum. Regardless of the inoculum, our results revealed that the organic matter removal efficiency and current generation were enhanced at shorter HRT (1 day vs. 3 and 6 days). This effect was related to variations in the diversity, abundance, and composition of bacterial and archaeal communities in the anodic microbiome. For example, increases in OM removal % and energy production at HRT1 were linked to greater abundance and diversity of bacterial communities and increased number of KOs involved in electron transfer compared to longer HRTs. Limited bioelectricity generation was associated with reduced ORR and a greater abundance of archaeal communities, particularly at longer HRTs. Together, our findings have significant bioengineering implications, as they can help improve the performance of MFCs treating saline effluents, such as those from the seafood industry. The inoculum type was found to determine variations in bioelectricity generation and OM removal efficiency in an MFC fed with saline wastewater and operated in continuous mode. Regardless of the HRT, the FCI inoculum led to the greatest values of voltage production, power and current densities, and ORR compared to inocula from domestic WWTPs. This could be due to the activated sludge from the fish-canning industry enriched with electroactive and non-electroactive microorganisms that may rapidly adapt to saline conditions in the wastewater. This assertion is supported by the rapid increases observed in the abundance of bacterial communities and the diversity of prokaryotic communities in the anodic microbiome following inoculation of the MFC with FCI, in contrast to GRA and NEV treatments. In general, the values of voltage production and power density in this study are higher than those reported in other studies using MFCs without saline effluents (Kim et al. 2015 , 2016 ; Sobieszuk et al. 2017 ; Fazli et al. 2018 ; Ye et al. 2020 ). Previous research has demonstrated the potential of saline effluents to support bioenergy generation in MFCs due to their high conductivity (Lefebvre et al. 2012 ; Li et al. 2013 ; Guo et al. 2021 ). However, we extend those findings by showing inocula from industry effluents such as the fish canning industry can not only enhance bioelectricity generation but also the treatment performance of continuous-flow MFCs (You et al. 2010 ; Jayashree et al. 2016 ; Jamal and Pugazhendi 2021 ). Together, these results highlight the potential of continuous-flow MFC for practical applications. Our findings also showed that the inoculum type determined changes in the composition of the anodic microbiome which was tightly linked to variations in OM removal % and current production. A group of known electroactive microorganisms such as Defluviicoccus , Desulfovibrio , Fusibacter , Geobacter , and Rhodobacter were enriched in the anode of the MFC inoculated with FCI compared to GRA and NEV treatments. All these genera contain species that are known to transfer electrons to anodes (Koch and Harnisch 2016 ; Ying et al. 2017 ; Castellano-Hinojosa et al. 2022a , b ). The presence of these taxa was correlated with increased current production, thus suggesting inoculum selection is critical to enhance bioelectricity generation of continuous-flow MFC fed with saline wastewater. Wastewater from the seafood industry has been reported to be enriched with electroactive microorganisms. This is thought to be due to salinity favoring the growth of electroactive microorganisms compared to domestic wastewater (Guo et al. 2021 ; Xin et al. 2022 ). Notably, a total of 16 predicted KOs involved in electron transfer (e.g., cytochromes, pilus, and quinones) in the anodic microbiome were identified, which were significantly more abundant in the MFC inoculated with FCI compared to GRA and NEV treatments. Together, these results suggest that inoculum selection not only determines differences in the diversity and composition of the anode microbiome but also in its functionality, all of which are critical traits for efficient electron transfer. It is interesting to note that there were differences in the abundance, diversity, and composition of prokaryotic communities between GRA and NEV treatments which could be related to the location of the domestic WWTPs. For example, the NEV inoculum was collected from a WWTP located in a high-mountain area above 2000 m a.s.l. whereas the GRA inoculum was taken from a WWTP in Granada city (about 670 m a.s.l.). Previous studies have shown that temperature can impact the microbial ecology of activated sludge in WWTPs (Jung and Regan 2007 ; Griffin and Wells 2016 ). Our findings suggest that differences in microbial ecology between inocula from domestic WWPTs can determine variations in the composition of electroactive microorganisms which may subsequently impact the performance of MFCs. For example, Desulfomicrobium , Pseudomonas , and Rhodobacter were the most abundant exoelectrogenic microorganisms in the MFC treated with GRA whereas Gordonia and Rhodoccocus were the dominant electroactive bacteria after treatment with NEV. Regardless of the inoculum, HRT was found to determine changes in the efficiency of removal of organic compounds and energy production, an effect that appears to be also related to variations in the anodic microbiome. The ORR was increased at shorter HRT (1 day vs. 3 and 6 days) when the ORL was higher. These results can be explained by a lower concentration of organic compounds for use by exoelectrogenic microorganisms at lower OLR, which can result in lower ORR and decreased current generation. This agrees with other studies that showed decreases in power density with reduced OLR at longer HRTs (Liu et al. 2008 ; Sharma and Li 2010 ; Ye et al. 2020 ; Castellano-Hinojosa et al. 2024 ). Jamal and Pugazhendi ( 2021 ) reported that inoculation of MFCs with halophilic consortium can achieve OM removal % of about 84% (initial concentration of 1.21 g COD L −1 ) at 20 days of HRT. Here, it was detected an OM removal % of about 65% at an HRT of only 6 days with double the initial COD (2.46 g COD L −1 ) when the MFC was inoculated with FCI. Together, these results suggest that activated sludge from the seafood industry could be a suitable inoculum for enhancing both organic removal and bioenergy generation in MFCs treating saline effluents. Increases in OM removal % and voltage production at HRT1 were linked to greater abundance and diversity of bacterial communities, along with an increased number of KOs involved in electron transfer, compared to longer HRTs. This could be due to the shorter HRT favoring fast colonization of the anode and subsequent growth of microbial communities, as supported by qPCR results and predicted functionality. Interestingly, it was also noted that limited bioelectricity generation was correlated with an increased abundance of archaeal communities, particularly evident at longer HRTs. Previous studies have shown that the proliferation of archaeal communities at longer HRTs can reduce bioelectricity generation in MFCs, an effect that has been related to competition between exoelectrogenic microbes and methanogens for electrons (Castellano-Hinojosa et al. 2024 ). Alternative processes such as direct or mediated interspecies electron transfer (IET and MIET, respectively) between electroactive bacteria and archaea within the anode can also diminish voltage production (Quéméner et al. 2021 ). We acknowledge that our findings were obtained using synthetic wastewater and real effluents may contain more complex organic compounds than the acetate used in this study. Future studies should explore the impact of inoculum selection and HRT on bioenergy generation and prokaryotic communities of MFCs treating real wastewater. Nonetheless, our findings remain useful as all comparisons in this study were performed under similar operational conditions (e.g., influent composition, temperature, HRT, etc.). This study shows that the inoculum type and the HRT can impact the removal efficiency of organic compounds, bioelectricity generation, and the anodic microbiome of a continuous-flow MFC fed with saline wastewater. Such insights hold the potential for enhancing the valorization of saline effluents from both industrial and domestic wastewater treatment plants (WWTPs) through continuous-flow MFC technology. Our findings show that activated sludge from the fish canning industry increased the electricity generation and organic matter removal compared to the inocula from domestic WWTPs. These effects were tightly linked to variations in the anodic microbiome at different HRTs. For example, the inoculum from the seafood industry favored greater absolute and relative abundances of electroactive microorganisms and predicted electron transfer genes likely due to better adaption to salinity conditions by microbial communities of this inoculum type. Additionally, it was found that the removal efficiency of organic compounds and current generation is enhanced at shorter HRTs (1 day vs. 3 and 6 days) regardless of the inoculum type. This effect was linked to changes in the abundance, diversity, and composition of bacterial and archaeal communities in the anodic microbiome. Despite our findings, scaling up H-cell MFC technology for industrial use still presents several key challenges that should be addressed in future studies. For instance, the construction and maintenance of large-scale reactors involve significant costs, especially in saline environments that require corrosion-resistant materials (Prathiba et al. 2024 ). Cost-effective and life cycle assessment studies are needed to compare the suitability of continuous-flow MFCs for treating saline effluents versus other existing technologies. Operational challenges, such as maintaining stable and efficient microbial populations over time, are also critical, as fluctuations in microbial communities could impact treatment performance and electricity generation. Moreover, shorter HRTs, while beneficial for efficiency, may increase operational complexity and costs. Addressing these challenges will require advancements in reactor design, such as improved materials that reduce costs and enhance durability. Additionally, developing strategies to maintain stable microbial populations, such as optimizing inoculum types and environmental conditions, will be crucial. The potential for bioelectricity generation offers an economic incentive, which could be maximized by integrating MFC systems with existing industrial processes. Collaborative efforts across research, industry, and policy will be essential to overcoming these barriers and realizing the full potential of H-cell MFC technology in practical applications."
} | 5,138 |
25303102 | PMC4193761 | pmc | 2,018 | {
"abstract": "Advancing the size and complexity of neural network models leads to an ever increasing demand for computational resources for their simulation. Neuromorphic devices offer a number of advantages over conventional computing architectures, such as high emulation speed or low power consumption, but this usually comes at the price of reduced configurability and precision. In this article, we investigate the consequences of several such factors that are common to neuromorphic devices, more specifically limited hardware resources, limited parameter configurability and parameter variations due to fixed-pattern noise and trial-to-trial variability. Our final aim is to provide an array of methods for coping with such inevitable distortion mechanisms. As a platform for testing our proposed strategies, we use an executable system specification (ESS) of the BrainScaleS neuromorphic system, which has been designed as a universal emulation back-end for neuroscientific modeling. We address the most essential limitations of this device in detail and study their effects on three prototypical benchmark network models within a well-defined, systematic workflow. For each network model, we start by defining quantifiable functionality measures by which we then assess the effects of typical hardware-specific distortion mechanisms, both in idealized software simulations and on the ESS. For those effects that cause unacceptable deviations from the original network dynamics, we suggest generic compensation mechanisms and demonstrate their effectiveness. Both the suggested workflow and the investigated compensation mechanisms are largely back-end independent and do not require additional hardware configurability beyond the one required to emulate the benchmark networks in the first place. We hereby provide a generic methodological environment for configurable neuromorphic devices that are targeted at emulating large-scale, functional neural networks.",
"conclusion": "Conclusions In this study, we have presented a systematic comparison between neural network simulations carried out with ideal software models and a specific implementation of a neuromorphic computing system. The results for the neuromorphic system were obtained with a detailed simulation of the hardware architecture. The core concept is, essentially, a functionalist one: neural networks are defined in terms of functional measures on muliple scales, from individual neuron behavior up to network dynamics. The various neuron and synapse parameters are then tuned to achieve the target performance in terms of these measures. The comparison was based on three cortically inspired benchmark networks: a layer 2/3 columnar architecture, a model of a synfire chain with feed-forward inhibition and a random network with self-sustained, irregular firing activity. We have chosen these specific network architectures for two reasons. First of all, they implement very different, but widely acknowledged computational paradigms and activity regimes found in neocortex: winner-take-all modules, spike-correlation-based computation, self-sustained activity and asynchronous irregular firing. Secondly, due to their diverse properties and structure, they pose an array of challenges for their hardware emulation, being affected differently by the studied hardware-specific distortion mechanisms. All three networks were exposed to the same set of hardware constraints and a detailed comparison with the ideal software model was carried out. The agreement was quantified by looking at several chosen microscopic and and macroscopic observables on both the cell and network level, which we dubbed “functionality criteria”. These criteria were chosen individually for each network and were aimed at covering all of the relevant aspects discussed in the original studies of the chosen models. Several hardware constraint categories have been studied: the dynamics of the embedded neuron and synapse models, limited parameter ranges, synapse loss due to limited hardware resources, synaptic weight noise due to fixed-pattern and trial-to-trial variations, and the lack of configurable axonal delays. The final three effects were studied in most detail, as they are expected to affect essentially every hardware-emulated model. The investigated distortion mechanisms were studied both individually, as well as combined, similarly to the way they would occur on a real hardware substrate. As expected, above certain magnitudes of the hardware-specific distortion mechanisms, substantial deviations of the functionality criteria were observed. For each of the three network models and for each type of distortion mechanism, several compensation strategies were discussed, with the goal of tuning the hardware implementation towards maximum agreement with the ideal software model. With the proposed compensation strategies, we have shown that it is possible to considerably reduce, and in some cases even eliminate the effects of the hardware-induced distortions. We therefore regard this study as an exemplary workflow and a toolbox for neuromorphic modelers, from which they can pick the most suitable strategy and eventually tune it towards their particular needs. In addition to the investigated mechanisms, several other sources of distortions are routinely observed on neuromorphic hardware. A (certainly not exhaustive) list might include mismatch of neuron and synapse parameters, shared parameter values (i.e., not individually configurable for each neuron or synapse) or limited parameter programming resolution. These mechanisms are highly back-end-specific and therefore difficult to generalize. However, although they are likely to pose individual challenges by themselves, some of their ultimate effects on the target network functionality can be alleviated with the compensation strategies proposed here. Our proposed strategies aim at neuromorphic implementations that compete in terms of network functionality with conventional computers but offer major potential advantages in terms of power comsumption, simulation speed and fault tolerance of the used hardware components. If implemented successfully, such neuromorphic systems would serve as fast and efficient simulation engines for computational neuroscience. Their potential advantages would then more than make up for the overhead imposed by the requirement of compensation. From this point of view, hardware-induced distortions are considered a nuisance, as they hinder precise and reproducible computation. In an alternative approach, one might consider the performance of the system itself at some computational task as the “fitness function” to be maximized. In this context, some particular architecture of an embedded model, together with an associated target behavior, would then become less relevant. Instead, one would design the network structure specifically for the neuromorphic substrate or include training algorithms that are suitable for such an inherently imperfect back-end. The use of particular, “ideal” software models as benchmarks might then given up altogether in favor of a more hardware-oriented, stand-alone approach. Here, too, the proposed compensation strategies can be actively embedded in the design of the models or their training algorithms. The hardware architecture used for our studies is, indeed, suited for both approaches. It will be an important aspect of future research with neuromorphic systems to develop procedures that tolerate or even actively embrace the temporal and spatial imperfections inherent to all electronic circuits. These questions need to be addressed by both model and hardware developers, in a common effort to determine which architectural aspects are important for the studied computational problems, both from a biological and a machine learning perspective.",
"introduction": "Introduction 1.1 Modeling and Computational Neuroscience The limited availability of detailed biological data has always posed a major challenge to the advance of neuroscientific understanding. The formulation of theories about information processing in the brain has therefore been predominantly model-driven, with much freedom of choice in model architecture and parameters. As more powerful mathematical and computational tools became available, increasingly detailed and complex cortical models have been proposed. However, because of the manifest nonlinearity and sheer complexity of interactions that take place in the nervous system, analytically treatable ensemble-based models can only partly cover the vast range of activity patterns and behavioral phenomena that are characteristic for biological nervous systems [1] . The high level of model complexity often required for computational proficiency and biological plausibility has led to a rapid development of the field of computational neuroscience, which focuses on the simulation of network models as a powerful complement to the search for analytic solutions [2] . The feasibility of the computational approach has been facilitated by the development of the hardware devices used to run neural network simulations. The brisk pace at which available processing speed has been increasing over the past few decades, as allegorized by Moore's Law, as well as the advancement of computer architectures in general, closely correlate to the size and complexity of simulated models. Today, network models with tens of thousands of neurons are routinely simulated on desktop machines, with supercomputers allowing several orders of magnitude more [3] , [4] . However, as many authors have pointed out (see e.g. [5] , [2] ), the inherently massively parallel structure of biological neural networks becomes progressively difficult to map to conventional architectures based on digital general-purpose CPUs, as network size and complexity increase. Conventional simulation becomes especially restrictive when considering long time scales, such as are required for modeling long-term network dynamics or when performing statistics-intensive experiments. Additionally, power consumption can quickly become prohibitive at these scales [6] , [7] . 1.2 Neuromorphic Hardware The above issues can, however, be eluded by reconsidering the fundamental design principles of conventional computer systems. The core idea of the so-called neuromorphic approach is to implement features (such as connectivity) or components (neurons, synapses) of neural networks directly in silico : instead of calculating the dynamics of neural networks, neuromorphic devices contain physical representations of the networks themselves, behaving, by design, according to the same dynamic laws. An immediate advantage of this approach is its inherent parallelism (emulated network components evolve in parallel, without needing to wait for clock signals or synchronization), which is particularly advantageous in terms of scalability. First proposed by Mead in the 1980s [8] – [10] , the neuromorphic approach has since delivered a multitude of successful applications [11] – [14] . By far the largest number of neuromorphic systems developed thus far are highly application-specific, such as visual processing systems [15] – [18] or robotic motor control devices [19] . Several groups have focused on more biological aspects, such as the neuromorphic implementation of biologically-inspired self-organization and learning [20] , [21] , detailed replication of Hodgkin-Huxley neurons [22] or hybrid systems interfacing analog neural networks with living neural tissue [23] . These devices, however, being rather specialized, can not match the flexibility of traditional software simulations. Adding configurability comes at a high price in terms of hardware resources, due to various hardware-specific limitations, such as physical size and essentially two-dimensional structure. So far there have only been few attempts at realizing highly configurable hardware emulators [24] – [28] . This approach alone, however, does not completely resolve the computational bottleneck of software simulators, as scaling neuromorphic neural networks up in size becomes non-trivial when considering bandwidth limitations between multiple interconnected hardware devices [29] – [33] . 1.3 The BrainScaleS Hardware System A very efficient way of interconnecting multiple VLSI (Very Large Scale Integration) modules is offered by so-called wafer-scale integration. This implies the realization of both the modules in question and their communication infrastructure on the same silicon wafer, the latter being done in a separate, post-processing step. The BrainScaleS wafer-scale hardware [27] uses this process to achieve a high communication bandwidth between individual neuromorphic cores on a wafer, thereby allowing a highly flexible connection topology of the emulated network. Together with the large available parameter space for neurons and synapses, this creates a neuromorphic architecture that is comparable in flexibility with standard simulation software. At the same time, it provides a powerful alternative to software simulators by avoiding the abovementioned computational bottleneck, in particular owing to the fact that the emulation duration does not scale with the size of the emulated network, since individual netowrk components operate, inherently, in parallel. An additional benefit which is inherent to this specific VLSI implementation is the high acceleration with respect to biological real-time, which is facilitated by the high on-wafer bandwidth. This allows investigating the evolution of network dynamics over long periods of time which would otherwise be strongly prohibitive for software simulations. 1.4 Hardware-Induced Distortions: A Systematic Investigation Along with the many advantages it offers, the neuromorphic approach also comes with limitations of its own. These have various causes that lie both in the hardware itself and the control software. We will later identify these causes, which we henceforth refer to as distortion mechanisms . The neural network emulated by the hardware device can therefore differ significantly from the original model, be it in terms of pulse transmission, connectivity between populations or individual neuron or synapse parameters. We refer to all the changes in network dynamics (i.e., deviations from the original behavior defined by software simulations) caused by hardware-specific effects as hardware-induced distortions . Due to the complexity of state-of-the-art neuromorphic platforms and their control software, as well as the vast landscape of emulable neural network models, a thorough and systematic approach is essential for providing reliable information about causal mechanisms and functional effects of hardware-induced distortions in model dynamics and for ultimately designing effective compensation methods. In this article, we design and perform such a systematic analysis and compensation for several hardware-specific distortion mechanisms. First and foremost, we identify and quantify the most important sources of model distortions. We then proceed to investigate their effect on network functionality. In order to cover a wide range of possible network dynamics, we have chosen three very different cortical network models to serve as benchmarks. In particular, these models implement several prototypical cortical paradigms of computation, relying on winner-take-all structures (attractor networks), precise spike timing correlations (synfire chains) or balanced activity (self-sustained asynchronous irregular states). For every emulated model, we define a set of functionality criteria, based on specific aspects of the network dynamics. This set should be complex enough to capture the characteristic network behavior, from a microscopic (e.g., membrane potentials) to a mesoscopic level (e.g., firing rates) and, where suitable, computational performance at a specific task. Most importantly, these criteria need to be precisely quantified, in order to facilitate an accurate comparison between software simulations and hardware emulations or between different simulation/emulation back-ends in general. The chosen functionality criteria should also be measured, if applicable, for various relevant realizations (i.e. for different network sizes, numbers of functional units etc.) of the considered network. Because multiple distortion mechanisms occur simultaneously in hardware emulations, it is often difficult, if not impossible, to understand the relationship between the observed effects (i.e., modifications in the network dynamics) and their potential underlying causes. Therefore, we investigate the effects of individual distortion mechanisms by implementing them, separately, in software simulations. As before, we perform these analyses over a wide range of network realizations, since - as we will show later - these may strongly influence the effects of the examined mechanisms. After having established the relationship between structural distortions caused by hardware-specific factors and their consequences for network dynamics, we demonstrate various compensation techniques in order to restore the original network behavior. In the final stage, for each of the studied models, we simulate an implementation on the hardware back-end by running an appropriately configured executable system specification, which includes the full panoply of hardware-specific distortion mechanisms. Using the proposed compensation techniques, we then attempt to deal with all these effects simultaneously. The results from these experiments are then compared to results from software simulations, thus allowing a comprehensive assertion of the effectivity of our proposed compensation techniques, as well as of the capabilities and limitations of the neuromorphic emulation device. 1.5 Article Structure In section Neuromorphic testbench and investigated distortion mechanisms, we describe our testbench neuromorphic modeling platform with its most relevant components, as well as the essential layers of the operation workflow. We continue by explaining the causes of various network-level distortions that are expected to be common for similar mixed-signal neuromorphic devices. In the same section, we also introduce the executable system specification of the hardware, which we later use for experimental investigations. Section Hardware-induced distortions and compensation strategies contains the description of the three benchmark models. We start the section on each of the models with a short summary of all the relevant findings. We then describe its architecture and characteristic aspects of its dynamics which we later use as quality controls. We continue by discussing the effects of individual hardware-specific distortion mechanisms as observed in software simulations, propose various compensation strategies and investigate their efficacy in restoring the functionality of the network model in question. Subsequently, we apply these methods to large-scale neuromorphic emulations and examine the results. Finally, we summarize and discuss our findings in the Conclusions section."
} | 4,796 |
30498977 | PMC6394480 | pmc | 2,022 | {
"abstract": "The development of robust microorganisms that can efficiently ferment both glucose and xylose represents one of the major challenges in achieving a cost-effective lignocellulosic bioethanol production. Candida intermedia is a non-conventional, xylose-utilizing yeast species with a high-capacity xylose transport system. The natural ability of C. intermedia to produce ethanol from xylose makes it attractive as a non-GMO alternative for lignocellulosic biomass conversion in biorefineries. We have evaluated the fermentation capacity and the tolerance to lignocellulose-derived inhibitors and the end product, ethanol, of the C. intermedia strain CBS 141442 isolated from steam-exploded wheat straw hydrolysate. In a mixed sugar fermentation medium, C. intermedia CBS 141442 co-fermented glucose and xylose, although with a preference for glucose over xylose. The strain was clearly more sensitive to inhibitors and ethanol when consuming xylose than glucose. C. intermedia CBS 141442 was also subjected to evolutionary engineering with the aim of increasing its tolerance to inhibitors and ethanol, and thus improving its fermentation capacity under harsh conditions. The resulting evolved population was able to ferment a 50% ( v / v ) steam-exploded wheat straw hydrolysate (which was completely inhibitory to the parental strain), improving the sugar consumption and the final ethanol concentration. The evolved population also exhibited a better tolerance to ethanol when growing in a xylose medium supplemented with 35.5 g/L ethanol. These results highlight the potential of C. intermedia CBS 141442 to become a robust yeast for the conversion of lignocellulose to ethanol. Electronic supplementary material The online version of this article (10.1007/s00253-018-9528-x) contains supplementary material, which is available to authorized users.",
"introduction": "Introduction Concerns about global warming and uncertainties in the future energy supply are important driving forces behind the development and implementation of a sustainable bio-based economy. Biorefineries will be used in the future to transform biomass into a wide range of products, including fuels and other energy forms, high value-added chemicals, and other materials, similar to current petroleum-based refineries (Olsson and Saddler 2013 ). Lignocellulosic biomass is the most abundant form of organic matter in nature (Sánchez and Cardona 2008 ) and is hence considered the foremost source of raw material for use in biorefineries. Among biomass-derived products, lignocellulosic bioethanol represents an important bulk chemical for the replacement of fossil fuels in the short to medium term. Although significant progress has recently been made towards the commercialization of lignocellulosic bioethanol, several bottlenecks must be resolved to achieve economic and sustainable conversion processes (Balan 2014 ; Gubicza et al. 2016 ). Lignocellulose is composed of three main polymers: cellulose, hemicelluloses, and lignin, linked by covalent and non-covalent bonds to form a complex matrix. Biochemical conversion of lignocellulose includes three major steps: (1) pretreatment, to open up the inherent recalcitrant structure, (2) enzymatic hydrolysis, for the saccharification of carbohydrate polymers, and (3) microbial fermentation, to transform the hydrolyzed sugars into useful products such as ethanol. The pretreatment step is usually performed under harsh conditions (such as high temperatures and pressures and/or low/high pH), which leads to extensive biomass degradation, generating various lignocellulose-derived compounds such as aliphatic acids, furan derivatives, and several phenolic compounds (Alvira et al. 2010 ). These compounds inhibit hydrolytic enzymes as well as fermentative microorganisms, thereby decreasing both the overall conversion yields and the microbial fermentation capacity (Taherzadeh and Karimi 2011 ; Jönsson et al. 2013 ). For instance, the furan derivatives furfural and 5-hydroxymethylfurfural have a direct impact on fermentative cells by inhibiting glycolytic and/or fermentative enzymes, thus reducing biomass formation and ethanol yields (Horváth et al. 2003 ). Furthermore, inhibitory compounds have been shown to exert a synergistic action on fermentative microorganisms, which can inhibit their growth, even at low concentrations (Oliva et al. 2004 ; Alvira et al. 2013 ). Another relevant stress factor affecting fermentation is the end product, ethanol, as it impedes cell growth and reduces viability, thus limiting the fermentation productivity and final conversion yields (Stanley et al. 2010 ). Most fermentative microorganisms are capable of removing and/or tolerating inhibitory compounds (Ask et al. 2013 ; Lindberg et al. 2013 ; Adeboye et al. 2017 ). This natural robustness can be improved by subjecting the microorganism to evolutionary engineering (Sauer 2001 ), by exposing the cells to sublethal inhibitory concentrations of a specific stressor to promote enrichment of cells with a selective advantage at the expense of the initially dominant cells. The advantages of using evolutionary engineering to increase microbial robustness are (1) no detailed knowledge of the microorganism is necessary, (2) resistance to multiple stress factors may be promoted at the same time, and (3) no external genetic modifications are introduced, thus preserving the non-GMO classification of the microorganism (Sauer 2001 ; Koppram et al. 2012 ). Saccharomyces cerevisiae is currently the most commonly used fermentative microorganism in the starch-based bioethanol industry due to its superior fermentation capacity of hexose sugars, particularly glucose. Moreover, in comparison with most other microorganisms characterized to date, S. cerevisiae exhibits a high tolerance to ethanol as well as lignocellulose-derived inhibitors (Piskur et al. 2006 ; Stanley et al. 2010 ; Parawira and Tekere 2011 ; Koppram et al. 2014 ). However, the major disadvantage of using S. cerevisiae strains to produce bioethanol from lignocellulosic materials is its inability to ferment pentoses such as D-xylose and L-arabinose (Sun and Cheng 2002 ; Hahn-Hägerdal et al. 2007 ). As xylose is the second most prevalent sugar monomer after glucose in lignocellulosic hydrolysates, and hence a highly important substrate, extensive research efforts have been made to introduce heterologous genes for xylose metabolism into S. cerevisiae (Moyses et al. 2016 ). These metabolic engineering approaches are often followed by evolutionary engineering and/or inverse metabolic engineering to optimize the xylose uptake and fermentation capacity. Although considerable progress has been achieved, engineered strains still suffer from inefficient xylose uptake and sequential fermentation of glucose and xylose (Subtil and Boles 2012 ). Furthermore, inefficient cofactor recycling during the catalysis of the NADPH-preferring xylose reductase and the NAD + -dependent xylitol dehydrogenase enzymes results in the accumulation of xylitol as a by-product, thus reducing the overall yield of ethanol on xylose (Jeffries and Jin 2004 ). Native xylose-fermenting yeasts, including species of the genera Scheffersomyces ( Scheffersomyces stipitis and Scheffersomyces shehatae ), Spathaspora ( Spathaspora passalidarum ), Pachysolen ( Pachysolen tannophilus ), and Candida ( Candida tropicalis and Candida tenuis ), can be considered as alternatives to these genetically modified S. cerevisiae strains (Sánchez et al. 2002 ; Gárdonyi et al. 2003 ; Su et al. 2015 ). The yeast Candida intermedia is also an interesting pentose-fermenting microorganism since it exhibits similar specific growth rates in glucose and xylose (Gárdonyi et al. 2003 ), expresses potent xylose transporters (Leandro et al. 2006 ), and has been shown to ferment glucose and xylose at high concentrations (Saito et al. 2017 ). Furthermore, it harbors multiple isoforms of xylose reductases, one of which has dual cofactor specificity, which may contribute to a better redox balance (Nidetzky et al. 2003 ). A new strain of C. intermedia , CBS 141442, was recently isolated in our lab from the liquid fraction of a steam-pretreated wheat straw hydrolysate (Moreno et al. 2017 ). The aim of this study was to determine the capacity of CBS 141442 to ferment glucose and xylose and its tolerance to lignocellulose-derived inhibitors and ethanol. Moreover, the strain was subjected to evolutionary engineering to improve its robustness and its ability to ferment xylose under limiting conditions.",
"discussion": "Discussion In order to achieve cost-effective lignocellulosic bioethanol production, the fermenting microorganisms must convert glucose and xylose efficiently into ethanol in a highly challenging environment. Despite intensive efforts to develop xylose-fermenting strains of S. cerevisiae , the xylose uptake and the fermentation performance of this yeast have still not reached satisfactory levels (Moyses et al. 2016 ). The continuous study of native xylose-fermenting yeasts is therefore of the utmost importance to improve our understanding of xylose conversion and to identify key metabolic steps. In this study, we have characterized the fermentation performance of the in-house isolated C. intermedia strain CBS 141442. Simultaneous co-fermentation of glucose and xylose was observed, although xylose was consumed at a lower rate than glucose. Glucose/xylose co-fermentation is a highly desirable trait in simultaneous saccharification and fermentation of lignocellulosic feedstocks, since this process configuration is characterized by high initial xylose concentrations (due to the solubilization of hemicelluloses during the pretreatment step), whereas glucose is continuously released during the fermentation process (Olsson et al. 2005 ). The preference for glucose over xylose in glucose/xylose mixtures has been observed for most yeast species studied, including S. cerevisiae and natural pentose-fermenting yeasts such as S. stipitis , S. shehatae , and also other C. intermedia strains (Panchal et al. 1988 ; Saito et al. 2017 ). In many cases, this is due to a repression mechanism that impedes the utilization of other carbon sources while glucose is available (Gancedo 1998 ). Gárdonyi et al. ( 2003 ) reported that C. intermedia exhibited strong inhibition of xylose utilization in the presence of glucose. Nonetheless, the results presented in the present paper suggest that C. intermedia CBS 141442 can assimilate xylose in the presence of glucose, which is a trait worth further exploration. Moreover, microbial robustness to lignocellulosic inhibitors is of crucial importance for lignocellulosic bioethanol production. C. intermedia CBS 141442 proved to be more sensitive to lignocellulose-derived inhibitors in the xylose-fermenting phase than in the glucose-fermenting phase. Moreno et al. ( 2013 ) have also reported limited xylose consumption by the xylose-recombinant yeast S. cerevisiae strain F12 after glucose depletion. In their study, the low xylose consumption was found to be associated with a decrease in cell viability due to the stress exerted by the inhibitors on the yeast during fermentation. Similarly, C. intermedia CBS 141442 might suffer from a loss of cell viability during fermentation of 40% and 50% ( v / v ) hydrolysate due to exposure to high concentrations of lignocellulose-derived compounds. In spite of having inherent mechanisms for tolerating/converting some of the inhibitory compounds such as furfural and 5-HMF (these compounds can be converted into less inhibitory products by the fermentative microorganisms through oxidation and/or reduction processes), at a certain concentration the synergistic action of the cocktail of inhibitors in a hydrolysate becomes sufficiently high to stop fermentation completely. At the selected fermentation pH range of 5–5.5, acetic acid (pKa = 4.75) and formic acid (pKa = 3.75) present in the hydrolysate are predominantly in their non-protonated forms and thus expected not to dominate the toxicity effects on cells. However, steam-exploded wheat straw hydrolysates usually contain also phenolics and other compounds, not evaluated in this study, that might play significant roles and promote synergistic effects during the inhibition of C. intermedia . Resulting ethanol concentrations must be sufficiently high (above 40 g/L) for the biomass conversion process to be economically viable. Ethanol concentrations of 23.7 g/L and 31.6 g/L severely affected the growth of C. intermedia CBS141442 in media containing glucose and xylose, while concentrations of 43–53 g/L were predicted to be completely inhibitory. Ethanol concentrations in the range of 5–7% ( v / v ) have also been reported to be completely inhibitory to species belonging to the Hanseniaspora , Candida , Pichia , Kluyveromyces , Metschnikowia , Torulaspora , and Issatchenkia genera (Ciani et al. 2016 ). In the case of S. stipitis , which is considered to be one of the best natural xylose-fermenting microbes, the maximum ethanol concentrations at which cells could not grow were estimated to be 33.6 g/L (glucose) and 44.7 g/L (xylose) (Lee et al. 2000 ), which are about 20% lower than those observed in the present study for C. intermedia CBS 141442 (Fig. 3 c, d). According to the results of the growth and fermentation tests performed, we can conclude that C. intermedia CBS 141442 has a modest tolerance to lignocellulose-derived compounds and ethanol, especially when utilizing xylose. Therefore, xylose conversion by this yeast strain might be improved by increasing its robustness to these compounds. Evolutionary engineering has proved to be a highly suitable method for enhancing traits in various microorganisms, including non-conventional yeast species for which we lack a detailed knowledge about their physiology as well as tools for targeted genome editing. Moreover, evolutionary engineering makes it possible to target complex polygenic phenotypes, such as tolerance to the cocktail of inhibitors present in lignocellulosic hydrolysates, without prior knowledge about the genes responsible for the trait (Steensels et al. 2014 ). For example, Nigam ( 2001 ) reported overcoming a 65-h lag phase in S. stipitis during fermentation of 60% ( v / v ) red oak acid hydrolysate through adaptation of cells on hardwood hemicellulose acid prehydrolysate, and the acetic acid tolerance of a S. passalidarum yeast strain was improved by an evolutionary engineering strategy based on UV-mutagenesis and subsequent selection by continuous cultivation (Morales et al. 2017 ). Similarly, the evolutionary engineering strategy employed in the present study resulted in the final evolved population C. intermedia EVO 2, which showed improved xylose-to-ethanol conversion in highly challenging environments. Here, the fermentation of 50% ( v / v ) hydrolysate, which was completely inhibitory to the parental stain, showed similar final ethanol concentrations and xylose consumption rates to those observed at lower inhibitor concentrations (30–40% ( v / v ) hydrolysate), independently of the evolved population used. This result suggests that stable genetic changes are responsible for the corresponding evolved phenotype, since population EVO 2 was obtained in absence of lignocellulose-derived inhibitors. Furthermore, the EVO 2 population showed improved ethanol tolerance when growing in xylose medium with an ethanol concentration of 35.5 g/L, a concentration that caused complete growth inhibition of CBS 141442 after 24 h. Although we cannot completely rule out the possibility that the increased ethanol tolerance is influenced by the rich medium used in the experimental setup, these results highlight the benefit of sequentially applying lignocellulose-derived compounds and ethanol as selective pressures during evolutionary engineering to produce strains with enhanced robustness to these inhibitory compounds. Future studies involve determining the underlying genetic changes responsible for the improved phenotypes, which can yield valuable insights into how the specific traits are established. The native xylose-fermenting yeast C. intermedia has several traits of interest for the conversion of lignocellulosic biomass into bio-based products such as ethanol. These include rapid xylose uptake (Leandro et al. 2006 ), xylose reductases for better redox balance (Nidetzky et al. 2003 ), and the simultaneous utilization of both glucose and xylose during fermentation, as shown here. Furthermore, cells with enhanced robustness to lignocellulosic-derived inhibitors and ethanol were obtained by subjecting C. intermedia CBS141442 to evolutionary engineering. We believe that the findings of this study may contribute to our understanding of important physiological/genetic traits which will help in the further development and implementation of optimal lignocellulosic biomass conversion processes."
} | 4,264 |
34094298 | PMC8162394 | pmc | 2,024 | {
"abstract": "Antifouling properties of materials play crucial roles in many important applications such as biomedical implants, marine antifouling coatings, biosensing, and membranes for separation. Poly(ethylene glycol) (or PEG) containing polymers and zwitterionic polymers have been shown to be excellent antifouling materials. It is believed that their outstanding antifouling activity comes from their strong surface hydration. On the other hand, it is difficult to develop underwater glues, although adhesives with strong adhesion in a dry environment are widely available. This is related to dehydration, which is important for adhesion for many cases while water is the enemy of adhesion. In this research, we applied sum frequency generation (SFG) vibrational spectroscopy to investigate buried interfaces between mussel adhesive plaques and a variety of materials including antifouling polymers and control samples, supplemented by studies on marine animal (mussel) behavior and adhesion measurements. It was found that PEG containing polymers and zwitterionic polymers have very strong surface hydration in an aqueous environment, which is the key for their excellent antifouling performance. Because of the strong surface hydration, mussels do not settle on these surfaces even after binding to the surfaces with rubber bands. For control samples, SFG results indicate that their surface hydration is much weaker, and therefore mussels can generate adhesives to displace water to cause dehydration at the interface. Because of the dehydration, mussels can foul on the surfaces of these control materials. Our experiments also showed that if mussels were forced to deposit adhesives onto the PEG containing polymers and zwitterionic polymers, interfacial dehydration did not occur. However, even with the strong interfacial hydration, strong adhesion between mussel adhesives and antifouling polymer surfaces was detected, showing that under certain circumstances, interfacial water could enhance the interfacial bio-adhesion.",
"conclusion": "Conclusion In this research, SFG has been applied to investigate the molecular structures of buried solid/liquid and solid/solid interfaces in situ in real time to understand antifouling and bio-adhesion, supplemented by the marine animal (mussel) activity study and adhesion measurements. It was found that both antifouling zwitterionic polymer SBMA and PEG containing polymer OEGMA have strong surface hydration in water. Because of the strong surface hydration, mussels do not want to settle on the SBMA or OEGMA surface to deposit adhesive plaque ( Fig. 5a ). Oppositely, on control sample surfaces, only weak surface hydration was detected. It is easy for mussels to settle on such surfaces, dehydrate the surfaces, and deposit adhesive plaques on these surfaces. Therefore surface hydration is important for antifouling activity of a material. Fig. 5 Schematics (not drawn to scale) showing proposed mechanisms of antifouling (a) and strong bio-adhesion (b) of mussels on SBMA or OEGMA surfaces. Because of the strong surface hydration, mussels do not want to stay on the SBMA or OEGMA surface, leading to the excellent antifouling performance of SBMA and OEGMA (a). If mussels are forced to stay on SBMA or OEGMA ( e.g. , tied with a rubber band), the interfacial water can strongly interact with both surface and mussel adhesive proteins, resulting in strong bio-adhesion (b). It is widely observed that bio-adhesion is weak in water. Our SFG studies show that dehydration occurs when mussels deposit adhesives on control sample surfaces, leading to strong adhesion, as expected. It has been shown in the literature that mussel adhesive proteins can replace interfacial water. 16,69 When interfacial dehydration occurred, the interfacial SFG water O–H stretching signal was replaced by the interfacial protein N–H stretching signal. Interestingly, it was found that for a small amount of mussels which deposited adhesives on SBMA or OEGMA, interfacial dehydration did not occur. Under such circumstances, strong adhesion was also measured between the mussel adhesive and SBMA or OEGMA surface, with the presence of interfacial water. We therefore concluded that under certain conditions, interfacial water can enhance the interfacial adhesion, with the help of interfacial proteins ( Fig. 5b ). The detailed mechanism on how water and mussel adhesive protein interact at the interface to enhance adhesion needs further investigation in the future. It is well known that cross-linking of 3,4-dihydroxylphenylalanine (DOPA) containing mussel adhesive proteins plays an important role in bio-adhesion of mussels. We believe that on SBMA and OEGMA, interfacial water molecules can be strongly hydrogen bonded with both cross-linked mussel adhesive proteins and polymer surfaces, leading to strong bio-adhesion. Certainly other related or different effects such as chelating or capillary forces may also play roles. 15,66,67,70 This study demonstrates the power of using SFG to probe interfacial hydration/dehydration, and the importance of surface hydration/dehydration for antifouling and bio-adhesion. The knowledge obtained from this study will help the design and development of antifouling materials with improved performance and polymer materials with desired bio-adhesion properties.",
"introduction": "Introduction Antifouling properties are important for many extensively used materials in various applications and devices such as biomedical implants, drug delivery vehicles, marine antifouling coatings, sensors, and membranes. 1–4 Extensive research demonstrated that zwitterionic polymers and polymers containing poly(ethylene glycol) components (referred to as “PEG materials” in this article) exhibit excellent antifouling activities. 5–9 It is believed that the strong surface hydration of such polymers mediates their excellent antifouling performance. 10–13 With strong surface hydration, it is difficult for biological media ( e.g. , protein molecules, bacteria, marine organisms, etc. ) to displace strongly bonded surface water molecules to stick to (or to foul on) the surfaces of zwitterionic polymers and PEG materials. In contrast, for non-antifouling materials, surface hydration is not as strong. 12,13 Then biological media can displace or repel surface water molecules to dehydrate the surface to stick to the surface. It is reasonable to believe that strong interfacial hydration leads to antifouling, while interfacial dehydration results in fouling. Similarly, almost all the man-made adhesives and glues work on a dry surface but fail to stick to a wet hydrated surface, because such adhesives cannot dehydrate the surface. The hard, dry adhesives of insects and geckos also stick well when dry and less well or not at all when substrates are underwater. 14,15 Different from insects and geckos, a trip to the beach will find mussels, barnacles, tube worms, sea stars, oysters, and many more animals bonding to rocks that are quite wet. Presumably, marine biology has devised clever ways clearing water from surfaces to dehydrate the surface in order to create adhesive–substrate bonds. 16–18 In addition to the understanding on the molecular mechanisms of antifouling, understanding on bio-adhesion in wet environments is equally important, which will greatly aid in the design and development of underwater glues for many important medical and other applications. We believe that bio-adhesion is also mediated by interfacial hydration/dehydration. However, it is difficult to test the above “beliefs” or “hypothesis” on the relationships between “interfacial hydration/dehydration” and “antifouling/fouling” experimentally because it is challenging to study buried solid/liquid interfaces in situ nondestructively to probe interfacial hydration or examine buried solid/solid interfaces to test interfacial dehydration. Traditional surface sensitive techniques require high vacuum to operate. It is therefore impossible to use them to test surface structures in a liquid environment. Recently, sum frequency generation (SFG) vibrational spectroscopy has been developed into a powerful tool to probe buried solid/liquid and solid/solid interfaces in situ in real time at the molecular level. 19–22 We have applied SFG to study the interfacial molecular behavior of proteins including adhesive proteins. 23–26 Here in order to elucidate the effects of interfacial hydration/dehydration on antifouling and bio-adhesion, we probed buried interfaces between mussel adhesive plaques and a variety of antifouling and fouling materials using SFG, in combination with mussel behavior study and adhesion measurements. Our results demonstrated that strong hydration is the key to achieve excellent antifouling performance of a polymer material, and dehydration leads to fouling. Surprisingly, we also found that hydration can enhance bio-adhesion under certain circumstances.",
"discussion": "Results and discussion Mussel behavior Mussels were placed atop a series of substrates ranging from high to low surface energies. The animals attached by depositing a flowing adhesive precursor that cures into a final plaque formation. 33 Control materials investigated in this study on which biofouling occurs included fused silica (water contact angle of ∼20°), PMMA (∼71°), aluminum (∼75°), and PS (∼87°). Also included in this research were the antifouling zwitterionic polymer (grafted SBMA, ∼20°) and PEG material (grafted OEGMA, 42°). 10 As discussed above, zwitterionic 34 and OEGMA 35 surfaces have gained a measure of fame for antifouling properties ( i.e. , preventing adhesion) within biomedical as well as marine shipping contexts. When placed onto surfaces, marine mussels attach by depositing their adhesive plaques ( Fig. 1a and b ). 36 Prior to settlement, the animals may move around in search of the most appealing location. Consequently, we used rubber bands to keep the animals in one place and force adhesion to a given substrate. For the aluminum ( Fig. 1c ), fused silica, PS, and PMMA substrates, mussels remained in place and attached with their plaques. The zwitterionic SBMA ( Fig. 1d ) and OEGMA surfaces, by contrast, resulted in different animal behavior. Most of the mussels were able to work their way off from these antifouling substrates. In order to understand the origin of this unusual action, we monitored the animals over several days. ESI Video 1 † shows that the mollusks opened and closed their shells repeatedly. In doing so, they were able to wiggle free from their constraints, but only when on the zwitterionic SBMA and OEGMA surfaces. Fig. 1 Mussel adhesive and animal behavior on different surfaces. (a) Mussels using an adhesive for attaching to each other and the side of an aquarium tank. (b) A mussel bonding to a sheet of polytetrafluoroethylene (Teflon), showing the adhesive plaques and threads. (c) Photograph showing mussels two days after binding to aluminum substrates. Note how all animals remain attached. Top down views of mussels on the substrates are provided in Fig. S1 (ESI † ). Each mussel and substrate is resting on a piece of plastic pipe to prevent the animal from placing its adhesive on the more stable aquarium bottom. (d). Mussels after two days on a zwitterionic SBMA surface. Most animals have moved off the substrates. The above observation of the mussel behavior demonstrated clearly that animals could easily settle on the four “control” or “fouling” surfaces including silica, PMMA, PS, and aluminum, but do not want to settle down on the two antifouling polymer SBMA and OEGMA surfaces. We believe that the drastically different animal behaviors on fouling and antifouling surfaces are caused by different surface hydration on these materials. Likely the control materials have weak surface hydration, while the antifouling SBMA and OEGMA have strong surface hydration, which will be investigated in more detail below. Surface hydration in water SFG spectra could be collected from the material/water interfaces in situ in real time to probe the molecular structure of a solid/liquid interface. Extensive research has been performed to investigate the interfacial water structure using SFG. 37–45 Typically two broad peaks centered around 3200 and 3400 cm −1 could be detected, sometimes along with a relatively narrower peak at around 3700 cm −1 . 37–45 It is agreed that the ∼3700 peak is due to the free OH stretching, while debates still exist for the assignment for the other two peaks. It is generally believed that the 3200 cm −1 and 3400 cm −1 peaks come from the strongly and weakly (or loosely) hydrogen bonded water molecules at the interface. We have collected SFG spectra from a variety of surfaces in water. 11,46–51 It was found in our previous research that stronger surface hydration leads to better antifouling behavior. 10–13,52,53 The strong hydration is characterized by a strong 3200 cm −1 peak, a weak (or absence of) 3400 cm −1 signal, and a strong overall water signal intensity. 10–13 For example, SFG spectra collected from the antifouling SBMA and OEGMA surfaces in water are both dominated by the ∼3200 cm −1 peak, showing that the surfaces are covered by strongly bonded water molecules, leading to excellent antifouling properties. 10 For a control PMMA surface, both 3200 cm −1 and 3400 cm −1 peaks were detected, indicating that the PMMA surface has a substantial amount of loosely bonded water molecules and thus PMMA is not a good antifouling material. 12 \n Fig. 2a shows a picture of a mussel adhesive plaque on a surface, while Fig. 2b shows the schematic of the SFG sample geometry. SFG spectra can be collected from the sample/water interface (position A in Fig. 2b ) and the sample/mussel adhesive plaque interface (position B in Fig. 2b ). Fig. 2 Mussel adhesive plaque and SFG sample geometry. (a) Picture of a mussel adhesive plaque on a surface. (b) Schematic showing the SFG sample geometry used to collect SFG spectra from the sample/water interface (position A) and the sample/mussel adhesive plaque interface (position B). The input laser beams penetrate the substrates (silica or silica with polymer thin films) to reach the sample/water and the sample/mussel adhesive interfaces. Aluminum was not studied because it is not transparent for the input laser beams. \n Fig. 3 shows SFG spectra collected from the surfaces and interfaces of five samples, while aluminum was not included owing to a lack of optical transparency. Note that the y -axes are on different scales to aid visualization of spectral features. We first discuss the SFG spectra collected from the material/water interface in Fig. 3 , shown in blue. It is worth mentioning that these SFG spectra were not collected from the “clean” material/water interface, instead, they were detected from the samples after the mussel deposition experiments, from the surface regions where there were no adhesives deposited ( e.g. , position A in Fig. 2b ). Fig. 3 SFG spectra of the substrate/water interface (region A in Fig. 2(b) , blue curves), substrate/mussel adhesive interface (region B in Fig. 2(b) , black curves), and the substrate/mussel adhesive interface after placing the sample in D 2 O for a while (red curves). Note that the y -axes differ between panels in order to better visualize differences in spectra. The SFG intensities in different panels are different because of the different water orientation and ordering at various interfaces due to different interfacial interactions. SFG spectra collected from mussel adhesive plaque/substrate interfaces for the different plaques on the same type of substrate are similar and the results are reproducible. (a) Fused silica. General water and amine peak assignments are shown in green. These same assignments apply to all other spectra shown. (b) PMMA. In the water spectrum, prior to adhesive deposition, a substrate methyl peak was observed. (c) PS. The adhesive spectrum also shows a phenyl peak from the substrate. (d) Zwitterionic SBMA. (e) OEGMA. The O–H stretching SFG spectra collected from three control samples, fused silica, PMMA, and PS in water show some common features (blue spectra in Fig. 3a–c ). In addition to the signals at around ∼3200 cm −1 , all three blue spectra have strong ∼3400 cm −1 signals, showing substantial coverage of loosely bonded water molecules on the surfaces, or weak hydration. This observation matches our previous observation on the “clean” PMMA/water interface well. 12 In contrast, SFG spectra collected from the SBMA and OEGMA surfaces are dominated by the ∼3200 cm −1 signal (blue spectra in Fig. 3d and e ), showing strong surface hydration, similar to those observed from the “clean” SBMA/water and OEGMA/water interfaces. 10 The surface hydration detected by SFG can be well correlated with the above animal behavior on various sample surfaces: On the weakly hydrated silica, PMMA, and PS surfaces, mussels easily settled and deposited adhesives. On the strongly hydrated SBMA and OEGMA surfaces, mussel tried to move away from the surfaces, even after tying with rubber bands. Our previous research shows that for SBMA, protein molecules cannot disrupt the surface hydration of SBMA at the SBMA/protein solution interface, 10 due to the strong SBMA surface hydration. On the OEGMA/protein solution interfaces, protein molecules could disrupt the interfacial water structure and loosely deposit onto the surface, which can be easily washed off by water. Both SBMA and OEGMA are antifouling materials. For fouling surfaces, protein molecules can disrupt the surface hydration and deposit to the surfaces permanently. Here instead of protein molecules, we used SFG to study interfacial hydration between various surfaces and mussel adhesives deposited by real live animals, the results of which will be presented below. Dehydration at interfaces between mussel adhesives and control surfaces To understand bio-adhesion and biofouling in detail, SFG spectra were collected from the buried interfaces between the three control samples (silica, PMMA, and PS) and mussel adhesives respectively (black spectra in Fig. 3a–c ). After mussel attachment and partial plaque coverage, the interfacial SFG spectra of all three buried solid/solid interfaces between the control samples and adhesives exhibit some common features: They all shifted from showing strong water signals (O–H stretching at ∼3200 cm −1 and ∼3450 cm −1 ) to those of amines from the protein-based adhesive (N–H stretching at ∼3280 cm −1 ). Spectral fitting results in Fig. S2 (ESI † ) showed that a single amine peak can describe the spectra well, indicating that no ordered water exists at the three control material/mussel adhesive interfaces, and likely dehydration occurred at these interfaces. SFG detects signals from a medium with no inversion symmetry. The absence of the SFG water signal at the control sample/mussel adhesive interfaces indicates that there is no ordered water at the interfaces. To confirm the interfacial dehydration and to exclude the existence of disordered water ( e.g. , water molecules adopting a random orientation) at the interfaces, the above three samples were placed in D 2 O and SFG spectra were then collected. The SFG spectral regions shown in Fig. 3 will not detect O-D and N-D stretching signals. If an interface is hydrated by disordered water molecules, then changing a sample from residing in H 2 O (black spectra) to D 2 O (red spectra) will diminish the interfacial N–H signals because of the H-D exchange between the N–H groups and D 2 O molecules accessing the interfaces. Placement of adhesive-fused silica into D 2 O ( Fig. 3a ) shrank the amine N–H signal with some intensity persisting. This change indicated that water (D 2 O here) could access some of the interface to exchange protons with proteins ( e.g. , N–H to N–D). The remaining N–H signal indicated that water could not access the entire interface with the rest of the system being dehydrated, without ordered or disordered water. For PMMA, the amine N–H peak for the adhesive/sample interface only decreased a little ( Fig. 3b ), showing that the buried PMMA/mussel adhesive interface is mostly dehydrated and cannot be accessed by D 2 O, preventing the H-D exchange for interfacial N–H groups. Further support for such a conclusion was obtained by having the animals deposit glue atop deuterated PMMA to yield similar results (Fig. S2, ESI † ). PS ( Fig. 3c ) was generally similar to PMMA – a small N–H decrease when in D 2 O, and analogous results when using a deuterated plastic surface (Fig. S2, ESI † ). The H–D exchange experiments using D 2 O clearly demonstrated the dehydration of the buried interfaces between the three control samples and mussel adhesives. Therefore we concluded that for biofouling to occur, mussels displaced the water on the sample surface and dehydrated the surface while depositing adhesives on the surface. These happened when a material surface does not have strong hydration. Hydration at interfaces between mussel adhesives and antifouling SBMA as well as OEGMA As presented above, most mussels left the SBMA or OEGMA surfaces without depositing adhesive plaques even when tied with rubber bands. Only a small amount of mussels stayed on the SBMA or OEGMA surfaces and deposited adhesives. SFG spectra were collected from the buried interfaces between these mussel adhesives and antifouling polymers, SBMA or OEGMA. For zwitterionic SBMA ( Fig. 3d ), plaque deposition decreased the SFG water signal, but only to a small extent. Furthermore, no prominent amine N–H stretching peak was observed. Quantitative spectral fitting results in Fig. S2 (ESI † ) attested to the absence of the amide N–H signal. A water signal at ∼3200 cm −1 indicated water that was hydrogen bonded strongly. These data showed hydrogen bonded interfacial water residing between the mussel adhesive and the surface. From the OEGMA spectra ( Fig. 3e ), it can be seen that plaque binding decreased the water signal to yield a continued presence of water with a weak N–H contribution (Fig. S2, ESI † ). Here, too, interfacial water appeared to persist between the adhesive and surface. Soaking the adhesive–SBMA sample in D 2 O eliminated almost all of the SFG signal (red spectrum in Fig. 3d ), indicating that the entire interface was accessible by water (or D 2 O). Furthermore, there appeared to be no dehydrated, ordered N–H functionalities at this adhesive–SBMA interface. A similar result was found for OEGMA ( Fig. 3e ). A weak N–H may have been present for adhesive–OEGMA (Fig. S2, Table S6, ESI † ). Nonetheless, the entire SBMA and OEGMA systems appeared to be water accessible, with no regions of dehydration. The interfacial hydration between mussel adhesives and control samples vs. antifouling materials is drastically different. At the mussel adhesive/control sample interfaces, dehydration occurred. At the antifouling polymer/mussel adhesive interfaces, dehydration was not observed. Adhesion testing To further understand antifouling and bio-adhesion, we measured the adhesion strengths of mussel adhesives on control surfaces and antifouling surfaces. The adhesive performance of mussel glue on each substrate was quantified using an established method ( Fig. 4 ). 36,54 Briefly, adhesion values (in kPa) were obtained by pulling each plaque perpendicular to the surface until failure and the maximum force was divided by the plaque contact area on the surface. Data in Fig. 4 and Table S1 † show that this adhesive functioned in a generally classic manner, with stronger binding to substrates with higher surface energies. Such behavior is in agreement with how adhesives stick better to, for example, metals than hydrophobic plastics. 55 However, mussel adhesion does not track perfectly with surface energy. Aluminum brings about particularly high strengths, possibly due to observations of changes in the surface ( e.g. , shiny to dull) after hours of residence in sea water. 36 Condition index tests ensured that animal health was not influenced by any substrate (Table S2 † ). 56,57 Fig. 4 Performance of mussel adhesives on different substrates. Average adhesion of mussel plaques on each surface. The substrates are ordered from the highest to lowest surface energies. Error bars shown are 99% confidence intervals. A particularly interesting finding here is that, in the rare instances when mussels attached to the OEGMA and zwitterionic SBMA substrates, adhesion strengths were quite substantial. This was different from what we expected. Since most adhesives and glues fail in wet environments, we thought that the interfacial hydration would lead to weak adhesion between mussel adhesives and SBMA or OEGMA. This was not observed and will be discussed in more detail below. Discussion on antifouling Our previous studies demonstrated that the surface hydration is well correlated with the antifouling activity of a material. 10–13 For control samples or materials which do not possess antifouling properties, surface hydration is weak, as evidenced by the observation of both 3200 cm −1 and 3400 cm −1 peaks. Mussels can easily settle on such surfaces, dehydrate the surfaces, and deposit adhesives. Interfacial dehydration was observed in situ using SFG between mussel adhesives and the three control samples. The dehydration process and the adhesive deposition process could occur at the same time: The adhesive proteins are deposited onto the surface to replace interfacial water molecules. Adhesive proteins could also bind water molecules initially, but when they came into contact with the control sample, they more favorably interacted with the control sample surface directly, squeezing the initially bound water from the protein and the adhesive/control material interface, leading to interfacial dehydration. For the two antifouling materials SBMA and OEGMA, strong surface hydration was observed, as evidenced by the dominating SFG signals at 3200 cm −1 . With strong surface hydration, mussels do not like to settle on these surfaces and try to escape even when bound by rubber bands. The established antifouling properties of these surfaces may derive primarily from mussels avoiding the hydrating surfaces. Once attached, however, bonding tracked well with the high surface energies, which is not related to the antifouling behavior of the material. Discussion on bio-adhesion How might we explain the attachment strength of mussels to substrates without consistently strong protein peaks and without completely dehydrated surfaces? The first column of Table 1 lists these five substrates in the order of decreasing surface energy or increasing water contact angles (column 2). After mussel attachment, the SFG spectra collected from the buried mussel adhesive/sample interfaces were fit to approximate relative intensities of water O–H versus protein amine N–H signals (columns 3 and 4). Measured adhesion values are in column 5. Quite striking is that, contrary to what one might expect, adhesion did not correlate with the detected presence of protein (N–H signal) on all surfaces. We can explain this underwater bonding with either water O–H or protein N–H signals. Such results lead to the conclusion that adhesion may be brought about either by the protein alone, which can dehydrate some substrates, 16–18 or a combination of disordered protein and ordered interfacial water. Such ordered interfacial water can be the initial hydration layer on the SBMA or OEGMA surface and/or the protein bound water at the interface (protein bound water in the adhesive bulk should not contribute the SFG signal due to inversion symmetry). Water appears to help adhesion on some surfaces. Examination of substrates as functions of water contact angles, SFG signal intensities, and mussel adhesion. Arbitrary units are denoted by “a.u.” Surface Water contact angle (°) Water intensity with plaque (a.u.) N–H intensity with plaque (a.u.) Adhesion (kPa) Fused silica 20 0 2.2 ± 0.2 115 ± 17 SBMA 20 1.2 ± 0.2 0 99 ± 23 OEGMA 42 1.1 ± 0.1 0 90 ± 21 PMMA 71 0 1.4 ± 0.1 52 ± 11 PS 87 0 1.5 ± 0.2 75 ± 14 Speaking generally, water inhibits most adhesion phenomena. Water bound to surfaces prevents proteins or polymers from generating strong contacts. Cohesive forces can also be eliminated by hydration, thus “hiding” proteins or polymers from the counterparts and minimizing interactions. Reducing the amount of water present at surfaces appears to be a logical means of generating adhesive contacts. However, the practical aspects of such drying are not easy to achieve. Placement of hydrophobic species such as oils atop submerged high energy substrates may be able to remove bulk hydration, but water remains attached at the surface. 58–60 Our current understanding of the importance of removing water to gain adhesive function may be in flux. Having a hydrophobic polymer appears to be helpful, if not essential, for displacing water and achieving strong adhesion between submerged substrates. 61 One of the highest performing underwater adhesives is made from an all hydrophobic polymer host with pendant catechol groups. 62 Yet some data indicate that increasing the polymer dielectric constant aids underwater bonding. 63 These results may imply that hydrophilicity is actually beneficial for underwater bonding. Results found here indicate that viewing surfaces to be only wet and bad at adhesion versus dry and good for bonding is likely too simplistic of a view. The hard, dry adhesives of insects and geckos provide a potential analogy to explain how interfacial water might contribute to the bonding of a wet, flowing adhesive. These terrestrial organisms attach temporarily to surfaces, via van der Waals forces, using hard and structured materials often shaped into nanoscale hair 64 or pad 65 structures. In general, they stick well when dry and less well or not at all when substrates are underwater. 14,15 However, when in between dry and submerged, high humidity environments can actually increase the bonding of hard bioadhesives. 66,67 Just the right amount of water fills gaps between the hard adhesive and substrate, contributing capillary forces and strengthening bonding. 15 At an even smaller scale, nanoconfined water between the surface and adsorbate can be structured, no longer behaving in the way that we usually think of bulk solvent and possibly illustrating how surface hydration can be beneficial. 68 With data shown here, we may now be seeing the first known instance of interfacial hydration contributing to the underwater bonding of a curing bio-adhesive.\n\nDiscussion on antifouling Our previous studies demonstrated that the surface hydration is well correlated with the antifouling activity of a material. 10–13 For control samples or materials which do not possess antifouling properties, surface hydration is weak, as evidenced by the observation of both 3200 cm −1 and 3400 cm −1 peaks. Mussels can easily settle on such surfaces, dehydrate the surfaces, and deposit adhesives. Interfacial dehydration was observed in situ using SFG between mussel adhesives and the three control samples. The dehydration process and the adhesive deposition process could occur at the same time: The adhesive proteins are deposited onto the surface to replace interfacial water molecules. Adhesive proteins could also bind water molecules initially, but when they came into contact with the control sample, they more favorably interacted with the control sample surface directly, squeezing the initially bound water from the protein and the adhesive/control material interface, leading to interfacial dehydration. For the two antifouling materials SBMA and OEGMA, strong surface hydration was observed, as evidenced by the dominating SFG signals at 3200 cm −1 . With strong surface hydration, mussels do not like to settle on these surfaces and try to escape even when bound by rubber bands. The established antifouling properties of these surfaces may derive primarily from mussels avoiding the hydrating surfaces. Once attached, however, bonding tracked well with the high surface energies, which is not related to the antifouling behavior of the material.\n\nDiscussion on bio-adhesion How might we explain the attachment strength of mussels to substrates without consistently strong protein peaks and without completely dehydrated surfaces? The first column of Table 1 lists these five substrates in the order of decreasing surface energy or increasing water contact angles (column 2). After mussel attachment, the SFG spectra collected from the buried mussel adhesive/sample interfaces were fit to approximate relative intensities of water O–H versus protein amine N–H signals (columns 3 and 4). Measured adhesion values are in column 5. Quite striking is that, contrary to what one might expect, adhesion did not correlate with the detected presence of protein (N–H signal) on all surfaces. We can explain this underwater bonding with either water O–H or protein N–H signals. Such results lead to the conclusion that adhesion may be brought about either by the protein alone, which can dehydrate some substrates, 16–18 or a combination of disordered protein and ordered interfacial water. Such ordered interfacial water can be the initial hydration layer on the SBMA or OEGMA surface and/or the protein bound water at the interface (protein bound water in the adhesive bulk should not contribute the SFG signal due to inversion symmetry). Water appears to help adhesion on some surfaces. Examination of substrates as functions of water contact angles, SFG signal intensities, and mussel adhesion. Arbitrary units are denoted by “a.u.” Surface Water contact angle (°) Water intensity with plaque (a.u.) N–H intensity with plaque (a.u.) Adhesion (kPa) Fused silica 20 0 2.2 ± 0.2 115 ± 17 SBMA 20 1.2 ± 0.2 0 99 ± 23 OEGMA 42 1.1 ± 0.1 0 90 ± 21 PMMA 71 0 1.4 ± 0.1 52 ± 11 PS 87 0 1.5 ± 0.2 75 ± 14 Speaking generally, water inhibits most adhesion phenomena. Water bound to surfaces prevents proteins or polymers from generating strong contacts. Cohesive forces can also be eliminated by hydration, thus “hiding” proteins or polymers from the counterparts and minimizing interactions. Reducing the amount of water present at surfaces appears to be a logical means of generating adhesive contacts. However, the practical aspects of such drying are not easy to achieve. Placement of hydrophobic species such as oils atop submerged high energy substrates may be able to remove bulk hydration, but water remains attached at the surface. 58–60 Our current understanding of the importance of removing water to gain adhesive function may be in flux. Having a hydrophobic polymer appears to be helpful, if not essential, for displacing water and achieving strong adhesion between submerged substrates. 61 One of the highest performing underwater adhesives is made from an all hydrophobic polymer host with pendant catechol groups. 62 Yet some data indicate that increasing the polymer dielectric constant aids underwater bonding. 63 These results may imply that hydrophilicity is actually beneficial for underwater bonding. Results found here indicate that viewing surfaces to be only wet and bad at adhesion versus dry and good for bonding is likely too simplistic of a view. The hard, dry adhesives of insects and geckos provide a potential analogy to explain how interfacial water might contribute to the bonding of a wet, flowing adhesive. These terrestrial organisms attach temporarily to surfaces, via van der Waals forces, using hard and structured materials often shaped into nanoscale hair 64 or pad 65 structures. In general, they stick well when dry and less well or not at all when substrates are underwater. 14,15 However, when in between dry and submerged, high humidity environments can actually increase the bonding of hard bioadhesives. 66,67 Just the right amount of water fills gaps between the hard adhesive and substrate, contributing capillary forces and strengthening bonding. 15 At an even smaller scale, nanoconfined water between the surface and adsorbate can be structured, no longer behaving in the way that we usually think of bulk solvent and possibly illustrating how surface hydration can be beneficial. 68 With data shown here, we may now be seeing the first known instance of interfacial hydration contributing to the underwater bonding of a curing bio-adhesive."
} | 9,139 |
29973728 | null | s2 | 2,025 | {
"abstract": "Change history: In this Letter, Alexander Groisman should have been listed as an author. This error has been corrected online."
} | 31 |
28243071 | PMC5317333 | pmc | 2,027 | {
"abstract": "Durable superhydrophobic coatings were synthesized using a system of silica nanoparticles (NPs) to provide nanoscale roughness, fluorosilane to give hydrophobic chemistry, and three different polymer binders: urethane acrylate, ethyl 2-cyanoacrylate, and epoxy. Coatings composed of different binders incorporating NPs in various concentrations exhibited different superhydrophobic attributes when applied on polycarbonate (PC) and glass substrates and as a function of coating composition. It was found that the substrate surface characteristics and wettability affected the superhydrophobic characteristics of the coatings. Interfacial tension and spreading coefficient parameters (thermodynamics) of the coating components were used to predict the localization of the NPs for the different binders’ concentrations. The thermodynamic analysis of the NPs localization was in good agreement with the experimental observations. On the basis of the thermodynamic analysis and the experimental scanning electron microscopy, X-ray photoelectron spectroscopy, profilometry, and atomic force microscopy results, it was concluded that localization of the NPs on the surface was critical to provide the necessary roughness and resulting superhydrophobicity. The durability evaluated by tape testing of the epoxy formulations was the best on both glass and PC. Several coating compositions retained their superhydrophobicity after the tape test. In summary, it was concluded that thermodynamic analysis is a powerful tool to predict the roughness of the coating due to the location of NPs on the surface, and hence can be used in the design of superhydrophobic coatings.",
"conclusion": "Conclusion To obtain the necessary surface morphology for superhydrophobicity (exposed NPs) combined with durability (embedded NPs), a balance should be achieved between wetting of the NPs by the binder and migration of the NPs to the surface. Thus, the use of thermodynamic analysis as an analytical tool to predict the localization of NPs is essential for the selection of binders, NPs, and substrates for different systems. Three different binders, ECA, epoxy, and UA, and two substrates, PC and glass, were included in the study. The interfacial tensions and spreading coefficients of the as prepared solutions and the evaporation during spin coating were calculated to predict the localization of the NPs. The thermodynamic investigation was in good agreement with the SEM analysis with respect to the localization of the NPs as function of the binder concentration and substrate type. ECA showed a high negative spreading coefficient value with NPs resulting in localization of the NPs on the surface. Epoxy and UA showed moderately negative spreading coefficient values that can be overcome by heat during curing. In all three systems, with 5 and 10 binder wt%, although the spreading coefficient is positive, the NPs were on the surface due to the lack of sufficient binder to wet the NPs. While the ECA formulations exhibited superhydrophobicity on glass from 5 to 25 binder wt%, they showed superhydrophobicity on PC only from 5 to 10 binder wt%. Epoxy exhibited superhydrophobicity on glass from 5 to 10 binder wt%, but did not form superhydrophobic structures on PC substrates. The UA formulation showed superhydrophobicity at 5 and 10 binder wt% for both glass and PC. In general, the glass substrates provided superhydrophobicity over a wider range of compositions compared to PC. SEM analysis revealed the presence of nanoroughness from the presence of NP localization on the surface, which is critical for the creation of superhydrophobic surfaces. In all cases, the 5 and 10 binder wt% showed superhydrophobicity from insufficient binder to wet the NPs. Some of the coatings retained their superhydrophobic behavior after durability testing demonstrating the potential of preparing durable superhydrophobic coatings for a variety of applications.",
"introduction": "Introduction Superhydrophobic surfaces have been studied for a variety of applications, such as self-cleaning, 1 , 2 anticorrosion, 3 antipollution, 4 oil/water separation, 5 , 6 self-healing, 7 anti-fouling, 8 and ice repellant surfaces. 7 , 9 – 12 Superhydrophobic surfaces have static contact angles >150° and sliding angles <10°, allowing easy rolling of water droplets along the surface. The two key parameters for superhydrophobicity include hydrophobic chemistry and a nano/micro-hierarchical surface structure. Hydrophobic chemistry can be imparted by different low surface energy materials, such as fluorinated and long alkyl chain compounds. 13 A self-assembled monolayer of alkyltrichlorosilane (ATS) was shown to modify the surface energy of W 18 O 49 nanowire substrates by changing the carbon length of ATS. Results showed an increase in contact angle with increasing the number of carbons in ATS. 14 The wettability of a zinc oxide nanowire array was monotonically changed from hydrophilic to hydrophobic by increasing the carbon chain length of a chemisorbed fatty acid (FA). 15 A self-assembling film of fatty acids with different carbon chain lengths on hydroxyapatite discs surfaces showed that contact angles increased significantly with increases in the FA carbon chain length. 16 Additionally, a hierarchical structure composed of nanoscale texture on microscale texture is needed to obtain superhydrophobic characteristics. The rough structure can support a liquid droplet, in the Cassie–Baxter state, and display high contact angles as air is trapped between the rough asperities. Such roughness may also provide lower sliding angles as the small contact area leads to reduced drop pinning (lower resistance to droplet movement). Roughness can reduce the apparent surface free energy (SFE) of the superhydrophobic surface due to the reduced contact area between the liquid drop and the surface. 17 This roughness can enhance the ability of superhydrophobic surfaces to repel low surface tension liquids, such as oils. In order to prepare superhydrophobic surfaces, a variety of materials and techniques have been used including top-down methods, such as lithography 18 – 21 and plasma etching, 22 and bottom-up methods using different nanoparticles (NPs), such as silica, 23 – 25 silica/polyhedral oligomeric silsesquioxane, 26 functionalized aluminum oxide NPs, 27 , 28 titania, 29 , 30 carbon nanotubes, 31 – 34 ZnO particles, 8 , 35 – 37 and silanized calcium carbonate. 38 – 41 Regardless of the technique or the substrates used, each method provided micro/nanometer features for roughness and a low surface energy chemistry. Although superhydrophobic surfaces have been studied previously, their applications are limited due to their high production costs and low abrasion resistance. 26 , 30 A particularly attractive approach is a coating, which can be applied to a wide variety of surfaces. It should be mentioned that, in many studies, the coatings are limited to specific substrates due to the specific composition and/or preparation techniques involved. 42 – 48 The interactions between the surface/substrate and the coating as well as between the binder/NPs affect the structure of the coating and any resultant hierarchical structure needed to create superhydrophobicity in addition to the coatings’ durability. Thus, a fundamental study of these parameters is necessary to understand the effect of the substrate and coating properties for preparing new coatings on a wide array of substrates in the future. Another key attribute is durability, such that the surface retains its superhydrophobic characteristics, despite abrasion or wear. A previous study 22 indicated that a stable superhydrophobic urethane acrylate (UA) coating was obtained using silica NPs of various diameters grafted with photoreactive benzophenone groups and methylated fumed silica NPs with a fluorosilane top layer. The coatings showed good abrasion resistance under air flow and good durability in accelerated weathering conditions. Consequently, this work is aimed at investigating the thermodynamics of the system based on the interfacial tension and spreading coefficient parameters. Hence, the effects of binder type, binder concentration, and substrate wetting properties on the coating properties and morphology were studied. Two substrates, polycarbonate (PC; hydrophobic) and glass (hydrophilic), coated with a variety of formulations incorporating hydrophobic silica NPs dispersed in three different binders, ethyl cyanoacrylate (ECA), epoxy, and urethane acrylate (UA), were studied for the formation of superhydrophobic coatings having good durability. In the present case where low viscosity binders have been used, the mixing and coating processes are thermodynamically driven and kinetics do not play a role in the thermodynamics that lead to localization of the NPs.",
"discussion": "Results and discussion Thermodynamic analysis For this study, hydrophobic silica NPs and three low-molecular-weight binders ECA, epoxy, and UA were used. As the presence of the NPs on the surface is critical to obtain nanoscale roughness and thus superhydrophobic surfaces, the thermodynamics in the liquid state of the different formulations was investigated to explain the localization of the NPs in different binder formulations. In an ideal emulsion having low molecular weight and hence low viscosity, the localization at equilibrium of the silica particles is governed by thermodynamics, as kinetic effects do not play a significant role. 51 , 52 Thus, in this system (dispersed NPs in diluted binders) the localization of the NPs will be thermodynamically driven to its lowest energy state. Consequently, provided that an interaction between the dispersed NPs in diluted binder will reduce the total free energy of the system, better spreading will be achieved and favor the location of the NPs in the binder phase. If the interaction between the dispersed NPs in diluted binder will increase the total free energy of the system, spreading will not occur and will favor the location of the NPs on the binder surface. To investigate the localization of the NPs as a result of spreading considerations, the surface tensions of the neat and diluted binders (uncured) were measured using the pendant drop method and are shown in Tables 1 and S1 . In the case of neat ECA, the surface tension was difficult to measure due to its rapid polymerization in humidity, and thus, was assumed to be 26.67 mN/m using the trend line extrapolation ( Table S1 ). The results in Table S1 indicate that the surface tensions of the diluted binders are lower compared to the neat ones. No significant change in surface tension, however, was observed with increasing acetone concentration from 50 to 90 wt%. Addition of the dispersed NPs into the diluted binders is a critical step to insure good mixing and compatibility between the NPs and the binder. If the addition is thermodynamically favored, then NPs will be mixed spontaneously; however, if mixing is not thermodynamically favored, then NPs will not distribute well and remain on the surface of the binder. In this system (containing a binder polymer, NPs, and two solvents), the interfacial tension (γ AB ) between the dissolved or neat binders (γ A ) and the dispersed NPs (γ B ) was calculated and compared using both Antonoff’s rule (immiscible system) and the Girifalco and Good method (miscible system with dispersion attraction). The results are shown in Table 2 . The average surface tension for the dispersed NPs in IPA was 21.45 mN/m using the pendant drop measurement. Both Antonoff’s rule and the Girifalco and Good method showed a decrease in interfacial tension (IFT) for the diluted binder as compared to the neat binder. However, the Girifalco and Good method exhibited lower interfacial tension values compared to Antonoff’s rule, probably because acetone and IPA are miscible with each other. Comparison of the interfacial tension between the different binders showed that UA system exhibited the lowest interfacial tension values, whereas epoxy possessed the highest interfacial tension values. The interfacial tension values, however, do not reflect the full energy balance of the system. If the system (binder and solvent) is at a high free energy state (high surface tension) and the addition of a second component will decrease the total free energy of the system, it will favor spontaneous spreading. In contrast, if the system (binder and solvent) is at low free energy state (low surface tension), and the addition of second component will increase the total free energy of the system, the addition of the second component is not favored. To quantify the spreading phenomenon, the spreading coefficient parameters were determined. Accordingly, if the surface tension of component A is greater than the sum of surface tension of component B and interfacial tension AB, then the spreading coefficient is positive and the total free energy will be reduced, resulting in spontaneous spreading. If the surface tension of component A is lower than the sum of surface tension of component B and interfacial tension AB, then spreading is negative since the total free energy of the system is increased and partial or no spreading will occur. A zero spreading coefficient means that the total free energy of the system before and after adding component B has not changed and spreading will occur provided continuous mixing is used. As mentioned earlier, since complete wetting between the dispersed NPs and the diluted binders is important, the spreading coefficient during the mixing stage should be positive to allow spontaneous spreading and good mixing of the system. The spreading coefficient using Antonoff’s rule was zero for NPs containing formulations, which indicated that these systems are not immiscible. Thus, the calculated spreading coefficients were based on the Girifalco and Good method and are shown in Table 3 . All formulations showed positive spreading coefficients; epoxy, however, showed significantly higher values compared to UA and ECA. The spreading coefficient decreased with increasing acetone (binder solvent) wt%. The more positive the spreading coefficient value the greater the tendency for spontaneous spreading of NPs dispersed in IPA will occur in the binder. This behavior would suggest good wetting between the binder solution and the dispersed NPs and favor the localization of the NPs in the binder phase, which is an important stage before the coating process. Partial or no spreading in the mixing stage can affect the NPs localization in the final coating process. Finally, the thermodynamics of the coating step was studied. During the coating process, the diluted formulations were spin coated on the substrate while acetone (binder solvent) and IPA (NPs carrier solvent) flowed to the surface and evaporated. As acetone has a higher evaporation rate compared to IPA (5.6 and 1.5, 53 respectively), it will evaporate first and IPA will complete evaporation last. This dynamic process is critical for determining the final localization of the NPs. During IPA evaporation the specific binder can either spread on the NPs or not. If the spreading will result in a reduction of the total free energy, then the spreading coefficient will be positive, and NPs are expected to be located within the binder phase. If the spreading, however, results in an increase in the total free energy, then the spreading coefficient will be negative, and NPs are expected to be located on the surface. In order to calculate the interfacial tension between the binders and the solid NPs a special experimental step was used. Accordingly, glass slides were coated with the dispersed NPs in IPA and heated for 1 hour to ensure complete evaporation of IPA. Then, contact angles were measured between the NP-coated surface (γ s ) and neat binders (γ L ). The interfacial tension was calculated using Young’s equation. The results are summarized in Table 4 . The SFE of the solid NPs was taken as 22.4 mN/m (discussed later). The results in Table 4 demonstrate that all the binders have negative spreading coefficients. The spreading coefficient between ECA and NPs was extremely low and indicates that the NP will be localized on the surface for the ECA/NP system. This results in the formation of superhydrophobicity if the “right” roughness is obtained. For UA and epoxy the spreading coefficients were moderately negative however could reach zero if energy, such as heat (for curing), is supplied. These results predict the NPs localization inside the binder and loss of the nanoroughness, which is needed for superhydrophobicity. Hence, we would expect that only the ECA system could produce a superhydrophobic coating. It should be emphasized that the binder loading, in addition to the wetting of the substrate, will affect the final coating behavior and ultimate superhydrophobicity due to the development of the specific surface morphology. To study the effect of binder loading, different binder wts% ranging from 5 to 25 were formulated. All formulations were spin coated on glass (hydrophilic) and PC (hydrophobic) substrates to understand the effect of substrate wetting behavior on superhydrophobicity. SEM was used to study the coatings’ topographies with different binder wt% on two different substrates ( Figures 1 – 3 ). The SEM images indicate that a variety of hierarchical structures are developed with respect to the NP location in the coatings; these changes were due to changing interactions between the binder and the NPs, the substrate and the coating system, as well as the binder concentration. For ECA formulations on glass, for all binder concentrations, the hydrophobic NPs appeared to locate on the top layer of the coating as predicated in the analysis. On PC, the results also supported the prediction and NPs appeared to be located on the top layer; however the ECA did not fully wet the substrate surface leading to non-coated regions. In the case of the epoxy binder on glass, >10 wt% epoxy, NPs mostly resided in the binder as predicted. Below 10 wt% binder, NPs were identified on the surface, probably due to lack of sufficient binder to completely surround the particles. On PC, most NPs appeared to be embedded in the epoxy binder as predicted. Coatings with UA binders provided similar trends as those observed for coatings with epoxy binders where NPs appeared to be embedded in the cases of high UA concentrations. As opposed to the other two binders, the NPs were not seen to spread on the surface, but were kept in the binder boundary surface. The thermodynamic-based predictions for the NPs location for the ECA system were consistent with the SEM images showing that the NPs were mostly located on the surface for all ECA concentrations. For the UA and epoxy systems where the spreading coefficient was close to zero, the interaction between the binder and NPs in addition to the heat is sufficient to result to localization in the binders, with 15–25 wt% binder. For 5 and 10 binder wt%, although the spreading coefficient was negative, the NPs were on the surface due to the lack of sufficient binder to wet the NPs. The substrate wetting behavior affects the wetting of the binders. All three binders showed complete wetting on glass, as expected, as glass has high SFE compared to the binders. On PC substrates having lower SFE, epoxy showed full wetting; however, ECA and UA did not wet the surface even at higher binder wt% and uncoated areas were observed. The calculated surface tensions (γ t is the total surface tension, γ d is the dispersive component of the surface tension, and γ p is the polar component of the surface tension), using the pendant drop method, of the neat uncured binders are given in Table 1 . To quantify the SFE of the PC substrate, both Zisman’s 49 and LW/AB methods were used ( Table 5 ). The surface tension of glass (high surface energy) was taken as 146 mN/m (based on the literature for microscope soda lime slides). 54 A comparison of the SFE value shows that the Zisman method had lower value for PC than LW/AB method. As Zisman’s method uses only nonpolar liquids and LW/AB uses both polar and nonpolar liquids, polar interactions are not considered in Zisman’s method. As PC has two aromatic rings and a carbonyl group, which are electron donors, the SFE was taken as 44.36 mN/m. All three binders exhibited lower SFE values than glass, which could be attributed to full wetting of the glass by the binders as seen previously by SEM ( Figures 1 – 3 ). For PC, mixed results were observed because SEM showed only full wetting of PC by the epoxy while ECA and UA did not wet the PC. To explain why ECA and UA having lower surface energy than PC, did not fully wet the surface, further work has been carried out as follows. In spin coating, when the interaction energy between a substrate and a solvent overcomes that between a substrate and a polymer, the films become rough and segregate. On the contrary, when the interaction energy between a substrate and a polymer is stronger than that between a substrate and a solvent, or when both interaction energies are weak, the films obtained are homogeneous and flat. 55 Hence, the calculated interaction energy between the substrate/binder and substrate/solvent can explain the experimental results. The higher the interfacial tension the lower is the interaction between the liquid and the solid. Subsequently, the interfacial tensions were calculated using Young’s equation for contact angle measurements. As acetone was used as the binders’ solvent, the interfacial tension between acetone and the substrates was also measured using the same method. The results are presented in Tables 6 and 7 . The results in Table 6 , for glass substrate, show very high interfacial tension values for both acetone/glass and binders/glass. Because high interfacial tension values means weak interactions, the solvent and binders demonstrate weak interaction energies, resulting in a flattened and more homogeneous film as evident by the respective SEM. The results in Table 7 can explain why ECA and UA did not fully wet the PC substrate. The interfacial tensions between ECA/PC and UA/PC were higher (lower interaction) than the acetone/PC interfacial tension, resulting in segregation and uncoated areas as shown by the respective SEM. For epoxy, the interaction energy with PC was lower than the acetone/PC interaction, meaning stronger interaction energy, resulting in a flattened and more homogeneous film as shown in the respective SEM. Superhydrophobicity study The superhydrophobicity of glass and PC substrates coated with silica NPs dispersed in different binders was evaluated by contact and sliding angle measurements. Three samples were prepared for each formulation, and the average values are shown in Figures 4 and 5 . The contact angle results correlate well with the SEM images and with the formation of nanoroughness, as shown in Figures 4 and 5 . ECA system exhibited superhydrophobicity for all ECA wt% on glass due to the presence of hydrophobic NPs on the top layer as shown in the respective SEM. On PC, however, superhydrophobicity was not obtained when ECA concentration was >10 wt%. SEM showed NPs located on the top layer; however ECA did not fully wet the substrate surface leading to non-coated regions. The presence of coated and non-coated regions on the PC can explain the high standard deviations of the contact angle results for high ECA concentrations – that is, some areas were superhydrophobic and some were not. Epoxy exhibited superhydrophobicity when coating glass from 5 to 10 binder wt%, but superhydrophobicity was lost at higher epoxy wt%. SEM showed that with an increase in binder wt%, it appeared that more NPs penetrated into the binder, resulting in the loss of nanoroughness, and hence, decreasing contact angle values. In the case of the PC substrate, contact angles increased with decreasing epoxy wt% due to the appearance of NPs on the surface; however, none of the epoxy formulations were superhydrophobic. This is probably because the appropriate roughness was not obtained due to the lack of sufficient numbers of NPs on the surface. The UA formulation showed superhydrophobicity at 5 and 10 binder wt% for both glass and PC. SEM showed that <10 wt% UA, NPs were on the surface. In general, the glass substrates provided superhydrophobicity over a wider range of compositions compared to PC. As the formation of nanoroughness is essential for superhydrophobicity, it was necessary to characterize better the localization of NPs on the coating surface as they impart the nanoscale roughness to the surface. Hence, XPS was carried out to analyze whether the NPs were located on the surface, as they appeared to be in the SEM images. First, the as-supplied fumed silica NPs (control sample) and modified (with FAS) silica NPs were mounted on carbon tape and analyzed. The results are shown in Figure 6A . The XPS spectra detected the presence of silicon, oxygen, carbon, and fluorine on the surface, whereas no fluorine was detected for the as-supplied fumed NPs. Thus, the detection of FAS via fluorine atoms can be used to determine if the treated NPs are on the surface or embedded in the matrix. Because the coverage of the NPs on the carbon tape was not complete, the detection of carbon on the as-supplied fumed silica NPs was also related to the carbon tape. Figures 6B, C present XPS data for all glass and PC samples coated with 5 wt% binder concentration. The XPS data confirmed the presence of NPs with FAS on the coating surface as was also confirmed by the respective SEM analysis. It should be emphasized that all these coatings were superhydrophobic except 5 wt% epoxy on PC, although NPs were on the surface for this formulation as well. This non-superhydrophobic surface could be attributed to the lack of sufficient NPs to create the appropriate surface roughness for superhydrophobicity. To conclude this part of the work, spreading coefficient calculation can predict NPs localization for formulations with high binder wt% (≥15) on high surface energy substrates, such as glass. At low binder loading (15 wt%) and with low surface energy substrates, NPs localization did not follow our prediction. The above-mentioned thermodynamic analysis showed that different binder systems demonstrate different behaviors due to the different spreading coefficients, which was shown to be an effective tool to predict the localization of the NPs within the binder used. This allows one to potentially predict which binder systems would create superhydrophobic surfaces when combined with the NPs. Surface free energy (SFE) Thermodynamic analysis was used to study the SFE of the different cured coatings leading to a variety of hierarchical structures with respect to the NPs location in the coatings. The SFE expresses the tendency of a system to increase its surface area resulting in a lower SFE. Hence, the presence of NPs on the surface increases the surface area and causes a reduction of SFE. SFE is further reduced by the presence of fluoro groups on the surface. To quantify the interactions, the SFEs of the component materials were calculated using both Zisman’s and LW/AB methods ( Table 8 ). The surface tension of glass (high surface energy) was taken as 146 mN/m (based on the literature for microscope soda lime slides). 44 A comparison of the SFE values shows that Zisman’s method has the best correlation with the LW/AB method for the epoxy binder. As Zisman’s method uses nonpolar liquids and LW/AB uses polar and nonpolar liquids, these values indicate that polar interactions are insignificant for epoxy. The Zisman SFE values for PC, ECA, and UA were lower than LW/AB values, indicating significant polar interactions in these cases. PC has two aromatic rings and a carbonyl group, ECA has a cyano group and ester group, and UA has urethane and acrylate groups, which are electron donors so their SFEs were taken as 44, 34, and 46 mN/m, respectively. The LW/AB values were used for comparison. A similar procedure was used to determine the SFE for the binder formulations (cured binder, NPs, and fluoroalkylsilane). In this case, the determined LW/AB values were not suitable for highly hydrophobic surfaces, so Zisman’s method was used. Zisman’s values for all binder formulations on glass and PC are reported in Table 9 and graphs are shown in Figures S1 – S4 . As expected, all the surfaces showed lower SFE compared to the neat binders ( Table 5 ). SFE results, however, did not indicate a clear distinction between the superhydrophobic and non-superhydrophobic surfaces. In fact, some non-superhydrophobic surfaces exhibited lower SFE than the superhydrophobic ones. These results confirmed that low SFE values alone are not sufficient to define superhydrophobicity and that surface morphology has an important role in achieving superhydrophobicity. As observed in Table 9 , for ECA formulations coated on glass no significant change in SFE was obtained due to the presence of NPs on the surface (as was shown in the SEM images). For PC, a small change in SFE was obtained between the formulations due to some non-coated areas and exposure of the PC substrate, which has a lower SFE value. As the droplet used to measure the contact angles was large enough (1 mm diameter) to encompass both the non-coated and coated areas, the size of the non-coated areas and their structure must be considered. In the case of epoxy formulations applied on glass, higher SFE values were found for the superhydrophobic formulations with the NPs on the top surface, compared to the non-superhydrophobic ones. For PC, there was no significant change in SFE for all formulations except 5 wt% (where the NPs were on the surface and thus gave a higher SFE value). Superhydrophobic UA formulations (5 and 10 wt%) on glass and PC showed higher SFE values than the non-superhydrophobic ones, as expected. It should be noted that the surface tension of neat ECA and UA measured by the pendant drop method was not consistent with the calculated SFE of the cured binders using the LW/AB method; 25 and 27 mN/m compared to 34 and 47 mN/m, respectively. This difference was attributed to the low molecular weight of the monomers compared to the cured binder as the surface tension of polymers tends to increase with increasing molecular weight. 56 In the case of the two-part epoxy, no significant change was observed between the surface tension and SFE of the cured binder, 38 and 39 mN/m, respectively. This result indicated that the uncured epoxy oligomers may have high enough molecular weight. Durability evaluation and analysis Superhydrophobic coatings can lose their nanoroughness by shear forces due to abrasion or other mechanical forces. As durability is critical for real applications, the mechanical integrity of the coatings was studied using the tape test. The superhydrophobicity after the tape test was determined and the results are shown in Figures 7 and 8 . Tape tests, which are often used to study coating adhesion, were used as a preliminary method to evaluate the adhesion and durability of the superhydrophobicity. No detachment of the coating was visually observed after the tape test. The results indicated that epoxy exhibited the highest durability on glass and PC compared to ECA and UA. All epoxy formulations (the superhydrophobic and non-superhydrophobic) showed only small changes in the contact angle and sliding angle values before ( Figures 7 and 8 ) and after the tape test. This durability was attributed to the negative interfacial tension values between the NPs and epoxy (calculated in Table 4 ), resulting in good adhesion. The tape test results for the ECA/glass system indicated low durability for low binder wt%. This behavior was attributed to excess NPs on the top surface layer that can be detached easily. At higher binder wt%, high contact angles and low sliding angles were unchanged following the tape test due to the low interfacial tension of 4.14 mN/m between ECA and the NPs, which resulted in good adhesion. For the PC substrate, with increasing binder wt% some of the areas were coated, while some were not ( Figure 3 ), a condition that resulted in high standard deviations of the measured contact angles. UA coatings demonstrated low durability even though NPs were embedded in the binder. This poor durability may be due to the increased SFE of the cured UA (46 mN/m) compared to the non-cured UA (27 mN/m), which resulted in higher interfacial tension and low adhesion between the binder and the NPs. Of interest is the fact that some of the coatings are able to retain their superhydrophobic behavior after durability testing. The ECA formulations show good retention of superhydrophobic properties on glass. Clearly further work is needed to improve the behavior on PC."
} | 8,246 |
31110470 | PMC6499189 | pmc | 2,031 | {
"abstract": "Spike-Timing-Dependent Plasticity (STDP) is a bio-inspired local incremental weight update rule commonly used for online learning in spike-based neuromorphic systems. In STDP, the intensity of long-term potentiation and depression in synaptic efficacy (weight) between neurons is expressed as a function of the relative timing between pre- and post-synaptic action potentials (spikes), while the polarity of change is dependent on the order (causality) of the spikes. Online STDP weight updates for causal and acausal relative spike times are activated at the onset of post- and pre-synaptic spike events, respectively, implying access to synaptic connectivity both in forward (pre-to-post) and reverse (post-to-pre) directions. Here we study the impact of different arrangements of synaptic connectivity tables on weight storage and STDP updates for large-scale neuromorphic systems. We analyze the memory efficiency for varying degrees of density in synaptic connectivity, ranging from crossbar arrays for full connectivity to pointer-based lookup for sparse connectivity. The study includes comparison of storage and access costs and efficiencies for each memory arrangement, along with a trade-off analysis of the benefits of each data structure depending on application requirements and budget. Finally, we present an alternative formulation of STDP via a delayed causal update mechanism that permits efficient weight access, requiring no more than forward connectivity lookup. We show functional equivalence of the delayed causal updates to the original STDP formulation, with substantial savings in storage and access costs and efficiencies for networks with sparse synaptic connectivity as typically encountered in large-scale models in computational neuroscience.",
"conclusion": "5. Conclusions There are multiple forms of organizing data structures for storing synaptic weights. Among these different memory arrangements, pointer-based models are capable of data compression by storing only the existent connections in the network. In pointer-based models, weights are stored, in a high-level sense, as lists of post-synaptic addresses and weights, where the pointer to the list is defined by the pre-synaptic neuron address. Biologically relevant neural networks are typically unstructured and sparsely connected, making pointer-based architectures particularly efficient at storing these network topologies. In this work, we studied the storage costs (in bits) of each data structure and identified the most efficient based on network parameters (e.g., network size and weight bit-length) and connectivity density. For the different data structures, we analyzed the computational complexity (in number of memory accesses) of obtaining synaptic address and weight when accessing the tables in forward and reverse directions. Though efficient in terms of storage for a wide range of connectivity density values, pointer-based models natively present only forward connectivity access, making them inefficient when implementing spike-time-based local learning rules such as STDP—which requires both forward (pre-to-post) and reverse (post-to-pre) connectivity access. Therefore, we devised a novel means of efficiently implementing STDP by forward-only synaptic connectivity access, benefiting from the reduced memory storage property of pointer-based data structures. In the traditional STDP algorithm, causal updates are performed at the onset of post-synaptic spikes, demanding reverse access at this instant. Our proposed method operates by delaying the causal weight updates until the instant of expiration of the pre-synaptic STDP timer. With this, forward access is performed for both causal and acausal updates, driven by pre-synaptic events. Natural drawbacks arise when delaying the causal updates, particularly with respect to high-firing post-synaptic neurons. All the drawbacks can be addressed by a very simple rule: the number of STDP timers for each neuron should be equal to the number of spikes which can occur inside the STDP learning window. This rule can be obtained by using multiple timers when T refr < T stdp , with each timer lasting T refr time steps. Using this strategy results in the possibility of implementing nearest-neighbor and all-to-all temporal spike interaction. Additionally, by extending the number of timers, the more complex triplet-based temporal interaction can also be deployed. Lastly, besides the comparison of storage and access costs and efficiencies for each data structure, we devised a budget efficiency figure of merit for a trade-off analysis of the benefits of each model depending on application requirements and storage and access budget. In sum, we feel our work is unique in that it presents a methodology for identifying the optimal memory arrangement solution based on system requirements and network topology, including also the cost of memory access, and supplying the first viable and exact solution for implementing STDP learning in systems organized with either crossbar arrays or forward-only connectivity tables.",
"introduction": "1. Introduction Extensive research in the field of artificial neural networks (ANNs) in the past decade has given rise to diverse neuron functions, network topologies, and training techniques (Nair and Hinton, 2010 ; Krizhevsky et al., 2012 ; Goodfellow et al., 2014 ; Kingma and Ba, 2014 ; Ioffe and Szegedy, 2015 ), capable of solving complex cognitive tasks, such as image classification (Krizhevsky et al., 2012 ), sequence generation (Graves, 2013 ), speech recognition (Graves et al., 2013 ), and game playing (Silver et al., 2016 ). However, the components of these algorithms are normally only loosely based on actual biological neural networks, particularly with respect to the non-local learning rules (e.g., the widely used backpropagation algorithm, Rumelhart et al., 1986 ) and the continuous activation functions (e.g., sigmoid unit and rectified linear unit). Spiking neural networks (SNNs), in contrast, incorporate multiple aspects of biological nervous systems into its components (Gerstner and Kistler, 2002 ), including biologically relevant neuron models, binary activation functions and communication, event-driven processing, and local learning rules (i.e., where all the information required for adjusting parameters between neurons is collocated with these neurons). The neuron models can range from simple single-variable differential equations (e.g., McCulloch-Pitts and integrate-and-fire), to complex systems with dynamics more homologous to real neurons (e.g., Hodgkin-Huxley). In SNNs, neurons communicate between each other via a binary event known as an action potential (or spike ), which is elicited whenever a neuron variable (typically, the membrane potential) crosses a threshold value. Whenever a neuron produces an action potential, this spike event information is conveyed to its population of downstream post-synaptic neurons, resulting in an update of their respective internal variables based on the values of synaptic efficacy (or weight ). Due to their binary nature, the time at which spikes occur is essential information when training SNNs. The origins of hardware designed to emulate the biological nervous system, also known as neuromorphic systems Mead ( 1990 ), targeted design of neural properties at the device level, with natural focus on analog circuits (Maher et al., 1989 ; Andreou et al., 1995 ; Koch and Mathur, 1996 ). More recently, however, neuromorphic systems such as TrueNorth (Merolla et al., 2014 ), SpiNNaker (Furber et al., 2014 ), and Loihi (Davies et al., 2018 ) were designed with purely digital components, being capable of emulating large-scale SNNs with real-time dynamics in the millisecond timescale. Additionally, large digital systems have the advantage of being more readily verifiable in simulation and a software-hardware equivalence is typically possible. While ANNs operate in a sequential manner, where data propagates through the network one layer at a time, neuromorphic systems typically present multiple cores running in parallel at biological timescales, with synaptic memory local to each core. Systems with distributed processing and memory move away from the traditional von Neumann architecture, where memory is centralized and a high-frequency global clock is responsible for fast computation and memory access (Merolla et al., 2014 ). Among the bio-inspired learning mechanisms, spike-timing-dependent plasticity (STDP) is perhaps the most widely considered form of induced synaptic modification (Markram et al., 1997 ). STDP originated from experimental data collected in cultures of dissociated rat hippocampal neurons, where scientists observed that a causal relationship between spike times of pre- and post-synaptic neurons could induce synaptic strengthening or weakening, and this change was correlated with the relative temporal difference of spikes (Bi and Poo, 1998 ). The experiments showed that long-term potentiation and long-term depression could both be induced in synapses depending on the order of spike occurrence, where a causal relationship (i.e., pre-synaptic neuron spikes before post-synaptic neuron) potentiated the synapse, while an acausal relationship (i.e., post-synaptic spikes before pre-synaptic) weakened the synapse. The authors then approximated the measured synaptic modification with a mathematical model. In the model, the STDP function (or kernel ) defines the change of the weight as a function of the relative time between pre- and post-synaptic action potentials, and the duration of the causal (and acausal) influence of spikes is called the STDP learning window (Sjöström and Gerstner, 2010 ). An important aspect of STDP is that, though it is a local learning rule, weight updates occur at the onset of both pre- and post-synaptic spikes, requiring for the algorithm to be able to not only identify all neurons which the pre-synaptic neuron sends its spikes to, but also locate all the neurons which the post-synaptic neuron receives its spikes from. This is a fundamental property of STDP, and throughout our work we will refer to reading the neuron addresses and weights from pre-to-post connectivity as forward access and reading from post-to-pre connectivity as reverse access . In traditional ANNs, the typical data structure used to represent the weights between neurons is a dense matrix, constituting a fully connected topology. However, more realistic and biologically relevant neural networks, such as small-world and locally connected random networks (Bassett and Bullmore, 2006 ; Bullmore and Sporns, 2009 ; Seeman et al., 2018 ), do not conform to this structured topology. In these cases, synaptic weight storage costs can benefit greatly using compressed representations. For physical realizations of the STDP learning rule, the arrangement used to organize the synaptic weights in memory has a direct impact on the ease of forward and reverse access. As we will later show, dense matrices typically have the advantage of natively facilitating both types of connectivity access. Conversely, compressed memory arrangements suffer greatly when trying to access in the reverse direction, making causal STDP weight updates in these structures computationally intensive. In this work, we discuss the complexity of storing and accessing synaptic weights in different types of data structures and their impact on implementations of the STDP algorithm, and propose a novel method of performing STDP using only single-direction connectivity access, consequently taking advantage of compressed structures. Storage costs associated to synaptic weight memory arrangements have been previously studied (Moradi et al., 2013 ; Pedroni et al., 2016 ; Joshi et al., 2017 ; Kornijcuk et al., 2018 ). In Materials and Methods, we give an overview of four typical data structures used for representing synaptic weights, and analyze storage costs based on different network parameters (number of neurons and weight bit-length) and varying degrees of network connectivity density. We extend our analysis to verify the memory access cost and efficiency associated to each data structure, focusing particularly on the computational complexity and requirements for performing STDP. Inspired by our previous work (Pedroni et al., 2016 ), we propose a definite pre-synaptic-driven solution for obtaining a quantitatively equivalent algorithm to STDP. Previous attempts in approximating STDP using forward-only connectivity include (1) simplifying the STDP rule by equally updating all the synaptic weights based on recent spike activity (Bichler et al., 2012 ; Yousefzadeh et al., 2017 ), (2) using other variables (usually post-synaptic membrane potential) as a proxy for the post-synaptic spike times when computing causal updates (Brader et al., 2007 ; Davies et al., 2012 ; Lagorce et al., 2015 ; Sheik et al., 2016 ), and (3) delaying the weight updates (Jin et al., 2010 ; Davies et al., 2018 ). In the discussion, we compare our method to these, particularly with the third type, currently present in SpiNNaker and Loihi, and explain how our solution can produce exact STDP while previous methods rely on particular balanced firing rate conditions in the network or simply produce qualitative approximations to STDP. In Results, a network composed of 256 pre-synaptic and 256 post-synaptic neurons is simulated using our proposed method and compared against the original STDP learning rule, showing that our method produces the same post-synaptic membrane potentials, resulting in identical spiking activity and synaptic weights.",
"discussion": "4. Discussion Storage costs associated to synaptic weight memory arrangements have been previously studied. In Moradi et al. ( 2013 ), the authors describe a network clustering scheme which uses a two-stage routing architecture to reduce the overall memory storage requirements. This method is also mentioned in Joshi et al. ( 2017 ) and is referred to as “clustered addressing.” In both of these studies, the storage savings comes at the cost of reduced flexibility in network connectivity, since a specific topology must exist for groups of neurons to be clustered together. Instead, we decided not to constrain our networks to any structured topology. In Joshi et al. ( 2017 ), the authors describe the data structures we have presented, highlighting, particularly, the storage cost savings obtained for a large range of connectivity density when using the PB-BMP architecture. However, the impact of pointer-based models on learning algorithms was only briefly mentioned, and memory access costs were not analyzed. More recently, the impact of using different memory arrangements on spike routing and network traffic congestion was described in Kornijcuk et al. ( 2018 ). Though the work describes a theoretical means of routing-rate evaluation and results for maximum network sizes for each of their memory arrangements, it does not target any specific learning algorithm, and the experimental results focus only on an inference task without synaptic plasticity. More recently, the authors in Kim et al. ( 2018 ) proposed a modified SRAM which enables transposable memory access. The method is interesting as it facilitates the reverse (post-to-pre) access for causal updates; however, it can only be applied to fully connected network topologies (i.e., crossbars), and, thus, are not efficient for representing sparse networks since compressed data structures are typically not transposable. In terms of spike-driven learning, there have been multiple attempts to replicate or approximate STDP with forward-only connectivity. The motivation for storing synaptic weights in a pre-synaptic perspective (i.e., pre-to-post) is because post-synaptic-driven systems are not as efficient in terms of number of memory accesses as pre-synaptic-driven systems; this is mainly because, as we sweep through neurons to update their states during a system time step, Δ t , for each post-synaptic neuron we must verify the spike state of every pre-synaptic neuron, even if none of these has spiked. Conversely, pre-synaptic-driven systems operate in an on-demand fashion, accessing the pre-synaptic spike states only as needed. In Pedroni et al. ( 2016 ); Detorakis et al. ( 2018 ), we described a less-detailed version of our method; yet, we did not study all the data structures nor were we able to address all of the drawbacks incurred by delayed causal updates (as we have shown in the current paper). One of the earliest works which evaluated the complexity of implementing the STDP learning algorithm in a neuron address domain was presented in Vogelstein et al. ( 2003 ). The authors discussed how the address-event representation (AER) protocol could support STDP learning in the address domain. Being pioneering work, the paper considered only small networks, consequently not addressing the different possible arrangements for organizing synaptic weights in memory and the implications of requiring reverse access for performing causal updates. Methods that approximate STDP learning by equally updating all the synaptic weights based on recent spike activity have been proposed. In Bichler et al. ( 2012 ), the authors use a special form of STDP which equally depresses all the synapses that did not recently contribute to the post-synaptic spike activation regardless of their activation time; in contrast, synapses that were activated with a pre-synaptic spike a short time before post-synaptic spikes are strongly potentiated. The authors in Yousefzadeh et al. ( 2017 ) created a more hardware-friendly version of this model by limiting the number of synapses to be potentiated (instead of limiting the STDP time window duration), eliminating the need for time-stamping the spikes. Though efficient in terms of memory access, with both of these methods it is not possible to depress synapses whose activation time is precisely not correlated with the post-synaptic spike, and the methods only work if LTD is systematically applied to synapses not undergoing an LTP. Additionally, the methods are post-synaptic-driven, undergoing the aforementioned drawbacks of this mechanism. Another alternative to approximating STDP is by using other variables (usually post-synaptic membrane potential) as a proxy for the post-synaptic spike times when computing causal updates. This learning rule was proposed in Brader et al. ( 2007 ) and has even been incorporated in the SpiNNaker system (Davies et al., 2012 ; Lagorce et al., 2015 ). More recent work describes how to use the rule for learning sequences of spikes (Sheik et al., 2016 ). Once again, though very efficient in terms of memory access and spike time storage, in this method exact STDP is not possible as post-synaptic potential serves only as a [deterministic (Lagorce et al., 2015 ) or probabilistic (Sheik et al., 2016 )] proxy of the post-synaptic spike time and, in many cases, is not capable of capturing the subtle spike time causalities of STDP. The third category of methods for approximating STDP consists on delaying the weight updates, and is the category which our proposed method falls under. In the Loihi system, the authors adopt a less event-driven method where synaptic modification is performed in an epoch-based mechanism (Davies et al., 2018 ). Their method delays the updating of all synaptic states to the end of a periodic learning epoch time, and, to avoid receiving more than one spike in a given epoch, the epoch period is normally set to the minimum refractory delay of all neurons in the network. Though Loihi implements forward connectivity tables for supporting generalized STDP rules, the periodic servicing (i.e., non-event-driven methodology) can result in inexact weights being delivered to post-synaptic neurons since multiple pre-synaptic spikes may occur before a weight update takes place. Therefore, certain conditions in firing rates must be guaranteed for their method to be equivalent to STDP. In the current version of the SpiNNaker system, STDP learning is approximated using a trace-based approach via delayed updates (Mikaitis et al., 2018 ). Since in trace-based STDP each spike leaves an exponentially decaying trace (Morrison et al., 2008 ), this renders possible linearly accumulating the spike traces into a single variable, representing the total current effect of all past spikes. In this manner, weight updates can then be performed in an online fashion at the onset of either pre- or post-synaptic spikes. In SpiNNaker, however, the updates only occur at the onset of pre-synaptic spikes, meaning that, for the method to follow rather closely to original STDP, the system relies on frequently firing pre-synaptic neurons. This issue can be observed in the case when a post-synaptic neuron spikes multiple times soon after a pre-synaptic spike (typically resulting in large causal updates): if the pre-synaptic neuron spikes again in a much later time, then the causal updates will be practically null due to the almost completely decayed traces (somewhere along the lines of the problem encountered in case 4 in Figure 4D ). Additionally, besides serving only as an approximation to STDP, the trace-based method requires an exponentially decaying kernel, and, thus, other kernels such as those in Figure 1C cannot be implemented. Perhaps the most similar work to ours has been presented in Jin et al. ( 2010 ), which uses a deferred-event approach and stores spike times for postponed processing at the time of the next event following them. This method has been previously implemented in the SpiNNaker system under their “deferred event driven model” (Rast et al., 2008 ; Diehl and Cook, 2014 ; Galluppi et al., 2015 ). It is similar to our proposed method in that weight updates are driven by pre-synaptic spikes and causal updates are delayed; however, some important distinctions should be highlighted: A neuron's spike history is stored as a bitmap in an array. However, as presented in Appendix A1 , using multiple timers is at least as efficient as using a bitmap array, and becomes extremely more efficient for large T refr . Acausal updates are not immediately processed and are also deferred to the future, once more pre-synaptic spikes have arrived. This implies that larger arrays are required to store spikes on both sides of the STDP window for post-synaptic neurons. In fact, in their work the post-synaptic bitmap array is three times larger than the pre-synaptic array. In our solution, applying the acausal updates immediately at the onset of pre-synaptic spikes demands that we use timers that must cover only one side (i.e., the longest side) of the STDP window. Since the bitmap array is only updated at the onset of new spikes (but not necessarily at the expiration of the pre-synaptic STDP timer) and STDP updates can only take place when an “old” pre-synaptic spike eventually exits the bitmap array, this means that both causal and acausal updates rely on frequently firing pre-synaptic neurons. This demands that pre-synaptic spikes arrive at a high enough rate to ensure that the pre-synaptic spike time bitmap array is frequently updated so weight updates are not lost. In their work, the minimum firing rate for pre-synaptic neurons is 10.4 Hz. Since multiple pre-synaptic spikes may occur before an “old” pre-synaptic spike eventually exits the bitmap array, this implies that the weights being used for updating post-synaptic neuron variables at each pre-synaptic spike event could (or most likely will) be an “old” set of weights since the causal and acausal updates have been deferred. Therefore, though qualitatively similar, a quantitative equivalence with the original STDP algorithm will probably not occur."
} | 5,965 |
29570696 | PMC5951332 | pmc | 2,033 | {
"abstract": "Poly(chloro-p-xylylene) (PPXC) film has a water contact angle (WCA) of only about 84°. It is necessary to improve its hydrophobicity to prevent liquid water droplets from corroding or electrically shorting metallic circuits of semiconductor devices, sensors, microelectronics, and so on. Herein, we reported a facile approach to improve its surface hydrophobicity by varying surface pattern structures under different temperature and relative humidity (RH) conditions on a thermal curable polydimethylsiloxane (PDMS) and hydrophobic silica (SiO 2 ) nanoparticle coating. Three distinct large-scale surface patterns were obtained mainly depending on the contents of SiO 2 nanoparticles. The regularity of patterns was mainly controlled by the temperature and RH conditions. By changing the pattern structures, the surface wettability of PPXC film could be improved and its WCA was increased from 84° to 168°, displaying a superhydrophobic state. Meanwhile, it could be observed that water droplets on PPXC film with superhydrophobicity were transited from a “Wenzel” state to a “Cassie” state. The PPXC film with different surface patterns of 200 μm × 200 μm and the improved surface hydrophobicity showed wide application potentials in self-cleaning, electronic engineering, micro-contact printing, cell biology, and tissue engineering.",
"conclusion": "4. Conclusions In summary, PDMS could be well thermally cured at a temperature of 60 °C or higher for 90 min. Meanwhile, the content of SiO 2 nanoparticles and the relative humidity (55% and 95%) showed little effect on the complete curing of PDMS at 60 °C or higher temperature. However, among the samples treated at different temperatures and RH conditions, the cross-linking reaction process was different and resulted in different surface morphologies. By changing the contents of SiO 2 nanoparticles, thermal cross-linking temperatures, or RH treatment conditions, three different types of large-scale surface patterns (RBS, TPS, and NAS) were obtained on PPXC film. Meanwhile, the number, distribution, and dimension of these three pattern structures were tuned by changing the experimental conditions. With the evolution of surface patterns, the surface wettability was changed from weak hydrophobicity to superhydrophobicity and the WCA increased from 84° to 168°. A typical transition of water droplets from a “Wenzel” state to a “Cassie” state on the surface was observed. The PPXC film with different surface patterns of 200 μm × 200 μm and the improved surface hydrophobicity showed wide application potentials in self-cleaning, electronic engineering, micro-contact printing, cell biology, tissue engineering, and so on.",
"introduction": "1. Introduction Poly(chloro-p-xylylene) (PPXC) film has been extensively applied in the fields of semiconductor devices, sensors, microelectronics, material moisture protection, and so on [ 1 , 2 , 3 , 4 ]. However, it has weak hydrophobicity with a water contact angle (WCA) of only about 84° [ 5 ]. It is necessary to improve its hydrophobicity to prevent liquid water droplets from corroding or electrically shorting metallic circuits of the devices. Superhydrophobic coating [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 ] with a WCA greater than 150° and a sliding angle (SA) of less than 10° was used to prevent water droplets from wetting PPXC film. A facile strategy of improving surface hydrophobicity to superhydrophobicity of PPXC film is important in industrial applications of PPXC film [ 15 , 16 , 17 ]. The surface wettability can be tuned by changing surface morphologies or surface chemical compositions [ 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 ]. Regulation of surface morphologies had been extensively explored for pattern structures [ 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 ], which could be obtained by printing [ 35 , 36 ], self-assembly of block copolymer [ 37 ], and femtosecond laser structuring [ 38 ]. However, these methods could not satisfy the requirements of industrial applications due to some key defects, such as the small area of patterns, multi-step processes, and high cost. Breath figures (BF) based on self-assembled water droplet arrays as dynamic templates had been extensively used to produce ordered honeycomb polymer films. This method had obvious advantages in cost and versatility [ 39 , 40 , 41 , 42 ], but the surface morphologies of BF films were closely related to the self-assembled water droplet arrays, which could be tuned by changing solvents, polymers, the concentration of solution, the temperature, and relative humidity (RH) [ 43 , 44 ]. Due to the complexity of these influencing factors, a facile and inexpensive strategy to obtain large-scale patterns on PPXC film still remains a challenge. Controlling the treatment conditions of polymer coating such as temperature and RH was considered as a facile strategy to achieve tunable surface morphology and wettability. However, the effects of temperature and RH on the anti-stick plastic substrates such as PPXC film were seldom reported. Besides, most of previous studies on the BF method were focused on polymers without thermal curing such as polystyrene, and thermal curable polymers such as polydimethylsiloxane (PDMS) used as replica materials were seldom explored [ 45 , 46 ]. Meanwhile, the introduction of hydrophobic silica (SiO 2 ) nanoparticles under temperature and RH treatments would have great effects on the surface morphology and surface wettability of PPXC film. Therefore, we herein reported a facile strategy to obtain large-scale patterns of 200 μm × 200 μm and tune surface wettability of PPXC film with a thermal curable PDMS/SiO 2 coating. The pattern structures obtained by the BF method were usually porous honeycomb structures, but herein three structures (special raised bowl-shaped structures, traditional porous structures, and nanoparticle aggregation structures) were obtained on PPXC film. Moreover, the effects of different treatment conditions such as the content of SiO 2 nanoparticles, thermal cross-linking temperature of PDMS, and RH on the morphology evolution of PDMS/SiO 2 coatings were explored systematically, and the surface wettability of PPXC film could be improved from weak hydrophobicity with a WCA of about 84° to superhydrophobicity with a WCA of about 168°. Meanwhile, a typical transition of water droplets from a “Wenzel” state to a “Cassie” state on the surface was observed after tuning the treatment conditions. This transition was demonstrated by the sliding behavior of water droplets on the surface and the temporal changes of WCA on samples. To the best of our knowledge, it is the first time this engineering strategy to obtain large-scale different patterns on PPXC film, without destroying the original PPXC film, has been reported.",
"discussion": "3. Results and Discussion 3.1. Effects of Temperature, RH, and Content of SiO 2 on the Curing Behavior of PDMS Owing to the thermal curability of PDMS, the temperature is an important factor in the cross-linking reaction process. As shown in Figure 2 a, the non-isothermal exothermic curve characterized by DSC indicates that the PDMS curing reaction occurs at about 60 °C and reaches a maximum reaction rate at near 120 °C with the appearance of an exothermic peak. To further investigate the cross-linking reaction behavior of PDMS at different temperatures, isothermal DSC curves of PDMS at 120, 80, 60, and 50 °C for 90 min were determined ( Figure 2 b). The curing reaction reaches a maximum rate within 40 s at 120 °C with the appearance of an obvious exothermic peak, but only a small exothermic peak occurred within 60 s at 80, 60, or 50 °C (red rectangle frame in Figure 2 b and enlarged image in Figure 3 a). Therefore, the curing reaction rate was obviously decreased at a lower curing temperature, and the required curing time of PDMS was increased. At 120 °C, the cross-linking reaction ends within 500 s, but it extended to 3000 s at 80, 60, or 50 °C (red rectangle frame in Figure 2 b, and enlarged image in Figure 3 b). Therefore, the curing reaction time for all samples in this study was set as 90 min to ensure the complete curing of PDMS. Based on the elasticity of cured PDMS, we further investigated whether the PDMS was completely cured or not by a repeated stretching test of the sample after curing at 80, 60, and 50 °C for 90 min under 55% RH (abbreviated as 80 °C-55% RH). The completely cured PDMS with good elasticity after curing at 80 °C and 60 °C could retain its original shape after several stretches ( Figure 2 c,d). However, the sample cured at 50 °C showed a typical viscous state ( Figure 2 e) indicating that PDMS was not completely cured. Therefore, the cross-linking reaction temperature of PDMS with a completely cured state for 90 min under 55% RH should not be lower than 60 °C. Furthermore, FTIR results shown in Figure 4 are used to investigate the effects of higher RH (95%) and different contents of SiO 2 nanoparticles on the cross-linking reaction of PDMS. As for the sample (0.0 wt % of SiO 2 ) without curing, the reactive groups of Si–H (2162 cm −1 ) and C=C (1614 cm −1 ) are obviously observed in green dash line. However, there are no typical peaks for these reactive groups after the cross-linking reaction at 60 °C-95% RH, indicating that PDMS can be thermally cured completely under a higher RH compared to a lower RH (55%) shown in Figure 2 d. Similarly, the reactive groups of Si–H and C=C also cannot be observed after curing at 60 °C-95% RH for the samples with the increased contents of SiO 2 nanoparticles (1.0 and 2.5 wt % of SiO 2 ). Meanwhile, the characteristic peaks indicated the existence of typical groups in PDMS molecular chains at 2965 cm −1 , 1404 cm −1 , 1082 cm −1 , and 797 cm −1 ascribed to -CH 3 and Si-O-Si groups, respectively. Thus, the higher RH and different contents of SiO 2 nanoparticles had no obvious effect on the complete thermal curing of PDMS at 60 °C for 90 min. Therefore, the surface morphologies of samples treated for 90 min in the study were in a stable and ultimate state and the surface morphology evolution is discussed below. 3.2. Surface Morphology Evolutions of PDMS/SiO 2 Coating on PPXC Film under Different Treatment Conditions Although PDMS with different contents of SiO 2 (from 0.0 to 2.5 wt %) could be thermally cured completely at different cross-linking reaction temperatures (120, 80, and 60 °C for 90 min) and different RH treatment conditions (55% and 95%), the cross-linking reaction processes might differ from each other and result in different surface morphologies. SEM and AFM images were implemented to confirm this prediction. Surface morphology evaluation of the samples with different contents of SiO 2 (0.0, 0.5, 1.0, and 2.0 wt %) treated at 80 °C-55% RH is shown in Figure 5 . A large number of special raised bowl-shaped structures (RBS) could be found on the surface of PPXC film with pure PDMS. The special RBS showed the homogeneous distribution on the whole PPXC film ( Figure 5 (a-0.0)), but a similar phenomenon was not reported in previous studies on PPXC film with different temperatures or RH treatment conditions. Furthermore, a higher magnification of RBS is characterized by the AFM image (black circles in Figure 5 (b-0.0). Most of the RBS had a similar typical dimension of about 7 μm in length, and the periphery of RBS was obviously higher than the center of RBS, as indicated by the altitude scale on the right of the AFM image. When low contents (0.5 and 1.0 wt %) of SiO 2 nanoparticles were added in PDMS coating, RBS disappeared and the traditional porous structures (TPS, similar to the pores formed via a traditional “Breath Figures” method) appeared ( Figure 5 (a-0.5) and (a-1.0). The number of TPS showed no significant variation compared to that of RBS, but the dispersion of TPS is much larger than that of RBS ( Figure 5 (a-0.0)). Meanwhile, the dimension of TPS becomes inhomogeneous with a distribution between 1 and 7 μm (red dash circles in high-magnification AFM images, Figure 5 (b-0.5) and (b-1.0). With a high content (2.0 wt %) of SiO 2 nanoparticles in PDMS coating ( Figure 5 (a-2.0)), RBS or TPS did not appear, but a lot of nanoparticle aggregation structures (NAS) occurred, like previous studies [ 47 , 48 ]. The distribution of NAS is very homogeneous, as shown in a high-magnification AFM image ( Figure 5 (b-2.0)), due to the good dispersion of PDMS and SiO 2 in hexamethylene solution. As mentioned above, lower temperature increased the time required for the cross-linking reaction of PDMS. In this section, we further investigated the surface morphology evaluation of the samples with different contents of SiO 2 (0.0, 0.5, 1.0 and 2.0 wt %) at a lower curing temperature (60 °C for 90 min under 55% RH, Figure 6 ). A large number of RBS with homogeneous distribution could be found on the whole PPXC film with pure PDMS ( Figure 6 (a-0.0)), showing a similar distribution to that in Figure 5 (a-0.0). However, the number of RBS was much greater and the dimension of RBS became smaller with the length of only about 5 μm after the treatment at 60 °C ( Figure 6 (b-0.0), black circles) compared to that obtained after the treatment at 80 °C ( Figure 5 (b-0.0)). After the low content (0.5 and 1.0 wt %) of SiO 2 nanoparticles were added in PDMS coating, RBS disappeared, and TPS appeared ( Figure 6 (a-0.5) and (a-1.0)), showing the same morphology as that obtained after the treatment at 80 °C ( Figure 5 (a-0.5) and (a-1.0)). The number of TPS increased, but the dispersion of TPS showed no significant change compared to that of RBS ( Figure 6 (a-0.0)). Meanwhile, the dimension of TPS became smaller with a homogeneous distribution close to 1 μm (red dash circles in AFM images, Figure 6 (b-0.5) and (b-1.0)) compared to that of RBS ( Figure 6 (b-0.0)). After high contents (2.0 wt %) of SiO 2 nanoparticles were added into PDMS coating ( Figure 6 (a-2.0) and (b-2.0)), lots of NAS with homogeneous distribution were found, similar to that in Figure 5 (a-2.0) and (b-2.0). Moreover, the effect of a higher RH on the surface morphology evaluation of samples with different contents of SiO 2 (0.0, 0.5, 1.0 and 2.0 wt %) was also investigated at a lower temperature (60 °C for 90 min under 95% RH) ( Figure 7 ). Compared with the samples in Figure 5 (a-0.0) and Figure 6 (a-0.0), a large number of RBS could be found on the entire surface of PPXC film with pure PDMS ( Figure 7 (a-0.0)). However, the number of RBS obviously decreased and the dimension of some RBS with a severely inhomogeneous distribution became as long as about 45 μm (or near 100 μm) (black circles in Figure 7 (a-0.0) and (b-0.0)). For the coating with the low content (0.5 and 1.0 wt %) of SiO 2 nanoparticles, RBS disappeared, and TPS occurred. The number of TPS shows little change ( Figure 7 (a-0.5) and (a-1.0)), showing the same morphology as the samples treated at 55% RH ( Figure 6 (a-0.5) and (a-1.0)). However, the dispersion and dimension of TPS increased a little (red dash circles in AFM images, Figure 7 (b-0.5) and (b-1.0)) due to the appearance of some large-scale TPS compared to that in Figure 6 (b-0.5) and (b-1.0). After high contents (2.0 wt %) of SiO 2 nanoparticles were added in PDMS coating ( Figure 7 (a-2.0) and (b-2.0)), lots of NAS still could be found, but there were still a few TPS, shown as red dash circles in Figure 7 (b-2.0) compared to the samples in Figure 5 (b-2.0) and Figure 6 (b-2.0). The TPS should be attributed to the higher RH adopted in the sample treatment at 60 °C-95% RH. To clearly reveal the mechanism of surface morphology evolution, the effects of cross-linking reaction temperature (upper row for 80 °C and lower row for 60 °C), RH (left column for 55% RH and right column for 95% RH), and content of SiO 2 (left column for pure PDMS without SiO 2 , middle column for PDMS with a low content of SiO 2 , and right column for PDMS with a high content of SiO 2 ) are shown in Figure 8 . Based on the results in Figure 5 , Figure 6 and Figure 7 , the conclusions can be drawn below. Three different types of surface patterns (RBS, TPS, and NAS) with an area over 200 μm × 200 μm were generated on PPXC film mainly by controlling the content of SiO 2 nanoparticles. Firstly, as for the coating on PPXC film without SiO 2 nanoparticles, only RBS could be found on the surface even at different cross-linking reaction temperatures and under different RH treatment conditions. It should be attributed to the repellence of PDMS molecular chains to condensed water droplets; the PDMS molecular chains migrated and finally cured at the three-phase contact line (the interface of coating phase, condensed water droplets phase, and air phase) to form typical RBS after the evaporation of condensed water droplets, as shown in the upper right image of the left column in Figure 8 . Secondly, as for coatings with low contents of SiO 2 nanoparticles (between zero and 2.0 wt %) shown in the middle column of Figure 8 , RBS disappeared but TPS occurred under different cross-linking reaction temperatures and RH treatment conditions. The changed phenomenon might be interpreted as follows. The introduction of SiO 2 nanoparticles greatly hindered the migration of PDMS molecular chains to the three-phase contact line. Nonetheless, PDMS molecular chains underneath condensed water droplets could still migrate to SiO 2 nanoparticle aggregations due to the repellence to the sinking condensed water droplets. Therefore, many pores were formed with the evaporation of condensed water droplets, similarly with the formation process of pores via the traditional “Breath Figures” method. Thirdly, with the increase of SiO 2 nanoparticles, TPS were gradually transformed into NAS and there were almost no TPS when the content increased to no less than 2.0 wt % (the right column in Figure 8 ). Condensed water droplets hardly sank into the large number of aggregated SiO 2 nanoparticles due to the lower movability of nanoparticles than PDMS molecular chains. These three typical physical structures can be further demonstrated on sample treated by a high thermal cross-linking temperature (80 °C) and RH treatment condition (95%) shown in Figure S1 in Supplementary Materials . There were still only RBS on the sample without SiO 2 nanoparticles ( Figure S1 (a-0.0) and (b-0.0), black circles). TPS could be found with a low content of SiO 2 nanoparticles (between zero and 2.0 wt %, red dash circles) ( Figure S1 (a-0.5), (a-1.0) and (b-0.5), (b-1.0)). Meanwhile, TPS were gradually changed to NAS and there were almost no TPS when the content of SiO 2 nanoparticles increased above 2.0 wt % ( Figure S1 (a-2.0) and (b-2.0)). Under the same content of SiO 2 nanoparticles, the number, distribution, and dimension of RBS showed similar variations with TPS and these variations could be tuned by changing thermal cross-linking temperature and RH treatment condition. At a higher temperature and under the same RH treatment conditions, the faster thermal curing of PDMS could not provide enough time for the formation of a stable RH condition to generate a large number of homogeneous condensed water droplets on the coating surface. Meanwhile, water vapor had more energy to show a more drastic Brownian Motion, which resulted in the formation of larger condensed water droplets with a decrease in the regularity. Therefore, the number was smaller and the distribution was less homogeneous. The dimension was larger for RBS and TPS at 80 °C than that at 60 °C. Furthermore, compared to the lower RH condition (55%), the higher RH condition (95%) with more water vapor also resulted in the formation of larger condensed water droplets and a decrease in the number and regularity of RBS and TPS. We compared the samples ( Figure S1 ) obtained with a higher thermal cross-linking temperature (80 °C) and higher RH treatment condition (95%) with the samples ( Figure 5 , Figure 6 and Figure 7 ) obtained with a lower thermal cross-linking temperature (60 °C) and/or lower RH treatment condition (55%) and found similar variations in RBS and TPS: the number of RBS or TPS became smaller; the distribution was less homogeneous, the dimension was increased. Moreover, the number, distribution and dimension of TPS were affected by the content of SiO 2 nanoparticles. The number decreased and the distribution became irregular. The dimension became smaller with the increase in the content of SiO 2 nanoparticles ( Figure 5 (a-0.5)–(a-1.0), Figure 6 (a-0.5)–(a-1.0), Figure 7 (a-0.5)–(a-1.0) and Figure S1 (a-0.5)–(a-1.0)). If the content of SiO 2 nanoparticles reached 2.0 wt %, TPS nearly disappeared and were transformed into NAS because the migration of PDMS molecular chains had been seriously hindered by the high content of SiO 2 nanoparticles. Therefore, the thermal cross-linking temperature and RH treatment condition showed little effect on NAS, which seemed to be only controlled by the content of SiO 2 nanoparticles (the right column in Figure 8 ). 3.3. Surface Wettability of PDMS/SiO 2 Coating on PPXC Film under Different Treatment Conditions According to previous studies, the surface wettability was usually determined by the rough physical structures and low surface energy chemical compositions. As the materials used in this study (both PDMS and hydrophobic SiO 2 shown in Figure 4 ) were hydrophobic under different treatment conditions, the evolution of surface morphologies ( Figure 5 , Figure 6 and Figure 7 ) should be responsible for the variations in surface wettability ( Figure 9 ). RMS calculated from AFM images is shown in Figure 9 a. The surface roughness increased monotonously and rapidly with the increase in the content of SiO 2 nanoparticles for the samples treated at 80 °C, but RMS showed a tiered growth for the samples treated at 60 °C (similar RMS of 85 nm for the samples with zero, 0.5, and 1.0 wt % SiO 2 ; similar RMS of 110 nm for the samples with 1.5 and 2.0 wt % SiO 2 ; and similar RMS above 150 nm for the samples with 2.5 wt % SiO 2 ). This special variation of RMS was ascribed to a faster thermal curing of PDMS at 80 °C than at 60 °C, so the surface physical structures were quickly immobilized without enough time to transform to homogeneous structures. The RMS variation of the samples treated at 60 °C-55% RH and 60 °C-95% RH indicated that RH had little effect on the statistic surface roughness but showed obvious effects on the number, distribution, and dimension of surface structures ( Figure 5 , Figure 6 , Figure 7 and Figure 8 and Figure S1 ). Due to the increase in RMS, WCA also increased gradually with the increase in the content of SiO 2 nanoparticles ( Figure 9 b–d). As for the samples with RBS (pure PDMS without SiO 2 ), WCA ( Figure 9 b,c) was about 118° and showed little increase compared to that on smooth PDMS [ 48 ] due to the increased roughness caused by RBS patterns ( Figure 5 or Figure 6 ). Furthermore, WCA in Figure 9 d was only about 108° for the sample treated at 60 °C-95% RH due to the much smoother RBS with a lower RMS ( Figure 7 , about 65 nm) compared to that on the samples treated at 80 °C-55% RH and 60 °C-55% RH ( Figure 5 and Figure 6 , about 89 nm). Water droplets could not slide off all the samples with RBS even after 180° reversal of the samples (the upper left profiles in Figure 9 b–d), indicating that water droplets on the surface were at a typical “Wenzel” state [ 49 ]. This should result from a lack of enough nano-scale roughness and the existence of only some micro-scale RBS (black circles in Figure 10 , Figures S2 and S3 ) on the surface. Moreover, water droplets on the samples with TPS also could not slide off the surface (the left of red dash circles in Figure 9 b–d). As shown in the surface morphologies of the samples treated at 60 °C-55% RH in Figure 10 (a-0.5), (b-0.5), (a-1.0), and (b-1.0) and the samples treated at 80 °C-55% RH and 60 °C-95% RH in Figures S2 and S3 , micro-scale TPS (red dash circles) and nano-scale roughness based on SiO 2 nanoparticle aggregations were increased, thus improving the surface hydrophobicity. However, water droplets were pinned on these samples with a WCA lower than 150° for two reasons. Firstly, the low nano-scale roughness was caused by the too-low content of SiO 2 nanoparticles (0.5 or 1.0 wt %) and the submersion of some SiO 2 nanoparticles into PDMS (white circles in Figure 10 , Figure S2 and Figure S3 ). Secondly, the capillary effect of the pores from TPS and the gaps formed between SiO 2 nanoparticle aggregations shown in red dash circles in Figure 10 , Figure S2 , and Figure S3 led to the phenomenon of water droplets. When WCA on the samples was higher than 150° (the right of red dash circles in Figure 9 b–d), water droplets began to slide off the surface with an SA less than 40°, showing a transition of water droplets from a “Wenzel” state to a “Cassie” state [ 49 , 50 , 51 ]. Samples in the left area of the circles were in a “Wenzel” state because water droplets could not slide off the surface even after 180° reversal of the surface, and samples in the right area of the circles were in a “Cassie” state because water droplets began to slide off the surface with the help of the release of air bubbles underneath the droplets. This transition is consistent with the increase of surface RMS ( Figure 9 a). The samples treated at 60 °C-55% RH and 60 °C-95% RH showed a typical tiered growth of RMS and a sharp increase under 1.5 wt % of SiO 2 nanoparticles, displaying a WCA close to 150°. The transition started under the conditions of 1.5 wt % of SiO 2 nanoparticles and a WCA close to 150°. The special tiered growth of RMS should be interpreted as follows. Curing time required at 60 °C was longer than that at 80 °C so that the migration of PDMS could achieve the more homogeneous morphologies. When the content of SiO 2 nanoparticles reached 2.5 wt %, in the sample treated at 80 °C-55% RH, RMS continued to increase to about 270 nm; in the samples treated at 60 °C-55% RH and 60 °C-95% RH ( Figure 9 a), RMS sharply increased, thus resulting in a further increase of WCA and water droplets easily rolling off the surface with an SA of only about 8°. All the samples showed a typical “Cassie” state ( Figure 9 b–d). The changes should be ascribed to the special NAS similar to the sample containing 2.0 wt % of SiO 2 nanoparticles ( Figure 10 (a-2.0) and (b-2.0)) and our previous studies. Moreover, the precise transitions of water droplets from a “Wenzel” state to a “Cassie” state for samples treated at different conditions are shown in red dash circles in Figure 9 b–d. Meanwhile, the profiles of water droplets for WCA and SA on different samples after being left in room conditions for about 6 months were shown in Figure 11 . All the WCAs were close to 150° and water droplets could not slide off the surfaces even after 180° reversal of the samples if the samples were in a “Wenzel” state shown in the left columns of Figure 11 a–c. However, water droplets began to slide off the surfaces due to the repellency of air bubbles underneath the water droplets if the samples were in a “Cassie” state, as shown in the right columns of Figure 11 a–c. WCAs and the sliding behavior of water droplets on these samples showed little change compared with the as-prepared samples in Figure 9 . Therefore, these samples showed a stable hydrophobicity or superhydrophobicity. This stability should be attributed to the complete cross-linking of PDMS after thermal treatment as demonstrated in Figure 2 , Figure 3 and Figure 4 . Thus, these structures, RBS, TPS, and NAS, were achieved with one-step treatment within 90 min. The temporal changes of WCA on hydrophobic or superhydrophobic surfaces usually show only a little change, for example, less than 10° decrease in 25 h for samples in the reference [ 19 ], and less than 5° change in 30 days for superhydrophobic samples in previous publications [ 47 ]. In consideration of different kinds of structures in this manuscript, the temporal changes of WCA on samples with RBS, TPS, and NAS are shown in Figure 12 . However, WCAs on RBS (0.0 wt % SiO 2 ) in Figure 12 a,b decreased about 17° with 20 min. This obvious decrease should be attributed to the firm adhesion of water droplets on RBS, and the three-phase contact line showed almost little change during the evaporation of water droplets. As for WCA on RBS (1.5 wt % SiO 2 ) in Figure 12 a,c, it showed only about 5° change, which should result from the shorter length of the three-phase contact line with a higher initial WCA. Meanwhile, the WCAs within 20 min for samples with lower contents of SiO 2 (less than 2.0 wt %) showed a steady decrease, which should be due to the samples in a “Wenzel” state and the firm adhesion of water droplets on RBS and TPS. When the contents of SiO 2 increased to 2.0 wt % and 2.5 wt %, WCA showed wavy change, which should be due to the samples in a “Cassie” state. With a “Cassie” state, air bubbles underneath the water droplets would escape from the edge of the three-phase contact lines during the evaporation of water droplets. The release of air bubbles would shorten the length of the three-phase contact line (blue dash lines in Figure 12 d) and result in an increased WCA shown as red dash upward arrows (2.0 wt %) and red downward arrows (2.5 wt %) in Figure 12 a."
} | 7,378 |
21259034 | PMC3082026 | pmc | 2,034 | {
"abstract": "Peptides that bind to silkworm-derived silk fibroin fiber were selected from a phage-displayed random peptide library. The selected silk-binding peptides contained a consensus sequence QSWS which is important for silk-binding as confirmed by binding assays using phage and synthetic peptides. With further optimization, we anticipate that the silk-binding peptides will be useful for functionalization of silk for biomaterial applications.",
"introduction": "Introduction Silkworm silk is composed of two classes of proteins fibroin and sericin. It is one of the oldest and most versatile biomaterials with many favorable properties such as biocompatibility and biodegradability. Silk is also easily processed into various forms such as fibers, yarns, textiles, sponges, and films (Altman et al. 2003 ). Not only has silk been extensively used as sutures in medical settings for centuries but silk-based biomaterials have been developed for modern biomedical applications such as matrices for tissue engineering (Altman et al. 2003 ; Wang et al. 2006 ). Molecular functionalization of silk is an attractive strategy for further expanding the versatility of silk as biomaterials. However, relatively few previous reports on chemical functionalization of silk-derived materials can be found. Most of these efforts involve covalent chemical modifications of the silk proteins using reactive reagents (Murphy and Kaplan 2009 ). Genetic modifications of silkworms have also been recently reported (Mori and Tsukada 2000 ). However, genetically modified organisms are still subject to stringent regulatory approval, and the possible modifications are limited to those that can be genetically encoded and nontoxic to the host silkworm. New universal molecular tools such as peptide tags that can be fused to other functional moieties (e.g. proteins, organic molecules, or nanoparticles) to impart new functions to silk via noncovalent interactions would be highly attractive. Affinity selection using phage libraries that display random peptides fused to a coat protein has proven to be a highly successful strategy for discovering potent peptide ligands (Pande et al. 2010 ). Many peptide ligands with affinities to various biomolecules (Pande et al. 2010 ) and inorganic (Sarikaya et al. 2003 ) and organic materials (Serizawa et al. 2005 ) have been reported. Here, we describe our initial results of the selection and screening of peptide ligands that bind to silk fibroin fibers from a phage-displayed random peptide library. Sequencing of the selected phage clones revealed a short consensus sequence QSWS which appears to be important for binding based on binding assays using phage and synthetic peptides. We envision that these peptide ligands will be useful for noncovalent functionalization of various silk-based biomaterials.",
"discussion": "Results and discussion Selection and identification of silk-binding peptides Results of the phage selection process are summarized in Table 1 . Stringency of selection was increased each round by decreasing the binding time and/or increasing the number of washes, except for the fifth round in which the binding and washing conditions were kept the same as the fourth round. We noticed that the number of phage recovered decreased by an order of magnitude after the fourth round, probably due to the higher stringency. However, the phage recovery increased in the fifth round suggesting that the clones with higher affinities to silk fiber were being selected. We randomly picked and analyzed the sequences of 12 clones from the phage selected after the third round. However, the result revealed no consensus or enriched sequences (data not shown). On the other hand, 11 clones analyzed after the fifth round revealed two sequences YN42 and YN43 that appeared three times each (Table 2 ). Moreover, both of these enriched sequences contained a common tetrapeptide sequence QSWS. A closer examination of the selected sequences indicates other characteristic patterns such as the high frequency of tryptophan residues. Table 2 Amino acid sequences of the phage-displayed peptides after silk affinity selection Phage a \n Number of clones Sequence YN42 3 SYTFHWH QSWS S YN43 3 \n QSWS WHWTSHVT YN41 1 WTWRWAHVTNTR YN48 1 QDVHLTQQSRYT YN49 1 HKAHEYDPWISP YN50 1 SYSQHYGIPNPW YN52 1 SSWQMSWSWMGS \n a The sequenced phage clones were randomly picked from the fifth round selection. A total of 11 clones were sequenced \n Characterization of silk-binding peptides We focused our initial studies on one of the two highly enriched peptides YN42, because both peptides (YN42 and YN43) contained the common QSWS motif. First, the silk-binding property of the phage clone displaying the YN42 peptide was confirmed by phage binding assay. YN42 phage were recovered in 100 times better yield than the unselected phage population, indicating that the selection indeed enriched phage displaying silk-binding peptides (Fig. 1 ). To gauge the contributions, if any, of the QSWS motif, two single mutants of YN42 were constructed that display QAWS or QSAS instead of QSWS. While the recovery of the phage YN42/QAWS was moderately lower (by approximately 30%) compared to YN42, the recovery of YN42/QSAS phage were reduced by more than 90%, indicating that the tryptophan in QSWS motif plays a significant role in silk-binding. Fig. 1 Phage binding assay. Silk fibroin fibers (10 mg) were incubated with phage (5 × 10 7 pfu) in 1 ml solution. After washing as described in “ Materials and methods ” section, bound phage were eluted in 1 ml buffer and titered. The assays were performed twice for each phage clone and the error bars indicate the range of the measurements \n Results of the peptide binding assays using synthetic biotinylated peptides generally agreed with the phage binding assay (Fig. 2 a). In particular, a control peptide with the same amino acid composition but with a randomly shuffled amino acid sequence (pepYN42/Shuffled) showed significantly lower affinity to silk compared to pepYN42, indicating that the sequence, not the amino acid composition, is important for the observed silk-binding property. To demonstrate how silk-binding peptides can be used to functionalize silk, silk fibers treated with pepYN42 and streptavidin-HRP conjugate were stained by an insoluble substrate (3,3′,5,5′-tetramethylbenzidine, TMB) for HRP. While the fibers treated with pepYN42/Shuffled were only minimally stained, the fibers treated with pepYN42 were stained significantly (Fig. 2 b). Concentration dependent binding assay using pepYN42 suggests an apparent submicromolar affinity of the peptide to silk fibroin fibers (Fig. 3 ). Fig. 2 Synthetic peptide binding assay. Peptides with the amino acid sequence corresponding to the one displayed on the YN42 phage and its variants were synthesized and conjugated to biotin via GK (biotin) at the C -terminus. pepYN42/Shuffled: H 2 N-WYSHSHSQTSFWGK(biotin)-CONH 2 . a Peptides (3 μM) were incubated with silk fibers (1 mg) and the bound peptides were quantified by ELISA as described in “ Materials and methods ” section using ABTS as the chromogenic substrate. The data shown are averages of four replicate experiments and the error bars indicate ±SD. b Visual detection of biotinylated silk-binding peptides using streptavidin-HRP conjugate and an insoluble HRP substrate (TMB). Left : silk fibers treated with pepYN42; center : untreated fibers; right : fibers treated with pepYN42/Shuffled \n Fig. 3 Concentration dependent binding of pepYN42 to silk fibers. Each data point represents an average of two independent experiments and the error bars indicate the range of the two measurements\n\nDiscussion Our goal was to discover small peptides with significant affinity to silk fibroin fibers that can be used to noncovalently modify silk-based biomaterials with various functional units such as enzymes, hormones, or signaling molecules. We identified YN42 as one of the two highly enriched peptides in silk affinity selection from a phage displayed peptide library. The YN42 peptide exhibited significant affinity to silk fibroin fiber as a coat-protein fusion displayed on M13 phage, as well as a C -terminally biotinylated synthetic peptide. Therefore, YN42 or its derivatives are promising candidates for generic silk-binding tags that can be used to impart new functions to silk-based materials. We anticipate that our silk-binding peptides can be used in a number of forms, for example, as synthetic peptide conjugate demonstrated in this work and as genetically encoded fusion tags to other proteins. Further modifications of the peptide sequence may enable fine-tuning of the silk-binding property for individual applications."
} | 2,176 |
35518755 | PMC9054385 | pmc | 2,035 | {
"abstract": "Taking advantage of the triboelectrification effect and electrostatic induction, triboelectric nanogenerators (TENGs) provide a simple and efficient path to convert environmental mechanical energy into electric energy. Since the generation of surface charges and their density on triboelectric materials are the key factors in determining TENG performance, many efforts have been undertaken to engineer the structures and chemistry of triboelectric materials. Among others, dielectric control of triboelectric materials is an emerging approach because the dielectric constant is intimately correlated with the capacitance of materials. In this regard, we prepared porous polydimethylsiloxane (PDMS) composites decorated with Au nanoparticles (NPs), which was designed to engineer the compressibility and dielectric constant of PDMS elastomer. To this end, a polydopamine layer was synthesized on the PDMS surface to facilitate the homogeneous deposition of Au NPs. Unlike untreated PDMS sponges, Au NPs were efficiently coated onto polydopamine-treated PDMS sponges to increase the dielectric constant. When the resulting porous NP-PDMS composites were assembled into TENG devices, the electrical output of the TENGs initially improved but decreased with the amount of Au NPs. This trade-off relationship has been discussed in terms of charge generation on the air surface and pores of NP-PDMS composites based on a recent experimental model.",
"conclusion": "Conclusions In this study, we demonstrated a simple and versatile method for enhancing TENG performance with composite PDMS sponges with Au NPs. Since the direct coating of Au NPs onto the PDMS surface was severely restricted by the chemical inertness of PDMS, PDA polymers have been introduced as an adhesive interface to promote homogenous adsorption of Au NPs on the inner surface of PDMS pores. Owing to the PDA layer, the amount of adsorbed NPs on the porous PDMS elastomers could be easily engineered by the immersion time, which resulted in a monotonic increase in the dielectric constant of the PDMS/PDA/Au composite sponges. The simultaneous control of both ε and d values with the composites PDMS/PDA/Au sponges substantially modified the TENG performance. When the amount of Au NPs was relatively small, the adsorbed NPs enhanced the electric output because of the increased dielectric constant as well as the additional charge generation inside the pores. However, further attachment of Au NPs to the PDMS surface deteriorated the TENG performance because they could interfere with contact charge generation in the triboelectric process. Because a variety of inorganic NPs could be adsorbed onto the PDA layer, further improvement in the TENG performance can be expected by introducing other high dielectric constant NPs to the porous PDMS structures, which would provide a better understanding of the triboelectricity and help the development of high-performance TENG devices.",
"introduction": "Introduction The extraction of electric energy from environmental and biological sources has attracted a great deal of attention in recent years in order to supply power for micro-electronics as well as solve the global energy crisis. 1–7 Among many energy-harvesting techniques, triboelectric nanogenerators (TENGs) represent a simple and efficient approach that is based on triboelectrification and electrostatic induction. 3,4,8–15 In TENGs, surface charge transfer occurs by mechanical contact between two materials with different triboelectric polarities. 3,4,16 When separated, the two oppositely charged materials induce an electric potential difference, which in turn drives the electron flow through an external load. 3,4,8–15 Since the electric potential difference in TENGs is regulated by the distance between the two charged surfaces, electricity can be generated in response to the successive contact and separation of component structures by mechanical vibration. 3,4,8–15 Understanding that the generation of surface charges and their density on the triboelectric materials are the key factors in determining the output performance of TENGs, many research groups have engineered interfacial structures and/or surface chemistry to optimize the charge transfer process. 8–15 For instance, the formation of micro-patterns such as pyramid, cube, and lines, 8,9 utilization of electro-spun nanofibers, 10,11 and introduction of inorganic structures 12,13 have been employed to increase the contact area. In addition, surface modification with self-assembled monolayers (SAMs) 14 or polymers 15 has engineered the triboelectric sequence of the given materials for better control of the triboelectric properties. Recently, it has been further demonstrated that the charge density on triboelectric materials is closely correlated with their capacitance. 17–25 Since capacitance, which is the capability to hold charges at a given potential, is proportional to the ratio of the dielectric constant ( ε ) to the thickness of triboelectric materials ( d ), the ratio of ε / d has been adjusted to optimize the electric output from TENGs. 17–25 For instance, polydimethylsiloxane (PDMS) elastomers have been blended with inorganic nanoparticles (NPs), 17,18 or new polymers such as poly( tert -butyl acrylate)-grafted polyvinylidene difluoride copolymers 19 have been synthesized to increase the dielectric constant. With the increased dielectric property, the surface charges on the triboelectric materials increased substantially, which resulted in improved power generation. 17–19 Comparable approach for the dielectric control has been also investigated in piezoelectric nanogenerators to enhance their output performance. In particular, inorganic fillers such as carbon nanotubes, graphene oxide, metal nanoparticles have been introduced in poly(vinylidene fluoride) (PVDF) matrix to increase the dielectric property of the PVDF composites and also to induce piezoelectric crystal phases. 26–28 Alternatively, nano- or micro-scale pores have been introduced into PDMS elastomers to reduce the thickness of PDMS under external pressure. 20,21 Because of the fact that the charge transfer occurs under mechanical contact between two materials, namely, PDMS and metal, the compressed PDMS can accommodate more surface charges due to the increased capacitance, in comparison with a flat film-based PDMS. From this perspective, one can further anticipate that the ε / d value could be better engineered by introducing inorganic NPs with a high dielectric constant and pores simultaneously in the triboelectric materials. Recently, Baik et al. prepared Au NP-embedded mesoporous PDMS by a simple evaporation method, in which Au NPs were placed at the bottom side of the pores by gravitational force. 22 Alternatively, PDMS elastomers were blended with inorganic NPs and NaCl (or sugar) powder, from which pores were created by removing NaCl (or sugar) by the H 2 O rinsing step. 23,24 These porous NP-PDMS composites are interesting because they can provide increased capacitance and additional charge can be generated from each pore by mechanical contact between the NPs and the inner surface of the pores. Understanding that charge generation from individual pores in the porous NP-PDMS composites is an important factor for enhancing the TENG performance, it is necessary to decorate the surface of PDMS pores by inorganic NPs. However, the direct attachment of inorganic NPs to the surface of PDMS pores is strongly limited by the chemical inertness of PDMS. Although plasma treatment has been previously utilized for the surface modification of PDMS for the NP attachment, 29,30 this method cannot be applicable for the porous structures. In this study, in order to address this issue, we utilized a polydopamine (PDA) layer as an adhesive interface for PDMS and Au NPs. Dopamine is a small molecule consisting of catechol and amine and can be spontaneously polymerized into a thin film onto numerous types of surfaces. 31–33 Since PDA coating is based on the diffusion and in situ polymerization of dopamine monomers, PDA coating would be suitable for the surface modification of porous materials. 31,32 In addition, the coated PDA thin film has intrinsic metal binding ability owing to the catechol moiety. 31–33 This can facilitate the deposition of Au NPs onto PDA-coated PDMS. It also needs to be noted that the utilization of PDA for the uniform coating of NPs on the inner surface of PDMS pores has not been reported in TENG devise. Based on this perspective, we present a simple and versatile method for the preparation of NP-decorated PDMS sponges with the aid of a PDA polymer. In particular, we demonstrate that the amount of NPs deposited on the PDMS sponges could be engineered by simply adjusting the immersion time of PDA-coated PDMS sponges into an aqueous solution of Au NPs. We found that the open-circuit voltage and short-circuit current initially increased with the amount of Au NPs, but decreased under higher NP content. This trade-off relationship has been discussed in terms of charge generation on the air surface and pores of the NP-PDMS composites.",
"discussion": "Results and discussion To enhance the capacitance of triboelectric materials, we attempted to simultaneously introduce porous structures and inorganic nanoparticles to triboelectric materials to increase the value of the ε / d ratio. For this purpose, we first prepared porous PDMS elastomers using the sugar-template method. Although several techniques have been proposed for the preparation of porous PDMS elastomers, 17,20–24,34 the sugar-template method represents a simple, cost-effective, and shape/size-tunable approach. 20,34 In this method, the sugar template is first prepared by pouring the kneaded sugars into a pre-made mold. By infiltrating PDMS prepolymers into the pores of the sugar template, porous PDMS elastomers can be replicated by curing the PDMS prepolymers, followed by rinsing the sugar template with water. As shown in Fig. 1a , the replicated PDMS sponge can be repeatedly compressed and recovered without mechanical deterioration. In addition, the white appearance of PDMS, which would be caused by light scattering from the internal surface of the PDMS elastomer, may indicate the presence of porous structures. When the micro-structures of the PDMS replica were examined by FE-SEM ( Fig. 1b ), the presence of pores (300–500 μm) can be clearly discernible from the three-dimensionally interconnected PDMS frame. It should be noted that the SEM image in Fig. 1b was collected from the air surface of the PDMS sponge; however, mainly identical structure was observed from the cross-sectional area of the PDMS replica. Fig. 1 (a) Photographs of the sugar template and replicated PDMS elastomer. (b) Plan-view SEM image of the porous PDMS sponge. (c and d) TEM image (c) and UV-Vis spectrum (d) of the synthesized Au NPs. The DLS results with Gaussian fittings before (red color) and after (blue color) centrifugation-based purification are shown in Fig. 1d . Confirming the formation of the porous PDMS elastomer, we synthesized Au NPs using the citrate-reduction method as an inorganic counterpart. 35 Although other inorganic NPs could be utilized, Au NPs were selected as model NPs because of their chemical stability, easy synthesis, and structural uniformity. As shown in the TEM image ( Fig. 1c ), the synthesized Au NPs were uniform in size and shape. Since the Au NPs could be aggregated during centrifugation-based purification process, we compared the size distribution of Au NPs before and after the centrifugation. In the DLS result (inset of Fig. 1d ), the size distribution of Au NPs was remained the same after the centrifugation, ruling out the possible NP aggregation. From the Gaussian fitting of the DLS result, the average diameter of Au NPs was determined to be 12 nm. In addition, Au NPs exhibited a well-known localized surface plasmon resonance (LSPR) peak at 520 nm ( Fig. 1d ), 36 which further confirmed NP synthesis. In the UV-Vis spectra, the absorbance peak of Au NPs at 520 nm was adjusted as 1.0. Since the molar extinction coefficient of 12 nm Au NPs is 1.89 × 10 8 M −1 cm −1 , 37 the NP concentration was calculated as 5.3 nM by Lambert–Beer law. It needs to be noted that the synthesized Au NPs have negative charges on the surface (zeta potential = −33.2 mV) owing to the citrate molecules. This strongly restricts the adsorption of NPs onto the hydrophobic surface of the PDMS sponges. In this regard, we introduced polydopamine (PDA) layer as an adhesive interface for PDMS and Au NPs from the following experiment. To synthesize the PDA polymer, PDMS sponges ( Fig. 1b ) were immersed in a dopamine solution in Tris buffer (pH = 8.5) for 4 h. Note that dopamine monomers spontaneously undergo oxidative polymerization in alkaline conditions to form a PDA coating on virtually all types of surfaces. 31,32 In particular, it has been reported that PDA can be coated on the PDMS surface to reduce the hydrophobicity of PDMS as well as to promote chemical and/or physical interactions with other molecules by the catechol and amine groups in PDA molecules. 38–40 Therefore, it can be anticipated that the PDA layer behaves as an adhesive interface for binding Au NPs and PDMS surfaces. After the polymerization, the white color of the PDMS sponge turned brown, which strongly indicated the synthesis of PDA on the PDMS surface. The same brown color also appeared when the cross-sectioned PDMS surface was examined. Therefore, it is reasonable to consider that dopamine monomers diffused inside the PDMS sponges for the homogenous PDA coating on the inner surface of the pores. To further confirm the synthesis of the PDA polymer, ATR-FTIR spectra were measured ( Fig. 2a ). For pristine PDMS sponges (red color), characteristic vibrational peaks corresponding to Si–CH 3 , Si–O–Si, and Si–CH 3 groups were detected at 1260, 1011, and 790 cm −1 , respectively, (denoted as circles). 39 After the synthesis of PDA polymers on a PDMS sponge (blue color), the characteristic PDMS peaks were maintained (denoted as circles), and new vibrational peaks appeared at 1621 and 1508 cm −1 (denoted by arrows), which can be assigned to phenylic C \n \n\n<svg xmlns=\"http://www.w3.org/2000/svg\" version=\"1.0\" width=\"13.200000pt\" height=\"16.000000pt\" viewBox=\"0 0 13.200000 16.000000\" preserveAspectRatio=\"xMidYMid meet\"><metadata>\nCreated by potrace 1.16, written by Peter Selinger 2001-2019\n</metadata><g transform=\"translate(1.000000,15.000000) scale(0.017500,-0.017500)\" fill=\"currentColor\" stroke=\"none\"><path d=\"M0 440 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z M0 280 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z\"/></g></svg>\n\n C stretching vibrations and N–H shearing vibrations of the PDA polymer. 40 Fig. 2 (a) ATR-FTIR spectra of PDMS sponges with the introduction of PDA polymers and Au NPs. The PDMS/PDA was immersed in the NP solution for 24 h. (b) Time-dependent UV-Vis spectra after the immersion of PDMS (red color) and PDMS/PDA (blue color) in the NP solution. (c) Photographs of PDMS (i), PDMS/PDA (ii), PDMS/PDA/Au with different immersion times (iii), and PDMS/Au in the absence of PDA (iv). (d–f) Cross-sectional FE-SEM images of PDMS/PDA/Au sponges after immersion in NP solution for different times. (g) Cross-sectional FE-SEM images of PDMS/Au sponges after immersion in NP solution for 60 h. To introduce Au NPs into the PDMS sponges, PDA-coated PDMS sponges (denoted as PDMS/PDA for simplicity) were immersed in an aqueous solution of Au NPs for different periods of time (0–60 h). In this set of experiments, both PDMS sponges before and after PDA synthesis were immersed into the NP solution, and the absorbance peak of Au NPs at 520 nm was monitored at regular time intervals ( Fig. 2b ). It needs to be noted that the UV-Vis spectrum of Au NPs in H 2 O was very stable and remained the same for an extended period of storage (more than one week). On the other hands, when PDMS/PDA or PDMS sponges were immersed in the NP solution, the absorbance peak of the NP solution was gradually reduced during the immersion. Since the adsorption of NPs onto the PDMS sponges reduces the concentration of Au NPs in a solution state, the decrease in peak intensity can be indicative of the amount of adsorbed NPs. In the case of PDA-coated PDMS, the absorbance value of colloidal Au NPs monotonically decreased and then saturated upon immersion (blue color in Fig. 2b ). From the decrease of UV-Vis spectra, the concentration of Au NPs in PDMS/PDA/Au composites were evaluated as 0.013, 0.017, 0.023, 0.033, 0.038, 0.037 wt% with the immersion time of 6 h, 12 h, 24 h, 36 h, 48 h and 60 h, respectively. On the other hand, for pristine PDMS without PDA layer (red color), the absorption value was more or less maintained after the initial decrease. This comparison verified that PDA coating is necessary for the attachment of Au NPs onto the hydrophobic PDMS surface. The adsorption of Au NPs was further supported by the color changes of the PDMS sponges ( Fig. 2c ). As discussed, the color of the PDMS sponges changed from white ((i), in Fig. 2c ) to pale brown (ii) after the PDA coating. Upon immersion in the NP solution, the color of PDMS became darker with the immersion time (iii) owing to the increased amount of Au NPs. In stark contrast, the color of PDMS was slightly changed to light purple in the absence of PDA coating due to the restricted adsorption of NPs onto the hydrophobic PDMS surface (iv). This slight color change can be ascribed to physically arrested NPs during water evaporation. The same result was also obtained by analyzing the cross-sectional FE-SEM images. In the presence of the PDA layer, the density of the adsorbed Au NPs on the PDMS surface noticeably increased with longer immersion time ( Fig. 2d–f ). However, the adsorption of NPs was strongly restricted on the hydrophobic PDMS surface without the adhesive PDA coating ( Fig. 2g ). Based on the results, PDA-coated PDMS sponges with the adsorbed Au NPs will be simply denoted as PDMS/PDA/Au NPs hereafter. We also noted that the PDMS/PDA/Au composites showed the same characteristic FTIR peaks of PDA and PDMS (green line in Fig. 2a ). After verifying the preparation of PDMS/PDA/Au sponges, we prepared TENG devices by assembling the composite PDMS sponges, Cu electrodes, and other components, as shown in the schematics of Fig. 3 (see details for the Experimental section). In the TENG structure, there was an air gap (∼3 mm) between the top electrode and the air interface of the PDMS sponges to allow successive contact and separation under periodic compressive force. In addition, a flat PDMS film with the same thickness as the PDMS sponges (5 mm) was utilized for the TENG device for comparison. Fig. 3a shows the open-circuit voltage ( V OC ) of the TENG with a series of composite PDMS sponges. As expected, the flat PDMS film without pores produced a very low output voltage with an average value of 33 V. With the successive introduction of pores, PDA coating, and Au NPs into the PDMS elastomers, the output performance of TENGs was systematically increased. In particular, PDMS/PDA/Au sponges (immersion time in NP solution = 24 h) produced V OC with an average value of 180 V, which was enhanced by ∼5.5-fold in comparison with flat PDMS. Also note, PDMS/Au sponges without PDA coatings enhanced the output voltage (65 V) in comparison with that of the pristine PDMS sponges (47 V), indicating the positive effect of the physically adsorbed Au NPs. A similar trend was observed in the short-circuit current ( I SC ) measurement. In Fig. 3b , one can clearly observe the gradual current enhancement, which was accompanied by the introduction of pores, PDA coating, and Au NPs in the PDMS elastomers. Unfortunately, the precise measurement of I SC was rather limited by (a) the sensitivity of our machine to detect the low current from TENGs and (b) non-identical local contacts inside the pores at each contact-separation step (this will be discussed later), which resulted in the non-uniform current response in Fig. 3b . Nevertheless, the gradual current enhancement from the corresponding TENG devices was identical to many independent preparations/measurements, confirming the enhanced TENG performance with pores and inorganic NPs. Fig. 3 (a) Open-circuit voltage and (b) short-circuit current of TENG devices prepared by flat PDMS (black), PDMS sponge (red), PDMS/PDA (blue), PDMS/Au (green), and PDMS/PDA/Au (purple) composites. (c) Open-circuit voltage and (d) short-circuit current of the TENG devices prepared by PDMS/PDA/Au composites under different immersion times. The structure of the TENG device is also included in the scheme. To better understand the result, it should be noted that (a) the enhanced output performance is correlated with the charge density created on triboelectric materials during physical contact 3,4,8–15,17–25 and (b) TENG structures can be treated as an analogy of flat-panel capacitors. 17–19,22–25 This indicates that the created charge density ( σ ) is proportional to the capacitance of triboelectric materials ( C ). 17–19,21–25 Moreover, because the capacitance is proportional to the ratio of the dielectric constant ( ε ) to the thickness ( d ) of triboelectric materials as C ∼ ε / d , experimental control on ε and d can increase the charge density as well as the TENG performance. 17–19,21–25 In our experiment, the introduction of pores can effectively reduce the thickness of PDMS sponges under external pressure. In addition, the introduction of Au NPs with a higher dielectric constant ( ε Au = 6.0) to PDMS ( ε PDMS = 3.0) can increase the dielectric constant of composite materials ( ε ). 17,18,22 To verify the enhancement in the dielectric constant, we measured the frequency-dependent dielectric permittivity of the PDMS composites over the frequency range of 10 3 to 10 7 Hz ( Fig. 4a ). For the porous PDMS sponges (green color), the average dielectric constant was 2.94. On the other hand, with the introduction of PDA layer (blue color) and Au NPs (red color, immersion time = 24 h), the dielectric constants of the composites PDMS sponges were gradually increased to 3.21 and 3.70, respectively. Meanwhile, the dielectric loss of the PDMS sponges were remained nearly unaltered after the addition of PDA layer and Au NPs ( Fig. 4a ), which indicated that the degree of conductance loss was not deteriorated. Therefore, it can be validated that ε / d has been collectively engineered to enhance the capacitance of composite PDMS sponges and the corresponding TENG performance without increasing dielectric loss. Fig. 4 (a) Frequency-dependent dielectric constant and loss tangent of PDMS sponge (green), PDMS/PDA (blue), and PDMS/PDA/Au (red) composites. (b) Average dielectric constant of PDMS/PDA/Au composites with different immersion times in aqueous solution of Au NPs. (c) Open-circuit voltage and (d) short-circuit current of gapless TENG devices based on flat PDMS (black), PDMS sponge (red), and PDMS/PDA/Au composites (blue) with immersion time = 24 h. To better study the influence of Au NPs, we utilized the composite PDMS/PDA/Au sponges prepared under different immersion conditions in the NP solution. As shown in Fig. 3c and d , both the output voltage and current initially increased and then decreased with the immersion time. According to the previous discussion, the introduction of more Au NPs into PDMS sponges allows for an increase in the dielectric constant, thereby enhancing the TENG performance. Indeed, the dielectric constant of the PDMS/PDA/Au sponges monotonically increased with immersion time ( Fig. 4b ). Also note, all the PDMS/PDA/Au composites in Fig. 3c and d have the same pore structures and thicknesses. Therefore, the decreased TENG performance under higher NP content could not be explained by the ε / d ratio value alone. Nevertheless, the trade-off trend can still be understood in terms of the created charge density. As discussed, the introduction of Au NPs can increase the charge density mainly by increasing the dielectric constant of the PDMS/PDA/Au composites. However, the presence of Au NPs on the surface of PDMS alternatively reduced the contact area between the top electrode and the air interface of the PDMS/PDA sponges. As shown in Fig. 2 , the adsorbed Au NPs on the surface of the PDMS composites increased with the immersion time (further SEM images are shown in ESI Fig. S1 † ). Because the adsorbed Au NPs interfere with the effective triboelectric contact between PDMS and the Cu electrode, the high density of Au NPs on the surface would reduce the charge generation process. Consequently, Au NPs can be considered as either enhancers or reducers for TENG performance depending on the concentration. The triboelectric charges were produced not only from the air surface but also inside the individual pores. To reveal the contribution of porous structures to the TENG performance, we prepared the same TENG devices as in the schematics of Fig. 3 , but removed the air gap by directly attaching the air interface of the PDMS composite to the top electrode. Then, the periodic compressive force was applied to the gapless TENG devices to measure the output voltage and current ( Fig. 4c and d ). In the case of the flat PDMS film (black color), both V OC and I SC were quite negligible because the contact-separation process, which is mandatory for charge generation in triboelectric materials, was essentially inhibited. When a PDMS sponge is utilized (red color), a local contact-separation process can be allowed for each pore to produce additional charges on the pore surface. However, triboelectrification between homogenous material surfaces is inefficient, which results in a small enhancement in the electric output. In stark contrast, the PDMS/PDA/Au composite sponges produced strongly enhanced V OC and I SC (blue color), which clearly revealed the positive effect of Au NPs on the TENG performance. In the case of the PDMS/PDA/Au composites, the surface of each pore was decorated with Au NPs. Therefore, local contact between the Au NPs and PDA layer can be induced upon compression, which results in efficient charge transfer from Au NPs to the pore surface. As a result, the charge density in the pores could be strongly increased to improve the TENG performance. Nevertheless, it must be noted that the local contact inside the pores of PDMS/PDA/Au composites cannot be identical at each contact-separation cycle. Because the size and orientation of pores are irregular (see Fig. 1b ), the vertical compression of the PDMS composites allows not only squeezing the pores but also sliding between pore surfaces. As a result, different types of contacts between (i) NPs and PDA, (ii) NPs and NPs, and (iii) PDA and PDA could be made at each contact-separation step. Since only the physical contact between NPs and PDA surfaces could efficiently contribute to the charge transfer, the generated charge density in the pores would be different at each contact-separation cycle. Therefore, the measured current would have different intensities in Fig. 3 and 4 . In the similar way, the open-circuit voltage would be also affected by the different amount of charge generation in the pores, resulting in the non-uniform minimum value of V OC in Fig. 4c . Comparable results on gapless TENGs have been previously reported by the Baik group. 22 In their study, Au NPs were introduced to mesoporous PDMS sponges by evaporation. Since no specific chemical interaction has been utilized, Au NPs were placed at the bottom side of the pores by gravitational force. In this configuration, the local contact in the pores allowed charge transfer from Au NPs to PDMS, which produced negative charges on the top surface of the pores and negative charges on the Au NPs placed at the bottom side of the pores. As a result, the opposite charges were aligned with respect to each other, which brought about electrostatic induction to the electrodes for the TENG enhancement. However, in our study, Au NPs were homogeneously attached to the PDMS pores by the adhesive PDA layer, and the resulting positive/negative charges would be randomly distributed inside the pores. In this regard, it is more or less unclear how the pore charges could contribute to the electric output of TENG devices at the current stage. To assess the prospects of the prepared TENGs, the voltage, current, and output power of TENGs with PDMS/PDA/Au composites (immersion time = 24 h) were measured under various external loads. As shown in Fig. 5a , the voltage increased with increasing external resistance, while the current decreased. As a result, the maximum power of the TENG reached a value of 115 μW at an external resistance of 10 MΩ. In addition, when the durability of the TENGs was examined under periodic contact-separation processes, the electric output was stable without noticeable degradation over 1500 cycles ( Fig. 5b ). This result strongly indicates the long-term stability of the composite PDMS/PDA/Au sponges. To better demonstrate the practical aspects, the TENGs with composite PDMS/PDA/Au sponges were connected to an array of commercial light-emitting diodes (LEDs) through a bridge rectifier. The power generated from the TENGs was able to turn on LEDs without charging capacitors. As shown in ESI Fig. S2, † the number of lighted LEDs initially increased but decreased with the immersion time, which additionally supported the discussed dual function of Au NPs in Fig. 3 . Fig. 5 (a) Output voltage (black), current (red), and power (blue) of the TENG with composite PDMS/PDA/Au sponge on the external load resistance. (b) Stability and durability test of the TENG under periodic contact-separation processes."
} | 7,484 |
26776218 | null | s2 | 2,036 | {
"abstract": "Technological advances are making large-scale measurements of microbial communities commonplace. These newly acquired datasets are allowing researchers to ask and answer questions about the composition of microbial communities, the roles of members in these communities, and how genes and molecular pathways are regulated in individual community members and communities as a whole to effectively respond to diverse and changing environments. In addition to providing a more comprehensive survey of the microbial world, this new information allows for the development of computational approaches to model the processes underlying microbial systems. We anticipate that the field of computational microbiology will continue to grow rapidly in the coming years. In this manuscript we highlight both areas of particular interest in microbiology as well as computational approaches that begin to address these challenges."
} | 228 |
34461907 | PMC8406616 | pmc | 2,037 | {
"abstract": "L-valine is an essential amino acid that has wide and expanding applications with a suspected growing market demand. Its applicability ranges from animal feed additive, ingredient in cosmetic and special nutrients in pharmaceutical and agriculture fields. Currently, fermentation with the aid of model organisms, is a major method for the production of L-valine. However, achieving the optimal production has often been limited because of the metabolic imbalance in recombinant strains. In this review, the constrains in L-valine biosynthesis are discussed first. Then, we summarize the current advances in engineering of microbial cell factories that have been developed to address and overcome major challenges in the L-valine production process. Future prospects for enhancing the current L-valine production strategies are also discussed.",
"conclusion": "Conclusions and perspectives The branched-chain amino acid, L-valine has wide applications and its market capacity requirement constantly increases, but the production level of this essential biochemical is still far from satisfactory to meet the industrial demands. In this perspective, we describe several strategies that comprise system metabolic engineering for producing L-valine. These strategies encompass carbon source and nutrient utilization, flux distribution at branch points, cofactors and energy requirements, regulatory engineering, product export, gene manipulation, system metabolic engineering, fermentation conditions and characteristics. As engineering of microbial cells for L-valine production is applied more widely, we expect these strategies can be effective in addressing the above challenges. Although, the use of renewable carbon source, sugar in fermentation with aid of microbial factories offers many advantages as can be seen from the relevant discussion above, several challenges need to be overcome for efficient production. Methanol is a promising raw material for the production of fuels and chemicals. Therefore, people have devoted a great deal of efforts to the design of microorganisms that utilize methanol as an unnatural methyl nutrition platform. The main industrial C. glutamicum , after reasonable design and experimental engineering, can be used as a methanol-dependent synthesis of methyl nutrition organisms. The cell growth of methanol-dependent strains depends on the co-utilization of methanol and xylose, most notably methanol is an essential carbon source [ 76 ]. The use of abundant and cheap carbon sources can effectively reduce production costs and improve economic feasibility. Acetate is a promising carbon source for realizing cost-effective microbial processes. Recently, some researchers have designed an E. coli strain to produce itaconic acid from acetate. In order to increase the yield, the acetic acid assimilation pathway and the glyoxylic acid shunt pathway are amplified through the overexpression and deregulation of pathway genes. After 88 h of fermentation, acetic acid quickly assimilated, and the resulting strain WCIAG4 produced 3.57 g L −1 itaconic acid (16.1% of the theoretical maximum yield). These efforts support that acetate may become a potential raw material for engineered E. coli production biochemistry [ 77 ]. At present,the average selling price of L-valine is $5.4 per kg and the current maximum yield of valine is about 172.2 g L −1 L-valine with a yield of 0.63 mol mol −1 of glucose for 24 h [ 59 ]. The raw materials such as glucose, formic acid and acetic acid are $0.772 per kg, $0.386 per kg and $0.463 per kg respectively. The raw materials such as formic acid and acetic acid obtained from industrial waste are even lower. Therefore, the formate and acetate converted from the reduction of carbon dioxide emissions has great potential and can be used as a sustainable raw material for the biological production of biofuels and biochemical substances. However, due to the toxicity of formate or lack of metabolic pathways, its use for the growth and chemical production of microbial species is limited. The formate assimilation pathway in E. coli has been constructed and applied to adaptive laboratory evolution to improve the use of formate as a carbon source under sugar-free conditions [ 78 ]. Optimizing the metabolic flow of engineering precursor is frequently complicated between target product and byproduct, gene expression and complex enzyme regulation. Although metabolic problems can be handled by regulating key enzymes, more often than not, the complexity of the branched-chain amino acids synthesis pathway requires more precise regulation. Gene regulation based on CRISPRi has emerged as a powerful tool in synthetic circuits. CRISPRi will have enormous potential as a primary means of technology to reconstruct gene regulatory networks. With the advancement of novel strategies and tools in system metabolic engineering, it is possible to achieve optimal production of L-valine on an industrial scale. The avenues that promise the major impact on L-valine production are: First, stain development by biosensor-mediated adaptive laboratory evolution and high-throughput screening (using biosensors specifically developed for L-valine) [ 73 ]. Second, applying the omics strategies, including transcriptomes, proteomes and metabolomes, the physiological information of L-valine producer strain will possibly add insight to hidden constraints. Third, 13 C-metabolic flux analysis combined with a genome-scale metabolic model (GSM) can lead to the development of optimal process through systematically identifying genetic targets for overexpressions, downregulation and deletions [ 79 ]. To this end, all these strategies will provide engineering with insights for improving microbial production for L-valine in the future. Nonetheless, one of the major challenges of metabolic engineering is to transfer the CO 2 fixation capacity of autotrophic organisms to heterotrophic organisms (such as E. coli ). The knowledge of more than 100 metabolic pathways of autotrophic E. coli can be used, which have been optimized for the production of fuels and chemicals, and are produced through CO 2 . Gleizer et al. in their 2019 contributions type of E. coli that can use CO 2 as the sole carbon source is designed. The author used metabolic engineering technology to construct E. coli with disrupted EMP and PPP. The bacteria can fix CO 2 through the Calvin cycle and use formate as an electronic resource. Then within 350 days, the strain was optimized for CO 2 fixation through adaptive evolution, thereby generating mutations that regulate gene expression and pathway regulation. Formic acid is introduced into the growth medium. As a method to generate reducing power, formate dehydrogenase is proven to be a key additive to increase the biomass in carbon dioxide from 35 to 100% [ 43 ]. All those feedstocks have more possibility to use in microbial production for L-valine. Industrial production of L-valine requires system-wide engineering, while considering culture conditions in fermentation. It can be summarized as several strategies. Back to the high-yield L-valine, iterative utilization and different orders make more sense. As an increasing body of knowledge on microbial species and metabolic systems which can be implemented in producing L-valine, we expect adopting these strategies can address more challenges, and meanwhile inspirit L-valine development to meet the needs of society."
} | 1,867 |
24844569 | null | s2 | 2,038 | {
"abstract": "Carbon concentrating mechanisms (CCMs) are common among microalgae, but their regulation and even existence in some of the most promising biofuel production strains is poorly understood. This is partly because screening for new strains does not commonly include assessment of CCM function or regulation despite its fundamental role in primary carbon metabolism. In addition, the inducible nature of many microalgal CCMs means that environmental conditions should be considered when assessing CCM function and its potential impact on biofuels. In this study, we address the effect of environmental conditions by combining novel, high frequency, on-line (13)CO2 gas exchange screen with microscope-based lipid characterization to assess CCM function in Nannochloropsis salina and its interaction with lipid production. Regulation of CCM function was explored by changing the concentration of CO2 provided to continuous cultures in airlift bioreactors where cell density was kept constant across conditions by controlling the rate of media supply. Our isotopic gas exchange results were consistent with N. salina having an inducible \"pump-leak\" style CCM similar to that of Nannochloropsis gaditana. Though cells grew faster at high CO2 and had higher rates of net CO2 uptake, we did not observe significant differences in lipid content between conditions. Since the rate of CO2 supply was much higher for the high CO2 conditions, we calculated that growing cells bubbled with low CO2 is about 40 % more efficient for carbon capture than bubbling with high CO2. We attribute this higher efficiency to the activity of a CCM under low CO2 conditions."
} | 411 |
34014980 | PMC8136678 | pmc | 2,039 | {
"abstract": "Iron reduction and sulfate reduction are two of the major biogeochemical processes that occur in anoxic sediments. Microbes that catalyze these reactions are therefore some of the most abundant organisms in the subsurface, and some of the most important. Due to the variety of mechanisms that microbes employ to derive energy from these reactions, including the use of soluble electron shuttles, the dynamics between iron- and sulfate-reducing populations under changing biogeochemical conditions still elude complete characterization. Here, we amended experimental bioreactors comprised of freshwater aquifer sediment with ferric iron, sulfate, acetate, and the model electron shuttle AQDS (9,10-anthraquinone-2,6-disulfonate) and monitored both the changing redox conditions as well as changes in the microbial community over time. The addition of the electron shuttle AQDS did increase the initial rate of Fe III reduction; however, it had little effect on the composition of the microbial community. Our results show that in both AQDS- and AQDS+ systems there was an initial dominance of organisms classified as Geobacter (a genus of dissimilatory Fe III -reducing bacteria), after which sequences classified as Desulfosporosinus (a genus of dissimilatory sulfate-reducing bacteria) came to dominate both experimental systems. Furthermore, most of the ferric iron reduction occurred under this later, ostensibly “sulfate-reducing” phase of the experiment. This calls into question the usefulness of classifying subsurface sediments by the dominant microbial process alone because of their interrelated biogeochemical consequences. To better inform models of microbially-catalyzed subsurface processes, such interactions must be more thoroughly understood under a broad range of conditions.",
"introduction": "Introduction The biogeochemical cycling of carbon (C), iron (Fe), and sulfur (S) in aquatic and terrestrial environments is driven largely by microbially-catalyzed redox reactions. Such reactions by definition involve the transfer of electrons, so it is necessary to assess the thermodynamic and kinetic constraints on electron transfer in appropriate model systems in order to understand the metabolic processes that drive these biogeochemical cycles. For example, in many environments ferric iron (Fe III ) is primarily present as relatively insoluble Fe III oxides. These minerals provide an important electron sink during anaerobic respiration by a variety of dissimilatory Fe III -reducing bacteria (DIRB) and archaea. These phylogenetically diverse microorganisms are able to obtain energy by coupling the oxidation of organic compounds or molecular hydrogen to the reduction of Fe III to Fe II under suboxic and anoxic conditions [ 1 – 3 ]. Fe III -reducing microorganisms and the reactive ferrous species they produce play a major role in controlling water quality [ 4 , 5 ], the dissolution and precipitation of minerals [ 6 – 8 ], nutrient availability [ 9 ], and the fate and transport of contaminants [ 10 ]. Due to the relative insolubility of ferric minerals in most environments, DIRB must employ different mechanisms to respire using these terminal electron acceptors than those used for soluble terminal electron acceptors such as dissolved oxygen (O 2 ), nitrate (NO 3 − ), and sulfate (SO 4 2− ) [ 11 ]. Some DIRB such as Geobacter and Shewanella can transfer electrons directly to Fe III oxide surfaces by means of reductases located on their outer cell membrane [ 12 ] or via electrically conductive pili or nanowires [ 13 – 17 ]. The need for physical contact between Fe III oxide minerals and microbial cells, however, can be readily overcome. The dissolution of Fe III oxides is promoted by exogenous and endogenous ligands and the resulting soluble Fe III complexes can diffuse away and be reduced by DIRB at a distance [ 18 , 19 ]. Likewise, the transfer of electrons from the cell to external electron acceptors (e.g., Fe III oxides) can be facilitated by soluble electron shuttles, i.e., compounds that can be reversibly oxidized and reduced. In this scenario, an oxidized electron shuttle is reduced by the organism, which can transfer electrons to a remote acceptor. Since electron shuttles can be oxidized and reduced repeatedly, they can have a substantial effect on both the rate and extent of Fe III oxide reduction even when present at trace concentrations [ 20 ]. A wide variety of endogenous and exogenous organic and inorganic compounds have been shown to function as electron shuttles in the bioreduction of Fe III oxides, including quinones, flavins, phenazines, and reduced sulfur species [ 20 – 31 ]. In addition, humic substances—a class of naturally occurring, chemically heterogeneous organic oligoelectrolytes derived primarily from the decomposition of bacteria, algae, and higher plant material that are ubiquitous in aquatic and terrestrial environments—can also be utilized as electron shuttles in the bioreduction of Fe III oxides [ 18 , 30 , 32 – 35 ]. The ability of humic substances to act as electron shuttles has largely been attributed to the presence of quinone groups within their structures [ 36 – 38 ]. However, given the polymorphic nature of humic substances, their structure and functional characteristics are highly variable, including the type and number of reversibly redox active moieties, resulting in a distribution of redox potentials and inherent variability in the redox properties observed among them [ 38 – 42 ]. Therefore, model quinones with well-defined redox characteristics (e.g., reduction potentials) such as 9,10-anthraquinone-2,6-disulfonate (AQDS) have been widely used as analogs for the redox active moieties in humic substances [ 6 , 32 , 43 ]. A phylogenetically diverse range of bacteria and archaea can transfer electrons to model quinones (e.g., AQDS) and humic substances [ 44 , 45 ]. Indeed, many microorganisms that are not able to reduce Fe III oxides directly, can cause the reduction of Fe III oxides in the presence of a suitable electron shuttle, including organisms that are not primarily categorized as Fe III -reducing microbes (i.e., fermenters, sulfate reducers, and methanogens) [ 26 , 27 , 44 , 46 – 49 ]. The ubiquity of humic/quinone-reducing microorganisms in aquatic and terrestrial environments [ 50 ] and the ability of reduced humics/quinone to shuttle electrons to Fe III oxides suggests that the presence of electron shuttles provides the potential for bacteria that are not metabolically capable of reducing Fe III minerals to contribute to Fe III oxide bioreduction, thereby increasing microbial diversity under iron-reducing conditions. Although electron shuttles have been shown to significantly enhance Fe III oxide reduction in systems with complex, multispecies microbial communities [ 51 – 55 ], the effects of electron shuttles on microbial community development have not been explicitly examined. In this study, we investigate the effects of the presence and absence of a soluble electron shuttle (AQDS) on biogeochemical dynamics and microbial community development under Fe III - and sulfate-reducing conditions.",
"discussion": "Discussion The addition of the electron shuttle AQDS did increase the initial rate of Fe III reduction ( Fig 1 and Table 1 ); however, it had little effect on the composition of the microbial community during this phase ( Fig 5 ). The increase in Fe II in both AQDS+ and AQDS–bioreactors corresponded directly to an increase in the relative abundance of sequences classified as Albidiferax and Geobacter . A single OTU dominated the Geobacter in the AQDS–bioreactors throughout the experiment and in the AQDS+ bioreactors through day 56 ( Fig 7 ). Both Albidiferax and Geobacter are known to reduce ferric minerals like the goethite present in the natural sienna amendment. Geobacter in particular has been shown to predominate in sedimentary environments where Fe III reduction is occurring [ 86 – 91 ], typically in response to acetate biostimulation [ 56 , 92 – 97 ]. This pattern is well-established, as Geobacter blooms have been observed in Fe III - and acetate-amended bioreactors using material from rice paddies [ 98 ], marshes [ 99 ], and aquifer sediment [ 100 , 101 ]. Geobacter has also been enriched in acetate-amended experimental systems where AQDS (not Fe III ) was added as the sole electron acceptor [ 102 ]. Some other taxa (e.g., Acidovorax ) were abundant at day 2, but these were quickly overtaken in dominance by Geobacter . Given the diversity of organisms capable of reducing synthetic (e.g., AQDS) and naturally occurring quinones (e.g., humic substances) [ 44 ], we had hypothesized that the presence of AQDS would lead to greater microbial diversity in the AQDS+ bioreactors during Fe III reduction due to recruitment of non-metal-reducing, quinone-respiring organisms. However, this was not the case under our experimental conditions. In both AQDS+ and AQDS–bioreactors diversity declined at the onset of Fe II production, with sequences classified as Geobacter becoming more abundant. Indeed, Geobacter were even more dominant in the AQDS+ systems than in AQDS–; which in retrospect is perhaps not unexpected given their ability to use both Fe III and AQDS as terminal electron acceptors for anaerobic respiration [ 2 ]. A similar enhancement in Geobacter abundance was reported by Rowland et al. [ 100 ] in microcosm studies where sediments were amended with acetate alone or acetate and AQDS (50 μM; an order of magnitude lower AQDS concentration than in our study) and by Chen et al [ 103 ] where the relative abundance of Geobacter in rice paddy soil incubations increased with increasing AQDS concentration. It is possible that AQDS may be toxic to some members of the Fe III -reducing community, potentially decreasing diversity. Direct evidence of the potential toxicity of AQDS to bacteria is lacking; however, circumstantial evidence suggests that AQDS toxicity is not likely to be an issue in our systems containing such low (100 μM) concentrations of the quinone molecule. In acetate-amended sediment incubations, lower Fe II concentrations were observed in systems containing 250 μM AQDS compared with 50 μM, the next lowest concentration examined [ 21 ]. However, AQDS concentrations as high as 20 mM showed no inhibition of Fe II production relative to concentrations as low as 10 μM in rice paddy soil incubations [ 103 ]. Furthermore, pure culture studies with Geobacter sulfurreducens and Shewanella putrefaciens CN32 showed no inhibition of Fe II production with 500 μM and 1000 μM AQDS, respectively [ 20 , 104 ]. Similar to the trajectory observed following acetate injection in a uranium-contaminated aquifer in Rifle, Colorado [ 56 ] (the same location from which the sediment used to inoculate the bioreactors was obtained), the initial bloom of Geobacter was followed by a period of sulfate reduction and the increased relative abundance of sequences associated with sulfate-reducing bacteria. At both Rifle and in our experiment, the most abundant taxa associated with sulfate reduction were of the genus Desulfosporosinus . Initially at less than <1% of the total community in these systems, Desulfosporosinus sequences eventually came to dominate the microbial community during the latter stages of the experiment, where they accounted for roughly one-third of all sequences in both the AQDS+ and AQDS–bioreactors ( Fig 5 ). As with the Geobacter OTUs, a single Desulfosporosinus OTU dominated in both AQDS+ and AQDS–bioreactors ( Fig 7 ). In addition to the bioreactors in this study, as well as an earlier study from our group [ 105 ], Desulfosporosinus have previously been found in acetate-amended sediments [ 78 , 106 , 107 ]. The preponderance of Desulfosporosinus in these acetate-amended systems is perhaps unexpected given the inability of all previously cultivated Desulfosporosinus spp. to couple acetate oxidation to dissimilatory sulfate reduction [ 108 – 118 ]. However, the increase in the relative abundance of Desulfosporosinus in our bioreactors was coincident with a sharp increase in the rate of sulfate consumption ( Fig 1 and Table 1 ), suggesting that they are active contributors to sulfate reduction in this system. G . sulfurreducens can oxidize acetate by syntrophic association with hydrogen-oxidizing anaerobic partners [ 119 , 120 ] and many Desulfosporosinus spp. can use H 2 as an electron donor for dissimilatory sulfate reduction [ 109 , 110 , 114 , 116 , 118 , 121 ] suggesting the possibility for sulfate reduction via a syntrophic association between Geobacter and Desulfosporosinus . Of particular note is that the majority of Fe II production actually occurred during the “sulfate-reducing” latter phase of the experiment. During the early phase of the experiment dominated by Albidiferax and Geobacter , <10% of the total amount of Fe III added was reduced. Indeed, 64% and 78% of the total amount of Fe II produced occurred during the “sulfate-reducing” phase in AQDS+ and AQDS–bioreactors, respectively, likely driven by the reduction of Fe III by the sulfide produced by Desulfosporosinus and other SRB [ 122 ]. These results are consistent with recent findings highlighting the potential importance of sulfur-driven reactions in the biogeochemical cycling of iron in sedimentary environments. Because iron reduction is strongly pH-dependent, experimental and modeling evidence suggests that under sulfidic, alkaline conditions, Fe III reduction by metal-reducing bacteria likely proceeds primarily via an electron shuttling pathway mediated by S 0 [ 31 ]. Even under circumneutral conditions, laboratory and field studies suggest that sulfur cycling can play a significant role in Fe III reduction in freshwater and marine environments [ 122 – 125 ]. These results call into question the paradigm of parsing out geomicrobiological reactions into “iron-reducing” and “sulfate-reducing” phases. This traditional conception of terminal electron accepting processes in the subsurface, while long-established (e.g., [ 126 ]), has increasingly been called into question by both theoretical and experimental observations in both the field and laboratory. Iron reduction and sulfate reduction have frequently been observed to co-occur in sedimentary environments [ 86 , 87 , 127 – 129 ], and modeling results predict that the co-occurrence of these processes may even benefit both groups [ 130 , 131 ]. In our results, sulfate reduction began essentially consequent with iron reduction. While some of the sulfate may have been consumed by assimilatory sulfate reduction during the growth of other organisms, this is unlikely as there was no difference in the rate of sulfate consumption between AQDS+ and AQDS–systems. If the growth of iron reducers was responsible for the decrease in the concentration of sulfate, the rate of sulfate consumption would have been greater during the early phase in AQDS+ bioreactors. While initially higher rates of growth, particularly in the presence of electron shuttles, may give iron reducers an early advantage under growth-stimulating conditions, such dynamics may ultimately bear little resemblance to the processes that occur in most aquifers."
} | 3,828 |
25264452 | PMC4173114 | pmc | 2,040 | {
"abstract": "Collective decision-making in biological systems requires all individuals in the group to go through a behavioural change of state. During this transition fast and robust transfer of information is essential to prevent cohesion loss. The mechanism by which natural groups achieve such robustness, though, is not clear. Here we present an experimental study of starling flocks performing collective turns. We find that information about direction changes propagates across the flock with a linear dispersion law and negligible attenuation, hence minimizing group decoherence. These results contrast starkly with current models of collective motion, which predict diffusive transport of information. Building on spontaneous symmetry breaking and conservation laws arguments, we formulate a new theory that correctly reproduces linear and undamped propagation. Essential to the new framework is the inclusion of the birds’ behavioural inertia. The new theory not only explains the data, but also predicts that information transfer must be faster the stronger the group’s orientational order, a prediction accurately verified by the data. Our results suggest that swift decision-making may be the adaptive drive for the strong behavioural polarization observed in many living groups."
} | 319 |
37970006 | PMC10634217 | pmc | 2,041 | {
"abstract": "A liquid Ga-based synaptic device with two-terminal electrodes\nis demonstrated in NaOH solutions at 50 °C. The proposed electrochemical\nredox device using the liquid Ga electrode in the NaOH solution can\nemulate various biological synapses that require different decay constants.\nThe device exhibits a wide range of current decay times from 60 to\n320 ms at different NaOH mole concentrations from 0.2 to 1.6 M. This\nresearch marks a step forward in the development of flexible and biocompatible\nneuromorphic devices that can be utilized for a range of applications\nwhere different synaptic strengths are required lasting from a few\nmilliseconds to seconds.",
"conclusion": "4 Conclusions Synaptic current behavior\nwas demonstrated via a surface electrochemical\nredox reaction on a liquid Ga electrode in a NaOH solution. By control\nof the NaOH molar concentration, a wide range of decay time constants\nwere obtained from 60 to 130 ms at 1.6 and 0.2 NaOH molar concentrations,\nrespectively. Synaptic devices with long decay times can be used to\nstore information for extended periods, learning, and adaptation in\nresponse to input patterns, while synaptic devices with short decay\ntimes can be used to filter out noise and unwanted signals by rapidly\nreducing the strength of synapses. Our research marks a step forward\nin the development of flexible and biocompatible neuromorphic devices,\nwhich can be utilized for a range of applications where different\nsynaptic strengths are required that can last from a few milliseconds\nto seconds.",
"introduction": "1 Introduction The von Neumann architecture\nhas served as the basis of modern\ndigital electronics for the last few decades. 1 It mainly consists of a central processing unit (CPU) and a memory\nunit. The CPU accesses the memory through a shared bus, which facilitates\nthe transfer of data between the CPU and memory in both directions.\nThis data transfer through the bus suffers from large latency and\nhigh power consumption, causing interconnect bottlenecks when working\nwith large data sets. 2 , 3 With the advent of artificial\nintelligence (AI) that can perform cognitive tasks and offer predictions\nfor the future, the von Neumann bottlenecks limit the performance\nof advanced electronics requiring real-time processing and computing\nsuch as big data analytics and machine learning. 4 To address the challenges of latency and power consumption\ninherent in the von Neumann architecture, new computational methodologies\nbased on in-memory processing units (IMPUs) or processing in memory\n(PIM) have been proposed. 5 In the IMPUs\nor PIM architecture, the processing elements are integrated directly\ninto the memory units, allowing computation to be performed on the\ndata stored in the memory without data transfer between the CPU and\nmemory through the bus. In short, this computing architecture allows\nCPU and memory functionality to occur within the same unit without\nthe interconnect. Compared to the traditional von Neumann architecture,\nthe IMPU/PIM architecture can considerably minimize latency and power\nconsumption since there is no requirement to transfer data repeatedly\nbetween the CPU and memory. 6 This computational\nmethodology replicates brain functions, and the resulting devices\nare accordingly referred to as neuromorphic computing and synaptic\ndevices because of the functional similarity with neural synapses. 7 Synaptic devices are a key component of neuromorphic\ncomputing systems and can serve to achieve ultrafast and energy-efficient\ncomputing in neuromorphic systems. These synaptic devices use analog\ncircuits to model the variable strength of synaptic connections in\nthe brain, enabling them to perform functions such as learning, memory,\nand pattern recognition. Unlike traditional memory devices that use\nbinary information, those reported synaptic devices are not just binary\non or off processes. In the biological system, information is transferred\nfrom one neuron to another through synapses. The strength of each\nsynapse is represented by its synaptic weight that is represented\nas analog values and is used to control the amount of influence that\none neuron has on another. The synaptic weight plays a crucial role\nin neuromorphic devices, which enable the system to process information\nin a way that is more like natural processing. So far, various synaptic\ndevices have been reported with its own advantages and challenges,\nincluding resistive switching memory, 8 − 12 phase-changing memory, 13 − 15 ferroelectric memory, 16 − 19 electrochemical memory, 20 − 25 and charge trap memory. 26 − 28 Resistive switching memory implemented\nsynaptic weights in neuromorphic systems, where the strength of a\nsynapse is proportional to the resistance of the memory element. 8 − 12 Phase-change memory implemented synaptic weights by varying the\ncrystalline state of the material to adjust the strength of the synapse,\nwhich can switch between amorphous and crystalline states to store\ndata. 13 − 15 Ferroelectric memory can vary the polarization state\nof the material to adjust the strength of the synapse, which can switch\nbetween two stable polarization states to store data. 16 − 19 Electrochemical memory implemented synaptic weights by controlling\nthe flow of ions across the junction, 20 − 25 where the strength of a synapse can be adjusted by changing the\nconcentration of ions at the synaptic junction. These devices are\nelectronic devices that use ions rather than electrons to conduct\nelectricity. Charge trap memory varied synaptic weights by controlling\nthe number of trapped charge carriers to adjust the strength of the\nsynapse. 26 − 28 These reported synaptic devices have shown promise\nin terms of energy efficiency, scalability, and compatibility with\nthe current technology. Very recently, an ionic device using\nliquid Ga was made to resemble\nneural spike signals, which is beneficial for organ cell–machine\ninterface systems because Ga is biocompatible and flexible at room\ntemperature. 29 , 30 Depending on the specific application,\nsynaptic functions with different decay constants are required for\nsynaptic devices because synaptic functions with different decay constants\ncan affect the temporal dynamics of the network and its ability to\nperform certain computations, such as short-term memory or pattern\nrecognition. 31 Synaptic decay refers to\nthe gradual decrease in the strength of a synaptic connection over\ntime. For example, different decay constants can help the network\ndistinguish between phonemes that have similar temporal patterns but\ndifferent durations. In fact, the human brain utilizes a wide range\nof synaptic current decay from 1 ms to 1 s. 32 Different types of synaptic currents serve different functions in\nthe brain. Fast decay time involves the rapid transmission of information\nbetween neurons, which can be used to filter out noise and unwanted\nsignals. On the other hand, slow decay times are involved in more\ncomplex neural processes such as learning and memory. In previous\nstudies, different decay characteristics have been demonstrated using\ndifferent materials, 33 crystallinity, 34 defects, 35 , 36 and bias voltages. 29 In this study, we demonstrate a liquid Ga-based\nsynaptic device using two-terminal electrodes in different NaOH solutions\nthat can generate different synaptic current responses. The proposed\ndevices exhibited a wide range of current decay times from 60 to 320\nms in different NaOH moles.",
"discussion": "3 Results and Discussion When a NaOH-based\nionic channel was established between the Ga\nand Cu electrodes, as shown in Figure 1 a, spontaneous half-cell reactions took place on both\nsides, resembling those observed in a galvanic cell. As a result,\nthe Ga electrode underwent oxidation and released electrons that were\nassumed to have been consumed by the Cu electrode via oxygen reduction.\nFollowing the oxidation of the Ga electrode, Ga-based oxide was subsequently\nformed on the Ga electrode. 38 It is known\nthat a Ga-based oxide layer can also be dissolved in a NaOH solution.\nUnder steady-state conditions, the reaction between the formation\nand dissolution of Ga-based oxide on the Ga surface eventually produces\nan extremely thin layer of Ga-based oxide. It is worth noting that\nwhen a voltage was applied to the Cu electrodes, similar oxidation\nand reduction peaks were observed as well, 39 indicating that Cu could also be dissolved in the NaOH solution.\nThe Cu electrode was grounded while a voltage was applied to the Ga\nelectrode. The Ga liquid metal moved toward the cathode when the applied\nvoltage was higher than 1 V. This movement could be explained by several\nfactors such as differences in the electrochemical potential or surface\ntension of the metal and the solution. In our previous work, 29 we demonstrated that this Ga-based oxide layer\ncould be removed electrically by applying −2 to −5 V\nto the Ga electrode in the NaOH solution, which is called an electrochemical\nreduction. In the present study, when a negative voltage was applied\nto the Ga electrode, positively charged ions from the NaOH solution\nwere attracted to the Ga-oxide surface. These electrically attracted\nions reacted with the Ga-oxide layer, causing it to break down and\ndissolve in the NaOH solution. Conversely, when a positive voltage\nwas applied to the Ga electrode, a Ga-based oxide was formed, leading\nto an anodic current peak (oxidation currents). Figure 1 b shows the current–voltage curve\nof the Ga electrode in a NaOH solution at different molar concentrations.\nThe voltage was adjusted from −1 to 1 V at a speed of 0.1 V/s.\nThe results showed a small current hump at around −0.25 V regardless\nof molar concentration. This hump current represented anodic peaks\nand was attributed to the formation of a Ga-based oxide layer on the\nGa surface. As the NaOH molar concentration increased, the peak current\ntended to increase, indicating that more oxidation occurred with the\nhigher molar concentrations. Figure 1 (a) Grounded Cu electrode with the voltage applied\nto the Ga electrode.\nThese two electrodes, separated by 3 mm with a canal width of 1 mm,\nwere submerged in a NaOH solution at different molar concentrations.\nThe dimensions were determined because it could be achieved by the\n3D printer. The NaOH container with Ga and Cu electrodes was heated\nto a hot chuck at 50 °C. (b) Current–voltage curve of\na Ga electrode at different NaOH molar concentrations. After analysis of the electrochemical oxidation\ncurrent, as shown\nin Figure 1 b, artificial\nsynaptic behavior was demonstrated, imitating the behavior of a biological\nsynapse. This is shown in Figure 2 , which depicts the plasticity characteristics of the\nGa electrode in the 1 M NaOH solution under different pulse conditions. Figure 2 a presents a definition\nof a pulse-shaped stimulus including the rise and fall time, delay,\nand pulse width. When the voltage was changed from −2 to 1\nV as shown in Figure 2 b, the current level remained relatively constant. However, when\nthe voltage was adjusted from −5 to 1 V in Figure 2 b, the current value changed\nwith an external stimulus exceeding the threshold value. Figure 2 (a) Definition\nof the pulse-shaped stimulus. It is defined in terms\nof the rise and fall time, delay, and pulse width. (b) Synaptic behaviors\nof the Ga electrode in 1 M NaOH concentrations at a pulse width of\n0.04 ms, a rise and fall time of 0.01 ms, and a delay time of 0.002\nms. The results in Figure 2 b indicated that the Ga electrode in the\nNaOH solution is\nanalogous to presynaptic and postsynaptic terminals; the frequency\nof −5 V acted as a neuronal stimulus. The nonlinearity factor\nof the potentiation was 0.35 at −5 V as shown in Figure S2 . Thus, the voltage condition from −5\nto 1 V was applied in the system unless otherwise noted. It was found\nthat synaptic characteristics including nonvolatile behavior, low-power\nconsumption, high-speed operation, programmability, and analog behavior\ncould be optimized depending on the pulse width, duty cycles, and\nvoltage amplitude. Figure 3 compares\nthe generation of synaptic spike responses at two different pulse\nwidths. It showed that synaptic behavior was observed at a pulse with\n0.04 ms in Figure 3 a, while the synaptic behavior was not observed at a pulse with 0.08\nms in Figure 3 b. It\nindicated that an optimized pulse condition needs to be developed\nfurther. There are several key requirements for synaptic devices including\nnonvolatile behavior, low-power consumption, high-speed operation,\nprogrammability, and analog behavior. By satisfying these requirements,\nsynaptic devices could serve as an efficient device that can perform\ncomplex tasks such as image and speech recognition with high accuracy\nand speed. Typical learning and forgetting characteristics of the\nsynaptic device are shown in Figure S3 . Figure 3 Generation\nof synaptic spike responses with a rise and fall time\nof 0.01 ms and a delay time of 0.002 ms at a pulse width of (a) 0.04\nand (b) 0.08 ms. Figure 4 a shows\nthe transient current response generated by the formation of a Ga-based\noxide layer on the Ga surface at different NaOH molar concentrations.\nOnce the oxide layer formed on the Ga surface when the voltage changed\nfrom −5 to 1 V, a further oxidation reaction was spontaneously\nprevented. This caused a current reduction over time, and the current\nfinally reached a stable state. The current response showed that both\nthe peak current and the saturation current increased with higher\nNaOH molar concentrations. The higher peak current was attributed\nto the higher oxidation reactions, while the higher saturation current\nwas presumably due to the higher ionic conduction in the NaOH solution.\nAs shown in Figure 4 b, the current response curve was normalized to compare the current\ndecay behavior at different NaOH molar concentrations. Figure 4 (a) Step response, (b)\nnormalized step response of oxidation current,\n(c) current decay time, and (d) cumulative probability of Ga electrode\nin the NaOH solution at different NaOH molar concentrations when the\nvoltage is adjusted from −5 to 1 V. The current decay time was defined as the time\nit took for the\ncurrent to decay from 90 to 10% of its maximum peaks. The results\nshowed that the decay time decreased with higher NaOH molar concentrations,\nas shown in Figure 4 c. The average decay time at 0.2 M was 320 ms, and the value decreased\nto as low as 60 ms at 1.6 M. Synaptic devices with either long (320\nms) or short (60 ms) decay time can be utilized depending on the specific\ncase and requirement. Synaptic devices with long decay times can be\nused to store information for extended periods, allowing for the implementation\nof long-term memory in neuromorphic systems. Synaptic devices with\nlong decay times can also be used for learning and adaptation in response\nto input patterns. On the other hand, synaptic devices with short\ndecay times can be used to filter out noise and unwanted signals by\nrapidly reducing the strength of synapses. The synaptic device with\ndifferent decay times allows for the implementation of a wider range\nof synaptic plasticity, which can lead to more efficient neuromorphic\ncomputing systems. Figure 4 d exhibits the cumulative probability of the current decay\ntime at different molar concentrations. Cumulative probability refers\nto the probability indicating the likelihood of a random variable\nto assume a value that is equal to or less than a specified value.\nIn other words, it represents the probability of an event occurring\nup to a certain point in a distribution. This showed that the current\ndecay time in this work could be predicted at certain molar concentrations.\nThe ability to implement different decay times is significant because\nit can represent a variety of STP/LTP behaviors. This enables the\nsimulation of various synaptic behavior characteristics. This result\ncould be useful for neuromorphic computing applications, which may\nrequire specific decay time to function properly. Figure 5 a,b illustrates\na schematic drawing of the formation and dissolution reactions of\nthe Ga-based oxide on the Ga electrode at high and low NaOH molar\nconcentrations, showing the different oxidation reactions. The different\noxidation mechanisms resulted in various current decay times. It was\npresumed that a mixed layer of stoichiometric Ga 2 O 3 and nonstoichiometric Ga 2 O x ( x <3) was able to form at lower NaOH molar\nconcentrations. Oxygen atoms in the Ga-oxide layer were supplied from\nOH – ions in the NaOH solution. Thus, it was expected\nthat the formation of a nonstoichiometric Ga 2 O x ( x <3) layer resulted from the\nlow concentration of OH – ions in the lower NaOH\nsolution at lower NaOH molar concentrations. The nonstoichiometric\nGa 2 O x ( x <3) layer in the mixture would be vulnerable to the NaOH solution and\neasily dissolve, which would cause new oxidation to occur. This dissolution\nand partial oxidation would continue at lower NaOH molar concentrations\nuntil the surface of the Ga electrode would be entirely covered with\nthe stoichiometric Ga 2 O 3 layer, resulting in\na slower current decay behavior. In contrast, with higher NaOH molar\nconcentrations, abundant OH – ions could be supplied\nnear the surface of the Ga electrode when the voltage changed from\n−5 to 1 V, which could cause the formation of the stoichiometric\nGa 2 O 3 layer on the Ge electrode. The formed\nstoichiometric Ga 2 O 3 layer might be sufficiently\nrobust in the NaOH solution and thereby prevent further oxidation\nreactions, which would result in faster current decay behavior. Synaptic\nproperties can be classified into two major categories: short-term\nplasticity (STP) and long-term plasticity (LTP). STP is a form of\nsynaptic plasticity associated with rapid and transient changes that\ncan last from milliseconds to seconds. STP can filter out noisy and\nirrelevant information. LTP, on the other hand, is a longer-lasting\nchange in synaptic strength that can last from hours to days or even\nlonger. LTP is thought to relate to many forms of learning and memory.\nRegarding STP, various types of synaptic devices have been demonstrated\nto emulate biological synapses. Their typical decay time constants\nare listed in Figure S1 . The results show\nthat the proposed electrochemical oxidation device using liquid Ga\nin NaOH would be able to cover a wide range of decay times by changing\nthe NaOH molar concentration. Like other liquid-based devices, high-temperature\nconditions in practical use would be detrimental to the functionality\nof the device. For example, water in the NaOH solution will start\nto evaporate at high-temperature conditions. To improve long-term\nstability such as reliability and endurance characteristics, passivation\ntechnology would be further required. The proposed Ga-based device\nin the NaOH solution was designed to operate under certain temperature\nconditions that are compatible with the liquid Ga electrode and the\nNaOH solution. The proposed electrochemical redox device using the\nliquid Ga electrode in the NaOH solution could replicate the function\nof biological synapses by utilizing the properties of the liquid Ga\nelectrode and the NaOH solution to produce different decay constants.\nThis could potentially have applications in fields such as artificial\nintelligence and neural networks, where the ability to replicate the\nfunction of biological synapses could be useful. Further research\nand development to fully understand the potential capabilities and\nlimitations of such a device are necessary. Figure 5 Schematic illustration\nof the proposed (a) reduction and (b) oxidation\nmechanisms of the Ga electrode in the NaOH solution at either high\nor low M."
} | 4,926 |
35943212 | PMC9426500 | pmc | 2,042 | {
"abstract": "ABSTRACT The soil fungal community plays pivotal roles in soil nutrient cycling and plant health and productivity in agricultural ecosystems. However, the differential adaptability of soil fungi to different microenvironments (niches) is a bottleneck limiting their application in agriculture. Hence, the understanding of ecological processes that drive fungal microbiome assembly along the soil-root continuum is fundamental to harnessing the plant-associated microbiome for sustainable agriculture. Here, we investigated the factors that shape fungal community structure and assembly in three compartment niches (the bulk soil, rhizosphere, and rhizoplane) associated with tobacco ( Nicotiana tabacum L.), with four soil types tested under controlled greenhouse conditions. Our results demonstrate that fungal community assembly along the soil-root continuum is governed by host plant rather than soil type and that soil chemical properties exert a negligible effect on the fungal community assembly in the rhizoplane. Fungal diversity and network complexity decreased in the order bulk soil > rhizosphere > rhizoplane, with a dramatic decrease in Ascomycota species number and abundance along the soil-root continuum. However, facilitations (positive interactions) were enhanced among fungal taxa in the rhizoplane niche. The rhizoplane supported species specialization with enrichment of some rare species, contributing to assimilative community assembly in the rhizoplane in all soil types. Mortierella and Pyrenochaetopsis were identified as important indicator genera of the soil-root microbiome continuum and good predictors of plant agronomic traits. The findings provide empirical evidence for host plant selection and enrichment/depletion processes of fungal microbiome assembly along the soil-root continuum. IMPORTANCE Fungal community assembly along the soil-root continuum is shaped largely by the host plant rather than the soil type. This finding facilitates the implementations of fungi-associated biocontrol and growth-promoting for specific plants in agriculture practice, regardless of the impacts from variations in geographical environments. Furthermore, the depletion of complex ecological associations in the fungal community along the soil-root continuum and the enhancement of facilitations among rhizoplane-associated fungal taxa provide empirical evidence for the potential of community simplification as an approach to target the plant rhizoplane for specific applications. The identified indicators Mortierella and Pyrenochaetopsis along the soil-root microbiome continuum are good predictors of tobacco plant agronomic traits, which should be given attention when manipulating the root-associated microbiome.",
"introduction": "INTRODUCTION Plants and microbes have been interacting with each other and evolving for their mutual benefit ( 1 , 2 ). Consequently, the ability of root-associated microbiota (i.e., rhizomicrobiota) to facilitate the growth and health of the host plant via phytohormone production and competition with pathogens, respectively, is a subject of intense research ( 3 – 5 ). Furthermore, harnessing the root-associated microbiome is increasingly perceived as a sustainable approach to facilitate agricultural production ( 6 , 7 ). Understanding of fundamental ecological processes that shape the microbiome assembly along the soil-root continuum is prerequisite for the precise manipulation of the microbiome in a specific niche. The complexity of root-associated microbial communities is governed largely by the attractant and repellent activities of the host plant ( 8 – 10 ). The host root provides a nutrient-rich niche for microbes, in which the plant-microbe interactions are fostered by plant innate immunity ( 11 ). Meta-transcriptome analysis revealed differences in the bacterial microbiomes in the bulk soil and rhizosphere of several plant species ( 12 ), with a higher proportion of active bacteria associated with roots than with the bulk soil ( 13 ). Furthermore, bacterial and fungal communities associated with wheat “total roots” (i.e., including the endosphere and rhizoplane compartments) are clearly distinguished from those in the bulk soil and rhizosphere ( 14 ). Apart from the filtering effect of the host plant, the microbial community is also shaped by other factors, i.e., cropping practices, soil types, and nutrients ( 14 – 17 ). For instance, soil type is thought to account for early microbial community assembly in the plant rhizosphere ( 18 – 20 ), as the soil is a reservoir of diverse microbes. While, to date, studies have focused mainly on bacterial members of the overall microbial community, many issues remain concerning the fungal microbiome assembly by the host plant and in relation to the soil type in different niches (i.e., the bulk soil, rhizosphere, and rhizoplane) ( 14 , 21 , 22 ). Understanding of the ecological processes that shape the soil fungal community is essential, as fungi play important roles in the soil ecosystem, e.g., symbiosis ( 23 , 24 ), nutrient cycling ( 25 ), decomposition ( 26 , 27 ), pathogenesis ( 28 , 29 ), and N 2 O production ( 29 , 30 ). For instance, the versatile lifestyle and nutritional adaptability of Trichoderma enable several members of this genus to establish symbiotic interactions with the host plant, with Trichoderma -based products commercialized as biopesticides and biofertilizers ( 31 – 33 ). Fungi appear to be more sensitive to microenvironmental variations than bacteria ( 34 , 35 ) probably because most soil fungi are saprophytic and are constantly searching for available nutrients, e.g., by developing mycelia. Furthermore, symbiotic fungi aid plant nutrient absorption by forming mutualistic associations with plant roots, as the fungal hyphae extend beyond the area of nutrient depletion in root vicinity ( 36 , 37 ) and grow in soil pores whose diameters are considerably smaller than that of the root to exploit nutrients ( 38 ). Additionally, some fungal species survive in a yeast form or as a mycelium depending on the environmental and some internal conditions, which facilitates their adaptation to the environment ( 39 ). Finally, different ecological niches impact the taxonomic groups and functions of the microbiomes within them ( 13 , 22 , 40 ). Considering the above information, it is imperative to elucidate fluctuations in fungal communities along the soil-root continuum and to decipher the effects of host plant selection and soil variables therein. These data would add another dimension to the notion that host plant selection acts as a driver of microbial community variation along the soil-root continuum ( 11 , 22 ). Accordingly, we established a greenhouse pot experiment involving four different soil types from major tobacco-producing areas in China and examined fungal community assembly in three distinct compartment niches therein (the bulk soil, rhizosphere, and rhizoplane). We hypothesized that (i) the contributions of plant selection and soil type to fungal community assembly along the soil-root continuum would shift across the bulk soil to the rhizosphere and the rhizoplane, (ii) the diversity and network complexity of the fungal community would decrease with an increasing plant selection effects, and (iii) the indicator taxa in the compartment niches would serve as a good predictor of plant agronomic traits.",
"discussion": "DISCUSSION The plant-associated microbiome is closely related to plant growth and traits ( 2 , 44 ). Few studies to date have focused on fungal community variations in the niches along the soil-root continuum, let alone simultaneously evaluated fungi in different soil types. In the present study, the pattern of fungal community separation in each niche was consistent with a spatial gradient from the bulk soil through to the rhizosphere and the rhizoplane, with similar patterns of fungi alpha diversity in each niche ( Fig. 1 and Table 1 ). This result was in line with findings of a previous study, which demonstrated marked differences in fungal abundance patterns in the bulk soil and rhizocompartments in rice ( 8 ), indicating that the niche impacts plant fungal community assembly. However, the soil type was a lesser source of fungal community variation than the compartment niche ( Table 1 ), with only a marginal difference in fungal alpha diversity in different soil types ( Table S3 ). Intriguingly, although the chemical properties of different soil types were highly variable, none of the soil factors significantly affected the rhizoplane community (see Table 2 and Table S6 in the supplemental material). For instance, the effective soil properties, i.e., soil pH and nutrient stoichiometry (TN/TP, TN/TK), that fundamentally contributed to fungal community variation in the bulk soil and the rhizosphere were not associated with fungal community assembly in the rhizoplane. Hartman et al. ( 14 ) reported that fertilization drives the differences in fungal communities in the bulk soil but not in the root compartments (including the rhizoplane and endorhizosphere). Thus, the legacies of soil properties that drive the fungal community assembly are inherited by the plant rhizosphere rather than the rhizoplane, probably because the rhizoplane contains sufficient nutrients for the colonizing fungi. In addition, mutualistic associations between certain fungi and plant roots are probably involved in fungal community assembly in the rhizoplane ( 36 , 37 ). These points should be addressed to allow crop microbiome manipulation in the future. 10.1128/msystems.00361-22.9 TABLE S6 Soil chemical characteristics of bulk soil of all soil types in a pot experiment. Data presented are means (SD); n = 3. Different lowercase letters within the same column indicate significant differences among soil types at a P value of <0.05 (Tukey’s HSD test). Download Table S6, DOCX file, 0.01 MB . Copyright © 2022 Li et al. 2022 Li et al. https://creativecommons.org/licenses/by/4.0/ This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . The abundance of most of the identified fungal phyla sequentially decreased from the bulk soil to the rhizoplane. Ascomycota and Basidiomycota were the most abundant phyla in the bulk soil ( Table S2 ), broadly corresponding to the extensive surveys of soil fungal communities ( 45 – 47 ). On the other hand, in the current study, the abundance of Basidiomycota was lower than that of Mortierellomycota in both the rhizosphere and the rhizoplane. These observations could be explained by the notion that Basidiomycota are considered K-strategists ( 48 ), with difficulty in surviving in nutrient-abundant rhizocompartments (i.e., in comparison with the oligotrophic bulk soil) that are rich in root exudates (e.g., carbohydrates, organic acid ions, amino acids, and vitamins) ( 49 , 50 ). In contrast, Mortierellomycota, Chytridiomycota, and Glomeromycota were more abundant in the rhizosphere than in the other compartment niches, indicating that the rhizosphere compartment, which is affected largely by root exudates, is the most suitable niche for these taxa, whereas they are unable to overcome the plant host immune system to successfully colonize the rhizoplane ( 51 , 52 ). Indeed, in the present study, the abundance of some genera that were significantly enriched in the rhizoplane compartment was relatively low (0.01 to 0.30%) ( Fig. S2 ). The distinctiveness of the root-associated microbiome has been substantiated by various lines of evidence ( 11 , 14 , 53 , 54 ). In the present study, we used the co-occurrence patterns to further examine the increasing effect of plant selection on fungal communities from the bulk soils to the rhizosphere to the rhizoplane. We recorded the highest DI and DSI values in the rhizoplane network, with the network complexity decreasing with the sequentially declining number and abundance of fungal taxa, especially those representing Ascomycota (i.e., the Ascomycota abundance decreased by 89.7%) ( Fig. 2 ; Table 3 and S2). Similar co-occurrence patterns in the bacterial community across the above compartments have been reported previously ( 22 ). Furthermore, the higher DI value in the rhizoplane than in the rhizosphere network (7.83 versus 0.94) may indicate the attractions for some specific taxa by root exudates but a selective inhibition of their colonization of the rhizoplane by the host plant. The distribution pattern of fungal community in the rhizoplane was similar in all soil types ( Fig. 1A ), suggesting a considerable contribution of the host plant to the microbiota rhizoplane specialization. Niu et al. ( 55 ) reported that a simplified synthetic bacterial community reproducibly assembled on the maize root surface by incorporating the distinctive microbiota assembled by maize roots. Hence, in the present study, the depletion of complex ecological associations in fungal communities associated with the rhizoplane highlights the potential of community simplification to target their application in the plant rhizoplane. Microbiome multifunctionality in an ecosystem is positively correlated with the overall community complexity ( 56 ). Hence, the depletion of both community complexity and certain fungal taxa in the rhizoplane compartment observed in the present study is conducive to directional regulation of their target functions. Interactions between different species determine the character of microbial community assembly and feedback effects on ecosystem function related to nutrient cycling ( 42 , 56 – 59 ). In the current study, positive interactions (74% to 77%) between microbes dominated the co-occurrence network in the three compartments, with a higher proportion of positive interactions in the rhizoplane than those in the bulk soil or the rhizosphere ( Fig. 2 ). Positive interactions between microbes at the same trophic level can be derived from facilitation, a process that is mutually beneficial to the interacting partners ( 60 ). Negative interactions can be an outcome of a competition for resources ( 61 ) or direct inhibition (antagonism) ( 62 ). Furthermore, in less diverse communities, fungi invest less energy in the inhibition of other microbes (e.g., by releasing secondary metabolites) ( 63 , 64 ) than that in vegetative growth and secretion of decomposition enzymes ( 65 , 66 ). Hence, the enhancement of facilitations among fungal taxa in the rhizoplane noted in the present study provides an opportunity to directly regulate the individual taxa to impact functional community performance. Furthermore, the interspecies relationships between fungi in the rhizoplane were closer, with a higher average clustering coefficient (0.451) than those in the bulk soil (0.417) and the rhizosphere (0.414) ( Fig. 2 ). Collectively, these observations confirm the existence of efficient interactions between fungal taxa in the rhizoplane, which contribute to the microflora effects related to both functional features and host plant traits. Indicator taxa play an important ecological role in microbiome assembly and ecosystem function ( 43 , 67 ). The differential abundance of the top-20 indicator genera identified in the compartment niches in the present study by using LEfSe was indicative of their sensitivity to host-mediated selection ( Fig. 3A ). Particularly, the indicator genera Humicola , Mortierella , Acremonium , Aspergillus , Podospora , Cercophora , and unclassified_Strophariaceae were more abundant in the rhizosphere than in the other two niches ( Fig. 3A ), which might indicate a selective inhibition of their colonization of the rhizoplane by the host plant. In the current study, Fusarium , Humicola , Mortierella , and Pyrenochaetopsis accounted for much of the microbial community variation along the soil-root continuum, with Fusarium and Mortierella more abundant (1.40% to 1.90%) than the other indicator genera ( Fig. 3A and Fig. S2 ). Furthermore, the relative abundances of Mortierella and Pyrenochaetopsis in the rhizoplane were good predictors of plant agronomic traits ( Fig. 3B ). That finding was not surprising, as Mortierella fungi grow rapidly on organic substrates, participate in soil nutrient cycling ( 68 ), and act as hub nodes in microbial networks for other plant species ( 40 ). Coincidently, in the present study, network hub nodes (ASV level) representing Mortierella , Pyrenochaetopsis , and Humicola showed high sensitivity to host plant selection in the overall community (De-taxa) ( Table S7 ). The hub nodes are thought to exert a disproportionately large effect on the ecosystem function and microbiome assembly ( 43 , 67 ). Furthermore, Fusarium fungi served as a hub node and important network hub in the bulk soil and the rhizoplane, respectively ( Table S7 ). The high abundance of Fusarium in the soil or root frequently indicates the occurrence of plant root rot ( 47 ). Thus, the identification of these indicator taxa provides critical information for future root-associated microbiome manipulation, facilitating the application of bioinoculants for plant growth. In conclusion, in the current study, our results demonstrated that the soil-root microbiome continuum is shaped largely by the host plant, which attracts and repels specific microbes, with a marginal influence of soil-type-dependent environmental factors. Furthermore, the depletion of complex ecological associations in the fungal community along the soil-root continuum and the enhancement of interactions among fungi in the rhizoplane provide empirical evidence for the potential of community simplification as an approach to target the plant rhizoplane for specific applications. Notably, the identified indicator taxa along the soil-root microbiome continuum provide critical information for the manipulation of the root-associated microbiome. However, sequencing data obtained in this study suggest the existence of several deeply divergent class-level fungal lineages that have not yet been described or sequenced previously, especially in the rhizoplane compartment. More efforts are needed for the continued research into fungal diversity and interactions for specific applications in the future."
} | 4,580 |
19842620 | null | s2 | 2,043 | {
"abstract": "The hydrophobicity of a surface can be enhanced by physical textures. However, no existing theories of surface wetting can provide guidance to pinpoint the texture size requirement to achieve super/ultrahydrophobicity. Here, we show that the three-phase contact line tension, tau, is an important link to understand the dependence of macroscopic wetting on physical texture size in an ideal Cassie regime. Specifically, we show that texture size is the dominant parameter in determining surface hydrophobicity when the size approaches a limiting physical length scale, as defined by tau and the surface tension of the liquid."
} | 156 |
34064293 | PMC8224282 | pmc | 2,044 | {
"abstract": "The Florida Keys, a delicate archipelago of sub-tropical islands extending from the south-eastern tip of Florida, host the vast majority of the only coral barrier reef in the continental United States. Abiotic as well as microbial components of the surrounding waters are pivotal for the health of reef habitats, and thus could play an important role in understanding the development and transmission of coral diseases in Florida. In this study, we analyzed microbial community structure and abiotic factors in waters around the Florida Reef Tract. Both bacterial and eukaryotic community structure were significantly linked with variations in temperature, dissolved oxygen, and total organic carbon values. High abundances of copiotrophic bacteria as well as several potentially harmful microbes, including coral pathogens, fish parasites and taxa that have been previously associated with Red Tide and shellfish poisoning were present in our datasets and may have a pivotal impact on reef health in this ecosystem.",
"conclusion": "5. Conclusions A growing body of research has introduced the concept of using microorganisms as bioindicators, which can provide an immediate and sensitive measure of water quality that can and should supplement abiotic water quality measurements. Waters surrounding reefs interact with different components of coral reefs and thus may have a strong impact on reef health. Our survey of the waters around the Florida Keys uncovered a high abundance of copiotrophic microbial taxa, including opportunistic pathogens. This survey represents only a snapshot of the potential factors in the waters that might influence reef health in the Florida Keys’ archipelago, and regular monitoring of microbes in conjunction with abiotic stressors will be pivotal to understand possible threats to reef health and can thus guide informed ecosystem management.",
"introduction": "1. Introduction Coral reefs around the Florida Keys constitute the main part of the third largest barrier reef ecosystem in the world [ 1 ]. In addition to the reefs, the area around the Keys is comprised of diverse habitats such as shallow seagrass meadows and mangrove forests. These ecosystems are constantly threatened by global climate change (e.g., ocean warming and ocean acidification), human activities (e.g., fishing and pollution), hurricanes, and tropical storms. In 1990, the Florida Keys National Marine Sanctuary (FKNMS) was established to protect the only coral barrier reef in the continental United States, which provides essential ecosystem services and represents a very important source of food and income for coastal communities [ 2 ]. FKNMS annually attracts nearly five million visitors who collectively contribute to its $4.4 billion economic value (data from 2017; https://marinesanctuary.org/wp-content/uploads/2019/07/FKNMS-Report-Final-072819.pdf , accessed on 11 May 2021) through marine-related activities in the sanctuary, including fishing, snorkeling, diving, wildlife viewing, boating and other activities. The Florida Keys Reef Tract has experienced several major disease outbreaks over the past four decades that have drastically changed the reef ecosystems [ 3 ]. Therefore, the preservation of the Florida Keys has become a national priority in the USA [ 4 ] and unprecedented restoration efforts are on the way to restore parts of the nearly 90% of original coral cover that was lost ( https://www.fisheries.noaa.gov/southeast/habitat-conservation/restoring-seven-iconic-reefs-mission-recover-coral-reefs-florida-keys , accessed on 11 May 2021). Most recently, stony coral tissue loss disease (SCTLD), which was first identified in 2014 off the coast of Virginia Key, has affected at least 23 reef-building coral species, especially on the outer reef parts [ 5 , 6 , 7 , 8 ]. The disease often results in whole colony mortality [ 5 , 9 , 10 ]. Aquaria studies have shown that disease transmission can occur through direct contact and through the water column. Additionally, disease lesions are significantly impacted (stopped or slowed) by antibiotic treatment, indicating a bacterial origin of the disease [ 7 ]. However, thus far, no single pathogen has been identified as the cause of this outbreak. Coral reefs and their well-structured associated microbial communities are extremely complex and should be seen as parts of an ecosystem with a strong benthic–pelagic exchange [ 11 , 12 , 13 ]. Therefore, the impact of abiotic and biotic components of reef waters on corals and coral health cannot be overestimated. While presumably not related to the current SCTLD outbreak in Florida, ocean warming, pH decrease, overfishing and coastal pollution are the main threats to coral reefs worldwide [ 14 , 15 , 16 ]. Increases in sea surface temperatures cause coral bleaching, which is recognized as one of the main concerns over the coming decades; however, an even greater threat can arise from the increasing frequency and impact of coral diseases [ 17 ]. The increased prevalence of potential coral diseases is presumed to be driven by nutrient enrichment in nearshore waters [ 14 , 18 ] and is usually also correlated to higher temperatures and increased total suspended solids (TSS; [ 19 ]). The FKNMS is directly influenced by water masses with distinct nutrient content, including the Florida Current, the Gulf of Mexico Loop Current, inshore currents of the SW Florida Shelf, discharge from the Everglades through the Shark River Slough, as well as by tidal exchange with both Florida Bay and Biscayne Bay [ 20 , 21 , 22 ]. In the present study, abiotic measurements were combined with microbial community analyses to analyze water quality in waters around the Florida Reef Tract.",
"discussion": "4. Discussion The waters around the Florida Reef Tract are generally oligotrophic and nutrient-deplete [ 22 ]. Year-long seasonal monitoring of abiotic water quality parameters (including the dataset presented here) has shown that waters have usually slightly higher turbidity and nutrient concentrations on the Gulf of Mexico side of the Keys than on the Atlantic side, along the reef tract [ 22 ]. Despite complex water patterns and significant differences in anthropogenic pressure along the Florida Reef Tract, the microbial community composition, at least during our sampling event, was not strikingly different in the different waters around the Keys and no distinct coherent clustering by longitude/latitude or sample type was apparent. Elevated organic carbon concentrations can directly impact coral microbiomes and increase coral mortality [ 35 ]. The importance of increased organic carbon for microbial community structure in reef waters was well demonstrated by our data, as TOC concentrations did have a significant impact on both prokaryotic and eukaryotic microbial community composition, together with temperature and DO concentrations. The data presented in this study show that the total abundances of unpigmented cells are not necessarily correlated with organic carbon concentrations, and even though higher abundances of HNA bacteria were more likely to be found in areas with elevated TOC concentrations, this trend was not consistent. More detailed studies on microbial functions rather than relative abundances of certain microorganisms are necessary to address questions related to changes in functional diversity of planktonic microbes near healthy and unhealthy reefs. Nevertheless, previous data on microbial community composition in reef environments have identified potential bioindicator species and their relationship to abiotic stressors. Existing monitoring data that combine microbial data and abiotic data in reef habitats demonstrated that high temperatures are usually correlated to an increase in taxa belonging to Rhodobacteraceae , Cryomorphaceae , Synechococcaeae , Vibrio and Flavobacterium (which include putative coral pathogens and opportunistic bacteria). Flavobacteriaceae -affiliated taxa are significant indicators correlated to high chlorophyll a (Chl a), TSS and particulate organic carbon (POC) concentrations. Halomonadaceae are significantly correlated with high Chl a and TSS, and representatives of the phylum Verrucomicrobia are significant indicators correlated with high TSS levels [ 19 ]. Opportunistic copiotrophic taxa, such as Cryomorphaceae , Flavobacteriaceae and Rhodobacteraceae, are usually more prevalent in the higher nutrient nearshore waters (e.g., [ 36 , 37 ]). Recent studies have pinpointed Flavobacteriaceae -affiliated taxa as indicators for increased organic nutrients at the Great Barrier Reef [ 19 , 38 ]. We found that the relative abundances of Cryomorphaceae PS008 were significantly correlated with TOC concentrations in the water (R 2 = 0.77, p < 0.001). Interestingly, several bacterial groups within the order Flavobacteriales , including Cryomorphaceae , are more likely to be present within SCTLD-diseased coral tissue than within apparently healthy coral tissue [ 6 ]. Greater abundances of Rhodobacteraceae and Cryomorphaceae have been also shown at inlet-influenced coastal waters of southeast Florida, where the SCTLD outbreak began [ 39 , 40 ]. PS001, classified as Rhodobacteraceae , was the most abundant prokaryotic species-level taxa in our dataset and was found relatively abundant throughout the Florida Keys archipelago. Rhodobacterales have been also detected at higher relative abundances in SCTLD lesion samples [ 8 ]. Usually, bacteria associated with coral disease are rarely detected in seawater due to their low concentrations [ 41 , 42 ]. In this case, the same taxa that are abundant in corals with SCTLD symptoms are abundantly found in the water column, which further supports the hypothesis that some of these organisms may act as secondary opportunistic pathogens associated with progression of SCTLD [ 43 ]. Terrestrial runoff that leads to organic enrichment of coastal waters and sediments has been identified as a key process in the degradation of coral reefs [ 44 , 45 ]. Increases in organic matter result in reduced O 2 concentrations, lower pH, and formation of hydrogen sulfide, a potent toxin to most organisms, which can accelerate the spread of reef colony mortality [ 44 ]. The high relative abundance of potentially sulfur-oxidizing Thioglobaceae (PS007) in the water column indicates high sulfide production in areas where SCTLD had spread at the time of the study (Upper and Middle Keys; [ 40 ]). Not only prokaryotes, but also microbial eukaryotes are being used to assess water quality in coral reef ecosystems [ 46 ]. Microalgae are generally enriched nearshore due to high nutrient and resuspension requirements [ 47 ]. The most abundant eukaryotic group in our dataset, Scrippsiella , is a non-toxic, cosmopolitan marine dinoflagellate that can be found in both cold and tropical waters, where it is known to produce “red tide” events. Scrippsiella blooms can lead to oxygen depletion, resulting in fish kills [ 48 ], and have been reported in the Southern Gulf of Mexico and the coastal United States [ 49 ]. Dinoflagellates of the order Suessiales , which also contains the genus Symbiodinium , the main phototrophic coral symbiont, were also found in the water column. High nutrient concentrations [ 50 , 51 , 52 , 53 ], but also high numbers of dinoflagellates in reef waters impose potential threats to Symbiodinium . Firstly, it may lead to the spread and proliferation of viruses that could also infect zooxanthellae species [ 54 , 55 , 56 ], and secondly, it may increase activity and impact of algicidal bacteria against dinoflagellates [ 57 ]. While Mayali and Azam [ 58 ] initially reported that species belonging to the Bacteroides are the most abundant and widely isolated algicidal bacteria, a recent study identified culturable algicidal bacteria from a wide range of taxa [ 59 ]. Dinoflagellates are known as one of the major components of diverse marine ecosystems (e.g., [ 60 ]). They often form red tides or harmful algal blooms that sometimes cause human illness and large-scale mortality of fin-fish and shellfish [ 61 , 62 , 63 ]. Thus, the high abundance of dinoflagellates in the waters of a region that has a large tourism industry is of critical concern, not only in relation to SCTLD, but also in relation to other human interests. Our study indicates that dinoflagellates contributed to a larger fraction of the EMC compared to abundances reported by a study that was carried out in the Florida Keys three years earlier [ 64 ], although this could reflect seasonal variation or differences in the composition of sampling locations. Several abundant eukaryotic OTUs were assigned to well-described pathogens. Four OTUs were classified as Syndiniales , an order of dinoflagellates exclusively composed of marine parasites that infect a wide range of hosts, from fish larvae to dinoflagellates, including Scrippsiella [ 65 , 66 ]. Other parasitic eukaryotes in our dataset included Paradinium sp. (parasites of various copepod hosts, [ 67 ]), Rozella ( Cryptomycota ), a genus of endoparasites of a broad range of hosts [ 68 ], including other parasites [ 69 ], and Labyrinthulids, (endophytic net slime molds, most of which are opportunistic pathogens found in association with marine vegetation, including seagrasses and mangroves [ 70 , 71 ]). As our dataset only provides a snapshot of the microbial communities in these waters, establishing a regular monitoring program would be essential to assess changes in the relative abundances of these organisms."
} | 3,390 |
27199545 | PMC4869604 | pmc | 2,045 | {
"abstract": "Microbiomes are ubiquitous and are found in the ocean, the soil, and in/on other living organisms. Changes in the microbiome can impact the health of the environmental niche in which they reside. In order to learn more about these communities, different approaches based on data from multiple omics have been pursued. Metagenomics produces a taxonomical profile of the sample, metatranscriptomics helps us to obtain a functional profile, and metabolomics completes the picture by determining which byproducts are being released into the environment. Although each approach provides valuable information separately, we show that, when combined, they paint a more comprehensive picture. We conclude with a review of network-based approaches as applied to integrative studies, which we believe holds the key to in-depth understanding of microbiomes.",
"conclusion": "Conclusion and Future Directions In this article, we have discussed how three different omic approaches – metagenomics, metatranscriptomics, and metabolomics – provide useful information toward understanding microbiomes. We also discussed how the value of an integrative approach is greater than the sum of its parts. Biological networks have long been used to model interactions between biological entities, with applications to areas, such as gene regulation, metabolic and signaling pathways, protein–protein networks, and food webs in ecology. 156 – 159 With its proven application to analyzing interrelationships and their critical role in multiomics, we believe biological network analysis will be critical to future multiomic approaches to studying the microbiome. In addition, network analyses offer the possibility of exploring both local (eg, relationship with neighbors) as well as global properties (eg, connectivity) of a community. Dutkowski et al. 160 studied the assignment of ontologies using networks and developed tools, such as Cytoscape, 161 to perform these analyses. Metagenomic studies have shown that interactions within a microbiome can be naturally modeled using a network representation, 14 , 42 , 162 with properties closely related to social networks . 15 , 24 Macroscale community structures have been observed in these types of networks, indicating clubs (ie, groups of co-occurring bacteria) as well as rival clubs (ie, groups of bacteria that tend to not co-occur). 15 , 42 In order to integrate data from various omic sources, microbiomes can also be modeled as heterogeneous networks ( Fig. 3 ), which provides a visual description of what such a network in the context of the microbiome would look like. A heterogeneous network would allow researchers to generate new interesting hypotheses that involve entities from the different omics described in this article (represented in the figure by nodes with different shapes and colors). For instance, we could potentially have a club that includes genes, microbes, and metabolites. Heterogeneous networks have been used in other applications, such as associations between genetic interactions and protein–protein interactions in order to infer cellular function. 163 Another study couples these same types of networks to infer gene dependencies and new processes, such as DNA damage repair, and also different types of co-expression networks. 164 Many types of omic networks were also integrated to study gene regulation in the bacterium Mycobacterium tuberculosis . 165 Other omic areas not included in this study include metaproteomics, metalipidomics, and metaglycomics. We believe that analyzing heterogeneous networks built across multiple omic datasets is critical to linking the different levels of complexity inherent to biological systems, thus establishing a more comprehensive understanding of the nature and dynamics of microbiomes.",
"introduction": "Introduction Communities of microbes are found in diverse environmental niches, such as the ocean, soil, and inside host organisms, including all animals, plants, and lower eukaryotes. 1 These communities show characteristics, such as complexity, diversity, interaction, cooperation, dynamism, generosity, danger, and competition. 2 In such communities, microbes may compete for nutrients, 3 share functional genes through horizontal gene transfer, 4 produce toxins that can kill other microbes, 5 produce various metabolites and signaling molecules for sharing and communication, 6 and combine forces to fight common enemies, such as the host immune system. 7 In short, the importance of the microbial community stems from the fact that they are critical to the health of the environmental niche in which they reside, 8 and an imbalance in the community could be harmful. 9 Traditionally, a microbiome has been defined as a microbial community occupying a reasonably well-defined habitat. 10 One of the most common approaches to studying a microbiome is analyzing its constituent microbial genomes through meta-genomics. More recently, this definition has evolved to include not only the microbes and their genomes but also the aggregate of environmental and host factors. The inclusion of the host environment as part of the microbiome significantly expands its implications, with the interactions between the host and its associated microbial community now relevant to understanding the dynamics of the microbiome. For evolutionary and functional studies of the microbiome, modifications in the host environment (eg, a diet shift in the host organism or a compositional change in the environmental matrix under study) now become critical and must be taken into consideration. Coevolution processes can then be identified, providing valuable information to understand the relationship of the microbial community with its host. This apparent conceptual shift is accompanied by the recognition that, in order to achieve a more comprehensive study of microbiomes, metagenomics must be combined with other omic approaches. Many relevant omic approaches have been proposed for microbiome studies. In this article, we discuss metatranscriptomics and metabolomics, which are rapidly becoming critical to microbiome studies. Metagenomics is the study of the genomes in a microbial community and constitutes the first step to studying the microbiome. As seen in the “Metagenomics” section, metagenomics comes in different flavors. However, its main purpose is to infer the taxonomic profile of a microbial community. Although whole-metagenome sequencing (WMS) provides a partial glimpse into the functional profile of a microbial community, it is better inferred using metatranscriptomics, which involves sequencing the complete (meta)transcriptome of the microbial community. Metatranscriptomics informs us of the genes that are expressed by the community as a whole. With the use of functional annotations of expressed genes, it is possible to infer the functional profile of a community under specific conditions, which are usually dependent on the status of the host. While metagenomics helps address the question “what is the composition of a microbial community under different conditions?”, and metatrascriptomics helps answer the question “what genes are collectively expressed under different conditions?”, the question considered by metabolomics is “what byproducts are produced under different conditions?”. The metabolites released by the microbial community are largely responsible for the health of the environmental niche that they inhabit. Regardless of whether microbiome studies are biomedical or environmental in their focus, it is clear that the different omic approaches provide invaluable information. However, the best results are obtained by performing integrative studies that involve all available omic datasets. 11 While such efforts hold promise, the integration must be done carefully. 12 As suggested by a variety of different analyses, 13 – 16 we believe that network-based approaches can lead to a sophisticated in-depth analysis of microbiomes, particularly when applied to integrative studies, and consequently lead to critical insights into the world of microbiomes. Major microbiome initiatives Human microbiome studies The National Institute of Health has funded a major initiative that aims to generate resources for a comprehensive characterization of the human microbiome to understand its impact on human health and disease. The first phase, known as the Human Microbiome Project (HMP), 17 focuses on the study of microbial communities that inhabit the human body of healthy individuals, 18 , 19 with particular emphasis on nasal, oral, skin, gastrointestinal, and urogenital areas. 17 , 18 , 20 – 23 It is known that the amount of microbial cells present in the human body is notably larger than the amount of human cells. These bacterial communities play critical roles, such as assisting in the digestion of food, synthesizing necessary vitamins, and aiding the immune system in defending our body from pathogenic invaders. 24 Human microbiome studies have revealed strong correlations between changes in microbial community profiles and diseases. 22 , 25 – 27 These studies have also shown that the structure of the microbial community is significantly different in five areas of the human body (gut, mouth, airways, urogenital, and skin), and that this seems to be independent of gender, age, and ethnicity. 18 , 19 All the data and protocols associated with this project are available at the HMP Data Analysis and Coordination Center (DACC). 28 The Integrative HMP (iHMP) 27 is the second phase of this initiative, going a step further by gathering multiple omic data from both the microbiome and the host. This is part of a longitudinal study with a broader objective of understanding host–microbiome interactions using integrative analyses. Another related initiative focused on the human microbiome is the Metagenomics of the Human Intestinal Tract (MetaHIT) project. 29 This project was funded by the European Seventh Framework Programme until 2012. Its goal was to understand the link between the human intestinal microbiota and human health/disease. For this purpose, they focused on two disorders of increasing incidence in Europe: obesity and inflammatory bowel disease. Similarly, the Human Food Project and the American Gut Project 30 focus on the gut microbiome with the aim of determining how to acquire a healthy microbiome through food. Environmental microbiome studies The Earth Microbiome Project (EMP) is a remarkable effort started in 2010 to characterize the diversity, distribution, and structure of microbial ecosystems across the planet and has already gathered over 30,000 samples. 31 Their focus is on diverse ecosystems, including not only the ones within the bodies of humans, animals, and plants but also terrestrial, marine, freshwater, sediment, air, and constructed environments, as well as every intersection of these ecosystems. J. Craig Venter Institute’s (JCVI) Global Oceanic Sampling (GOS) expeditions and the European Tara Oceans initiatives 32 – 36 have focused on understanding and cataloging the marine microbiome diversity across the planet. JCVI’s vessel, Sorcerer II , has made multiple oceanic expeditions to collect samples from oceans across the globe. Their multistage processing allows them to exploit size differences to separate different groups of microbes, including large micro-zooplankton and phytoplankton (3–20 µm), picoplankton and large cyanobacteria (0.8–3 µm), prokaryotes and large viruses (0.1–5 µm), and viroplankton (below 0.1 µm).",
"discussion": "Discussion In summary, metataxonomics helps us to compute the taxonomic profile of a microbial community, while metagenomics helps us to compute the functional profile by focusing on the gene content and using the available functional annotations of the corresponding proteins. While metagenomics is powerful, solely using it to study a microbiome is limited in value. Many experts have confirmed that the percentage of documented bacteria is very low compared to the estimate of bacterial species on our planet. 82 This may be due partially to the impossibility of culturing complex environments or replicating in the laboratory the real conditions in which the microbiome exists. Either way, the reference databases used to classify and label bacteria are limited to what has been cataloged. Current methods typically either discard reads from undocumented microbes or label them based on the closest documented microbe from the database. Thus, inevitably, results will be based on a biased percentage of bacteria present in the samples, representing the first shortcoming of these methods. Another limitation is that metagenomics cannot reveal dynamic properties, such as the spatiotemporal activity of the community and the impact of the environment on these activities. The only information that can be obtained at a functional level is the potential of the microbiome to display functional properties associated with the presence of genes with no information about their expression levels or lack thereof. The need to monitor gene expression patterns brings us to the topic of our next section, metatranscriptomics.\n\nDiscussion Although current metatranscriptomic techniques are promising, there are still several obstacles that limit their large-scale application. First, much of the harvested RNA comes from ribosomal RNA, and its dominating abundance can dramatically reduce the coverage of mRNA, which is the main focus of transcriptomic studies. Some efforts have been made to effectively remove rRNA. 107 Second, mRNA is notoriously unstable, compromising the integrity of the sample before sequencing. Third, differentiating between host and microbial RNA can be challenging, although commercial enrichment kits are available. This may also be done in silico if a reference genome is available for the host, as in the work of Perez-Losada et al. 108 who consider the impact of host–pathogen interactions on the human airway microbiome. Finally, transcriptome reference databases are limited in their coverage. WMS approaches provide information on the taxonomic profile of a microbial community as well as its potential functional profile; in contrast, whole metatranscriptome sequencing describes the active functional profile. This would help in studying the dynamics of functional profiles with varying conditions. We now discuss metabolomics, which studies the consequences of the shifts in the collective gene expression of the microbial community that modifies the very medium where the microbial community must feed, grow, reproduce, and cooperate or compete to survive.\n\nDiscussion By cataloging all metabolites present in a sample, metabolomics offers a powerful way to relate the metabolites to the cellular processes of which they are the byproducts. The combination of metabolomic and pathways information can lead to new hypotheses. One important challenge of this approach is difficulty in determining whether a metabolite was generated by the host or by the microbiome. In addition, if conclusions are to be made about which genes, enzymes, or pathways are associated with a specific metabolite, the results obtained from a metabolomic study must be combined with other omic data. This highlights the need for new approaches that deal with integrated omics, as discussed in the “Integrating multiomic data” section."
} | 3,850 |
34774761 | null | s2 | 2,046 | {
"abstract": "Cyanobacteria hold promise for renewable chemical production due to their photosynthetic nature, but engineered strains frequently display poor production characteristics. These difficulties likely arise in part due to the distinctive photoautotrophic metabolism of cyanobacteria. In this work, we apply a genome-scale metabolic model of the cyanobacteria Synechococus sp. PCC 7002 to identify strain designs accounting for this unique metabolism that are predicted to improve the production of various biofuel alcohols (e.g. 2-methyl-1-butanol, isobutanol, and 1-butanol) synthesized via an engineered biosynthesis pathway. Using the model, we identify that the introduction of a large, non-native NADH-demand into PCC 7002's metabolic network is predicted to enhance production of these alcohols by promoting NADH-generating reactions upstream of the production pathways. To test this, we construct strains of PCC 7002 that utilize a heterologous, NADH-dependent nitrite reductase in place of the native, ferredoxin-dependent enzyme to create an NADH-demand in the cells when grown on nitrate-containing media. We find that photosynthetic production of both isobutanol and 2-methyl-1-butanol is significantly improved in the engineered strain background relative to that in a wild-type background. We additionally identify that the use of high-nutrient media leads to a substantial prolongment of the production curve in our alcohol production strains. The metabolic engineering strategy identified and tested in this work presents a novel approach to engineer cyanobacterial production strains that takes advantage of a unique aspect of their metabolism and serves as a basis on which to further develop strains with improved production of these alcohols and related products."
} | 444 |
33737724 | null | s2 | 2,048 | {
"abstract": "To circumvent the von Neumann bottleneck, substantial progress has been made towards in-memory computing with synaptic devices. However, compact nanodevices implementing non-linear activation functions are required for efficient full-hardware implementation of deep neural networks. Here, we present an energy-efficient and compact Mott activation neuron based on vanadium dioxide and its successful integration with a conductive bridge random access memory (CBRAM) crossbar array in hardware. The Mott activation neuron implements the rectified linear unit function in the analogue domain. The neuron devices consume substantially less energy and occupy two orders of magnitude smaller area than those of analogue complementary metal-oxide semiconductor implementations. The LeNet-5 network with Mott activation neurons achieves 98.38% accuracy on the MNIST dataset, close to the ideal software accuracy. We perform large-scale image edge detection using the Mott activation neurons integrated with a CBRAM crossbar array. Our findings provide a solution towards large-scale, highly parallel and energy-efficient in-memory computing systems for neural networks."
} | 290 |
30814502 | PMC6393428 | pmc | 2,049 | {
"abstract": "Bioinspired ceramics with micron-scale ceramic “bricks” bonded by a metallic “mortar” are projected to result in higher strength and toughness ceramics, but their processing is challenging as metals do not typically wet ceramics. To resolve this issue, we made alumina structures using rapid pressureless infiltration of a zirconium-based bulk-metallic glass mortar that reactively wets the surface of freeze-cast alumina preforms. The mechanical properties of the resulting Al 2 O 3 with a glass-forming compliant-phase change with infiltration temperature and ceramic content, leading to a trade-off between flexural strength (varying from 89 to 800 MPa) and fracture toughness (varying from 4 to more than 9 MPa·m ½ ). The high toughness levels are attributed to brick pull-out and crack deflection along the ceramic/metal interfaces. Since these mechanisms are enabled by interfacial failure rather than failure within the metallic mortar, the potential for optimizing these bioinspired materials for damage tolerance has still not been fully realized.",
"introduction": "Introduction New materials for structural applications in aerospace, energy, and transportation often have the requirement to operate safely at high temperatures in aggressive environments and, for aviation applications, have low density. Ceramic materials in many respects represent an ideal solution to this problem, but their use has been severely compromised by the fact that they invariably display near-zero tensile ductility and low fracture toughness values, which makes them prone to sudden catastrophic failure. Nature, however, is particularly adept at designing damage-tolerant ceramic-like materials with excellent strength and toughness, using a small palette of individual constituents with relatively meager mechanical properties. Nature develops remarkable materials through sophisticated hierarchical, multiple length-scale architectures that optimize the mechanical properties of the hard mineral and soft organic phases, often with compositional, orientation, or structural gradients and graded interfaces 1 – 4 . A notable example here is nacre, which is known to have a fracture toughness three orders of magnitude higher (in energy terms) than its constituents. Nacre is found in mollusk shells, such as abalone, sea snails, and various bivalves. The particularly well-studied variety found in abalone shells is comprised of ~95 vol.% aragonite mineral (calcium carbonate) and ~5 vol.% biopolymer 1 , 2 , 4 , 5 , formulated into a brick-and-mortar microstructure that enables multiple toughening and strengthening mechanisms. The brick-and-mortar microstructure facilitates the creation of damage tolerance. The mineral “bricks” provide strength, whereas limited (a few micrometer) displacements within the biopolymeric “mortar” act to dissipate locally high stresses, thereby providing a degree of ductility that promotes toughness. The salient toughening mechanisms are principally extrinsic 6 and involve crack deflection and primarily brick pullout leading to crack bridging 4 . The sliding behavior within the mortar is essential for toughening, but it must be limited to retain strength 1 . It is restricted due to the roughness on the platelet surfaces and the tensile and shear strength of the biopolymers that act as a glue between the platelets 7 ; in certain organisms, tablet interlocking due to the dovetail geometry 8 of the platelets and mineral bridges that connect the platelets 4 , 9 can also inhibit platelet sliding. Nature’s precise tailoring of its structures and constituents’ properties, e.g., the use of high-aspect mineral bricks with a tensile/shear-resistant mortar can generate mechanical performance that is comparable with advanced engineering ceramics 5 . Computational models and biomimetic approaches therefore suggest that there is potential to create new lightweight structural materials with current engineering ceramics by utilizing the design principles and strength/toughening mechanisms active in nacre and other biological materials 8 – 10 . A promising method to recreate nacre-like materials using ceramics is freeze-casting followed by infiltration of a compliant (mortar) phase, such as a polymer or metal 11 – 13 . Freeze-casting is a method of creating a porous ceramic scaffold by mixing ceramic particles in water and then freezing the water with a temperature differential to induce lamellar ice growth. Once the ice freezes, it can be sublimated leaving a porous ceramic structure in the negative image of the ice structure 14 . This technique can create ceramic materials with complex hierarchical microstructures that can be controlled using various casting conditions and additives in the solvent 15 – 20 . One notable example is a freeze-cast alumina scaffold that was pressed and infiltrated with polymethyl methacrylate as a mortar to mimic the nacre architecture 11 . With 80 vol.% ceramic in a brick-and-mortar structure, it displayed a strength of ~225 MPa and an extremely high toughness in excess of 30 MPa·m ½ , making it one of the toughest ceramic materials reported to date. Indeed, there have been comprehensive review papers on various experiments to freeze-cast 21 or to use other methods 22 , 23 to create such nacre-like ceramics. There has been sustained interest in creating ceramic structures with a nacreous microstructure, particularly as computational models have suggested that the strength of nacre-inspired ceramics could increase, without sacrificing toughness, if the polymer-compliant phase was replaced by a metal 8 . This direction has the best potential to create exciting new lightweight structural materials with high strength and toughness, and a capability for high-temperature functionality. To make such biomimetic materials, methods such as coextrusion 24 – 26 , spark-plasma sintering 27 , and additive manufacturing 28 have been attempted with varying success due to issues with undesirable reactions, dewetting between the metal and the ceramic phases, or coarseness of the ceramic brick phases. Melt-infiltration after freeze-casting a ceramic scaffold represents a liquid-processing technique that has the potential to fabricate ceramic–metal hybrid materials with strong interfacial bonding 29 . Hybrid ceramics with a lamellar structure have been fabricated by pressure-assisted infiltration with conventional Al alloys 30 – 32 , but those contained less than 40 vol.% of ceramic, which is significantly lower than the mineral content in nacre. Furthermore, as the volume fraction of reinforcement increases and the pore spacing decreases in a scaffold, the pressure required to completely infiltrate the scaffold increases, leading to deformation or cracking of the scaffold 33 , 34 . Since the magnitude of the applied pressure is closely related to the wettability of the ceramic by a molten alloy, strategies such as the addition of a reactive element in the matrix alloy or coating the surface of reinforcements have been applied to improve the wettability and therefore feasibility of infiltration between the alloy and ceramic 33 . For example, Mg was used as a reactive element to fabricate bioinspired hybrid materials using pressureless infiltration 28 , 35 – 37 , but the microstructure was lamellar with a low ceramic content compared to the brick-and-mortar structure. In an attempt to solve this problem, we use reactive wetting to create a nacre-like, high volume-fraction ceramic material containing a metallic-compliant phase. We selected a Zr-based bulk-metallic glass (BMG) as the mortar phase, as this alloy shows excellent wettability with alumina 38 , 39 . BMGs are metallic alloys with a disordered, noncrystalline atomic structure that can be cast in bulk (over 1 mm) layers. Zr-based BMGs have been shown to have high strength with some degree of toughness, which represent ideal properties as a compliant-phase (mortar) material. As a result, the combination of perfect wettability and mechanical properties of the BMG makes it an interesting candidate for infiltration into a brick-like ceramic scaffold. However, processing of BMGs introduces challenges as they can oxidize at elevated temperatures 40 and require a critical cooling rate to solidify in the amorphous state. Moreover, the final material must be used below the glass-transition temperature as these glasses can embrittle upon crystalization 41 . While fibrous or particulate BMG-matrix composites have been developed by melt-infiltration 42 , 43 , this study relates their infiltration behavior to the thermophysical properties and wettability of the BMG, in order to tailor the properties of the ceramic/metal interfaces while maintaining the BMG in the fully amorphous state as a metallic mortar in nacre-like material. Here, we synthesize “nacre-like” (high-volume fraction) alumina structures, containing a glass-forming alloy as the compliant phase, processed via pressureless infiltration. Contact-angle measurements and electrostatic levitation were used to identify the length scale of the ceramic scaffolds and their excellent wettability with a liquid alloy as crucial factors to successfully synthesize these materials. Near-perfect wetting within seconds for the Zr-based BMG on alumina was observed, indicating that a conformal bond can be attained at the interface of the two materials, which in turn implies that high strength may be realized in an Al 2 O 3 /BMG hybrid material 44 , 45 . However, the interfacial strength can markedly change due to the formation of brittle interfaces upon solidification of the metallic phase, which can compromise the damage tolerance of the bioinspired material. Finally, we examine how processing conditions can affect the ceramic–metal interfaces to discern the fundamental mechanisms underlying the flexural strength and fracture toughness of these bioinspired nacre-like ceramics.",
"discussion": "Discussion In this work, we have created a freeze-cast alumina ceramic with a bioinspired, nacre-like, brick-and-mortar structure with a Zr-based BMG as the infiltrated compliant-phase mortar. Our motivation is that nacre-like alumina infiltrated with a polymer-compliant phase has been shown to have exceptional toughness, and theoretical modeling has predicted that the mechanical properties may be even better with a metallic mortar 8 . To overcome the issues associated with infiltrating a metal into a ceramic scaffold, we have employed the concept of reactive wetting by utilizing a Zr-based BMG as the infiltrating phase. We discuss below the features of our processing procedures to create these materials, our associated attempts at tailoring the ceramic–metal interfaces, and finally how these factors can result in enhanced damage-tolerant properties in nacre-like structural ceramics. The excellent wettability allowed spontaneous infiltration of the nacre-like alumina scaffolds at very rapid rates (10 min hold time) without the need for applied pressure. This was possible due to the high bonding strength between the alumina and the BMG melt, which can be up to an order of magnitude higher than the nonreactive metal/alumina couples. This was measured using the work of adhesion, which was calculated from surface tension and the final contact angles, as shown in Supplementary Note 1 of the Supplementary Information . This high bonding strength also leads to high capillary pressures induced by the BMG melt. Based on the calculations in Supplementary Note 2 , we found that the capillary pressures induced by the BMG melt are the same or higher than the external pressure applied for complete infiltration in other experiments, leading to spontaneous infiltration. This is true despite the high viscosity of the BMG melt because the length scale of the ceramic scaffold is sufficiently large to accommodate the infiltration kinetics of the system according to Darcy’s law, as shown in Supplementary Note 3 . However, infiltrating finer-scale scaffolds (similar to the dimensions found in natural nacre) with this BMG would present a problem as the high viscosity of the BMG melt would prevent complete infiltration (Supplementary Fig. 2 ). This indicates that for future synthesis of nacre-like materials using infiltration, it is critical to observe the length scale of the scaffolds along with the viscosity and wettability of the melt to find feasible processing conditions. The temperature used for infiltration had a marked effect on the flexural strength and fracture toughness of both the lamellar and brick-and-mortar structures. This is particularly clear for the samples with a nacre-like architecture, where samples infiltrated at 1273 K compared with the materials infiltrated at 1153 K displayed a factor-of-four lower strength, yet a factor-of-three higher toughness, caused by stable crack growth (and rising R-curve behavior) rather than sudden catastrophic fracture. The latter characteristic of sustaining stable cracking following crack initiation, without immediate catastrophic failure, represents an essential property of structural materials. For the case of the 1273 K infiltrated brick-and-mortar materials, this resulted from significant crack propagation along the BMG-mortar/alumina–ceramic interfaces. Indeed, all these strength and toughness properties are primarily related to the nature of these interfaces. In metal–ceramic systems showing reactive wetting, the interfacial bonding is affected by the chemical reaction between the melt and ceramic 50 . Furthermore, as noted above, the BMG melt infiltrated into the scaffolds can be partially or fully crystallized during solidification due to the interface that acts as a site for heterogeneous nucleation. Thus, the effect of temperature on the reaction and crystallization of the BMG mortar are vital issues to be considered to evaluate the interfacial bonding in these nacre-like alumina/glass-forming hybrid materials. Figure 10a, b shows the interface of the hybrids with a lamellar structure that were infiltrated at 1153 and 1273 K. The material infiltrated at 1153 K (Fig. 10a ) shows sharp interfaces between the metallic mortar and alumina, and a faceted crystalline (Zr 2 Cu) phase with a thickness of 0.3 ± 0.2 μm. Reaction products like ZrO 2 were not observed at the resolution of the SEM, indicating the slow kinetics of the reaction (3Zr + 2Al 2 O 3 → 4Al + 3ZrO 2 ) at 1153 K. The material infiltrated at 1273 K (Fig. 10b ) shows erosion of the interface, which produces a rough boundary, due to the reaction between the BMG melt and alumina. It should be noted that voids (marked by yellow arrows) and cracks are concentrated in the layer with a thickness of less than 2 μm between the faceted Zr 2 Cu phase and alumina (marked as A in Fig. 10b ). The layer also contained round-shaped pores (marked by blue arrows) that were generated from grain pullout during mechanical polishing. Examining this morphology, it is clear that the interfacial intermetallic layer provides a weak interface that creates a tortuous path for crack propagation. This is further supported by the micro-cantilever experiments, which showed a clearly weaker interface strength for the materials infiltrated at 1273 K (Fig. 9d, f ), as compared with those infiltrated at 1153 K (Fig. 9c, e ). Fig. 10 SEM micrographs showing the interface of the compliant-phase alumina/BMG-forming materials with a lamellar structure infiltrated at 1153 and 1273 K. a The material infiltrated at 1153 K displays a sharp interface between the metallic mortar and alumina, and faceted crystalline (Zr 2 Cu) phase with a thickness of 0.3 ± 0.2 μm. b In contrast, the interface of the material infiltrated at 1273 K reveals erosion of the interface, which produces a rough boundary, due to the significant reaction between the BMG melt and alumina. Scale bar: 1 μm During solidification, the undercooled BMG melt experiences abrupt volume contraction accompanied by crystallization. The Zr-based BMG used in this study shows about 1.5% volumetric contraction if the BMG melt is fully crystallized at its melting point 51 . To estimate the degree of volume contraction in the metallic mortar, the crystallinity of the metallic mortar was evaluated using a comparison of the heat of crystallization from the alumina/glass materials with the heat of crystallization from the monolithic BMG from the DSC analysis 52 (Fig. 5 ). The alumina/glass-forming alloy materials with a lamellar structure infiltrated at 1153 K and 1273 K had a crystallinity of 6% and 47%, respectively, which is comparable with the ratio between the thickness of the crystalline phase and that of metallic mortar in the materials with a lamellar structure (5% and 48%, respectively), as confirmed by microstructural analysis. Thus, a volume contraction of more than 0.7% takes place in the metallic mortar when the lamellar scaffold is infiltrated at 1273 K and then water-quenched. Furthermore, the alumina/glass-forming alloy materials with the brick-and-mortar structure infiltrated at 1273 K experience a volume contraction of the metallic mortar of about 1.5% due to their high crystallinity (97%). The significant volume contraction within this material leads to the formation of interfacial cracks or voids, resulting in a low flexural strength in the nacre-like materials. As the brittle interface forms at higher temperatures, the crack patterns are more tortuous, making it harder to propagate through the sample. The fracture toughness tests further support this hypothesis because the brick-and-mortar samples infiltrated at 1273 K show significant crack deflection compared with those infiltrated at 1153 K, as shown in Fig. 7 . The brick-and-mortar material infiltrated at 1153 K has a much higher flexural strength associated with its stronger ceramic/metal interface, but it appears to be too strong as the ceramic bricks fracture. In contrast, the brick-and-mortar samples infiltrated at 1273 K show several extrinsic toughening mechanisms (typical of natural nacre 1 , 4 ), such as inter-brick displacements, brick pullout, and crack deflection, leading to a significant increase in fracture toughness, compared with the corresponding structures infiltrated at 1153 K. However, the decreased flexural strength of samples infiltrated at 1273 K compared with those infiltrated at 1153 K is a strong indication that the alumina/BMG interfacial strength decreases with higher infiltration temperatures. So, what do we learn from this work with respect to ceramic brick-and-mortar structures? First, we have shown that it is possible to rapidly infiltrate a ceramic scaffold with a metal mortar without any applied pressure using reactive wetting. Moreover, this can result in a high-strength structure due to a strong ceramic/metal (brick/mortar) interface in a high-volume-fraction ceramic structure, as shown for infiltration at 1153 K. However, for toughness, we also require some degree of ductility or shielding, ideally with a structure comprising a high-aspect ratio, fine scale, and bricks in conjunction with a mortar that permits some degree of inter-brick displacements to alleviate any locally high stresses. We achieved this with the structure infiltrated at 1273 K. Instead of outright catastrophic failure, the 1273 K infiltrated structures displayed markedly rising R-curve behavior (Fig. 7 ), where the inter-brick displacements serve to stabilize the growth of incipient cracks, over crack extensions up to 1 mm, with crack-growth toughness rising to the order of 14 MPa·m ½ . This is what nacre-like structures can offer. The limited inter-brick displacements can lead to crack deflection and most importantly brick pullout and crack bridging, which act to stabilize crack growth; there is little change in crack-initiation toughness as the extrinsic toughening mechanisms solely enhance the crack-growth toughness 6 . This is also underlined by the microfracture experiments (Fig. 9 ), which clearly show that the interface strength is rather poor and most toughening arises from the microstructural alignment with respect to the crack path. A significant increase in initiation toughness is conceivable, but must involve strengthening of the interface properties. However, the Achilles’ heel of these structures is that the inter-brick displacements need to be within the mortar and not along the interface, as in the present high-toughness structure infiltrated at 1273 K. There are two problems with interface cracking. Whereas it still provides a means to alleviate high local stresses and confer good toughness (which is the basic concept underlying the toughness of nacre), weak interfaces invariably result in low strength, which one can readily appreciate by comparing the flexural strength of the 1273 K infiltrated structure to that infiltrated at 1153 K. Additionally, as noted above, the properties of metal-infiltrated nacre-like structures are predicted to be far superior to those of polymer-infiltrated structures because of the higher tensile/shear strengths of metal mortars; 10 however, for this to be realized, the inter-brick displacements naturally have to be within the mortar. For brick-and-mortar architectures, such behavior has been achieved in the high-toughness alumina–PMMA structures, where the ceramic–polymer interfaces were strengthened by grafting techniques 11 , and in coextruded (but very coarse-grained) alumina–nickel structures 24 , but to our knowledge, it has yet to be achieved with nacre-like, fine-grained, metal-infiltrated ceramic structures. One possible method to attain such materials is to select a BMG with higher ductility, such as a Pt-based BMG 53 , and freeze-cast a substrate the metal can readily wet to overcome the strength and toughness trade-off. Concerning the damage-tolerant behavior of nacre-like (brick-and-mortar) alumina ceramics infiltrated with metallic mortar of a Zr-based BMG, we therefore draw the following conclusions. Alumina/BMG-forming hybrid materials, with nacre-like structures containing a high alumina fraction up to 80 vol.%, were rapidly synthesized using freeze-casting followed by melt-infiltration. The excellent infiltration behavior, showing rapid and complete infiltration without any applied pressure, was attributed to the strong chemical interaction and excellent wetting behavior between the BMG melt and alumina. These results show that BMGs, or any alloys with fast and excellent wettability on ceramic, can be infiltrated in a high-volume-fraction ceramic preform with a complex architecture to synthesize advanced bioinspired structural materials with exceptional mechanical properties. Mechanical testing results indicated how infiltration temperature can have a marked effect on the interfacial properties of the resulting hybrid materials. The reactive wetting of BMG with alumina provided a strong interfacial bonding when infiltrated at a lower temperature (1153 K), but the chemical reaction at a higher temperature (1273 K) and crystallization of the BMG melt resulted in the formation of weak ceramic–metal interfaces. These effects of infiltration temperature resulted in a trade-off between the strength and toughness of the compliant-phase ceramic materials. In samples infiltrated at 1153 K, the high interfacial strength led to higher strength in the alumina/BMG-forming hybrids, but low toughness due to the brittle nature of the metallic matrix. The samples infiltrated at 1273 K exhibited high fracture toughness, with stable crack growth, and hence resistance-curve behavior, up to stress intensities of 9–14 MPa·m ½ (the latter value being not strictly valid by ASTM Standards 49 ), due to interphase displacements between the bricks. However, their strength was lower because these displacements (and ultimately failure) occurred along the interface, as opposed to within the mortar layer, such that nacre-like failure mechanisms were not perfectly mimicked."
} | 6,000 |
21684003 | null | s2 | 2,051 | {
"abstract": "A minimum in the biological response to materials that is observed to occur within a narrow surface energy range is related to the properties of water at these biology-contacting surfaces. Wetting energetics are calculated using a published theory from which it is further estimated that water molecules bind to these special surfaces through a single hydrogen bond, leaving three other hydrogen bonds to interact with proximal water molecules. It is concluded that, at this Goldilocks Surface, the local chemical environment of surface-bound water is nearly identical to that experienced in bulk water; neither deprived of hydrogen bond opportunities, as it is in contact with a more hydrophobic surface, nor excessively hydrogen bonded to a more hydrophilic surface. A minimum in the biological response occurs because water vicinal (near) to the Goldilocks Surface is not chemically different than bulk water. A more precise definition of the relative terms hydrophobic and hydrophilic for use in biomaterials becomes evident from calculations: >1.3 kJ/mole-of-surface-sites is expended in wetting a hydrophilic surface whereas <1.3 kJ/mole-of-surface-sites is expended in wetting hydrophobic surfaces; hydrophilic surfaces wet with >1 hydrogen bond per water molecule whereas hydrophobic surfaces wet with <1 hydrogen bond per water molecule."
} | 336 |
34712394 | PMC8515068 | pmc | 2,053 | {
"abstract": "Environmental pressure to reduce our reliance on agrochemicals and the necessity to increase crop production in a sustainable way have made the rhizosphere microbiome an untapped resource for combating challenges to agricultural sustainability. In recent years, substantial efforts to characterize the structural and functional diversity of rhizosphere microbiomes of the model plant Arabidopsis thaliana and various crops have demonstrated their importance for plant fitness. However, the plant benefiting mechanisms of the rhizosphere microbiome as a whole community rather than as an individual rhizobacterium have only been revealed in recent years. The underlying principle dominating the assembly of the rhizosphere microbiome remains to be elucidated, and we are still struggling to harness the rhizosphere microbiome for agricultural sustainability. In this review, we summarize the recent progress of the driving factors shaping the rhizosphere microbiome and provide community-level mechanistic insights into the benefits that the rhizosphere microbiome has for plant fitness. We then propose the functional compensatory principle underlying rhizosphere microbiome assembly. Finally, we suggest future research efforts to explore the rhizosphere microbiome for agricultural sustainability.",
"introduction": "1 Introduction Plants harbor diverse microbiome inhabitants, including bacteria, fungi, viruses, protists, and nematodes. These microorganisms are key components of the host plant and can colonize outside and inside of the plant tissue, referring to the rhizosphere (a narrow zone influenced by plant roots), phyllosphere (aboveground plant parts, particularly the leaves), anthosphere (a zone around the flowers, a subdivision of phyllosphere), spermosphere (a habitat surrounding the seeds where the soil, germinating seeds, and the microbial communities interact) and endosphere microbiomes (inside plant parts) [1] . The plant and its associated microbiomes are proposed to function as a holobiont, which is a consequence of evolutionary selection between plants and microorganisms [2] . Compared to the other plant compartments and the bulk soil, the rhizosphere, in which the microbial abundance, density and activity are largely increased, is considered the second genome of plants [3] . Therefore, the rhizosphere is a hotspot for plant-microbiome-soil interactions and serves as the gateway for plants to uptake nutrients and the first line of defense against different biotic and abiotic stresses [4] . Consequently, the rhizosphere microbiome can potentially be manipulated to increase crop yields and to reduce chemical fertilizer and pesticide inputs [5] . The beneficial roles of some plant growth-promoting rhizobacteria (PGPR) have been extensively studied over the past decades [6] . For example, some Bacillus spp. and Pseudomonas spp. members can not only stimulate induced systemic resistance (ISR) of the host plant and produce antibiotics to suppress pathogens, but also secrete secondary metabolites to promote growth or to enhance abiotic stress tolerance of the host plant [7] , [8] . However, these functions depend on the abiotic conditions and the biological interactions for these members to exert their plant beneficial properties in a rhizosphere community with high species diversity [9] , [10] . To achieve a more comprehensive understanding of how to make full use of the plant beneficial functions of rhizosphere microorganisms, first, it is important to characterize the principles that govern the assembly process and drive the diversity and composition of the rhizosphere microbiome. Next, the complicated interactions within the rhizosphere microbiome and between the rhizosphere microbiome and its host plant and their consequences on promoting plant growth and development, improving plant nutrient acquisition, increasing abiotic stress tolerance and enhancing pathogen suppression of the plants need to be identified. Furthermore, integrated strategies exploiting microbial functions and plant traits need to be implemented in the sustainable development of agriculture. In this review, we summarize the recent progress of the driving factors underlying the assembly of the rhizosphere microbiome. Next, we review mechanistic insights into the benefits of the rhizosphere microbiome on plant fitness. Finally, we discuss the new reductionist approaches for the development of synthetic microbial inoculants and the challenges for exploiting the beneficial outcomes of the rhizosphere microbiome."
} | 1,138 |
28377751 | PMC5359226 | pmc | 2,054 | {
"conclusion": "Conclusions and future outlook The most challenging hurdle of producing biofuels using “microbial factories” is to generate a large amount of fuel on a comparatively lower budget and greater efficiency as compared to the conventional fossil fuels. In other words, for replacing petrol with bioethanol, the latter should be cheaper, which could be a highly challenging task in terms of meeting the daily requirement (quantity). For example, in USA, approximately 19 million barrels of petrol is consumed per day; generating this amount on the industrial scale could be an arduous task. Therefore, to increase the acceptability of microbial biofuel, its productivity should be prioritized in the future."
} | 175 |
27172092 | null | s2 | 2,055 | {
"abstract": "The power of a single engineered organism is limited by its capacity for genetic modification. To circumvent the constraints of any singular microbe, a new frontier in synthetic biology is emerging: synthetic ecology, or the engineering of microbial consortia. Here we develop communication systems for such consortia in an effort to allow for complex social behavior across different members of a community. We posit that such communities will outpace monocultures in their ability to perform complicated tasks if communication among and between members of the community is well regulated. Quorum sensing was identified as the most promising candidate for precise control of engineered microbial ecosystems, due to its large diversity and established utility in synthetic biology. Through promoter and protein modification, we engineered two quorum sensing systems (rpa and tra) to add to the extensively used lux and las systems. By testing the cross-talk between all systems, we thoroughly characterized many new inducible systems for versatile control of engineered communities. Furthermore, we've identified several system pairs that exhibit useful types of orthogonality. Most notably, the tra and rpa systems were shown to have neither signal crosstalk nor promoter crosstalk for each other, making them completely orthogonal in operation. Overall, by characterizing the interactions between all four systems and their components, these circuits should lend themselves to higher-level genetic circuitry for use in microbial consortia."
} | 385 |
35880077 | PMC9307998 | pmc | 2,057 | {
"abstract": "In nature, bacteria form biofilms in very diverse environments, involving a range of specific properties and exhibiting competitive advantages for surface colonization. However, the underlying mechanisms are difficult to decipher. In particular, the contribution of cell flagellar motility to biofilm formation remains unclear. Here, we examined the ability of motile and nonmotile E. coli cells to form a biofilm in a well-controlled geometry, both in a simple situation involving a single-species biofilm and in the presence of co-colonizers. Using a millifluidic channel, we determined that motile cells have a clear disadvantage in forming a biofilm, exhibiting a long delay as compared to nonmotile cells. By monitoring biofilm development in real time, we observed that the decisive impact of flagellar motility on biofilm formation consists in the alteration of surface access time potentially highly dependent on the geometry of the environment to be colonized. We also report that the difference between motile and nonmotile cells in the ability to form a biofilm diminishes in the presence of co-colonizers, which could be due to motility inhibition through the consumption of key resources by the co-colonizers. We conclude that the impact of flagellar motility on surface colonization closely depends on the environment properties and the population features, suggesting a unifying vision of the role of cell motility in surface colonization and biofilm formation.",
"introduction": "Introduction Biofilms represent the preferred lifestyle for bacteria ( Flemming et al., 2016 ). In these three-dimensional structures, the aggregated cells prosper in a self-produced polymer extracellular matrix that protects them from shear stress, grazers and biocides ( Hall-Stoodley et al., 2004 ; Karygianni et al., 2020 ). The formation of a biofilm is a highly multifactorial process in which the cell properties and the details of the environment are both important, bringing about a tremendous diversity in behavior. In the planktonic state, most bacteria swim in a series of runs and tumbles and rotate their flagella assembled in bundles ( Berg and Anderson, 1973 ; Silverman and Simon, 1974 ; Nakamura and Minamino, 2019 ), which provides a significant adaptive advantage in nutrient search ( Colin et al., 2021 ). The importance of this cell motility to biofilm formation has been investigated in a large spectrum of bacterial species and environmental conditions, resulting in differing viewpoints. For example, early research determined that motility was crucial for biofilm development in Pseudomonas aeruginosa ( O’Toole and Kolter, 1998 ), Listeria monocytogenes ( Lemon et al., 2007 \n ) and Escherichia coli ( Pratt and Kolter, 1998 ; Wood et al., 2006 ). Furthermore, flagellar motility is often associated with increased virulence in pathogenic species, with motile bacteria exhibiting facilitated host colonization ( Josenhans and Suerbaum, 2002 ). Different functional mechanisms may actually be involved in motility effect to biofilm formation which do not necessarily engage flagella-assisted adhesion ( Haiko and Westerlund-Wikstrom, 2013 ). Motility per se can support aerotaxis, inducing interface accumulation which indirectly promotes biofilm formation ( O’Toole and Kolter, 1998 ; Suchanek et al., 2020 ). It has also been shown that elongated motile bacteria could accumulate near the surface due to hydrodynamic trapping. However this happens only at a reduced distance of the surface ( Frymier et al., 1995 ; Wood et al., 2006 ; Giacche et al., 2010 ). Besides, motility regulation overlaps with a complex signaling network that controls functions involved in biofilm formation such as quorum sensing or exopolysaccharide production ( Merritt et al., 2007 ; Shrout et al., 2011 ). It has therefore been difficult to determine both a clear causal relationship and the mechanisms that could support a hypothesis of motility as an advantageous feature for biofilm-forming bacteria. In E. coli , flagellar activity has been proposed to facilitate the initial contact of the cell with the surface, potentially helping to overcome repulsive forces on the surface ( Pratt and Kolter, 1998 ). Nevertheless, when other surface appendages such as Curli or conjugative pili are constitutively expressed, flagella become dispensable for the initial adhesion and biofilm development ( Prigent-Combaret et al., 2000 ; Reisner et al., 2003 ), suggesting motility per se might not be the adhesion promotion factor. On the other hand, flagella mechanosensory function as a surface-sensing tool has been proposed to govern the planktonic-sessile transition underlying biofilm formation ( Laganenka et al., 2020 ; Wong et al., 2021 ). However, in this case the surface detection by the flagella culminates in motility downregulation. Consistently, high concentrations of the second messenger cyclic diguanylate (c-di-GMP) have been shown to correlate with motility downregulation and the development of a thick biofilm ( Valentini and Filloux, 2016 ; Jenal et al., 2017 ). These results seem paradoxical with regard to the positive effect of motility on biofilm development, highlighting motility and biofilm development as mutually exclusive events. Nevertheless, bacteria may also swim within a mature biofilm ( Houry et al., 2012 ). Motility has also been suggested to influence biofilm maturation and architecture ( Wood et al., 2006 ; Barken et al., 2008 ), although reports about the impact of cell motility on later stages of biofilm development are scarce. In Vibrio cholerae , motility has been proposed to favor the invasion of resident biofilms ( Nadell et al., 2015 ). Ultimately, the motility effect on biofilm formation significantly changes depending on the environment, including surface properties and hydrodynamics ( Zheng et al., 2021 ). Therefore, despite its obvious competitive fitness advantage in planktonic life (particularly regarding nutrient pursuit), the question of whether cell motility is a superior trait in surface-colonizing competition remains open. To address this issue, we examined biofilm formation by motile and nonmotile E. coli cells in the controlled geometry of a millifluidic device from a kinetic perspective, covering both the short and long time scales of the adherent community development. This allowed us to search for mechanistic information that could distinguish the colonizing ability of swimming cells from their nonmotile counterparts. In this study, we take motility to mean flagellar motility apart from surface-associated motions such as swarming or twitching ( Wadhwa and Berg, 2021 ). We thus examined the simple situation of motile vs. nonmotile E. coli strains colonizing a bare glass surface, followed by a more complex and more naturally relevant situation involving the same strains in the presence of other species in a co-colonization test. The latter included a 4-species assemblage that we previously showed to form a deterministic community in about 40 hours of growth under continuous nutrient flow in a millifluidic channel ( Monmeyran et al., 2021 ). The use of this continuous flow growth mode makes it possible to control the physicochemical properties of the environment throughout the development of the biofilm, and ensures that the biofilms can be thoroughly compared. Our results reveal a clear-cut effect of motility on surface colonization, which consists in introducing a lag time of several hours to the biofilm development. Meanwhile, the motile and nonmotile cells display a similar biofilm growth rate. We interpret this effect in terms of a spatial exploration discrepancy. Finally, we reveal how the presence of co-colonizers affects this behavior and discuss the strong dependence of flagellar motility effects on specific environmental conditions.",
"discussion": "Discussion Bacterial flagellar motility has been investigated for decades now, and significant advances in the understanding of the involved mechanisms have been made. However, many open questions remain about the impact of this function on the ability of bacteria to colonize surfaces. Here, we report a kinetic analysis of E. coli surface colonization in a millifluidic channel, with a focus on active vs. non-active flagellar motility. Our results show that motility introduces a significant delay to the colonization of bare surfaces by E. coli alone. This is in contrast to the intuitive concept encountered in the literature that the absence of motility reduces the chances of bacterial cells coming into contact with the surface ( Pratt and Kolter, 1998 ; Zheng et al., 2021 ). Based on mathematical modeling of the settling and diffusion of the bacteria in the channel, we found that the nonmotile cells landed on surface following a simple settling law. In contrast, random diffusion did not account for motile cell on surface over time. In our experimental configuration, the ‘race’ towards the surface is won by nonmotile cells which is not predicted by the calculations. To explain this discrepancy, we suggest that motile cells may repeatedly bounce on the surface before attaching which would slow down the dwelling kinetic. Besides, taking into account the asymmetry of the experimental channel, we make the hypothesis that chemical gradients may also bias bacteria swimming and delay surface colonization. The channel geometry that we used is relevant for many natural situations in which bacteria dwell in channels and pores in the millimetric range under continuous or intermittent flow with irregular nutrient spatial distribution and chemical gradients. This highlights the importance of the geometry of the environment, which has generally been overlooked in colonization assays, in the competition between motile and nonmotile cells. We observed that local dynamics and biomass growth rates were very similar for both motile and nonmotile cells, suggesting that flagellar motility does not alter anchorage or surface persistence once the cells have reached the surface, an issue that has remained controversial to date. As detailed in previous investigations of this question ( Pratt and Kolter, 1998 ), it is difficult to decipher the primary factors involved in the initial attachment. A particular challenge is to disentangle the contribution of motility per se from the contribution of flagella as a surface appendage and the contingent involvement of other structures such as type I pili. Under the conditions used in this study for bare surface colonization, it is likely that the attachment step is dominated by the overexpression of the F-pilus, which is crucial in promoting the initial adhesion ( Ghigo, 2001 ; Reisner et al., 2003 ; Beloin et al., 2008 ). In this case, flagellar motility simply reduces surface abundance, which is established during the inoculation period, and does not affect biofilm development. This is an interesting finding to take into account for multispecies colonization processes where surface access kinetics are crucial in the competitive dynamics that shape the attached community ( Eigentler et al., 2022 ). In the presence of co-colonizers (here, the members of a 4-species community able to build a stable biofilm), we observed that motile and nonmotile E. coli cells exhibit very similar colonization profiles that differ from the profiles displayed in the single-species colonization experiments. Specifically, there is a lower surface-bound E. coli global biomass, consistent with the intrinsic competition for the surface expected from the presence of other adhesive species ( Lloyd and Allen, 2015 ). Two striking features stand out here: the emergence of E. coli kinetic colonization phases that match the four-species biofilm climaxes; and the reduced lag observed between motile and nonmotile cells in the characteristic time of colonization, primarily due to the receding of the nonmotile cells in comparison to the E. coli colonizing ability in the absence of co-colonizers. In a previous study, the four-species biofilm climaxes were interpreted as species responses to the oxygen depletion induced by biofilm development, suggesting that the reduction in oxygen might also contribute to the diminished colonizing efficiency of E. coli ( \n Monmeyran et al., 2021 ). Moreover, knowing that the lack of oxygen strongly affects E. coli motility ( Douarche et al., 2009 ) by inducing a motile to nonmotile transition in the bacterial population, we can hypothesize that the environmental oxygen scarcity caused by the co-colonizers accounts for the convergence of the motile and nonmotile cell colonization kinetics in this multispecies context. Nevertheless, the co-colonizers could also induce a shift in the limiting step of the colonizing process by increasing the characteristic time of attachment on the surface, which would result in abolishing the difference between motile and nonmotile cells that ultimately dwell on the surface. These results stress the importance of studying the processes from a kinetic perspective in order to acquire mechanistic information. Our report establishes that the impact of flagellar motility on surface colonization is not necessarily an intrinsic trait associated with this function; instead, it closely depends on an environment defined by both topology and population composition. Importantly, we demonstrate that cell swimming can regulate the surface access time. However, a shift in the environmental conditions (such as the presence of co-colonizers) can drastically alter the outcome of this distinctive property and abolish the asymmetry between motile and nonmotile cells. We thus propose here a study with the potential to enlighten the long-standing controversy over the role of cell motility in surface colonization and biofilm formation."
} | 3,462 |
40053579 | PMC11887813 | pmc | 2,058 | {
"abstract": "Marine sediments are highly bioactive habitats, where sulfate-reducing bacteria contribute substantially to seabed carbon cycling by oxidizing ~77 Tmol C org year −1 . This remarkable activity is largely attributable to the deltaproteobacterial family Desulfobacteraceae of complete oxidizers (to CO 2 ), which our biogeography focused meta-analysis verified as cosmopolitan. However, the catabolic/regulatory networks underlying this ecophysiological feat at the thermodynamic limit are essentially unknown. Integrating cultivation-based (80 conditions) proteogenomics of six representative Desulfobacteraceae spp., we identify molecular commonalities explaining the family’s environmental relevance and success. Desulfobacteraceae genomes are specifically enriched in substrate uptake, degradation capacities, and regulatory functions including fine-tuned sulfate uptake. Conserved gene arrangements and shared regulatory patterns translate into strikingly similar (sub-)proteome profiles. From 319 proteins, we constructed a meta-network for catabolizing 35 substrates. Therefrom, we defined a Desulfobacteraceae characteristic gene subset, which we found prevalent in metagenomes of organic-rich, marine sediments. These genes are promising targets to advance our mechanistic understanding of Desulfobacteraceae -driven biogeochemical processes in marine sediments and beyond.",
"introduction": "INTRODUCTION Sulfate-reducing bacteria (SRB), such as Desulfobacteraceae , couple the oxidation of organic carbon to the reduction of sulfate to sulfide (dissimilatory sulfate reduction), thereby linking the carbon and sulfur cycles. This process is particularly important in global marine environments due to very high sulfate concentrations in the oceans [for overview, see ( 1 , 2 )]. Here, continental margins, coastal ranges, and shelf sediments stand out by their high input of organic matter, and more than 50% of their mineralization is achieved in the upper sediment layers, coupled to sulfate reduction ( 3 , 4 ). Furthermore, organic carbon richness of upwelling regions generates oxygen minimum zones in the waterbody where SRB are involved in carbon turnover and a cryptic sulfur cycle ( 5 ). Globally, of the total carbon flux reaching the ocean floor, 12 to 29% are oxidized via sulfate reduction, as estimated from steady-state sulfate profiles ( 6 ). However, considering sulfate-reduction rates calculated from 35 S radiotracer measurements as well as re-oxidation and cryptic sulfur cycle, this share is markedly higher, accounting for an estimated 77 Tmol C org year −1 ( 7 ). These high mineralization rates may only be achieved by SRB capable of complete oxidation to CO 2 ( 8 , 9 ). However, the well-studied family Desulfovibrionaceae cannot achieve this turnover because these organisms only incompletely oxidize organic substrates to acetate. The discovery of the family Desulfobacteraceae , encompassing completely oxidizing SRB ( 10 – 12 ), solved this biogeochemical paradox. Members of this family use a large variety of organic substrates, ranging from small molecules (e.g., fermentation end products) to long-chain fatty acids and aromatic compounds, including recalcitrant hydrocarbon compounds, such as n -alkanes, alkylbenzenes, and -phenols ( 1 , 13 , 14 ). The degradation routes of these different organic carbon substrates ultimately converge at the level of acetyl–coenzyme A (CoA) that is oxidized to CO 2 via the Wood-Ljungdahl pathway (WLP) in most cases ( 1 ). In agreement with their proposed environmental role, Desulfobacteraceae members, particularly those of the Desulfosarcina / Desulfococcus cluster, were shown to dominate the SRB communities in marine shelf sediments (e.g., 15 – 17 ). Generally, SRB thrive at the thermodynamic limit of life, due to the very low redox potential of the sulfate/sulfide redox pair (−228 mV), allowing the generation of only ~10% of the energy obtained as compared to aerobic heterotrophs applying the oxygen/water pair (+818 mV) ( 18 ). To cope with this challenge, SRB evolved a number of specialized enzymes/complexes that harness biochemically intriguing mechanisms to facilitate endergonic reactions without spending adenosine 5′-triphosphate (ATP), e.g., reduction of sulfite to sulfide via a DsrC trisulfide ( 19 ), an electrogenic redox loop involving QrcABCD in sulfate reduction ( 20 ), or ATP-independent reductive dearomatization applying electron bifurcation ( 21 ). However, this research focus on single remarkable enzymatic mechanisms needs to be complemented by global analyses of the involved catabolic networks and their regulatory modulations to explain the complexity involved in the environmental success and functions of SRB, both in general and for the Desulfobacteraceae in particular. To achieve a holistic understanding of the role of Desulfobacteraceae in the marine carbon/sulfur cycles, we first conducted a meta-analysis (literature based) to capture globally the biogeography of this family. Then, we conducted comparative proteogenomic analyses across six members of the Desulfobacteraceae , which we selected to cover the versatility, phylogeny, and lifestyles of this family. Among them, Desulfosarcina variabilis 3be13 is a particularly versatile, cultivated representative of the Desulfosarcina/Desulfococcus cluster, which is why we sequenced its genome and determined the proteomes of 29 substrate conditions. Overall, a total of 80 substrate conditions across the six studied strains allowed us (i) to gain unprecedented insights into commonalities of consistently formed (constitutive) versus substrate-specifically formed (regulated) subsets of the proteome (subproteomes), the formed transportome (entirety of transmembrane transport systems), and the regulation of sulfate uptake and (ii) to construct a catabolic meta-network (synthesis of all reactions/proteins involved in substrate degradation and respiratory energy conservation). Last, genes encoding various enzymes of key pathways in the meta-network were recognized as widespread across 43 Desulfobacteraceae reference genomes and also detectable in available metagenomes from geographically far apart and geochemically distinct marine sediments.",
"discussion": "DISCUSSION Our proteogenomic analysis was devised to understand the catabolic basis of the long-appreciated role of SRB and Desulfobacteraceae , particularly in the interwoven carbon and sulfur cycles of marine sediments. The five genera selected for our comparative approach are indeed cosmopolitan thriving in diverse marine habitats, underscoring the transferability of the present findings to the family of Desulfobacteraceae in general. We identified two family-defining metabolic key properties: (i) the pathway-module inherent coupling of substrate-degradation via electron transfer proteins (tailored to individual oxidation reactions) to membrane-embedded redox complexes and dissimilatory sulfate reduction; (ii) a common catabolic network architecture, where multiple substrate specifically regulated pathway modules (peripheral degradation routes) feed into few constitutively formed central modules of degradation and energy metabolism. Both properties together contribute to an energy efficient exploitation of diverse substrates, ultimately enabling life at the thermodynamic limit and fostering environmental success. The constructed meta-network reveals a broad range of shared and strain-specific degradation capacities among the six studied strains. This may even be underestimated given the breadth of the transporter repertoire, which greatly exceeds the number of tested substrates (plus required nutrients) and may therefore provide access to hitherto unknown substrates/pathways. Furthermore, the inclusion of Desulfobacteraceae strains with not yet considered physiologies, e.g., n -alkane ( 54 ) and polymer degraders ( 55 ), will expand the catabolic diversity of our meta-network even more. Hence, the high sulfate reduction rates in organic-rich, marine sediments do not rely on individual key species, but rather result from the additive degradation capacities of site-specific SRB communities (communal accomplishment). Yet, the Desulfosarcina and Desulfococcus species currently stand out by exceptionally numerous degradation modules, both reflecting their known broad substrate-spectrum and rationalizing their dominance in SRB communities of marine sediments. The prevalence and conservation of the degradation modules across the 43 representative Desulfobacteraceae genomes hints at their niche-defining role. We speculate that the wealth of these modules was achieved by comprehensive horizontal gene transfer implicated by the genomes’ richness in mobile elements. Overall, these congeneric yet tailored genomic, regulatory, and catabolic capacities of Desulfobacteraceae shape their environmental function and success, ultimately imprinted in their global biogeography. In the light of the burgeoning large-scale metagenomic studies [particularly metagenome-assembled genomes (MAGs) [e.g., ( 56 )], the here presented metabolism-centered insights into the Desulfobacteraceae provide a “treasure-trove” from which to select target genes for functional analysis of SRB communities in natural and technical environments. First, it can complement incubation experiments with labeled substrates that target the active part of the community [e.g., ( 57 )]. Second, given their enormous carbon turnover in the seabed, Desulfobacteraceae should substantially shape the dissolved organic matter (DOM) ( 58 ), e.g., by depleting a broad range of organic substrates while enriching the recalcitrant fraction. Such microbial activities in the sediment should also feedback on DOM in the water column ( 59 ), particularly in shelf seas, when considering vertical exchange processes across the bottom boundary layer ( 60 ). Third, symbiotic/commensal relations of marine SRB with, e.g., marine oligochaete worms ( 61 ), benthic foraminifera ( 62 ), or seagrass/salt marsh cordgrass ( 63 ) and non-marine SRB with, e.g., the human gut [e.g., ( 64 )], can be studied on a functional level. Fourth, the mechanistic understanding of deleterious effects of SRB on technical installations can be improved, e.g., production and processing facilities of the gas and oil industry [e.g., ( 65 )], as well as sites of metal and concrete corrosion [e.g., ( 66 , 67 )]."
} | 2,613 |
18415009 | PMC2441489 | pmc | 2,059 | {
"abstract": "The brain processes underlying cognitive tasks must be very robust. Disruptions such as the destruction of large numbers of neurons, or the impact of alcohol and lack of sleep do not have negative effects except when they occur in an extreme form. This robustness implies that the parameters determining the functioning of networks of individual neurons must have large ranges or there must exist stabilizing mechanisms that keep the functioning of a network within narrow bounds. The simulation of a minimal neuronal architecture necessary to study cognitive tasks is described, which consists of a loop of three cell-assemblies. A crucial factor in this architecture is the critical threshold of a cell-assembly. When activated at a level above the critical threshold, the activation in a cell-assembly is subject to autonomous growth, which leads to an oscillation in the loop. When activated below the critical threshold, excitation gradually extinguishes. In order to circumvent the large parameter space of spiking neurons, a rate-dependent model of neuronal firing was chosen. The resulting parameter space of 12 parameters was explored by means of a genetic algorithm. The ranges of the parameters for which the architecture produced the required oscillations and extinctions, turned out to be relatively narrow. These ranges remained narrow when a stabilizing mechanism, controlling the total amount of activation, was introduced. The architecture thus shows chaotic behaviour. Given the overall stability of the operation of the brain, it can be concluded that there must exist other mechanisms that make the network robust. Three candidate mechanisms are discussed: synaptic scaling, synaptic homeostasis, and the synchronization of neural spikes.",
"introduction": "Introduction Already in 1893 the Italian psychiatrist Tanzi ( 1893 ) postulated that the formation of memories was carried by the growth or strengthening of interactions in the brain (D’Anguilli and Dalenoort 1996 ), an idea that some fifty years later was stated more explicitly by Hebb ( 1949 ). He formulated the rule that the efficiency of a synapse is increased when a pair of neurons involved are simultaneously active. Although the well-known learning rule that originated from the ideas of Tanzi and Hebb, has been extensively studied over the last fifty years, there have been relatively few studies directed at the consequence of the rule: that cell-assemblies are the carriers of our memory traces. Hebb saw this as one of his major contributions to our understanding of cognitive brain functioning. The question of the robustness of cell-assemblies was not raised before Milner ( 1957 ), who introduced inhibitory interactions in the simulations of cell-assemblies in order to make them more stable. Since then studies on cell-assemblies have remained relatively scarce. The issue of robustness was studied in Hopfield networks ( 1982 ) by means of analytical methods. For such relatively simple networks, with simple models of neurons, it is possible to draw conclusions from analytical studies (Gerstner and Kistler 2002 ). Provided the model of the network is not too complex, the equations describing the dynamics of such networks can be approximately analysed, on the basis of arguments that are mainly heuristic. These equations can only be analysed for networks that are of infinite size, that have some properties of symmetries, or that consist of simplified neurons, for example such that they all have the same threshold, and the same numbers of interconnections. Moreover, the system must be random to allow statistical arguments. The equations can then be handled in a statistical fashion, in manners known from statistical physics. Unfortunately, or perhaps fortunately, these statistical analyses cannot be used for the study of the properties of networks that are to serve as substratum for cognitive tasks that have some relevance for the study of cognition, such as doing simple arithmetic, or producing and understanding language. These networks can in principle not be analysed fully in a statistical manner. Even a basic requirement, the implementation of binding—a topic to be discussed later on—seems to be impossible in an analytical model. For these inhomogeneous and non-uniform networks, only the process itself can be simulated. This is in contrast to the numerical analysis of a simple network for which the equations can be so far approximated that the equations can be evaluated by numerical techniques for different cases of parameter values, and for different types of networks. For the architecture and dynamics of neural networks that can serve as the substratum of specific cognitive tasks, only simulations are possible of the network itself. (Dalenoort, personal communication). The last two decades have shown an increase in the interest of cell-assemblies (Dalenoort 1985 ; Pulvermüller 1996 , 1999 ; Huyck 2004 ) and also in analytical studies, where they are represented in terms of attractors (Amit 1995 ; Amit and Mongillo 2003 ). As we argued above an analytical representation in terms of attractors is not suitable to answer questions about the specific network structure of cell-assemblies necessary from a cognitive point of view (Dalenoort and de Vries 1995 ). The approach of this paper is that cognitive requirements expressed in terms of the functional notion of memory traces are used to design simulations at the neural level. On the basis of these simulation studies new phenomena in the model can be distinguished and compared with what is known neurophysiologically. It is possible that this will lead to the discovery of actual new phenomena. In earlier work (de Vries 1995 ), we referred to this approach as ‘downward emergence’. Quintessential to the approach is that a strict bottom-up study of brain functioning will not be sufficient. Cognition—the top-down approach—has to be taken into account as well (Dalenoort 1990 ; Dalenoort and de Vries 1998a ).",
"discussion": "Discussion We have explored the parameter space of a minimal model of cognitive brain functioning. The results of the simulation experiments indicate that the desired behaviour of the model occurs in several subspaces of the parameter space. In every subspace, however, two typical phenomena can be observed. On the one hand the behaviour strongly depends on the selection of specific parameter values from a single narrow range, like the mean excitatory threshold, the mean strength of excitatory connections internal to a cell-assembly, and the decay. On the other hand, value ranges of parameters can be quite fragmented. For a single parameter there exist value ranges that produce adequate behaviour, arbitrarily mixed with ranges exhibiting unexpected effects. These findings indicate that the key to the solution of the robustness question is not to be found in the parameter space of the brain. In the following we will pursue some alternative answers. The robustness/flexibility dilemma Suppose that we would have found the large parameter ranges necessary for the robust behaviour of the minimal architecture. This would then raise the question of how the architecture could adapt itself to new changes in its environment or could develop new cognitive structures, e.g. corresponding to creative thought. A necessary condition for these things to happen is that the brain is capable of producing a sufficient variation of excitation patterns. The observed deviations in the autonomous-growth and extinction conditions may be part of this variation. Stabilizing mechanisms Even if we take into account that the unexpected behaviours observed in simulation experiments may have a function, the role of stabilizing mechanisms—other than the discussed arousal control system—is not excluded. We will review three candidate mechanisms, of which the first two are discussed in Turrigiano ( 1999 ) whereas the last one is a hypothesis of the authors. The three mechanisms have to be distinguished from learning mechanisms because they do not depend on any external events playing a role in learning. \n Synaptic scaling is the regulation by cortical and hippocampal neurons of their own firing rates by scaling their synaptic inputs up or down as a function of activity. This mechanism operates relatively slowly: requiring hours or days of altered activity to modify synaptic strengths. As a solution to the robustness problem formulated this paper, it may therefore be insufficient. A mechanism that keeps the architecture within its bounds should also be able of an instantaneous reaction since sudden changes in parameter values should not disrupt cognitive functioning. \n Synaptic homeostasis refers to the capability of a neuron to maintain relative constant firing properties although it is subject to many changes: growth, changes in shape, loss and gain of synapses, and the constant turnover of the ion-channels that determine its electrical-firing properties. The underlying mechanism probably makes use of the intracellular concentration of certain ions. A change in this concentration triggers a compensatory reaction that modifies ionic conductance such that the level of neuronal activity remains constant. Accordingly, distortions—like the failures occurring in the described simulation experiments—are immediately repaired. However, the effects of the mechanism of synaptic homeostasis have only been found in neuro-muscular synapses. \n Synchrony of firing implies that the spikes produced by two or more neurons are in phase. This phenomenon is relevant to the issue of robustness because phase synchrony may be a condition for the propagation of neural excitation: two neurons will only activate a third one if their spikes are in phase. Such a propagation may become robust if it takes the form of a loop, in which the spikes are interlocking. This synchronous firing of neurons could then trigger a biochemical process that compensates for changes in parameters of neural functioning. If robustness is based on synchronous firing, one would expect specific spike patterns on the presentation of a stimulus, although not every stimulus needs to have a unique pattern (see the discussion on the identity of a memory trace in the second section of this paper). In addition, repeated presentations of the same stimulus should reproduce the same spike patterns. Data compatible with this hypothesis have been reported by Fellous et al. ( 2004 ). Accordingly, synchronization could also play a role in the robustness of permanent memory structures next to its role in the temporal coupling of neuronal activity (‘binding’), a hypothesis fundamental to the already cited work on synfire chains and to many neurophysiological studies such as Singer et al. ( 1994 ), Roelfsema et al. ( 1997 ), and Freiwald et al. ( 2001 ). The list of stabilizing mechanisms presented here, is not meant to be exhaustive. Moreover, the three mechanisms on the list are not mutually exclusive and each of them requires further study. Their discussion is the offspring from the exploration of the parameter space of a minimal architecture, by means of which we have tried to lay down some important questions on cognitive brain functioning."
} | 2,808 |
34991714 | PMC8740439 | pmc | 2,062 | {
"abstract": "Background Soil microbial communities are major drivers of cycling of soil nutrients that sustain plant growth and productivity. Yet, a holistic understanding of the impact of land-use intensification on the soil microbiome is still poorly understood. Here, we used a field experiment to investigate the long-term consequences of changes in land-use intensity based on cropping frequency (continuous cropping, alternating cropping with a temporary grassland, perennial grassland) on bacterial, protist and fungal communities as well as on their co-occurrence networks. Results We showed that land use has a major impact on the structure and composition of bacterial, protist and fungal communities. Grassland and arable cropping differed markedly with many taxa differentiating between both land use types. The smallest differences in the microbiome were observed between temporary grassland and continuous cropping, which suggests lasting effects of the cropping system preceding the temporary grasslands. Land-use intensity also affected the bacterial co-occurrence networks with increased complexity in the perennial grassland comparing to the other land-use systems. Similarly, co-occurrence networks within microbial groups showed a higher connectivity in the perennial grasslands. Protists, particularly Rhizaria, dominated in soil microbial associations, as they showed a higher number of connections than bacteria and fungi in all land uses. Conclusions Our findings provide evidence of legacy effects of prior land use on the composition of the soil microbiome. Whatever the land use, network analyses highlighted the importance of protists as a key element of the soil microbiome that should be considered in future work. Altogether, this work provides a holistic perspective of the differential responses of various microbial groups and of their associations to agricultural intensification. Supplementary Information The online version contains supplementary material available at 10.1186/s40793-021-00396-9.",
"conclusion": "Conclusions Using a holistic microbiome investigation of bacterial, fungal and protist communities in a long-term field experiment managed under different levels of land use intensity, we showed that land management only affected α -diversity of the bacterial community, with increased diversity in the continuous cropping system. However, we identified a clear shift in the structure and the composition of all communities in response to land use, in particular between the continuous cropping and the perennial grassland. Moreover, our results showed legacy effects of cropping on the structure of the soil microbiome that lasted after three years of temporary grasslands. This highlights that prior land use can shape the present-day community for multiple microbial groups across domains. The perennial grassland system led to more complex bacterial as well as inter-domain networks, which can have implication for the contribution of microbes to ecosystem multifunctionality [ 16 ]. Inter-domain networks also revealed the predominant role of the protist as key taxa in soil microbiome networks across all land-use types. Future work need to validate the importance of protists in shaping soil microbial communities, directly through biotic interactions and/or indirectly through changes in abiotic factors."
} | 833 |
36221148 | PMC9555204 | pmc | 2,063 | {
"abstract": "Background The dramatic increase in greenhouse gas (GHG) emissions, which causes serious global environmental issues and severe climate changes, has become a global problem of concern in recent decades. Currently, native and/or non-native C1-utilizing microbes have been modified to be able to effectively convert C1-gases (biogas, natural gas, and CO 2 ) into isobutanol via biological routes. Even though the current experimental results are satisfactory in lab-scale research, the techno-economic feasibility of C1 gas-derived isobutanol production at the industrial scale still needs to be analyzed and evaluated, which will be essential for the future industrialization of C1-gas bioconversion. Therefore, techno-economic analyses were conducted in this study with comparisons of capital cost (CAPEX), operating cost (OPEX), and minimum isobutanol selling price (MISP) derived from biogas (scenario #1), natural gas (scenario #2), and CO 2 (scenario #3) with systematic economic assessment. Results By calculating capital investments and necessary expenses, the highest CAPEX ($317 MM) and OPEX ($67 MM) were projected in scenario #1 and scenario #2, respectively. Because of the lower CAPEX and OPEX from scenario #3, the results revealed that bioconversion of CO 2 into isobutanol temporally exhibited the best economic performance with an MISP of $1.38/kg isobutanol. Furthermore, a single sensitivity analysis with nine different parameters was carried out for the production of CO 2 -derived isobutanol. The annual plant capacity, gas utilization rate, and substrate cost are the three most important economic-driving forces on the MISP of CO 2 -derived isobutanol. Finally, a multiple-point sensitivity analysis considering all five parameters simultaneously was performed using ideal targets, which presented the lowest MISP of $0.99/kg in a long-term case study. Conclusions This study provides a comprehensive assessment of the bioconversion of C1-gases into isobutanol in terms of the bioprocess design, mass/energy calculation, capital investment, operating expense, sensitivity analysis, and minimum selling price. Compared with isobutanol derived from biogas and natural gas, the CO 2 -based isobutanol showed better economic feasibility. A market competitive isobutanol derived from CO 2 is predicable with lower CO 2 cost, better isobutanol titer, and higher annual capacity. This study will help researchers and decision-makers explore innovative and effective approaches to neutralizing GHGs and focus on key economic-driving forces to improve techno-economic performance.",
"conclusion": "Conclusions and prospective In this study, TEA was applied to calculate the OPEX and CPAEX of the proposed biological routes to evaluate the economic feasibility and industrialization potential of the bioconversion of C1-gaseous substrates for isobutanol production. With the lowest OPEX and CAPEX, the CO 2 -derived isobutanol presents the lowest MISP of $1.38/kg isobutanol compared with that derived from biogas and natural gas. By employing single/multiple-point sensitivity analyses, the annual plant capacity, gas utilization rate, and CO 2 price are determined to be key cost drivers. With the expected research targets, the promising MISP of $0.99/kg isobutanol can be achieved by reducing CO 2 cost and enhancing the production performance of isobutanol production. Currently, biogas, natural gas, and CO 2 are the main sources of GHG emissions, of which the amounts are projected to further increase in the coming decades. By adopting the proposed biological routes, the C1-gases, including biogas, natural gas, and CO 2, can be used as an alternative inedible substrate for isobutanol production, which is consistent with the best interest of global environmental and economic sustainability. It is expected that this study may provide engineering practice guidance and cost optimization strategies for the future biological conversion of C1 greenhouse gases to platform chemicals."
} | 998 |
36605606 | PMC9642934 | pmc | 2,068 | {
"abstract": "Biofuel cells (BFCs) are an environmental friendly technology that can simultaneously perform wastewater treatment and generate electricity. Peculiarities that hinder the widespread introduction of this technology are the need to use artificial aeration and chemical catalysts, which make the technology expensive and cause secondary pollution. A possible solution to this issue is the use of biocathodes with microalgae and cyanobacteria. Microalgae in the biocathodic chamber produce oxygen as the terminal electron acceptor. Various BFC technologies with algal biocathode (microbial fuel cells, microbial desalination cells, and plant microbial fuel cells) can address a variety of issues such as wastewater treatment, desalination, and CO 2 capture. The main technological parameters that influence the performance of the biocathode are light, pH, and temperature. These technological parameters affect photosynthetic production of oxygen and organic compounds by microalgae or cyanobacteria, and hence affect the efficiency of electricity production, wastewater treatment and production of added-value compounds in microalgae biomass like lutein, violaxanthin, astaxanthin. The ability to remove carbon, nitrogen, and phosphorus compounds; antibiotics; and heavy metals by pure cultures of microalgae and cyanobacteria and by mixed cultures with bacteria in the cathode chamber can be used for wastewater treatment.",
"conclusion": "Conclusions Microalgae and cyanobacteria as pure cultures and in mixed cultures with bacteria can be used in biocathode to remove carbon compounds, nitrogen compounds, phosphorus compounds, antibiotics, and heavy metals. Morover, biocathodes with microalgae can be used for desalination of water, electricity generation, and wastewater treatment after passing through the anode chamber along with removal of carbon dioxide. The development of BFCs with microalgae and cyanobacteria as biological agents of biocathodes could enable energyefficient wastewater treatment and production of addedvalue compounds in microalgae biomass. Furthermore, the obtained biomass of microalgae can be used as a substrate at the anode of BFC or to produce added-value products such lutein, violaxanthin, astaxanthin, and cantaxanthin and biodiesel, which could confirm that BFC is economically viable. Modification of electrodes by polymers can improve the performance of BFC and enhance the adhesion of bacteria on the anode or cathode. Further studies on the synergistic effect of light (photoperiod and illuminance), pH, temperature, modification of electrodes, etc. are necessary to determine the optimal technological parameters of the full BFC. It should be noted that estimation their effect on technology (in order to increase power density or carbon capture and nitrogen compounds removal) will depend on the target product, namely the accumulation of biomass, oxygen production.",
"introduction": "Introduction Water and energy deficiency is currently a very serious global issue (Ashwaniy and Perumalsamy, 2017 ). Growing demand for fossil fuel energy may intensify global warming (Zhang et al., 2019 ). Therefore, search for alternative sustainable technologies of energy production and wastewater treatment, especially by energydependent countries, is a necessary and urgent task today (Kuzminskiy and Shchurska, 2018 ). Biofuel cells (BFCs) are an environmental friendly and promising technology that may facilitate to resolve the abovementioned issues (Kokabian and Gude, 2015 ), because they can be applied in both wastewater treatment and electricity generation or to obtain energy carriers (Kuzminskiy and Shchurska, 2018 ). The major reasons that prevent the introduction of this technology on an industrial scale is the use of expensive catalysts such as platinum and toxic chemical agents such as ferricyanide (Gude, 2016 ), which can be overcome by using microalgae as biocathodes. Microalgae and cyanobacteria can perform aeration at the cathode, which is advantageous for reducing the aeration cost (Huang et al., 2011 ; Arun et al., 2020 ). The purpose of this review was to investigate the possibility of using microalgae as a biological agent for biocathodes in microbial fuel cells. Microalgae in the biocathodic chamber produce oxygen as the terminal electron acceptor. Electrons pass through an external electrical circuit from the anode to the cathode (Mohan et al., 2014 ). Electrons transmitted to the anode are released through the bioelectrochemical decomposition of organic compounds from wastewater by exoelectrogenic bacteria. Microalgae- and cyanobacteria-based BFCs can address various issues such as desalination, wastewater treatment, bioremediation, bioenergy production, CO 2 capture, and synthesis of high added-value products (Saratale et al., 2017 ; Enamala et al., 2020 )."
} | 1,203 |
27616809 | null | s2 | 2,070 | {
"abstract": "Division of labor in insect societies relies on simple behavioral rules, whereby individual colony members respond to dynamic signals indicating the need for certain tasks to be performed. This in turn gives rise to colony-level phenotypes. However, empirical studies quantifying colony-level signal-response dynamics are lacking. Here, we make use of the unusual biology and experimental amenability of the queenless clonal raider ant "
} | 109 |
38605161 | PMC11009343 | pmc | 2,071 | {
"abstract": "A worldwide increase in the prevalence of coral diseases and mortality has been linked to ocean warming due to changes in coral-associated bacterial communities, pathogen virulence, and immune system function. In the Mediterranean basin, the worrying upward temperature trend has already caused recurrent mass mortality events in recent decades. To evaluate how elevated seawater temperatures affect the immune response of a thermophilic coral species, colonies of Astroides calycularis were exposed to environmental (23 °C) or elevated (28 °C) temperatures, and subsequently challenged with bacterial lipopolysaccharides (LPS). Using immunolabeling with specific antibodies, we detected the production of Toll-like receptor 4 (TLR4) and nuclear factor kappa B (NF-kB), molecules involved in coral immune responses, and heat shock protein 70 (HSP70) activity, involved in general responses to thermal stress. A histological approach allowed us to characterize the tissue sites of activation (epithelium and/or gastroderm) under different experimental conditions. The activity patterns of the examined markers after 6 h of LPS stimulation revealed an up-modulation at environmental temperature. Under warmer conditions plus LPS-challenge, TLR4-NF-kB activation was almost completely suppressed, while constituent elevated values were recorded under thermal stress only. An HSP70 up-regulation appeared in both treatments at elevated temperature, with a significantly higher activation in LPS-challenge colonies. Such an approach is useful for further understanding the molecular pathogen-defense mechanisms in corals in order to disentangle the complex interactive effects on the health of these ecologically relevant organisms related to global climate change.",
"introduction": "Introduction Direct and indirect impacts of global warming are proving detrimental effects to the health of marine species and are the primary cause of coral death worldwide 1 – 3 , as well as affecting the Mediterranean 4 , 5 . The coral holobiont is an obligate association between the coral animal and a plethora of mutualists (i.e., an appropriate commensal microbiota), including dinoflagellate endosymbionts of zooxanthellate species 6 ; however, this association is delicately balanced, and coral species worldwide are increasingly threatened by warming seawater and temperature-driven disease outbreaks 1 , 7 – 9 . Indeed, the risk of climate anomalies in the Mediterranean basin has increased sharply over the last few decades, with warming temperatures exceeding the range of normal fluctuations historically experienced by marine organisms 5 , 10 , 11 . This trend has boosted well-documented disease outbreaks and, consequently, mortality events that have affected several anthozoan species across varying geographic extents 4 , 12 – 14 . Despite recent advances, much remains to be understood about the molecular processes that underpin coral immune responses to pathogens. Consequently, it is crucial to determine the immune mechanisms involved and how they may be influenced by environmental signals (e.g., temperature). From an anatomical point of view, corals have a simple body structure consisting of only two true layers of tissue, and no organs. Depending on the stage of development – developing polyp or adult coral—these tissues are called ectoderm/epithelium and endoderm/gastroderm, respectively, interlined by the mesoglea (an amorphous and practically acellular extracellular matrix) 15 , 16 . In zooxanthellate species, Symbiodinium spp. (coral-dinoflagellate symbiont) reside mainly in the gastrodermal areas between the gut and the external epithelium (barrier to the surrounding environment) in the oral end, and more rarely in the gastrodermal areas between the intestine and the skeleton of the organism. Instead, nematocysts are present only in the epithelium of the oral end, but not in the calicoblastic epithelium 16 . The majority of host corals have only species-specific associations with specific bacterial phylotypes (including Symbiodinium spp.) which populate the different habitats of coral anatomical compartments, such as the surface mucus, tissues, and skeleton 17 , 18 . This suggests the presence of a well-developed self/non-self-recognition system in which appropriate strains multiply by establishing a stable symbiosis, while unsuitable strains are actively removed through cellular processes 6 , 19 , 20 . Corals and other cnidarians contain a surprisingly high degree of genetic complexity 21 – 24 , with homologs of many proteins involved in vertebrate immunity being described for these organisms in the last two decades 16 , 23 , 25 , 26 . The primary function of the anthozoan innate immune system is to recognize specific patterns of non-self-entities 27 , 28 . These patterns are called pathogen-associated molecular patterns (PAMPs), or microbe-associated molecular patterns (MAMPs), such as lipopolysaccharide (LPS), peptidoglycan, and mannan components of the microbial cell wall. These and other PAMPs are recognized by proteins called pattern recognition receptors (PRRs), which are molecules that include both membrane-bound proteins, such as Toll-like receptors (TLRs), and soluble proteins, such as lectins 1 , 16 , 25 , 29 . Corals possess TLRs and nucleotide oligomerization domain (NOD)-like receptors (NLRs) with intracellular Toll/interleukin-1 receptor (TIR) domains which can interact with genetic homologs of myeloid differentiation primary response protein 88 (MyD88), IL-1R-associated kinase (IRAK), receptor-associated factor 6 of TNF (TRAF), and IkB kinase (Ikk), which cleave nuclear factor kappa B (NF-kB) inhibitors and allow NF-kB protein dimers to translocate into the nucleus, enhancing the expression of inflammatory cytokines, the production of antimicrobial peptides (AMPs), cell survival, and apoptosis 16 , 25 , 29 , 30 . For example, the innate immune pathway from TLR to NF-kB in the tropical species Orbicella faveolata was recently characterized. Compared to human TLRs, the intracellular TIR domain is very similar to TLR4, and treatment of O. faveolata tissue with lipopolysaccharides (LPS; a common ligand for mammalian TLR4) led to changes in gene expression consistent with the mobilization of the NF-kB pathway 26 . One group of proteins used as ubiquitous and putative markers of temperature-induced cellular stress in corals are the heat shock proteins (HSPs) 31 – 35 . HSPs are differentiated by molecular weight into several major chaperone families (HSP40, HSP60, HSP70, HSP90, HSP100, and the small HSPs), with specific intracellular localization and function 36 . As molecular chaperones, they support protein homeostasis by facilitating proper protein folding and translocation, and by aiding in the folding or degradation of proteins damaged by heat or other environmental stresses 37 , 38 . The involvement of these proteins in the immune system has been widely reported in both vertebrates and invertebrates, increasing after exposure to biotic challenges such us during the development of infections and/or diseases 39 , 40 . For example, human HSP70 activates the TIR receptor signaling pathway during a highly inflammatory response 39 , 40 . For corals, the increased genetic expression of HSP70 in tropical species colonies showing signs of disease indicates their potential involvement in the immune and stress responses to pathogenic challenges 41 , 42 . The present study, using a manipulative aquarium-based experiment, aims to assess the impacts of warmer conditions on the activity of TLR4, NF-kB, and HSP70 markers under bacterial LPS challenge in an azooxanthellate coral species, Astroides calycularis (Pallas, 1766). This species is commonly found in the central-southern part of the Mediterranean Sea, covering vertical rocky reefs, overhangs, and caves below the intertidal fringe 43 . A. calycularis occupies both well-lit and sciaphilous habitats and is considered a tolerant thermophilic species, thriving at relatively high temperatures 43 . However, recent field observations indicate that this orange coral is impacted by seawater temperatures reaching and exceeding 28 °C 44 , 45 . In this experiment, the use of bacterial LPS allowed us to avoid the well-documented difficulties of infecting coral with live strains 46 , while testing the induction/suppression of immune pathways by PRRs 15 , 47 , 48 .",
"discussion": "Discussion The present study investigated the effects of warmer seawater conditions and LPS-challenge (simulating a potential pathogen aggression) on the expression of selected immunological (TLR4 and NF-kB) and stress (HSP70) markers in the Mediterranean scleractinian species A. calycularis . Both these types of abiotic and biotic factors are recognized as being among the most important contributors to the worldwide decline of coral habitats and mass mortality events 4 , 5 , 7 , 8 , 49 . Indeed, in recent decades, efforts to understand the cellular processes involved in the immunological and physiological responses to environmental stresses that cause mortality have increased. This is significant, especially considering that these animals cannot move to new optimal environmental conditions and that the earliest steps of the organism’s response occur at the cellular level 50 , 51 . To date, the regulation of the TLR-NF-kB pathway performed in corals under bacterial challenge has confirmed that these organisms possess temporally dynamic and responsive cellular machinery to counteract stresses 26 , 52 – 54 . Consistently with this, the results of our experiment showed significant activity modulation in corals at environmental temperature, for both TLR4 and NF-kB markers. Immunostaining using specific antibodies on sections of coral colonies exposed to LPS resulted in strong staining of nematocysts in the epithelia compared to control. Dense cellular aggregates lining the gastroderm directly adjacent to the lining of the gut in both oral (facing the mouth and external environment) and aboral (lining the skeleton) tissue were also stained. This wide activation is coherent with the expression levels in LPS-challenged samples detected by the Western blot analysis in the present study. Localization of the TLRs to both the nematocysts and the gastroderm would ensure that pathogens are exposed to receptor proteins regardless of how they enter the corals, thus activating the organism’s immune response. Given the binding capabilities of bacterial cell walls by TLRs, this exposure could guarantee an effective means to inhibit further tissue colonization. The activation times observed in this study (i.e., after 6 h of LPS exposure) are also consistent with previous observations, which showed up-regulated levels of mRNA encoding several TLR-NF-kB pathway components approximately 4 h after treatment of a tropical coral species with LPS 26 . Under healthy control conditions, A. calycularis therefore appears to possess the resources necessary to activate the TLR-NF-kB pathway after 6 h of LPS exposure. These immune pathways are essential in coral response to disease, and generally their tight regulation is a balanced consequence of signaling, organismal conditions, and overall immune strategy 55 . Under elevated temperature conditions, a significant up-regulation in TLR4 and NF-kB markers of LPS-unchallenged specimens was shown. Increased expression of several components involved in the TLR-NF-kB pathway has also been described in several anthozoans following heat stress treatment 56 – 59 . This provides further evidence that the innate immune system of corals is sensitive to environmental changes 48 , 60 , 61 . Conversely, when exposed to LPS challenge and warmer seawater conditions simultaneously, a significant suppression in both immune markers’ activities was observed, with a lack of extensive immunostaining in both the epithelial and gastrodermal tissues of the corals. Since the TLR-NF-kB signaling pathway also regulates AMP expression, the orange coral may reduce the production of these key molecules and, therefore, be unable to respond effectively to disease under thermal stress. Although no AMPs have yet been identified in A. calycularis , genes encoding AMPs have been characterized for several scleractinian corals and in other cnidarians 62 – 66 . Indeed, AMPs have been shown to be crucial in actively regulating and maintaining the health of the tissue-associated bacterial community in several species of anthozoans 62 , 67 , 68 . The suppression of these immuno-dynamics could have serious implications for this Mediterranean species during the summer months, when seawater temperatures are higher and the risk of disease outbreaks increases in the basin 4 , 12 – 14 . Despite these considerations, a limitation of this work is that it only considers one-time post-exposure to LPS, providing partial information about the organism's responses. Understanding and/or identifying the relevant timing of immune activities will be crucial for avoiding underestimates of the response capability to stress, and therefore the survival, of these ecologically relevant organisms. With regard to HSP70, a significant up-regulation was detected in LPS-challenged corals at environmental temperatures. Compared with the control slides, the immunohistochemical analysis showed an effective, wide activation in all tissue districts of the organisms. These findings suggest that changes in HSP70 occur in corals in response to microbial aggressions, also in the absence of thermal stress, further supporting their role in the coral immune response 34 , 42 . Indeed, several studies conducted on invertebrates (including corals) have already shown that modulations of HSP70 production appear to protect the organism from pathogenic infection 34 , 40 , 42 , 69 . For example, evidence that HSP70 enhances resistance to pathogens by priming and enhancing the expression of the pro-phenoloxidase system has already been presented 69 . Extracellular HSPs could also represent the ancestral danger signal of cell death or lysis-activating innate immunity 70 . In the present study, an increase in HSP70 protein production following LPS recognition could be an attempt to protect the organism, possibly by activating other constitutive components of the coral effector immune systems (e.g., via activation of the pro-phenoloxidase cascade). Another novel result from this study is a significant modulation of HSP70 activity at elevated seawater temperature, for both control and LPS-challenged treatments. The unchallenged corals showed significantly higher values than the controls at environmental temperature, though not reaching those of LPS-challenged colonies (both at environmental and elevated temperature). Indeed, under thermal stress, the key role of cytoplasmic HSPs (including HSP70) in several cellular processes has already been widely demonstrated; these include the folding of newly synthesized and misfolded proteins, the stabilization of the cytoskeleton, and protein transfer to other cellular compartments 38 , 71 . When LPS-challenged, A. calycularis showed a significantly up-modulation than control corals, also occurred in terms of increased areal activation of immunostaining tissues. The emerging dynamics suggest that, under warmer seawater conditions, the organism is able to activate a moderate response to thermal stress alone, probably in order to maintain homeostasis; instead, corals stimulated by LPS are able to further implement their response through the sensitive regulation of HSP70 production similar to that at environmental temperature. However, to confirm such immune dynamics under warmer conditions and pathogen eliciting in corals, studies over longer time intervals and with higher sampling resolution would be required. In conclusion, these results demonstrate that the immune response of A. calycularis exhibited 6 h post LPS-challenge is significantly affected by warmer seawater conditions. While thermal stress alone stimulated the activation of the coral TLR-NF-kB pathway, responses to LPS under elevated temperature were almost completely suppressed. HSP70 activity was up-modulated for both treatments under thermal stress, with the response of LPS-challenge colonies significantly stronger compared to control corals. These non-linear, temperature-induced responses of the examined markers could be the result of energetic trade-offs between maintaining homeostasis and the costs incurred to implement an effective immune response by the organism, occurring within the predetermined constraints of evolution 55 , 72 , 73 . Such an immuno-ecological approach represents a challenging path for evaluating immunocompetence and understanding natural patterns of disease among coral species across habitats. This study provides new biological information on an endemic Mediterranean species that can be used to better understand ecological patterns and, therefore, increase the accuracy of predicted responses to future climate-related events."
} | 4,265 |
29977334 | PMC6013992 | pmc | 2,072 | {
"abstract": "Background Bioelectrochemical systems (BESs) are an innovative technology developed to influence conventional anaerobic digestion. We examined the feasibility of applying a BES to dark hydrogen fermentation and its effects on a two-stage fermentation process comprising hydrogen and methane production. The BES used low-cost, low-reactivity carbon sheets as the cathode and anode, and the cathodic potential was controlled at − 1.0 V (vs. Ag/AgCl) with a potentiostat. The operation used 10 g/L glucose as the major carbon source. Results The electric current density was low throughout (0.30–0.88 A/m 2 per electrode corresponding to 0.5–1.5 mM/day of hydrogen production) and water electrolysis was prevented. At a hydraulic retention time of 2 days with a substrate pH of 6.5, the BES decreased gas production (hydrogen and carbon dioxide contents: 52.1 and 47.1%, respectively), compared to the non-bioelectrochemical system (NBES), although they had similar gas compositions. In addition, a methane fermenter (MF) was applied after the BES, which increased gas production (methane and carbon dioxide contents: 85.1 and 14.9%, respectively) compared to the case when the MF was applied after the NBES. Meta 16S rRNA sequencing revealed that the BES accelerated the growth of Ruminococcus sp. and Veillonellaceae sp. and decreased Clostridium sp. and Thermoanaerobacterium sp., resulting in increased propionate and ethanol generation and decreased butyrate generation; however, unknowingly, acetate generation was increased in the BES. Conclusions The altered redox potential in the BES likely transformed the structure of the microbial consortium and metabolic pattern to increase methane production and decrease carbon dioxide production in the two-stage process. This study showed the utility of the BES to act on the microbial consortium, resulting in improved gas production from carbohydrate compounds. Electronic supplementary material The online version of this article (10.1186/s13068-018-1175-z) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusions We assessed the application of a BES in the first stage of a two-stage process to recover hydrogen and methane using glucose as a model organic substrate. The BES used low-cost carbon sheets and applied electric current that was low enough to prevent water electrolysis by electrode polarisation. The initial pH in the first stage was relatively high, 6.5 or 7.3, to decrease the cost of lowering the pH. The second stage included CFTs in the reactor for efficient methane generation. The effect of electrode polarisation by the BES on the suppression of methanogenesis was clear at a relatively long HRT. Moreover, the electrode polarisation changed the microbial consortium structure and metabolic patterns. Particularly at a relatively short HRT, the BES increased the species richness of the microbial consortium and the relative abundance of microorganisms related to the Ruminococcus genus and Veillonellaceae family, corresponding to an increase in the generation of ethanol and propionate. A decreased relative amount of microorganisms related to the Clostridium and Thermoanaerobacterium genera corresponded to a decrease in the generation of hydrogen and butyrate in the BES. In addition, a greater amount of acetate was generated in the BES. These changes were likely triggered by changing the redox potential of the electrode; however, future clarification of this mechanism is necessary. The BES reduced the amount of gas produced in first stage and increased the amount produced in the second stage. This resulted in an increase in the generation of methane and decrease in the generation of carbon dioxide in the two-stage process.",
"discussion": "Discussion Methane production in the second stage was increased owing to a change in the microbial consortium, which was driven by the application of the BES in the first hydrogen fermentation stage due to a change in the microbial consortium. The two-stage process including the BES decreased the total production of carbon dioxide under an HRT of 2 days by lowering the amount of gas in the first stage with a relatively high content of carbon dioxide and increasing the amount of gas in the second stage with a relatively low content of carbon dioxide, which contributed to a decrease in the cost of upgrading biogas to remove carbon dioxide. The number and diversity of microbial species were high owing to the low electric current in the reactor under this HRT condition, although the cell production was similar for the BES and NBES (Fig. 4 ). High microbial diversities can enhance resistance to environmental stresses, such as high organic loads [ 26 ]. During dark hydrogen fermentation, the upper limit of the hydrogen yield of glucose is 4 mol H 2 per mole of hexose during acetic fermentation [Eq. ( 1 )], while 2 mol H 2 per mole of hexose is recorded in the butyrate pathway [Eq. ( 2 )] [ 16 , 25 ]: 1 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ {\\text{C}}_{ 6} {\\text{H}}_{ 1 2} {\\text{O}}_{ 6} + {\\text{ 2H}}_{ 2} {\\text{O }} \\to {\\text{ 2CH}}_{ 3} {\\text{COOH }} + {\\text{ 2CO}}_{ 2} + {\\text{ 4H}}_{ 2} $$\\end{document} C 6 H 12 O 6 + 2H 2 O → 2CH 3 COOH + 2CO 2 + 4H 2 \n 2 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ {\\text{C}}_{ 6} {\\text{H}}_{ 1 2} {\\text{O}}_{ 6} \\to {\\text{ CH}}_{ 3} {\\text{CH}}_{ 2} {\\text{CH}}_{ 2} {\\text{COOH }} + {\\text{ 2CO}}_{ 2} + {\\text{ 2H}}_{ 2} . $$\\end{document} C 6 H 12 O 6 → CH 3 CH 2 CH 2 COOH + 2CO 2 + 2H 2 . \n Thus, the amount of hydrogen produced as calculated from the concentration of acetate and butyrate in the NBES at an HRT of 2 days with a substrate pH in of 6.5 and 7.3 was 42.2 mM [= 2 × (17/2) + 2 × (25.2)/2] and 35.3 mM, comparable to the actual hydrogen production of 44.4 and 27.8 mM, respectively. However, the actual amount of hydrogen produced, 24.3 and 16.7 mM, in the BES was much lower than the values calculated from the acetate and butyrate concentrations, 39.4 and 36.1 mM, at an HRT of 2 days with a substrate pH in of 6.5 and 7.3, respectively. These results showed that a hydrogen-consumption reaction occured in the BES, although the ratio of hydrogen to carbon dioxide in the gas was unchanged in the BES and NBES. This was also supported by the fact that the hydrogen yield was lower in the BES (0.60–0.87 mM/mM glucose ) than in the NBES (1.00–1.60 mM/mM glucose ) at an HRT of 2 days, as the typical hydrogen yield ranges from 1 to 2.5 mM/mM glucose [ 25 ]. Next, we considered the reason for the low hydrogen yield caused by the low electric current induced by electrode polarisation in the BES operated at an HRT of 2 days. It is reasonable that microorganisms belonging to Bacillus , Clostridium , and Thermoanaerobacterium were dominant in both the BES and NBES, because these organisms include typical anaerobic fermentative bacteria that convert monosaccharides into hydrogen [ 27 – 30 ]. Clostridium and Thermoanaerobacterium are the dominant hydrogen producers during acetate/butyrate fermentation under thermophilic conditions [ 24 , 31 ]; thus, a decrease in these genera corresponds corresponded with decreased butyrate production in the BES. Interestingly, microorganisms belonging to the Ruminococcus species that reportedly produce ethanol in addition to hydrogen and acetic acid [ 32 ], increased in the BES. Accordingly, this corresponded with increased ethanol production in the BES. Microorganisms belonging to the Veillonellaceae family, which is known to produce propionate as a major fermentation product [ 33 , 34 ], also increased in the BES. This result corresponds with increased propionate production in the BES, which contributes to hydrogen consumption [Eq. ( 3 )]: 3 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ {\\text{C}}_{ 6} {\\text{H}}_{ 1 2} {\\text{O}}_{ 6} + {\\text{ 2H}}_{ 2} \\to {\\text{ 2CH}}_{ 3} {\\text{CH}}_{ 2} {\\text{COOH }} + {\\text{ 2 H}}_{ 2} {\\text{O}} . $$\\end{document} C 6 H 12 O 6 + 2H 2 → 2CH 3 CH 2 COOH + 2 H 2 O . \n An increase in volatile fatty acids, except butyrate, led to a decrease in hydrogen production in the BES, whereas acetate production increased. One explanation for the increased acetate production irrespective of the lower hydrogen yield is that an acetogenic hydrogen-consuming reaction, homoacetogenesis [Eq. ( 4 )] [ 35 , 36 ], may occur in the BES; however, this has not been clarified 4 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ 2 {\\text{CO}}_{ 2} + {\\text{ 4H}}_{ 2} \\to {\\text{ CH}}_{ 3} {\\text{COOH }} + {\\text{ 2H}}_{ 2} {\\text{O}} . $$\\end{document} 2 CO 2 + 4H 2 → CH 3 COOH + 2H 2 O . \n The mechanism of the change in the structure of the microbial community is interesting. Direct electron transfer between the electrode and microorganisms [ 37 ] had a low impact in the reactions because of the low current density. At an HRT of 4 days, the BES inhibited methanogenic archaea, corresponding to the results of our previous research [ 15 ]. Previous studies have shown that the redox potential in the fermentation culture affects the growth of methanogenic archaea [ 38 , 39 ]. We speculate that a high redox potential owing to the anodic reaction suppressed the growth and/or methanogenesis by methanogenic archaea [ 15 ], considering the fact that the redox potential of the anode was 0.85 V in this study. Villano et al. [ 40 ] showed that cathodic reaction increased isobutyrate production in the microbial consortium cultured in the BES with a proton exchange membrane to separate the cathode and anode. Thus, the construction of an environment with different redox potentials in the reactor could change the microbial consortium structure, leading to increased growth of microorganisms related to the Ruminococcus genus and Veillonellaceae family. The species richness was higher in the second stage (i.e., MF) compared to that in the first stage, probably due to the neutral pH conditions during methane fermentation. For methanogenesis, hydrogenotrophic methanogens (i.e., Methanobacterium and Methanothermobacter ) and acetoclastic methanogen (i.e., Methanosarcina ) [ 12 ], increased in the MF cultures. It is reasonable that an increase in microorganisms related to Pelotomaculum and Syntrophomonas was observed in the MF cultures, because these microorganisms reportedly grow via syntrophy with methanogens to degrade propionate and butyrate, respectively [ 41 , 42 ]. Interestingly, the microbial community structure in the fermentation cultures did not differ significantly between second-stage and single-stage MF, probably due to the retention of major microorganisms in the CFT [ 18 , 43 ]."
} | 2,893 |
21350596 | PMC3028068 | pmc | 2,073 | {
"abstract": "Honey bee queens ( Apis mellifera ) who mate with multiple males produce colonies that are filled with numerous genetically distinct patrilines of workers. A genetically diverse colony benefits from an enhanced foraging effort, fuelled in part by an increase in the number of recruitment signals that are produced by foragers. However, the influence of patriline diversity on the attention paid to these signals by audiences of potentially receptive workers remains unexplored. To determine whether recruitment dances performed by foragers in multiple-patriline colonies attract a greater number of dance followers than dances in colonies that lack patriline diversity, we trained workers from multiple- and single-patriline colonies to forage in a greenhouse and monitored their dance-following activity back in the hives. On average, more workers followed a dance if it was performed in a multiple-patriline colony rather than a single-patriline colony (33% increase), and for a greater number of dance circuits per follower. Furthermore, dance-following workers in multiple-patriline colonies were more likely to exit their hive after following a dance, although this did not translate to a difference in colony-level exit rates between treatment types. Recruiting nest mates to profitable food sources through dance communication is critical to a colony’s foraging success and long-term fitness; polyandrous queens produce colonies that benefit not only from increased recruitment signalling, but also from the generation of larger and more attentive audiences of signal receivers. This study highlights the importance of integrating responses of both signal senders and receivers to understand more fully the success of animal-communication systems.",
"introduction": "Introduction Eusocial insect taxa in the Order Hymenoptera are characterized by work forces of non-reproductive individuals who share high levels of intracolonial relatedness (Wilson, 1971 ). Typically, close relatedness among colony members is maintained through monandry (Strassmann, 2001 ), when a queen mates with a single male. Monandry is ancestral in the Hymenoptera and remains widespread across the majority of its extant social species (Hughes et al., 2008a ). Because of the haplodiploid system of sex determination, a monandrous queen produces a work force of female offspring who have on average 75% of their genes in common, a condition that is viewed as an important precursor for the repeated evolution of altruism in hymenopteran societies (Hughes et al., 2008a ). However, not all eusocial Hymenoptera have work forces that are populated by highly related kin. Colonies of a minority of species of social bees, ants and wasps are comprised of genetically distinct patrilines of workers because the queens that head them have each mated with multiple males (polyandry). Extremely polyandrous mating behaviour has evolved repeatedly in eusocial groups such as the army and driver ants ( Eciton , Dorylus ; Kronauer et al., 2004 , 2006 ), leaf-cutter ants ( Atta , Acromyrmex ; Villesen et al., 2002 ), harvester ants ( Pogonomyrmex ; Wiernasz et al., 2004 ), yellow jacket wasps ( Vespula ; Foster and Ratnieks, 2001 ) and honey bees ( Apis ; Oldroyd and Wongsiri, 2006 ; Palmer and Oldroyd, 2000 ; Tarpy et al., 2004 ). Thus, extreme polyandry presents a fascinating challenge to the predicted scenario of high levels of intracolonial relatedness in eusocial hymenopteran colonies. Polyandry in honey bees is particularly intriguing in contrast to the ancestral state of monandry in the Hymenoptera because, across species, it occurs universally and to a tremendous extent. Queens of the European honey bee ( Apis mellifera ) typically mate with an average of 12 different males (drones) (Tarpy et al., 2004 ), with a record of 44 mates reported for one queen (Moritz et al., 1996 ). One A. dorsata colony was found to have a staggering 102 patrilines (Wattanachaiyingcharoen et al., 2003 ). Such exceptionally promiscuous honey bee queens create work forces that are so genetically diverse that within-colony relatedness is diminished to a degree seemingly at odds with the promotion of altruism based on inclusive-fitness benefits (Hamilton, 1964 ). However, honey bee workers have lost full reproductive totipotency and the option to gain significant direct fitness, which forces them to act as helpers and accept reductions in indirect fitness should queens mate multiply and levels of within-colony relatedness decline (Beekman et al., 2006 ). That extreme polyandry is so pervasive in Apis suggests that it is favoured strongly by selection once workers are locked into a non-reproductive role (Boomsma, 2009 ; Brown, 2003 ; Crozier and Fjerdingstad, 2001 ; Crozier and Page, 1985 ; Hughes et al., 2008a , b ; Oldroyd and Fewell, 2008 ). Indeed, a growing body of evidence shows that honey bee colonies reap numerous fitness benefits from the genetic diversity that is introduced into their population by the polyandrous behaviour of their queens. Honey bee queens who mate multiply minimize the fitness load incurred by colonies as a result of the production of sterile diploid males (Page, 1980 ; Shaskolsky, 1976 ; Tarpy and Page, 2002 ) and produce work forces with an enhanced ability to overcome the effects of parasites and pathogens (Baer and Schmid-Hempel, 2001 ; Palmer and Oldroyd, 2003 ; Seeley and Tarpy, 2007 ; Sherman et al., 1988 ; Tarpy and Seeley, 2006 ). Importantly, intracolonial genetic diversity has been linked to an increase in colony-level productivity and long-term fitness (Fuchs and Schade, 1994 ; Jones et al., 2004 ; Oldroyd et al., 1992 ; Mattila and Seeley, 2007 ). Similar increases in colony growth and foraging productivity have been found in harvester ants (Cole and Wiernasz, 1999 ; Wiernasz et al., 2004 , 2008 ) and wasps (Goodisman et al., 2007 ) as mate number per queen rises. For honey bees, colony productivity is boosted in part by an enhanced use of recruitment communication in genetically diverse, multiple-patriline colonies compared with those that lack the same degree of diversity because they have monandrous queens and single patrilines (Mattila et al., 2008 ; Mattila and Seeley, 2010 ). Honey bees use elaborate signals, such as the waggle dance and shaking signal, to communicate to nest mates information about the availability and location of profitable food resources (Seeley, 1995 ; von Frisch, 1967 ). Given the same foraging environment, workers in multiple-patriline colonies, as a collective, produce longer and more frequent waggle-dance signals; they advertise food resources that are farther away from the nest, and they produce more shaking signals at the start of the day of foraging than do workers in single-patriline colonies (Mattila et al., 2008 ). Foragers in genetically diverse colonies show universally a heightened responsiveness to available resources, including colony-wide decreases in dance-response thresholds (i.e., across all patrilines) and increases in per capita rate of resource visitation and duration of dance response relative to colonies with minimal diversity (Mattila and Seeley, 2010 ). Taken together, these studies highlight some of the mechanisms by which greater signal output contributes to increases in the foraging effort of colonies with polyandrous queens. Mattila and Seeley ( 2010 ) focused on the recruitment signals that are produced by honey bee foragers in multiple- and single-patriline colonies without considering the other half of the equation––the effect of these signals on their intended audience. Recruitment can be enhanced not only by the production of stronger signals by senders, but also by an increase in the number of receivers who heed signals and subsequently respond to the information that they contain. By examining the response of honey bee workers to the dances that they follow, the effects of increased signal production by dancers can be integrated with the response of the workers that follow those dances to more completely understand how a colony’s foraging and recruitment effort benefits from genetic diversity within its work force. Hence, the aim of this study was to compare the behaviour of workers as they attend dances that are performed by foragers in multiple-patriline colonies to that of workers exposed to signals in single-patriline colonies. We trained workers from both types of colonies to forage in a greenhouse where a single feeder contained known concentrations of sucrose solution and, as the foragers danced to advertise the food source back in their hive, we determined (1) the number of workers that followed a dance, (2) the number of dance circuits that were followed per dance follower, (3) the action taken by followers after they stopped attending a dance, and (4) the rate at which workers exited the hive as recruitment to the feeder began. Using these data, we assess how differences in foraging performance can arise between genetically diverse and uniform colonies from the perspective of those individuals who are motivated to seek out and to act on the recruitment signals that are available to them in their colonies.",
"discussion": "Discussion This study demonstrates that a honey bee colony with multiple patrilines makes greater use of recruitment-signal communication than a colony that lacks such genetic diversity. Communication is enhanced through two important mechanisms and complementary perspectives when a colony has numerous patrilines: first, by the production of more signals by each sender (Mattila et al., 2008 ; Mattila and Seeley, 2010 ) and, second, by heightened exposure of receivers to those signals (shown here). Signal receipt was increased because dances that advertised a “nectar source” (i.e. a feeder stocked with sucrose solution) in multiple-patriline colonies generated larger audiences of more attentive dance followers than dances in single-patriline colonies. On average, in multiple-patriline colonies compared with single-patriline colonies, 33% more workers followed each focal dance and for almost 2 more circuits per worker. In addition to a greater number of workers following each dance and for a longer time on average, dance followers from multiple-patriline colonies were 2.7 times more likely to leave the hive within 2 min after they stopped attending a focal dance compared with their counterparts in single-patriline colonies. Conversely, a significantly greater proportion of dance followers from single-patriline colonies remained on the dance floor after they stopped following a dance, perhaps having become disengaged from the recruitment process, as suggested by their lower tendency to seek out subsequent dances. That such strong differences were found between colony types with a modest sample size (a necessary trade of colony number for detailed behavioural observations) is compelling evidence that these differences are real. Furthermore, it reinforces the trend that single-patriline honey bee colonies consistently underperform relative to multiple-patriline colonies (Fuchs and Schade, 1994 ; Jones et al., 2004 ; Mattila and Seeley, 2007 , 2010 ; Mattila et al., 2008 ; Oldroyd et al., 1992 ), rather than showing similar mean performance with a relative increase in variation around that mean. For reasons that were not made clear by this study, relative exit rates from the colony during the 30 min that the feeder was open did not reflect this greater tendency of workers in multiple-patriline colonies to leave the hive after following a dance. These results were surprising, given that concurrent increases in waggle dancing and foraging rates have been found in multiple-patriline colonies relative to single-patriline colonies in other contexts (Mattila and Seeley, 2007 ; Mattila et al., 2008 ). These conflicting findings may reflect our inability to disentangle in a greenhouse-foraging assay counts of workers who were exiting their hive in search of the feeder from those workers who were leaving for other reasons (e.g., searching for the water source, orientating to the entrance and greenhouse, cleansing flights). Further study is needed to determine whether potential recruits in multiple-patriline colonies are not only more receptive to dance signals, but are also more successful after they leave the hive in the use of the information that they receive––that is, whether they are more likely to find a food source after following a dance that advertises it because they have more information about it. Such an experiment should be executed in a natural foraging environment with a greater distance between the feeder and the hive so that the extent to which extra information helps dance followers to locate a food source can be determined definitively. It is important to emphasize that we found increases in dance-following activity on a per - dance - circuit basis as foragers from multiple-patriline and single-patriline colonies visited a strictly controlled, singular resource. Therefore, the greater number of bees attending dances in multiple-patriline colonies is not only the result of greater signal production by dancing foragers in these colonies (i.e. relatively longer focal dances; see also Mattila and Seeley, 2010 ), but also a reflection of increased exposure of potential recruits to each dance circuit. This colony-level finding extends our understanding of the reasons why, in honey bee colonies, work force productivity and foraging effort are enhanced when colonies have a high degree of genetic diversity (Mattila and Seeley, 2007 ; Mattila et al., 2008 ). Such boosts in the responsiveness of workers to recruitment signals may explain a similar report of increased foraging productivity in genetically diverse ant colonies with polyandrous queens (Wiernasz et al., 2008 ). Differences in the dance-following activity of workers between multiple-patriline and single-patriline colonies may be explained by two hypotheses that are not mutually exclusive. First, it may be the case that dances performed by foragers in multiple-patriline colonies are more attractive to their nest mates, prompting a greater number of workers to orient and remain attentive to them. Our analysis of focal dancers did not indicate that dances performed in genetically diverse colonies conveyed more “excitement” or “enthusiasm” [at least, based on their tempo (Seeley et al., 2000 )] than those performed in genetically uniform colonies, although they could have been more appealing to potential recruits based on parameters that we did not evaluate, such as the amplitude of the brief waggle movements, the release by dancers of pheromones that modulate nest mate behaviour (Thom et al., 2007 ), or a overall levels of activity on the dance floor. The fact that the focal dances were longer on average in multiple-patriline colonies may reflect an aspect of attractiveness that was not captured by our study. The second hypothesis considers the nature of the dance followers themselves. It is possible that dance-following activity was enhanced in some proportion of the worker population in multiple-patriline colonies for one or a number of the following reasons: because some workers had (1) a genetic propensity to participate in dance following (Arnold et al., 2002 ), (2) generally lower thresholds for initiating a dance-following response to dance-signal stimuli (Robinson and Page, 1988 ), (3) faster response times to dances––which are often short––once they began, (4) greater attention spans, which could result in greater time spent attending a dance, (5) differences in the probability of response once thresholds were exceeded (Weidenmüller, 2004 ), or (6) a strong inclination to advertise a nectar resource over other types of forage, such as pollen (Hunt et al., 1995 ; Oldroyd et al., 1991 ). Any combination of these trait differences would generate a boost in the dance-following activity of workers and increase the size of the audience that heeds the signals that are produced in multiple-patriline colonies. None of these hypotheses can be ruled out by this study, and we believe that it is likely that several are reasons why communication is enhanced in colonies with a high degree of patriline diversity. Above all, an interesting question that remains unanswered is the extent to which patrilines of workers differentially participated as audience members in recruitment signalling. Here, we have demonstrated substantial differences in colony-level exchange of recruitment signals in multiple- and single-patriline colonies; however, we were not able to determine the role that patriline membership played in forming the response of a colony’s forager work force to recruitment dances when numerous patrilines were present in a colony, primarily because destructively sampling focal followers would have interfered with our behavioural observations. This line of investigation is important to pursue given that, within the same experimental setup, small subsets of patrilines were responsible for producing a majority of the dance signals in these colonies (Mattila and Seeley, 2010 ). Foragers who successfully locate a food source that is advertised by a waggle dance need to follow a mean of at least eight waggle runs of a dance to obtain sufficient information about the location of that resource (Judd, 1995 ). The same information about distance and direction that is encoded in waggle dances for faraway food sites is also found in so-called “round dances” (i.e. dance signals that contain brief waggle phases) that broadcast the location of resources that are near the colony (Kirchner et al., 1988 ; Jensen et al., 1997 ; Gardner et al., 2007 ). It is not clear whether successful recruits need to follow more or less than eight circuits of a recruiting “round” waggle to find nearby sites with the same degree of success, but it is interesting that the average number of circuits followed by dance followers in this study hovers around this 8-circuit mark (Fig. 1 b), the threshold number of informative waggle signals shown by Judd ( 1995 ) to promote successful food-site discovery by dance-following recruits. Furthermore, calculated across concentrations, the average number of circuits followed in single-patriline colonies was less than eight per follower. Although dance-following workers in multiple-patriline colonies followed an average of only 1.6 more circuits per dance compared with workers in single-patriline colonies, this difference may have noteworthy consequences if it creates large differences in recruitment success simply because more workers in multiple-patriline colonies are exposed to a threshold number of dance signals, either a number that prompts them to exit the nest (as seen here) or one that allows sufficient time for recruits to glean enough information to successfully locate the resource. Had dance followers already known the location of the feeder because they had visited it recently, perhaps the number of circuits that they followed may be of lesser importance if attendance at a dance only jogged followers’ memories through reactivation. However, the design of our study greatly limited the experience that dance followers had with the feeder and, thus, we expected that workers who were stimulated to locate the resource would follow sufficient numbers of dances signals to allow them to learn adequate information about the feeder before leaving the hive. Based on such a scenario, we infer that workers from multiple-patriline colonies are more motivated to learn about the food source than workers from single-patriline colonies. Furthermore, the tendency exhibited by workers in multiple-patriline to follow a second dance (for the feeder) before exiting the hive may also reflect increased motivation to learn about food sources in general, or our feeder specifically, relative to dance followers in single-patriline colonies. It is becoming abundantly clear that the execution of a large-scale, self-organized foraging effort in genetically diverse honey bee colonies is enhanced both by the production of more signals by waggle-dancing foragers and by increased responsiveness of the audience of workers for whom the dance signals are intended. By raising the activity of both signal producers and receivers, multiple mating expedites the transfer among nest mates of information about available resources. The advantage of such amplification of communication is also observed in simultaneous group departures that are characteristic of gorillas (Stewart and Harcourt, 1994 ) and flocks of whooper swans (Black, 1988 ). Initially, a few individuals make their intention to leave known through the production of signals, which are imitated by other members of the group who are similarly motivated to leave. For both gorillas and whooper swans, a threshold level of vocalizations or body movements must be produced by the collective before group action is taken. Hence, communication is enhanced by positive feedback because increased signalling draws the attention of more group members, who in turn become signallers. A similar positive feedback is probably at work in honey bee colonies: longer and more frequent dance signals, coupled with a more receptive audience of dance followers, likely begets even more dancing as these motivated audience members search for and discover advertised resources and, presumably, become signal producers (with some probability) upon return to their hive. Beyond the influence of genetic diversity on recruitment signalling in honey bee colonies, this study highlights the need to examine not only the perspective of signal producers, but also the perspective of signal receivers when evaluating the efficacy of communication in animal systems. After all, it is the action, or the lack thereof, taken by a signal producer’s audience that determines the success of the effort to communicate."
} | 5,509 |
25495940 | null | s2 | 2,074 | {
"abstract": "Mine tailings in semiarid regions are highly susceptible to erosion and are sources of dust pollution and potential avenues of human exposure to toxic metals. One constraint to revegetation of tailings by phytostabilization is the absence of microbial communities critical for biogeochemical cycling of plant nutrients. The objective of this study was to evaluate specific genes as in situ indicators of biological soil response during phytoremediation. The abundance and activity of 16S rRNA, nifH, and amoA were monitored during a nine month phytostabilization study using buffalo grass and quailbush grown in compost-amended, metalliferous tailings. The compost amendment provided a greater than 5-log increase in bacterial abundance, and survival of this compost-inoculum was more stable in planted treatments. Despite increased abundance, the activity of the introduced community was low, and significant increases were not detected until six and nine months in quailbush, and unplanted compost and buffalo grass treatments, respectively. In addition, increased abundances of nitrogen-fixation (nifH) and ammonia-oxidizing (amoA) genes were observed in rhizospheres of buffalo grass and quailbush, respectively. Thus, plant establishment facilitated the short term stabilization of introduced bacterial biomass and supported the growth of two key nitrogen-cycling populations in compost-amended tailings."
} | 352 |
25905817 | PMC5386109 | pmc | 2,076 | {
"abstract": "Condensed liquid behavior on hydrophobic micro/nano-structured surfaces is a subject with multiple practical applications, but remains poorly understood. In particular, the loss of superhydrophobicity of hydrophobic micro/nanostructures during condensation, even when the same surface shows water-repellant characteristics when exposed to air, requires intensive investigation to improve and apply our understanding of the fundamental physics of condensation. Here, we postulate the criterion required for condensation to form from inside the surface structures by examining the grand potentials of a condensation system, including the properties of the condensed liquid and the conditions required for condensation. The results imply that the same hydrophobic micro/nano-structured surface could exhibit different liquid droplet behavior depending on the conditions. Our findings are supported by the observed phenomena: the initiation of a condensed droplet from inside a hydrophobic cavity, the apparent wetted state changes, and the presence of sticky condensed droplets on the hydrophobic micro/nano-structured surface."
} | 281 |
38512481 | PMC10957709 | pmc | 2,077 | {
"abstract": "Abstract Thermophilic cyanobacteria are prokaryotic photoautotrophic microorganisms capable of growth between 45 and 73 °C. They are typically found in hot springs where they serve as essential primary producers. Several key features make these robust photosynthetic microbes biotechnologically relevant. These are highly stable proteins and their complexes, the ability to actively transport and concentrate inorganic carbon and other nutrients, to serve as gene donors, microbial cell factories, and sources of bioactive metabolites. A thorough investigation of the recent progress in thermophilic cyanobacteria reveals a significant increase in the number of newly isolated and delineated organisms and wide application of thermophilic light-harvesting components in biohybrid devices. Yet despite these achievements, there are still deficiencies at the high-end of the biotechnological learning curve, notably in genetic engineering and gene editing. Thermostable proteins could be more widely employed, and an extensive pool of newly available genetic data could be better utilised. In this manuscript, we attempt to showcase the most important recent advances in thermophilic cyanobacterial biotechnology and provide an overview of the future direction of the field and challenges that need to be overcome before thermophilic cyanobacterial biotechnology can bridge the gap with highly advanced biotechnology of their mesophilic counterparts. Key points \n • Increased interest in all aspects of thermophilic cyanobacteria in recent years \n \n • Light harvesting components remain the most biotechnologically relevant \n \n • Lack of reliable molecular biology tools hinders further development of the chassis \n Graphical Abstract",
"introduction": "Introduction to thermophilic cyanobacteria Thermophilic cyanobacteria, usually found in hot springs, are prokaryotic microorganisms capable of growth between 45 and 73 °C and utilising oxygenic photosynthesis as a main source of energy and carbon. In this work, we define thermophilic cyanobacteria as members of cyanoprokaryota for which “part or all of their optimal growth temperature range is above 45 °C” according to a well-established definition by Castenholz ( 1969 ). In addition to their high growth temperature, they sometimes exhibit polyextremophilic characteristics regarding high pH or elevated concentration of metal ions. These organisms are also essential primary producers of geothermal ecosystems that significantly contribute to carbon and nitrogen fixation and hot spring productivity. Together with other microorganisms, they usually form stratified microbial mats in thermal springs (Esteves-Ferreira et al. 2018 ; Kees et al. 2022 ). These robust photosynthetic microbes possess features that make them highly biotechnologically relevant (Fig. 1 ). Potentially useful natural products from thermophilic cyanobacteria include bioactive metabolites and polymers. They can also serve as thermostable gene donors for other organisms and as microbial cell factories for waste valorisation. The biggest advantages that the thermophilic cyanobacteria may have over their mesophilic counterparts are directly coupled to their high growth temperature. The most notable aspects include protection from microbial contamination and grazers that are incapable of withstanding high temperatures, biosynthesis of highly stable proteins and their complexes, and better prospects for transgene selection and biocontainment (Liang et al. 2019 ; Patel et al. 2019 ). Fig. 1 Overview of biotechnological applications of thermophilic cyanobacteria. P: promoter; RBS: Ribosome Binding Site; GOI: Gene of Interest; T: Terminator; PS II: Photosystem II; PS I: Photosystem I. The figure was partly generated using Servier Medical Art, provided by Servier, licensed under a Creative Commons Attribution 3.0 unported license In recent years, there has been a renewed interest in these organisms leading to a significant expansion of knowledge about their taxa, properties, metabolism, and biotechnological potential (Saini et al. 2021 ). This is particularly evident in the recent surge in the delineation of novel thermophilic taxa such as the following Thermoleptolyngbya (Sciuto and Moro 2016 ), Thermostichus (Komárek et al. 2020 ), Leptothermofonsia (Tang et al. 2022b ), Trichothermofontia (Tang et al. 2023 ), or Thermocoleostomius (Jiang et al. 2023 ). These are often combined with the collection and analysis of extensive genomic data about representative thermophilic isolates (Tang et al. 2022c ). Simultaneously, widespread deployment of next-generation sequencing (NGS) technologies allowed further expansion of the sequence space among the thermosphere and depositing of numerous metagenome-assembled genomes (MAGs) from hot spring sequencing projects, opening more possibilities for biotechnological exploration of this resource. Despite the significant increase in the sequence information, thermophilic cyanobacteria remain relatively understudied in several aspects compared to their mesophilic counterparts, most notably in genetic engineering. In this work, we attempt to synthesise the most important recent advances in thermophilic cyanobacterial biotechnology and provide an outlook into the future direction of the field. Such includes challenges that must be overcome before thermophilic cyanobacterial biotechnology can bridge the gap with more advanced biotechnology of their mesophilic counterparts."
} | 1,370 |
36384960 | PMC9668812 | pmc | 2,079 | {
"abstract": "Neuromorphic machines are intriguing for building energy-efficient intelligent systems, where spiking neurons are pivotal components. Recently, memristive neurons with promising bio-plausibility have been developed, but with limited reliability, bulky capacitors or additional reset circuits. Here, we propose an anti-ferroelectric field-effect transistor neuron based on the inherent polarization and depolarization of Hf 0.2 Zr 0.8 O 2 anti-ferroelectric film to meet these challenges. The intrinsic accumulated polarization/spontaneous depolarization of Hf 0.2 Zr 0.8 O 2 films implements the integration/leaky behavior of neurons, avoiding external capacitors and reset circuits. Moreover, the anti-ferroelectric neuron exhibits low energy consumption (37 fJ/spike), high endurance (>10 12 ), high uniformity and high stability. We further construct a two-layer fully ferroelectric spiking neural networks that combines anti-ferroelectric neurons and ferroelectric synapses, achieving 96.8% recognition accuracy on the Modified National Institute of Standards and Technology dataset. This work opens the way to emulate neurons with anti-ferroelectric materials and provides a promising approach to building high-efficient neuromorphic hardware.",
"introduction": "Introduction In the past few decades, neuromorphic computing, mimicking the human brain’s architecture and operation with electronic devices, has attracted great interest due to its high biomimetic and high-energy efficiency 1 – 3 . Artificial neurons are the core components of neuromorphic computing implementation, emulating biological neurons functions of potential accumulation and firing 4 , 5 . For the hardware implementation of neurons, hardware overhead, energy efficiency, and reliability are the critical evaluation criteria 5 , 6 . Yet, current hardware demonstrations of neurons struggle to satisfy these key metrics simultaneously. Generally, the complementary metal-oxide-semiconductor (CMOS) circuit is the most mature and stable scheme for emulating biological neurons. Nevertheless, due to the lack of intrinsic biological resemblance and the complexity of circuits, CMOS neurons face many challenges in density or energy efficiency 7 – 9 . Recently, various emerging devices have been extensively explored to emulate biological neurons benefiting from their biological resemblance and scalability. Memristive neurons have trigged the most interest among them, including redox memristors 10 – 13 , Mott memristors 14 – 19 , phase-change memristors (PCM) 20 – 22 , magnetic random access memory (MRAM) 23 , 24 , etc. These neurons utilize the gradual switching of conductance to mimic membrane potential evolution, successfully emulating essential biological neuron functions with low hardware cost. However, high electroforming voltage and limited reproducibility due to temporal and spatial variations are still open questions 25 , 26 . In addition, capacitors are usually needed to realize the integration in memristive neurons, which limits their practical applications in large-scale neuromorphic computing systems 10 , 17 , 27 . In the very recent research, novel ferroelectric polarization-based neurons are proposed and experimentally demonstrated 28 – 32 . They utilize gradual polarization to mimic the integration process of biological neurons without additional capacitors 33 . Moreover, polarization is the intrinsic property of ferroelectric materials, which is recognized to be reproducible, reliable, and energy-efficient 29 , 34 . These features are promising to implement neurons. However, ferroelectric devices are nonvolatile, and thus need a feedback path 28 – 30 or a special design of ferroelectric layer 31 , 32 to achieve spontaneous reset after firing. The feedback path will increase the hardware cost and energy consumption of neuron implementation. In addition, it will increase the complexity of the operation, as each new input must wait for the completion of the previous reset process, especially in a system with a rate coding scheme. Thus, demonstrating an ideal electronic device that processes advanced and balanced neuronal performance without additional capacitors and reset feedback path deserves more attention. In this work, we report a leaky integrate-and-fire (LIF) neuron based on a CMOS-compatible anti-ferroelectric field-effect transistor (AFeFET). The intrinsic polarization/depolarization processes of the Hf 0.2 Zr 0.8 O 2 AFeFET successfully emulate the integrate/leaky neuronal functions without any capacitors and reset peripheral circuits. Furthermore, attributing to the plentiful merits of ferroelectric materials, AFeFET neuron exhibits many superiorities: electroforming-free, ultra-low-energy consumption (37 fJ/spike), high endurance (>10 12 ), small cycle-to-cycle variation (as low as 3.93%) and device-to-device variations (7.57%). Also, we present that the temporal integration speed in such an AFeFET neuron depends on the intensity of postsynaptic potential, illustrating the fundamental features for performing classification tasks. Subsequently, we demonstrate a two-layer spiking neural network (SNN) with full-ferroelectric architecture for learning and recognizing the Modified National Institute of Standards and Technology (MNIST) datasets by simulation, obtaining the maximum recognition accuracy 96.8% comparable to ideal neurons. These results demonstrate that the proposed AFeFET neuron is a competitive candidate for constructing neuromorphic systems.",
"discussion": "Discussion SNNs, inspired by the human brain, are powerful platforms for enabling low-power event-driven neuromorphic hardware. In SNNs, spiking neurons are the key units that enable spikes, which exchange information through connected plastic synapses. With rich physical dynamics, memristive devices are considered promising devices to emulate spiking neurons. However, the high-energy consumption or limited reliability hinders the applications of memristive neurons in neuromorphic computing. In this work, we demonstrated a leaky integrate-and-fire neuron based on an AFeFET. The dynamic relationship between the intrinsic polarization/depolarization process of the Hf 0.2 Zr 0.8 O 2 AFeFET and integrate/leaky neuronal functions are successfully built. The AFeFET neuron features CMOS-compatible, tunable firing frequency, ultra-low hardware cost (no capacitance and additional reset circuit), ultra-low-energy consumption (37 fJ/spike), high endurance (>10 12 ), and high uniformity among different cycles and devices, showing advanced overall performances compared with emerging devices-based neurons in literature. To verify the feasibility of the neuron, we constructed a two-layer SNN combined with FeFET synapses, achieving high recognition accuracy (96.8%), low-energy consumption, and high robustness on MNIST datasets. These results demonstrate that the AFeFET neuron is a promising candidate for constructing high-efficient SNN systems and may promote the industrial landing of neuromorphic machines based on anti-ferroelectric materials."
} | 1,762 |
20486672 | null | s2 | 2,082 | {
"abstract": "During the past decade, the field of structural DNA nanotechnology has grown enormously, not only in the number of its participants but also qualitatively in its capabilities. A number of goals evident in 2001 have been achieved: These include the extension of self-assembled crystalline systems from 2D to 3D and the achievement of 2D algorithmic assembly. A variety of nanoscale walking devices have been developed. A key unanticipated development was the advent of DNA origami, which has vastly expanded the scale of addressable DNA structures. Nanomechanical devices have been incorporated into 2D arrays, and into 2D origami structures, as well, leading to capture systems and to a nanomechanical assembly line. DNA has been used to scaffold non-DNA species, so that one of its key goals has been achieved. Biological replication of DNA nanostructures with simple topologies has also been accomplished. The increase in the number of participants in the enterprise holds great promise for the coming decade."
} | 252 |
36577077 | PMC9910604 | pmc | 2,083 | {
"abstract": "Significance Terpenoids are a class of high-value natural products with wide application in both industrial and human health spaces, but sourcing them from nature or deriving from petrochemicals is no longer sustainable. Microbial biosynthesis of terpenoids has emerged as the most commercially viable option for their large-scale production. However, economically viable titers and productivities are mostly hampered by the limited availability of the main precursors in their biosynthesis, prenol phosphates, in Saccharomyces cerevisiae . Here, we overcome this challenge by establishing the isopentenol utilization pathway–based precursor-forming routes in S. cerevisiae to augment the native mevalonate pathway, circumventing the competition from sterol biosynthesis. Our findings provide a universal and effective yeast platform for the production of diverse terpenoids.",
"discussion": "Discussion Overexpression of endogenous genes and heterologous pathways are often undermined by native regulatory networks and competition with native metabolism. This drawback is especially evident in the synthesis of compounds requiring long and complex pathways for their biosynthesis (e.g., natural products such as terpenoids). In this regard, utilization of orthogonal pathways decoupled from native metabolism has shown promise for improved pathway control and overall performance ( 9 ). This is especially important when the biosynthetic pathway of interest competes with endogenous pathways that are essential for cell growth. In the specific case of terpenoid synthesis, the MVA pathway for IPP/DMAPP synthesis starts from acetyl-CoA derived from central carbon metabolism, thereby competing with other cellular processes for resources, which can complicate attempts to increase terpenoid pathway flux. Furthermore, the MVA pathway requires 2 molecules of NADPH and 3 molecules of ATP per molecule of IPP synthesized. This nontrivial demand for cofactor sharpens competition for cellular resources and often limits the flux achievable by overexpressing endogenous genes ( 43 ). Besides carbon and cofactor limitations, current microbial production of terpenoids also faces challenges associated with highly regulated enzymes. Pathway intermediates or downstream products have been reported to inhibit enzymes in the MVA pathway: acetoacetyl-CoA thiolase can be inhibited by free CoA ( 44 ), HMG-CoA synthase can be inhibited by acetylacetyl-CoA, HMG-CoA, and CoA ( 45 ), HMG-CoA reductase is capable of being inhibited by HMG, free CoA, and NAD(P) + /NADPH ( 46 , 47 ), and IPP, DMAPP, GPP, and FPP have been shown to inhibit mevalonate kinase ( 48 ). Those complex regulations often hinder attempts to up-regulate the MVA pathway for efficient terpenoid synthesis. To overcome these limitations, we here introduced the two-step IUP, which can directly convert isoprenol and prenol to IPP and DMAPP, respectively, into S. cerevisiae . Our results demonstrated that instead of extensively engineering the native pathways, the IUP can serve as a powerful augmentation to the MVA pathway, strengthening the flux toward IPP formation and downstream pathways. As a metabolic pathway decoupled from central carbon metabolism, the IUP not only allowed shortcut access to IPP, circumventing the bottlenecks encountered in the native pathways, but also achieved minimal cofactor requirements. Since the IUP consists of two phosphorylation steps, only two molecules of ATP are required per molecule of IPP synthesized, reducing the competition for NADPH with the decoration of terpene scaffold by cytochrome P450 enzymes. These benefits underline the potential of IUP in yeast metabolic engineering for terpenoid synthesis. In this study, we first demonstrated the utility of IUP in the model industrial microbial host S. cerevisiae . The introduction of IUP led to a significant increase (147-fold) in intracellular IPP/DMAPP, which subsequently translated to increased terpenoid precursors by overexpressing ERG20 and introducing ERG20 F96W-N127W and SaGGPPs, respectively. Further optimization was achieved by designing a cofeeding strategy with both isoprenol and prenol, making use of the desirable characteristics of S. cerevisiae that, in yeast, have similar tolerance to isoprenol and prenol. Its stronger tolerance than that of the other two common hosts namely E. coli and Y. lipolytica also suggests that IUP utilization in S. cerevisiae is more competitive ( SI Appendix , Fig. S1 ). This is also supported by the best performance in S. cerevisiae in terms of content and productivity of IPP/DMAPP augmented by the IUP than the other two organisms harboring ScCK and AtIPK only ( SI Appendix , Table S4 ). In addition, engineering efficient terpene scaffold production in S. cerevisiae is of particular interest because yeast cells have good capability for the functional expression of cytochrome P450 enzymes, which are required for the synthesis of complex terpenoids. Last but not least, the established three-step pathway based on the IUP enabled our strain to achieve a 374-fold increase in the GGPP level compared with the native pathway. As such, this three-step pathway has a promising potential for diterpene and tetraterpene biosynthesis. Overall, our exploration demonstrated a universal and effective platform for supporting plant-derived terpenoid synthesis in S. cerevisiae ."
} | 1,348 |
38453967 | PMC10920631 | pmc | 2,084 | {
"abstract": "Insect antennae facilitate the nuanced detection of vibrations and deflections, and the non-contact perception of magnetic or chemical stimuli, capabilities not found in mammalian skin. Here, we report a neuromorphic antennal sensory system that emulates the structural, functional, and neuronal characteristics of ant antennae. Our system comprises electronic antennae sensor with three-dimensional flexible structures that detects tactile and magnetic stimuli. The integration of artificial synaptic devices adsorbed with solution-processable MoS 2 nanoflakes enables synaptic processing of sensory information. By emulating the architecture of receptor-neuron pathway, our system realizes hardware-level, spatiotemporal perception of tactile contact, surface pattern, and magnetic field (detection limits: 1.3 mN, 50 μm, 9.4 mT). Vibrotactile-perception tasks involving profile and texture classifications were accomplished with high accuracy (> 90%), surpassing human performance in “blind” tactile explorations. Magneto-perception tasks including magnetic navigation and touchless interaction were successfully completed. Our work represents a milestone for neuromorphic sensory systems and biomimetic perceptual intelligence.",
"introduction": "Introduction Biological tactile sensory organs, encompassing skin, whiskers, antennae, among others, have manifested in diverse forms, each possessing unique anatomical structures, sensory functions, and neuronal encoding or processing mechanisms. Mainly, these organs demonstrate proficiency in mechano-sensation related to pressure and vibration, utilizing spatiotemporal encoding methods to interpret somatosensory information obtained from various mechanoreceptors. These intricate systems endow a refined perception of textures, profiles, and shapes during both tactile interaction and active exploration 1 – 3 . Inspired by nature, artificial tactile sensory systems, including e-skin and e-whisker, have been reported, contributing to the ongoing development of skin electronics and epidermal electronics 4 – 7 . State-of-the-art artificial mechanoreceptors further combine neuromorphic devices/circuits and biomimetic tactile sensors to impart processing and memory functions 8 – 11 . However, previous efforts to emulate natural tactile sensory organs and nervous systems mainly focused on the planar, multilayer design of sensor (e-skin) and the multi-directional sensation of force and strain (e-whisker) 6 , 7 , 12 – 15 . These systems mostly rely on mimicking the skin and hair in mammals, imposing limits on their structures and functions. Insects’ tactile sensory organs, despite their diminutive scale and paucity of neurons compared to mammals, exhibit efficient processing and multimodal sensory functions encompassing mechano-perception, magneto-perception, audio-perception, and chemo-perception 2 , 16 , 17 . These capabilities could serve as blueprints for the development of biomimetic sensory platforms. Hallmarked by their segmented, flexible, three-dimensional architecture, insect antennae deliver exceptional mechano-sensory performance in response to deflections and vibrations 18 . Some research grounded in cellular and molecular evidence tentatively hypothesizes that specific insects, such as ants, enable magneto-reception via their antennae’s mechanically sensitive, magnetite-infused magnetoreceptors 19 – 27 . The exquisite antennal structures, densely innervated with sensory receptors and neurons, emit spatiotemporally-encoded neural spike sequences, allowing for the detection of vibrotactile and magnetic stimuli. The perceptual acuity is comparable to, or even surpasses, that of human skin, thereby enabling insects to execute complex tasks, including foraging, object identification, and navigation 2 , 22 , 28 , 29 . Nevertheless, tactile sensory systems inspired by insect antennae (antennal sensillum as well) are yet to be fully explored. It is envisioned that artificial tactile sensory systems, mimicking the structural, functional, and neuronal characteristics of insect antennae, will enable multimodal perception in highly efficient and biologically plausible manners. The development of these insect-inspired systems has the potential to break the design constraints of skin electronics, leading to the realization of tactile intelligence and perceptual augmentation for advanced robotics and human-machine interfaces. In this work, we report a neuromorphic antennal sensory system designed for vibrotactile- and magneto-perception. The electronic-antennae sensor in this system features biomimetic, flexible, three-dimensional (3D) structures. The sensor’s responses to vertical compression, lateral scanning, and magnetic proximity are separately measured in different operation modes to assess its sensing capabilities for pressure, vibration, and magnetic stimuli. The artificial synaptic device in the system, designed with dual planar gates, is fabricated through the liquid-phase adsorption of two-dimensional (2D) nanoflakes of transition metal dichalcogenide (TMD) onto a metal oxide film. Subsequently, the device’s performance in processing spatiotemporal spiking signals is investigated. This system adopts the connection architecture of receptors and neurons and also employs the encoding strategy of fast-adapting (FA; sensitive to dynamic stimulation) and slowly-adapting (SA; sensitive to static stimulation) mechanoreceptors to imitate the neural pathway and neuronal coding observed in biological antennae. Neuromorphic tactile- and magneto-perception experiments, including chess profile classification, Braille code recognition, surface discrimination, magnetic material classification, magnetic navigation, and touchless interfacing, were conducted to validate the system’s potential applications in tactile cognition, sensory robotics, and smart interfaces. Unlike the extensively studied “e-skin” system, our neuromorphic antennal sensory system (“electronic antennae”) incorporates biologically plausible designs that mimic the insect antennae in terms of structure, function, and neural encoding/processing.",
"discussion": "Discussion We present a neuromorphic antennal sensory system utilizing electronic antennae sensor and artificial synaptic device. By emulating the structural, functional, and neuronal features of ant antennae, our system achieves neuromorphic vibrotactile and magneto perception through multifunctional sensing, spike encoding, and synaptic processing. The system follows the labeled-line model of the receptor-neuron pathway, where SA and FA spike trains are separately sent to two synaptic transistors functioning as specialized neurons. This enables event-based, highly efficient, parallel processing of spatiotemporal patterns of sensory stimuli. Our system can operate in different sensing modes that allow contact detection of profile and texture, as well as non-contact detection of magnetic/ferromagnetic material. Applications in profile classification, surface discrimination, material classification, magnetic navigation, and touchless interfacing were demonstrated to validate the cognitive intelligence in tactile and magneto sensations. Distinguishing itself from state-of-the-art artificial sensory systems (both tactile and visual), our system exhibits unique features in system architecture, sensor structure, signal processing, sensory functions, and neuronal characteristics (Table S5 , Table S6 , Supplementary Note 12 ) 6 , 8 , 12 , 13 , 15 , 43 – 45 . In our future work, we aim to integrate a flexible actuator with the sensor to enable antennal movement and active tactile exploration (Supplementary Note 13 ). The insect antennae play fundamental roles in vibrotactile and magneto-perception, serving as a diminutive organ that facilitates active exploration of surfaces and objects through contact-based interactions, while enabling navigation and orientation in non-contact modes. These diverse sensory functions owe their efficiency to the antennal nerve, which is composed of receptor-neuron networks. These networks transmit spatiotemporal spikes of sensory stimuli, effectively deciphering and conveying environmental cues. By imitating the anatomical and neuronal characteristics of insect antennae, our work addresses the challenging issues concerning tactile cognition, contactless perception, and sensory processing, and further expands the capability of mechano-perception to magneto-perception. The system developed here serves as an artificial sensory platform with neuromorphic, insectomorphic characteristics and hardware-level perceptual intelligence. It can be integrated with sensory robots and interactive devices, contributing to the development of augmented perception beyond human senses."
} | 2,188 |
40108655 | PMC11924602 | pmc | 2,085 | {
"abstract": "Background 2,5-Furandicarboxylic acid (FDCA) is a promising building block for biobased recyclable polymers and a platform for other potential biobased chemicals. The common route of its production is by oxidation of sugar-derived 5-hydroxymethylfurfural (HMF). Several reports on biocatalytic oxidation using whole microbial cells or enzymes have been reported, which offers potentially a greener alternative compared to the chemical process. HMF oxidases and aryl alcohol oxidases are the only enzymes able to catalyse the complete oxidation to FDCA, however at low concentrations and are subject to inhibition by the FFCA (5-formylfuran-2-carboxylic acid) intermediate. The present report presents a study on the oxidation of FFCA to FDCA using the obligately aerobic bacterium Gluconobacter oxydans and identification of the enzymes catalyzing the reaction. Results Screening of three different strains showed G. oxydans DSM 50049 to possess the highest FFCA oxidation efficiency. Optimal reaction conditions for obtaining 100% conversion of 10 g/L (71 mM) FFCA to FDCA at 100% reaction yield were at pH 5, 30 °C and using 200 mg wwt /mL cells harvested at mild-exponential phase. In a reaction run at a 1 L scale using a total of 15 g/L (107 mM) FFCA supplied in a fed-batch mode, FDCA was obtained at a yield of 90% in 8.5 h. The product was recovered at 82% overall yield and 99% purity using a simple recovery process. Screening of several oxidoreductase enzymes from the gene sequences identified in the bacterial genome revealed two proteins annotated as membrane-bound aldehyde dehydrogenase (MALDH) and coniferyl aldehyde dehydrogenase (CALDH) to be the enzymes catalyzing the oxidization of FFCA. Conclusion The study shows G. oxydans DSM 50049 and its enzymes to be promising biocatalysts for use in the FDCA production process from biomass. The high reaction rate and yield motivate further studies on characterization of the identified enzymes exhibiting the FFCA oxidizing activity, which can be used to construct an enzyme cascade together e.g. with HMF oxidase or aryl alcohol oxidase for one-pot production of FDCA from 5-HMF. Supplementary Information The online version contains supplementary material available at 10.1186/s12934-025-02689-x.",
"conclusion": "Conclusions Through this study, we show G. oxydans DSM 50049 to be a promising source of oxidative enzymes for furan transformations under mild conditions. The wild-type bacteria could be used directly for selective and efficient production of FDCA from FFCA, and the enzymes involved in the oxidation were identified. We noted that the FFCA oxidation activity was not common to all the G. oxydans strains under the conditions used. Although the biocatalytic activity was sensitive to inhibition by FFCA, FDCA, and low pH, fed-batch bioconversion with pH control and sufficient aeration provided a sufficiently efficient process for FDCA production. Further enhancement can be achieved by improving the tolerance of the bacterium to the furan derivatives, incorporating a cofactor regeneration system, and designing the process to enable in situ product removal. The low aqueous solubility of FDCA, particularly at low pH, provided a simple route for its recovery at high overall yield and purity. The wild-type G. oxydans DSM 50049 lacks the enzyme(s) able to catalyze oxidation of the hydroxyl group linked to the furan ring and hence does not allow the production of FDCA directly from 5-HMF, which is oxidized instead to HMFCA and not further to any other oxidation product [ 35 ]. The identified enzymes in G. oxydans DSM 50049 exhibiting the FFCA oxidizing activity will enable the construction of an enzyme cascade together e.g. with HMF oxidase or aryl alcohol oxidase for one-pot production of FDCA from 5-HMF in a heterologous host [ 24 , 27 ]. Alternatively, FDCA production by the G. oxydans can be achieved by engineering the cells with a gene encoding the enzyme oxidizing the hydroxyl group in HMFCA, e.g. HMF/furfural oxidoreductase (hmfH) (to be reported in another study) [ 48 ].",
"discussion": "Results and discussion Oxidation of FFCA to FDCA using G. oxydans Initial experiments were performed to screen different G. oxydans strains, DSM 2003, 2343, and 50049, for their ability to oxidize FFCA to FDCA (Fig. 1 A and S1A-C). The cells were grown in a glycerol medium that seems to have an activating effect on the expression of the sugar/polyol oxidizing enzymes [ 24 ]. The reaction against 5 mg/mL FFCA in 0.1 M acetate buffer pH 5 at 30 °C using the cells (collected from 4 mL culture and equivalent to 52 mg wet weight/mL of the reaction volume) showed that the strains 2003 and 50049 produced FDCA at a reaction yield > 90%, 100% selectivity and productivity of 0.2 g/L·h compared to G. oxydans DSM 2343 that gave only ~5% reaction yield (Fig. 1 A and Figure S1 ). G. oxydans DSM 50049 cells were used for further investigations because not only was its activity slightly higher than that of DSM 2003 (Figure S1 ), but the strain also exhibited a shorter lag phase during cell cultivation as compared to the other strain. Screening of the optimal reaction parameters showed that G. oxydans DSM 50049 harvested at mid-exponential phase, i.e. 16 h of growth (Fig. 1 B and S2), exhibited the highest activity in the reaction with 5 mg/mL FFCA at pH 5 to give FDCA with 100% reaction yield (Fig. 1 C and S3). \n Fig. 1 Optimization of different reaction parameters for oxidation of FFCA to FDCA catalyzed by resting Gluconobacter oxydans DSM 50049 cells in 0.1 M sodium acetate buffer pH 5 at 30 °C. The parameters included ( A ) G. oxydans strains (DSM 2003, DSM 2343, and DSM 50049), ( B ) cell cultivation time, ( C ) pH value, and ( D ) cell amount. The experiments for B, C and D were performed only with G. oxydans DSM 50049. The FFCA concentration used was 5 mg/mL in A, B and C, and 10 mg/mL in D \n Increasing the cell amount from 52 to 196 mg wwt/mL (corresponding to the culture volumes of 4 to 32 mL) and FFCA concentration to 8 mg/mL showed 100% conversion of FFCA, and > 80% yield of FDCA (Fig. 1 D & S4). No other coproduct was observed; the lower reaction yield may be attributed to the enzyme inhibition and/or the lower aqueous solubility of FDCA [ 36 ]. The results obtained are in agreement with our earlier observations on the DSM 50049 strain oxidizing only the aldehyde group in 5-HMF to form HMFCA [ 35 ]. The organism does not exhibit any activity against the hydroxyl group on the furan ring. The inhibitory effect of FFCA on the catalytic activity of the G. oxydans cells was clearly observed at FFCA concentration of 10 mg/mL (at a cell concentration of 52 mg wwt/mL) when only 60% of the substrate was converted after 6 h of the reaction, and no more conversion was observed even with longer incubation up to 24 h (Fig. 2 A and S5). No FDCA formation was noted with a further increase in the FFCA concentration to 15–20 mg/mL (Fig. 2 A). According to earlier reports, FFCA concentration above 15 mM (2 mg/mL) has an inhibitory effect on fungal aryl alcohol oxidases (AAO) and HMF oxidase [ 15 , 24 ]. Hence, G. oxydans and its enzymes seem to exhibit higher resistance to inhibition since FFCA could be used up to a concentration of 71 mM (10 mg/mL). Increasing the cell concentration for the reaction could overcome the inhibition to a certain extent (Fig. 1 D). \n Fig. 2 Effect of different concentrations of ( A ) FFCA and ( B ) FDCA on the oxidation activity of G. oxydans DSM 50049 against FFCA in 0.1 M acetate buffer pH 5 at 30 ºC. The amount of cells used for the reactions was 52 mg wet weight/mL, and in case of ( B ) the FFCA concentration used was 5 mg/mL \n The inhibitory effect of the product was also studied by including different concentrations of FDCA in the reaction with FFCA. It is important to note that FDCA has a limited solubility in aqueous solution (1.72 g/L at 313.5 K). Nevertheless, Fig. 2 B and S6 show a slight inhibitory effect on the oxidative activity of G. oxydans DSM 50049 when the FDCA concentration was increased to 5 mg/mL, resulting in ~ 80% conversion of FFCA. Further increase in FFCA concentration to 7.5 and 10 mg/mL FDCA led to significant inhibition, and 32 and ~ 61% of FFCA remained unconverted, respectively, even after 24 h of reaction (Fig. 2 B). This inhibition was ascribed to the decrease in pH of the reaction (to pH 4.3) due to the accumulation of FDCA. A similar observation was made by Wang et al. (2020) on inhibition of vanillin dehydrogenase activity used as a biocatalyst for 5-HMF oxidation to FDCA, and the inhibition was significantly reduced by neutralizing the pH [ 37 , 38 ]. Fed-batch biotransformation of FFCA to FDCA using G. oxydans DSM 50049 To overcome the inhibitory effect of FFCA, fed-batch biotransformation was performed in a flask using a feed of 5 mg/mL FFCA in 50 mL acetate buffer pH 5 and without controlling pH and aeration. It took 9 h to achieve 90% FFCA conversion after the first feed and 12 h after the second feed when FDCA concentration reached around 8 g/L giving rise to increased acidity. A dramatic loss in cell activity was seen during the third feed, resulting in less than 10% FFCA conversion after 24 h (Fig. 3 A). Only 8.1 g out of 15 g FFCA per liter added to the reaction, was converted to 9 g/L FDCA by the end of the experiment. \n Fig. 3 Fed-batch oxidation of FFCA (♦) to FDCA (▲) using resting G. oxydans DSM 50049 cells at 30 ºC in: ( A ) 50 mL 0.1 M acetate buffer pH 5 in 250 mL shake flask with uncontrolled pH and aeration, and ( B ) 1 L reaction volume in 3 L bioreactor maintained at pH 5 and 70% dissolved oxygen. The reaction was fed with FFCA stock solution to achieve the initial concentration of 5 mg/mL \n Subsequently, the fed-batch experiment was done in 1 L reaction volume with pH controlled at 5 and dissolved oxygen at 70%. As seen in Fig. 4 B, FFCA in feeds 1 and 2 was converted efficiently to FDCA at 100% yield and selectivity within only 2 h, while it took 3 h for complete conversion of the last 5 g/L FFCA in the third feed (Fig. 3 B). Overall, around 15 g/L FDCA with around 90% total reaction yield was obtained from 15 g/L FFCA within 8.5 h. Comparison of the formation of FDCA over time in the first batch shows that the initial activity of G. oxydans is 5-fold higher under controlled conditions of pH and aeration applied in the bioreactor compared to that under uncontrolled conditions (Fig. 3 A and B) [ 37 , 38 ]. The productivity during the first batch was 2.8 g/L·h compared to only 0.5 g/L·h FDCA obtained from the first batch in the shake flask. However, the drop in the overall productivity to 1.8 g/L·h in the controlled reactor and 0.2 g/L.h in the shake flask is due to the inhibitory effect of both product and substrate as described above, and the low solubility of FDCA could account to some extent for the observed drop in the reaction yield after the third feed. The decrease in activity may also be ascribed to insufficient regeneration of the cofactor required by the enzyme(s) catalyzing the reaction [ 39 ]. Hence, identifying the enzyme(s) involved in the oxidation process will provide valuable insights for optimizing biocatalyst and reactor design in future studies. \n Fig. 4 Analysis of the purification of FDCA product: ( A ) HPLC chromatograms of the final reaction solution obtained by G. oxydans DSM 50049 catalyzed oxidation of FFCA (1), FDCA recovered by extraction in ethyl acetate (2), FDCA remaining in the aqueous phase (3), and standard FDCA (4); ( B ) a picture of the purified FDCA, and ( C ) 1 H-NMR of the purified FDCA, indicating > 99% purity \n Recovery and purification of FDCA Since FFCA is converted selectively and nearly quantitively to FDCA, the product recovery and purification from the reaction solution at high purity was successfully achieved via a simple procedure involving first increasing the pH to 9 to ensure the solubilization of the product, centrifugation for separating the insoluble material including cells, filtering the supernatant to remove any remaining particulate matter before lowering the pH to 1.5. Around 80% of the FDCA was precipitated out from the solution, and > 70% of the remaining FDCA in the supernatant was recovered by liquid/liquid extraction (Table 1 ; Fig. 4 ). The final FDCA product was obtained with an overall yield of 82%, which was higher than that (76%) reported by Koopman et al. (2010) during the purification of FDCA from the reaction medium [ 40 ] and can be further improved to 87% by avoiding the washing step (Table 1 ). The purity of the recovered FDCA exceeded 99%, as verified by HPLC and proton NMR (Fig. 4 A and C), indicating the high product quality for its use, especially in polymer synthesis. \n Table 1 Purification of FDCA from the final reaction solution obtained from the fed-batch process of FFCA oxidation in 1 L reaction volume using the resting cells of G. oxydans DSM 50049. The solution contained 15 g/L of FDCA and no FFCA Purification step FDCA step yield (%) FDCA overall yield (%) Reaction solution 100 100 Cell removal 96.15 96.14 Concentration and precipitation 90.5 87 Washing and drying 94.3 82 \n Identification of FFCA oxidizing enzyme(s) The potential enzymes involved in the oxidation of FFCA to FDCA were screened from the G. oxydans DSM 50049 genome based on the conserved amino acid residues that are commonly present in the NAD(P) + , FAD, and PQQ-dependent oxidoreductases [ 40 ]. Over 100 genes encoding for oxidoreductases were identified, annotated, and classified based on the theoretical isoelectric point (pI) of the encoded protein and the cofactor identified for each annotated enzyme. Knowing from the literature that the G. oxydans whole cells carry out the oxidation reactions at pH in the acidic range and the cofactor dependence of enzymes catalyzing similar reactions, fourteen genes encoding oxidoreductases were selected (Table 2 ), amplified from the G. oxydans DSM 50049 genome and cloned into proper vectors. The nucleotide sequences of the selected genes and their corresponding amino acid sequences are shown in Figure S7. Seven constructs annotated as FAD-dependent oxidoreductase (39.9 kDa), alcohol dehydrogenase (ADH) (51.1 kDa), membrane-bound aldehyde dehydrogenase (MALDH) (83.1 kDa), xanthine dehydrogenase (XDH) (25.3 kDa), coniferyl aldehyde dehydrogenase (CALDH) (31.3 kDa), cyclohexadienyl dehydrogenase (CHDH) (24.1 kDa), and aldehyde oxidoreductase iron-sulfur binding (ALOD-Fe/SB) (54.8 kDa) were so far successfully cloned and transformed into different E. coli expression strains (BL21(DE3) and CodonPlus) for protein production (Table S3). Five of the seven proteins including MALDH, FAD-dependent oxidoreductase, ADH, CHDH, and ALOD-Fe/SB were successfully expressed in E. coli BL21 (DE3) grown in LB medium at 16 °C and induction with 0.5 mM IPTG. The expression of the other two proteins, i.e. xanthine dehydrogenase and coniferyl aldehyde dehydrogenase (CALDH) was only possible in E. coli CodonPlus grown in LB (and induced with 0.5 mM IPTG) and in autoinduction medium, respectively (Table S3, Figure S8A-C). Screening of enzyme activities using the whole recombinant E. coli cells showed only two sets of cells - expressing the enzymes annotated as MALDH and CALDH, respectively – to display activity against FFCA to form FDCA. \n Table 2 Fourteen genes are selected from G. oxydans based on pI values, and their conserved amino acid residues. Names, annotations and expected sizes are indicat No. Gene Name Annotation Theoretical pI Anticipated cofactor Size (bp) 1 ddmA1 FAD dependent oxidoreductase 5.1 FAD 887 2 adhB2 Alcohol dehydrogenase NAD + 1461 3 GO50049LU-2_1_00206 GMC family oxidoreductase 8.5 FAD 1227 4 GO50049LU-2_1_02677 Membrane-bound aldehyde dehydrogenase 8.8 FAD molybdopterin guanine dinucleotide [4Fe-4 S] cluster 2334 5-1 sldA_1 Glycerol dehydrogenase large subunit 6.6 PQQ 321 5-2 sldA_2 1659 6 ctcP FAD-dependent oxidoreductase 8.5 FAD 903 7 rfbD FAD-dependent oxidoreductase 6.2 FAD 1164 8 GO50049LU-2_1_02764 Xanthine dehydrogenase 5.1 NAD + 461 9 GO50049LU-2_1_02763 294 10 GO50049LU-2_1_01181 GMP reductase 4.1 NAD(P) + 377 11 adhA1 Alcohol dehydrogenase 6.6 NAD + 750 12 calB Coniferyl aldehyde dehydrogenase 9.6 FAD 879 13 GO50049LU-2_1_01307 HMF oxidase 8.8 FAD 2072 14 tyrC Cyclohexadienyl dehydrogenase 5.0 NAD + 678 15 paoA Aldehyde oxidoreductase iron-sulfur-binding 6.9 FAD [4Fe-4 S] cluster 531 16 GO50049LU-2_1_00575 116 17 paoB 935 \n The membrane-bound aldehyde dehydrogenases are mostly PQQ-dependent enzymes [ 41 ]. The G. oxydans DSM 50049 MALDH has the conserved amino acid residues (GxGxxG) common for flavoproteins and was identified with 66.6% identity to an uncharacterized pyrroloquinoline quinone dependent MALDH (using non-redundant UniProtKB/SwissProt sequences database) (Figure S9A) [ 42 ]. However, when the NCBI Protein Reference Sequences database was used, a molybdopterin-dependent oxidoreductase was identified with 100% identity [ 43 ]. The latter enzyme is a dimer of heterotrimer comprising 88.7 kDa molybdoprotein large subunit, 30.2 kDa flavoprotein medium subunit, and 17.8 kDa iron-sulphur domain small subunit [ 44 ]. On the other hand, G. oxydans DSM 50049 MALDH comprised only 83.1 kDa protein domain. Experimentally, the E. coli cells with overexpressed MALDH showed activity in the presence of FAD, and the addition of sodium molybdate to the culture during protein expression enhanced the enzyme activity against FFCA (data not shown). This result fits with the blast search and conserved motif indicating that the enzyme is molybdopterin-dependent rather than PQQ-dependent. The CALDH enzymes are NAD + -dependent oxidases that oxidize coniferyl aldehyde to ferulic acid [ 45 ]. The closest related enzyme to CALDH is a coniferyl aldehyde dehydrogenase from Pseudomonas sp. HR199 (sequence ID: O86447.3) [ 45 ] with a sequence coverage of 95% and sequence identity of 46.3% (Figure S9B). Interestingly, both MALDH and CALDH align, although with only 25% and 28% sequence identity, with the molybdenum-dependent periplasmic oxidoreductase and a cytoplasmic dehydrogenase enzyme, respectively, in Pseudomonas taiwanensis VLB120 and Pseudomonas putida KT2440 that were identified recently by extended gene deletions to be involved in oxidative detoxification of HMF [ 46 ]. The pterin cytosine dinucleotide (MCN) and dioxothiomolybdenum (VI) (MOS) binding sites were found to be conserved in MALDH and the periplasmic oxidoreductase as in the crystal structure of an molybdoenzyme (aldehyde oxidase) from E. coli (PDB:5G5G) (Figure S10) [ 47 ]. Similarly, the active site residue and NAD + binding site in CALDH and the Pseudomonas cytoplasmic dehydrogenase were conserved and matched those in the crystal structure of a related aldehyde dehydrogenase from Bacillus cereus (PDB: 5GTK) (Figure S11). This provides additional evidence of the role and cofactor dependence of the enzymes identified from G. oxydans DSM 50049. The recombinant E. coli cells expressing MALDH exhibited activity against FFCA, however, on cell lysis, only the insoluble fraction showed activity (Fig. 5 A, C). This observation may confirm that MALDH is a membrane bound enzyme that includes a signal peptide (data not shown), and hence the activity is located in the insoluble fraction (containing membrane) of the cell lysate. On the other hand, the CALDH activity resided in the soluble fraction of the cell lysate (Fig. 5 B, C). Activity measurements using the same concentration of the recombinant E. coli expressing the two enzymes showed that MALDH-containing cells exhibited 2.7 times higher activity against 0.5 mg/mL FFCA compared to CALDH (Fig. 5 C). This could partly be due to a possible higher level of MALDH expression, however, CALDH exhibited significantly higher activity when FFCA concentration was increased to 5 mg/mL, giving 100% conversion of FFCA to FDCA compared to only 6% conversion with MALDH (Fig. 5DI and II). The enzymes need to be investigated in more detail to provide deeper insight into the enzyme structure, function, and their roles in the oxidation of furans. \n Fig. 5 Oxidation of FFCA to FDCA using recombinant E. coli (lysate and whole cells) bearing G. oxydans DSM 50049 membrane-bound aldehyde dehydrogenase (MALDH) and coniferyl aldehyde dehydrogenase (CALDH), respectively. ( A ) HPLC analysis of samples from the reaction catalysed by the insoluble fraction of the cells expressing MALDH (green chromatogram), in comparison with the reaction using negative control E. coli lysate (black chromatogram), ( B ) HPLC analysis of samples from the reaction performed with the soluble fraction of CALDH containing cells (green chromatogram), in comparison with control reaction using negative control E. coli lysate (black chromatogram), ( C ) FDCA concentration after 24 h reactions catalysed by 1.94 mg cell dry weight/mL of the whole cells overexpressing MALDH and CALDH, respectively, against 0.5 mg/mL FFCA and the corresponding initial reaction rates, ( D ) Activity of MALDH (DI) and CALDH (DII) against different concentrations of 0.5, 1, 2 and 5 g/L of FFCA"
} | 5,323 |
37687353 | PMC10489935 | pmc | 2,086 | {
"abstract": "Arbuscular mycorrhizal fungi (AMF) form symbiotic relationships with the roots of nearly all land-dwelling plants, increasing growth and productivity, especially during abiotic stress. AMF improves plant development by improving nutrient acquisition, such as phosphorus, water, and mineral uptake. AMF improves plant tolerance and resilience to abiotic stressors such as drought, salt, and heavy metal toxicity. These benefits come from the arbuscular mycorrhizal interface, which lets fungal and plant partners exchange nutrients, signalling molecules, and protective chemical compounds. Plants’ antioxidant defence systems, osmotic adjustment, and hormone regulation are also affected by AMF infestation. These responses promote plant performance, photosynthetic efficiency, and biomass production in abiotic stress conditions. As a result of its positive effects on soil structure, nutrient cycling, and carbon sequestration, AMF contributes to the maintenance of resilient ecosystems. The effects of AMFs on plant growth and ecological stability are species- and environment-specific. AMF’s growth-regulating, productivity-enhancing role in abiotic stress alleviation under abiotic stress is reviewed. More research is needed to understand the molecular mechanisms that drive AMF-plant interactions and their responses to abiotic stresses. AMF triggers plants’ morphological, physiological, and molecular responses to abiotic stress. Water and nutrient acquisition, plant development, and abiotic stress tolerance are improved by arbuscular mycorrhizal symbiosis. In plants, AMF colonization modulates antioxidant defense mechanisms, osmotic adjustment, and hormonal regulation. These responses promote plant performance, photosynthetic efficiency, and biomass production in abiotic stress circumstances. AMF-mediated effects are also enhanced by essential oils (EOs), superoxide dismutase (SOD), peroxidase (POD), ascorbate peroxidase (APX), hydrogen peroxide (H 2 O 2 ), malondialdehyde (MDA), and phosphorus (P). Understanding how AMF increases plant adaptation and reduces abiotic stress will help sustain agriculture, ecosystem management, and climate change mitigation. Arbuscular mycorrhizal fungi (AMF) have gained prominence in agriculture due to their multifaceted roles in promoting plant health and productivity. This review delves into how AMF influences plant growth and nutrient absorption, especially under challenging environmental conditions. We further explore the extent to which AMF bolsters plant resilience and growth during stress.",
"conclusion": "4. Conclusions Most study papers have already provided evidence of the positive effect AMF has in enhancing plant growth in conditions considered stressful. As a result, the current material related to the role of AMF has been combined cohesively in this review to understand the symbiotic relationship that AMF has with various plants when exposed to stressful situations. In the past, the AMF has been primarily discussed as a beneficial entity for the uptake of nutrients from the soil. However, in more recent times, it has been depicted that plants inoculated with AMF can effectively combat various environmental cues, such as salinity, drought, nutrient stress, alkali stress, cold stress, and extreme temperatures, and can therefore help increase the yield per hectare of a wide variety of crops and vegetables. Arbuscular mycorrhizal fungi (AMF) can increase the absorption of nutrients such as phosphorus, nitrogen, and zinc and increase the availability of some essential micronutrients. By colonizing the soil, AMF can improve the soil structure and boost the activity of beneficial microorganisms, which can help plants become more resistant to drought and disease. Therefore, enhancing the colonization of AMF in the soil can help to improve crop productivity and yield in the future. By identifying and improving the characteristics of AMF accessibility, functionality, and climate resilience in new cultivars, farmers can maximize their productivity and increase the sustainability of their production. This will also ensure food production, which is essential for global food security. By integrating this system, we can significantly reduce the amount of energy and artificial input necessary while enhancing the effectiveness of beneficial organisms and the sustainability of agricultural systems. In addition to improving yields, this method can significantly reduce crop losses due to pests, diseases, and extreme weather events since it optimizes using natural resources, such as water and nutrients. Many researchers believe that by administering mycorrhiza to plants, they can absorb more soil nutrients and water. In addition, they will be able to withstand diseases and stressors better since they can absorb more nutrients and water. As a result, reducing fertilizer and other agricultural inputs contributes to enhanced sustainability and cost-effectiveness by concurrently improving crop yields and quality. As well as providing essential nutrients for crop production, such as by adding nitrogen, phosphorus, and potassium to the soil, AMF also retains moisture in the soil, which is crucial to the growth of crops. Aside from these benefits, there are also other benefits, including the reduction of erosion, the reduction of weeds, and the improvement of the soil’s organic matter level. These benefits will, in turn, likely lead to increased agricultural productivity and an improved environment due to all of these benefits. Because of a symbiotic relationship between mycorrhizal fungi and plants, the plants have an easier time accessing nutrients and water. By doing this, the plants can grow faster and healthier and reduce toxic elements in the soil, which may result in higher yields, improved soil quality, and a more stable ecosystem.",
"introduction": "1. Introduction Industrialization, urbanization, and globalization are shrinking the arable lands and declining agricultural production, leading to increased food demand for the rapidly growing population. In addition, changing climatic conditions have resulted in extreme weather conditions that ultimately lead to more droughts, high temperatures, and floods, affecting the food supply from agricultural systems [ 1 ]. There are several reasons, including that export crops often have a higher economic value. In contrast, biofuel crops offer a substitute for conventional fossil fuels as a source of energy. Furthermore, these crops are often more resilient to changing climates, making them a viable option in some areas [ 2 ]. These consequences have had far-reaching implications for our planet, with the effects of climate change becoming more pronounced with each passing day [ 3 ]. Sustainable agriculture, or agricultural systems that bind the use of natural resources to produce food in a way that minimizes the adverse effects of the production process on the environment is crucial to achieving this balance [ 4 ]. It is becoming increasingly clear that soil is an integral part of food production and a critical resource that must be managed carefully to ensure long-term economic and environmental sustainability [ 5 , 6 ]. A comprehensive overview of what we currently know about AMF and how it may be used to improve crop yields under both optimal and stressed conditions will be provided as part of our study. As part of our discussion, AMF was discussed as a way to improve crop growth. By secreting more enzymes, bacteria facilitate nitrogen uptake by plants. However, there is an issue since oxygen destroys nitrogen-free enzymes. To combat this issue, bacteria interact with plants to develop a nodule root structure [ 7 , 8 , 9 ]. Many studies demonstrated the beneficial effects of AMF on crop yields, which will be discussed in this review with an in-depth analysis. These symbiotic relationships allow plants to obtain nutrients from the unavailable soil. In addition to boosting plant growth and well-being, fungi absorb phosphorus and other essential minerals from the soil. Due to the release of enzymes by the fungus, the plant can gain access to nutrients that would otherwise be inaccessible to it. Moreover, the fungal part releases several hormones that aid the plant in developing an extensive root system. In exchange, the plant provides the fungus with sugar, the main component for energy and survival [ 10 ]. Agroecosystems are designed to minimize the number of external inputs, such as pesticides, chemical fertilizers, and water, while maximizing natural resources, such as sunlight and soil nutrients. This makes them more sustainable than traditional farming methods [ 11 ]. For instance, beneficial bacteria can fix nitrogen from the atmosphere, make it available to plants, and suppress soil-borne diseases [ 12 ]. They establish a mutualistic relationship with plant roots, providing essential minerals and water from the soil in exchange for photo-synthetically fixed carbon from the plant [ 13 ]. Furthermore, these fungi are also known to provide critical benefits to the plants they associate with, such as improved water and nutrient uptake, increased resistance against diseases [ 10 ], and nutrient mobilization from organic substrates [ 14 ]; these interactions can lead to dramatic changes in the composition, structure, and functioning of plant communities, and a comprehensive understanding of these processes is essential for successful ecosystem conservation and management [ 15 , 16 , 17 , 18 ]. The imbalance of beneficial microorganisms caused by the disruption in the microbial population has led to soil deterioration and decreased crop production. Alterations in the composition of microbial communities can also result in changes in soil nutrient cycling and a reduction in soil water retention, leading to further soil degradation [ 19 ]. Microbes are essential for sustainable soil fertility management in nutrient cycling, pest management, and soil structure. This study aims to draw attention to the significant role that endomycorrhiza symbiosis can play as a provider of ecosystem services to ensure crop yield and assist in developing sustainable agriculture systems. Consequently, the general goal of this study is to review the work of AM growths on agricultural yield efficiency and biological system administration. Hence, understanding microbial interactions and how they interact with plants is critical to developing sustainable management of soil fertility and crop production ( Figure 1 ) [ 20 ]. In return, the microbes receive a continuous supply of nutrients, allowing them to flourish and provide a healthier environment for the proper growth of plants [ 21 ]; for instance, plant roots often provide nitrogen and other essential minerals to their associated fungi, while the fungi provide the plants with the essential nutrients and water [ 22 ]. Moreover, AMF can also help plants resist environmental stress, such as drought and soil salinity, resulting in higher levels of plant growth and an increased rate of survival; by making the most of these technologies, smallholder farmers can not only maximize the yields of their crops but also enhance the sustainability of their farming practices for future generations [ 23 ] ( Table 1 ). Arbuscules, internal fungal structures in the root cortical cells, allow arbuscular mycorrhizal (AM) fungi to form strong relationships with a host plant [ 50 ]. According to current estimates, AM fungus started cooperating with host plants between 400 and 480 million years ago, facilitating the first terrestrial plant colonization of the land [ 51 ]. Approximately eighty per cent of terrestrial plant species are in relationships of close symbiosis with AM fungus [ 52 ] for various factors that benefit plants, such as nutrient acquisition, crop mass, yield increases, and reduced stress from abiotic pressures. AMF is a crucial and helpful gathering of soil accumulation that may significantly increase crop efficiency and ecological continuity in the production methods of new plants [ 53 ]. Endomycorrhiza fuofs allows the start of a mutualism relationship along with the root structure of eighty per cent of plant families; it just does not better the development of plants via enhanced absorption of phosphorus (P) available in the soil and other on-labile mineral nutrients necessary for the development of the plant; it also has ‘unhealthy’ effects on maintaining the collected soil, intended to stop erosion, and overcomes stress in plants due to abiotic and biotic factors [ 54 , 55 ]. AM fungi’s positive effects on plant execution and soil well-being are fundamental for agricultural ecosystems to be managed sustainably [ 56 ]. 1.1. Economic Importance of Soil Microbes However, as the “first green revolution,” beneficial soil microbes, generally, and AM in particular, have received significantly less attention [ 57 , 58 , 59 , 60 ]. Even though some offerings stand outside the marketplace and are hard to measure, the lowest approximate equivalent or outstrip worldwide gross countrywide outcomes [ 61 ]; a price tag of USD 190 billion has been calculated based on the value of goods and services provided by nature, including clean water, fertile soils, and pollination. This is double the worth of the gross national items of the world. According to present research, two critical environmental offerings, ‘formation of the soil’ and ‘cycling of nutrients, were expected to highlight USD 17.1 and 2.3 trillion US dollars. Simultaneously, most nations consume their charge frameworks to save the climate by limiting the number of exercises involved in the pollution they allow (such as the carbon tax) or to motivate the growth of rules that are in favour of the environment (Ecological Tax Reform); Costa Rica is the first nation to made a countrywide attempt to safe atmosphere offerings [ 62 ]. In 1996, this nation followed the rule (Forestry Law No. 7575), spotting four vital services supplied through the public woods: carbon elimination, hydrological offerings, biodiversity security, and attractive charm [ 63 ]. The rule set up a structure for the fee for atmosphere offerings, as outlined in a program authorized by the pagos por service Ambien (PSA) and managed through the fund provided by the National Forestry (FONAFIFO), which includes landlords and all other succeeding customers of the earth who agreed to charge environment offerings for twenty years and offer opposition through replanting, continuous management, conservation, and strategies of rebirth [ 64 ]. The availability of farming goods and environmental services is essential to human existence and the standard of life. Nevertheless, new farming techniques that have significantly enhanced the world’s food contribution have unintentionally negatively affected the environment and resources [ 65 , 66 ]. In the context of advancement, new and beneficial procedures are needed to operate the Earth’s surrounding services and counter the lack of effort to hold necessary outcomes for renewable food manufacturing in the face of the enhancing populace worldwide [ 67 ]. Horticulture is the most significant connection between people and the climate; accommodating yield creation and natural respectability, reasonable harvest, and crop production is difficult for agribusiness and predetermination for landowners [ 68 ]. This demonstrates extending yield executives methodologies that enhance soil fertility, organic assortment, and crop production through developing agroecosystems that favour standard ecological methods and support efficiency in the long term [ 69 ]. Ecological administrations sustain soil attributes and plant well-being; soil flexibility in this environment is incredibly relevant [ 70 ]. Specifically, soil microorganisms that structure cooperative interactions with the roots of plants have attracted growing attention in rural examination and improvement since they provide a natural substitute to stimulate plant development and lower contributions to feasible editing arrangements [ 71 ]. The omnipresence of arbuscular mycorrhizal growths at the point of interaction among soil and plant roots makes them an essential and valuable association of soil biota. Their wholesome and non-dietary attributes significantly affect surrounding methods promoting farming yield creation and agroecological environmental administrations [ 72 , 73 ]. The proper management of environmental services provided by AM will positively influence the utilization and conservation of natural resources for the benefit of human societies [ 74 ]. Biomolecules have a backbone made of carbon, which is a necessary component of living organisms. Nevertheless, too much CO 2 in the atmosphere is thought to be dangerous. Thus, it is involved among the primary greenhouse gases [ 75 ]. Carbon sequestration captures carbon dioxide to lower atmospheric carbon dioxide levels [ 76 ]. The amount of carbon dioxide in the atmosphere directly impacts global climate changes, seriously threatening the entire biosphere [ 77 ]. The amount of CO 2 has steadily risen in the atmosphere, with an average of approximately 385 ppm [ 78 ]. Some research teams have been performing evaluations to determine the most effective carbon sequestration technologies for a few decades. Two primary geological and biological carbon sequestration processes have been documented [ 79 , 80 ]. One of the critical processes of biological carbon sequestration is the assimilation of atmospheric CO 2 through the biological activity known as photosynthesis. To reduce atmospheric CO 2 levels, microbe-mediated CO 2 uptake in the soil and plants is a crucial problem. The increase in photosynthetic rate is caused by bacterial populations in the leaf endosphere, phyllo sphere, and rhizosphere, despite photosynthesis being a natural process that consumes atmospheric CO 2 [ 81 ]. Similar to how mammals develop symbiotic connections with microorganisms and have probiotics, plants have a variety of microbial communities and do the same [ 82 ]. 1.2. Molecular Mechanisms of Plant-Microbe Interactions Microbial communities linked with plants impact the operation of intricate bio-networks, such as complete food chains in many ecosystems [ 83 ]. Entophytic and endophytic interactions between plants and microorganisms have generally been discovered. However, these interactions are context-dependent, bidirectional, and complicated [ 84 ]. Despite some understanding of the relationship between plants and bacteria, scientific research has restricted attempts to use microbes as bio-stimulants or biofertilizers. Plant-associated microorganisms’ different molecular pathways to increase photosynthetic rate are being investigated [ 85 ]. By supplying specific bioactive chemicals, endophytes colonized in the interior tissues of plant parts help improve plant growth, yield, and disease resistance [ 86 ]. In contrast, the host plant provides carbon substrates to the microorganisms. Interestingly, a certain microbial strain prefers to colonize the rhizosphere when atmospheric CO 2 levels are rising over those of other strains [ 87 ]. Improving carbon absorption by plants that rely on particular microbial strains would also help lower atmospheric CO 2 . For instance, the bacterial strain Pseudomonas simiae colonizes the rhizosphere much more quickly than the strain P. putida when the CO 2 concentration is high [ 88 ]. In addition, well-documented microbe-mediation increases in photosynthetic rate in the tissues of plant leaves. The rhizosphere’s microbial population is solely reliant on plant signals. In addition to their fundamental tasks, plants’ roots are involved in various additional systems. Organic macromolecules called “root exudates” are discharged into the rhizosphere by roots, making up 5 to 40% of the plant’s photosynthetically absorbed carbon [ 89 ]. For instance, the root exudates contain substances like citric acid, fumaric acid, flavonoids, tryptophan, etc., that function as signalling molecules and draw advantageous microorganisms to the rhizosphere [ 90 ]. Under high CO 2 conditions, plant roots release root exudates that attract beneficial microbes to colonize in the rhizosphere or within the plant tissues [ 88 ]. Similarly, when plants’ carbon sinks, such as respiration, growth, and storage, require them to store photosynthetically fixed carbon, they enlist the help of beneficial bacteria to improve the rate at which they fix carbon through photosynthesis [ 87 ]. When bacteria colonize the root system, root exudates play a role in the selection processes that favor certain bacteria over others. The photosynthetic pigments chlorophyll and carotenoids are the most abundant in the plant’s aerial portions, and it has been suggested that bacteria work as bio-stimulants to increase this concentration. By increasing the chlorophyll content, plant growth-promoting rhizobacteria (PGPR), Enterobacter sp., considerably accelerates the development of the Okra plant overall [ 91 ]. Similar advantages can be attained by intentionally introducing fungal endophytes into the plants. For instance, adding fungal endophytes to Coleus forskolin plants increased their yield by increasing their total photosynthetic rate [ 92 ]. Microbes increase chlorophyll content to increase the photosynthetic rate even under stress circumstances caused by metal elements. Lycopersicon esculentum, a plant growing under cadmium (Cd) stress, has a dramatically higher total chlorophyll content after being inoculated with P. aeruginosa or Burkholderia gladioli, according to Khanna et al.’s (2019) analysis. Microbes connected with plants can increase the expression of genes involved in photosynthesis and boost the effectiveness of the enzymes needed for biological carbon fixation [ 93 ]. In the plant Sedum alfredii, growing under Cd stress, Wu et al. (2018) showed that bacterial inoculants might enhance the effectiveness of photosynthetic enzymes including Rubisco, Ca 2+ -ATPases, and Mg 2+ -ATPase and upregulate the expression of photosystem related genes (SaPsbS and SaLhcb2) [ 94 ]. One of the primary purposes of stomata is carbon fixation, which enables the plant to absorb CO 2 and exhale O 2 . There have been claims that bacteria can control stomatal conductance in plant leaves. Interestingly, some research discovered enhanced stomatal conductivity [ 95 ], while others found that it was reduced when the plants were colonized by certain microbes [ 96 ]. Carbon is removed from the environment and deposited in the soil due to increased CO 2 sequestration from the air, improved photosynthesis, and augmented soil organic matter [ 97 ]. By boosting plant root development, microbial inoculants improve photosynthesis and deposit soil organic matter in the rhizosphere. Plant growth hormones are produced by root-associated microorganisms [ 98 ], enhance the expression of genes involved in root growth and development [ 92 , 99 ], and increase the concentration of photosynthetic pigments, chlorophyll, and carotenoids [ 100 ], all of which contribute to considerable organic matter deposition in the soil in the form of increased root yield. Reactive oxygen species (ROS) are produced in leaf tissues when photosynthetic pigments are overexcited or when plants are exposed to environmental challenges [ 101 ]. The released ROS reduces the photosynthesis rate by inhibiting photosynthetic enzyme activity [ 102 ]. Interestingly, Rhizobiaceae bacteria, some endophytes, Trichoderma sp., Piriformospora indica , arbuscular mycorrhizal fungi, and other beneficial microbes boost the expression of genes that make proteins that detoxify reactive oxygen species, as shown in Table 2 [ 103 ], thereby protecting the plants’ photosynthesis efficiency from ROS. Overall, plant development and physiology are significantly impacted by the CO 2 uptake in plants mediated by microbes. For instance, it has been demonstrated that higher CO 2 concentrations promote plant development by enhancing photosynthetic C input, nutrient and water utilization efficiency, and biomass product [ 104 , 105 ]. C fixation is therefore utilized in modern agriculture, especially in focused agricultural systems, as a “fertilizer” to increase plant yields. In order to increase plant productivity and lessen the effectiveness of the greenhouse gas, CO 2 , released by human activities, we must better understand the molecular mechanisms of plant–microbe interactions that enable plants to increase the efficiency of their photosynthetic processes as well as the role of various microbe-related factors in CO 2 fixation."
} | 6,175 |
36756571 | PMC9890961 | pmc | 2,087 | {
"abstract": "The fabrication of mechanically robust multifunctional nanocomposite (NC) films using simple but effective strategies is a long-term challenge. Inspired by natural nacre, we designed and fabricated high-performance nacre-like NC films (Na-MTM/HBP) through the self-assembly of the hyperbranched poly(amido amine) (HBP) and montmorillonite (Na-MTM) using a vacuum filtration approach. The optimal Na-MTM/HBP NC film shows excellent mechanical strength (106 MPa), which can be attributed to the formation of numerous hydrogen bonds and the electrostatic interactions between hyperbranched HBP and Na-MTM nanosheets. Such films also exhibit excellent gas barrier and fire–fire-retardant owing to the high aspect ratio of the Na-MTM nanosheets. In this work, a class of high-performance NC films exhibiting good mechanical, gas barrier, and flame retardancy properties have been developed. These NC films have great potential in packing or coating materials.",
"conclusion": "4. Conclusions We successfully prepared nacre-like NC films using vacuum filtration through the self-assembly of HBP molecules with Na-MTM nanosheets. The Na-MTM/HBP NC films exhibited superior mechanical properties owing to the strong interactions between HBP and Na-MTM. The tensile strength of Na-MTM/HBP-5 NC film reached 106 MPa, which is higher than that of many MTM-based composites. The XPS, FTIR, and zeta potential results verified that the self-assembly was driven by hydrogen bonding and electrostatic reactions between Na-MTM and HBP. Moreover, because of the high aspect ratio of Na-MTM, Na-MTM/HBP NC films have good gas barrier and flame fire-retardant. As a result of their excellent mechanical, gas barrier, and flame fire-retardant, these films are promising candidates for use in fire-protection packaging or coating materials.",
"introduction": "1. Introduction The development of new generations of materials exhibiting high performance and multiple functionalities for diverse strategic fields, such as building construction, transportation, aerospace, and biotechnology, is constantly in demand. 1–3 Through evolution, nature has found smart ways to create lightweight, strong, and tough structural materials with multiple functionalities, such as nacre, tooth, and bone. 4–6 These structural materials usually exhibit elegant and complex architectures at multiple length scales. 7,8 Thus, mimicking the architecture of natural materials is a clever approach to designing new materials. The nacreous layer of mollusks comprises alternating layers of brittle inorganic calcium carbonate platelets and biopolymers, and it exhibits remarkably high toughness and resilience. 9–11 The outstanding merits of nacre can be attributed to the hierarchical arrangement of its soft and hard constituents that shape a brick-and-mortar structure. 7,12 The nacre-inspired design principle is to arrange the hard, reinforcing platelets and energy-dissipating soft polymers into ordered structures. 13,14 Novel methods, including layer-by-layer assembly, 15–17 spray casting, 18 vacuum-filtration-induced self-assembly, 19,20 and magnetic-field-assisted additive manufacturing, 21 have been developed to replicate multifunctional nacres using polymers and inorganic platelets. For instance, Du et al. 22 fabricated a nacre-mimetic composite with intrinsic self-healing and shape-programming capabilities by infiltrating a thermally reversible Diels–Alder network polymer into a long-range-ordered lamellar scaffold of alumina platelets. Li et al. 9 prepared nacre-like poly (vinyl alcohol)/graphene composite films with superior mechanical, electrical, and biocompatible properties using simple solution casting. In mimicking a nacre film, the interaction between the soft polymer and hard 2D nanoplatelets is a key factor in the combination of outstanding mechanical properties and functionalities. 23,24 Introducing multiple hydrogen bonds and chemical crosslinks can effectively enhance the interactions between soft polymers and nanoplatelets. 25,26 However, the fabrication of nacre-like materials with superior mechanical properties is usually tedious and time-consuming. Dendric polymers, including hyperbranched polyglycerol and poly(amido amine), are alternative soft constituents for fabricating strong nacre-like composites because they have various polar groups capable of forming hydrogen bonds and electrostatic interactions with nanoplatelets. 27–30 Here, a strong biomimetic artificial nacre with gas barrier and fire-retardant functions was successfully fabricated using vacuum filtration of sodium montmorillonite (Na-MTM) and hyperbranched poly(amido amine) (HBP) ( Fig. 1a ). Na-MTM and HBP are self-assembled into a biomimetic laminated structure during vacuum filtration with the help of hydrogen bonds and electrostatic interactions. The Na-MTM/HBP nanocomposite (NC) films have mechanical strengths as high as 106 MPa and exhibit excellent gas barrier and flame–fire-retardant. The oxygen permeability (OP) rate is 0.03 mL μm m −2 d −1 kPa −1 . Because of their excellent mechanical, gas barrier, and flame–fire-retardant, these films are promising candidates for diverse applications, including barrier materials for encapsulation and coating. Fig. 1 (a) Schematic diagram for the preparation of Na-MTM/HBP artificial nacre, (b) a photograph of Na-MTM/HBP film, (c) TGA traces for Na-MTM, HBP and Na-MTM/HBPs, (d) Na-MTM weight of Na-MTM/HBPs.",
"discussion": "3. Results and discussion 3.1 Preparation of HBP and Na-MTM/HBP films For the preparation of Na-MTM/HBP nacre-like films, HBP was first synthesized using a one-pot Michael addition reaction between MBA and HAD with a molar ratio of 1 ( n MBA : n HAD = 1) (ESI Fig. S1 † ). HBP has a molecular weight of 99 253 g mol −1 and a polydispersity of 1.46. FTIR and 1 H NMR were used to examine the molecular structure of HBP (ESI Fig. S2 † ). As shown in the FTIR spectrum of MBA, 3305 cm −1 was assigned to the strong stretching vibration peak of the N–H group; 1658 and 1453 cm −1 were assigned to the stretching vibration peaks of the C \n \n\n<svg xmlns=\"http://www.w3.org/2000/svg\" version=\"1.0\" width=\"13.200000pt\" height=\"16.000000pt\" viewBox=\"0 0 13.200000 16.000000\" preserveAspectRatio=\"xMidYMid meet\"><metadata>\nCreated by potrace 1.16, written by Peter Selinger 2001-2019\n</metadata><g transform=\"translate(1.000000,15.000000) scale(0.017500,-0.017500)\" fill=\"currentColor\" stroke=\"none\"><path d=\"M0 440 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z M0 280 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z\"/></g></svg>\n\n O (amide I) and N–H (amide II) groups, respectively; and 992 and 967 cm −1 were assigned to the bending vibration peaks of the C C. As seen in the FTIR spectrum of HBP, the stretching vibration peak of –NH– widened; the characteristic peaks of amide I and amide II were observed at 1648 and 1540 cm −1 , respectively; and the characteristic peak of C C disappeared. As seen in the 1 H NMR spectrum of HBP, the typical proton signals of C C at 6.25 and 5.70 ppm disappeared, and a new signal belonging to the methylene of –CH 2 –CH 2 –CONH– in HBP appeared at 2–3 ppm. The FTIR and 1 H NMR results confirmed that all vinyl groups were consumed during the Michael addition reaction, and HBP was successfully synthesized. The percentages of primary amine, secondary amine, and tertiary amine units calculated by integrating the characteristic peaks of HBP in the NMR spectra were 20.5%, 59%, and 20.5%, respectively (ESI Fig. S3 † ). The fabrication process of the Na-MTM/HBP NC films is shown in Fig. 1a . The Na-MTM suspension and different concentrations of HBP aqueous solutions were used to prepare a series of Na-MTM/HBP dispersions, which were then redispersed using centrifugation. Finally, Na-MTM/HBP NC films were formed using vacuum infiltration. After their self-assembly during vacuum infiltration, the Na-MTM/HBP NC films exhibited certain light transmittance, as indicated by the logo of Sichuan University visible through the Na-MTM/HBP NC films in Fig. 1b . The transmittance spectra of the Na-MTM/HBP NC films showed that the transmission decreased with HBP concentration in the range of 400–800 nm. The transparency of the Na-MTM/HBP NC films was because of the uniform distribution and orderly arrangement of Na-MTM nanosheets. According to the TGA curves shown in Fig. 1c , the residue weights of the Na-MTM/HBP NC films obtained at 800 °C were 48.07%, 49.03%, 54.56%, 56.61%, and 63.17% for Na-MTM/HBP-1, Na-MTM/HBP-2, Na-MTM/HBP-3, Na-MTM/HBP-4, and Na-MTM/HBP-5, respectively. Further, the approximate amounts of clay in the hybrid NC films were 55.73%, 56.84%, 63.25%, 65.63%, and 73.23% for Na-MTM/HBP-1, Na-MTM/HBP-2, Na-MTM/HBP-3, Na-MTM/HBP-4, and Na-MTM/HBP-5, respectively, as shown in Fig. 1d . 3.2 Structure and morphology of the nacre-like composites The fracture surfaces of the Na-MTM/HBP NC films are examined using SEM. Fig. 2a and b show that the Na-MTM/HBP building blocks were stacked together, forming a compact lamellar microstructure. Similar to the special “brick-and-mortar” structure of natural nacre, the Na-MTM nanoplatelets acted as bricks and HBP as mortar. Fig. 2 SEM images of Na-MTM/HBP nacre-mimetic cross section, the scale bars indicate 4 μm (a) and 500 nm (b), 1D SAXS curves of the Na-MTM and Na-MTM/HBP (c), and their Lorentz corrected SAXS plots (d). Prominent interlayer gaps were observed between lamellae. SAXS was conducted to quantify the structural periodicities of the pure Na-MTM nanosheets and Na-MTM/HBP NC films to investigate the effect of HBP concentration on the interlayer gaps of Na-MTM/HBP NC films, as shown in Fig. 2c and d . The pure Na-MTM layers had a high q value, which gradually decreased with increasing HBP concentration. Therefore, the spacing between the Na-MTM layers of the Na-MTM/HBP NC films was larger than that of the pure Na-MTM nanosheets. As calculated using Bragg's formula, 32 the specific spacings of the Na-MTM layers of Na-MTM/HBP-1, Na-MTM/HBP-2, Na-MTM/HBP-3, Na-MTM/HBP-4, and Na-MTM/HBP-5 NC films were 3.15, 2.90, 2.43, 2.05, and 1.95 nm, respectively. The polymer layer thickness can be calculated to be 1.95, 1.70, 1.23, 0.85, and 0.75 nm by deducting the thicknesses of the clay nanosheets (1.2 nm) from the interlayer distances. These results indicated that each Na-MTM/HBP NC film exhibited periodic structure and that HBP was absorbed into the Na-MTM nanosheet layers. 3.3 Self-assembly mechanism of the nacre-like composites In mimicking a nacre with superior mechanical properties and multiple functionalities, the interactions between the soft polymer and the hard 2D nanoplatelets play a key role. The hypothetical mechanism for the fabricated Na-MTM/HBP NC film is shown in Fig. 1a . The self-assembly of the NC film was mainly accomplished by the hydrogen and electrostatic interactions between the amino groups in HBP and Na-MTMs, making the structure stable. The TEM images of the Na-MTM nanosheets before and after absorbing the HBP molecules are shown in Fig. 3a and b . The pure Na-MTM nanosheets easily aggregated, but the aggregation of Na-MTM weakened after the integration of HBP molecules. This phenomenon was attributed to the strong interactions between HBP and Na-MTMs that may result in the adsorption of HBP into Na-MTM nanoplatelet laminates. Fig. 3 (a and b) TEM images of MTM nanosheets before and after adsorbing HBP molecules, (c) XPS wide spectrum of Na-MTM/HBPs, (d) FT-IR spectra of Na-MTM and Na-MTM/HBP nanosheets, (e) zeta potentials of Na-MTM, HBP and Na-MTM/HBPs, (f) photographs of Na-MTM suspension liquid, HBP solutions and Na-MTM/HBP suspension liquid. The interfacial interactions were further investigated by analyzing the chemical structure of the Na-MTM/HBP NC films using XPS and FTIR. As shown in the XPS wide spectrum of Na-MTM/HBP NC films in Fig. 3c , the relative content of Si increased with the amount of Na-MTM. This result was attributed to the Si component of Na-MTM. Three characteristic peaks were observed in the C 1s core-level spectrum of Na-MTM/HBP (ESI Fig. S4a † ) and assigned to C–C/C–H (284.8 eV), C–N (285.8 eV), and C O (287.8 eV) for HBP. In addition, the relative contents of primary amine (401.5 eV), secondary amine (399.6 eV), and tertiary amine (398.9 eV) were determined from the N 1s core-level spectrum of Na-MTM/HBP (ESI Fig. S4b † ) and calculated to be 19.3%, 61.4%, and 19.3%, respectively. These values were consistent with the NMR results. The large number of amino groups in HBP caused the formation of hydrogen bonds between HBP and Na-MTM. As shown in the C 1s core-level spectrum of Na-MTM/HBP (ESI Fig. S4c † ), (HBP) C O (531.3 eV) and (Na-MTM) O (532.0 eV) appeared simultaneously. This finding illustrated that HBP molecules were absorbed by Na-MTM. FTIR measurements were also used to investigate the absorption of HBP in Na-MTM. Fig. 3d shows that the FTIR spectrum of pure Na-MTM displayed stretching bands at 1039 cm −1 for Si–O–Si stretching and 1639 cm −1 for –OH bending. In addition, the FTIR spectrum of HBP exhibited two characteristic peaks: the bending vibration peak of –NH– at 1556 cm −1 and the stretching vibration peak of C–N at 1414 cm −1 . For the Na-MTM/HBP NC films, the FTIR signals combined the characteristic peaks of HBP and Na-MTM. A comparison of peak positions in HBP and Na-MTM/HBP revealed that the absorption peaks of the amide I band and the amide II band shifted in Na-MTM/HBP NC films (ESI Fig. S5 † ). This finding further verified that HBP molecules were attached to Na-MTM nanosheets through hydrogen bond interactions. HBP aqueous solutions and Na-MTM and Na-MTM/HBP suspensions were investigated using a zeta potential instrument to disclose the absorption mechanism between HBP and Na-MTM. Fig. 3e shows that Na-MTM exhibited a strong negative zeta potential ( ζ = −39.4 mV). The overall negative charge was due to ionic substitutions in the nanoclay structure, leading to charged basal planes and concomitant Na + counterion release during delamination in water. The presence of a large number of amino groups on the surface of HBP resulted in positive zeta potential ( ζ = 16.0 mV). Upon mixing Na-MTM with HBP, the zeta potential of the Na-MTM/HBP suspension reversed to strongly positive owing to the formation of polyelectrolyte NC films (Na-MTM/HBP). Hence, the entropy driving force of the adsorption of HBP onto Na-MTM was generated by the hydrogen bonding and the release of ordered water molecules at the surface of nano-Na-MTM. Positively charged HBP can integrate strongly with Na-MTM after the release of Na + counter ions. As shown in Fig. 3f , the Na-MTM/HBP suspension darkened in color when the negatively charged Na-MTM absorbed HBP. This finding further indicated that the integration of Na-MTM and HBP was driven by hydrogen bond interaction and electrostatic interaction. 33 The NA-MTM/HBP dispersion is self-assembled into a film with a nacre-like structure using vacuum filtration. 3.4 Mechanical properties of Na-MTM/HBP nacre films Tensile tests were conducted to investigate the influence of HBP on the mechanical properties of nacre-like Na-MTM/HBP NC films. The typical stress–strain curves of HBP and nacre-mimetic NC films are shown in Fig. 4 . The tensile strength and fracture strain of HBP were only 16 MPa. The tensile strength improved gradually with the increase in Na-MTM content. When the Na-MTM content was up to 50%, the tensile strength reached 106 MPa, which is equivalent to that of reported nacre-like Na-MTM-based NC films. 15,34–36 When a high content of inorganic fillers is incorporated into a composite without a periodic structure, the mechanical properties of the materials are usually damaged. In this work, this dilemma was overcome by using vacuum infiltration to create ordered nacre-like structures. Upon the addition of Na-MTM nanoplatelets, the tensile strength was improved owing to the formation of hydrogen bonds and electrostatic interactions between Na-MTM and HBP “brick-and-mortar” structures. Fig. 4 Typical tensile stress–strain curves of Na-MTM/HBP nacre-like composites. As the Na-MTM content increases from Na-MTM/HBP-1 to Na-MTM/HBP-5, the fracture toughness first increases and then decreases, with a maximum value of 1.61 MJ m −3 at Na-MTM/HBP-3. This trend is closely related to Na-MTM content and Na-MTM arrangement regularity in the samples. Because HBP acts as an energy-dissipating soft constituent, it has the highest fracture toughness at 7.69 MJ m −3 . We also proposed a fracture model of the robust Na-MTM/HBP NC films. In the first stage, when mechanical stress was applied to the NC film, the slippage and displacement of layers were resisted by the Na-MTM nanosheets. In the next stage, with the increment of stress, the hydrogen bonds and electrostatic interactions began to break, and the Na-MTM nanosheets started to slide over each other, followed by the initiation of cracks that resulted in energy dissipation. Meanwhile, the HBP molecules stretched from the surface of Na-MTM. When the load continued to increase, the nanosheets were pulled out of the lamella, and fractures occurred. Compared with the complex techniques usually used for Na-MTM nacre-like composites, 25,26 such as the introduction of chemical crosslinking, our method is more easily able to obtain such high-strength nacre-like NC films. Hyperbranched HBP can form numerous hydrogen bonds and electrostatic interactions with Na-MTM nanosheets, resulting in superior mechanical properties. Researchers introduce strong interaction groups with Na-MTM nanosheets via complex monomer design and synthesis steps or cross-link the nacre films using a complex curing process to achieve nacre with high strength. 3.5 Multifunction properties of Na-MTM/HBP films The use of 2D nanosheets usually endows the composites with multiple functionalities, such as electrical, gas barrier, and flame fire-retardant. For the Na-MTM/HBP films, Na-MTM nanoplatelets with a high aspect ratio should be promising owing to their gas barrier and flame fire-retardant. The OP of the Na-MTM/HBP NC films decreased with the addition of Na-MTM nanosheets. In particular, the OP decreased to 0.15, 0.11, 0.09, 0.06, and 0.03 mL μm m −2 d −1 kPa −1 with the introduction of 10%, 20%, 30%, 40%, and 50% Na-MTM nanosheets, respectively. The OP of Na-MTM/HBP NC film with 50% Na-MTM was close to that of commercial ethylene vinyl alcohol film (0.01 mL μm m −2 d −1 kPa −1 ). 37 The superior gas barrier properties of the Na-MTM/HBP NC films were closely related to the high aspect ratio and orderly arrangement of Na-MTM nanosheets in the NC films. We then investigated the flame fire-retardant of the Na-MTM/HBP NC films by placing a cotton swab behind the samples in an alcohol flame that had a temperature of 600–800 °C, as shown in Fig. 5 . The cotton swab did not catch flame even after 5 min of exposure to the Na-MTM/HBP NC films. Meanwhile, the cotton swab behind pure HBP ignited immediately upon contact with the flame, and the film burned off quickly. The Na-MTM/HBP NC film containing 50% Na-MTM kept its original shape after the sample was placed in the flame for 5 min. However, the sample became charred and turned black because of HBP combustion. The residue material showed good flame resistance. These results suggested that the nacre-like Na-MTM/HBP NC films exhibited superior flame fire-retardant. Fig. 5 Fire-retardant test of the pure HBP film (a) and the composite film with 50% Na-MTM content (b) in which the film acts as a fire shield to protect a cotton swab."
} | 4,913 |
38906891 | PMC11192760 | pmc | 2,090 | {
"abstract": "To adapt to the complex belowground environment, plants make trade-offs between root resource acquisition and defence ability. This includes forming partnerships with different types of root associating microorganisms, such as arbuscular mycorrhizal and ectomycorrhizal fungi. These trade-offs, by mediating root chemistry, exert legacy effects on nutrient release during decomposition, which may, in turn, affect the ability of new roots to re-acquire resources, thereby generating a feedback loop. However, the linkages at the basis of this potential feedback loop remain largely unquantified. Here, we propose a trait-based root ‘acquisition-defence-decomposition’ conceptual framework and test the strength of relevant linkages across 90 angiosperm tree species. We show that, at the plant species level, the root-fungal symbiosis gradient within the root economics space, root chemical defence (condensed tannins), and root decomposition rate are closely linked, providing support to this framework. Beyond the dichotomy between arbuscular mycorrhizal-dominated versus ectomycorrhizal-dominated systems, we suggest a continuous shift in feedback loops, from ‘high arbuscular mycorrhizal symbiosis-low defence-fast decomposition-inorganic nutrition’ by evolutionarily ancient taxa to ‘high ectomycorrhizal symbiosis-high defence-slow decomposition-organic nutrition’ by more modern taxa. This ‘acquisition-defence-decomposition’ framework provides a foundation for testable hypotheses on multidimensional linkages between species’ belowground strategies and ecosystem nutrient cycling in an evolutionary context.",
"introduction": "Introduction Since the early Devonian, the progressive emergence of root organs has had profound consequences for land colonisation and diversification of plants 1 – 3 . During evolution, plants have balanced costs and benefits for better adaptation to complex and ever-changing environments. Sessile growth may have led plants to evolve particularly complex and diverse belowground strategies. When investing carbon (C) belowground, for example, plants may build a root with trait syndromes that range from the very efficient acquisition of soil resources on the one side to effective defences against biotic and abiotic threats for their survival and prosperity on the other side 4 . They can also rely on microbial partners to acquire resources or benefit from protection 5 – 7 . Such trade-offs between different types of ecological strategies have consequences for the fate of allocated C and nutrients after organs have senesced and died 8 – 10 . Roots decompose where they grow and die 11 , implying that plants may access and preempt the nutrients released by their root litter. Therefore, the trade-off in plant resource allocation to belowground resource acquisition versus defence 12 has legacy effects on root decomposition, which may further feedback to nutrient reutilization by plants 13 , 14 . By taking a holistic view of these intrinsic linkages, we propose a feedback loop of ‘acquisition-defence-decomposition’ (hereafter abbreviated as ADD) operating on root systems driven by the multidimensional belowground adaptation strategies of species (Fig. 1 ), which may facilitate a comprehensive understanding of plant belowground strategies, species coexistence, and diversity maintenance. Fig. 1 Conceptual framework for the ‘acquisition-defence-decomposition (ADD)’ loop in the root system. a To adapt to the complex soil environment, plants either rely on their absorptive roots (i.e., root pathway) or cooperate with symbiotic fungi (arbuscular mycorrhizal or ectomycorrhizal pathways) to acquire soil resources (e.g., nitrogen N; phosphorus P; potassium K) while producing defensive substances (e.g., secondary metabolites) for protection against pathogens and herbivores. There could be a trade-off between ‘acquisition’ and ‘defence’ for given photosynthetic carbon, which may exert a legacy effect on carbon (C) loss and nutrient release during decomposition by mediating root chemistry. These released nutrients can further feedback to nutrient reutilization by root systems considering that absorptive roots decompose where they live and die. We propose that these root processes are intrinsically linked to form an ADD loop. b Differences in belowground strategies with respect to mycorrhizal associations (type and intensity) are assumed to drive a continuum of feedback loops from a ‘high AM symbiosis-low defence-fast decomposition’ loop to a ‘high EcM symbiosis-high defence-slow decomposition’ loop with several intermediate transition loops. The left diagram is inspired by Heike et al. 79 and Salas-González et al. 80 . The ADD conceptual framework is essentially determined by the relationships among root traits, which are largely influenced by root-fungal symbiosis 15 . The vast majority of land plants form symbiotic associations with mycorrhizal fungi 16 , 17 . Plants allocate photosynthate to roots or fungal partners to acquire edaphic resources 6 , and to different types of mycorrhizal partners (e.g., arbuscular mycorrhizal (AM) and ectomycorrhizal (EcM)), which results in varying root traits that represent the type of association and their intensity 18 . Mounting evidence has shown strong evolutionary controls over root traits related to mycorrhizal symbiosis, such as diameter 19 – 21 , cortical tissue 22 , and branching 20 . Furthermore, condensed tannins, an important root trait related to chemical defence 23 , are less abundant in the early diverging magnoliids than in modern rosids 24 . These findings imply possible covariations among root traits associated with ‘acquisition’ and ‘defence’ as taxa evolve, yet it remains unclear whether these evolutionary patterns exert influences on the ‘acquisition-defence-decomposition’ linkages within the ADD conceptual framework. The two most common mycorrhizal associations in woody plants, AM and EcM, have distinct evolutionary histories and symbiotic pathways (intracellular vs. intercellular) 15 , 25 , leading to large differences in their root traits 18 . Belowground resource acquisition strategies are reflected by the trait-based root economic space (RES), where the root-fungal symbiotic pathway and intensity differentiate AM from EcM plant species 18 , 26 . In the evolutionarily ancient AM fungal species, hyphae can penetrate cortical cells, forming intracellular arbuscules, and thus, host plants tend to build roots with a broader cortex space for colonization by AM fungi to acquire limiting soil resources 18 . It has been well documented that thicker AM roots with a greater cortex and higher mycorrhizal dependency are associated with lower concentrations of condensed tannins 24 , a phenolic compound slowing down root decomposition 27 . These linkages imply that with the constraint of C, AM tree species relying more on symbiotic fungi may invest less in their belowground chemical defences, thus forming root litter with fewer tannins and faster decomposition. These patterns in AM plant species may not necessarily hold for more modern EcM species with intercellular symbiosis. EcM plants tend to enhance their collaboration with mycorrhizal fungi by constructing roots with high branching to maximize the area of the symbiotic interface for acquiring limiting soil resources 6 , 28 . The evolutionary transition from AM to EcM symbiosis implies stronger control of the host plants over symbiotic fungi 16 . For instance, EcM angiosperms develop wall-thickened exodermis with Casparian bands and suberin lamellae and deposit phenolic compounds (e.g., tannins), which function as a physical and chemical barrier to prevent further penetration of symbiotic fungi and pathogens into the inner part of the cortex 29 – 31 . Therefore, EcM fungi may trigger higher levels of physical and chemical defences in their host plants than AM fungi 32 , which may lead to a symbiosis-defence relationship distinct from that in AM species. Furthermore, tannin-rich roots in EcM species could associate with chitin in fungal necromass to form more complex macromolecular polymers that are recalcitrant to decomposers 33 , 34 . These findings indicate that the intensified chemical defences associated with EcM symbiosis could probably result in slower root decomposition. The classical mycorrhizal-associated nutrient economy (MANE) hypothesis proposes that AM- and EcM-dominated forests with contrasting nutrient-cycling modes differ in litter quality, which feeds back to plant nutrient uptake 9 , 25 . Fast-decomposing litters with higher nutrient contents also release nutrients faster and in larger amounts for use by soil organisms and complexation with the soil matrix 35 . Soil microbial communities fed with organic matter inputs of lower C:nutrient ratio further tend to release a larger proportion of nutrients for the use of plants 36 . Such fast-decomposing litter also favours the development of copiotrophic (bacterial-dominated) microbial communities speeding up nutrient cycling in soils, whereas slower-decomposing litter favours oligotrophic (fungal-dominated) communities better adapted to degrade organic compounds with low nutrient content 37 . In this context, EcM tree species, characterized by an organic-nutrient economy, benefit from habitats with slow organic matter decomposition and nutrient release. Conversely, AM tree species thrive in conditions of faster C and nutrient cycling that they promote. However, the MANE framework remains disconnected from current theories in root ecology and aspects of the root economics space 26 , belowground litter decomposition and nutrient recycling 38 , and is based on a dualistic view of monolithic AM versus EcM mycorrhizal associations. The ADD conceptual framework (Fig. 1 ) intends to fill in these gaps by specifically focusing on belowground aspects of root economics, particularly the trade-off in root traits, and developing a more quantitative trait-based approach to account for the large variation in trait values across and within tree species mycorrhizal type. In this study, we focused on absorptive roots (the most distal 1st- and 2nd-order roots within the branching root system) of 90 angiosperm tree species (65 AM and 25 EcM) spanning two temperate and two subtropical forests in China. These species represent a subset of species from Yan et al. 18 . To reveal the strategies for belowground resource acquisition, we characterized the RES of these species by integrating root morphological (root diameter (RD), specific root length (SRL), root tissue density (RTD)), architectural (branching intensity (BI)), anatomical (cortex thickness (CT)), and chemical (nitrogen (N) concentration) traits with the concentration of condensed tannins, which indicate the chemical defence of roots against herbivores 23 , 24 , 39 . We also performed a microcosm experiment to quantify the decomposition rate of absorptive roots in these tree species. Our overarching hypothesis is that there are strong linkages and trade-offs between resource acquisition strategy, defence capacity, and decomposition rate in the absorptive root system. Considering the distinct evolutionary histories, symbiotic pathways, and nutrient economies of AM and EcM species 9 , 25 and large variation in trait values within each mycorrhizal type, we further hypothesized that evolutionary shifts from AM to EcM association mediate the ‘acquisition-defence-decomposition’ linkages, thus forming a progressive gradient of nutrient-cycling modes. Here, we provide evidence for our ‘acquisition-defence-decomposition’ conceptual framework by documenting a continuum in root and mycorrhizal traits responsible for multiple feedback loops among plant strategies of nutrient acquisition, defence and afterlife effects on decomposition. This work substantiates the classical MANE hypothesis and, using a trait-based approach expands it beyond the restrictive dichotomy between AM-dominated versus EcM-dominated systems.",
"discussion": "Discussion We hypothesized here that several key root processes, including nutrient acquisition, chemical defence, and root decomposition, are linked via trade-offs, which can be characterized by root trait syndromes 18 , 40 , 41 . By comparing root traits closely related to these processes across 90 tree species, we found a suite of important belowground trait gradients and linkages. First, root resource acquisition can be reflected by a bi-dimensional RES, consistent with the findings in Yan et al. 18 . The SRL-RTD axis represented a ‘lifespan gradient’ from acquisitive-strategy species with short root lifespan to conservative-strategy species with long root lifespan, while the RD/CT-BI axis represented a ‘symbiosis gradient’ showing the dependency of plant nutrient acquisition on their fungal partners (Fig. 2a ). Second, a trade-off between the type of mycorrhizal association (AM vs EcM) and some aspects of chemical defence were indicated by the negative relationship of the score of RD/CT-BI axis (symbiosis gradient) in the RES with the concentration of condensed tannins (Fig. 2b , Supplementary Fig. 2 ). Our results further suggest that this trade-off may have exerted a legacy effect on root decomposition, as indicated by the negative relationship between concentrations of condensed tannins and root mass loss (Fig. 2c , Supplementary Fig. 2 ). These trade-offs and legacy effects jointly bring preliminary evidence for the role of tree roots and mycorrhizal type in driving a feedback loop (ADD conceptual framework; Fig. 1 ). The RES revealed here differed from the axis representations in Bergmann et al. 26 where RD-SRL axis represents a “collaboration” gradient and RTD-N axis represents a “conservation” gradient. Here, we argue that these two representations of the RES are overall consistent with each other, and the fundamental divergence stems from our sampling of tree species displaying little variation in RD. Based on the theoretical formula: SRL = 4/(π × RTD × RD 2 ), any change in RD has much stronger weight on SRL than change in RTD because RD is quadratic to SRL 42 . In contrast to the global dataset of Bergmann et al. 26 , who used a broad definition of fine roots, our dataset is restricted to a narrow range of root orders (1st- and 2nd-order roots). It also focuses on a narrower group of plants (‘trees’ only rather than the ‘woody species’ category), and lineage (angiosperms). This stringent selection is likely responsible for the narrow interspecific variation in RD, which led to the decoupling between RD and SRL. In contrast, we found a large variation in RTD across species leading to a strong coupling between SRL and RTD. Interestingly, our results, based on a strong representation of subtropical forests (60%, Supplementary Data 1 ), support previous observations that changes in SRL may be more strongly regulated by RTD compared to RD in tropical forests 43 . Nonetheless, we also acknowledge that the low variation in RD observed in our study may not be representative of global tree root variation, which deserves further research. Tree species position along the ‘symbiosis gradient’ is associated with their evolutionary history, indicated by the close link between the RD/CT-BI axis scores and divergence time at the family level (Fig. 2e ; Supplementary Fig. 6 ). Along evolutionary lines, AM species may have reduced their dependence on symbiotic fungi, represented here by the decreased RD and CT with divergence time. Meanwhile, following the emergence of EcM associations, EcM species typically show a consistently strong reliance on EcM fungi, as reflected here from high BI 20 – 22 . Considering the tight linkages among root-fungal symbiosis, chemical defence, and root decomposition, we emphasize that the evolutionary trend of root-fungal symbiosis has major consequences on multidimensional strategies of the root system (Fig. 2e ), although we did not find here direct evidence that evolution acts upon chemical defence and root decomposition (Supplementary Fig. 7 ). Therefore, the root ADD strategies may have evolved from ‘high AM symbiosis-low defence-fast decomposition’ (e.g., Magnoliaceae) to ‘low AM symbiosis-high defence-slow decomposition’ (e.g., Sapindaceae) and to ‘high EcM symbiosis-high defence-slow decomposition’ (e.g., Betulaceae). This strategic transition may result from trade-offs and long-term adaptation to complex climatic and edaphic environments over the evolutionary history of plants. Notably, non-mycorrhizal (NM) plants that may evolve from facultative AM or NM-AM plants often emerge in ultra-poor soils 44 . To exclude mycorrhizal fungi from their roots, NM plants have evolved potent chemical defence mechanisms by accumulating relatively advanced secondary metabolites (e.g., alkaloids and cyanogens), contrasting with mycorrhizal plants that are more likely to contain primitive secondary metabolites, such as tannins 16 , 44 . This basic symbiosis-defence difference may result in contrasting root ADD strategies in NM plants compared to mycorrhizal plants, requiring further investigation. Considering each tree mycorrhizal type separately, the linkages between the ‘symbiosis gradient’ and chemical defence were distinct (Fig. 3b, f ). This finding is largely consistent with previous knowledge of AM versus EcM differences in anatomical and chemical root trait organization 17 . We found negative relationships of the dominant traits (RD and CT, Supplementary Fig. 4 ) and the RD/CT-BI axis scores with the concentrations of condensed tannins in AM species only (Fig. 3b ), which concurs with a recent study reporting greater chemical protection associated with lower mycorrhizal dependency in AM species 24 . AM species with more primitive symbiotic pathways tend to construct thicker roots with a larger cortex to provide habitat for symbiotic AM fungi 18 , 28 . Compared to the nonmycorrhizal root system, mycorrhizal symbiosis may consume a large amount of C for the construction and maintenance of the symbiont habitat (wide cortex) and for producing the extraradical mycelium 44 . Therefore, roots would face selection pressure in optimizing host symbionts or enhancing chemical protection 45 , leading to trade-offs between mycorrhizal dependency and chemical defence. Our findings, in parallel with those of Xia et al. 24 , indicated that roots of AM species would maintain a low level of condensed tannins if they rely strongly on symbiotic partners for acquiring nutrients, and vice versa. In contrast to AM species, we found a synergy between the reliance of EcM species on mycorrhiza and root chemical defence, indicated by the positive relationships of the dominant traits (BI, Supplementary Fig. 5 ) and the BI-RD axis scores (the PC2 scores were made negative) with the concentration of condensed tannins (Fig. 3f ). This coordination could be fundamentally driven by the evolutionary shift in the root-fungal symbiosis interface from AM to EcM associations 16 . Increasing control of associations by the host reflects an important aspect of mycorrhizal evolutionary trend 16 . Unlike AM species, EcM angiosperms increase the complexity of the symbiotic interface by developing extracellular symbiotic pathways and enhance colonization sites by greater root branching 18 , 28 , 46 . To constrain the possible cheating behaviour from mycorrhizal fungi (i.e., low nutrient delivery and/or high photosynthate demand) 16 , EcM hosts can modify the exodermis cells by forming suberized Casparian bands 46 and depositing phenolic compounds such as condensed tannins to restrict the further penetration of EcM fungi into the inner cortex 46 – 48 . These specialized features could be associated with local cell death (in and around the initial infection site) and the release of toxins and phenols compounds (e.g., tannins) caused by ‘hypersensitive response’ during the formation of root-fungal symbiosis 30 , 47 – 49 . Overall, this finding implies that as roots collaborate closely with mycorrhizal fungi, they will also exhibit high levels of chemical defence in EcM angiosperms. Nonetheless, further investigation is needed to examine how symbiosis and defence interact to affect the root cell wall at the microscale. Aside from mediating mycorrhizal symbiosis, chemical defence components are also important determinants of root decomposition 27 , 50 . Consistent with the results from a previous study 27 , condensed tannins had a negative relationship with root decomposition, irrespective of the mycorrhizal type (Figs. 2c and 3c, g ), indicating their important role in slowing root decomposition. Furthermore, root-derived condensed tannins, particularly in EcM species, can form stable complexes with compounds from fungal polymers (e.g., chitin and melanin) 33 . Hence, the encased root-derived organic matter may have little opportunity to interact with extracellular enzymes 34 , leading to slow root decomposition. In addition to condensed tannins, lignin, bound phenolics, and other non-structural secondary compounds are also important chemical traits driving root decomposition 24 , 27 . Further incorporating these chemical drivers would thus help give a more comprehensive picture of the afterlife effects of root defence within the ADD framework. Notably, dominant root traits on the symbiosis gradient, particularly RD, also largely influenced root decomposition. We found a consistently positive relationship between RD and root decomposition rate for both AM species ( R 2 = 0.097, P = 0.012, Supplementary Fig. 4 ) and EcM species ( R 2 = 0.262, P = 0.009, Supplementary Fig. 5 ). Interestingly, a significant negative correlation occurred between the root diameter and condensed tannin concentration, which was robust across mycorrhizal types (Supplementary Figs. 3 – 5 ). Such a tight coupling among RD, condensed tannins, and root decomposition further supported the ADD conceptual framework. Although evidence is growing that RD along with BI and CT are closely associated with mycorrhizal symbiosis 24 , 39 , 51 , it is still necessary to further examine how a suite of traits with more direct links with the intensity of symbiotic associations (e.g., colonization rates and mycorrhizal C investments) fit within the ADD framework, particularly over long-time scales. The positive correlations between root ‘decomposition’ and ‘acquisition’ depicted here agree with plant‒soil feedback theories 38 , 52 and the MANE hypothesis 9 . The strong linkage between root decomposition rates and the RD/CT-BI axis scores, as opposed to a lack of relationship between root decomposition and the SRL-RTD axis scores (Fig. 3 , Supplementary Table 6 ), suggests a major role for plant mycorrhizal association type in soil nutrient cycling, beyond typical indicators of root physical defences such as lignin and dry matter content 8 . In turn, the positive linkage between root decomposition and acquisition relying more on AM fungi may induce a positive feedback loop that favours the presence of AM tree species. This strong reliance on AM fungi could be related to the lower cost and higher benefit of producing AM hyphae relative to constructing roots 25 . Compared to roots, hyphae grow faster and hence benefit more from quicker access to ephemeral inorganic nutrient patches 9 , 53 . In addition, the rapid turnover of hyphae allows plants more rapid cessation of resource allocation to hyphae when nutrient hotspots become depleted 54 , 55 . Furthermore, the variation of root decomposition rate along the ‘symbiosis gradient’ within the AM tree species suggests a transition from a fast-cycling belowground system that relies on AM symbiosis (ancient taxa) to a slower-cycling belowground system with lower dependence on AM symbiosis (young taxa). This pattern suggests an evolutionary strategy whereby plants tend to invest more in the root itself, resulting in higher chemical defence and slower nutrient cycling. For EcM species, no linkage was observed between the RD-BI axis scores (assumed to represent a stronger reliance of species on EcM association with increasing BI) and root decomposition rates (Fig. 3h ), but there was a strong positive relationship between BI and condensed tannins and a negative relationship between condensed tannins and mass loss (Supplementary Fig. 5 ). According to the MANE hypothesis, EcM fungi are characterized by an organic-nutrient economy, a trait partly retained from ancestral saprophytic fungi, and thus can obtain organic-bound nutrient either by symbiosis with plants or by decomposing or stimulating the decomposition of soil organic matter 9 , 56 . Our results further indicate large variation among EcM tree species, with a gradient from a slow-cycling belowground system that strongly relies on EcM symbiosis to a faster-cycling belowground system with lower dependence on EcM symbiosis. Overall, considering the large variations in root traits that occur within both the AM and EcM species groups and largely overlap with the classical AM versus EcM differences (Figs. 2 and 3 ) and their consequences for decomposition processes, our findings imply that the nutrient-cycling mode of plant species is more likely a progressive gradient rather than the clearcut AM versus EcM association described by the MANE hypothesis. In conclusion, the ADD conceptual framework depicts how plants coordinate root resource acquisition, chemical defence, and nutrient reutilization (nutrient release via root decomposition), thereby forming a nutrient feedback loop in the root-soil system. The general evolutionary trend in the root system towards less dependence on symbiotic AM fungi and the emergence of the EcM association has profound consequences for the multidimensional belowground strategy of resource acquisition and plant defence and for belowground nutrient cycling. We provide here some elements that further suggest a continuum between contrasting feedback loops, ranging from a loop favouring AM species (i.e., “high AM symbiosis-low defence-fast decomposition-inorganic nutrition”) to a loop favouring EcM species (i.e., “high EcM symbiosis-high defence-slow decomposition-organic nutrition”). Nonetheless, recent works depicting root trait effects on soil organic matter beyond the simple metrics of decomposability suggest that both fast decomposing roots 57 and slow decomposing roots 58 can contribute to the stabilization of soil organic matter, which limits nutrient accessibility to roots and fungi and may partly decouple root litter decomposition rate and N release. These uncertainties emphasize the necessity to provide further experimental support for the development of the ADD conceptual framework, particularly based on plant-soil feedback experiments and a broader set of taxa."
} | 6,715 |
28555626 | PMC5494993 | pmc | 2,091 | {
"abstract": "Methanogenic archaea are major players in the global carbon cycle and in the biotechnology of anaerobic digestion. The phylum Euryarchaeota includes diverse groups of methanogens that are interspersed with non-methanogenic lineages. So far methanogens inhabiting hypersaline environments have been identified only within the order Methanosarcinales . We report the discovery of a deep phylogenetic lineage of extremophilic methanogens in hypersaline lakes, and present analysis of two nearly complete genomes from this group. Within the phylum Euryarchaeota , these isolates form a separate, class-level lineage “Methanonatronarchaeia” that is most closely related to the class Halobacteria . Similar to the Halobacteria , “Methanonatronarchaeia” are extremely halophilic and do not accumulate organic osmoprotectants. The high intracellular concentration of potassium implies that “Methanonatronarchaeia” employ the “salt-in” osmoprotection strategy. These methanogens are heterotrophic methyl-reducers that utilize C 1 -methylated compounds as electron acceptors and formate or hydrogen as electron donors. The genomes contain an incomplete and apparently inactivated set of genes encoding the upper branch of methyl group oxidation to CO 2 as well as membrane-bound heterosulfide reductase and cytochromes. These features differentiates “Methanonatronarchaeia” from all known methyl-reducing methanogens. The discovery of extremely halophilic, methyl-reducing methanogens related to haloarchaea provides insights into the origin of methanogenesis and shows that the strategies employed by methanogens to thrive in salt-saturating conditions are not limited to the classical methylotrophic pathway.",
"conclusion": "Conclusions We discovered an unknown, deep euryarchaeal lineage of moderately thermophilic and extremely halo(natrono)philic methanogens that thrive in hypersaline lakes. This group is not monophyletic with the other methanogens but forms a separate, class-level lineage “Methanonatronarchaeia” that is most closely related to Halobacteria . The “Methanonatronarchaeia” possess the methyl-reducing type of methanogenesis, where C 1 -methylated compounds serve as acceptor and formate or H 2 are external electron donor, but differ from all other methanogens with this type of metabolism in the electron transport mechanism. In contrast to all previously described halophilic methanogens, “Methanonatronarchaeia” grow optimally in saturated salt brines and probably employ potassium-based osmoprotection, similar to extremely halophilic archaea and Halanaerobiales . This discovery is expected to have substantial impact on our understanding of biogeochemistry, ecology and evolution of the globally important microbial methanogenesis.",
"introduction": "Introduction Methanogenesis is one of the key terminal anaerobic processes of the biogeochemical carbon cycle both in natural ecosystems and in industrial biogas production plants 1 , 2 . Biomethane is a major contributor to global warming 3 . Methanogens comprise four classes, “ Methanomicrobia ”, Methanobacteria , Methanopyri and Methanococci , and part of the class Thermoplasmata , within the archaeal phylum Euryarchaeota 4 – 7 . The recent metagenomic discovery of putative methyl-reducing methanogens in the Candidate phyla “Bathyarchaeota” 8 and “Verstaraetearchaeota” 9 indicates that methanogenesis might not be limited to Euryarchaeota . Three major pathways of methanogenesis are known 1 , 2 : hydrogenotrophic (H 2, formate and CO 2 /bicarbonate as electron acceptor), methylotrophic (dismutation of C 1 methylated compounds to methane and CO 2 ) and acetoclastic (dismutation of acetate into methane and CO 2 ). In the hydrogenotrophic pathway, methane is produced by sequential 6-step reduction of CO 2 . In the methylotrophic pathway, methylated C 1 compounds, including methanol, methylamines and methylsulfides, are first activated by specific methyltransferases. Next, one out of four methyl groups is oxidized through the same reactions as in the hydrogenotrophic pathway occurring in reverse, and the remaining three groups are reduced to methane. In the acetoclastic pathway, methane is produced from the methyl group after activation of acetate. The only enzyme that is uniquely present in all three types of methanogens is methyl-CoM reductase, a Ni-corrinoid protein catalyzing the last step of methyl group reduction to methane 10 – 12 . The recent discovery of methanogens among Thermoplasmata \n 5 , 13 – 15 drew attention to the fourth, methyl-reducing, pathway, previously characterized in Methanosphaera stadtmanae ( Methanobacteria ) and Methanomicrococcus blatticola (“Methanomicrobia”) 16 – 20 . In this pathway, C 1 methylated compounds are used only as electron acceptors, whereas H 2 serves as electron donor. In the few known representatives, the genes for methyl group oxidation to CO 2 are either present but inactive ( Methanosphaera ) 16 or completely lost ( Thermoplasmata methanogens) 6 – 7 . Recent metagenomic studies have uncovered three additional, deep lineages of potential methyl-reducing methanogens, namely, Candidate class “Methanofastidiosa” within Euryarchaeota 21 and Candidate phyla “Bathyarchaeota” and “Verstraetearchaeota” 8 , 9 , supporting the earlier hypothesis that this is an independently evolved, ancient pathway 22 . The classical methylotrophic pathway of methanogenesis that has been characterized in moderately halophilic members of Methanosarcinales \n 23 , apparently dominates in hypersaline conditions 23 – 25 . In contrast to the extremely halophilic haloarchaea, these microbes only tolerate saturated salt conditions but optimally grow at moderate salinity (below 2–3 M Na + ) using organic compounds for osmotic balance (“salt-out” strategy) 26 , 27 . Our recent study of methanogenesis in hypersaline soda lakes identified methylotrophic methanogenesis as the most active pathway. In addition, culture-independent analysis of the mcr A gene, a unique marker of methanogens, identified a deep lineage that is only distantly related to other methanogens 28 . We observed no growth of these organisms upon addition of substrates for the classical methanogenic pathways and concluded that they required distinct growth conditions. Here we identify such conditions and describe the discovery and physiological, genomic and phylogenetic features of a previously overlooked group of extremely halophilic, methyl-reducing methanogens."
} | 1,621 |
35380743 | PMC9323447 | pmc | 2,092 | {
"abstract": "Abstract A common way for bacteria to cooperate is via the secretion of beneficial public goods (proteases, siderophores, biosurfactants) that can be shared amongst individuals in a group. Bacteria often simultaneously deploy multiple public goods with complementary functions. This raises the question whether natural selection could favour division of labour where subpopulations or species specialize in the production of a single public good, whilst sharing the complementary goods at the group level. Here we use an experimental system, where we mix engineered specialists of the bacterium Pseudomonas aeruginosa that can each only produce one of the two siderophores, pyochelin or pyoverdine and explore the conditions under which specialization can lead to division of labour. When growing pyochelin and pyoverdine specialists at different mixing ratios under different levels of iron limitation, we found that specialists could only successfully complement each other in environments with moderate iron limitation and grow as good as the generalist wildtype but not better. Under more stringent iron limitation, the dynamics in specialist communities was characterized by mutual cheating and with higher proportions of pyochelin producers greatly compromising group productivity. Nonetheless, specialist communities remained stable through negative frequency‐dependent selection. Our work shows that specialization in a bacterial community can be spurred by cheating and does not necessarily result in beneficial division of labour. We propose that natural selection might favour fine‐tuned regulatory mechanisms in generalists over division of labour because the former enables generalists to remain flexible and adequately adjust public good investments in fluctuating environments.",
"introduction": "1 INTRODUCTION Division of labour is a defining feature of major transitions in evolution (Bourke, 2011 ; Ispolatov et al., 2012 ; Rueffler et al., 2012 ; Szathmáry & Smith, 1995 ). At all levels of biological organization, division of labour underlies specialization, whereby a generalist entity transits to multiple specialist entities, performing complementary tasks. Examples include the transition from (i) generalist cells to specialized soma and germlines in multicellular organisms and (ii) generalist individuals to specialized castes in eusocial insect societies (Bourke, 2011 ; Cooper & West, 2018 ). Division of labour entails that specialized individuals (or entities) perform cooperative tasks that benefit others in the group and that specialization leads to an inclusive fitness benefit to all individuals (West & Cooper, 2016 ). Consequently, division of labour is more likely to evolve and remain stable when the interests of the specialized individuals are aligned, which is typically the case when relatedness is high and the traits individuals specialize on are essential for survival and reproduction (Fisher et al., 2013 ; West & Cooper, 2016 ). Despite this clear conceptual framework, it is often difficult to experimentally demonstrate that a particular form of specialization constitutes division of labour. Due to their high amenability, bacterial (and other microbial) systems have become popular models to test basic theory and to experimentally investigate division of labour and the underlying evolutionary forces (Ackermann et al., 2008 ; Armbruster et al., 2020 ; Dragoš, Kiesewalter et al., 2018 ; Dragoš, Martin et al., 2018 ; van Gestel et al., 2015 ; Giri et al., 2019 ; Kim et al., 2016 ; West & Cooper, 2016 ; Yanni et al., 2020 ). Bacteria are social organisms, and the most common form of cooperation involves the secretion of beneficial compounds that can be shared as public goods (West et al., 2007 ). For example, bacteria secrete enzymes to digest extra‐cellular polymers and proteins, quorum‐sensing molecules for communication, biosurfactants to enable swarming on wet surfaces and siderophores to scavenge iron (Drescher et al., 2014 ; Kramer et al., 2020 ; Roman et al., 2017 ; Yan et al., 2019 ). Because multiple public goods might be required at the same time, a key question is whether bacterial populations can segregate into subpopulations, each specializing in one of the public goods and mutually sharing the beneficial compounds at the population level (Frank, 2013 ; Schiessl et al., 2019 ; West & Cooper, 2016 ). Indeed, the evolution of such division of labour has been described in several laboratory systems. For example, Bacillus subtilis populations segregate in biosurfactant‐producing and matrix‐producing cell types, which synergistically interact to enable sliding motility (van Gestel et al., 2015 ). Similarly, populations of Pseudomonas fluorescens rapidly segregate into two cell types, one producing a lubricant at the front of the colony and the other one pushing cells forward (Kim et al., 2016 ). Other studies enforced specialization through genetic engineering to study its fitness consequences. In Bacillus subtilis , enforced specialization between two strains producing either the structural protein TasA or the exopolysaccharide EPS lead to stable division of labour that outperformed the generalist wildtype in forming pellicle biofilms (Dragoš, Kiesewalter et al., 2018 ). Another study showed that a poor‐performing engineered cross‐feeding system between two Escherichia coli strains evolved into a productive, mutually beneficial division of labour during experimental evolution (Preussger et al., 2020 ). In contrast to these positive results, there are also examples where division of labour was either not stable or did not evolve (Dragoš, Martin et al., 2018 ; Schiessl et al., 2019 ). For example, a follow‐up study on the TasA vs. EPS specialization in B . subtilis showed that division of labour collapsed through the evolutionary reversion to an autonomous lifestyle (Dragoš, Martin et al., 2018 ). Moreover, a comparative study on hundreds of Pseudomonas isolates from natural pond and soil communities revealed high variation in public goods production amongst isolates, but little evidence for specialization and division of labour (Kramer et al., 2020 ;). Finally, a common phenomenon linked to public goods production is cheating, where mutants stop contributing to public goods production, yet exploit the public goods produced by others (Özkaya et al., 2017 ; Smith & Schuster, 2019 ). Co‐operators and cheaters often co‐exist in populations, which reflects a form of specialization (Driscoll et al., 2011 ; Maclean et al., 2010 ), but does often not represent division of labour because cheaters typically compromise group productivity (West & Cooper, 2016 ). The conditions that tip the balance in favour of division of labour as opposed to cheating are often unclear, but likely involve intrinsic (genetic architecture) and extrinsic (environmental) factors, determining the costs and benefits of specialization. In our study, we tackle this issue by using a model system where we experimentally enforce specialization between two bacterial strains and measure its fitness consequences across a range of environmental conditions. Specifically, we worked with two engineered mutants of the bacterium Pseudomonas aeruginosa , each specialized in the production of one of its two siderophores, pyoverdine or pyochelin (Schalk et al., 2020 ; Serino et al., 1997 ; Visca et al., 2007 ; Youard et al., 2011 ). Siderophores are produced in response to iron limitation, to scavenge this essential nutrient from the environment (Kramer et al., 2020 ; Kramer et al., 2020 ). Both pyoverdine and pyochelin can be public goods and shared between cells (Ross‐Gillespie et al., 2015 ) and although they are simultaneously produced, they take on slightly different functions (Dumas et al., 2013 ). Due to its relatively low iron affinity, pyochelin is primarily beneficial and increasingly produced under moderate iron limitation, whilst pyoverdine has a very high iron affinity and is more relevant and produced in high amounts under severe iron limitation (Dumas et al., 2013 ; Mridha & Kümmerli, 2021 ; Ross‐Gillespie et al., 2015 ). Using this experimental system, we tested whether enforced specialization alleviates the metabolic burden associated with the simultaneous production of both siderophores (intrinsic factor), whilst the benefits of the two molecules can still accrue to all individuals through molecule sharing at the population level. To this end, we mixed the pyochelin and pyoverdine specialists across a range of mixing ratios and compared the growth of the entire population relative to the fitness of the generalist wildtype strain producing both siderophores. If the two strains engage in division of labour, we expect population growth to be higher for specialized mixes than for generalist cultures. Conversely, if the two specialists engage in mutual cheating, then we expect population growth of mixed cultures to be compromised relative to generalist cultures. Moreover, we tested whether extrinsic environmental factors influence the propensity of specialization leading to division of labour or mutual cheating by manipulating the iron availability in the medium. We predict that the potential for division of labour is highest at intermediate iron availability where both siderophores are beneficial, whereas division of labour should not occur under high iron availability (where siderophores are no longer required) and is less likely to occur under severe iron limitation, as pyochelin is less useful. Finally, we explored whether the two specialists can co‐exist or whether one type drives the other one to extinction. For this purpose, we calculate the relative fitness of the two strains in co‐culture across a range of mixing ratios. Co‐existence would result in negative frequency‐dependent fitness patterns, where both specialists win the competition when initially rare in the population and consequently settle at an equilibrium frequency.",
"discussion": "4 DISCUSSION In our study we tested whether enforced specialization in the production of public goods can lead to division of labour. We used the siderophores pyochelin and pyoverdine in the bacterium P . aeruginosa as our model system. We mixed specialist strains producing either of the two siderophores at different starting frequencies and subjected them to a gradient of iron limitations. The productivity of the mixes was compared to that of the generalist wildtype and the relative frequency of the specialist were quantified after a 24‐h growth cycle. We found that (i) specialization increased group productivity under iron‐rich conditions where siderophores are less important, (ii) specialist and wildtype cultures performed equally well under moderate iron limitation where both pyochelin and pyoverdine are useful, (iii) specialization reduced group productivity in the environments where iron availability was low and pyoverdine is the more important siderophore, (iv) strains socially interacted with each other via the sharing of siderophores and lastly (v) specialists co‐existed in all environments settling on an intermediate ratio. Our observations indicate that enforced specialization does not lead to an efficient division of labour but fosters co‐existence through negative frequency‐dependent selection based on mutual cheating under most conditions tested. We found that mixtures of the two specialists performed better than the generalist wildtype in iron‐rich environment (CAA + 100 µM FeCl 3 ). However, this finding unlikely represents division of labour because siderophores are not important for iron scavenging under such iron‐replete conditions (Dumas et al., 2013 ; Ochsner et al., 1995 ; Serino et al., 1997 ). Instead, our results suggest that being a generalist is costly under iron‐replete conditions. This interpretation is supported by recent findings at the single cell level showing that the wildtype expresses both pyochelin and pyoverdine synthesis genes at background levels even in iron‐rich environments (Mridha & Kümmerli, 2021 ). Consequently, specialist communities are able to minimize these residual metabolic costs, whilst the generalist cannot. We further found that specialist communities were as productive as the generalist wildtype cultures in a medium where the overall iron availability is low, but iron is not bound to a chelator (plain CAA). This observation matches our prediction that division of labour is most likely to arise in environments where both siderophores are beneficial. Pyochelin is beneficial because it is relatively cheap to produce yet is still efficient in iron scavenging despite its relatively low iron affinity (K f = 10 5 M −1 ) (Youard et al., 2011 ). Pyoverdine is beneficial in any case due to its high iron affinity (K f = 10 24 M −1 ) (Visca et al., 2007 ), but is produced at moderate amounts presumably because of its relatively high production costs (Dumas et al., 2013 ). Our data, therefore, suggest that the two iron‐scavenging strategies successfully complement each other under moderate iron limitation. However, we might ask why the two specialists only reach wildtype productivity and do not engage in a more efficient division of labour? One explanation is that our strains are imperfect specialist. Pyoverdine and pyochelin synthesis occurs via non‐ribosomal peptide synthesis, which involves a series of enzymes (Visca et al., 2007 ; Youard et al., 2011 ). Whilst our mutants lack parts of the siderophore synthesis machinery, which turns them into phenotypic specialists, molecular studies have shown that the remaining part of the synthesis machinery is still partially active resulting into residual metabolic costs (Tiburzi et al., 2008 ). We propose that naturally evolving specialists might alleviate these costs, which could indeed give rise to division of labour under moderate iron limitation. In contrast, we observed that specialist communities had significantly lower productivity relative to the generalist wildtype cultures in media where iron was bound to a strong iron chelator (CAA + 2–2’‐bipyridyl). Whilst the importance of pyoverdine increases under strong iron limitation the one of pyochelin diminishes (Dumas et al., 2013 ). This misbalance in their relevance might prevent division of labour, as is demonstrated by our results that community productivity sharply declines with higher initial frequencies of pyochelin specialists in the mixes, showing that they are a burden for the community because they do not contribute to the pool of the more beneficial pyoverdine. Altogether, our productivity analyses show that the conditions under which division of labour, involving siderophore sharing, can evolve is narrow and restricted to environments with moderate iron limitation. A key prerequisite for division of labour to occur is that specialists interact, and in our case, exchange the two siderophores at the community level. Our results support this view as we found that the relative fitness (Figure 2 ) and the growth rate (Figure S3 ) of each specialist is dependent on the presence and frequency of the other specialist in the community. These findings corroborate previous results that both pyoverdine and pyochelin can be shared between cells and confer fitness benefits to the receivers (Griffin et al., 2004 ; Ross‐Gillespie et al., 2015 ; Sathe et al., 2019 ). Despite the sharing of the two siderophores, enforced specialization in our system did not result in division of labour but rather stimulated cheating. We reason that this is because the two siderophores have different costs and benefits, especially under strong iron limitation. Whilst pyochelin is relatively cheap to produce, it comes with little benefit to the community when iron is bound to a strong chelator (see also Figure 1 ). Conversely, pyoverdine is relatively costly for the producers yet generates a greater benefit to the community as it manages to scavenge iron even when it is bound to a strong chelator. This imbalance in costs and benefits puts pyochelin producers at an advantage as they can exploit the beneficial pyoverdine whilst saving costs by making the cheaper (but less useful) pyochelin. Consequently, pyochelin producers act predominantly as cheaters under severe iron limitation, reaping relative fitness benefits (Figure 2 ) whilst compromising community productivity (Figure 3 ). But our effect size analysis suggests that pyoverdine producers can also cheat on pyochelin producers, especially when rare (Figure S3 , Table S5 ). At low pyoverdine producer frequencies, the balance tips and pyochelin producers reach lower frequencies than expected based on their monoculture growth, especially under moderate (plain CAA) and strong (CAA + 2–2’‐bipyridyl) iron limitation (Figure S3 , Table S5 ). FIGURE 3 Relative fitness ln( v) of pyochelin producers in competition with pyoverdine producers declines with higher starting frequencies of pyochelin producers in the population. The colour code depicts mixing ratios from 0% pyochelin producers (pyoverdine monoculture) in blue to 100% pyochelin producers (pyochelin monoculture) in red, with the increase in frequency of pyochelin producers in the mixes being denoted by a gradual shift from blue to red. The findings suggest that there is negative frequency‐dependent selection in all media conditions. Competitions occurred between pyochelin (PAO1Δ pvdD :: gfp ) and pyoverdine (PAO1Δ pchEF :: mcherry ) producers over 24 hours in media differing in iron availability. Values of ln( v ) > 0, ln( v ) = 0, or ln ( v ) < 0 indicate that pyochelin producers won, did equally well or lost in competition against pyoverdine producers respectively. Boxplots represent the median with 25th and 75th percentiles, and whiskers show the 1.5 interquartile range across 12 independent replicates The observed fitness dynamics between the pyochelin and pyoverdine specialists are analogous to the snowdrift game, a popular concept from game theory (Doebeli & Hauert, 2005 ). The two siderophore specialists behave as opponents and the best strategy for each specialist is to minimize cooperation (i.e. produce little siderophores) and rely on the cooperative acts (i.e. siderophores) of the opponent. However, if the opponent does not cooperate it is better to cooperate oneself (i.e. produce and rely on one's own siderophores), otherwise there would be no benefit at all (i.e. no iron would be scavenged). Snow‐drift games are characterized by the co‐existence of the two strategies (cooperation and cheating) via negative frequency‐dependent selection as observed in our system. Taken together, these results suggest that interactions between specialists are characterized by mutual cheating reminiscent of snowdrift game dynamics and not division of labour under iron limitation. The question is whether co‐existence would be stable in the long run? One conceivable option is that a double mutant would arise that cheats on both siderophores, potentially bringing the communities to collapse under strong iron limitation. However, a previous experimental evolution study yielded no evidence for such a scenario (Ross‐Gillespie et al., 2015 ), but showed stable co‐existence of the wildtype generalist with a pyoverdine mutant that overproduced pyochelin and a second mutant that produced both siderophores at intermediate levels. An alternative option is that the interactions between specialists are lost through the evolution of an independent lifestyle, as shown to occur in the enforced biofilm specialization system in Bacillus subtilis (Dragoš, Martin et al., 2018 ). In our case, this could happen through the evolution of a superior siderophore system that is neither shareable nor exploitable by other community members (Lee et al., 2012 ). We showed that enforced specialization saves metabolic costs associated with the production of the two siderophores (Figure 3 , iron‐rich conditions). However, this cost saving did not result in an efficient division of labour under iron‐limited conditions. We propose that this is because P . aeruginosa has evolved sophisticated regulatory mechanisms to fine‐tune siderophore investment in response to environmental conditions. Such regulatory mechanisms also save costs (Darch et al., 2012 ; Kümmerli & Brown, 2010 ; Xavier et al., 2011 ) and could represent an alternative adaptive strategy to division of labour. Siderophore regulation in P . aeruginosa involves three levels. At the top level, Fur (ferric uptake regulator) blocks siderophore synthesis when iron accumulates within cells but loses its inhibitory effect when intra‐cellular iron reserves become depleted (Escolar et al., 1999 ; Leoni et al., 1996 ; Ochsner & Vasil, 1996 ; Youard et al., 2011 ). Fur repression seems to act more strongly on pyoverdine than pyochelin, allowing for preferential pyochelin production under moderate iron limitation (Dumas et al., 2013 ). The second level of regulation involves membrane‐embedded signalling cascades, where incoming siderophore‐iron complexes trigger a positive feedback loop that increases siderophore production in response to successful scavenging events (Lamont et al., 2002 ; Michel et al., 2007 ). The third level is poorly explored at the molecular level, yet is hierarchical in nature, whereby pyoverdine production suppresses pyochelin synthesis intracellularly (Dumas et al., 2013 ), allowing for preferential pyoverdine production under strong iron limitation. Such regulatory circuits could outperform genetic division of labour because individuals remain generalists and thus flexible in responding to environmental fluctuations (Butaitė et al., 2018 ), whilst genetic division of labour requires specialist partners to be present in the environment in adequate ratios and being able to exchange their goods, conditions that might often not be met. In conclusion, our study indicates that enforcing specialization with regard to siderophore production does not lead to beneficial division of labour in P . aeruginosa but leads to the stable co‐existence of the two specialists through mutual cheating. Our results suggest that there are at least three factors that could constrain the evolution of division of labour with regard to public good exchange in bacterial population. First, an imbalance in the costs and benefits of two public goods can promote cheating rather than an efficient division of labour. Second, division of labour might only be beneficial under specific environmental conditions and because environmental fluctuations are common in natural habitats generalist strategies might be favoured. Third, the evolution of fine‐tuned regulatory circuits can compensate for the costs associated with pursuing a generalist strategy, and at the same time allows generalist to flexibly respond to environmental fluctuations. Given these constraints, genetic division of labour might be harder to evolve in bacteria than previously thought, especially when environmental conditions fluctuate in unpredictable ways, making the evolution of fine‐tuned regulatory circuits the better option."
} | 5,813 |
36506817 | PMC9730831 | pmc | 2,093 | {
"abstract": "With the development of technology, Moore's law will come to an end, and scientists are trying to find a new way out in brain-like computing. But we still know very little about how the brain works. At the present stage of research, brain-like models are all structured to mimic the brain in order to achieve some of the brain's functions, and then continue to improve the theories and models. This article summarizes the important progress and status of brain-like computing, summarizes the generally accepted and feasible brain-like computing models, introduces, analyzes, and compares the more mature brain-like computing chips, outlines the attempts and challenges of brain-like computing applications at this stage, and looks forward to the future development of brain-like computing. It is hoped that the summarized results will help relevant researchers and practitioners to quickly grasp the research progress in the field of brain-like computing and acquire the application methods and related knowledge in this field.",
"introduction": "Introduction Achieving artificial intelligence as the major goal of mankind has been at the top of the heated debate. Since the Dartmouth Conference in 1956 (McCarthy et al., 2006 ), the development of AI has gone through three waves. They can be roughly divided into four basic ideas: symbolism, connectionism, behaviorism, and statism. These different ideas have captured some of the characteristics of “intelligence” in different aspects, but only partially surpassed the brain of humans in the aspect of function. In recent years, the computer hardware base has become more perfect, and deep learning has revealed its huge potential (Huang Y. et al., 2022 ; Yang et al., 2022 ). In 2016, AlphaGo defeated Lee Sedol, the ninth-degree Go master, which marked that the third wave of artificial intelligence technology revolution has reached its peak. In particular, the realization of AI has become one of the wrestling points of national power competition. In 2017, China released and implemented a new generation of artificial intelligence development planning. In June 2019, the United States released the latest version of the National Artificial Intelligence Research and Development Strategic Plan (Amundson et al., 1911 ). Europe has also identified AI as a priority development project: in 2016, the European Commission proposed a legislative motion on AI; in 2018, the European Commission submitted the European Artificial Intelligence (Delponte and Tamburrini, 2018 ), and published Coordinated Plan on Artificial Intelligence with the theme “Made in Europe with Artificial Intelligence.” Achieving artificial intelligence requires more powerful information processing capabilities, but relying on the current classical computer architecture cannot meet the huge amount of data computing. The classical computer system has encountered two major bottlenecks in its development: the storage wall effect due to von Neumann structure and Moore's law will fail in the next few years. On the one hand, traditional processor architecture is inefficient and energy intensive. When dealing with intelligent problems in real-time, it is impossible to construct suitable algorithms for processing unstructured information. In addition, the mismatch between the rate of programs or data transferred back and forth and the rate of the central processor processing information leads to a storage wall effect. On the other hand, as the chip's size assembly gets closer to the size of a single atom, the devices are getting closer to the limits of their respective physical miniaturization. So, the cost of performance enhancement will become higher and the technical implementation will become more difficult. Therefore, researchers put their hopes on brain-like computing in order to break through the current technical bottleneck. Early research in brain-like computing followed the traditional computer manufacturing process that we first recognize how the human brain works and develop a neuromorphic computer based on the theory. But after more than a decade of research, mankind is almost standing still in the field of brain science. So, the path of theory before technology was abandoned by mainstream brain-like research. Looking back at human development, we see that many technologies precede theories. For example, in the case of airplanes, we can build the physical object before conducting research to refine the theory. Based on it, researchers adopted structural brain analogs: using existing brain science knowledge and technology to simulate the structure of the human brain, and then refining the theory after success. This article first introduces the idea behind the research significance of brain-like computing in a general way. Then we summarize the research history and compare the current research progress with analysis and outlook. The article structure is shown in Figure 1 . Figure 1 The structure of the article is as follows the analysis of relevant models, the establishment of related platforms, implementation of related applications, challenges, and prospects."
} | 1,282 |
28469244 | PMC5431162 | pmc | 2,094 | {
"abstract": "In legume- Rhizobium symbioses, specialised soil bacteria fix atmospheric nitrogen in return for carbon. However, ineffective strains can arise, making discrimination essential. Discrimination can occur via partner choice, where legumes prevent ineffective strains from entering, or via sanctioning, where plants provide fewer resources. Several studies have inferred that legumes exercise partner choice, but the rhizobia compared were not otherwise isogenic. To test when and how plants discriminate ineffective strains we developed sets of fixing and non-fixing strains that differed only in the expression of nifH – essential for nitrogen fixation – and could be visualised using marker genes. We show that the plant is unable to select against the non-fixing strain at the point of entry, but that non-fixing nodules are sanctioned. We also used the technique to characterise mixed nodules (containing both a fixing and a non-fixing strain), whose frequency could be predicted using a simple diffusion model. We discuss that sanctioning is likely to evolve in preference to partner choice in any symbiosis where partner quality cannot be adequately assessed until goods or services are actively exchanged.",
"introduction": "Introduction Across the globe primary productivity is nitrogen limited 1 . This limitation has been overcome for plants in the family Fabaceae (commonly known as legumes) through a mutualistic association with nitrogen-fixing bacteria collectively called rhizobia 2 . The nitrogen provided through this symbiosis makes legumes rich in protein and important crops in human diets 3 . But, as ineffective strains will inevitably arise through mutation, there is the potential for the relationship to break down. Ineffective strains are known to be common, at least in some situations, which for agricultural legumes means poor yields and reduced nutritional quality 4 , 5 . Theory predicts that ineffective strains could be successful within legume – Rhizobium symbioses for two reasons 6 , 7 . First, rhizobia are not transmitted directly from parent plant to offspring. Instead, plants acquire rhizobia from the soil through an intricate signalling process in which bacteria enter specialized root nodules, where they fix nitrogen in return for plant-derived carbon 8 . This horizontal transmission means that rhizobial fitness is not perfectly aligned with the fitness of the host plant 7 . Second, although each nodule is usually occupied by the clonal descendants of a single Rhizobium \n 9 , 10 , a plant is usually infected by multiple rhizobial strains 11 , 12 . Thus, a non-fixing strain can potentially thrive by taking plant resources while leaving the costly process of nitrogen fixation to others 7 , 13 . To prevent losing resources to ineffective rhizobial strains that provide little or no nitrogen, legumes have two options: partner choice or sanctions 7 . Partner choice is usually defined as any mechanism that allows detection of suitable partners before a mutualistic relationship is established 7 , 13 – 15 , while sanctioning is a mechanism to discriminate against low-quality partners once the relationship is underway 6 , 7 , 13 – 16 (although confusingly ‘partner choice’ has also been used to describe a broader concept which includes sanctioning) 13 , 17 , 18 . Partner choice might seem to be the more attractive option as resources are not wasted setting up a relationship that is doomed to fail. But crucially, effective partner choice requires accurate assessment of the quality of partners in advance 13 . This is likely to be problematical for any symbiosis in which key traits are not manifested prior to the relationship being established. For example, in the legume- Rhizobium symbiosis, nitrogen fixation does not begin until the bacteria have entered the roots and nodule formation is sufficiently advanced for rhizobia to have differentiated into nitrogen-fixing bacteroids 8 . Once nodules are established they can be sanctioned, if they prove to be ineffective, by cutting off their supply of carbon, oxygen or other nutrients and this has been demonstrated empirically using argon gas to force nodules to fix less nitrogen 19 – 21 . Despite the empirical support for sanctioning and its apparent advantages, there are nevertheless several studies that claim evidence for partner choice 7 , 14 , 15 . However, the interpretation of these studies is problematic because the tested strains are rarely isogenic – meaning that strains differ in several traits, and not just in how much nitrogen they provide. Most importantly, strains are likely to differ in their competitiveness in colonizing plant roots and forming nodules. A range of traits affects competitiveness: examples include motility 22 , production of antibiotics 23 and the secretion of proteins and polysaccharides involved in biofilm formation and root attachment 24 . Such differences in competitiveness explain why poorly-fixing strains can also end up occupying a higher proportion of nodules – a problem that is often encountered when developing effective strains for use in agricultural settings 4 , 25 , 26 . Thus comparing the nodulation success of naturally occurring strains is difficult to interpret as a test of partner choice. To test whether plants can directly assess the effectiveness of potential rhizobia prior to nodulation we created a non-fixing mutant from a fixing strain and compared their success in colonising pea plant nodules. There are several key genes involved in nitrogen fixation 8 , any of which could undergo a mutation that would render the gene non-functional and hence transform the fixing into a non-fixing strain. We chose the nifH gene in Rhizobium leguminosarum bv. viciae ( Rlv ) 3841, and created a non-fixing mutant that was otherwise identical to its fixing parent strain. We then assessed when and how the plant discriminated between the two strains. To identify strains, they were marked with gusA or celB marker genes, rendering strains magenta or blue (respectively) following the application of a simple post-harvest staining protocol. Insertion of marker genes solves a secondary problem as it is usually extremely time-consuming to identify different strains using antibiotic markers and this limits the number of nodules that can be assessed. One possible complication is that a non-fixing strain can potentially thrive via mixed nodules (where two different bacteria have entered and colonised). If a non-fixing strain can take advantage of mixed nodules to increase its fitness at the expense of the fixing strain, then this would provide a route by which non-fixing strains could increase in frequency. Currently little is known about the frequency at which mixed nodules occur, and the relative fitness of strains within mixed nodules. The staining protocol rendered mixed nodules easily visible, so we assessed the frequency of mixed nodules under different inoculation densities. We discuss mixed nodules, partner choice and sanctioning in the context of the evolutionary stability of the legume- Rhizobium symbiosis.",
"discussion": "Discussion We created a non-fixing but otherwise isogenic mutant to mimic a process that might occur in nature, where a mutation arises in a fixing rhizobial strain, rendering it ineffective. We found that pea plants could not discriminate between these fixing and non-fixing strains prior to nodule formation. Pea plants were therefore unable to detect whether the potential partner was effective at fixing nitrogen and could not prevent the formation of non-fixing nodules. Our results are supported by an early study using similar isogenic strains 27 , but this study was severely limited in sample size and has therefore been overlooked. Our results indicate that partner choice is not a robust mechanism against ineffective strains as pea plants were unable to prevent non-fixing strains from entering. It could be argued that legumes may use genes other than nifH to assess the nitrogen fixation capacity of fixing strains before nodule formation; however we believe that this is unlikely. A mutation rendering a strain less effective can arise in any gene and effective partner choice would then require a mutation in the plant genome to detect this change. If the new mutation stops the cheat from entering, then it will spread through the plant population; however, given that rhizobial generation times are much faster than host plant generation times 28 , it seems that the host plants will be locked in an evolutionary arms race that they are doomed to lose; hence partner choice seems to be an ineffective way to stabilise the mutualism in the long term 21 . Furthermore, partner choice is susceptible to dishonest signals 13 . In contrast, we found that pea plants did discriminate against ineffective strains via sanctioning, which has been previously reported using argon gas to replace atmospheric nitrogen 19 – 21 . In our experiment, nodules containing the non-fixing strain were roughly half the size of fixing nodules, indicating reduced plant resources. In contrast to the case of partner choice, sanctioning can stabilise the mutualism in the long-term. If a mutation arises that allows a plant to detect and sanction a partner that is not delivering the goods it would be effective against a wide variety of future ineffective strains. Thus, sanctioning allows an instantaneous response to ineffective strains and does not require specific recognition genes or rely on honest signals. Sanctioning is therefore a more robust 13 mechanism against ineffective strains and can provide long-term stability to legume- Rhizobium mutualisms 16 , 29 . However, any discussion of sanctioning should take into account both plant and rhizobial fitness. Sanctioning can only be selected for when it saves plant resources and thus increases plant fitness. The reduced nodule size that we and others 20 , 21 have seen indicates that plants allocate fewer resources to non-fixing nodules. Whether this reduced resource allocation also reduces rhizobial fitness and thus stabilizes the mutualism on evolutionary time-scales is more difficult to establish. This may depend on whether or not the nitrogen-fixing bacteroids can still reproduce (usually determinate nodules), or are terminally differentiated and unable to reproduce (indeterminate nodules) 30 . In studies using argon gas, reduced rhizobial fitness has been shown in both determinate 19 , 20 and indeterminate 21 non-fixing nodules. However, in a study using a non-fixing isogenic strain, the fitness of non-fixing rhizobia was not reduced in determinate soybean nodules up to five weeks old 31 . Because effects on rhizobial viability may emerge later in the sanctioning process, perhaps the best test would be a multigenerational experiment, where in the first generation plants are inoculated with both fixing and non-fixing strains, and new plants are then repeatedly grown in the same soil for several generations to see how quickly the non-fixing strain is eliminated. It would be of special interest to perform such an experiment with both indeterminate and determinate species. Evidence from other mutualisms suggests that whether partner choice evolves in preference to sanctioning critically depends on how well partners can assess quality prior to establishment of the mutualistic relationship including the potential for dishonest signalling. For example, clients of the cleaner fish Labroides dimidiatus have evolved partner choice to counter cheating by individuals that take healthy tissue while removing parasites 32 . Partner choice is highly effective in this mutualism because the quality of service is known from previous experience and there are repeated interactions between individuals. In contrast, sanctioning has evolved in mutualisms between yucca moths and fig wasps and their respective plant hosts. In both cases the insects deposit seed-eating larvae in the flowers of host plants in return for pollination. In these mutualisms, plants cannot prevent eggs being laid nor assess partner quality, hence sanctioning has evolved: flowers containing too many eggs 33 , or too little pollen 33 , 34 are selectively aborted. While our study shows that partner choice is not a robust mechanism to exclude ineffective strains, legumes do not form symbioses with all potential rhizobial strains. Instead, an extensive signalling process 35 – 37 between legumes and their rhizobial partners can impose a high degree of selectivity on the relationship 36 , 38 , 39 , although the degree of selectivity varies greatly among hosts 39 . There are two explanations for this selectivity that are commonly proposed and are not mutually exclusive. First, specificity may arise in order to prevent the entry of pathogenic bacteria which utilise similar signalling pathways to gain access to host roots 38 , 40 , 41 . Second, by fine-tuning signalling pathways to target rhizobia that are particularly effective for a specific host, legume species might achieve greater nitrogen-fixation efficiency 42 , 43 . This is likely to be true if host environments are sufficiently different that specialization by rhizobia is selected for. Support for specialization comes from the observation that a single rhizobial strain can vary greatly in its effectiveness among hosts 42 . This type of co-evolutionary process is separate from the need to avoid non-fixing rhizobia, which can arise by mutation at any time, in any strain, even those that are usually highly effective. That these two processes are indeed separate is supported by the fact that the genes involved in nitrogen fixation ( nif and fix genes) are only expressed once the symbiosis has been established 44 and are different from the signalling genes involved in infection ( nod genes) 45 . Currently, genomic analyses are shedding more light on the selective pressures affecting both legume and rhizobial genes 46 – 48 . Sanctioning is a robust mechanism against ineffective strains, but requires hosts to monitor partner quality and provide resources accordingly. Currently, little is known about the exact mechanism behind sanctioning in legume- Rhizobium symbioses, and whether it only takes place at the nodule level, or also occurs within nodules 13 , 16 . If sanctioning takes place at the nodule level, mixed nodules could be a way for ineffective strains to avoid sanctions 6 , 13 , 49 . Indeed certain endosymbionts, even those belonging to different genera and lacking any genes for nitrogen fixation, have been shown to co-infect nodules by “piggybacking” on the genuine symbionts as they infect the root hairs 50 . Whether mixed nodules allow ineffective strains to persist depends on the frequency of mixed nodules, and on the relative fitness of fixing and non-fixing strains within mixed nodules. Estimates of the frequency of mixed nodules in the literature range from 2% to 74% 6 , 51 . Our findings at least partly explain this variability as the frequencies we found could be adequately represented by a simple diffusion model, which predicts that more mixed nodules are expected: (1) at high rhizobial densities; (2) when the proportions of different strains are similar; and (3) when rhizobia diffuse more easily, which might occur, for example, under wet conditions. Although the staining technique is valuable in identifying strains and characterising mixed nodules, it could not be used reliably to assess fitness of rhizobial strains within mixed nodules. Further work on the mechanism of sanctioning, how it is affected by external conditions such as soil nitrogen, and how it affects rhizobial fitness will help illuminate how the legume- Rhizobium mutualism has persisted for much longer than humans have been around to reap its benefits."
} | 3,964 |
29560024 | PMC5858145 | pmc | 2,096 | {
"abstract": "Background Photobiological H 2 production has the potential of becoming a carbon-free renewable energy source, because upon the combustion of H 2 , only water is produced. The [Fe–Fe]-type hydrogenases of green algae are highly active, although extremely O 2 -sensitive. Sulphur deprivation is a common way to induce H 2 production, which, however, relies substantially on organic substrates and imposes a severe stress effect resulting in the degradation of the photosynthetic apparatus. Results We report on the establishment of an alternative H 2 production method by green algae that is based on a short anaerobic induction, keeping the Calvin–Benson–Bassham cycle inactive by substrate limitation and preserving hydrogenase activity by applying a simple catalyst to remove the evolved O 2 . Cultures remain photosynthetically active for several days, with the electrons feeding the hydrogenases mostly derived from water. The amount of H 2 produced is higher as compared to the sulphur-deprivation procedure and the process is photoautotrophic. Conclusion Our protocol demonstrates that it is possible to sustainably use algal cells as whole-cell catalysts for H 2 production, which enables industrial application of algal biohydrogen production. Electronic supplementary material The online version of this article (10.1186/s13068-018-1069-0) contains supplementary material, which is available to authorized users.",
"discussion": "Discussion The potential energy conversion efficiency from sunlight to H 2 by green algae is in the range of 10–13% [ 14 , 72 ]. However, in nature H 2 production lasts only for a few minutes due to the inhibition of hydrogenases by the evolved O 2 [ 20 , 27 ]. Early studies on algal H 2 production based on dark anaerobic incubation were typically unable to sustain the initial high rates of H 2 production for more than a few hrs, if not resorting to continuous flushing with helium [ 31 , 65 , 68 ]. Later, sulphur deprivation became the method of choice to induce long-term H 2 production [ 15 , 17 , 42 ]. However, sulphur deprivation has several drawbacks which impede its industrial application [ 20 , 72 ]: the procedure requires several washing steps; H 2 production starts with a delay of about 2 days; it is largely dependent on acetate (H 2 production can be induced under photoautotropic conditions as well, but with a much lower efficiency [ 18 – 20 ]); it necessitates the inactivation of PSII; and it results in the degradation of the photosynthetic machinery. Recovery following the terminal phase of H 2 production by re-additions of sulphur was incomplete and could be performed only a few times [ 73 ]. The future of this biotechnology relies on the development of a novel approach at least as efficient as the sulphur deprivation procedure, and which could solve most of the present issues limiting its applicability. Here we report on the establishment of a photoautotrophic and sustainable H 2 production system in C. reinhardtii , and demonstrate the applicability of algal cells as whole-cell catalysts for H 2 production. This new protocol shares the early approach by Gaffron and Rubin [ 31 ] to induce hydrogenase activity by dark anaerobic treatment and keep the CBB cycle inactive by substrate limitation. As an important addition, the protocol applies a simple O 2 absorbent that preserve hydrogenase activity for several days (Fig. 4 ). This protocol has fundamental advantages relative to the earlier methods, namely that (i) following a few hours of anaerobic dark incubation, H 2 production starts promptly upon illumination (Fig. 1 b); (ii) as opposed to sulphur deprivation, it does not require media exchange (Fig. 7 ); (iii) it does not depend on starch degradation and does not require acetate, thus it is photoautotrophic; (iv) because no organic carbon source is required, the risk of bacterial contamination is low; (v) the cultures remain photosynthetically active during the H 2 production phase (Fig. 6 ) and they can be easily recovered afterwards (Fig. 7 ); (vi) it is based on linear electron transport and the electrons originate mostly from the water-splitting activity of PSII, as demonstrated by a DCMU-treatment (Fig. 2 ), and has relatively high light-to-H 2 energy conversion efficiencies (Table 1 ); (vii) during the growth phase, CO 2 , an industrial by-product, can be utilized; and, (viii) it can make use of relatively high light intensities (here, approx. 320 µmol photons/m 2 /s). The maximum H 2 production yield achieved using this protocol was approx. 200 µl H 2 /ml culture in 96 h, which is almost four times higher than the yield of sulphur-deprived cultures at equal chl content and illumination conditions (compare Figs. 1 d, 4 b), and it is in the same range as observed earlier for sulphur deprivation experiments using PBRs with even illumination [ 42 ], but having the drawbacks listed above. We expect the yield of H 2 production achievable upon anaerobic induction in minimal media to be further improved using advanced PBR designs, including optimized gas-to-liquid ratio, illumination and mixing conditions and efficient removal of the produced gases. There is also a high potential in applying this protocol to various photosynthetic mutants possessing, e.g. truncated light-harvesting antennae [ 74 ], or a high PsbA protein content [ 75 ]. Cyclic electron transport competes for the electrons with HydA (reviewed by [ 9 ]), thus its downregulation may entail a further increase in H 2 production under our conditions as well. Keeping the CBB cycle inactive was achieved by substrate limitation; it has been shown earlier that the CBB cycle represents a competing pathway for H 2 production (e.g. [ 31 , 47 , 49 , 68 ]) and that redirecting the electrons towards HydA from FNR may enhance the rate of H 2 production [ 76 , 77 ]. The present findings show that by imposing substrate limitation on the CBB cycle, the electrons are largely transferred to HydA, with the lack of carbon sources facilitating the establishment of hypoxia. The effects of CO 2 and FCCP additions (Fig. 2 ) and the relatively reduced PQ-pool (Additional file 4 : Fig. S3) indicate that the mechanism occurs by “photosynthetic control” [ 51 , 54 , 78 ]: since the hydrogenases are less effective at accepting electrons than the CBB cycle, the lumen is acidified and the photosynthetic electron transport is decelerated at the cytb 6 f complex. This results in a reduced PQ-pool, which entails a high charge recombination rate in PSII, resulting in diminished O 2 evolution (Fig. 2 i, j). Another key factor to reach a sustained H 2 production is to protect the hydrogenases from O 2 , which may also shift the balance between O 2 and H 2 production, established by the above-mentioned “photosynthetic control”. We applied an iron-salt-based O 2 absorbent, which decreased the O 2 concentration in the headspace below 0.1%. This very low concentration of O 2 was still inhibitory for hydrogenases (Fig. 5 ), thus it is desirable to test even more advanced materials in the future, as for instance crystalline salts of cationic multimetallic cobalt complexes [ 79 ]. By further decreasing the O 2 level, we expect that hydrogenase activity would be better preserved and act as a more effective electron sink; as a result, lumen acidification and “photosynthetic control” will be attenuated, and the yield of H 2 production further increased. Engineering hydrogenases to tolerate a few percent of O 2 [ 14 ] could also be a successful strategy to further increase the efficiency of H 2 production."
} | 1,908 |
35096655 | PMC8795689 | pmc | 2,097 | {
"abstract": "Biofilms have been established as an important lifestyle for bacteria in nature as these structured communities often enable survivability and persistence in a multitude of environments. Francisella tularensis is a facultative intracellular Gram-negative bacterium found throughout much of the northern hemisphere. However, biofilm formation remains understudied and poorly understood in F. tularensis as non-substantial biofilms are typically observed in vitro by the clinically relevant subspecies F. tularensis subsp. tularensis and F. tularensis subsp. holarctica (Type A and B, respectively). Herein, we report conditions under which robust biofilm development was observed in a stochastic, but reproducible manner in Type A and B isolates. The frequency at which biofilm was observed increased temporally and appeared switch-like as progeny from the initial biofilm quickly formed biofilm in a predictable manner regardless of time or propagation with fresh media. The Type B isolates used for this study were found to more readily switch on biofilm formation than Type A isolates. Additionally, pH was found to function as an environmental checkpoint for biofilm initiation independently of the heritable cellular switch. Multiple colony morphologies were observed in biofilm positive cultures leading to the identification of a particular subset of grey variants that constitutively produce biofilm. Further, we found that constitutive biofilm forming isolates delay the onset of a viable non-culturable state. In this study, we demonstrate that a robust biofilm can be developed by clinically relevant F. tularensis isolates, provide a mechanism for biofilm initiation and examine the potential role of biofilm formation.",
"introduction": "Introduction \n Francisella tularensis is an intracellular Gram-negative bacterium found ubiquitously across the northern hemisphere and is the causative agent of tularemia. Tularemia is most common among small mammals, such as rabbits and voles, and can be transmitted via arthropod bites, inhalation or direct contact with an infected organism ( Ellis et al., 2002 ; Sjostedt, 2007 ). For humans, the glandular and ulceroglandular forms of tularemia are the most prevalent disease manifestations, typically occurring from an arthropod bite. Though less common, pneumonic forms of tularemia acquired from inhalation of aerosolized bacteria pose the most serious threat ( Oyston et al., 2004 ; Hepburn and Simpson, 2008 ). F. tularensis is of particular concern for human health due to its high morbidity, ease of aerosol inoculation and low infectious dose leading to the United States Centers for Disease Control classification as a Tier 1 select agent ( Dennis et al., 2001 ; Keim et al., 2007 ). Multiple Francisella tularensis subspecies have been identified, however, the most consequential to human health are F. tularensis subsp. tularensis (Type A) and Francisella tularensis subsp. holarctica (Type B) with the former generally regarded as being more virulent. Within North America, F. tularensis subsp. tularensis isolates are typically found in association with more terrestrial environments. In contrast, F. tularensis subsp. holarctica tends to be more widely distributed throughout both North America and Eurasia, often in association with aquatic environments ( Jellison, 1974 ; Oyston et al., 2004 ; Sjostedt, 2007 ). This divergence in environmental prevalence is also reflected in the associated arthropod vector. Namely, most tularemia cases in the United States are thought to occur from tick bites, while mosquitoes tend to be the drivers of European tularemia cases ( Zellner and Huntley, 2019 ; Tully and Huntley, 2020 ). Though tularemia is often associated with rodents and lagomorphs, these populations are not likely to serve as a long-term reservoir of F. tularensis as infected individuals either rapidly succumb to disease or clear the infection ( Oyston et al., 2004 ; Telford and Goethert, 2020 ). It is much more likely that F. tularensis persists in the environment outside a mammalian host as this bacterium has been found to maintain viability in an arthropod vector, as well as cold water for extended periods of time ( Forsman et al., 2000 ; Telford and Goethert, 2010 ; Mani et al., 2015 ). Further, protozoa have been shown to graze on both Type A and B strains of F. tularensis although it is unclear if F. tularensis is able to replicate within these host cells ( Abd et al., 2003 ; Thelaus et al., 2009 ; Buse et al., 2017 ). However, these environments present unique challenges for the bacterial cell to contend with, such as low nutrient availability, vector immune system and transstadial transmission, as well as environmental fluctuations including pH and temperature. Over the past few decades, an abundance of work has led to the conclusion that biofilms are a distinct lifestyle that is often integral for survival in array of environments. It is believed that many bacteria found in a natural setting, environmental or pathogenic, are likely in a biofilm state ( Stoodley et al., 2002 ; Hall-Stoodley et al., 2004 ). These bacterial communities encased within extracellular matrix material (ECM) showcase resilience when faced with barrage of adverse environmental conditions, such as rapid osmolarity changes, nutrient deprivation, or even predation. Most of what we know about biofilm development in Francisella comes from the use of F. novicida , a closely related opportunistic pathogen that is routinely used as a BSL-2 lab surrogate. It has been elegantly shown that F. novicida is able to form a robust biofilm in vitro on a variety of surfaces ( Margolis et al., 2010 ; van Hoek, 2013 ; Hennebique et al., 2019 ). While F. novicida can be a good model to study Francisella biology, the implied notion is that findings can be applied to F. tularensis despite there being stark differences between these two species; most notably virulence and ecology as F. novicida rarely causes disease in humans, is thought to be mainly waterborne, and lacks a known mammalian host or arthropod vector ( Kingry and Petersen, 2014 ). Differences also arise in the Francisella Pathogenicity Island (FPI) as F. tularensis typically harbors two FPIs while only one island is found in F. novicida ( Nano et al., 2004 ; Larsson et al., 2005 ). Contributing to the differences in pathogenicity, distinct structural modifications in O-antigen (O-ag) are found when comparing F. tularensis to F. novicida as the core oligosaccharide in F. tularensis lacks a glucose residue in the β-glucose branch and the tetra-saccharide repeat is flanked by distinct sugar moieties ( Vinogradov et al., 2002 ; Vinogradov et al., 2004 ; Gunn and Ernst, 2007 ). Further, phase variation of the O-ag between blue and grey forms has been described in F. tularensis , but this phenomena has yet to be observed in F. novicida ( Eigelsbach et al., 1951 ; Soni et al., 2010 ). Lastly, F. novicida has retained the genes necessary to produce cyclic dimeric GMP (cdGMP), a well-known secondary messenger that stimulates biofilm formation ( Zogaj et al., 2012 ). The genes required to synthesize and degrade cdGMP are absent in fully virulent F. tularensis , which is thought to confer a selective advantage to the intracellular life-cycle ( Zogaj et al., 2012 ). A limited amount of studies have examined biofilm formation in both Type A and Type B isolates of F. tularensis. These studies found that F. tularensis tends to form a biofilm with a sparse cell density over an extended period of time ( Margolis et al., 2010 ; Mahajan et al., 2011 ; Champion et al., 2019 ). Recently, Champion et al. used a targeted approach utilizing mutants with deficiencies in O-ag and capsule-like-complex to convincingly show that Type A and B strains are capable of forming a robust biofilm ( Champion et al., 2019 ). Interestingly, it has been well established that F. tularensis is able to phase vary components of the lipopolysaccharide (LPS), including the O-ag ( Cowley et al., 1996 ; Hartley et al., 2006 ; Soni et al., 2010 ). Indeed, very early studies also understood the importance of phase variation as heterogeneous cultures were found, and phase variants were sorted based on multiple colony morphologies ( Eigelsbach et al., 1951 ; Eigelsbach and Downs, 1961 ). Variants were sorted into blue (BV) and grey (GV) corresponding to “wild-type LPS” and “altered LPS”, respectively. It was also noted that virulence was severely impacted in GVs using a mouse model challenged intraperitoneally ( Eigelsbach et al., 1951 ). However, a GV identified by Soni et al. retained a similar level of virulence in a mouse model inoculated intranasally ( Soni et al., 2010 ). Given the potential virulence impact, it is unclear how F. tularensis may benefit from phase variation. In this study, we investigate the ability of F. tularensis to form biofilm. We show that both Type A and Type B isolates of F. tularensis are able to form a robust biofilm that is dependent on a heritable phenotypic switch. We provide evidence that pH may act as a regulator of biofilm development, and biofilm forming cultures remain culturable longer than the parental wild-type. These data presented in this manuscript are the first to describe a potential role for biofilm formation and phase variation of the O-ag in F. tularensis .",
"discussion": "Discussion Biofilm formation is thought to be the predominant lifestyle of bacteria found in the environment as the ECM affords the bacteria protection from hostile conditions and the altered metabolic activity associated with biofilm provides a buffer to nutrient stress. In particular, the ability of Francisella species to form biofilm has been questioned, namely because biofilms formed by pathogenic species have been described as being typically sparsely populated, erratic or weak when performed in a short time-frame in vitro ( Margolis et al., 2010 ; Mahajan et al., 2011 ; Champion et al., 2019 ). However as we show from this study, we were able to observe robust biofilm formation, but in a stochastic manner from a diverse set of F. tularensis strains to include both Type A and B strains. Other key takeaways from our study are demonstrating the ability of a F. tularensis culture to form biofilm was dependent on both pH and conversion to grey variants. Finally, we identified a subset of LVS F. tularensis grey variants as being able to constitutively form biofilm and also delay in the ability to revert to a viable non-cultural state when compared to the parent strain. The ability of the closely related Francisella surrogate strain F. novicida , which is a strict aquatic organism and not normally pathogenic for humans, to form biofilm is well established, and biofilm most likely provides this organism the ability to persist in the environment ( Durham-Colleran et al., 2010 ; Margolis et al., 2010 ; Champion et al., 2019 ). Also, the F. novicida genome has a gene cluster that encodes for proteins possessing diguanylate cyclase (DGC) and phosphodiesterase (PDE) domains involved in the synthesis and degradation of the secondary messenger cyclic di-GMP (cdGMP) ( Zogaj et al., 2012 ). cdGMP is a secondary messenger associated with controlling biofilm formation, along with other bacterial cellular processes ( Tamayo et al., 2007 ; Boyd and O’Toole, 2012 ). Type A and Type B F. tularensis isolates also have the ability to also reside in the environment, but in contrast to F. novicida , these Francisella species are highly pathogenic for higher level mammals. Furthermore, these pathogenic species of Francisella, as opposed to F. novicida described above, are missing the genes encoding the proteins involved in regulating the cdGMP ( Zogaj et al., 2012 ). Therefore, the ability and exact role of biofilm in survival or virulence for these Francisella species has been questioned. If biofilm plays a role in persistence and transmission of Francisella remains unknown, largely because most biofilm studies have been completed with F. novicida ( Tully and Huntley, 2020 ). In this study, we demonstrate that a robust biofilm can be formed quickly by Type A and Type B F. tularensis isolates despite the absence of the cdGMP signaling system. This finding alters the perception that these subspecies typically only form a low density, weak biofilm and has implications for future studies on the role of biofilm in virulent Francisella subspecies. Furthermore, this work also lends support to the notion that pathogenic Francisella species do not rely on a continuous infection cycle of vertebrates to serve as an environmental reservoir ( Telford and Goethert, 2020 ). Most bacteria in the environment are thought to be found in biofilm communities rather than free-living planktonically ( Costerton et al., 1978 ; Hall-Stoodley et al., 2004 ). Often with biofilm formation the bacterium inhabitants alter their metabolism, usually in favor of a quiescent state ( Stoodley et al., 2002 ). F. novicida has already been shown to readily form a robust biofilm on a variety of surfaces, including chitin, and it has been suggested that N-acetyl-D-glucosamine (GlcNAc), the end product of chitin hydrolysis, can serve as a carbon source in the absence of glucose ( Margolis et al., 2010 ). Notably, chitin is present within the exoskeleton of arthropods, such as ticks and mosquitoes, which are known vectors for transmission of tularemia. Thus, the question arises if F. tularensis present within arthropod vectors are in a biofilm? Additionally, we show in this study that while both Type A and B isolates stochastically formed biofilm, Type B strains appeared to more readily produce this product. Though, we only tested 5 isolates of Type B strains, perhaps this increased biofilm formation is due to the more aquatic based ecological niches typical for this subtype of F. tularensis . Distinct differences exist in the vector ecology between F. tularensis (Type A) and F. holarctica , (Type B), particularly in the United States. While Type A and B strains are found distributed across North America, Type A isolates are closely associated with lagomorphs and ticks often in more arid terrestrial settings. In contrast, Type B isolates are associated with deer flies, mosquitoes and rodents in aquatic conditions ( Staples et al., 2006 ; Keim et al., 2007 ; Kugeler et al., 2009 ). Mosquito larvae have also been found to readily graze on LVS biofilms in water, perhaps fostering environmental persistence as the bacteria were found to escape the midgut and colonize this host ( Mahajan et al., 2011 ). However, it has been suggested that mosquitoes and deer flies are not long term reservoirs of F. tularensis ( Sjostedt, 2007 ; Maurin and Gyuranecz, 2016 ). Microorganisms, such as protozoa, including free-living amoeba, have been shown to graze on both Type A and B strains of F. tularensis resulting in enhanced colonization of these organisms in water, though it is unclear if replication of F. tularensis occurs and/or is host specific ( Abd et al., 2003 ; Thelaus et al., 2009 ; Buse et al., 2017 ). A recent study by Golovliov et al. found that Type A and B isolates of F. tularensis failed to produce biofilm in 0.9% saline ( Golovliov et al., 2021 ). However, it is difficult to compare this study to ours given that biofilm formation is a metabolically active process employing the synthesis of macromolecules to build the ECM ( Mann and Wozniak, 2012 ; Hobley et al., 2015 ). Bacteria present in a natural aquatic setting would have more nutrients available than saline. Furthermore, bacteria are more likely to be present within the sediment in an aquatic environment which would also contain some level of nutrients. Certainly, further studies are needed to determine if F. tularensis forms biofilm in a true environmental setting and if biofilm aids or hampers the colonization and/ or predation of protozoans. Biofilm can provide protection form predation by effectively “bulking up” colony morphology, however, biofilm also concentrates bacteria allowing for a higher inoculum when contact occurs with other organisms ( Costerton et al., 1999 ; Darby et al., 2002 ; Matz and Kjelleberg, 2005 ). Curiously, another mechanism employed by bacteria to control predation pressure is the variation of the O-ag ( Wildschutte et al., 2004 ). LVS previously was shown to establish a biofilm of considerable biomass after 15 days incubation in Mueller-Hinton broth, though it was noted that significant variations were observed across the biofilm ( Mahajan et al., 2011 ). In our studies, the well to well variation observed was a unique phenotype that is likely due to phase variation and, as our data suggests, a sub-population of GVs are driving biofilm formation. Phase variation was described early during the characterization of Francisella virulence ( Eigelsbach et al., 1951 ), and it was later found that the O-ag was altered in GVs ( Cowley et al., 1996 ; Hartley et al., 2006 ). GVs have also been reported to revert back to the BVs, though the underlying genetic mechanism has not been elicited ( Eigelsbach et al., 1951 ; Soni et al., 2010 ) Given that the Francisella LPS structure is unique and is critical for pathogenesis ( Raynaud et al., 2007 ; Kim et al., 2012 ; Jones et al., 2014 ; Rasmussen et al., 2014 ), it is difficult to understand why phase variation readily occurs in Type A and B isolates of F. tularensis . Indeed, numerous studies have shown that GVs are attenuated to at least some degree ( Eigelsbach and Downs, 1961 ; Cowley et al., 1996 ; Hartley et al., 2006 ; Soni et al., 2010 ). Our data provides a possible explanation as F. tularensis may undergo phase variation, despite the cost of virulence, to enter a biofilm lifecycle. Heterogeneity in the O-ag produced has been observed in numerous species, including the intracellular pathogens, such as Brucella abortus , Legionella pneumophila , and Burkholderia pseudomallei ( Freer et al., 1995 ; Luneberg et al., 1998 ; Tuanyok et al., 2012 ). Though the exact mechanism responsible for phase variation of the O-ag promoting biofilm in Francisella is yet to be discovered, the O-ag is largely responsible for cell surface attributes, such as hydrophobicity and surface charge, and has been implicated in biofilm formation in Gram-negatives. For instance, Pseudomonas aeruginosa produces two distinct types of O-ag (common polysaccharide antigen [CPA] and O-specific antigen [OSA]) when grown planktonically, but as the transition to a robust biofilm occurs, the length of CPA is decreased or lost, ultimately promoting cell to cell adhesion and surface attachment ( Lam et al., 1989 ; Lindhout et al., 2009 ). Furthermore, it was found that the decrease of CPA occurs in a dependent cdGMP manner ( McCarthy et al., 2017 ). Additional studies have observed an increase in cell surface hydrophobicity and the secretion of outer membrane vesicles (OMVs) for P. aeruginosa ( Baumgarten et al., 2012 ). Perhaps a link between OMVs and biofilm formation exists for F. tularensis . In LVS, increased OMV secretion and biofilm formation was observed in a fupA mutant ( Siebert et al., 2019 ). Additional experiments performed by Siebert and colleagues demonstrated that the addition of OMVs increased biofilm formation in a dose dependent manner ( Siebert et al., 2019 ). Interestingly, a link between GVs and OMVs exist as a GV with extended O-ag was found to form more membrane vesicles compared to wild-type LVS ( Soni et al., 2010 ). Further supporting this link, O-ag mutants were found to make smaller vesicles ( Champion et al., 2018 ). In this manuscript, we also demonstrate that pH can act as an environmental checkpoint for biofilm formation. The result that F. tularensis cultured at pH 7 does not form biofilm, but upon sub-culture to a slightly acidic pH a robust biofilm is formed quickly suggests that matrix assembly is impeded and phase variation is unaffected. While at this time we are unable to rule out that pH effects a critical enzyme or signal molecule for biofilm formation, it is more likely that surface adherence is affected. In F. novicida , chitinase was found to affect the biophysical properties to control adhesion and biofilm production by increasing the cell surface charge to foster interactions with negatively charged surfaces ( Chung et al., 2014 ). A study by Champion et al. , convincingly demonstrated that a double mutant lacking O-ag and the capsule like complex adhered significantly better than wild-type and formed a robust biofilm ( Champion et al., 2019 ). The results of our western analysis shows that all of the naturally occurring variants identified in our study have an altered O-ag and capsule, consistent with the results presented by Champion et. al. In their study, Champion et al. found that biofilm formation was dependent upon growth medium and that BHI grown bacteria produced less biofilm ( Champion et al., 2019 ). Here, we show that biofilm can be quite robust in BHI if the pH is decreased, suggesting that nutritional differences alone do not account for observed lack of biofilm, but rather the chemical properties of the medium can impact biofilm formation. Given that it is unclear how phase variation that results in biofilm specifically alters the O-ag, further studies are needed to resolve the structural differences in these variants to understand the mechanistic details. However, both the growth environment and nutrient availability likely play a decisive role in cell fate as Francisella enters a VBNC. Free-living planktonic F. tularensis has been shown to quickly lose the ability to be detected by culturing when grown in fresh water ( Berrada and Telford, 2011 ). F. tularensis has been shown to remain metabolically active in water despite being undetectable by culturing on agar plates ( Forsman et al., 2000 ; Gilbert and Rose, 2012 ). A recent study by Siebert et al., has shown that biofilms of F. novicida and F. philomiragia allow the bacteria to survive longer than those grown planktonically ( Siebert et al., 2020 ). In agreement with this study, we found that LVS populations that contain GVs that constitutively produce biofilm, the onset of the VBNC state is delayed. While further studies are needed, this result suggests that the metabolic state of biofilm forming isolates is different from that of wild-type cells, providing metabolic heterogeneity to the population. One possibility is that heterogeneity of BV and GV is likely important particularly during overwintering when detection of infections in mammals are low ( Jellison, 1974 ; Mani et al., 2016 ). Indeed, F. tularensis has been found to modify the acyl chains of lipid A in response to temperature fluctuation ( Phillips et al., 2004 ; Li et al., 2012 ). Phenotypic heterogeneity can act as a buffer and ensure at least some sub-population of cells is suited for a changing environment. A survival strategy employed, such as this, is known as “bet hedging”, especially given that LPS phase variants are likely maladapted for infection of a vertebrate hosts ( De Jong et al., 2011 ; Grimbergen et al., 2015 ). Another possibility is that heterogeneity of BV and GV could provide an advantage during the transition from a vertebrate host back to the environment or vector reservoir where the selection against GVs may not be as strong. In conclusion, we demonstrate the ability of several strains of F. tularensis to consistently form biofilm in a stochastic manner due to the emergence of GV strains. These results shed light on several important facets of F. tularensis biology and have implications for how this pathogenic bacterium may reside in the environment in a VBNC form. Current studies are underway to determine the genetic differences of our GV strains that were hyper biofilm producers. We hope that this will lead to a basis of biofilm formation and/or variance switching. Furthermore, we are examining the fitness and virulence potential of our F. tularensis strains that are hyper biofilm producers in both in vitro and in vivo assays. In addition to allowing us to understand the survival of F. tularensis , these studies on the role of biofilm and phase variation may lead to better medical countermeasures to prevent tularemia."
} | 6,121 |
36575466 | PMC9795604 | pmc | 2,098 | {
"abstract": "Background Whole-cell biotransformation is a promising emerging technology for the production of chemicals. When using heterotrophic organisms such as E. coli and yeast as biocatalysts, the dependence on organic carbon source impairs the sustainability and economic viability of the process. As a promising alternative, photosynthetic cyanobacteria with low nutrient requirements and versatile metabolism, could offer a sustainable platform for the heterologous production of organic compounds directly from sunlight and CO 2 . This strategy has been applied for the photoautotrophic production of sucrose by a genetically engineered cyanobacterium, Synechocystis sp. PCC 6803 strain S02. As the key concept in the current work, this can be further used to generate organic carbon compounds for different heterotrophic applications, including for the whole-cell biotransformation by yeast and bacteria. Results Entrapment of Synechocystis S02 cells in Ca 2+ -cross-linked alginate hydrogel beads improves the specific sucrose productivity by 86% compared to suspension cultures during 7 days of cultivation under salt stress. The process was further prolonged by periodically changing the medium in the vials for up to 17 days of efficient production, giving the final sucrose yield slightly above 3000 mg l −1 . We successfully demonstrated that the medium enriched with photosynthetically produced sucrose by immobilized Synechocystis S02 cells supports the biotransformation of cyclohexanone to ε -caprolactone by the E. coli WΔ cscR Inv:Parvi strain engineered to ( i ) utilize low concentrations of sucrose and ( ii ) perform biotransformation of cyclohexanone to ε -caprolactone. Conclusion We conclude that cell entrapment in Ca 2+ -alginate beads is an effective method to prolong sucrose production by the engineered cyanobacteria, while allowing efficient separation of the cells from the medium. This advantage opens up novel possibilities to create advanced autotroph–heterotroph coupled cultivation systems for solar-driven production of chemicals via biotransformation, as demonstrated in this work by utilizing the photosynthetically produced sucrose to drive the conversion of cyclohexanone to ε -caprolactone by engineered E. coli .",
"conclusion": "Conclusions Our present work demonstrates the viability of a coupled autotrophic–heterotrophic production system where the sucrose-producing Synechocystis S02 entrapped within Ca 2+ -alginate beads provides sucrose as organic carbon source to drive biotransformation of cyclohexanone to ε -caprolactone by the E. coli WΔ cscR Inv:Parvi strain. By employing the immobilization approach, we increased the specific sucrose productivity of Synechocystis S02 by 86% compared to suspension cultures. Furthermore, we demonstrated that the sucrose production in immobilized cultures can be prolonged for at least 17 days by applying a semi-continuous production system, in which the medium is refreshed periodically. Immobilization has a clear advantage over the suspension cultivation since it simplifies the change of medium in photobioreactors and, thus, makes the process less energy demanding. To emphasize the applicability of coupled culturing of sucrose-producing Synechocystis S02 strain and a heterotrophic microbe, sucrose was produced over 7 days in BG11 + NaCl medium. Subsequently, the medium was removed and used to drive biotransformation of cyclohexanone to ε -caprolactone by the E. coli WΔ cscR Inv:Parvi strain. The biotransformation went to near completion in three hours without formation of cyclohexanol side-product, showing that the sucrose in the medium sustains the reaction without the downstream processing. Our present work paves the way for further designing and optimization of a coupled autotroph–heterotroph production systems for ε -caprolactone production. Furthermore, the platform could be applicable for biosynthesis and biotransformation of other compounds using E. coli -based cell factories .",
"discussion": "Results and discussion Specific production of sucrose increases in immobilized Synechocystis S02 In order to enhance sucrose production by the engineered Synechocystis S02 cells and induce its export to the medium, cultures were resuspended in BG11 + NaCl medium (OD 750 = 0.5) supplemented with 1 mM IPTG (see the Materials and methods section for more details). Suspension cultures of sucrose-producing Synechocystis S02 showed almost linear growth during 7 days of incubation reaching optical density at 750 nm (OD 750 ) of ~ 6.8 (Fig. 3 A). Simultaneously, chlorophyll a (Chl) concentration steadily increased and saturated at the 6th day (Fig. 3 B). Sucrose accumulated in the medium in a linear manner until the 7th day, after which it plateaued at the maximum concentration of 1910 mg l. −1 (Fig. 3 C), which corresponds to ~ 70 mg sucrose per mg of Chl (Fig. 3 D). These results are comparable to the data obtained by Thiel et al. [ 14 ] Fig. 3 Characterization of sucrose-producing Synechocystis S02 in suspension cultures. A Cell growth assessed by OD 750 ; B Chl content; C total sucrose concentration in the culture medium; D specific sucrose production yield [mg (mg Chl) −1 ]. Error bars represent the standard deviation of three independent biological and two technical replicates The entrapment of algae and cyanobacteria in the polymeric matrix is known to diminish cell division and formation of biomass, thus allowing to divert energy for production of targeted chemicals [ 23 , 25 ]. Such solid-state photosynthetic production systems transfer cells to long-lived biocatalytic production mode. Therefore, in the next step we immobilized the sucrose-producing cells in 3% (wt/v) alginate beads cross-linked with Ca 2+ -ions. The beads can endure vigorous shaking, which facilitates the mass transfer between the cells and the medium through the matrix. Different from suspension cultures, the immobilized cells showed the most pronounced Chl accumulation during the initial phase of the sucrose production (Fig. 4 A). Fig. 4 Characterization of alginate immobilized Synechocystis S02 cells. A Chl content of immobilized cells; B sucrose production in alginate immobilized Synechocystis S02 cultures; C specific sucrose production yields in suspension cultures and immobilized cells. Error bars represent the standard deviation of three independent biological and two technical replicates. The difference between suspension and alginate-entrapped samples is statistically significant for each time point ( P = 0.001–0.006) By the 3rd day of the experiment, the color of the beads became much darker than in the beginning due to the Chl accumulation. Then, the Chl content of the beads increased slower until the 5th day and started gradually declining after that (Fig. 4 A). The initial rise in the Chl content can be explained by the cell division and the increase of the biomass within the matrix until the point when the immobilized cultures start experiencing light and presumably nutrient limitations, resulting in inhibition of metabolic activity and cell growth by the end of the experiment. It is important to note that cell outgrowth from the beads was hardly noticeable during 7 days of sucrose production. Ca 2+ -alginate-entrapped cells produced sucrose similar to suspension cultures (Figs. 4 and 3 ). The specific productivity of both cultures steadily increased from the beginning of the experiment and reached the maximum on the 7th day, after which production ceased under both setups (Figs. 3 C and 4 B). It is noteworthy that the immobilized cells showed significantly higher specific production yields compared to suspension cells throughout the experiment reaching the 86% increase (1200 and 700 mg sucrose mg −1 Chl, respectively) by the 7th day (Fig. 4 C). The total maximum sucrose yield of immobilized cells was 1150 mg l −1 (Fig. 4 B). It was demonstrated in other works that the immobilization of Synechocystis is an effective way to increase the production yield of the cells. Immobilization in Ca 2+ -alginate beads was reported to increase succinate [ 26 ] and β -phellandrene [ 27 ] production, while immobilization in Ca 2+ -alginate thin film was shown to be effective for the increase of ethylene production [ 28 ]. The immobilization was previously reported to be effective to increase sucrose production as well, with engineered Synechococcus elongatus sp. PCC 7942 cells showing a 2- to 3-fold increase in specific sucrose production after their entrapment within Ba 2+ -alginate beads [ 8 ]. Our results obtained with Synechocystis S02 cells immobilized within Ca 2+ -alginate beads are in line with the above-mentioned studies. The production of sucrose by Synechocystis cells could be further enhanced by genetic engineering towards constitutive sucrose synthesis without application of the high salt stress and by redesigning the photosynthetic electron transport for enhanced carbon partitioning towards the sucrose production [ 14 , 18 ]. Further technological improvements are also possible, such as the design and use of specialized photobioreactors with the improved light distribution and the application of new immobilization materials with better porosity and mechanical stability. Sucrose production stimulates photosynthetic O 2 evolution and CO 2 fixation The real-time gas fluxes were monitored in immobilized cells and in cells grown in suspension by membrane inlet mass spectrometry (MIMS) to investigate the correlation between sucrose production and photosynthetic activity. After 3 days of sucrose production, net photosynthetic O 2 evolution both in suspension and immobilized cultures was significantly higher ( P = 0.02–0.04) in cultures producing sucrose (+ NaCl) compared to non-producing ones (-NaCl) (Fig. 5 A). Net CO 2 yield rates were also higher in the sucrose-producing cells, albeit the difference was statistically significant only in suspension cultures ( P = 0.006) (Fig. 5 B). These results suggest that sucrose production acts as a strong sink for photosynthetic CO 2 fixation, which increases the overall photosynthetic activity of the cells. The increase in net O 2 evolution and net CO 2 fixation yield in suspension cultures of sucrose-exporting cells was also described previously [ 14 , 29 ]. It is important to note that we observed higher gas-flux rates in suspension cultures compared to bead-immobilized cells, which can be attributed to the low gas permeability of the alginate matrix and other effects of immobilization on the cell metabolism [ 25 ]. By the 7th day, the photosynthetic O 2 evolution and carbon fixation rates dropped both in the suspension cultures and immobilized cells, demonstrating a decrease in photosynthetic activity. In line with the real-time gas exchange results the effective photosynthetic yield Y(II) in both cultures dropped from 0.37 to 0.08, which was accompanied by the decline of sucrose production. A possible reason for the decline in photosynthetic activity is the accumulation of sucrose, which switches the cell metabolism into photomixotrophy, when the cells start metabolizing sucrose as a carbon source while performing photosynthesis and CO 2 fixation [ 30 ]. This form of metabolism, which provides phototrophic cells with extra energy and carbon, often occurs when organic carbon sources are available in the environment, for example during phytoplankton blooms. Besides decreased photosynthetic activity, the transition to photomixotrophic growth is accompanied by increased cell respiration [ 31 ]. Indeed, we observed a tendency to enhanced respiration by the end of the experiment almost under all conditions, except in unstressed suspension cultures (-NaCl Susp) where respiration did not change (Fig. 5 C). However, based on this data we could not clearly state if enhanced respiration is linked to the transition to mixotrophic growth. Net CO 2 yield, which represents a difference between CO 2 consumption in the CBB cycle and CO 2 release in respiration, was low after 3 days in all cases and dropped below the respiration compensation point by the 7th day. The latter indicates on a significant drop in CO 2 fixation capacity and correlates well with high respiration rates (Fig. 5 C), which finally affect the sucrose productivity. The further studies are needed for understanding mechanisms leading to the transition to the photomixotrophic growth and the declined sucrose production capacity. In the long-term production process, photomixotrophy can be abolished by periodically refreshing the medium throughout. Fig. 5 Real-time gas exchange rates in suspension and immobilized cells. A Average steady state net O 2 evolution; B net CO 2 yield; C dark O 2 consumption. The cells were incubated in BG11 supplemented with NaCl and 1 mM IPTG (+ NaCl) to induce sucrose production. As a negative control the cells were incubated in BG11 medium supplemented with 1 mM IPTG, but not NaCl (-NaCl). Error bars represent the standard deviation of three independent biological replicates. Statistical significance between cultures in medium with and without NaCl is marked by * ( P = 0.006–0.04) The cultivation, thus preventing nutrient limitation and sucrose accumulation in the cultures leading to the end-product inhibition effect and further metabolization of sucrose by the cells. Semi-continuous production mode leads to prolongation of the sucrose production in beads To remove secreted sucrose, the medium was refreshed every 3–4 days after sampling throughout the experiment. By employing this method, the period of efficient sucrose production was considerably prolonged. At day 10, the cumulative production yield was close to 2200 mg sucrose l −1 (Fig. 6 , black squares) and then, the production activity started to decline gradually (Fig. 6 , yellow bars). As a result, the production yield dropped considerably by the 17th day, but even after that point negligible amount of sucrose was detected in the medium until the end of the experiment (Fig. 6 ). The beads remained stable during the 27 days of incubation and outgrowth from the beads remained inconsequential in the beginning of production but increased after 2 weeks of cultivation (up to OD 750 1.0 on the 13th day), and then continued to increase until the end of the experiment reaching the maximum OD 750 1.5 (Fig. 6 ). This can be attributed to the slow degradation and the gradual breach of the surface of the beads. Bead immobilization is a practical and effective method for semi-continuous cultivation since no energy-intensive centrifugation is needed and even on an industrial scale, the medium can be easily changed through a valve system. The cumulative maximum sucrose yield during the semi-continuous cultivation reached 3000 mg l −1 after 17 days (Fig. 6 ), which is almost three times higher compared to the amount obtained during 7 days of batch production. This clearly demonstrates that semi-continuous cultivation is a better alternative than batch cultivation to maximize the amount of harvested sucrose by prolonging the duration of the production period of Synechocystis S02 immobilized in Ca 2+ -alginate beads. Fig. 6 The cumulative amount of sucrose harvested from bead-immobilized cultures. Samples were taken and the medium was refreshed every 3–4 days. Error bars represent the standard deviations of two independent biological and two technical replicates. Cumulative values are obtained by summing up the averages obtained from individual time points. Outgrowth was evaluated by measuring OD 750 from the medium. OD 750 data points represent individual measurements Sucrose produced by Synechocystis drives biotransformation in E. coli The next step was to verify the capability of sucrose produced by immobilized Synechocystis S02 cells to sustain the biotransformation of cyclohexanone to ε -caprolactone in recombinant E. coli . For this purpose, we engineered E. coli WΔ cscR Inv, which is capable of effectively utilizing even low amounts of sucrose due to the deactivation of sucrose catabolism repressor gene ( cscR ) and heterologously expressed invertase enzyme with an N-terminal pelB leader sequence for export to the periplasm [ 32 ]. Expression of a heterologous Baeyer–Villiger monooxygenase from Parvibaculum lavamentivorans (BVMO Parvi ) in the WΔ cscR Inv background enabled the biotransformation of cyclohexanone, exogenously added substrate, to ε -caprolactone by utilizing the sucrose produced by immobilized Synechocystis S02 (Fig. 7 ). Fig. 7 Progression of E. coli driven biotransformation of 5 mM cyclohexanone substrate to ε -caprolactone in different media. A BG11 + NaCl medium enriched with sucrose by alginate immobilized Synechocystis S02 cells; B M9 medium supplemented with 10 mM sucrose as positive control and C BG11 + NaCl medium without sucrose as negative control Sucrose was produced in BG11 + NaCl by alginate immobilized Synechocystis S02 cells over a 7-day period (1150 mg l −1 ). Then the medium was removed from the beads and used as culture medium for the biotransformation by the E. coli WΔ cscR Inv:Parvi strain. We monitored the biotransformation by the engineered E . coli cells to evaluate the potential of coupled production. Under these conditions, we observed full biotransformation of cyclohexanone to ε -caprolactone within three hours (Fig. 7 A) with the average transformation rate of 0.9 mM h −1 . M9 minimal medium used for the cultivation of E . coli was supplemented with 10 mM sucrose and used as a positive control for the biotransformation. The control reaction in M9 was faster, and proceeded to completion within 2 h, with the average rate of 2.3 mM h −1 (Fig. 7 B). As a negative control fresh BG11 + NaCl medium without any additional sucrose was used to ascertain that the E. coli is not capable of performing the biotransformation reaction without sucrose, utilizing for example, energy stored during previous cultivation steps. Only minimal amount of cyclohexanone was converted to ε -caprolactone in the absence of sucrose over the 24-h time period (Fig. 7 C). Since both the substrate (cyclohexanone) and the product ( ε -caprolactone) are semi-volatile, the final concentration of the product can deviate from the initial substrate concentration. From these results it is evident that the E. coli WΔ cscR Inv:Parvi strain is capable of fast conversion of cyclohexanone to ε -caprolactone without the formation of side-product, cyclohexanol. The conversion rate is comparable to other E. coli strains harboring BVMOs when utilizing rich TB-medium [ 9 ]. This also confirms that an organic carbon source, in our case sucrose, is essential for the cyclohexanone biotransformation in E . coli , as expected based on the BVMO dependence on NADPH, which is generated via the glycolytic pentose phosphate pathway in this host. [ 33 ]. However, other components of the previously used TB-medium seem to be less important, while the high salt concentration is no hindrance for cells to perform the biotransformation under the used conditions. Altogether, the data unambiguously demonstrate that sucrose produced by Synechocystis S02 over a 1-week batch culture, is sufficient to drive the conversion of 5 mM cyclohexanone to ε -caprolactone by E . coli WΔ cscR Inv:Parvi without any downstream modification or manipulation of the medium. Our findings support the general concept of utilizing carbohydrates synthesized from CO 2 by engineered cyanobacteria as a source of energy for biotransformation catalyzed by heterotrophic microbes and lay the foundation for an alternative sustainable ε -caprolactone production platform. To effectively couple the photoautotrophic and heterotrophic production systems in actual co-cultures where the two process is simultaneously ongoing in one bioreactor, however, further optimization is needed. In contrast to the published co-cultures [ 8 , 19 – 22 ] biotransformation offers ‘the substrate-in-product-out’ concept, where besides the sucrose produced by cyanobacteria also the external substrate has to be fed to the E. coli cells. In the published examples the compounds produced by the heterotrophs are products derived from the carbon fixed by the phototroph from CO 2 and no compound is fed to the strain to be transformed. The CO 2 that the phototroph converts to sucrose is either utilized for the synthesis of compounds or for the growth of the heterotroph [ 8 , 19 – 22 ]. The issues that need to be addressed to establish a working co-culturing system are ( i ) the considerable difference between the rate of sucrose production and the biotransformation process. This challenge could be overcome by introducing the E. coli performing the biotransformation together with cyclohexanone at later stages of the sucrose production to ensure sufficient amounts of sucrose to sustain the biotransformation; ( ii ) the overall slow sucrose production rate. This could be addressed by further engineering the sucrose producer or by exploring different strains, such as the promising Synechococcus elongatus UTEX 2973 [ 18 ]; ( iii ) the effective collecting of the biotransformation product without interrupting the sucrose production and co-culturing as well as the ( iv ) ways to avoid losses of the substrate and the product of the biotransformation in a prolonged setup remain challenges, that need experimental trials to find ways to overcome them."
} | 5,401 |
36691741 | PMC9871512 | pmc | 2,100 | {
"abstract": "Abstract Biogenic waste (solid/liquid/gaseous) utilization in biological processes has disruptive potential of inclining towards carbon neutrality, while producing diverse products output. Anaerobic fermentation (methanogenesis and acidogenesis) routes are crucial bioprocesses for production of various renewable chemicals (carboxylate platform/organic acids, short/medium chain alcohols, aldehydes, biopolymers) and fuels (methane, hydrogen, hythane, biodiesel and electricity), while individual operations posing process limitations on their conversion efficiency. Advantageous benefit of using the individual bioprocess technicalities is of utmost importance in the context of sustainability to conceptualize and execute integrated waste biorefinery. The opinion article intends to document/familiarize the waste‐fed biorefinery potential with application of hybrid advancements towards multiple product/energy/renewable chemical spectrum leading to carbon neutrality bioprocesses. Unique and notable challenges with diverse process integrations along with electrochemical/interspecies‐redox metabolites‐materials synergy/enzymatic interventions are specifically emphasized on application‐oriented waste feedstock potential towards achieving sustainability.",
"introduction": "INTRODUCTION – WASTE FED BIOREFINERY Climate change is a constant alarming concern in the context of ever‐increasing environmental pollutants with anthropogenic activities and green‐house gas emissions (Jiang et al., 2019 ). Mitigation needs to be emphasized in the context of sustainability and aligning to the advancements in carbon capture, utilization and storage (CCUS) processes (Venkata Mohan et al., 2016 ). Biogenic waste (solid/liquid/gaseous) has huge potential to capture market, when cohesively involved with individual advantages of the bioprocesses in an integrated manner (Venkata Mohan et al., 2016 ; Venkata Mohan & Katakojwala, 2021 ). Non‐genetic biological approaches could be economically viable with higher reach for execution at variable operational scales for multiple products benefit (renewable chemicals, energy and fuels) with carbon neutralization (Tharak & Venkata Mohan, 2022 ). Hence, the advantage of biocatalysts with regulation on the process limitations and mechanism towards maximizing products conversion using waste feedstock could be considered a challenge to address in the context of waste biorefinery. Biological processes are considered sustainable options in waste utilization/conversion towards renewable chemicals, energy and fuels production (Tharak & Venkata Mohan, 2022 ; Van et al., 2020 ). Anaerobic fermentation (AF)/anaerobic digestion (AD) are potential bioconversion processes for treating diverse biogenic waste/wastewaters resulting in energy/product spectrum generation (Van et al., 2020 ; Zhao et al., 2020 ). Individual bioprocesses have conversion efficiency and system stability limitations, resulting in decreased overall process performance and products output (Baek et al., 2018 ; Zhao et al., 2020 ). Mechanism‐oriented regulations with process integrations relative to metabolic/electrochemical/enzymatic/material interventions are crucial for specified products generation (Bertsch et al., 2016 ; Chen et al., 2020 ; Cotton et al., 2020 ; Jiang et al., 2019 ; Lovley & Holmes, 2022 ). In this domain, the residual organics/inorganics conversion/utilization to increase the overall energy/product value requires acidogenesis and methanogenesis process integrations along with hybrid advancements based on required products orientation (Borrel et al., 2016 ). Acidogenesis process could be beneficial in establishing a carboxylate platform in terms of short/medium chain fatty acids production (Cotton et al., 2020 ). Unregulated acidogenesis could lead to methanogenesis (CH 4 production) during biological processes, which is beneficial yet leads to unspecified product output with lesser process understanding (Detman et al., 2018 ; Van et al., 2020 ). Hence, process regulations at individual level could be a criterion for establishing/understanding the biological mechanisms with metabolic/electrochemical/enzymatic/material interventions with channelized integrations towards carbon neutrality, residual carbon utilization with maximized products generation (Berghuis et al., 2019 ; Martins et al., 2018 ; Singh et al., 2020 ). Photosynthesis integration can also be aligned to acidogenesis and methanogenesis to achieve self‐sustainability, while maximizing resource recovery with carbon neutral footprints/near zero carbon discharge (Venkata Mohan et al., 2016 ). Photosynthetic integrations could additionally benefit in biomass, biodiesel, bioethanol, edible oil, protein, carbohydrates, nutraceuticals and pigments, while deoiled biomass can be used for biofertilizer and biochar/bioelectrode preparations (Jiang et al., 2019 ; Venkata Mohan et al., 2016 ). Conceptual circularity framework with domestic and industrial concerns is an accountable scenario with current environmental issues (Venkata Mohan & Katakojwala, 2021 ). Carbon neutrality maintenance needs to be considered as a state‐of‐art entity prioritizing the carbon capture, utilization and storage (CCUS) with green chemistry principles (Venkata Mohan et al., 2016 ; Venkata Mohan & Katakojwala, 2021 ). Integrated closed loop approach in the context of establishing waste biorefinery with multiple cohesive bioprocesses is ideal towards benefitted environmental and economic sustainability (Figure 1 ) (Venkata Mohan et al., 2016 ). Waste‐fed biorefinery emphasizes on the thorough utilization of residual carbon savings of individual bioprocesses within a channelized network as incentives for additional products benefit with further reduction in footprints. Waste‐fed biorefineries could be a techno‐economically viable benchmark platform technologies with customized streamlining for integration with respect to the waste/wastewater/biomass characteristics. System thinking and life cycle considerations in the context of industry 4.0 and societal benefits need to accounted for cohesively inter‐linking bioprocesses paving a futuristic way for establishing a circular economy concept. FIGURE 1 Conceptual circularity framework depicting possible integrations of biological and bioelectrochemical routes towards carbon neutrality."
} | 1,592 |
36132777 | PMC9418410 | pmc | 2,101 | {
"abstract": "With the development of flexible electronics and wearable devices, there is strong demand for flexible, superhydrophobic, and multifunctional coatings. Motivated by the promise of attractive multifaceted functionality, various techniques have been developed to fabricate flexible surfaces with non-wetting properties. However, until now, there have been few reports on superhydrophobic surfaces with condensate microdrop self-propelling (CMDSP) functionality on a carbon nanotube film. Here, we used a facile electrodeposition method to develop for the first time a new type of flexible superhydrophobic surface with CMDSP functionality based on carbon nanotube films. These flexible CMDSP surfaces are robust after multiple cycles of bending of the film-coated substrate, i.e. , without impacting the surface superhydrophobicity and CMDSP performance. The proposed light and flexible surface, combined with CMDSP, will support a novel generation of coatings that are multifunctional, flexible, smart, and energy saving. This new type of functional flexible interface not only opens new avenues in research into the fundamental structure–property relationships of materials, but also exhibits significant application potential for advanced technologies.",
"conclusion": "4. Conclusion We report the first robust and flexible superhydrophobic surfaces with the desired condensate microdrop self-propelling functionality. This ability may be used to develop CNTF-based flexible surfaces for moisture self-cleaning and anti-frosting behaviors, power generation, electrical energy harvesting, enhanced condensation heat transfer, and conventional superhydrophobicity applications with regard to macroscopic droplets. These low-cost and robust CNTF-based CMDSP surfaces are also suitable for other functional devices, which can advance the development of flexible electronics. Such CNTF-based flexible surfaces are very robust for both macro- and micro-condensation drops, which can be demonstrated by multiple cycles of bending of the film-coated substrate without impacting the surface superhydrophobicity and CMDSP performance. In addition, our method is facile, low-cost, scalable, and has the potential to evolve into practical nanoengineering technologies for CNTF-based surfaces. We plan to further optimize the functional layers and increase the tolerance to mechanical damage. The proposed light and flexible material combined with the CMDSP concept will lead to a novel generation of multifunctional and flexible energy-saving coatings.",
"introduction": "1. Introduction By appropriately designing and engineering surfaces with different nanostructures, a large variety of multifunctional composite materials can be created for a wide range of application areas, from chemistry to advanced biomedical science. Currently, a new type of superhydrophobic surface is attracting interest due to its condensate microdrop self-propelling (CMDSP) functionality on metal, silicon, or other rigid substrates. 1,2 Different from conventional superhydrophobic surfaces characterized by gravity-driven shedding or bouncing of macroscopic water drops, such novel surfaces can self-remove small-scale condensate microdrops as they jump away via coalescence-released excess surface energy. This phenomenon does not require other forces such as gravity or stream shear force; thus, the surfaces retain their robust superhydrophobicity under moisture, condensation, or steam conditions. 3–10 CMDSP surfaces are considered to have significant value in applications such as self-cleaning of moisture, 11 electrostatic energy harvesting, 12 and enhanced condensation heat transfer 13 for higher efficiency energy utilization and thermal management. Typically, the building of special surficial nanostructures is needed to obtain CMDSP functionality. 14–17 There is a prototype in nature. Wisdom et al. 18 reported that tiny condensate microdrops on the closely packed nanocone surface of cicada wings can self-remove by jumping via mutual coalescence; thus, the surface has a moisture self-cleaning function. Based on this bio-inspiration, many studies have shown that constructing arrays of inorganic nanostructures with sharp tips can indeed endow the surfaces with CMDSP functionality. In principle, any sub-microscale structures with a small feature size ( i.e. , tip size and interspace) and a certain height (or depth) can become effective candidates for creating CMDSP surfaces. Such structures are arrays of closely packed nanotips, such as nanocones, 19–22 nanoneedles, 23–26 nanopencils, 13 tip-like nanotubes, 2,27 and nanorod-capped nanopores, 16,28 as well as the porous structure of nanoparticles, 15 nanosheets 4,12,29,30 on copper substrates, nanowires, 31,32 and two-tier structures on silicon substrates. 33,34 Despite much effort made to date, most nanostructures are constructed on rigid inorganic or metal substrates, i.e. , few flexible CMDSP surfaces have been reported. However, there have been several reports about flexible superhydrophobic surfaces based on polymers. 35–37 These flexible superhydrophobic surfaces lack the CMDSP functionality, which limits their application in the areas of moisture self-cleaning, energy utilization, and thermal management. Despite there being some polymer CMDSP surfaces, the substrates are rigid glass, and several steps are required to obtain sharp polymer tips. 11 Furthermore, as most flexible polymers are not conductors, it is difficult to use the facile electrodeposition method to construct nanotip structures on these flexible surfaces. Another promising flexible material is called nanocellulose paper, but this material has poor shape stability in water and other solutions and a low decomposition temperature (320 °C). 38 Hence, the successful fabrication of robust superhydrophobic surface nanostructures with CMDSP functionality on a suitable flexible substrate using facile one-step methods is very appealing but remains a challenge. In this work, we make a breakthrough to show, for the first time, the fabrication of flexible CMDSP surfaces. We find that zinc oxide (ZnO) nanoneedles can endow flexible carbon nanotube films (CNTFs) with CMDSP functionality when fabricated using a facile and cheap electrodeposition method. CNTFs consist of carbon nanotubes that can sustain their flexibility even at low temperatures. They are good conductors if composed of metal CNTs. However, prior to our study, there were no reports on constructing nanoneedles on CNTFs by electrodeposition. One of the greatest challenges to overcome is that these carbon assemblies are easily dispersed in an electrolyte during the electrodeposition process. We pre-treated the CNTF using an efficient microwave method to strengthen the film and avoid dispersion during the electrodeposition process. Our results show that the ZnO–CNTF composite film can maintain its flexibility even after it is covered with rigid nanoneedles. The composite film is very robust and retains its superhydrophobicity, even after continuous up-and-down bending. Furthermore, due to the microscopically low-adhesive nature of its building blocks, the flexible composite film has been endowed with CMDSP functionality, which can not only realize high-density nucleation but also maintain the high frequency of self-propelled jumping. These findings helped in the development of novel flexible coatings with moisture self-cleaning, stretchable electrodes, and advanced heat and mass transfer nanomaterials and devices.",
"discussion": "3. Results and discussion To achieve the desired CMDSP functionality, it is necessary to minimize the dissipation of coalescence-released excess surface energy caused by solid–liquid adhesion. 25 According to previous research, 40 nanoneedles have an extremely low solid–liquid interface adhesion and can endow surfaces with the CMDSP functionality. Hence, based on our recent work in constructing nanostructures on CNTF, 41 we first design the flexible CMDSP film consisting of ZnO nanoneedles and CNTF. Fig. 1 shows the macroscopic and microscopic structures, together with the function of our ZnO nanoneedle and CNTF composite films. The lower image of Fig. 1 depicts our design for a ZnO–CNTF composite nanosurface. After fluorosilane modification, the water droplets can self-propel on the surface, driven by coalescence-released excess surface energy. The gray film, green cones, and blue circles represent CNTF, ZnO nanoneedles, and water droplets, respectively. The upper images show more atomic details of how CNTs assemble into bundles and bundles pack into fibers. The growth details are also shown, i.e. , zinc and oxygen adatoms adsorb, nucleate on surfaces of CNT bundles, and then grow into ZnO nanoneedles on CNT bundles. The next issues to be addressed are the growth of ZnO nanoneedle coatings on the flexible CNTF surface by using the facile, low-cost, and scalable electrodeposition method and verification of their functional utility. Fig. 1 Schematic showing the utilization of ZnO nanoneedles (green cones) to endow flexible CNTF surfaces (gray film) with a condensate microdrop self-propelling function. As shown by the lower images, water microdrops (blue circles) jump on the tips of ZnO nanoneedles. The upper images show that the flexible CNT films consist of CNTs. During the electrodeposition process, adatoms (yellow and blue balls) first adsorb on the surfaces of CNT bundles, and then they nucleate and grow into ZnO nanoneedles. It has been reported that electrodeposited ZnO nanoneedle films can be rapidly grown on copper or silicon surfaces. However, there have been no reports of electrodeposited ZnO nanoneedles on top of CNTF or their use for studying CMDSP functionality. One big challenge is that these carbon assemblies are easily dispersed in electrolyte during the electrodeposition process. Here, in order to successfully grow ZnO nanoneedles on CNTF surfaces, we first apply a microwave pretreatment to strengthen the CNTFs so they avoid dispersion during a 30 min electrodeposition process. Then, we use the facile electrodeposition method to deposit ZnO nanoneedles on CNTFs. Fig. 2a and b show typical scanning electron microscopy (SEM) top and side views of the as-grown nanoneedles, corresponding to a reaction time of 30 min. Nanoneedles directly grown on the surface of the CNTF are clearly seen in Fig. 2b . The CNTF is the fiber network structure seen below the nanoneedles. However, due to the uneven surfaces of the CNTF, nanoneedles are not as closely packed and straight as those grown on a rigid flat substrate. The average height and tip size of the nanoneedles are 1.5 μm and 20 nm, respectively. As shown in Fig. 2c , there are spiraling steps in the nanoneedles that indicate that their growth mechanism is induced by step climbing of screw dislocations, which is a common growth mechanism for nanoneedles or nanocones. The EDAX results show that these nanoneedles are ZnO. Fig. 2 Microscopic morphology of ZnO nanoneedles. (a) Top-view SEM with low magnification. (b) Side-view SEM. (c) Top-view SEM with high magnification. (d) EDAX result of the region in the red square in (a). Subsequently, we characterized the flexibility and superhydrophobicity at the macroscale level. As compared with a pristine CNTF ( Fig. 3a ), the ZnO–CNTF composite films still maintain their flexibility to be easily bent up and down ( Fig. 3b ). After modification with fluorosilane, they exhibit excellent superhydrophobicity (inset of Fig. 3b ), in contrast to the hydrophobic surfaces of pristine CNTFs (inset of Fig. 3a ). Furthermore, we use an injection set to inject macroscale water drops (diameters greater than ∼3 mm) on the up- or down-bent surfaces. The overlapped optical images in Fig. 3c–f show continuous moving trajectories of macroscale water drops. We can clearly see in Fig. 3c and d that surfaces of pristine CNTFs have very high adhesion; hence, when macroscale water drops fall on their surfaces, the drops firmly adhere to the surfaces. In contrast, the surfaces of nanoneedle CNTFs have much lower interfacial adhesion. Fig. 3e and f show that macroscale water drops fall on the nanoneedle surfaces and quickly jump off. Our results verify that the ZnO–CNTF composite films are very robust. Their macroscopic superhydrophobicity is maintained even under continuous up- or down-bent conditions (see the movies in the ESI † ). Fig. 3 Comparison of pristine and nanoneedle-covered CNTFs in terms of their flexibility and superhydrophobicity at the macroscale level. (a) Flexible pristine CNTF. (b) Flexible nanoneedle-covered CNTF. The insets show optical images of macroscale drops (∼3 mm in diameter) on the surface of pristine CNTF and nanoneedle-covered CNTF, respectively. (c–f) Overlapped optical images showing continuous macroscale drops (∼3 mm in diameter) dropping on the up- and down-bent surfaces of pristine CNTF and nanoneedle-covered CNTF, respectively. In (e) and (f), we only overlap the images in which the drops maintain their spherical shape to represent the superhydrophobic properties of surfaces. More details can be seen in the movies in the ESI. † We take a further step to test the CMDSP functionality of these films. Our studies indicate that the nanoneedle-covered CNT films have the desired CMDSP functionality after fluorosilane modification. Fig. 4a–d show the representative optical top views of the instant self-expulsion process of condensate microdrops on the sample surface. It is evident that the in-plane coalescence of adjacent microdrops, caused by their direct condensation growth, can trigger the out-of-plane jumping of the merged microdrop (as elaborated in Fig. 4e ). As shown in Fig. 4g and h , the merged microdrop can eject from the nanosample surface and then fall along a parabolic trajectory. In addition, these ejected microdrops can fall back to the sample surface and trigger impact-induced self-propelling events, as shown in Fig. 4f . This CMDSP mode differs from that caused by quasi-static growth and can greatly reduce the residence period of microdrops. Fig. 4 (a–d) Typical time-lapse optical images showing the self-propelling of condensate microdrops by their mutual coalescence (highlighted by dotted ovals). (e–h) Schematic, overlapped optical top views, and side views showing the coalescence-induced self-jumping of condensate microdrops on horizontal and vertical surfaces, respectively. (i and j) Statistical drop number percentage of condensed microdrops with diameters of <20 μm (black), 20–40 μm (red), and 40–60 μm (green) on the nanostructured surface. Clearly, such a nanostructured film has a remarkable high-density self-renewal ability of small-scale condensate microdrops. The samples are placed on a horizontal or vertical cooling stage with a substrate temperature of approximately 2 °C, ambient temperature of approximately 25 °C, and relative humidity of approximately 85%. (k) Average density per area of condensed microdrops on CMDSP surfaces. To quantify the self-removal ability of microdrops on the sample surface, we conducted statistical analyses of the diameter ( Fig. 4i ) and density ( Fig. 4k ) of condensate drops on the nanostructured surface varied with condensation time, the drop number distribution of residence microdrops with diameters (d) of <20 μm, 20–40 μm, and 40–60 μm ( Fig. 4j ). Approximately 90% of the microdrops have d < 20 μm and have a slight fluctuation with time; around 10% of the microdrops have d = 20–40 μm; and almost 4% of the microdrops have d = 40–60 μm. The percentage value only increases a little bit with a longer condensation time. Accordingly, the nanoneedle-covered CNT film can realize efficient self-removal of small-scale condensate microdrops, especially those with sizes below 20 μm. It is interesting to note that our CMDSP functionality is far superior even compared with robust and rigid CMDSP surfaces, such as clustered ribbed-nanoneedle structured copper surfaces 23 and porous films of nanoparticles on copper surfaces. 15 Furthermore, we also bent the composite film to test whether the robust CMDSP functionality is maintained. Fig. 5 shows an optical image of a bent film that was put on an arched ingot's surface. Fig. 5b–e show the CMDSP phenomenon from the top view, indicating that even in the bent parts the CMDSP function is still very robust. Hence, our results indicate that bending does not affect the microscopic superhydrophobicity, and the CMDSP phenomenon is very active and robust on the top of the bent part (as clearly shown in Fig. 5 ). Fig. 5 (a) Optical image of a bent film that was put on an arched ingot's surface. (b–e) Typical time-lapse optical images showing the self-propelling of condensate microdrops by their mutual coalescence (highlighted by dotted ovals). We know that the density and height of the ZnO nanoneedle array play an important role in the superhydrophobic CMDSP performances. The construction rules of CMDSP surfaces on rigid metal substrates have been thoroughly discussed in our review paper 1 and recent article. 40 According to previous analyses into the roles of these geometric parameters in governing their CMDSP performance, we know the following basic rules: (1) the increase of tip diameters can lower CMDSP efficiency; (2) the decrease of interspaces is beneficial to increase CMDSP efficiency; (3) the appropriate height can avoid the easy penetration of moisture. Hence, in this paper, we have constructed the ZnO nanoneedles with optimum parameters mentioned in a previous paper. 40 However, besides following those rules on rigid surfaces, the case of deformation of flexible surfaces should also be considered, and ZnO nanoneedles on flexible surfaces should not be as straight as those on rigid surfaces. This is because flexible films always work in case of deformation. In this paper, we have tested the CMDSP efficiency in case of bending; as shown in Fig. 5 , our results show that the CMDSP function can be maintained even when this film is bent. Very recently, Wang et al. designed a robust superhydrophobic surface on different types of rigid substrates; the water repellency of the resulting superhydrophobic surfaces is preserved even after abrasion by sandpaper and by a sharp steel blade. 42 Inspired by this work, our study on the improvement of durability of flexible superhydrophobic CMDSP film in real applications is underway."
} | 4,621 |
37340871 | PMC10460853 | pmc | 2,103 | {
"abstract": "Abstract Neuromorphic artificial intelligence systems are the future of ultrahigh performance computing clusters to overcome complex scientific and economical challenges. Despite their importance, the advancement in quantum neuromorphic systems is slow without specific device design. To elucidate biomimicking mammalian brain synapses, a new class of quantum topological neuristors (QTN) with ultralow energy consumption (pJ) and higher switching speed (µs) is introduced. Bioinspired neural network characteristics of QTNs are the effects of edge state transport and tunable energy gap in the quantum topological insulator (QTI) materials. With augmented device and QTI material design, top notch neuromorphic behavior with effective learning‐relearning‐forgetting stages is demonstrated. Critically, to emulate the real‐time neuromorphic efficiency, training of the QTNs is demonstrated with simple hand gesture game by interfacing them with artificial neural networks to perform decision‐making operations. Strategically, the QTNs prove the possession of incomparable potential to realize next‐gen neuromorphic computing for the development of intelligent machines and humanoids.",
"conclusion": "3 Conclusion In this study, we successfully developed a quantum topological insulator based neuristor with equivalent neuromorphic synaptic performance as the mammalian brain. From our investigations, the quantum topological insulator materials are well exposed with their potential in advanced low power electronic switching systems. Surface state conduction in topological insulator materials facilitates liquid‐like charge flow between pre‐ and postsynaptic neuron layers representing the biological synaptic cleft operations. To optimize the material design, we investigated different proportions of 2D layered SnTe in SnSe matrix, thereby controlling the threshold set voltage in the neuristors. Furthermore, we prove the mechanism involved in low voltage switching is via the trapped charge limited current supported by the edge state transport in the QTI nanograin structure. Also, our QTNs demonstrate an ultrastable multistage synaptic switching capacity for multifunctionality, enabling high‐density data storage and processing behavior without sacrificing the physical dimensions of the device. Top notch, first‐rate neuromorphic characteristics of our proposed QTNs are successfully demonstrated by applying external presynaptic spike pulse trains. We represent the bionic modulations in QTNs via the fluctuations in the synaptic weight, representing the simulation of human neural network functions, including PPF and PTP synaptic function with learning‐relearning‐forgetting cycle effectiveness. Here, we deliberately claim that the energy requirement for the QTNs to perform the synaptic functions responding to external stimuli is in the range of 10 pJ, the lowest for any reported neuromorphic synaptic devices and which is 90 times lower than conventional Si‐cMOS neuromorphic circuits. [ \n \n 68 \n , \n 75 \n \n ] Finally, to demonstrate outstanding bionic characteristics, we effectively developed a sensory‐neuromorphic concept that incorporates the ANNs through a synaptic human–computer interface based on human‐gesture recognition which demonstrates an exceptional feasibility and exceptional recognition rate of 99%. Thus, the proposed QTNs are the first‐of‐a‐kind artificial synaptic devices with incomparable potential to become the new modern standard for the bioinspired neuromorphic device for next‐generation artificial intelligence applications.",
"introduction": "1 Introduction Synapses are the basic processing units in human brain responsible for daily operations and memory. [ \n \n 1 \n , \n 2 \n \n ] With 10 15 synapse interconnecting 10 11 neurons and only 10 femtojoules/synaptic process, the brain itself requires less than one‐millionth of the power consumed by a supercomputer for a process. [ \n \n 3 \n , \n 4 \n , \n 5 \n \n ] In the fact, over 8% of consumed global electricity are for computing devices which is also doubling in every decade, there is a challenging desire for developing efficient design of new materials and devices with low energy requirements compared to the conventional Si‐cMOS based architecture. [ \n \n 6 \n , \n 7 \n , \n 8 \n , \n 9 \n \n ] In this direction, by biomimicking the human nervous system and their biological synapse units, great efforts are underway to produce an artificial neuromorphic synapse device to learn and respond to the electronic neural signals from other artificial sensory devices/systems. [ \n \n 10 \n , \n 11 \n , \n 12 \n , \n 13 \n \n ] \n 1.1 Conventional Design of Neuromorphic Electronics Traditionally, silicon‐based complementary metal–oxide semiconductor are used for the design of neuromorphic computing systems to mimic the human biological neural systems for energy efficient operations. [ \n \n 6 \n \n ] Recently, Si‐cMOS‐based prototypes for neuromorphic chips were demonstrated in major projects such as SpiNNaker, [ \n \n 14 \n \n ] TrueNorth, [ \n \n 15 \n \n ] NeuroGrid, [ \n \n 16 \n \n ] BrainScaleS, [ \n \n 17 \n \n ] and Braindrop. [ \n \n 18 \n \n ] Though these projects demonstrate a level of energy efficiency, they are limited in their dynamic scalability, device volatility and ineffective plasticity affects their applications in adaptivity learning and complex processes. [ \n \n 19 \n , \n 20 \n \n ] Also, the traditional silicon cMOS architecture has reached the limit of key quantities like Size, Weight and Power (SWaP) for further scalability in device performance. This arises an additional challenge to reduce SWaP in artificial synaptic device with augmented device architecture. [ \n \n 21 \n \n ] Advantageously toward artificial synapse, memristors possess a great advantage to satisfy the SWaP requirements of excellent scalability, nonvolatile devices integrated with logics and memory units. [ \n \n 22 \n , \n 23 \n , \n 24 \n \n ] \n 1.2 Postsilicon Neuromorphic Electronics On the contrary to cMOS devices, memristors are bioinspired electronic devices with exact functions of biological synaptic and neural systems. [ \n \n 25 \n , \n 26 \n \n ] Unlike in cMOS devices, the basic device architecture of memristors is relatively simple with an active layer material of choice between two electrodes. Concurrently, this device design also resembles the biological synapses interconnecting pre/postsynaptic neuron clusters. [ \n \n 20 \n , \n 27 \n \n ] With organic/inorganic and oxide‐based semiconductors as active materials in memristor design, various underlying mechanisms like filament formation, [ \n \n 28 \n \n ] ionic conduction, [ \n \n 29 \n \n ] trapping/detrapping, [ \n \n 30 \n \n ] defect migration, [ \n \n 31 \n \n ] and redox reactions [ \n \n 32 \n \n ] are being put forth as efficient synaptic switching process prove their neuromorphic functions. [ \n \n 24 \n \n ] Based on the material of fabrication, their characteristic response to an applied external (optical or electric or magnetic) stimulus varies owing to various factors such as energy consumption, electronic conduction, and phase fluctuations. [ \n \n 33 \n \n ] Despite the effective synaptic device design and mechanisms, the feasibility of training the memristive synaptic device to respond to external stimuli with ultralow power consumption (pJ) is poor, as they are specifically designed for memory and logic gates. Thus, it is well understood that the long‐standing issue with development of neuromorphic devices/systems for commercialization is the lack of efficient materials to mimic biological synapses with equivalent performance as the brain. [ \n \n 34 \n , \n 35 \n , \n 36 \n , \n 37 \n , \n 38 \n \n ] \n 1.3 Quantum Topological Insulators In this regard, quantum topological insulators (QTI) with topological phases illustrates a unique edge conducting states arising from their time reversal invariant band inversion. [ \n \n 21 \n , \n 39 \n \n ] These properties make them resistant to defect induced back scattering and spin–momentum locking at the surface states creates a topologically nontrivial narrow energy gap at room temperature. [ \n \n 40 \n \n ] The proven nontrivial electronic characteristics of topological insulator materials are distinct from metals and insulators, making them a special class of quantum materials. [ \n \n 41 \n \n ] The edge state transport with tunable energy gap in topological insulator materials makes them the perfect candidate for implementing artificial synaptic device for neuromorphic systems, as the synaptic switching current for turning of the device from nonconducting to conducting state flows via the edge states. Ultimately, the edge state transport consumes ultralow energy with high switching speed. [ \n \n 39 \n , \n 42 \n \n ] Most recently, QTIs like Bi 2 Se 3 and Bi 2 Te 3 are being explored deeply for their application in low power electronic devices and has proved to be most efficient in comparison to conventional silicon devices. [ \n \n 41 \n , \n 43 \n , \n 44 \n , \n 45 \n , \n 46 \n , \n 47 \n , \n 48 \n , \n 49 \n \n ] \n 1.4 Quantum Topological Neuristors Design In the stage of developing a quantum neuromorphic supercomputer with equivalent power requirements as human brain, the primary requirement of realization is to fabricate an effective artificial synaptic device unit. [ \n \n 50 \n \n ] By hybridizing QTI materials with memristive device design, here we introduce development of a new class of artificial synaptic device, neuristors, explicitly for neuromorphic systems. We use optimized proportions of 2D layered tin selenide (SnSe) and tin telluride (SnTe) QTIs via electrochemical process for the growth of synaptic clusters. Figure \n 1 a conceptually illustrates the hierarchical order in the bioinspired development of QTI‐based artificial synaptic neuristors. Here, our QTI‐based artificial synaptic neuristor design is an exact replica of biological synapses with postsynaptic and presynaptic electrode array/connections as silver contacts, neurotransmitters as trapped charge carriers. The conceptual Figure 1a explains the emulation process of QTI‐based artificial synaptic device from the biological synaptic cluster formations. Figure 1b depicts the structural resemblance of a neuron to our developed array of synaptic neuristors. Moreover, the communication signal pathways are more homogenously biomimicked between the synaptic dendrites in neurons and the artificial synaptic neuristor array. Figure 1c shows the material compatible advantages of using topological insulator materials for developing artificial synaptic neuristors. Optimizing the interplay of quantum mechanics in the topologically protected surface states of SnSe 1− \n \n x \n Te \n x \n will facilitate to achieve massive parallelism, high endurance and scalability in our quantum topological neuristors. With this approach, we meticulously emulate the biological synaptic actions in quantum topological neuristors (QTN) demonstrating high switching speed (µs) and ultralow power consumption (pJ). Moreover, we exhibit the neuromorphic potentials of our QTNs by their short‐/long‐term memory capacity with learning and forgetting cycle characteristics. The effective neuromorphic performance of the QTNs is also illustrated by associating with artificial neural networks (ANNs) through pattern recognition of hand gestures. With these first‐rate neuromorphic characteristics, QTNs prove to be the first of kind artificial synaptic device with incomparable potential for next generation artificial intelligence like prosthetics and humanoids. Figure 1 Biomimicking hierarchy from human brain to quantum topological neuristor. a) Conceptual illustration of the quantum topological neuristor design derivation from the synaptic neuron clusters function in human brain. b) Physical characteristics and communication pathway comparison of biological neuron to the artificial synaptic neuristor array. c) Advantages of quantum topological insulator materials for realizing artificial synaptic neuromorphic functions."
} | 2,997 |
36573789 | PMC9872165 | pmc | 2,106 | {
"abstract": "There has been substantial recent interest in the promise\nof sustainable,\nlight-driven bioproduction using cyanobacteria, including developing\nefforts for microbial bioproduction using mixed autotroph/heterotroph\ncommunities, which could provide useful properties, such as division\nof metabolic labor. However, building stable mixed-species communities\nof sufficient productivity remains a challenge, partly due to the\nlack of strategies for synchronizing and coordinating biological activities\nacross different species. To address this obstacle, we developed an\ninter-species communication system using quorum sensing (QS) modules\nderived from well-studied pathways in heterotrophic microbes. In the\nmodel cyanobacterium, Synechococcus elongatus PCC 7942 ( S. elongatus ), we designed,\nintegrated, and characterized genetic circuits that detect acyl-homoserine\nlactones (AHLs), diffusible signals utilized in many QS pathways.\nWe showed that these receiver modules sense exogenously supplied AHL\nmolecules and activate gene expression in a dose-dependent manner.\nWe characterized these AHL receiver circuits in parallel with Escherichia coli W ( E. coli W) to dissect species-specific properties, finding broad agreement,\nalbeit with increased basal expression in S. elongatus . Our engineered “sender” E. coli strains accumulated biologically synthesized AHLs within the supernatant\nand activated receiver strains similarly to exogenous AHL activation.\nOur results will bolster the design of sophisticated genetic circuits\nin cyanobacterial/heterotroph consortia and the engineering of QS-like\nbehaviors across cyanobacterial populations.",
"conclusion": "3 Conclusions In this study, we designed,\nbuilt, and tested an inter-species\ncommunication system based on genetic circuitry for quorum sensing.\nWe showed that the three systems (Lux, Tra, and Las) could sense and\nrespond to both exogenous and secreted AHL signals. Broadly, the circuit\nresponse patterns in S. elongatus were\ncomparable to those in E. coli , though\nwith increased levels of background expression led to lower induction\nratios. We demonstrated inter-species communication in direct co-cultivation,\nraising the prospect of this system for use in applications requiring\nmultiple species, such as the division of labor in bioproduction.\nThis is the first example of quorum sensing systems that have been\nused to generate inducible promoters and cross-species gene regulation\nin S. elongatus . This work contributes\nto the prospect of light-driven, sustainable bioproduction through\nthe coordination of microbial partners.",
"introduction": "1 Introduction Cyanobacteria are increasingly\nexplored as a potential chassis\nfor the bioproduction of valuable compounds from sustainable inputs\n(e.g., sunlight, CO 2 , and non-potable water streams). 1 , 2 The focus on increasing the cyanobacterial production of compounds,\nsuch as polymers, pigments, and biofuels, is dominated by genetic\nengineering. 3 − 5 While there is strong interest in chemical biosynthesis\nwithin a single photosynthetic species, 6 interest in designer consortia has recently increased. Multi-trophic\nconsortia of bioproduction-optimized cyanobacteria and heterotrophs\ncan distribute metabolic labor between carbon fixing cyanobacterial\nstrain(s) and co-cultivated heterotrophic microbes. Heterotrophic\nspecies (e.g., Escherichia coli ( E. coli ), Bacillus subtilis , and Saccharomyces cerevisiae ) within\nthese consortia utilize secreted photosynthates and contribute to\nthe productivity of the consortia by compartmentalizing key metabolic\nreactions or by cross-feeding interactions that improve the yield\nof cyanobacterial biomass. 7 Generally,\nengineered consortia offer valuable features for bioproduction such\nas improved robustness, efficient utilization of inputs, and metabolic\nderivatization. 8 , 9 However, the construction of robust\nmicrobial communities remains a grand challenge across many microbial\nbioengineering fields and requires the development of new genetic\ncircuits and engineering standards to coordinate activities across\npartner species. Adaptive, population-level control systems are vital\nfor the success of these future applications. By containing many mechanisms\nthat enhance community stability, natural microbial communities inspire\nmodern engineering strategies. Natural microbial communities\norchestrate collective behaviors\nwith cell–cell signaling processes collectively referred to\nas quorum sensing (QS). QS communication involves the production,\nsecretion, and accumulation of soluble signaling molecules, known\nas autoinducers. 10 These diffusible environmental\nsignals are monitored and indicate the abundance of neighboring microbes.\nWhen these molecules bind to their cognate receptors, a cascade of\ngene expression is typically affected, often including increased production\nof the autoinducer molecule itself. This feed-forward loop ( i.e. , positive feedback) helps to coordinate gene expression\nwithin the population. While the chemical nature of the diffusible\nautoinducer signal varies, many well-studied QS processes involve\nthe use of acyl-homoserine lactones (AHLs), which have a lactone ring\nand a 4–18 acyl carbon side chain and readily diffuse through\ncell membranes. AHLs were discovered in the marine bacterium Vibrio fischeri , where they synchronize the production\nof the bioluminescence pathway when a quorum is reached. 11 Since then, it has been revealed that AHL-dependent\nprocesses control a broad range of behaviors in Gram-negative bacteria,\nincluding biofilm production, pathogenicity, secondary metabolite\nproduction, and competence. 12 , 13 Bacteria often combine\nthe information transmitted by multiple types of QS autoinducers for\nsynchronous intra- and inter-species communication. 14 Recent synthetic biology circuits designed to program\npopulation-level\nbehaviors have been inspired by naturally occurring QS signaling pathways,\nresulting in a relatively well-characterized toolkit of genetic parts\n(e.g., promoters and genes) for AHL production and detection. 15 , 16 Modification of native systems changed sensitivity 17 , 18 and promoter orthogonality. 19 A recent\nstudy repurposed the underlying genetic circuitry to control native\nmicrobiomes in humans and plants as well as synthetic consortia. 20 Just as in natural communities, processes that\nimpose heavy metabolic burdens on individuals in an engineered microbial\nsystem can be futile if efforts fail to coordinate across the population.\nRecently reported demonstrations of the utility of such coordination\ninclude genetic circuits for signal oscillation, 21 differential gene expression, 22 maintenance of culture density, 23 and\ndefined social interaction. 24 Toward\nthe coordinated division of labor for light-driven bioproduction,\nwe created an inter-species communication system based on QS modules.\nWe installed and verified the functionality of three well-studied\nquorum sensing pathways (Lux, Las, and Tra) in Synechococcus\nelongatus PCC 7942 (hereafter S. elongatus ) with exogenously added AHLs. 25 − 29 The AHL synthases were expressed in E. coli W Δ cscR (hereafter E. coli ), which can utilize sucrose as a carbon source. Additionally, the cscB and sps genes were expressed under S. elongatus AHL promoters to regulate the secretion\nof sucrose in the growth medium by linking its export with E. coli population density. To the best of our knowledge,\nthis is the first time quorum sensing modules have been used in cyanobacteria\nto tune gene expression for cross-species communication.",
"discussion": "2 Results and Discussion 2.1 Receiver Construction and Characterization 2.1.1 Design and Rationale of AHL Cyanobacteria\nReceiver Strains Quorum sensing pathways exhibit diversity\nin the signal molecules that mediate coordination across a bacterial\npopulation. The AHL class of signals is small and relatively hydrophobic,\nwhich makes them diffusible across biological membranes. 14 In part due to this feature, AHLs have been\nused more extensively than other quorum sensing pathways in synthetic\ncircuit designs. In particular, 3-oxo-hexanoyl-HSL (3OC6-HSL), N -(3-oxooctanoyl)-HSL (3OC8-HSL), and 3-oxo-dodecanoyl-HSL\n(3OC12-HSL) have been effective in other heterologous circuits 21 , 28 , 29 and can be synthesized by the\nenzymes LuxI, TraI, and LasI, respectively ( Figure 1 ). The structural mechanism of activation\nof members of these transcription factors involves the binding of\nan AHL molecule in a hydrophobic pocket at the N-terminus of the transcriptional\nactivator (LuxR to 3OC6-HSL; TraR to 3OC8-HSL; LasR to 3OC12-HSL),\nwhich promotes correct folding of this sensing domain and prevents\nproteolytic degradation. 30 , 31 A C-terminal helix-turn-helix\nmotif in this protein also directs the stabilized protein to bind\nappropriate promoter sequences and activate downstream gene expression. 32 Figure 1 Overview diagram illustrating the quorum sensing-based\ngenetic\ncircuit design. AHL molecules (black dots) are synthesized in the\nsender cell (yellow; E. coli ) and\ndiffuse through the environment into the receiver cell (green; S. elongatus ). Within the receiver cell, AHL molecules\nare bound by the transcriptional activator gene product (white, i.e. , LuxR, LasR, and TraR), which activates the target\ngene of interest (orange; GOI) through the quorum sensing promotor\n(P QS ). To enable cross-species communication, we divided\nthe AHL signaling\npathway across E. coli “sender”\nand S. elongatus “receiver”\nstrains ( Figure 1 ).\nThe genes responsible for sensing the respective signaling molecules\nwere genomically integrated into S. elongatus under an IPTG-inducible trc promoter. In the context\nof characterizing the cross-species circuits, the inducible promoters\nwere used to determine if we could tune the sensitivity of the AHL\nreceiver circuits. 2.1.2 Tunable Gene Expression Responsive to AHL\nConcentrations is Driven in Engineered Receiver S.\nelongatus We first focused upon the capacity\nof S. elongatus to respond to AHL signals\nwhen expressing a cognate transcription factor by using the fluorophore\nmNG as a reporter protein under the control of an appropriate promoter\n( i.e. , P lux , P tra , and P las ). We monitored\nthe fluorescence in strains with the integrated AHL receiver module\nthat controls the expression of a mNeongreen (mNG) reporter ( Figure 1 ) using flow cytometry\nat different concentrations of the IPTG inducer and exogenously supplied\nAHL ( Figure 2 ). Figure 2 Characterization\nof quorum sensing receiver modules in S. elongatus . Expression of all LuxR family members\nwas controlled by IPTG-inducible P trc promoters\n(see Figure 1 ). (A–D)\nCharacterization of the Lux system. (A) Ridge plot of kernel-density\nestimation fits to peak intensity of cells induced with 5 μM\nIPTG and increasing concentrations of 3OC6-HSL. (B) Dose–response\ncurve of the 3OC6-HSL concentration vs mNG intensity, with fit to\nthe Hill equation (solid lines). Culture co-induced with 0, 5, or\n500 μM IPTG and increasing concentrations of 3OC6-HSL shown\nhere; other IPTG concentrations were omitted for clarity. (C) Heatmap\nof the induced mNG signal normalized to the highest intensity under\nexperimentally varied inducer concentrations. (D) Induction ratio\nvs IPTG concentration, calculated as the maximum over the minimum\nmNG intensity at specific IPTG concentrations relative to uninduced\ncontrols. (E–H) Characterization of the Tra system; (I–L)\ncharacterization of the Las system: same as panels (A–D) but\nwith specified AHL inducers, 3OC8-HSL and 3OC12-HSL, respectively.\n(A–L) All measurements were taken 24 h after induction with\nspecified concentrations of exogenous AHLs and IPTG. S. elongatus strains\nencoding LuxR\nexhibited a mNG signal that was positively correlated to the level\nof 3OC6-HSL added to the culture and with limited cell-to-cell variation\nacross the population ( Figure 2 A). As anticipated, the degree of LuxR expression altered\nthe sensitivity of S. elongatus strains\nto AHLs; the LuxR concentration is known to be correlated with 3OC6-HSL\nsensitivity. 19 While mNG expression was\nobserved at AHL concentrations ≥10 –7 M 3OC6-HSL\nat basal expression levels, increased IPTG was negatively correlated\nwith the amount of AHL required to induce a significant change in\nthe mNG reporter ( e.g. , ∼10 –9 M 3OC6-HSL at 500 μM IPTG; Figure 2 B and Figure S1A ). These trends of increased sensitivity at higher LuxR expression\nand higher reporter expression at higher AHL were consistent across\na broad range of inducer concentrations ( Figure 2 C). The ratio of mNG expression at maximal\nAHL concentrations was ∼8-fold higher than the basal expression\nin the absence of AHL ( i.e. , P lux ON/P lux OFF; Figure 2 D). Across all combinations\nof IPTG and AHL, the inter-cellular variation in reporter expression\nwas minimal ( Figure S2 ). By contrast,\nthe heterologous expression of TraR in S. elongatus did not confer dose-dependent reporter\nexpression in response to exogenous 3OC8-HSL, despite using both a\nmodified TraR (E192W) point mutant and synthetic promoter P tra* that were previously reported to increase the\nsensitivity and dynamic range ( Figure 2 E–H). 29 A ∼50%\nincrease in mNG reporter expression was observed at high concentrations\nof 3OC8-HSL relative to basal expression levels ( Figure 2 F), and the sensitivity of\nthe circuit was not changed by increasing TraR expression levels ( Figure 2 G). Finally,\nLasR-based circuits exhibited a higher dynamic range of\nresponse to the corresponding addition of 3OC12-HSL ( Figure 2 I–L). A lower basal\nlevel of reporter expression under P las was observed ( Figure 2 I), and the circuit was activated at substantially lower AHL concentrations\n( Figure 2 J; ∼10 –9 M 3OC12-HSL). However, this las promoter\nexhibited less tunability as expression levels of LasR were increased\nthrough IPTG addition. At IPTG concentrations ≥50 μM,\nthe dynamic range of induction was dramatically reduced ( Figure 2 K). At the highest\nlevels of IPTG and 3OC12-HSL, decreased growth and chlorosis of S. elongatus cultures were observed ( Figure S3 ), possibly indicating that LasR activity\nwas associated with aberrant activation of native genes. WT strains,\nor strains without induced LuxR family genes, displayed no fitness\ndefects in the presence of high concentrations of AHLs otherwise ( Figure S3 ). Altogether, LuxR and LasR exhibited\nthe capacity to promote inducible\ngene expression in S. elongatus in\nresponse to the cognate AHL signals. LasR exhibited the highest dynamic\nrange of expression (∼13-fold, relative to ∼8-fold for\nLuxR; Figure 2 D,K).\nHowever, high expression of LasR under conditions where this transcription\nfactor was stabilized led to impaired fitness—possibly due\nto incomplete orthogonality of downstream gene regulation. TraR-based\ncircuits did not perform as expected in S. elongatus , failing to induce more than a ∼50% change in reporter expression\neven over a large range of AHL concentrations tested ( Figure 2 E–H). To the best of\nour knowledge, the heterologous expression and characterization of\nLux/Las/Tra family members have not previously been reported in S. elongatus . 2.1.3 Species-Dependent and Species-Independent\nFeatures of AHL Circuit Design Some of the features of the\nAHL-dependent circuits characterized in S. elongatus were distinct from similar results described previously in E. coli , 28 , 29 which prompted us to\ninterrogate if these were species-specific effects of the performance\nof AHLs in cyanobacteria. We therefore expressed genetic receiver\nconstructs for the three systems described above in E. coli to evaluate if these effects were species-specific\nor were attributable to the genetic circuit design itself. E. coli strains expressing luxR under\nan IPTG-inducible trc promoter exhibited behavior\nthat was similar to the S. elongatus counterpart in sensitivity, magnitude of induction, and total fluorescence\nreporter yield ( Figure 3 A–C). As shown before ( Figure 2 A–D), increasing IPTG levels led to an enhanced\nsensitivity of reporter output to lower concentrations of cognate\nAHL, with a maximal induction of reporter output between 10 –8 and 10 –7 M ( Figure 3 A). The relative mNG fluorescence values of E. coli at maximal induction were similar to those\nobtained in S. elongatus , although\nthe basal level of fluorescence in the absence of inducer was considerably\nhigher in the cyanobacterial model, even when accounting for the auto-fluorescence\nof the photosynthetic pigments. Due to the higher overall expression\nin E. coli , the raw mNG intensities\nbetween species were not able to be directly compared and distinct\nfluorescence detector settings were used to avoid signal saturation\n(see section 4.4 ).\nNevertheless, there was broad agreement when comparing the relative\nfluorescence outputs of the two species to the same inputs, although E. coli with the basal expression of LuxR in the\nabsence of IPTG exhibited a higher sensitivity to 3OC6-HSL ( Figure 3 C). Additionally,\nthe induction ratio in E. coli was\nmuch higher, with nearly a ∼190-fold difference in the ON/OFF\nstates ( Figure S4 ), vs ∼8-fold in S. elongatus ( Figure 2 D), as the result of both lower basal and higher maximal\nexpression. Figure 3 Characterization of quorum sensing receiver modules in E. coli . Expression of all LuxR\nfamily members was controlled by IPTG-inducible P trc promoters. (A–C) Characterization of the Lux system.\n(A) mNG intensity of cells co-induced with 0, 5, or 500 μM IPTG\nand increasing concentrations of 3OC6-HSL, with fit to the Hill equation\n(solid lines). (B) Heatmap of the induced mNG signal normalized to\nthe highest intensity. (C) Comparison of relative expression between E. coli and S. elongatus . Panels (D–F) and (G–I) are the same\nas panels (A–C) but with a specified AHL inducer (see top).\n(A–I) All measurements were taken 24 h after induction with\nspecified concentrations of exogenous AHLs and IPTG. Similarly, the TraR-based reporter circuit in E.\ncoli exhibited many features that were observed in S. elongatus , including a relatively slight increase\nin total reporter output at maximal induction ( Figure 3 D–F). When IPTG was added to induce\nTraR, increasing levels of 3OC8-HSL were correlated with increased\nmNG output, albeit with a much less pronounced total accumulation\nof the fluorescence reporter relative to the LuxR-based system ( Figure 3 D). Again, the basal\nlevel of reporter expression from the TraR circuit was much lower\nin E. coli than in S.\nelongatus , leading to a maximal ON/OFF induction ratio\nin E. coli that was ∼4-fold\n( Figure S4 ), much higher than in the cyanobacterial\ncounterpart but in contrast with previous work that showed up to 10-fold. 29 Although we are unaware of any mechanistic rationales\nto explain the elevated basal level of expression in S. elongatus , this property makes the TraR system\neffectively unusable at present. Finally, the LasR-based reporter\ncircuit in E. coli responded similarly\nto the observations in S. elongatus , including reduced cellular growth at higher expression levels of\nthe receptor ( Figure S4 ). At lower levels\nof IPTG induction, E. coli exhibited\nsensitivity to 3OC12-HSL in the range of 10 –9 to\n10 –8 M ( Figure 3 G), similar to S. elongatus ( Figure 2 I), although E. coli basal expression of the mNG reporter was\nagain lower. We also observed a severe growth defect with the LasR\nconstruct that was expressed in E. coli with high IPTG inducer concentrations, where growth was negatively\naffected by 3OC12-HSL concentrations in a dose-dependent manner. Taken\ntogether, this suggests that conditions favoring a high accumulation\nof the Pseudomonas aeruginosa LasR\nprotein may be broadly cytotoxic in both species ( Figures S3 and S5 ). Curiously, LasR overexpression in its\nnative host ( P. aeruginosa ) has been\npreviously shown to cause a large growth burden in multiple environments, 33 suggesting that this protein itself may have\ncytotoxic properties, possibly due to transcriptional activation of\noff-target genes. In total, several features of the quorum sensing\ncircuits exhibited\nconsistency when installed in different model organisms, while a more\nlimited set of species-specific characteristics might explain other\nvariations observed. Generally, LuxR family members exhibited sensitivity\nsimilar to their corresponding cognate AHL when expressed in either S. elongatus or E. coli ( Figures 2 and 3 ). Across all three AHL circuits, E. coli strains exhibited a lower basal level of\nexpression in the absence of inducer molecules, and this “tighter\nOFF” state contributed to a larger induction ratio of all three\ncircuits relative to S. elongatus .\nCuriously, a higher expression level of LasR was associated with the\npoor performance of this circuit in both species, which may indicate\nbroad non-specific interference in gene regulation by this protein\nand/or increased metabolic burden. 33 2.2 Sender Construction and Characterization for\nAHL Production 2.2.1 AHL Production by E. coli Senders E. coli sender strains\nwere constructed by encoding each of the AHL synthases ( i.e. , LuxI, TraI, and LasI) on autoreplicative plasmids with the pBbA2c\nBglBrick backbone, with heterologous expression driven by an aTc-inducible\npromoter ( tet ). 34 To characterize\nthe AHL production of each strain, we first verified that these strains\ncould indeed biosynthesize the expected AHLs ( Figure S6 ). We subsequently measured the concentrations of\n3OC6-, 3OC8-, and 3OC12-HSLs in the supernatant after induction using\nLC–MS and found concentrations in the range of 300–350\nnM for each AHL ( Figure 4 A,B). A peak in the concentration of each AHL was observed around\n∼13 h post-induction, after which measurable AHLs in the supernatant\ndecreased ( Figure S7A ). To be able to quantify\nthe maximal AHL productivities, a standard curve correlating cell\ndensity with biomass was measured ( Figure S9 ). In this range, maximal AHL productivities were in the range of\n9–11 nmol g dw –1 h –1 ( Figure S7B ). Furthermore, growth curves of the E. coli sender strains were taken ( Figure S8 ). Interestingly, the AHL production did not cause\nany metabolic burden to E. coli as\nthey had the same growth with the wild type. Figure 4 Cross-species activation\nof QS circuits. (A) Diagram illustrating\nthe experimental workflow. E. coli cultures\nexpressing AHL synthases were grown for 13 h before measuring the\nAHL concentrations and transferring the supernatant to induce expression\nof mNG in S. elongatus cultures. (B)\nAHL concentrations produced by E. coli after 13 h of cellular growth in BG-11 co , measured by\nLC–MS. (C) Reporter intensity after induction with supernatants\nfrom the respective AHL producer cultures. 2.2.2 Sender Strains Drive Cross-Species Activation\nof Receiver Modules We experimentally validated the induction\nranges of our three S. elongatus receiver\nconstructs with biosynthetic AHLs recovered from the supernatant of E. coli sender strain cultures ( Figure 4 C). We observed a strong correlation\nbetween sender culture supernatant AHL and the level of mNG induction\nin the receiver Lux system ( Figure 4 C, left), which agreed with our previous observation\nwith exogenously added 3OC6-HSL ( Figure 2 A,C). By diluting the supernatant recovered\nfrom E. coli to defined AHL concentrations,\nwe also observed an IPTG-dependent increase in sensitivity to the\nAHL molecules, in agreement with prior results when supplying exogenous\n( i.e. , commercially obtained) AHLs. The Tra system\nalso showed an increase in expression with supernatant 3OC8-HSL but\nexhibited no significant sensitivity to AHL with increased IPTG ( Figure 4 C, center), which\nwas consistent with our previous data ( Figure 2 E–G). In the Las system, we observed\nminimal sensitivity to 3OC12-HSL levels but a strong negative correlation\nbetween the IPTG concentration and the circuit activation ( Figure 4 C, right), which\nwas similar to results seen in the exogenous AHL experiments ( Figure 2 I–K). These\nresults indicate that sender strain-produced AHLs can induce dose-dependent\nresponses within the receiver strains. The Lux system response can\nalso be tuned by changing IPTG concentrations. Given the minimal response\nof the Tra receiver module, we opted to omit it from further study. 2.3 Cross-Species Communication in the Mixed Culture 2.3.1 Sender Strains Activate Receiver Modules\nin the Co-Culture After characterizing both the sender and\nreceiver strains in axenic cultures, we sought to determine whether\nthese species could indeed be co-cultivated and if the underlying\ngenetic circuitry for communication would function as expected. To\nassess combined circuit function, S. elongatus and E. coli strains were grown together\nfor up to 48 h and samples were taken for flow cytometry to measure\nthe mNG reporter fluorescence ( Figure 5 A). The Lux sender generated a strong response in its\ncognate receiver ( Figure 5 B). The reporter fluorescence with the LuxI sender strain\nwas nearly 20% higher than with exogenously added 3OC6-HSL at 24 h\n( Figure 2 B). The basal\nlevel of expression also increased by nearly ∼2-fold, yielding\nan effective induction ratio of ∼6.5, which was similar to\nthe result under axenic growth and exogenously added 3OC6-HSL ( Figure 2 D). Likewise, co-cultures\nwith the LasI sender strain yielded strong mNG expression ( Figure 5 C). As with the Lux\nsystem, we observed a higher background with LasR/LasI and comparable\ncircuit performance to the axenic culture with exogenous 3OC12-HSL,\nwith a maximum induction ratio of 11.3 ± 1.4 after 48 h ( Figure 2 L). Figure 5 Direct co-cultivation\nof E. coli and S. elongatus . (A) Diagram depicting\nthe experimental design . E. coli and S. elongatus were cultured together\nin BG-11 co supplemented with 20 g L –1 sucrose. Samples were taken for flow cytometry measurements after\n24 and 48 h. (B) LuxI/LuxR co-cultivation, IPTG = 1 mM. (C) LasI/LasR\nco-cultivation, IPTG = 0 mM. IPTG was omitted to avoid growth inhibition\nobserved at high LasR expression ( Figure S3 ). (B, C) Control samples used E. coli W strains without the AHL synthase (LuxI/LasI) construct but were\notherwise treated identically. 2.4 Applications and Outlook for Inter-Species\nCoordination in Light-Driven Communities Although AHL-based\nsignaling pathways are widely utilized across many prokaryotes, the\nliterature possesses relatively little documentation of cyanobacterial\nspecies that utilize these—or other QS systems—to control\ntheir endogenous population-level behaviors. Readily identifiable\nhomologues of the LuxI/LuxR family are not encoded in most sequenced\ncyanobacterial genomes. 35 Nonetheless,\nAHLs are routinely found as extracellular metabolites in microbial\ncommunities dominated by cyanobacteria, such as cyanobacterial blooms, 36 , 37 and exogenously added AHL signals have been shown to alter growth\nor other physiological characteristics in a limited number of axenic\ncyanobacterial cultures. 38 − 42 The cyanobacterial species Microcystis aeruginosa and Gloeothece sp. PCC 6909 have been shown to\ndirectly secrete some AHL molecules, 42 , 43 although their\nphysiological function is poorly understood. Despite increasing indirect\nevidence that AHLs play an important role in controlling the microbial\ndynamics of natural cyanobacterial blooms, 37 , 44 there is limited mechanistic understanding of cyanobacterial AHL\nsensing or if such pathways are directly utilized by cyanobacteria\nto control QS behaviors in blooms or other natural contexts. More advanced forms of QS circuits in S. elongatus (in axenic or mixed cultures) could be designed if cyanobacteria\ncan be engineered to also secrete AHL signals in addition to sensing\nthem. Toward this goal, we performed preliminary experiments to express\nboth LuxI and LasI in S. elongatus under\nthe IPTG-inducible P trc promoter. We monitored\nthe abundance of 3OC6-HSL and 3OC12-HSL in the supernatant of these\ncultures during a 4 day time course following IPTG induction ( Figure S10 ). We observe maximal production levels\nof LuxI-expressing strains to be 6 nM 3OC6-HSL, while LasI-expressing\nstrains accumulate up to 500 nM 3OC12-HSL. These results demonstrate\nthat it is feasible to also install AHL production pathways in S. elongatus , although they also indicate that additional\noptimization may be necessary to utilize these for QS circuits that\nrely on cyanobacterial secretion of AHLs. For instance, it may be\nnecessary to boost 3OC6-HSL levels produced by S. elongatus for us to allow sufficient activation of LuxR receivers ( Figure S10A : ∼6 nM 3OC6-HSL is at the\nlower edge of detection for LuxR we report in Figure 2 ). Conversely, while 3OC12-HSL production\nfrom LasI-expressing S. elongatus is\n2 orders of magnitude higher ( Figure S10B ), LasR receiver strains exhibit a poorly understood toxicity at\nhigh activation levels ( Figure S3 ). Therefore,\nconstruction of QS self-inducing systems in S. elongatus is likely to require further research and optimization. Cyanobacterial\ngenetic circuits that can be coupled to sensing\nof population density could provide useful features for biotechnological\napplications, yet the limited information on any endogenous cyanobacterial\nQS pathways has hindered their development. As stated earlier, mixed\nmicrobial consortia are an emerging area of interest for many microbial\nbioproduction applications, and QS pathways are a primary mechanism\ninvolved in coordination and stabilization of natural microbial communities\nand symbiotic interactions. 45 Synthetic\nQS circuits have been developed in other contexts to drive advanced\nbehaviors in mixed-species consortia, such as dynamic oscillation\nof gene expression, 46 maintenance of a\ndesired ratio of partner species, 23 and\nadaptive gene expression responsive to physical spacing between different\nmicrobial strains. 47 Adapting such strategies\nto engineered microbial communities is likely to be necessary for\ntheir optimal performance and robustness in any real-world applications. To highlight the potential utility of QS circuits for bioproduction,\nwe used the Lux circuit to regulate expression of sucrose secretion\nmachinery ( Figure 6 A) that has been well characterized. 48 − 50 Briefly, expression\nof some forms of the bifunctional enzyme sucrose phosphate synthase\n( sps ) leads to the accumulation of cytosolic sucrose\nin the absence of salt stress, while heterologous expression of sucrose\npermease ( cscB ) allows for export of sucrose from\nthe cytosol. In combination with its respective AHL (3OC6-HSL) and\nLuxR expression, P luxI drove the expression\nof cscB and sps , leading to dose-dependent\nsucrose secretion in response to AHL induction ( Figure 6 B). Sucrose accumulation from S. elongatus P luxI :: cscB :: sps could also be observed when directly\nco-cultivated with E. coli K-12 strains\nbearing the relevant sender circuit ( i.e. , P tet :: luxI ; Figure S11 ), providing an additional proof-of-principle instance\nrelative to the mNG reporter lines above ( Figure 5 ). In this experiment, E.\ncoli K-12 can signal the cyanobacterial partner through\nthe secretion of 3OC6-HSL (100 nM aTc; Figure S12 ) but cannot directly consume the sucrose because it does\nnot encode the relevant transporter or invertases. However, sucrose\naccumulation in the co-culture was lower than that in reference axenic S. elongatus P luxI :: cscB :: sps strains with exogenously added\nAHL (compare Figure 6 and Figure S11 ), perhaps due to a lack\nof readily accessible carbon source to support E. coli K-12 growth or AHL synthesis. Figure 6 The Lux system can be used to control\nbioproduction and secretion\nof sucrose. (A) Sucrose production and secretion modules cscB and sps were placed under the control of the P luxI promoter. (B) Quantification of sucrose\n(μM) secreted after quorum sensing induction at 24 and 48 h\npost-induction with 3OC6-HSL (M) and 500 μM IPTG as these concentrations\nwere shown previously to show the full range of induction. Other applications of population-sensing circuits\nin cyanobacteria\ncould be useful for axenic cultures. Coordinating population-level\nbehaviors could have multiple applications in batch cultures, for\ninstance, to decouple the early stages of culture growth from later\nactivation of bioproduction pathways. Many engineered strains of cyanobacteria\ndeveloped for bioproduction express heterologous pathways under promoters\nresponsive to exogenously added inducers ( e.g. , IPTG\nand metals 2 ), but scaled application of\ninducer compounds is financially costly. 51 To bypass this cost, strong and constitutively active promoters\nare routinely used to drive engineered pathways, although this approach\nis vulnerable when expressing high-burden circuits that can impede\nearly growth of the culture and/or promote strong counter-selection\nagainst retention of the introduced genes. 52 QS pathways could also be used to alter cellular properties to adapt\nand overcome the challenges associated with large-scale cultivation\nand harvesting processes. For instance, constructing circuits that\nactivate in the later phases of growth that can stimulate cell settling 53 , 54 or express cell wall degrading enzymes can reduce costs associated\nwith harvesting cells and “pre-treating” biomass for\ndownstream processes. 55 To our knowledge,\nonly one other study has demonstrated synthetic\ngene regulation in cyanobacteria by using heterologous QS signals\nin Synechocystis sp. PCC 6803. 56 Further refinement and characterization of a toolkit of\ncyanobacterial genetic “parts” are likely to be important\ncomponents of ongoing efforts to develop cyanobacteria as a chassis\nfor sustainable biotechnologies."
} | 8,489 |
36438874 | PMC9686847 | pmc | 2,107 | {
"abstract": "Climate change, socioeconomical pressures, and new policy and legislation are driving a decarbonization process across industries, with a critical shift from a fossil-based economy toward a biomass-based one. This new paradigm implies not only a gradual phasing out of fossil fuels as a source of energy but also a move away from crude oil as a source of platform chemicals, polymers, drugs, solvents and many other critical materials, and consumer goods that are ubiquitous in our everyday life. If we are to achieve the United Nations’ Sustainable Development Goals, crude oil must be substituted by renewable sources, and in this evolution, biorefineries arise as the critical alternative to traditional refineries for producing fuels, chemical building blocks, and materials out of non-edible biomass and biomass waste. State-of-the-art biorefineries already produce cost-competitive chemicals and materials, but other products remain challenging from the economic point of view, or their scaled-up production processes are still not sufficiently developed. In particular, lignin’s depolymerization is a required milestone for the success of integrated biorefineries, and better catalysts and processes must be improved to prepare bio-based aromatic simple molecules. This review summarizes current challenges in biorefinery systems, while it suggests possible directions and goals for sustainable development in the years to come.",
"conclusion": "4 Conclusion Biomass is meant to be the most critical source of carbon materials, as our economy gradually abandons coal and crude oil for the extraction of raw materials and fuels. The new bio-based economy grows in complexity, alongside higher crude oil prices, environmental concerns, corporate commitments, policy, and legislation. Many fuels, chemical building blocks, polymers, and solvents are already being produced from biomass and biomass waste in the new generations of biorefineries, and this production is expected to grow as the biorefinery system reaches the industrial scale. However, the cost of bio-based production still exceeds the cost of its petrochemical alternative, and bio-based products need to be proven to outperform their petrochemical counterparts. Moreover, this growth of a bio-based economy also implies an increase in feedstock demand, technology development, and business opportunities. Nowadays, biorefineries have achieved enough degree of development to start providing reliable alternatives to most of the products needed in our daily day life. However, our biorefinery system is not yet at the point of displacing existing fossil refinery technologies. In fact, rather than displacing or even replacing existing petrochemical platforms, the biorefinery systems should aim at integrating with current infrastructure. The development of integrated biorefinery processes that produce both bio-based products and energy carriers is the most efficient strategy to valorize biomass in the future bio-based economy. The existence of several biorefinery platforms under development and the co-existence of competitive technologies are promising. Although some of these alternatives might not be successful in the long run, the fact that several options coexist only increases the chances of having full scale-up integrated biorefineries in the future. Among several challenges ahead, it is necessary to improve many of the processes of biomass transformations to make them economically feasible in a continuous scaled-up operation. Additionally, it is necessary to develop new processing technologies, catalysts, and chemical reactions that enable the synthesis of substances that are not currently achievable by other means. In the case of carbohydrate-containing biomass, the current technological developments can perform almost anything, but the same does not apply to the lignin platform. For lignin, neither chemically nor biotechnologically platforms provide satisfactory results, despite lignin being the most abundant source of aromatic moieties in nature. Enabling a scaled-up lignin platform to produce aromatic molecules is probably one of the most important challenges for the next decade in this field. This is part of a greater challenge, which is to arrive at carbon-neutral industrial processes. To achieve this objective, biorefineries must consume renewable energy to reduce their carbon footprint, which implies a dialogue between several stakeholders and a strong push for system integration with other sectors and relevant technologies.",
"introduction": "1 Introduction Complying with net-zero targets, reducing greenhouse gas emissions, and transitioning to an environmentally, economically, and socially sustainable development are complex and colossal challenges that require a highly efficient and cost-effective biorefinery system, among other interrelated factors. Biorefineries must integrate with current infrastructure, while enabling the transformation of a wide range of biological feedstocks and waste into a portfolio of bio-based materials and specialty chemicals, whose cost do not exceed the cost of their petrochemical counterparts and whose performance is proven to be at least as good as that of the petrochemical equivalents. For centuries, the traditional uses of biomass have been limited to animal feeding, energy generation, i.e., heating or cooking, and some engineering and commodity applications, including wood for construction and furniture and natural fibers for textiles. Traditional processing techniques also enabled the obtention of biomass-derived products such as vegetable oils or flours using milling, essences, and fragrances via steam distillation, valuable platform chemicals through liquid–liquid extraction, and yogurt, kefir, wine, or beer, from fermentation processes. Despite its abundance, applicability, and use, the list of substances, materials, and key building blocks available from biomass to produce fine and specialty chemicals is still quite limited compared to those provided by the petrochemical industry. Furthermore, the lack of cost-efficient scaled-up processes and the lower performance of many biorefinery products versus petrochemical alternatives have kept our economy still largely based on fossil sources. Following the United Nations (UN) definitions of sustainability, it is our responsibility to meet the needs of the present without compromising the ability of future generations to meet their own needs. Nowadays, among the 17 sustainable development goals (SDGs) highlighted by the UN, there are at least six goals, i.e., clean water and sanitation, affordable and clean energy, industry innovation and infrastructure, sustainable cities and communities, responsible consumption and production, and climate action, that are heavily dependent on two enormous challenges: 1) Drastically and rapidly reducing our atmospheric emissions of greenhouse gasses 2) Replacing crude oil as a non-renewable source, while maintaining a sustainable development in the longer term \n Addressing the first challenge requires, among other actions, the development of more efficient renewable energy platforms, as well as the electrification of the transportation sector, powered by an energy grid with zero CO 2 emissions. The second challenge calls for the development of technologies that shift our current material sources to more renewable raw materials, ideally from waste resources, while reducing many countries’ overdependency on fossil fuel imports or extraction. To this end, biorefineries emerge as a critical solution to co-produce materials, sustainable fuels, and platform chemicals from a great diversity of non-eatable biomass feedstocks. There are four main features that are commonly used as criteria for the classification of biorefinery systems: feedstock, processes, platforms, and products. In the following, we will briefly discuss the feedstock and process and further expand the discussion on biorefinery platforms and products through the subsequent sections. Concerning the feedstock, the definition of biomass depends on the source, and it is open to discussion about what to include or not to include under the consideration of biomass. In a very general way, it is any organic matter, derived from living, or recently living animals or plants such as crops, as well as the waste derived from them, or other residues from municipal wastes, wastewater treatment, and feedstock. Common feedstocks include grasses, starch crops from wheat and maize, sugar crops from beet and cane, lignocellulosic materials from wood, crops or residues, oil crops, algae and seaweeds, and many other organic wastes from industrial, agriculture, and livestock activities, as well as commercial and post-consumer waste ( Jong et al., 2012 ). In biorefineries, this feedstock can be obtained from 1) A dedicated production to obtain sugar and starch, lignocellulosic materials, oils, and marine biomass 2) Residual biomass formed by organic residues from urban waste, sludges, manure, industrial lignin residues, and oil-based residues \n The main pre-processing and processing technologies used in biorefinery are as follows: 1) Mechanical processes, e.g., pressing, fractionation, and size reduction 2) Chemical processes, e.g., acid hydrolysis, oxidations, and esterification 3) Thermochemical processes, e.g., hydrothermal processing, pyrolysis, and gasification 4) Biochemical processes, e.g., fermentation \n The mechanical and chemical processes are usually used as pre-processing techniques to breakdown biomass before getting into a thermochemical or biochemical process. These methods can be combined in different ways depending on the feedstock, the platform, and the intended application, for example, biofuels, chemicals, or a combination of both ( Cherubini et al., 2009 ). In the following sections, this review summarizes the origins of the biorefinery concept, the biorefinery platforms and technologies, some biorefinery products, and some of the current challenges and future directions in biorefinery development."
} | 2,516 |
36132850 | PMC9416850 | pmc | 2,108 | {
"abstract": "Understanding biobased nanocomposites is critical in fabricating high performing sustainable materials. In this study, fundamental nanoparticle assembly structures at the nanoscale are examined and correlated with the macroscale properties of coatings formulated with these structures. Nanoparticle assembly mechanisms within biobased polymer matrices were probed using in situ liquid-phase atomic force microscopy (AFM) and computational simulation. Furthermore, coatings formulated using these nanoparticle assemblies with biobased polymers were evaluated with regard to the hydrophobicity and adhesion after water immersion. Two biobased glycopolymers, hydroxyethyl cellulose (HEC) and hydroxyethyl starch (HES), were investigated. Their repeating units share the same chemical composition and only differ in monomer conformations (α- and β-anomeric glycosides). Unique fractal structures of silica nanoparticle assemblies were observed with HEC, while compact clusters were observed with HES. Simulation and AFM measurement suggest that strong attraction between silica surfaces in the HEC matrix induces diffusion-limited-aggregation, leading to large-scale, fractal assembly structures. By contrast, weak attraction in HES only produces reaction-limited-aggregation and small compact cluster structures. With high particle loading, the fractal structures in HEC formed a network, which enabled a waterborne formulation of superhydrophobic coating after silane treatment. The silica nanoparticle assembly in HEC was demonstrated to significantly improve adhesion, which showed minimum adhesion loss even after extended water immersion. The superior performance was only observed with HEC, not HES. The results bridge the assembly structures at the nanoscale, influenced by molecular conformation of biobased polymers, to the coating performance at the macroscopic level. Through this study we unveil new opportunities in economical and sustainable development of high-performance biobased materials.",
"conclusion": "Conclusion In this study we present a waterborne biobased nanocomposite coating system composed of HEC and silica nanoparticles. It was demonstrated that the polymer morphology has a dramatic effect on the assembly of nanoparticles within the polymer matrix. Silica nanoparticles form small closed packed clusters dispersed uniformly in HES. By contrast, silica nanoparticles in HEC form fractal structures with long range connectivity. Liquid phase AFM revealed that assembly in HEC is driven by a large attraction force at a longer onset distance. Using AFM derived parameters for interparticle potential, computational simulations replicated the experimentally observed structures in HEC and HES systems. Furthermore, the assembled structures demonstrated great influence over the resulting water repellency and robustness of the coating. Upon silanization of the dried coating film, the HEC coating demonstrated superhydrophobicity (160° contact angle), with comparable performance post immersion in water for 24 hours. The observation is due to a multiscale roughness, in combination of the fractal nanoparticle assembly structures. Crosshatch testing further confirmed the robust adhesion performance of HEC nanocomposite coating. On the contrary, HES coatings performed poorly in adhesion, water repellency, and immersion robustness tests. Through this study we demonstrate that biobased polymer morphology can be used to influence nanoparticle assembly structures. High performing superhydrophobic coatings can be fabricated with water dispersible biobased polymers.",
"introduction": "Introduction Biobased materials aim to provide sustainable alternatives to traditionally petroleum derived materials, such as coatings, adhesives, and construction chemicals. 1–4 Beyond the raw material cost, limited performance inhibits the replacement of petroleum-based polymers by their biobased counterparts. Taking coating materials as an example, the properties of biobased coatings must be improved, particularly their mechanical strength and water resistance. 5–7 One approach to build new functionality and improve performance is to fabricate biobased nanocomposites. 8–11 Nanoparticles have already been widely used to enhance the performance of the polymer matrix. Extensive research has been reported with petroleum based polymer matrices, though far fewer studies have been carried out with biobased polymer matrices. 12,13 More importantly, the understanding of molecular structures and interactions within biobased polymers is very limited, as these polymers often possess complicated architectures and vary from different parent sources. 14–18 The fundamental study and systematic comparison of nanoparticle assembly structures within different biopolymers are needed to provide insight on the interactions and assembly mechanism. Nanoparticles in polymer matrices can assemble into a number of structures ranging from clusters to networks and films. 19–23 Their assembly is dictated by intermolecular forces, such as hydrogen bonding, electrostatic and van der Waals forces. 24–26 The aforementioned forces vary in magnitude and depend on interparticle distance. In addition, polymer morphology and entropy (depletion force) may play important roles in modulating interparticle interactions. 27–32 To probe interparticle interactions and assembly mechanisms within biobased polymers, we adopt integrated advanced tools, including liquid-phase atomic force microscopy (AFM) and computational simulation. In in situ AFM, a fluid environment can be used to replicate the wet condition, where adhesion is measured with a microscale colloidal probe mounted on the scanning tip. 33 With a substrate opposite the probe, we can directly measure the adhesion and deflection forces between the two surfaces in polymer solutions as the probe approaches and retreats from the substrate. To further verify that the observed nanoparticle assembly structures indeed originate from interparticle forces, computational simulation is utilized to corroborate the force measurement with the assembly structures. So far, force measurement, theory, and simulation models on biobased polymers and their nanocomposites are scarce. 34 In this study, drastically different assembly structures of silica nanoparticles were observed in two biobased polymers, hydroxyethyl starch (HES) and cellulose (HEC). These polymers were correspondingly derived from starch and cellulose by randomly modifying the side hydroxy groups with hydroxyethyl groups. The minor modification (1.3–5 molar substitution) destroyed the crystallinity and renders these polymers completely water soluble. 35 HEC and HES share very similar monomer molecular structures that differ only in the orientation of glycosidic hydroxyethyl pendant groups. Intriguingly, the differences in chain conformation between HES and HEC mediate the formation of drastically different assembly structures of silica nanoparticles. In the coiled, cis conformation of HES matrix, silica nanoparticles formed close packed clusters. While in the extended, trans conformation of HEC, the same nanoparticles formed loose fractal structures. Liquid-phase AFM detected the obvious difference in interactions between silica surfaces within HEC and HES aqueous solutions. Based on the input from the force measurement, computational simulation successfully reproduced structures resembling the experimental observation. Utilizing the knowledge obtained from fundamental studies of HEC and HES polymers, we further applied the resulting nanoparticle assembly structures to achieve a waterborne formulation and create superhydrophobic coatings. A major challenge in developing a biobased waterborne coating system is to improve the water repellency, especially with water dispersible polymers. Approaches have been developed to impart hydrophobicity to polymers via surface treatments and controlling surface roughness. 36 Surface treatments often refer to silane vapor deposition or liquid cast. 37–41 Surface roughness can be fabricated via colloidal lithography, 42 hot press, 43–45 mould, 46–48 and selective evaporation. 49 In addition, switchable hydrophilic and hydrophobic patterns have been designed for lotus-like self-cleaning capabilities. 50–53 However, most of these methods require complicated fabrication and surface treatment, which often involve chemical reactions in organic solvent. 54–59 It is much more challenging to create a durable hydrophobic surface using simple and straightforward waterborne coating formulations. Taking advantage of the unique nanoparticle assemblies discovered in this study, we have created a waterborne formulation that can provide both superhydrophobicity and strong adhesion even after water immersion. In doing so, we offer solutions to address ongoing challenges in biobased polymer applications, including water resistance and adhesion performance. The results will inspire new opportunities for applying biobased materials to many more areas of study, including inks, additive manufacturing, and biomedical devices.",
"discussion": "Results and discussion In our nanoparticle assembly study, there are only two major components: the biobased polymer matrix and nanoparticles. This simple combination offers a model system in understanding how polymers drive the formation of complex assembly structures. For the matrix, hydroxyethyl modified cellulose (HEC) and starch (HES) polymers were used. To best capture the interactions in the silica nanoparticle suspension, in situ liquid phase AFM was performed. The colloidal probe was equipped with a glass bead, and the polymer solution (of HEC or HES) was added to a Petri dish with the glass bottom. With this setup, we can reproduce the silica–silica surface interactions in presence of polymer. The contrast in magnitude and range of forces between silica surfaces in HEC and HES is clearly demonstrated by the AFM measurement. The results of the experiment reveal a four-fold difference in attraction force between silica surfaces within HEC (1.0 × 10 −10 N) and HES (2.5 × 10 −11 N) polymer matrices ( Fig. 1a ). In addition, the distance range at which the attractive force persists is two times larger in HEC (∼100 nm) than in HES (∼50 nm). This means that there is a stronger force dictating silica interactions when HEC is used as the matrix. The discovery is intriguing as these polymers are almost identical in chemical composition but only differ in orientation of the anomeric bonds ( Fig. 1a inset). Based on the force measurement, the potentials between two silica particles were derived ( Fig. 1b ). Fig. 1 (a) In situ AFM colloidal force measurement between silica surfaces in HEC and HES solution, inset shows the molecular structure of HEC and HES; (b) potential between silica nanoparticles in HEC and HES solution derived from AFM results, inset shows the difference in chain conformation between HEC and HES; (c) fluorescent microscopy of silica nanoparticles in HEC solution; (d) fluorescent microscopy of silica nanoparticles in HES solution; (e) confocal 3D image of silica nanoparticle assembly in HEC; (f) simulation result of nanoparticle assembly in HEC; (g) simulation result of nanoparticle assembly in HES; (h) image of 3D view nanoparticle assembly in HEC by simulation. The primary conformation of HEC and HES is likely consistent with their cellulose and starch parent structures. The hydroxyethyl pendant groups limit movement and determine the conformation of the primary chain. The cellulose derivative has trans linkages (α-anomer), which separate the bulky pendant groups and allow the polymer to take on an extended conformation. On the other hand, hydroxyethyl starch has cis linkages (β-anomer), which place the hydroxyethyl groups in adjacent units close to each other, pushing the polymer to a coil conformation ( Fig. 1a inset). Therefore, the persistence lengths of starch and cellulose vary due to the orientation of the pendant alcohol groups. For starch, the persistence length is 6 nm due to the cis conformation. 61 By contrast, in cellulose structures, the trans conformation leads to an extended persistence length of 40 nm. 62 The extended conformation of HEC may help induce a stronger and longer attractive force between silica surfaces than HES, due to possible bridging and hydrogen bonding. Interestingly, the orientation of the pendant groups on the polymer backbone and chain conformation also have dramatic impacts on assembly structures of nanoparticles within the matrix. To track the assembly, the formulations were prepared in a closed system to mitigate any influence of evaporation. A dilute (0.16 wt%) suspension of fluorescently labelled silica nanoparticles (100 nm), initially dispersed, was allowed to self-assemble and observed in situ under a fluorescent microscope. The formation of a loose fractal structure was clearly observed in the HEC suspension within 12 hours ( Fig. 1c ). However, nanoparticles form small compact clusters in the HES suspension ( Fig. 1d ). Confocal laser scanning microscope further revealed the three-dimensional assembly structures ( Fig. 1e ). The comparison of nanoparticle assemblies at longer time point (24 hours) in Fig. S1 † consistently demonstrates the same trend as observed in 12 hours. In the HEC suspension, nanoparticles form larger fractal network clusters. Alternatively, HES remains dispersed as compact clusters throughout the 24 hour time lapse, only slightly settling to the substrate as time progresses. The difference in nanoparticle assembly structures in the aqueous suspension can also be viewed clearly even after solution is dried into a coating film. With dilute (0.16 wt%) concentrations of silica nanoparticles, we observe the formation of a fractal network structure in presence of HEC (4 wt%) in the dried film (Fig. S2 † ). By contrast, nanoparticles are dispersed as clusters in HES (4 wt%) matrices. To further study how differences in polymer–particle interactions within HEC and HES suspension could translate into the difference in nanoparticle assembly structures, computational modeling and simulation were carried out. Simulation parameters were established based on the AFM force measurement ( Fig. 1b and S3 † ). It is important to note that nanoparticles assemblies in HEC suspension sediment faster to the bottom of the container than individual nanoparticles due to differences in effective gravity. Through fluorescent intensity measurement, it is estimated that the concentration of nanoparticles in Fig. 1c is ∼0.8%, while concentration of nanoparticles in Fig. 1d is ∼0.16%. Therefore, these different concentration values were used in simulation for HEC and HES, correspondingly, to capture the experimental conditions more accurately. The simulation successfully replicated the experimentally observed structures in both cases of HEC ( Fig. 1f and h ) and HES ( Fig. 1g ). The results strongly support that the interparticle force in the presence of biobased polymer is the driving factor in the formation of network and dispersed structures with HEC and HES, respectively. Simulation confirmed that the strong attraction amongst silica nanoparticles in HEC solution may lead to diffusion limited aggregation (loose fractal structures), while the weak interactions among silica nanoparticles in HES solution result in reaction limited aggregation (close packed clusters). The qualitative differences observed in the simulated HEC and HES systems were complimented with a quantitative measure of cluster size at the 24 hour time point. The radius of gyration of individual clusters was calculated in the 3D simulations. The results show the average radii of gyration to be 23 μm in HEC and 0.7 μm in HES, which in general agrees with experimental measurement of average cluster radii: 32 μm in HEC and 1.5 μm in HES. Due to limitations in image resolution, the precise measurement of radii is expected to be very challenging. The significant variation in cluster size amongst HEC and HES matrices highlights key differences in assembly, emphasizing qualitative similarity to the simulation result. The drastic difference between the nanoparticle assembly structures inside HEC and HES suspensions, together with AFM and simulation results, suggest that the molecular architecture of the polymer backbone has profound effects upon the particle interactions and self-assembly. As discussed earlier, the morphology of HEC and HES in aqueous suspension can be drastically different, due to the conformational difference in how repeating units are connected. The extended conformation of HEC may induce stronger and longer-range attractions among silica surfaces than HES through hydrogen bonding and bridging effects. Another possible mechanism for attraction is the depletion force, which depends on many different factors including the polymer dimension and nanoparticle size. 39 It is known that larger polymer dimensions may induce stronger and longer-range depletion attractions amongst nanoparticles. 40 However, it is not entirely clear how polymer morphology would impact the depletion attractions. At this stage, the detailed mechanism is not yet clear how differences at the molecular level for HEC and HES lead to differences in the interactions medicated by these polymers, which warrants further studies in the future. The unique structures formed by nanoparticles within HEC were utilized to create superhydrophobic surfaces. When nanoparticle loading is increased from 0.16 to 3 wt% in the formulation and a vaporized fluorinated silane treatment is applied after coating film is dried, a contact angle of 160° is observed ( Fig. 2a ). Silane treated HEC becomes hydrophobic due to surface bound water, which allows for silane to polymerize prior to its attachment to the coating surface. The observed clusters form due to a mismatch in energy between the polymer and unreacted water particles. The silane deposits in clusters to reduce the surface tension consideration. 60 Water repellency of formulation with HES under the same treatment condition is greatly reduced in comparison to HEC, with an angle of 127°. Without nanoparticles, the contact angle is reduced by 45° and 60° in HEC and HES, respectively. This suggests that nanoparticle assembly structures are essential in providing superhydrophobicity to the coatings. Fig. 2 (a) Contact angle measurement of nanocomposite coating films made from HEC and HES, both treated and untreated; (b) SEM micrographs and AFM topography profiles of polymer alone samples; (c) SEM micrographs and confocal profiles of nanocomposite samples. Further characterization of the treated surface via SEM revealed that the silane vapor deposits as small clusters on the polymer coatings, in absence of nanoparticles. AFM topography measurements of the same surfaces reveal drastically different roughness profiles amongst HEC and HES treated coating surfaces ( Fig. 2b ). The HEC coatings presents much larger silane clusters than HES, likely contributing to its enhanced water repellency capability. With the introduction of the silica nanoparticles, SEM reveals multiscale roughness with fractal features in HEC. In HES, the nanoparticles are relatively dispersed, as they were in the dilute case. By contrast, in HEC the nanoparticles form heterogeneous assembly structures. The surface treated nanocomposites were then characterized using confocal microscopy as the surfaces are too rough for a successful AFM measurement. Confocal measurements corroborate the SEM data, exhibiting more than a three-fold enhancement in surface roughness with HEC compared to HES ( Fig. 2c ). The results are consistent with prior observations that surface roughness will enhance the hydrophobicity. 63–66 Further experiments demonstrate the importance of particle size, loading, and silane treatment. The results of these studies reveal that high loading of small particles promotes the best water repellency (Fig. S4 † ). Silica particle size does not drastically affect the contact angle, since surface roughness is mainly induced by particle assemblies, not individual particles. Non-fluorinated silane treatment, trichloro(octyl)silane, was also assessed in scope of this study (Fig. S5 † ). It was found that though this treatment can provide hydrophobicity, the fluorinated derivative is more effective. In all of the aforementioned cases, untreated samples showed poor water repellency, with and without nanoparticle filler. The enhanced water repellency exhibited in the treated nanocomposites prompted studies of their robustness under extended wet conditions. In this batch of experiment, the coatings were applied to a flexible PET substrate. To test the water repellency, treated and untreated samples of HEC and HES with silica nanofiller were immersed in deionized water for 24 hours. The contact angle was measured post-immersion to reevaluate the coating performance. Obviously, the silane treatment was essential in providing heightened hydrophobicity to HEC and HES, but only HEC demonstrated comparable performance post-immersion. The contact angle of HES is diminished by 40° after immersion, whereas HEC only loses 15° ( Fig. 3 ). With the trichloro(octyl)silane treatment, similar results are obtained pre- and post-immersion, with HEC contact angle reduced by 12°, and HES by 30° (Fig. S6 † ). Fig. 3 Contact angle measurements and corresponding confocal optical images: fluorinated silane treated and untreated samples with 3% 100 nm silica pre and post immersion in water for 24 hours. Scale bar is 200 μm. Confocal characterization of the surface reveals that adhesion is poor when the HEC and HES coatings are not treated with silane vapor, and the coating is washed away upon immersion. Interestingly, a similar phenomenon occurs even in treated HES, providing some explanation for its reduced contact angle post-immersion. HEC treated with fluorinated silane, on the other hand, retains its adhesion to the substrate after immersion for 24 hours ( Fig. 3 ). Treatment with trichloro(octyl)silane shows the same trend, with only treated HEC maintaining adhesion post-immersion (Fig. S6 † ). Therefore, the fractal network structures assembled by silica nanoparticle in HEC not only provide the roughness required for superhydrophobic properties, but also offer the enhancement in adhesion. The structures are essential in retaining the adhesion performance even after extended immersion in water. This is not possible in the HES system, demonstrating the superior performance of HEC–silica coatings for water repellency applications. An adhesive crosshatch test was further administered to further compare the adhesion performance of nanocomposite coatings. In order to carry out the test, the PET substrate was replaced with aluminum Q-panels. The crosshatch test is conducted by applying a crosshatch knife at a 90° angle. Subsequently, a piece of tape is quickly applied and removed from the crosshatched area. According to ASTM 3359, the performance of the coatings are assigned to one of the following rating based on their adhesion ( Table 1 ): 67 Crosshatch adhesion test performance metrics ASTM rating Surface identifiers 5A No peeling or removal 4A Trace peeling or removal along incisions or at their intersection 3A Jagged removal along incisions up to 1.6 mm (1/16 in.) on either side 2A Jagged removal along most of incisions up to 3.2 mm (1/8 in.) on either side 1A Removal from most of the area of the X under the tape 0A Removal beyond the area of the X Without silane treatment (pre-immersion), HEC samples exhibit strong adhesion, with no removal of the coating from the substrate. Post-immersion, the coating was washed away. Interestingly, significant coating removal is observed in HES even prior to water immersion. Like HEC, post-immersion the coating is washed away, leaving no surface to be removed with the adhesion crosshatch method. After silane treatment, adhesion performance of both HEC and HES samples were improved. The HEC coating film demonstrates no coating removal prior to immersion, and very little afterwards. HES, on the other hand, exhibits jagged removal along the crosshatch before and after immersion, though generally the adhesion is improved compared to the untreated HES sample ( Fig. 4 ). Again, the results of the adhesion test (along with contact angle and immersion) highlight the importance of nanoparticle assembly structures and polymer–nanoparticle interactions in determining macroscopic level material properties. The network structure formed with HEC polymers, and the roughness that is induced at high particle loading, have created a strongly adhered coating with superhydrophobic properties. The waterborne HEC coating demonstrates robustness to mechanical damage even under extended water immersion. The chemically identical HES polymer, was not capable of providing adhesion or superhydrophobic properties to the coating film. These results highlight the critical role of molecular conformation in determining the performance of products derived from biobased polymers. Fig. 4 Adhesive crosshatch test and corresponding optical images of fluorinated silane treated samples with 3% 100 nm silica pre and post immersion in water for 24 hours. Ratings are assigned according to ASTM 3359. Scale bar is 10 mm."
} | 6,362 |
25215180 | PMC4147103 | pmc | 2,109 | {
"abstract": "Background L-glutamic acid is one of the major amino acids that is present in a wide variety of foods. It is mainly used as a food additive and flavor enhancer in the form of sodium salt. Corynebacterium glutamicum ( C. glutamicum ) is one of the major organisms widely used for glutamic acid production. Methods The study was dealing with immobilization of C. glutamicum and mixed culture of C. glutamicum and Pseudomonas reptilivora ( P. reptilivora ) for L-glutamic acid production using submerged fermentation. 2, 3 and 5% sodium alginate concentrations were used for production and reusability of immobilized cells for 5 more trials. Results The results revealed that 2% sodium alginate concentration produced the highest yield (13.026±0.247 g/l by C. glutamicum and 16.026±0.475 g/l by mixed immobilized culture). Moreover, reusability of immobilized cells was evaluated in 2% concentration with 5 more trials. However, when the number of cycles increased, the production of L-glutamic acid decreased. Conclusion Production of glutamic acid using optimized medium minimizes the time needed for designing the medium composition. It also minimizes external contamination. Glutamic acid production gradually decreased due to multiple uses of beads and consequently it reduces the shelf life.",
"conclusion": "Conclusion Immobilization of C. glutamicum and mixed culture of C. glutamicum and P. repti-livora was used for glutamic acid production. First, RSM was used to manipulate the medium components for enhanced production. Hence, it was easy to standardize the alginate concentration for immobilization. Two percent alginate concentration was fixed and reusability study was carried out to analyze the stability of beads for production of glutamic acid. This study demonstrated the procedures for economic production of glutamic acid. Cell entrapment in a polymer matrix such as sodium alginate has been widely used for commercial production of various products. The simple method adopted for entrapment of cells reduced the costs of production. In fact, it is very simple to collect and estimate the end product in the medium. It minimizes the external contamination and subsequently other effects in the fermentation process are minimized as well.",
"introduction": "Introduction L-amino acids are major biological components commercially used as additives in food, feed supplements, infusion compounds, thera-peutic agents and precursors for peptides synthesis or agriculture based chemicals. The amino acids are the second most important category, after antibiotics, with fermentation products exhibiting the highest growth rates ( 1 ) . L-glutamic acid was the first amino acid produced commercially. The substance was discovered and identified in the year 1866 by the German chemist Karl Heinrich Leopold Ritt-hausen. L-glutamic acid was mainly produced by microbial fermentations and the chemical mode of synthesis is not widely preferred due to the formation of racemic mixture ( 2 ) . In biotechnological processes, Corynebacterium species are used for economic production of glutamic acid by submerged fermentation ( 3 ) . L-glutamic acid is produced per year using coryneform bacteria. A number of fermentation techniques have been used for the production of glutamic acid ( 4 – 6 ) . Glucose is one of the major carbon sources for production of glutamic acid. Glutamic acid was produced with various kinds of raw materials using sub-merged fermentation of palm waste hydrolysate ( 7 ) , cassava starch ( 8 ) , sugar cane bagasse ( 6 ) , date waste ( 9 ) . Immobilization of microbial cells in biological processes can occur either as a natural phenomenon or through artificial process. The method used for immobilization of cells was adsorption, cross linking, covalent bonding and encapsulation. These are all common methods employed for enzymes and microbial cells and usage of the methods depends on the cultures and conditions ( 10 ) . Artificial immobilization of cells results in restricted growth and facilitates the production process. In biotechnology, it has been recognized that immobilization and co-immobilization of cells/ enzymes facilitates the feasibility of two or multi-step conversions into a single-step conversion. Binding of the deficient enzyme from an external source to free or immobilized microorganisms or immobilization of mixed culture capable of carrying out two or multistep conversions into a single-step conversion, leads to co-immobilized cells. The co-immobilized cells can open up new possibilities of synergistic action and result in more yield/ conversion, which cannot be obtained to the same extent by separately immobilized cells ( 11 , 12 ) . Hence, the present report focused on immobilization of whole cells of C. glutamicum and mixed culture of C. glutamicum and P. reptilivora for the production of glutamic acid with an optimized medium and reusability of immobilized cells for the production of glutamic acid.",
"discussion": "Discussion The preliminary reports of the present investigation revealed that the basic medium used for production of L-glutamic acid was lower than the optimized medium used for production of L-glutamic acid. In immobilization studies, sodium alginate concentration had an influence on density of the beads; higher alginate concentration showed lower conversion efficiency which might be due to reduced pore size of the beads. The lower sodium alginate concentration affects the leakage of biomass from the beads which could be due to increased pore size of the beads. In other studies, it has been reported that natural isolates of C. glutamicum was used for glutamic acid production with free whole cells and with immobilization. Comparatively, whole cells produced more glutamic acid than immobilized cells. Moreover, regarding glutamic acid production among immobilized cells, agarose produced more glutamic acid as compared to alginate ( 16 ) . Another report emphasized that fed-batch and continuous fermentation process adopted for L-glutamic acid production with the cells of C. glutamicum entrapped in carrageenan gel. Higher yield was produced in batch fermentation rather than continuous fermentation process and repeated uses of immobilized cells resulted in lower glutamic acid production. Production was enhanced when the medium was supplemented with penicillin ( 6 ) . Sodium alginate concentration is also one of the factors influencing the productivity of immobilized cells. The reduction in productivity may be due to the increase in porosity which makes the leakage. Earlier investigations demonstrated that 3% alginate concentration enhances the productivity in co-immobilized culture of Brevibacterium roseum and E. coli among different concentrations of alginate ( 17 , 18 ) . The production of glutamic acid influenced immobilized cells due to ionic strength and stability in storage of beads ( 19 , 20 ) . There are some more studies focused on pH, temperature, agitation and other physical parameters used in glutamic acid production with immobilization ( 21 – 25 ) . This investigation analyzed reusability of immobilized cells for storage and usage in fermentation process. Furthermore, intensive studies are required for evaluating the methods in increasing glutamic acid production for immobilization in industrial fermentation. Immobilization is highly sensitive to pH, temperature and other factors such as ionic potency in long incubation periods for nonspecific adsorption. The surface adherent cells can be removed due to these factors. Activation of surfaces with cross-linkers such as glutaraldehyde could lead to covalent attachment of the cells through surface amine groups. Loss of cell activity and viability in immobilization may be due to the formation of bonds with metal activated supports. Immobilization can be achieved by entrapping the cells within the matrix formed by gels made from alginates, carrageenans, and polyacrylamide materials."
} | 1,994 |
34686763 | PMC8857181 | pmc | 2,110 | {
"abstract": "Plant root-associated bacteria can confer protection against pathogen infection. By contrast, the beneficial effects of root endophytic fungi and their synergistic interactions with bacteria remain poorly defined. We demonstrate that the combined action of a fungal root endophyte from a widespread taxon with core bacterial microbiota members provides synergistic protection against an aggressive soil-borne pathogen in Arabidopsis thaliana and barley. We additionally reveal early inter-kingdom growth promotion benefits which are host and microbiota composition dependent. Using RNA-sequencing, we show that these beneficial activities are not associated with extensive host transcriptional reprogramming but rather with the modulation of expression of microbial effectors and carbohydrate-active enzymes.",
"introduction": "Introduction Plant pathogenic fungi limit crop productivity globally. These threats are expected to increase with global warming [ 1 ]. Decades of advances in agrochemicals and plant breeding have expanded farmers’ toolkits with fungicides and resistant varieties to limit the detrimental effects of these organisms on crop yield. Yet, current tools are becoming environmentally unsustainable or ineffective against rapidly evolving pathogens [ 1 ]. A key example of this scenario is represented by the soil-borne plant pathogen Bipolaris sorokiniana (syn. Cochliobolus sativus , hereafter Bs ), the causal agent of spot blotch and common root rot diseases that threaten cereal production in warm regions [ 1 – 3 ]. Root rot normally originates from inoculum carried on the seed or from soil-borne conidia, but the fungus can infect plants at any developmental stage. However, as the importance of root-inhabiting pathogenic fungi has often been underestimated, very little is known about the molecular mechanism behind the detrimental interaction of Bs with roots [ 4 ]. Microbial communities living at the root−soil interface, collectively referred to as the plant root microbiota, have gained centre-stage in pathogen protection [ 5 ]. Past studies across a variety of plant species employed environmental sampling or controlled conditions in the field and laboratory to characterize the root microbiota [ 6 – 10 ], with an overall greater focus on bacteria than on filamentous fungi [ 11 ]. Microbial diversity and abundance gradually decrease between the soil and vicinity of the root (rhizosphere), and further between the rhizosphere and root internal compartments (endosphere). Moreover, a number of bacterial taxa (e.g., Proteobacteria, Actinobacteria, Bacteroidetes, and Firmicutes) consistently occur in the root endosphere of different examined plant species [ 10 ]. This latter feature underpins the “bacterial core microbiota” concept, in which strains from specific taxa are commonly selected as endophytes across plant species, soil types, and environmental conditions [ 12 ]. By contrast, studies of geographically distinct populations of Arabis alpina and Arabidopsis thaliana (hereafter Arabidopsis) showed that few fungal taxa are prevalent in the root endosphere, and that endophytic fungal communities are strongly influenced by location and climate [ 9 , 13 ]. The functions and benefits of root microbiota members in the context of abiotic or biotic stresses have been extensively investigated under laboratory conditions using single microbial strains and, more recently, synthetic bacterial communities (SynComs) [ 14 ]. Several bacterial and fungal isolates have the capacity to directly increase plant biomass via growth hormone production and/or by providing plants with limiting macro- or micro-nutrients [ 13 , 15 – 19 ]. Although diseases caused by pathogens have been shown to be directly or indirectly reduced by the addition of single or multiple beneficial microbes [ 4 , 8 , 20 – 23 ], how fungal root microbiota members with beneficial functions influence and are influenced by bacterial colonization remains less understood. Sebacinales fungi (Basidiomycetes) are a remarkable group of plant mutualists with worldwide occurrence in soils and as endophytes. While individual Sebacinales strains can interact with roots in the absence of differentiated structures, they can also form specialized interactions with distinctive morphological characteristics on relevant hosts, as in orchid- or ectomycorrhiza symbioses [ 24 ]. Root colonization by these fungi improved host growth and development, increased grain yield, and enhanced root phosphate uptake in several plant species [ 25 – 28 ]. The positive effects of Sebacinales on the host plant extend well beyond growth and development and cannot be explained by enhanced host nutrition alone [ 24 , 26 , 29 ]. Recently, it was shown that effector molecules derived from the Sebacinales root endophyte Serendipita indica contribute to the establishment of the fungus−host interaction [ 30 – 33 ]. S. indica effectors suppress plant defence responses and modulate plant metabolism to promote compatibility in the roots, but their contribution to beneficial outcomes is unclear. Similarly, the nature of host transcriptional programs and signalling networks that lead to a mutually beneficial fungus−plant partnership are not well understood. In the past few years, microbe−microbe interactions have emerged as an additional important element shaping plant host−microbe interactions [ 4 , 22 , 34 , 35 ]. Using a soil-based split-root system, we demonstrated that both local and systemic colonization by the Sebacinales endophyte Serendipita vermifera (syn. Sebacina vermifera , hereafter Sv ) afford protection against Bs infection and disease symptoms in Hordeum vulgare (barley) [ 4 ]. Here, we explore how Sv and Bs colonization capacities in two plant species, barley, and Arabidopsis, are modulated by the presence of individual members of the core bacterial microbiota or SynComs isolated from the barley rhizosphere [ 36 ] or Arabidopsis roots [ 37 ]. The finding that Bs also infects and causes disease symptoms in Arabidopsis roots motivated us to develop a set of physiological measurements to characterize disease severity and plant growth in Arabidopsis under different microbe treatment regimes. These measurements include ion leakage (quantified via electric conductivity) and photosynthetic activity (measured using pulse amplitude modulation fluorometry) as readouts for host cell death progression and biotic stress during the host−microbe interaction. Analyses of inter-kingdom activities in barley and Arabidopsis revealed that Sv can functionally replace the core bacterial component of the rhizosphere by mitigating pathogen infection and disease symptoms in both hosts. Additionally, we show that cooperation between bacteria and beneficial fungi leads to inter-kingdom synergistic beneficial effects, thereby providing insights into the complex relationships of the rhizosphere. Finally, RNA-seq experiments with selected bacterial strains alone or combined with Sv and/or Bs give insights to how microbes synergistically protect plants. We conclude that plants have evolved to preferentially accommodate communities that support their health and that root-associated prokaryotic and eukaryotic microbes can act synergistically with the plant host in limiting fungal disease.",
"discussion": "Discussion In complex environments, plant−microbe interactions are not only shaped by the plant immune system [ 20 , 74 , 75 ] but also by microbe−microbe competition and co-operation, acting directly on or as an extension to plant immunity [ 76 , 77 ]. Recent studies reveal the importance of root-associated bacteria for plant survival and protection against fungi and oomycetes [ 8 , 78 – 81 ]. Much less attention has been paid to the role of widely distributed beneficial endophytic fungi in a multi-kingdom context. Here we show that the effects on host growth and protection that are conferred by the Sebacinales member S. vermifera in bipartite and tripartite interactions [ 4 , 82 ] are retained in a community context. The observed robust protective function and stability of Sv colonization is likely due to its ability to adapt to changes in the plant host environment [ 4 ]. The strength of its protection against an aggressive root fungal pathogen ( Bs ) is underscored by the observation that Sv can functionally replace core bacterial microbiota members in mitigating pathogen infection and disease symptoms in distantly related plant hosts. This finding is in accordance with Arabidopsis root microbiota samplings across European habitats which shows Sebacinales fungi to be of low abundance but consistently present in the host roots and the rhizosphere. Our data highlight the potential importance of widespread root fungal endophytes in maintaining plant host physiological fitness in nature, thereby emphasizing that low-abundance microbes can play a significant role in microbiota beneficial functions and should be considered when designing SynComs with multiple traits, such as resilience and protective activities. Strikingly, the presence of Sv additionally stabilizes and potentiates the protective activities of root-associated bacteria and mitigates the negative effects caused by the non-native Hv SynCom in Arabidopsis (Fig. 3B–D and Fig. S3 ), revealing a more general protective activity of root endophytic fungi. The induction of cell death by the barley-derived SynCom in Arabidopsis could be due to the presence of specific bacterial strains that are absent in the At SynCom. One such bacterial group that is well represented in the Hv SynCom but absent in the At SynCom used in this study is the Pseudomonadales. Several members of this group are reported to be pathogenic whereas others with very few genome differences promote plant growth and exert biocontrol activities against different fungal pathogens [ 83 ]. However, we did not observe an increase in ion leakage upon inoculation with the Pseudomonas strain bi08 or other members of the Hv SynCom when inoculated alone (Fig. S5B–E ). The pathogenicity of a single bacterial strain is likely to be suppressed in a community context, as observed for Bs (Figs. 2 , 3 ). Thus, another explanation to the negative effects of the Hv SynCom in Arabidopsis but not in barley might be a lack of adaptation to Arabidopsis. This notion is supported by a recent analysis that detected a clear signature of host preferences among commensal bacteria from diverse taxonomic groups, including Pseudomonadales in Arabidopsis and Lotus japonicus [ 84 ]. Our transcriptomic analyses show that the effects of the tested bacterial strains in tripartite associations differ substantially. The general decreased barley transcriptional response to the pathogen driven by the Rhizobiales strain Root172 (Fig. 5B ) and the lysis of the fungal matrix at the host rhizoplane (Fig. 4 ) suggest that this bacterial strain most likely acts directly on Bs . This is also supported by the strong antagonism of Bs growth in confrontation assays irrespective of the presence of a host plant (Fig. 2B and Fig. S4 ). Taken together, these results point to Root172 as a possible biocontrol agent against Bs and potentially other root-infecting pathogens. The impact of Root172 contrasted strikingly with that of the Bacillales strain Root11, which did not limit Bs growth but rather enhanced Bs pathogenicity in barley. Notably, combining these two bacterial strains with Sv led to a restriction of Bs that exceeded the protective benefits of Sv and the bacteria alone (Fig. 2B ). These synergistic beneficial effects are decoupled from extensive host transcriptional reprogramming (Fig. 5B ) and cannot be solely explained by enhanced Sv growth (Fig. 2D and Fig. S4B ) as speculated for other fungal-bacterial synergistic beneficial effects [ 85 , 86 ]. Our transcriptional and phenotypic data further suggest that Sv —bacterial synergism in protecting host roots have also a component that is additive because the underlying antagonistic mechanisms displayed by the fungal root endophyte and the bacterial strains are likely to be distinct and explained mainly by direct microbe−microbe interactions outside the plant. Nonetheless, we have observed a higher level of inter-kingdom mediated antagonism on Bs in presence of the host (Fig. 2B and Fig. S4 ). This suggests a minor but relevant host-dependent effect that needs to be addressed. At the early time point of 6 dpi, growth promotion was only observed in the combined presence of Sv and certain bacterial strains with the strongest effect during co-inoculation with Root11 in barley and Root172 in Arabidopsis (Fig. 2E and Fig. S2E ). Furthermore, growth promotion required living microbes, as co-inoculation with heat-inactivated bacteria did not increase the root fresh weight in barley. Commensal bacteria in the rhizosphere can trigger plant growth promotion and resistance to pathogen [ 20 , 21 , 87 ]. Among them, strains belonging to the genus Bacillus are often used as bioagents due to their function in eliciting ISR (induced systemic resistance) as well as growth promotion [ 21 , 88 ]. However, plant growth-promoting bacteria (PGB) and Sebacinales mediated growth promotion are often reported during later stages of colonization. The early host growth enhancement observed with Sv and the bacteria might thus confer a competitive advantage for plants in nature. It is striking that the growth-promoting effect is not accompanied by an extensive host transcriptional response with only 14 barley DEG being specific to this condition (Table S6 ). Interestingly, several of these genes display differential expression across barley accessions (analyzed using Genevestigator) compared to the cultivar Golden Promise. It would therefore be informative to test growth outcomes of combined Sv and e.g., Root11 inoculation in different barley varieties/ecotypes. The resulting synergistic inter-kingdom benefits in plant protection against fungal disease and in plant physiology are in line with studies of the Sebacinales fungus S. indica with single bacterial strains on tomato [ 85 , 89 , 90 ], rice [ 91 ], barley [ 92 ], and chickpea [ 93 ] and underline the broad functional relevance in plant health for fungi of the order Sebacinales in multi-kingdom environments. Inter-kingdom benefits in plant–beneficial microbe interactions were reported also for native isolates of Trichoderma spp. and P. fluorescens against Ralstonia spp. in tomato and with B. velezensis against Fusarium in gooseberry [ 23 , 94 ], suggesting that the combined application of beneficial fungi and bacteria has strong potential as biocontrol agents. The deployment of microbiota as biocontrol agents for crop protection and enhancement is an ancient concept that is gaining increased relevance in modern agriculture [ 95 – 97 ]. Plant protection and growth promotion properties conferred by microbial consortia have been found to be more resilient than the use of single strains [ 95 ]. Moreover, Duran et al. 2018 showed that a complex SynCom consisting of bacteria, fungi, and Oomycetes led to the strongest beneficial effects on Arabidopsis growth and survival compared to mono-kingdom or small SynCom associations and hypothesized that selective pressures over evolutionary time favour inter-kingdom microbe−microbe interactions over interactions with single microbial strains [ 8 ]. Inter-kingom associations are frequently observed between members of the Sebacinales and bacteria. Different Sebacinales species host endobacteria of the orders Bacillales (genera Paenibacillus ), Pseudomonadales ( Acinetobacter ) and Actinomycetales ( Rhodococcus ) and its close relative S. indica hosts an endobacteria of the order Rhizobiales ( Rhizobium radibacter ) [ 98 ]. Beneficial effects of these intimate inter-kingdom interactions on the plant host and the fungus itself were described between S. indica and R. radibacter [ 98 , 99 ] and for interactions between arbuscular mycorrhizal fungi and bacteria belonging to different species of the orders Proteobacteria ( Rhizobiales ) and Firmicutes ( Bacillales ) [ 100 ]. Considering the pervasiveness of beneficial effects conferred by Sebacinales and bacteria compared to the vulnerability of Bs in a multipartite context, our data support the hypothesis that the establishment of beneficial inter-kingdom interactions in the plant microbiota is an evolutionary conserved and robust trait."
} | 4,148 |
31387934 | PMC6687939 | pmc | 2,111 | {
"abstract": "Predicting the metabolic potential and ecophysiology of mixed microbial communities remains a major challenge, especially for slow-growing anaerobes that are difficult to isolate. Unraveling the in situ metabolic activities of uncultured species may enable a more descriptive framework to model substrate transformations by microbiomes, which has broad implications for advancing the fields of biotechnology, global biogeochemistry, and human health. Here, we investigated the in situ function of mixed microbiomes by combining stable-isotope probing with metagenomics to identify the genomes of active syntrophic populations converting butyrate, a C 4 fatty acid, into methane within anaerobic digesters. This approach thus moves beyond the mere presence of metabolic genes to resolve “who is doing what” by obtaining confirmatory assimilation of the labeled substrate into the DNA signature. Our findings provide a framework to further link the genomic identities of uncultured microbes with their ecological function within microbiomes driving many important biotechnological and global processes.",
"conclusion": "Conclusions. In this study, stable-isotope-informed genome-resolved metagenomics was used to provide genomic insights into syntrophic metabolism during butyrate degradation in anaerobic digesters. The results obtained via genome binning and metabolic reconstruction showed that a 13 C-enriched Syntrophomonas genome contained the genetic capacity to convert butyrate into precursor metabolites for methane formation: acetate, hydrogen, and formate. A 13 C-enriched Methanothrix genome likely consumed the acetate produced during butyrate degradation, incorporating some 13 C into biomass. The presence of a CO 2 -reducing pathway, as well as formate dehydrogenase and hydrogenase genes, in the Methanothrix genome leaves open the possibility of flexible metabolism during methanogenesis. As syntrophic fatty acid-degrading populations are often slow-growing and thus difficult to isolate, this study demonstrates a new approach to link ecophysiology with genomic identity in these important populations involved in anaerobic biotechnologies as well as global carbon cycling. Advancing our understanding of in situ metabolic activities within anaerobic communities is paramount, as these microbiomes contain multiple interacting functional groups that, in cooperation, enable the processing of degradable organic carbon into methane gas. Coupling SIP-informed metagenomics with other activity-based techniques, such as metabolomics, transcriptomics, and proteomics, may further illuminate the structure of anaerobic metabolic networks as well as quantify metabolite fluxes, thus enabling newly informed process models to predict rates of anaerobic carbon transformation.",
"introduction": "INTRODUCTION Linking microbial genomic identity with ecological function is considered a “Holy Grail” in microbial ecology ( 1 ) and has broad implications for improving our ability to manage microbial communities in engineered biotechnologies. Anaerobic digestion is an example of a biotechnology that enables resource recovery from organic waste by generating methane gas as a renewable biofuel and thus plays a role in establishing a circular economy ( 2 ). The production of methane in anaerobic digestion is executed through a series of trophic interactions constituting a metabolic network of hydrolyzing and fermenting bacteria, syntrophic acetogens, and methanogenic archaea ( 3 , 4 ). Metabolic reconstructions based on shotgun metagenomic sequencing data have highlighted potential partitioning of functional guilds within anaerobic digester microbiomes ( 4 ). Yet, our understanding of the ecophysiology of the microorganisms present in anaerobic digesters is limited by the high community complexity and lack of cultured representatives ( 4 ). Elucidating the nature of interspecies interactions between different trophic groups in the anaerobic digester metabolic network may help to better understand and optimize the conversion of organic wastes into renewable methane. The terminal steps in the anaerobic metabolic network, syntrophy and methanogenesis, are responsible for a considerable portion of carbon flux in methanogenic bioreactors, as fatty acids are often produced during fermentation of mixed organic substrates ( 5 ). The accumulation of fatty acids in anaerobic digesters is often responsible for a reduction in pH and process instability ( 3 ). In particular, syntrophic degradation of the 4-carbon fatty acid butyrate can be a bottleneck for anaerobic carbon conversion, as this metabolism occurs at the thermodynamic extreme. Butyrate degradation to acetate and hydrogen is thermodynamically unfavorable under standard conditions (Δ G ° = 53 kJ/mol) and yields only −21 kJ/mol under environmental conditions typical of anaerobic bioreactors (pH 7, 1 mM butyrate and acetate, 1 Pa H 2 ) (see equation S1 in Table S2 in the supplemental material). Thus, cooperation between syntrophic bacteria and acetate- and hydrogen-scavenging methanogenic partners is necessary to maintain thermodynamic favorability. Cultured representative species carrying out syntrophic fatty acid oxidation are potentially underrepresented due to their slow growth and difficulty of isolation in the lab ( 6 ). So far, only two mesophilic ( Syntrophomonas and Syntrophus ) and two thermophilic ( Syntrophothermus and Thermosyntropha ) genera (12 bacterial species in total) have been shown to oxidize butyrate in syntrophic cooperation with methanogenic archaea, and they all belong to the families Syntrophomonadaceae and Syntrophaceae ( 6 ). Despite their major roles in processing carbon within anaerobic bioreactors, many syntrophic fatty acid-oxidizing bacteria have evaded detection with quantitative hybridization-based techniques ( 7 ), which is likely due to their low biomass yields ( 8 ) or our incomplete knowledge of active syntrophic populations within anaerobic digesters ( 9 ). Broad metagenomic surveys of anaerobic digester communities have similarly observed poor resolution of syntrophic populations, owing to their low abundance ( 4 , 10 ). Thus, highly sensitive culture-independent approaches are needed to expand our understanding of the ecophysiology of syntrophic populations to better control and predict metabolic fluxes in anaerobic environments. Recently, we demonstrated the potential of combining DNA–stable-isotope probing (DNA-SIP) with genome-resolved metagenomics to identify syntrophic populations degrading the long-chain fatty acid oleate (C 18:1 ) within anaerobic digesters ( 11 ). Stable-isotope-informed metagenomic sequencing can enrich metagenomic libraries with genomic sequences of actively growing microbes that incorporate 13 C into their biomass from an added labeled substrate ( 12 ) and thus allows for a “zoomed-in” genomic view of low-abundance populations, such as syntrophs. We also demonstrated that this approach was amenable for recovering high-quality microbial genomes using a differential coverage-based binning approach, as genomes from active microbes have low abundance in heavy DNA from 12 C controls but are enriched in heavy DNA from 13 C-amended treatments ( 11 ). Here, we applied stable-isotope-informed metagenomics to resolve the genomic makeup of active syntrophic butyrate-degrading populations within anaerobic digesters treating manure and sodium oleate ( 13 ). These same anaerobic digesters were previously used for DNA-SIP with oleate ( 11 ) at a similar time point, thus allowing for genomic comparisons using a multisubstrate SIP data set. This approach identified potential metabolic flexibility in a syntrophic bacterium implicated in the degradation of multiple fatty acids within the study anaerobic digesters, and elucidated an in situ syntrophic partnership between the acetogenic bacterium and an acetoclastic methanogen via interspecies metabolite transfer during butyrate degradation.",
"discussion": "RESULTS AND DISCUSSION DNA-SIP of methanogenic microcosms with [ 13 C]butyrate. Two laboratory-scale anaerobic digesters fed dairy manure were either pulse fed every 48 h or fed semicontinuously with sodium oleate (C 18:1 ) for over 230 days ( 13 ). Quantitative PCR and 16S rRNA gene amplicon sequencing indicated that Syntrophomonas became enriched within the reactors from oleate feeding ( 13 ). DNA-SIP-informed metagenomics confirmed that a majority of oleate-degrading bacteria in the two digesters belonged to Syntrophomonas ( 11 ). Here, we investigated whether any of the populations implicated in oleate degradation were also involved in the degradation of the short-chain fatty acid butyrate (C 4 ). Digestate from the pulse-fed and continuously fed anaerobic digester were incubated in duplicate microcosms, which were spiked with either [ 12 C]- or [ 13 C]butyrate (40 mM) for approximately 50 h. The added butyrate was converted into methane at a >80% conversion efficiency based on chemical oxygen demand (COD) recovery (see Fig. S1 in the supplemental material). After the 50-h incubation, the contents of the microcosms were sacrificed for DNA extraction, density gradient centrifugation, and fractionation. 10.1128/mSystems.00159-19.1 FIG S1 Cumulative methane production (minus blank controls) for the microcosms fed with 12 C- and 13 C-labeled butyrate over approximately 50 h. The black dashed line shows the theoretical methane potential of the added butyrate (25.3 ml CH 4 ; based on 1.82 g COD/g butyrate, 40 mM concentration, 10 ml sample, and 35°C temperature). Error bars represent the standard deviations of results from the biological replicates. Download FIG S1, PDF file, 0.01 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . The abundance of 16S rRNA genes of the known butyrate-degrading genus Syntrophomonas was quantified across density gradient fractions using quantitative PCR (qPCR) to identify DNA fractions that were enriched in 13 C. Density fractions with a buoyant density from 1.70 to 1.705 had 2.0 to 2.2 times more Syntrophomonas 16S rRNA genes (normalized to the maximum concentration) than the 12 C controls ( Fig. S2 ). Those DNA fractions were selected from each SIP microcosm for metagenomic sequencing, as well as for 16S rRNA gene amplicon sequencing. 10.1128/mSystems.00159-19.2 FIG S2 Ratios of Syntrophomonas 16S rRNA genes measured by qPCR in each density gradient fraction to the maximum observed across all density fractions. The points indicate the average values of the biological duplicates for 13 C-incubated microcosms and 12 C-labeled controls for both anaerobic digesters. The widths of error bars indicate the range of the biological duplicates. The filled points indicate density gradient fractions that were pooled for subsequent 16S rRNA gene amplicon sequencing and metagenomic sequencing. Download FIG S2, PDF file, 0.03 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . The microbial communities in the heavy density gradient fractions were assessed through paired-end 16S rRNA gene amplicon sequencing for all 12 C- and 13 C-incubated duplicate microcosms ( Fig. 1 ). Differential abundance analysis of operational taxonomic unit (OTU) read counts with DESeq2 ( 14 ) showed that approximately 50% (7 of 15) of the significantly enriched ( P < 0.05) OTUs in heavy [ 13 C]DNA samples relative to heavy [ 12 C]DNA were taxonomically classified as Syntrophomonas for the pulse-fed digester ( Fig. S3 ). For the continuously fed digester, approximately 40% of the 13 C-enriched OTUs (7 of 17) were assigned to Syntrophomonas ( Fig. S3 ). Additionally, two 13 C-enriched OTUs in both digesters were assigned to Methanothrix (formerly Methanosaeta ), which likely scavenges the [ 13 C]acetate generated by Syntrophomonas during [ 13 C]butyrate degradation. While one previous study observed that Syntrophaceae was enriched predominantly in anaerobic digester granular sludge incubated with [ 13 C]butyrate ( 9 ), various other studies also detected Syntrophomonadaceae populations (i) as active syntrophic butyrate degraders in anaerobic digester sludge using [ 14 C]butyrate and microautoradiography–fluorescent in situ hybridization (MAR-FISH) ( 15 ), (ii) in anaerobic digester sludge by means of SIP using [ 13 C]oleate ( 11 ), and (iii) in rice paddy soil with SIP using [ 13 C]butyrate ( 16 ). In the last two studies, acetate-scavenging partners ( Methanothrix and Methanosarcinaceae ) were also enriched. Indeed, syntrophic interaction with acetoclastic methanogens is beneficial, as acetate accumulation can thermodynamically hinder butyrate oxidation (e.g., the Δ G exceeds the theoretical threshold for catabolism (–10 kJ/mol) when acetate accumulates beyond 10 mM (pH 7, 1 mM butyrate, and 1 Pa H 2 ) (assumptions appear in Table S2 ). Notably, H 2 - and formate-consuming methanogens necessary for syntrophy were not detected during degradation of [ 13 C]butyrate, likely because these archaea utilize CO 2 as a carbon source and no [ 13 C]CO 2 is produced during butyrate oxidation. FIG 1 Relative 16S rRNA gene amplicon sequence abundances of the top 12 most abundant prokaryotic genera in heavy DNA from [ 13 C]butyrate-amended microcosms and the [ 12 C]butyrate-amended controls for the pulse-fed codigester (A) and the continuously fed codigester (B). Values for each biological duplicate are shown for each condition ( 12 C or 13 C) for both anaerobic digesters. 10.1128/mSystems.00159-19.3 FIG S3 The log 2 -fold change in abundance of 16S rRNA amplicon OTUs in the pulse-fed codigester (A) and the continuously fed codigester (B) that were identified as significantly enriched in 13 C samples versus 12 C samples using DESeq2 ( 14 ), along with their genus- and phylum-level taxonomic assignments. Each point represents a 13 C-enriched OTU. The significantly enriched OTUs were detected based on their abundance in duplicate 13 C samples relative to duplicate 12 C samples. Download FIG S3, PDF file, 0.03 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . Our results also found 13 C-enriched OTUs from lineages not known to degrade butyrate under methanogenic conditions: Treponema , Luteimonas , Thauera , Christensenellaceae ( Firmicutes ), and Anaerolineaceae ( Chloroflexi ) ( Fig. S3 ). Other studies using [ 13 C]butyrate also detected enrichment of populations likely unable to degrade butyrate, including Tepidanaerobacter and Clostridium , in a thermophilic anaerobic digester operated at 55°C ( 9 ) and Chloroflexi and Planctomycetes in rice paddy soil ( 16 ). Members of Tepidanaerobacter and Clostridium are known to syntrophically oxidize acetate under thermophilic conditions ( 17 ) and may have thus been enriched in [ 13 C]RNA from [ 13 C]acetate produced during the beta-oxidation of labeled butyrate in the study by Hatamoto et al. ( 9 ). Similarly, the Chloroflexi and Planctomycetes populations were hypothesized to have become enriched due to cross-feeding of intermediate metabolites, like acetate, in rice paddy soil ( 16 ). Thus, the “peripheral” populations detected in our study may grow on cellular-decay products, as genome-resolved metagenomics recently indicated that some uncultured Anaerolineaceae species are likely fermenters in anaerobic digesters ( 18 ). These results thus suggest that carbon cross-feeding may occur between multiple microbial trophic groups during the syntrophic degradation of butyrate in anaerobic digesters. 10.1128/mSystems.00159-19.4 FIG S4 Heatmaps of average nucleotide identity (ANI) between genomes from Syntrophomonadaceae (A) and Methanosarcinales (B). Genomes were clustered based on the ANI values using Ward’s minimum-variance method. The genome names shown in bold were identified in this study. Other genomes were obtained via the NCBI nr database (downloaded April 2018). Download FIG S4, PDF file, 0.1 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . Identifying active metagenome-assembled genomes (MAGs) in SIP metagenomes. Metagenomic sequencing of heavy DNA from duplicate [ 13 C]- and [ 12 C]butyrate-amended microcosms yielded an average of 30 million paired reads per sample for both digesters ( n = 8) ( Table S1 ). The filtered reads from heavy [ 13 C]DNA were coassembled, yielding a total assembly length of 516 Mb of contigs larger than 1 kb, with an average ( N 50 ) contig length of 5 kb. The fraction of filtered short reads that mapped to the coassembly were 66% ± 3% (standard deviation) and 69% ± 1% for the 12 C- and 13 C-labeled metagenomes, respectively ( n = 4 each) ( Table S1 ). The coassembly generated from 13 C reads thus captured much of the genomic information present in the heavy-DNA fractions. 10.1128/mSystems.00159-19.5 TABLE S1 Summary of SIP metagenomes utilized for the coassembly, including the number of raw and filtered reads and the fraction of reads that mapped to coassembly. Download Table S1, DOCX file, 0.01 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . The assembled metagenomic contigs were organized into 160 genomic bins at various levels of completion and redundancy ( Data Set S1 ). Differential abundance analysis of the mapped read counts for the bins across the 13 C- and 12 C-labeled metagenomes with DESeq2 ( 14 ) identified two genomic bins that were significantly ( P < 0.05) enriched in [ 13 C]DNA ( Table 1 ). These genomic bins were enriched in [ 13 C]DNA in both the pulse-fed and continuously fed bioreactors. Based on suggested completion and redundancy metrics for MAGs ( 19 ), one genomic bin is classified as a high-quality MAG (completion, >90%; redundancy, <10%), while the other is a medium-quality MAG (completion, >50%; redundancy, <10%). Taxonomic classification with CheckM ( 20 ) assigned one of the MAGs to the genus Syntrophomonas and the other to Methanothrix ( Table 1 ). TABLE 1 Genomic feature summary of the two metagenome-assembled genomes that were significantly enriched in [ 13 C]DNA after the degradation of [ 13 C]butyrate Name Bin ID Taxonomy a \n Size (Mb) GC (%) Completion (%) b \n Redundancy (%) b \n Syntrophomonas BUT1 Bin 26_1 Syntrophomonas 2.87 51.2 96.4 1.4 Methanothrix BUT2 Bin 26_2 Methanothrix 1.44 53.6 74.7 3.1 a Based on phylogenetic placement of single marker genes with CheckM ( 20 ). b Measured with anvi’o ( 71 ). 10.1128/mSystems.00159-19.7 DATA SET S1 Summary of all genomic bins identified in the DNA-SIP metagenomes, including genome size, completeness, redundancy, coverage, and taxonomy. Download Data Set S1, XLSX file, 0.07 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . The phylogenomic placement of the 13 C-enriched Syntrophomonas BUT1 MAG was consistent with its taxonomic assignment, as it was located in the Syntrophomonas genome cluster within the family Syntrophomonadaceae ( Fig. 2 ). The closest relative to Syntrophomonas BUT1 based on single-copy marker genes was Syntrophomonas PF07, which is a genomic bin enriched in 13 C from DNA-SIP with labeled oleate ( 13 C 18:1 ) using sludge from the same pulse-fed digester used in this study ( 11 ). A high average nucleotide identity (ANI) of 99% was observed between the Syntrophomonas BUT1 and Syntrophomonas PF07 genomes ( Fig. S4 ), suggesting that these two organisms likely originated from the same sequence-discrete population ( 21 ). The next-closest relative of Syntrophomonas BUT1 based on the phylogenomic analysis was Syntrophomonas zehnderi OL-4 ( Fig. 2 ), which was isolated from an oleate-fed anaerobic granular sludge bioreactor ( 22 ). However, the ANI between Syntrophomonas BUT1 and Syntrophomonas zehnderi OL-4 was below 95% ( Fig. S4 ), suggesting that these two organisms are different species ( 23 ). Thus, the active butyrate-degrading bacterial MAG identified in this study is distinct from any species obtained by isolation at this time. The detection of the sequence-discrete population of Syntrophomonas BUT1 within heavy [ 13 C]DNA from both experiments with universally labeled butyrate and oleate indicates that this syntrophic population may be metabolically flexible; that is, it may grow on fatty acids of various lengths and degrees of saturation. An alternative explanation may be that Syntrophomonas BUT1 was detected in the SIP experiment with universally labeled [ 13 C]oleate due to its degradation of shorter fatty acids, such as butyrate, excreted during oleate degradation by other community members. These findings have implications for current frameworks for mathematical modeling of anaerobic digesters, which typically assume that long-chain fatty acid (LCFA)-degrading and butyrate-degrading populations are distinct and do not cross-feed ( 24 ). Thus, the incorporation of genomic and functional characterization, as obtained through DNA-SIP genome-resolved metagenomics, may help to improve our ability to accurately model anaerobic digestion processes by accounting for metabolic flexibility or cross-feeding within key functional guilds. FIG 2 Phylogenomic tree showing the relationship of 13 C-enriched Syntrophomonas BUT1 to other genomes available from the Syntrophomonadaceae family in the NCBI nr database (downloaded April 2018). The tree is based on a concatenated alignment of 139 bacterial single-copy marker genes ( 77 ) obtained using anvi’o ( 74 ). Open reading frames were predicted with Prodigal v.2.6.3 (70) and queried against sequences in a database of bacterial single-copy marker genes using HMMER v.2.3.2 ( 81 ). The tree was calculated using FastTree ( 82 ). The Clostridium ultunense genome was used as the outgroup. A phylogenomic analysis of the 13 C-enriched Methanothrix BUT2 MAG based on archaeal single-copy marker genes placed the MAG within the genus Methanothrix , consistent with its taxonomic assignment with CheckM ( Fig. 3 ). Methanothrix BUT2 was closely clustered with the genome of Methanothrix soehngenii GP6, along with four MAGs reported in the study of Parks et al. ( 25 ). Congruently with the phylogenomic analysis, Methanothrix BUT2 shared an ANI of over 98% with Methanothrix soehngenii GP6 and the same with four MAGs from the work of Parks et al. ( 25 ) ( Methanothrix UBA243, Methanothrix UBA458, Methanothrix UBA70, Methanothrix UBA356), indicating that these genomes likely form a sequence-discrete population ( Fig. S4 ). A second, closely related population, including three MAGs from the work of Parks et al. ( 25 ) ( Methanothrix UBA372, Methanothrix UBA332, Methanothrix UBA533) shared an ANI of 96% with the Methanothrix BUT2 population ( Fig. S4 ). FIG 3 Phylogenomic tree showing the relationship of the 13 C-enriched Methanothrix BUT2 to other genomes within the order Methanosarcinales in the NCBI nr database (downloaded April 2018). The tree is based on a concatenated alignment of 162 archaeal single-copy marker genes ( 78 ) obtained using anvi’o ( 74 ). Open reading frames were predicted with Prodigal v.2.6.3 ( 70 ) and queried against sequences in a database of archaeal single-copy marker genes using HMMER v.2.3.2 ( 81 ). The tree was calculated using FastTree ( 82 ). The “ Candidatus Methanoperedens nitroreducens” genome was used as the outgroup. DNA-SIP using [ 13 C]oleate with the same anaerobic digester biomass as in this study did not identify any 13 C-enriched methanogenic archaea in the genome-resolved metagenomic analysis ( 11 ). One possible explanation for the higher relative enrichment of methanogens on [ 13 C]butyrate than on [ 13 C]oleate may be the higher fraction of overall free energy partitioned toward methanogens during anaerobic butyrate degradation than during oleate degradation. For the overall conversion of 1 mol of butyrate to CO 2 and CH 4 under environmental conditions in anaerobic digesters, the thermodynamic yields would be −21.1, −9.4, and −58.9 kJ for the acetogenic bacteria, hydrogenotrophic methanogens, and acetoclastic methanogens, respectively ( Table S2 ). For a similar conversion of 1 mol of oleate, the thermodynamic yields would be −219.9, −70.6, and −264.9 kJ, respectively ( Table S2 ). Thus, the acetogen would gain a lower percentage of the overall free energy yield from the conversion of butyrate (24%) than from that of oleate (40%). As cell yield can depend on free energy ( 26 ), the lower yield of the butyrate degradation likely leaves a higher fraction of acetate for assimilation by an acetoclastic methanogen. The relative growth yields may also be particularly relevant due to the compositional nature of genome abundance data from the DNA-SIP metagenomes. As the stable-isotope-informed analysis utilized in this study depended on incorporation of the added 13 C into biomass, it was not expected that autotrophic (i.e., hydrogenotrophic) methanogens would be highly enriched in the heavy [ 13 C]DNA because no CO 2 is produced during butyrate beta-oxidation and microcosms were preflushed with N 2 -[ 12 C]CO 2 ( Table S2 ). Comparing the enriched communities from DNA SIP with different fatty acids, along with bicarbonate, may highlight differences in energy partitioning between syntrophic bacteria and different archaeal partners. 10.1128/mSystems.00159-19.6 TABLE S2 Gibbs free energy for some of the acetogenic and methanogenic reactions likely involved in the syntrophic conversion of butyrate and oleate. Download Table S2, DOCX file, 0.01 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . Metabolic potential of 13 C-enriched MAGs. Functional annotation and metabolic reconstruction of the 13 C-enriched MAGs revealed their capacity to metabolize the [ 13 C]butyrate into methane through syntrophic cooperation. A complete pathway for butyrate β-oxidation was annotated for Syntrophomonas BUT1, indicating that this MAG was capable of metabolizing the added [ 13 C]butyrate ( Fig. 4 ). Notably, several homologues were detected for genes in the β-oxidation pathway. The Syntrophomonas BUT1 genome encodes 6 acyl coenzyme A (acyl-CoA) transferases, 7 acyl-CoA dehydrogenases, 8 enoyl-CoA hydratases, 5 3-hydroxybutyryl-CoA dehydrogenases, and 10 acetyl-CoA acetyltransferases ( Data Set S2 ). The presence of homologous β-oxidizing genes was also observed in the type strain Syntrophomonas wolfei subsp. wolfei Göttingen DSM 2245B ( 27 ). The large number of homologous β-oxidizing genes may afford Syntrophomonas BUT1 flexibility to metabolize multiple fatty acid substrates, as its genomic population was detected in heavy [ 13 C]DNA during SIP with both [ 13 C]butyrate (C 4 ) and [ 13 C]oleate (C 18 ) ( 11 ). The different homologous β-oxidizing genes may also have different kinetics and/or affinities, which may allow Syntrophomonas BUT1 to adapt to various substrate concentrations. Fluctuating environments are thought to lead to robustness toward gene loss within metabolic networks through an increase in multifunctional enzymes ( 28 ). Thus, the presence of various homologous genes for β-oxidation in Syntrophomonas BUT1 may have been selected for by fluctuating fatty acid concentrations, such as those imposed from pulse-feeding the anaerobic digester ( 13 ). It is also possible that the Syntrophomonas BUT1 population was enriched in 13 C from labeled oleate due to cross-feeding of shorter-chain intermediates during β-oxidation of the C 18 LCFA, as other syntrophic bacteria were enriched to a high degree during growth on [ 13 C]oleate ( 11 ). Yet, the enrichment of Syntrophomonas BUT1 on [ 13 C]butyrate, along with the presence of the complete butyrate β-oxidation pathway, strongly suggests that it is at least capable of β-oxidizing shorter-chain fatty acids (e.g., C 4 ) produced in anaerobic environments. FIG 4 Cell diagram showing proposed metabolic pathways for anaerobic butyrate degradation in syntrophic cooperation between Syntrophomonas BUT1 and Methanothrix BUT2. The H 2 /formate-utilizing methanogenic partner is shown for conceptual purposes but was not identified with [ 13 C]DNA-SIP in this study due to its autotrophic growth in the microcosms. Dotted lines indicate the direction of electron flow. Details of predicted proteins are given in Data Sets S2 and S3 . Enzyme abbreviations are as follows: Fd, ferredoxin; ( Syntrophomonas BUT1) Acd, acyl-CoA dehydrogenase; Crt, enoyl-CoA hydratase; HbdH, 3-hydroxybutyryl-CoA dehydrogenase; AtoB, acetyl-CoA acetyltransferase; AckA, acetate kinase; Pta, phosphate acetyltransferase; EtfA, electron transfer flavoprotein A; EtfB, electron transfer flavoprotein B; EtfD, EtfAB:quinone oxidoreductase; HydABC, bifurcating [Fe-Fe] hydrogenase; HyaABC, [NiFe] hydrogenase; FdhA-HylBC, formate dehydrogenase (electron bifurcating); FdnGHI, formate dehydrogenase (membrane bound, quinone reducing); FixC, electron transfer flavoprotein dehydrogenase; FixX, FixABC-associated ferredoxin; ( Methanothrix BUT2) Acs, acetyl-coenzyme A synthetase; CooS, carbon monoxide dehydrogenase; CdhA, acetyl-CoA decarbonylase/synthase complex; Mtr, methyltetrahydromethanopterin:CoM methyltransferase; McrABG, methyl-coenzyme M reductase; HdrED, coenzyme B-coenzyme M heterodisulfide reductase; FpoABDHIJKLMNO, F 420 H 2 dehydrogenase. 10.1128/mSystems.00159-19.8 DATA SET S2 Annotation information for key metabolic genes for the MAG of Syntrophomonas BUT1, including those for beta-oxidation, electron transfer, and others. Download Data Set S2, XLSX file, 0.02 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . 10.1128/mSystems.00159-19.9 DATA SET S3 Annotation information for key metabolic genes for the MAG of Methanothrix BUT2, including those for methanogenesis and others. Download Data Set S3, XLSX file, 0.02 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . Syntrophomonas BUT1 lacks genes for aerobic or anaerobic respiration, which is similar to genomes of S. wolfei and Syntrophus aciditrophicus that are capable of syntrophic butyrate degradation ( 27 , 29 ). Electrons derived from butyrate oxidation (reduced electron-transferring flavoprotein [ETF] from butyryl-CoA oxidation and NADH from 3-hydroxybutyryl-CoA oxidation) must be disposed of through reduction of CO 2 to formate and H + to H 2 via formate dehydrogenases and hydrogenases, respectively ( 30 – 33 ). In the Syntrophomonas BUT1 genome, we identified genes encoding butyryl-CoA dehydrogenase, ETF alpha and beta units (EtfAB), and two EtfAB:quinone oxidoreductases ( Data Set S2 ), indicating that this organism may transfer electrons from butyryl-CoA oxidation into membrane electron carriers using ETF. The Syntrophomonas BUT1 genome contains five gene clusters encoding formate dehydrogenases and four gene clusters encoding hydrogenases ( Data Set S2 ). These include a membrane-bound cytochrome b -dependent selenocysteine-containing formate dehydrogenase and [NiFe] hydrogenase, which may receive butyrate-derived electrons via menaquinol ( 30 ). The quinone-binding site of the selenocysteine-containing formate dehydrogenase was on the cytoplasmic side, indicating that it likely utilizes proton motive force to drive unfavorable electron transfer to CO 2 -reducing formate generation outside the cell. Energy investment via “reverse electron transport” is critical to drive the uphill electron transfer from the butyryl-CoA/crotonyl-CoA couple to CO 2 /formate or H + /H 2 couples. In contrast, the quinone binding site of the [NiFe] hydrogenase was on the periplasmic side, indicating that it couples outward vectorial proton transport with H 2 generation. Previous genomic and proteomic studies also highlight the importance of ETF-based electron transfer, membrane-bound formate dehydrogenases/hydrogenases, and reverse electron transport ( 5 , 27 , 33 – 36 ). To complete syntrophic butyrate oxidation, NAD + must also be regenerated through oxidation of NADH. However, NADH oxidation coupled with CO 2 /H + -reducing formate/H 2 generation is thermodynamically unfavorable. To address this obstacle, anaerobic organisms are known to utilize electron bifurcation (or confurcation), which involves the coupling of endergonic and exergonic redox reactions to circumvent energetic barriers ( 37 ). For instance, Thermotoga maritima utilizes a trimeric hydrogenase to couple the endergonic production of H 2 from NADH with the exergonic production of H 2 from reduced ferredoxin ( 38 ). Two trimeric formate dehydrogenase- and two trimeric [FeFe] hydrogenase-encoding gene clusters in Syntrophomonas BUT1 appear linked to NADH, as they all contained an NADH:acceptor oxidoreductase subunit ( Data Set S2 ). Yet, if the trimeric hydrogenases and formate dehydrogenases in Syntrophomonas BUT1 produce H 2 /formate via electron bifurcation with NADH and ferredoxin, it remains unknown how Syntrophomonas BUT1 regenerates reduced ferredoxin, as the known butyrate β-oxidation pathway does not generate reduced ferredoxin ( 30 ). Moreover, the Syntrophomonas BUT1 genome does not encode an Rnf complex, which would be necessary to generate reduced ferredoxin from NADH. Recently, the Fix (homologous to ETF) system was shown to perform electron bifurcation to oxidize NADH coupled with the reduction of ferredoxin and ubiquinone during N 2 fixation by Azotobacter vinelandii ( 39 ). The Syntrophomonas BUT1 genome encoded a Fix-related ETF-dehydrogenase, FixC, as well as its associated ferredoxin, FixX ( Data Set S2 ). A Fix system has also been detected in S. wolfei and was postulated to serve as a means of generating reduced ferredoxin for H 2 or formate production via the bifurcation mechanism ( 30 ). Yet, reduced ferredoxin production with the Fix system would be energetically costly, especially with regard to the low energy yields during syntrophic butyrate oxidation ( 40 ). Another mechanism was proposed for generating reduced ferredoxin in Rnf-lacking syntrophs that involves a heterodisulfide reductase complex (HdrABC) and an ion‐translocating flavin oxidoreductase (Flx or Flox) ( 41 ). The flxABCD‐hdrABC gene cluster was shown to be widespread among anaerobic bacteria, and the protein cluster (FlxABCD-HdrABC) is proposed to function similarly to the heterodisulfide reductase (HdrABC)–[NiFe]-hydrogenase (MvhADG) complex (HdrABC-MvhADG) involved in flavin-based electron bifurcation in hydrogenotrophic methanogenic archaea that couples the exergonic reduction of CoM‐CoB heterodisulfide (CoM‐S‐S‐CoB) with the endergonic reduction of ferredoxin with H 2 ( 42 ). A full floxABCD‐hdrABC gene cluster was detected in the genome of Syntrophomonas BUT1 ( Data Set S2 ). During the syntrophic growth of Syntrophomonas BUT1 on butyrate, the FlxABCD-HdrABC protein cluster may oxidize NADH with reduction of ferredoxin along with the reduction of a high‐redox‐potential disulfide acceptor ( 42 ). In Desulfovibrio vulgaris , it has been proposed that the DsrC protein serves as the high‐redox thiol-disulfide electron carrier that is reduced by the FlxABCD-HdrABC complex during growth ( 43 ). The DsrC protein was also detected in the syntrophic benzoate-degrading Syntrophorhabdus aromaticivorans strain UI, along with an flxABCD‐hdrABC gene cluster ( 41 ), suggesting that the reduction of a thiol-disulfide electron carrier may be a conserved mechanism for generating reduced ferredoxin in syntrophic bacteria. Yet, the Syntrophomonas BUT1 genome does not encode a DsrC protein, and thus an alternative and unknown thiol-disulfide electron carrier would be needed. Another possibility is that the trimeric hydrogenase can drive NADH-dependent H 2 generation, as shown in S. wolfei Goettingen ( 40 ). Nonetheless, this genomic analysis demonstrates that Syntrophomonas BUT1 has the potential capacity to overcome energetic barriers during syntrophic butyrate β-oxidation and contains multiple possible mechanisms for H 2 and formate production. In addition to interspecies electron transfer via molecular hydrogen and formate, a potential mechanism has been proposed for direct interspecies electron transfer (DIET), in which electrons are shared via electrically conductive nanowires ( 44 ). DIET activity has been suggested in enrichment communities degrading propionate and butyrate, in which Syntrophomonas was detected ( 45 , 46 ). However, DIET has not been demonstrated with pure cultures of Syntrophomonas to date. The direct transfer of electrons is thought to depend on electrically conductive type IV pili and external polyheme cytochromes ( 47 , 48 ). The Syntrophomonas BUT1 genome encodes a type IV pilin assembly protein, PilC, but no genes that encoded the structural protein PilA, which is associated with DIET ( 48 ), were found. Moreover, the type IV pilin genes identified in the Syntrophomonas BUT1 genome were of the type Flp (fimbrial low molecular protein weight), which are smaller than the Pil - type pilin utilized for DIET in Geobacter ( 49 , 50 ). A multiheme c -type cytochrome was detected in the Syntrophomonas BUT1 genome that had 59% amino acid identity (89% coverage) with the multiheme c -type cytochrome OmcS from G. sulfurreducens , which has been implicated in DIET ( 48 ) ( Data Set S2 ). However, that gene also had higher homology (69% identity, 94% coverage) with the cytochrome c nitrite reductase from S. wolfei (GenBank accession no. WP_081424886 ). Therefore, the roles of DIET in the metabolism of Syntrophomonas BUT1 remain unclear but warrant further attention via expression-based profiling. In addition to encoding potential genetic mechanisms for energy conservation during syntrophic growth, Syntrophomonas BUT1 encodes a capsule biosynthesis protein (CapA), which appears to be specific to syntrophic growth ( 51 ). The function of CapA in syntrophic growth is unclear but may be related to the production of exopolymeric substances that facilitate interactions with methanogenic partners ( 51 ). The Syntrophomonas BUT1 genome also contains the ftsW gene, which is related to shape determination and is also a postulated biomarker of a syntrophic lifestyle ( 51 ). Based on the presence of these “syntrophic biomarkers” along with genes for β-oxidization and H 2 /formate production, the genomic repertoire of Syntrophomonas BUT1 aligns with that of a syntrophic butyrate degrader. The genome of Syntrophomonas BUT1 was compared with published genomes of the Syntrophomonas genus ( S . wolfei subsp. wolfei , S. wolfei subsp. methylbutyratica , and S. zehnderi ) to investigate whether metabolic genes for beta-oxidation and energy conservation were conserved ( Data Set S4 ). Cutoffs of 42% amino acid similarity and 80% sequence overlap were employed based on the lowest first-quartile amino acid similarity that we observed for top BLAST hits (minimum of 20% amino acid similarity and 80% overlap) of Syntrophomonas BUT1 genes to each aforementioned Syntrophomonas genome (42.0%, 43.5%, and 43.5%, respectively). Based on these similarity thresholds, only 34% (1,050 out of 3,066) of protein-coding genes in the Syntrophomonas BUT1 genome have closely related homologs present in all of the other sequenced Syntrophomonas genomes. Notably, 40% of the Syntrophomonas BUT1 protein-coding genes have no homologs in other Syntrophomonas genomes that meet the similarity criteria described above. Reflecting this genomic diversity, Syntrophomonas BUT1 encodes several beta oxidation-related genes that have no homologs in the other Syntrophomonas genomes that meet the above criteria: one acetyl-CoA acetyltransferase, acyl-CoA dehydrogenase, acrylyl-CoA reductase, and acyl-CoA thioesterase ( Data Set S4 ). In addition, the Syntrophomonas BUT1 genome harbors putative isobutyryl-CoA mutase genes (SYNMBUT1_v1_1780025–27) highly similar to those of Syntrophothermus lipocalidus (65.0 to 83.4% amino acid similarity), suggesting that Syntrophomonas BUT1 may also be capable of syntrophic isobutyrate degradation. Hydrogenases, formate dehydrogenases, and energy conservation genes were generally conserved among Syntrophomonas BUT1 and the other Syntrophomonas genomes. Only the cytochrome b -dependent [NiFe] hydrogenase has no homologs in the S. wolfei subsp. wolfei genome. This implies that Syntrophomonas BUT1 may have distinct capabilities for fatty acid oxidation, but the levels of energy conservation necessary to drive syntrophic beta oxidation may not vary between Syntrophomonas species. 10.1128/mSystems.00159-19.10 DATA SET S4 Genomic comparison of published Syntrophomonas BUT1 genomes of the Syntrophomonas genus ( S . wolfei subsp. wolfei , S. wolfei subsp. methylbutyratica , and S . zehnderi ). Homologues were identified with a cutoff of 42% amino acid similarity and 80% sequence overlap. Download Data Set S4, XLSX file, 0.2 MB . Copyright © 2019 Ziels et al. 2019 Ziels et al. This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . A genomic analysis of the Methanothrix BUT2 genome indicated that it contained the complete pathway for methane production from acetate ( Fig. 4 ; Data Set S3 ). This observation agrees with the physiology of other Methanothrix species, which are known acetoclastic methanogens ( 52 , 53 ). Methanothrix BUT2 also contained genes that likely are involved in energy conservation during acetoclastic methanogenesis. The genome of Methanothrix BUT2 harbored acetyl-CoA synthetase for acetate activation, bifunctional CO dehydrogenase/acetyl-CoA synthase (CODH/ACS) to oxidatively split acetyl-CoA into CO 2 and CH 3 -H 4 MPT, tetrahydromethanopterin S -methyltransferase, and methyl-CoM reductase for methyl-CoM reduction to CH 4 ( Data Set S3 ). To couple acetyl-CoA oxidation and reductive CH 4 generation, BUT2 must transfer electrons from reduced ferredoxin to coenzyme M (CoM-SH) and coenzyme B (CoB-SH). We identified an FpoF-lacking F 420 H 2 dehydrogenase (Fpo) complex and heterodisulfide reductase (HdrDE) that could facilitate this ( Data Set S3 ) and also generate an ion motive force ( 54 ). This energy conservation system is highly similar to Methanothrix thermophila acetate oxidation ( 54 ). In previous studies, Methanothrix species have been observed to cooccur with Syntrophomonas in LCFA-degrading ( 13 ) and butyrate-degrading ( 55 – 57 ) anaerobic environments. In this study, the stable-isotope-informed metagenomic analysis strongly suggests that the labeling of Methanothrix BUT2 DNA was due to the incorporation of [ 13 C]acetate produced during the degradation of [ 13 C]butyrate by Syntrophomonas BUT1. A nearly complete pathway for methane production from CO 2 was also observed in the Methanothrix BUT2 genome ( Data Set S3 ). The only gene lacking in the CO 2 -reducing pathway was an F 420 -dependent N 5 , N 10 -methylene-tetrahydromethanopterin dehydrogenase (Mtd). While Methanothrix is thought to be an obligate acetoclastic methanogen ( 52 , 53 ), the presence and expression of the CO 2 -reducing pathway in Methanothrix were previously reported ( 58 – 60 ) and were hypothesized to be involved in methane formation via DIET. However, the mechanism through which Methanothrix directly accepts electrons from its syntrophic partner has not been identified ( 58 , 59 ). The other known electron donors for methane production from CO 2 are hydrogen and formate. A membrane-bound hydrogenase ( mbhAB ) was observed in the Methanothrix BUT2 genome ( Data Set S3 ). In other studies, negligible hydrogenase activity was observed with Methanothrix species ( 61 ). Two monomeric formate dehydrogenase enzymes (FdhA) were also encoded by Methanothrix BUT2 ( Data Set S3 ). Experiments with thermophilic Methanothrix sp. strain CALS-1 and mesophilic Methanothrix concilii showed that they displayed formate dehydrogenase activity by splitting formate into hydrogen and CO 2 ; however, the produced CO 2 was not used for methane generation ( 61 , 62 ). Yet, the mesophilic M. soehngenii did not show formate dehydrogenase activity ( 53 ). Thus, the roles of the hydrogenases, formate dehydrogenases, and CO 2 -reducing pathway for methane generation in Methanothrix BUT2 are not clear. Transcriptomic, metabolomic, and/or proteomic approaches are needed to elucidate the activity of the CO 2 -reducing methanogenesis production pathway during syntrophic growth on butyrate with Syntrophomonas BUT1. Conclusions. In this study, stable-isotope-informed genome-resolved metagenomics was used to provide genomic insights into syntrophic metabolism during butyrate degradation in anaerobic digesters. The results obtained via genome binning and metabolic reconstruction showed that a 13 C-enriched Syntrophomonas genome contained the genetic capacity to convert butyrate into precursor metabolites for methane formation: acetate, hydrogen, and formate. A 13 C-enriched Methanothrix genome likely consumed the acetate produced during butyrate degradation, incorporating some 13 C into biomass. The presence of a CO 2 -reducing pathway, as well as formate dehydrogenase and hydrogenase genes, in the Methanothrix genome leaves open the possibility of flexible metabolism during methanogenesis. As syntrophic fatty acid-degrading populations are often slow-growing and thus difficult to isolate, this study demonstrates a new approach to link ecophysiology with genomic identity in these important populations involved in anaerobic biotechnologies as well as global carbon cycling. Advancing our understanding of in situ metabolic activities within anaerobic communities is paramount, as these microbiomes contain multiple interacting functional groups that, in cooperation, enable the processing of degradable organic carbon into methane gas. Coupling SIP-informed metagenomics with other activity-based techniques, such as metabolomics, transcriptomics, and proteomics, may further illuminate the structure of anaerobic metabolic networks as well as quantify metabolite fluxes, thus enabling newly informed process models to predict rates of anaerobic carbon transformation."
} | 11,801 |
36386676 | PMC9659907 | pmc | 2,112 | {
"abstract": "Sensing surface topography, an upsurge of signaling biomolecules, and upholding cellular homeostasis are the rate-limiting spatio-temporal events in microbial attachment and biofilm formation. Initially, a set of highly specialized proteins, viz . conditioning protein, directs the irreversible attachment of the microbes. Later signaling molecules, viz . autoinducer, take over the cellular communication phenomenon, resulting in a mature microbial biofilm. The mandatory release of conditioning proteins and autoinducers corroborated the existence of two independent mechanisms operating sequentially for biofilm development. However, both these mechanisms are significantly affected by the availability of the cofactor, e.g., Copper (Cu). Generally, the Cu concentration beyond threshold levels is detrimental to the anaerobes except for a few species of sulfate-reducing bacteria (SRB). Remarkably SRB has developed intricate ways to resist and thrive in the presence of Cu by activating numerous genes responsible for modifying the presence of more toxic Cu(I) to Cu(II) within the periplasm, followed by their export through the outer membrane. Therefore, the determinants of Cu toxicity, sequestration, and transportation are reconnoitered for their contribution towards microbial adaptations and biofilm formation. The mechanistic details revealing Cu as a quorum quencher (QQ) are provided in addition to the three pathways involved in the dissolution of cellular communications. This review articulates the Machine Learning based data curing and data processing for designing novel anti-biofilm peptides and for an in-depth understanding of QQ mechanisms. A pioneering data set has been mined and presented on the functional properties of the QQ homolog in Oleidesulfovibrio alaskensis G20 and residues regulating the multicopper oxidase properties in SRB.",
"conclusion": "Conclusion Identification of ~700 microbial autoinducers corroborates a clear understanding of the later stages of microbial colonization resulting in a mature biofilm. In contrast, no structural and functional description is available on conditioning proteins regulating the first irreversible attachments of planktonic cells in natural landscapes. This extreme disparity in knowledge highlights the distinctive gap in scientific understanding of SRB’s communication channels and coping mechanisms active during stress conditions. The classification of the conditioning film’s component is the critical step in identifying the molecular determinants of microbial attachment. Precise monitoring of the microbial profile and microcosmic conditions is the prerequisite, which is otherwise rate-limiting variables in accurately characterizing the biomolecules. Likewise, given the diverse functionality of the Cu-ions in SRB growth and communication channels, an updated understanding of its sequestration pathways is needed. With the advent of the ML tools, numerous robust algorithms are under testing to annotate unexplored assignments of the Cu-ions. In addition, the central role of Cu as a cofactor in metabolic enzymes necessitates monitoring metabolic flux to establish a correlation between nutritional stress and biofilm formation by SRB. Revisiting the metabolic phenotypes will assist in biomanufacturing the Green solutions to prevent the catastrophic impacts of MIC caused by the SRB biofilms.",
"introduction": "Introduction Sulfate reducing bacteria (SRB) can be broadly categorized as mesophilic microbes, with the exception of thermophilic strains isolated from the oil and gas industry drainages, displaying autotrophic, litho-autotrophic, or heterotrophic respiration types under anaerobiosis ( Mayilraj et al., 2009 ; Fichtel et al., 2012 ; Jia et al., 2018 ). Numerous species of sulfate reducing bacteria viz . Desulfovibrio alaskensis ( Wikieł et al., 2014 ), Desulfovibrio vulgaris ( Ueki and Lovley, 2022 ), Desulfovibrio brasiliensis ( Warthmann et al., 2005 ), and Desulfovibrio ferrophilus ( Chatterjee et al., 2021 ) are reported for their biofouling, biofilm forming, and microbially induced corrosion (MIC) properties. The impact of the MIC will be more profound as it causes putrefaction of the metal pipelines, which may result in an oil spillage scenario by damaging supply pipelines. The detrimental impact of MIC is led by the biofilms originating from the microcosmic heterogenic microbial aggregations. This localized microbial aggregation is controlled by the surface properties, including microbe-metal interactions leading to the conditioning film development. Reports validate the conditioning film’s development within the first few minutes of microbial interactions with the surface. The nature of the conditioning films differs significantly depending on the natural landscapes, microcosmic environment, viz . metal type, nutrient availability, hydrophobicity, etc. ( Lorite et al., 2011 ). Conversely, the growing conditioning film also changes the surface properties, e.g., hydrophobicity, hydrophilicity, roughness, and charge ( Arnaouteli et al., 2016 ). Reports describe the complex proteinaceous nature ( Fong and Yildiz, 2015 ) of the conditioning films rich in exopolysaccharides, glycolipids, and other microbial secretions. Fibrinogen and fibronectins are the documented conditioning proteins (CP) dominating the pathogenic conditioning film on human implants ( Nandakumar et al., 2013 ; Souza et al., 2015 ; Khatoon et al., 2018 ). Human proteins such as albumin, vitronectin, and collagens are adsorbed on the surface of implants, making them suitable for bacterial attachment. Despite the complexity in the process of adsorption, the adsorption of proteins on the surface would increase with time, at least until the adsorbed proteins approach monolayer coating. Over time, without cellular interaction, absorbed layers are displaced by proteins such as kininogen with higher surface affinity. However, the family, structure, and properties of the homologous counterparts in biofilms growing under natural landscapes, e.g., hot springs, deep biosphere, water channels, and rhizospheres, are unexplored and undocumented. In addition to surface properties, micronutrient availability also influences SRB colonization. Microbes widely use copper (Cu) as a micronutrient and cofactor in numerous redox enzymes, e.g., cytochrome oxidase, transcription regulators, etc. At elevated concentrations, Cu is inhibitory toward SRB growth; hence, preferred for industrial pipelines, offshore installations, etc., where microbial colonization is a challenge. The presence of Cu restricts colonization by facilitating reactive oxygen formation or replacing Fe 2+ in the iron–sulfur (Fe-S) cluster of the critical enzymes. Contrary to the discernment of Cu as a quorum quencher (QQ), biofilms have been observed on Cu and nickel surfaces due to novel homeostasis mechanisms developed by a few species of SRB ( Rodrigues et al., 2008 ). Despite the reporting on Cu as a QQ, relevant mechanistic details are underexplored in the case of SRB. In the current scenario, a comprehensive review detailing initial cell attachment and cellular communications is well due as reports are available detailing biochemical events resulting in a mature SRB’s biofilm. With the broad-spectrum applicability of the Machine Learning (ML) platform, enormous opportunities have arisen in finding efficient QQ molecules and understanding common trends from the available scientific data. Therefore, relatively underexplored Oleidesulfovibrio alaskensis G20 has been selected as a model SRB to decode the roles of the multicopper oxidases and lactonases proteins actively involved in the colonization and communication stages. The phylogenetic information of the D . alaskensis G20 is given in Figure 1 , and several unique features of the SRB are given in Table 1 . The potentials of the ML platform have been investigated for facilitating the experimental validation of signaling molecules and designing anti-biofilm peptides. Given the ecological and industrial significance of the SRB biofilms, an assessment of literature from the past 15 years till June 2022 has been carried out to investigate the roles of (i) conditioning films and their constituents in microbial colonization, (ii) SRB adaptations towards elevated levels of copper, and (iii) quorum quenching mechanisms and role of ML in identifying novel quenching peptides for SRB. Figure 1 Phylogenetic position of the model organism Desulfovibrio alaskensis G20 among different sulfate reducing bacteria. Table 1 Overview of gene expression profiling and adaptations displayed by sulfur reducing bacteria in response to Cu-toxicity. Expression level Microbe Genes Cellular functionality References Downregulated Desulfovibrio vulgaris strain Hildenborough DVU0307, DVU1443, and DVU1444 Cell motility \n Copper (2018) DVU3132 Lactate oxidation dcrH , DVU0170 , CheW-1 , DVU0591, and DVU0700 Chemotaxis * Oleidesulfovibrio alaskensis G20 Dde_0356, Dde_ 2,958 Cell motility \n Hao et al. (1994) Dde_3239 Lactate oxidation CheD , CheW , CheC , Chez Chemotaxis * feoA , and Dde_2669 Inorganic ion transporter Upregulated Desulfovibrio vulgaris strain Hildenborough DVU2571, DVU2572, and DVU2574 Inorganic ion transporter \n Copper (2018) DVU1257, hup-2 DNA replication Oleidesulfovibrio alaskensis G20 Dde_2324 DNA replication \n Hao et al. (1994) * Includes methyl-accepting chemotaxis protein and cytosolic chemotaxis protein only."
} | 2,372 |
36750548 | PMC9905593 | pmc | 2,113 | {
"abstract": "Conventional artificial intelligence (AI) machine vision technology, based on the von Neumann architecture, uses separate sensing, computing, and storage units to process huge amounts of vision data generated in sensory terminals. The frequent movement of redundant data between sensors, processors and memory, however, results in high-power consumption and latency. A more efficient approach is to offload some of the memory and computational tasks to sensor elements that can perceive and process the optical signal simultaneously. Here, we proposed a non-volatile photomemristor, in which the reconfigurable responsivity can be modulated by the charge and/or photon flux through it and further stored in the device. The non-volatile photomemristor has a simple two-terminal architecture, in which photoexcited carriers and oxygen-related ions are coupled, leading to a displaced and pinched hysteresis in the current-voltage characteristics. For the first time, non-volatile photomemristors implement computationally complete logic with photoresponse-stateful operations, for which the same photomemristor serves as both a logic gate and memory, using photoresponse as a physical state variable instead of light, voltage and memresistance. The polarity reversal of photomemristors shows great potential for in-memory sensing and computing with feature extraction and image recognition for neuromorphic vision.",
"introduction": "Introduction The human vision system has a powerful capability in visual perception only consuming less than twenty watts of power. Such features are mainly attributed to the simultaneous sensing and early processing of visual information in the retina and parallel processing in the visual cortex 1 – 8 . For example, to efficiently discard the redundant visual data and accelerate subsequent processing tasks in the visual cortex, the human retina can extract critical features of visual data with plastic positive and negative photoresponse 9 . Inspired by the human vision system, AI machine vision technology has been developed to achieve the capability of perception. Usually, in traditional vision systems, the optical information is captured by a frame-based digital camera, and then the digital signal is processed afterward using machine-learning algorithms. In this scenario, a large amount of data (mostly redundant) has to be transferred from a standalone sensing elements to the processing units, which leads to a large latency and power consumption. To address this problem, much effort has been devoted to developing an in-sensor computing technology by emulating certain functions of the human retina, for example, metal-semiconductor-metal variable-sensitivity photodetectors (VSPDs), reconfigurable 2D semiconductor photodiodes, gate-tunable van der Waals heterostructures, etc. The above sensors constitute a built-in artificial neural network that can sense and process images simultaneously 10 – 17 . However, challenges still exist. The VSPDs based on the metal-semiconductor-metal structure had a bias-dependent dark current, and the reconfigurable 2D material-based neural network image sensors are volatile and need continuing gate voltage to update weights 9 , 18 – 21 . To create a non-volatile photodetector with tunable photoresponses requires complex device designs or manufacturing processes, for example, using a floating gate or a ferroelectric gate dielectric. Therefore, in order to efficiently process such a large amount of data and reduce power consumption, it is necessary to develop a non-volatile photodetector device with a simple architecture for high-density integration. The coupled electron-ion memristive system allows multiple resistive states being adjusted by memorizing the history of previous electrical inputs, thus simulating biological synapses 22 – 25 . The conductivity of memristors changes with an external bias voltage, while non-volatile resistive states are preserved 22 , 23 , 26 – 29 . Furthermore, the crossbar array based on the memristors can perform the matrix-vector product operation efficiently through Ohm’s law and Kirchhoff’s law with energy-efficient in-memory computing 30 , 31 . Inspired by memristive devices and the requirement for massively parallel computing in image processing, we have developed simple, two-terminal, non-volatile photomemristors with tunable photoresponsivity, in which the responsivity can be modulated by charge and/or photon flux through it. This new concept provides possibilities to achieve all-in-one sensing-memory-computing device with simple architecture for the implementation of in-sensor computing network. In this paper, simple two-terminal photomemristors based on 2D Graphene/MoS 2-x O x /Graphene (G/M/G) structures are experimentally demonstrated. Recorded high endurance (more than a hundred cycles) and reliable retention (more than a thousand seconds) indicate that our device can be used for multistate non-volatile photodetection. In addition, we demonstrate the photoresponse-stateful computationally complete logic with the photomemristors set, in which the same photomemristor serves simultaneously as logic gates and memory unit. Instead of physical state variables of light, voltage and conductance 32 – 34 , the non-volatile photoresponse is first served as the variable. This kind of photoresponse-stateful in-memory computing strategy can expand the functional diversity of edge-side neural networks such as binarized neural networks 33 . Furthermore, the proposed photomemristor arrays provide image pre-processing and recognition with multistate photoresponse, suggesting that a new type of photomemristor opens the possibility for the implementation of an in-sensory network in the future. This new type of two-terminal photomemristor not only provides versatile sensing-memory-computing approaches for neuromorphic vision hardware but also enables high-density integration.",
"discussion": "Discussion In order to investigate the mechanism of non-volatile photoresponsivity switching in two-terminal G/M/G devices, in-situ Raman analysis of structures with thinner MoS 2-x O x was performed to be able to characterize the underlying graphene electrodes (details in Figs. S12 – S15 in Supplementary Information Section C). Figure 2a shows the I-V characteristics of this sample (see also Fig. S12d , Supplementary Information). The photocurrent switches from LPS to HPS at a SET voltage of about 1.2 V, and the HPS switches to LPS at a RESET voltage of about -1.0 V. When the device switched from LPS to HPS, we measured the Raman modes of cathode and anode under MoS 2-x O x . Figure 2b shows the Raman scattering modes of the cathode. The I D /I G ratio decreased from 0.51 to 0.33 and the peak positions of the G- and 2D-bands show redshifts of 9 cm −1 and 7 cm −1 , respectively, as shown in Fig. 2b and c . Such a change in the Raman modes indicates the reduction of graphene 43 , 44 . After the RESET process, the I D /I G ratio increased to 0.49. In this case, a blue shift of the G- and 2D-bands was observed, demonstrating the oxidation process 43 – 45 . At the same time, the corresponding reduction and oxidation of the anode are observed for the Set and Reset processes (see Figs. S14 and S15 , Supplementary Information). An increase (decrease) in the short-circuit photocurrent of the device is accompanied by reduction (oxidation) of cathode electrodes, which indicates that photoresponsive switching correlates with reversible redox reactions at the MoS 2-x O x /G interface 46 . Note that such a reversible redox process at the required potential is observed only under illumination. Resistive switching in the dark requires a higher voltage of 12 V (Fig. S16 , Supplementary Information). Fig. 2 Photoresponse switching mechanism. a I-V characteristics of a G/M/G device with binary photoresponse switching when the voltage sweeps from 0 to 2 V, from 2 V to –2 V, and back to 0 V. b Raman curves showing the evolution of redox reactions on graphene electrodes when switching between HPS and LPS. c Correlation of changes in the ratio of the intensities of Raman scattering and shifts of the G mode of graphene electrodes and states of the non-volatile photoresponse. TCAD-simulated dark/photo current of the G/M/G device for LPS ( d ) and HPS ( e ). The insets of ( d ) and ( e ) demonstrate the qualitative device model for LPS and HPS, respectively. f The electron current density distribution of LPS (top) and HPS (bottom) under illumination, simulated with TCAD. g TCAD simulated the hole current density distribution of LPS (top) and HPS (bottom) under illumination, simulated with TCAD Reversible partial oxidation (reduction) of graphene (graphene oxide) leads to a reversible decrease (increase) in the mobility of charge carriers by more than an order of magnitude 43 . In addition, Fig. S17 confirms that the conductivity of our CVD graphene decreases when the oxidation degree increases. This can change the collection efficiency of photoexcited carriers at the cathode and anode. To investigate the relationship between photoresponse states and the redox process on graphene electrodes, a series of Sentaurus Technology Computer Aided Design (TCAD) simulations were run as shown in Fig. 2 d–g. When we read a G/M/G device in a low voltage range, the device acts like two back-to-back Schottky diodes connected in series, as shown in the insets of Fig. 2d and e . As the effective contact size increases (decreases), the resistance of the corresponding diode decreases (increases), and we simplify this change by the diode size 47 . The dark current is symmetric with a symmetrical oxidation degree of cathode and anode as shown in the black curve in Fig. 2d . When the bias voltage is 0, the current density of the anode and cathode have the same values with opposite polarity (top current density plot in Fig. 2f and g ) under illumination. Thus, the short-circuit photocurrent is 0, as shown by the red curve in Fig. 2d , which corresponds to the LPS of our G/M/G structure. After applying a positive bias voltage, oxygen vacancies migrate from MoS 2-x O x to the cathode and from the anode to MoS 2-x O x , leading to the reduction and oxidation of the cathode and anode, respectively. The forward current is less than the reverse current since the resistance of the anode (cathode) has increased (decreased). Under illumination, the TCAD simulated electron/hole current density of the anode and cathode demonstrate different values with opposite polarities, as shown in the bottom current density plots in Fig. 2f and g . The blue curve in Fig. 2e shows a short-circuit photocurrent of about 170 nA, which corresponds to the HPS of our G/M/G structure (Details in Table S2 ). To confirm the importance of the observed processes on graphene electrodes in our G/M/G devices for switching their photoresponse, we carried out control experiments on the Au/MoS 2−x O x (~100 nm)/Au (A/M/A) structure using the same deposition method. Such devices have not shown non-volatile photoresponsivity switching under illumination (Fig. S18 , Supplementary Information). These results show that the oxidation and reduction of graphene electrodes play an important role in the non-volatile photoresponsivity switching for our device. A computationally complete logical basis, using the fundamental elements of modern digital electronics and optoelectronics, can be formed by material implication (IMP) and FALSE operations (Table S3 , Supplementary Information) 32 , 48 . To implement the IMP operation, a set of three photomemristors p , q and s was developed with the corresponding connection scheme, as shown in Fig. 3a . Figure 3b shows the hysteresis curves of photomemristors for IMP operation. Device s is set to HPS and is illuminated during measurements with a light intensity of 6.4 mW cm -2 , resulting in no photoresponsive switching (Fig. 3b , black curve). The photoconductance of the photomemristor s is an order of magnitude lower than that of the photomemristor in an HPS and an order of magnitude higher than that of the photomemristor in an LPS. In this case, devices p and q are illuminated at a light intensity of 56 mW cm -2 , which leads to a clear switching of their photoresponse (Fig. 3b ). Fig. 3 “Photoresponse-stateful” strategy. a Illustration of an in-memory IMP operation based on a set of photomemristors triggered by light stimuli. b I-V characteristics of devices s , p and q under illumination with different light intensities. c Diagram of the optical and electrical pulses applied for the IMP operation. The blue and red curves show the optical and electrical signal of devices p and q before and during the IMP operation, and the orange curves show the change in photoresponse states after IMP operations, reproducing the IMP truth table. d Illustration of an in-memory FALSE operation triggered by light stimuli based on a set of photomemristors. e I-V characteristics of devices p and q under illumination. f Diagram of the optical and electrical pulses applied for the FALSE operation. The blue and red curves show the optical and electrical signal of devices p and q before and during the FALSE operation. The black curves show the change in photoresponse states after FALSE operations, reproducing the FALSE truth table Figure 3c demonstrates the photoresponse-stateful IMP operation. A positive voltage of 1.5 V (V SET ) applied for 6 s can switch devices from LPS (defined as 0) to HPS (define as 1), while a lower voltage of 0.8 V (V COND ) cannot change the photoresponse state. The IMP operation can be implemented by simultaneously applying light and electrical stimulation (V SET to q and V COND to p ) to achieve a conditional photoresponse switching, while reading can be performed by applying light stimuli. Under illumination, the unconditional operation of the set process of device q ( q ← 1) can be executed when V SET and V COND are separately applied to q and p , respectively. When the p photomemristor is in LPS ( p = 0), V COND has limited influence on the voltage across q , thus q is set while p remains unchanged. If the photomemristor p is in HPS ( p = 1), the voltage applied to q is below the threshold voltage at which q must remain without changing its photoresponse state. Figure 3d–f shows the scheme of photomemristors for implementing a FALSE operation (d) and I-V characteristics with a reset process for p and q (e). A negative voltage of –2.5 V (V RESET ) applied for 6 s can toggle the photoresponse state from HPS to LPS. However, the device does not show a negative photoresponse even a –2.5 V pulse is applied to the LPS device (Fig. S19 , Supplementary Information). Thus, the FALSE operation can be simply implemented by applying V RESET to p and q devices under illumination. The results and the corresponding truth table are shown in Fig. 3f . Thus, the IMP and FALSE operations which form the computationally complete logic are performed by G/M/G photomemristors, indicating a promising strategy for performing photoresponse-stateful logic operations triggered together by electrical and light stimuli. The parallel connection of two photomemristors with different polarities allows negative photoresponse states, which provide more freedom for neuromorphic computing functionality. Here we accommodate each device on a printed circuit board and connect them with wire and dial switches, setting a photoresponsive state individually and measuring a set of photomemristors assembled in parallel. Figure 4a shows the 7 distinguishable photoresponse states of our array of photomemristors (Fig. S20 , Supplementary Information). Using these 7 states, we emulate two types of neuromorphic vision functions: image pre-processing and classifier. Mimicking the neuromorphic vision preprocessing function of the human retina can speed up subsequent perception tasks and improve the image recognition rate 9 . These G/M/G photomemristors are combined into a 3 × 3 array that allows the simulation of the biological receptive field (RF) of the human retina controlled by different photoresponse states separately. Summing all photocurrents from each photomemristor of the emulated arrays performs the matrix-vector product operation: 1 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${{{\\boldsymbol{I}}}}_{{{{\\boldsymbol{m}}}},{{{\\boldsymbol{n}}}}} = \\mathop {\\sum}\\limits_{i,j}^{3,3} {{{{\\boldsymbol{R}}}}_{{{{\\boldsymbol{i}}}},{{{\\boldsymbol{j}}}}} \\times {{{\\boldsymbol{P}}}}_{{{{\\boldsymbol{i}}}},{{{\\boldsymbol{j}}}}}^{{{{\\boldsymbol{m}}}},{{{\\boldsymbol{n}}}}}}$$\\end{document} I m , n = ∑ i , j 3 , 3 R i , j × P i , j m , n where R i,j is the photoresponsivity matrix for various types of kernels and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${{{\\boldsymbol{P}}}}_{{{{\\boldsymbol{i}}}},{{{\\boldsymbol{j}}}}}^{{{{\\boldsymbol{m}}}},{{{\\boldsymbol{n}}}}}$$\\end{document} P i , j m , n is the vector of the optical signal of the input image, as shown in Fig. 4b and c , I m,n is the output vector which represents the dynamic current to the input signal. Various types of kernels for image pre-processing can be set by providing various states and polarities of the photomemristor array, allowing the image pre-processing function to be performed. With the working principle, we realized crucial functions that are widely used for image pre-processing with a photomemristor array consisting of a single photomemristor cell, such as the Gaussian blur. When we use photomemristors set as the base cell, as shown in Figs. S20 –S 23 (Supplementary Information), we can use more complex image pre-processing strategies with distinct kernels, such as the difference of the Gaussian, Prewitt, and Roberts operators. Figure 4d shows the logotype of the Chinese Academy of Sciences by computing a grayscale image using an emulated photomemristor array. Blurred and edge-enhanced images are similar to simulation results (Fig. S21 , Supplementary Information). Here, image pre-processing has been achieved without power consumption due to non-volatile photoresponse matrices and signal reading without external biases. Fig. 4 Image pre-processing and classifier based on G/M/G devices. a Photocurrent for different photoresponse states, in a different set of photomemristors. The inset shows schematically photomemristors installed with opposite polarity. b Schematic representation of the human visual system for sensing, memory, and computing. c Photomemristors with integrated sensing-memory-computing architecture. d Demonstration of image processing with various operators. These operations are realized by varying the different states and polarities of the photomemristors. e Schematic Illustration of the SLP photomemristors array for classifier emulation, all photomemristors with the same class (color) are connected in parallel to obtain the output current for the activation function. f SLP neural network architecture schematics. g Accuracy of the SLP classifier during training with floating-point weight and discrepant 7-level photoresponse state The working principle of the classifier is schematically shown in Fig. 4e . Each cell of the photodetector arrays consists of 5 photodetector sets corresponding to 5 classes (k = 0, 1, 2, 3, 4), summing all photocurrents from the cell with the same class from the emulated arrays performs the matrix-vector product operation: 2 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${{{\\boldsymbol{I}}}}_{{{\\boldsymbol{k}}}} = \\mathop {\\sum}\\limits_{i,j}^{m,n} {{{{\\boldsymbol{R}}}}_{{{{\\boldsymbol{i}}}},{{{\\boldsymbol{j}}}}}^{{{\\boldsymbol{k}}}} \\times {{{\\boldsymbol{P}}}}_{{{{\\boldsymbol{i}}}},{{{\\boldsymbol{j}}}}}}$$\\end{document} I k = ∑ i , j m , n R i , j k × P i , j where \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${{{\\boldsymbol{R}}}}_{{{{\\boldsymbol{i}}}},{{{\\boldsymbol{j}}}}}^{{{\\boldsymbol{k}}}}$$\\end{document} R i , j k is the photoresponsivity matrix for class k, P i,j is the vector of the optical signal of the input image, as shown in Fig. 4e and f , the output current I k is the input for the activation function. The network consists of a single-layer perceptron (SLP), together with SoftMax functions. This SLP represents a supervised learning algorithm that classifies the input images into 5 classes. The input layer of such SLP captures a 28 × 28-pixel image of 0, 1, 2, 3, 4 from the MNIST dataset, and the fully connected (FC) layer consists of 768 × 10 neurons. SLP is trained offline using 30596 training set images with a batch size 64 and 4000 iterations delivering the final output probability that classifies the input images in the test set (5139 images) into 5 classes with 97.66% accuracy. The weights in FC are discretized to accommodate the 7 photoresponse states. After discretization, the accuracy is about 96.44%, which is 1.22% lower than the pristine SLP. Further emulation of the 10 classes classifier based on our photomemristors demonstrates a 2% reduction in recognition rate compared to the pristine SLP, as shown in Fig. S24 (Supplementary Information). This indicates the potential applications of non-volatile in-sensory computing by using our photomemristors. Thus, here we have demonstrated a non-volatile responsivity matrix for simultaneous perception and processing of visual information using our two-terminal photomemristor without external power consumption. In summary, we have demonstrated tunable non-volatile photomemristors with a simple two-terminal G/M/G architecture, in which photoexcited carriers and ion migration are coupled leading to a displaced and pinched hysteresis of I-V characteristics. The devices can store and read multiple photoresponse states in a non-volatile mode at zero external voltage. Furthermore, the switching properties can be jointly controlled by the electric-field-driven migration of ions and photo-induced redox reactions at the asymmetric G/M/G contacts. By mimicking the biological functionalities of the human retina and designing specific device structures, the devices can act as a neural network for neuromorphic visual processing and implementation of completely photoresponse-stateful logic operations triggered by electrical and light stimuli together. This new concept of a two-terminal photomemristor not only provides versatile sensing-memory-computing approaches for neuromorphic vision hardware but also enables high-density integration."
} | 5,872 |
35642316 | PMC9185374 | pmc | 2,114 | {
"abstract": "Abstract Two main theories have been put forward to explain the origin of mitochondria in eukaryotes: phagotrophic engulfment (undigested food) and microbial symbiosis (physiological interactions). The two theories generate mutually exclusive predictions about the order in which mitochondria and phagocytosis arose. To discriminate the alternatives, we have employed ancestral state reconstructions (ASR) for phagocytosis as a trait, phagotrophy as a feeding habit, the presence of mitochondria, the presence of plastids, and the multinucleated organization across major eukaryotic lineages. To mitigate the bias introduced by assuming a particular eukaryotic phylogeny, we reconstructed the appearance of these traits across 1789 different rooted gene trees, each having species from opisthokonts, mycetozoa, hacrobia, excavate, archeplastida, and Stramenopiles, Alveolates and Rhizaria. The trees reflect conflicting relationships and different positions of the root. We employed a novel phylogenomic test that summarizes ASR across trees which reconstructs a last eukaryotic common ancestor that possessed mitochondria, was multinucleated, lacked plastids, and was non-phagotrophic as well as non-phagocytic. This indicates that both phagocytosis and phagotrophy arose subsequent to the origin of mitochondria, consistent with findings from comparative physiology. Furthermore, our ASRs uncovered multiple origins of phagocytosis and of phagotrophy across eukaryotes, indicating that, like wings in animals, these traits are useful but neither ancestral nor homologous across groups. The data indicate that mitochondria preceded the origin of phagocytosis, such that phagocytosis cannot have been the mechanism by which mitochondria were acquired.",
"conclusion": "Conclusions In the context of eukaryogenesis, our findings reject phagocytic models because the results indicate that the underlying premise of an ancestral phagocytic state for eukaryotes (in LECA) is unlikely to be true. Furthermore, our results indicate multiple independent origins of four of the five traits studied here, namely, plastids (including secondary plastids), the multinucleated state, phagocytosis, and phagotrophy. By contrast, mitochondria appeared with a clear single-origin in our analyses, tracing to LECA or prior. While recurrent acquisitions of photosynthetic plastids were previously described, multiple origins of multinucleate state, phagocytosis, and phagotrophy in eukaryotes are under-investigated issues. As such, our study here provides new insights into early eukaryote history and new methods for ASR that do not require the use of an agreed or accepted backbone species tree. All we require for this approach to work is codable information about traits, a sufficient number of genes present across members of the group in question, and taxonomic assignments regardless of phylogenetic relationship. Our results have implications for understanding the mechanism underlying the acquisition of mitochondria, a feature exclusive to eukaryotic cells. The broader significance of these findings is that the origin of mitochondria can be attributed to a fateful case of microbial symbiosis but cannot be attributed to a fateful case of indigestion.",
"introduction": "Introduction Phagocytosis is the process through which eukaryotic cells specifically recognize and engulf cell-sized particles (≥ 0.4 micrometer) via cytoskeleton-dependent invagination of the plasma membrane. Phagocytosis is a trait widely distributed among and exclusive to eukaryotes, serving as a strategy for internal digestion of food particles ( Martin et al. 2017 ; Mills 2020 ) as opposed to extracellular digestion via secreted enzymes. For as long as mitochondria have been discussed as endosymbionts, phagocytosis has been discussed in the context of mitochondrial origin. In her revitalization of the endosymbiotic theories of Mereschkowsky ( Kowallik and Martin 2021 , Martin and Kowallik 1999 ) and Wallin (1927) , Margulis, then named Sagan (1967) suggested in passing that phagocytosis was the mechanism by which the ancestral mitochondrion and its host became established. Cavalier-Smith proposed that phagocytosis directly gave rise to mitochondria and chloroplasts, but not via endosymbiosis, rather by origin of the organelles via restructuring of membranes in a cyanobacterial ancestor of eukaryotes ( Cavalier-Smith 1975 ). Many subsequent theories followed Margulis’s idea and emphasized phagocytosis as a mechanism for mitochondrial acquisition, hereafter collectively referred to as phagocytic models, and in most if not all such theories, the engulfed mitochondrial ancestor is interpreted as an undigested meal ( Doolittle 1998 ; Cavalier-Smith 2002 ; Roger et al. 2017 ; Poole and Gribaldo 2014 ). By contrast, a number of alternative theories for the origin of mitochondria do not entail a phagocytosing host, placing an emphasis on microbial interactions. Two kinds of microbial interactions are discussed: predatory bacteria and metabolic symbioses. The predatory bacteria class of theories posits mitochondria origin via predation by bacteria upon other bacteria. These theories lean on examples of predatory bacteria that enter the periplasm of their bacterial host, multiply there, and consume the hosts’s cytosolic content. Although initially proposed on the basis of Bdellovibrio predators from the deltaproteobacteria ( Guerrero et al. 1986 ), a number of alphaproteobacterial predators have been found and discussed in the context of mitochondrial origin ( Davidov et al., 2006 ; Davidov and Jurkevitch, 2009 ). In models involving predatory bacteria, the mitochondrion is seen not as an undigested meal but as an attenuated predator. Most current theories for mitochondrial origin involve metabolic symbioses among free living prokaryotes, though few take into account the low oxygen history of eukaryotic evolution, as recently reviewed by Mills et al. (2022) . Metabolic symbioses typically have a nutritional basis and often involve anaerobic syntrophy ( Schink 1997 ; Stams and Plugge 2009 ; Imachi et al. 2020 ) and hydrogen dependence ( Martin and Müller 1998 ; reviewed in Zimorski et al. 2014 ). Because phagotrophy is a feeding mechanism that supports day-to-day survival, its main function for cells is of physiological nature, involving the channeling of growth substrates from food vacuoles to mitochondria for ATP ( Martin et al. 2017 ). In non-phagocytic eukaryotes, such as fungi, digestive enzymes are secreted into the environment rather than into food vaculoes. Despite the popularity of the idea that phagocytosis was the key to eukaryote origin ( Cavalier Smith 1975 ; Embley and Williams 2015 ), physiological and cytological evidence suggests that the host was likely non-phagocytotic ( Gould et al. 2016 ; Martin et al. 2017 ) in line with fossil evidence indicating a late origin of phagocytosis ( Mills 2020 ). The main physiological evidence against the phagocytic orgin of mitochondria is 2-fold: (1) A mitochondrion-lacking phagotrophic archaeal host would have to ingest about 34 times its body weight in prokaryotic prey to obtain enough ATP to support one cell division at maximum energetic efficiency and (2) in contrast to all other archaea, it would lack ion gradients and chemiosmotic ATP synthesis at the plasma membrane, because phagocytosis and chemiosmotic ATP synthesis cannot coexist in the same membrane ( Martin et al. 2017 ). Furthermore, more recent observations show that the closest archaeal relatives to the host that acquired mitochondria are very small and simply organized archaeal cells ( Imachi et al. 2020 ), not phagocytotic proto-eukaryotes. Yet despite much evidence to the contrary ( Speijer, 2015 ), the phagocytic origin of mitochondria remains a very popular theory ( Dacks et al. 2016 ). The presence of mitochondria at the base of eukaryotic evolution ( Martin and Müller 1998 ; Embley and Martin 2006 ; Müller et al. 2012 ), combined with the lack of evolutionary intermediates, render the cell-morphological grade at the prokaryote-to-eukaryote transition steep and its phylogenetic reconstruction challenging. Most current theories agree that mitochondria and their related organelles—mitosomes and hydrogenosomes—descend from a proteobacterial symbiont that took up residence within its host ( Gray et al. 2001 ; Fan et al. 2020 ; Betts et al. 2018 ), whereby the host was a member of an ancient archaeal lineage ( Williams et al. 2013 ; Martin et al. 2015 ). A more debated issue concerns the timing of mitochondrial acquisition relative to the emergence of other eukaryotic traits, cell complexity in particular ( Lane and Martin 2010 ). In the context of the present study, if mitochondria were acquired via phagocytosis, then the host had already evolved a phagocytic lifestyle, meaning that large cell size, the endomembrane system, vesicle flux and cytoskeleton—the salient components of eukaryote cell complexity—had already arisen prior to mitochondrial acquisition ( Poole and Gribaldo 2014 ; Roger et al. 2017 ; Cavalier-Smith 2002 ; De Duve 2007 ). That is, according to phagocytic theories, eukaryotic complexity arose independent of mitochondrial functions or mitochondrial genes. However, no modern-day archaea grown in laboratory cultures or observed in nature are known to phagocytose. In current formulations, phagocytic models rely on inferences from metagenome-assembled genomes (MAGs) from uncultured asgard archaea, which are reported to encode homologs of phagocytosis-related genes in eukaryotes ( Zaremba-Niedzwiedzka et al. 2017 ; Spang et al. 2015 ), such as actin and tubulins. However, the purity of these MAGs has been questioned ( Garg et al. 2021 ), and the few phagocytosis-related genes found in asgard MAGs are arguably insufficient to confer full phagocytic capability as observed in eukaryotes today. This has been demonstrated with enriched cultures of Candidatus Prometheoarchaeum syntrophicum MK-D1 the only asgard archaeon that has been cultivated in the laboratory to date. MK-D1, the closest archaeon to the eukaryotic host, showed no evidence of phagocytic ability under the microscope, although it was able to generate membrane protrusions ( Imachi et al. 2020 ) which are feeding appendages that increase surface area for its fermentative lifestyle, similar to the function of hyphae in filamentous fungi ( Scannell et al. 2006 ). Alternatives to phagocytic models for the origin of mitochondria, the symbiotic models, have it that the archaeal host was not phagocytotic and that the mitochondrial ancestor established a symbiotic relationship living in close physical contact with its archaeal host ( Martin et al. 2001 , 2015 ). Over the course of time, the symbiosis of prokaryotes stabilized, the host became strictly dependent upon its symbiont (anaerobic syntrophy), leading to entry of the bacterial symbiont into the host’s cytosol (endosymbiosis). Several examples of prokaryotes that have taken up symbiotic relationships within the cytosol of another—nonphagocytotic—prokaryote are known ( Martin et al. 2017 ). In addition, modern-day archaea can undergo membrane fusions and cell fusions ( Naor and Gophna 2013 ), such that symbiogenic models do not require an origin of phagocytosis within archaea prior to mitochondrial origin. At face value, both phagocytotic and symbiogenic theories would predict the origin of the eukaryotic plasma membrane to be of archaeal origin, but the eukaryotic outer membrane is chemically more similar to that of bacteria. To account for this, symbiotic models have a corollary in which the lipids of the eukaryotic plasma membrane arose via secretion of membrane vesicles by the bacterial endosymbiont, which ultimately replaced the original host outer membrane ( Gould et al. 2016 ). Both phagocytic and symbiotic models for the origin of mitochondria are currently discussed and debated, whereby the role of environmental oxygen levels roughly 1% that of current oxygen levels during eukaryotic and mitochondrial origin as well as during the first billion years of eukaryotic evolution bear heavily upon these issues. For a balanced and comprehensive review, see Mills et al. (2022) . Discrimination between the theories requires more data and analyses, not more debate. One largely unexplored issue concerns the premise underlying phagocytic theories, namely that phagocytosis evolved prior to mitochondrial acquisition and hence was present in the last eukaryotic ancestor (LECA). An earlier study focused on the identification of phagocytosis-related genes in eukaryotic genomes followed by reconstruction of phylogenetic trees and used the gene trees as proxies to speculate about the origin of phagocytosis as a process ( Yutin et al. 2009 ). However, due to the multiplicity of functions a gene can have, identifying phagocytosis-related genes can lead to many false positives ( Gotthardt et al. 2006 ; Okada et al. 2006 ; Marion et al. 2005 ; Jacobs et al. 2006 ). Furthermore, genes that precipitated phagocytosis may have been lost or replaced, and eukaryotic genes that are currently known to be involved in phagocytosis may have originated prior to phagocytosis. Hence, inferences indicating that phagocytosis-related genes originated in LECA cannot be readily equated to an early-origin of phagocytosis. For example, both archaea and bacteria are known to possess tubulin homologues ( Erickson et al. 2010 ), but neither archaea nor bacteria are phagocytotic. Here, we address the origin of phagocytosis in eukaryotes within the framework of ancestral state reconstruction (ASR) analyses. By examining the presence of phagocytosis as a process, rather than the presence of a few phagocytosis-related genes, across a diverse sample of eukaryotic species we readdress the phagocytosis-origin problem from a novel empirical perspective. We specifically examine the timing of phagocytosis and phagotrophy in eukaryote evolution, in addition to the antiquity of the multinucleated (syncytial) state ( Skejo et al. 2021 ) and, as controls, the origin of mitochondria and plastids.",
"discussion": "Results and Discussion Framework and Data Our dataset consists of five eukaryotic traits—mitochondria, phagocytosis (the ability to engulf bacterial cells), phagotrophy (phagocytosis as a feeding habit), multinucleate organization, and plastids, as well as the distribution of these traits across 150 eukaryotic species that span six lineages: Opisthokonta, Archaeplastida, Hacrobia, Excavata, Stramenopiles, Alveolates and Rhizaria (SAR), and Mycetozoa ( fig. 1 ; see Materials and Methods for details). To evaluate the potential contribution of each of the traits to eukaryogenesis, we first set out to time their origin relative to the last eukaryotic common ancestor (LECA) using ASR. To perform ASR, two types of data are required: a table with the distribution of traits in some species and a phylogenetic tree upon which ASR is calculated for all internal nodes in the tree. Typically, the tree used for ASR is a species tree which is commonly reconstructed from sequences of single-copy genes common to all species under scope (universal orthologs). By clustering 1 848 936 protein-coding genes from the 150 eukaryotic genomes using a markov clustering algorithm (MCL) ( Enright et al. 2002 ), we obtained 239 012 gene families in total. Of the total, 313 gene families are present in at least in 140 genomes, 130 gene families are present in at least 145 genomes, and 15 gene families are present in 149 genomes. However, no gene family in our data is strictly universal, that is, with gene-copies present in all 150 eukaryotic genomes, because our species set includes species with highly reduced genomes including the parasite Giardia lambia and the unicellular photosynthetic species Nannochloropsis gaditana ( supplemental table 1, Supplementary Material online). Reconstructing a reliable species tree without universal orthologs is a challenging task, which is further complicated by the abundance of paralogues in the present data ( Tria et al. 2021 ), as a result of frequent gene duplications in eukaryotic evolution. Fig. 1. Presence (filled circle) absence (empty circle) distribution of five traits in 150 eukaryotic species. Species with no circle for a given trait indicate missing annotation. The reference tree was inferred from the alignment of 18S RNA sequences, rooted on the Excavates branch, with the sole purpose of data display (see Materials and Methods). Tip labels are species codes (see supplemental table 1, Supplementary Material online, for complete species names and detailed trait annotations). The first character of the species codes indicates supergroup affiliation of the species: Excavates ( E ), Mycetozoa ( M ), Hacrobia ( H ), Archaeplastida ( A ), SAR ( S ) and Opisthokonta ( O ). The shades of gray show the clades of the six eukaryotic supergroups. To harness phylogenetic information contained in genomes and bypass the reliance on a backbone species tree, we used 1789 gene families to reconstruct maximum-likelihood trees for each individual gene family. The 1789 gene families are distributed variably across eukaryotic genomes (mean = 105.3, median = 114, SD = 34.8) but are present in least one representative species from the six eukaryotic supergroups. The presence of the genes in six supergroups indicates that these gene families likely trace to LECA or prior to it. We then used each resulting gene tree to perform an independent ASR experiment, under the principle that each gene family is an independent data-sample. Gene trees are informative for ASR as long as the underlying gene families originated in LECA and not after it. Yet, the accuracy of the ASR may vary across trees due to tree errors and sampling effect generated by gene duplication and gene loss. We will address these issues along the paper. Notably, the gene trees used here served only as phylogenetic markers. We neither assumed nor expected the functions of the genes to be either directly or indirectly involved in the establishment of the eukaryotic traits investigated. LECA Had Mitochondria and Was Multinucleated, But It Was Neither Phagocytic nor Phagotrophic For each tree and trait, we labeled the species at the tips of the tree according to their trait-state annotations ( supplemental table 1, Supplementary Material online) and performed maximum-likelihood ASR (see Materials and Methods for details). In each tree, we could identify the trait-state (for example, presence or absence) that traced to LECA. A tree was only used for ASR of a given trait if the tree contained representative species for at least two trait-states. Trees displaying only one state for a given trait (for example, all taxa having mitochondria) were uninformative and not considered for ASR of that trait. A maximum-likelihood ASR yields probabilities for each possible trait-state at the root of the tree, where the result may be resolved or ambiguous when alternative trait-states are tied with equal probabilities. Because each tree spans all major eukaryotic lineages, its root corresponds to LECA. One way of summarizing ASR across trees is by counting the frequency in which each trait-state appeared in LECA across trees (the majority-rule). A trait-state occurring in LECA at a high frequency across trees likely reflects the true state in LECA, whereas trait-states occurring in low frequencies in LECA are the result of lineage specific origins for the trait or errors. It is important to note that the majority-rule method does not utilize trees with unresolved trait-states in LECA, and the magnitude of the difference in probabilities for alternative trait-states is not considered at all. Using the majority-rule, we found that the ASRs traced the presence of canonical mitochondria to LECA in 90% of the trees, recovering the (now) well-accepted notion of mitochondria being present in the LECA ( table 1 ) as posited by most current theories that address the origin of mitochondria. Alone, the presence of mitochondria in LECA has no weight in distinguishing current alternative theories for the origin of mitochondria in eukaryotes, because all current theories have mitochondria in LECA—a radical change from 20 years ago ( Martin et al. 2001 )—but it serves as a first validation of our approach. Another validation was obtained with the analyses of photosynthetic plastids, a trait uncontestably thought to have originated after the eukaryotic supergroups diverged, at the base of photosynthetic lineages (Archaeplastida). Our analyses indicated a late origin of plastids relative to LECA in 78% of the trees, in accordance to the expectation. The ASR placed the origin of photosynthetic plastids in LECA in only 6% of trees, with the remaining 16% trees having unresolved ASR. The 10% of trees that trace the origin of mitochondria after LECA and the 6% trees that traced plastids into LECA are clear deviations from the expected results, indicating a roughly 10% error rate underlying the majority-rule analyses. Our ASRs also show LECA as a multinucleate (syncytial) organism in 69% of the trees, in accordance with an independent study ( Skejo et al. 2021 ). Multinucleate species and stages, in which different nuclei divide independently both of each other and of cell division, are surprisingly common among eukaryotes ( Skejo et al. 2021 ). The results we obtained for mitochondria, plastids and the multinucleated state are in accordance with commonly accepted notions of eukaryotic trait evolution, serving as an internal control and validation for our analyses. Table 1 Maximum-likelihood ancestral reconstruction of five traits from 150 eukaryotic species, across a broad sample of gene trees as estimates of the underlying phylogeny. Absolute values indicate the number of trees with a trait state (presence/absence) tracing to LECA. The total number of trees used ( N ) as well as the number of trees with ambiguous reconstructions in LECA are indicated. Multinucleate, phagocytosis, and phagotrophy were modeled as binary traits, while mitochondria and plastids were modeled as traits with three states each (see supplemental table 1, Supplementary Material online and Materials and Methods for details). For mitochondria, “presence” indicates that canonical mitochondrion is the reconstructed ancestral state, while “absence” indicates that the reconstruction is either mitosome or hydrogenosome. Single-copy gene trees \n Trait \n Presence Absence Ambiguous Total ( N ) \n mitochondria \n 8 (100%) 0 0 8 \n plastid \n 1 (5%) 7 (33%) 13 (62%) 21 \n multinucleate \n 14 (67%) 0 7 (33%) 21 \n phagocytosis \n 1 (5%) 7 (33%) 13 (62%) 21 \n phagotrophy \n 1 (5%) 10 (48%) 10 (48%) 21 \n Multi-copy gene trees \n \n \n Trait \n \n Presence Absence Ambiguous Total ( N ) \n mitochondria \n 1191 (90%) 4 (0,3%) 123 (9%) 1318 \n plastid \n 106 (6%) 1372 (78%) 290 (16%) 1768 \n multinucleate \n 1234 (70%) 162 (9%) 372 (21%) 1768 \n phagocytosis \n 475 (27%) 779 (44%) 514 (29%) 1768 \n phagotrophy \n 323 (18%) 963 (54%) 482 (27%) 1768 The most relevant traits for investigating the question of how mitochondria entered the eukaryotic lineage are phagocytosis—the process of engulfing cells, like macrophages—and phagotrophy—engulfing cells as a feeding habit as opposed to osmotrophy, whereby enzymes are excreted outside the cell to digest and uptake digestion products. We analyzed each trait independently. Phagocytosis was defined as species harboring cells with the ability to actively internalize particles larger than 400 nm (the size of a small bacterium), whereas phagotrophy was defined as the special case of using phagocytosis as a feeding habit. For example, humans are phagocytic because of macrophage activity during infection but not phagotrophic, because we digest food in the intestine and uptake breakdown products via plasma membrane importers. Despite a wide distribution of phagocytosis and a moderate distribution of phagotrophy in the 150 eukaryotic species in our dataset ( fig. 1 ), the majority-rule across trees indicates that LECA was neither phagotrophic nor phagocytic. That is, the origin of phagocytosis was reconstructed after LECA in 44% of the trees, in LECA in 27% of the trees, whereas 29% trees had unresolved ASRs ( table 1 ). Phagotrophy, the trait that phagotrophic models for the origin of mitochondria require, appeared in LECA in only 18% of the trees, with 54% of trees placing the origin of phagotrophy after LECA. For 28% of the trees, the ASRs of phagotrophy were unresolved. The analyses for phagocytosis and phagotrophy yielded a higher proportion of unresolved ASRs in comparison to the traits mitochondria, multinucleate organization and plastids ( table 1 ). To assess the statistical significance of our results, we performed a test by matching the probabilities of the alternative trait-states for each tree regardless of outcome in LECA (trait-presence, trait-absence or tie) and assessed the differences in distributions using the Wilcoxon signed-rank test ( fig. 2 ). The test can be seen as a refinement of the majority-rule as it considers the magnitude of probabilities for all possible trait-states in LECA, which are directly obtained from the ASR, and integrate information from trees with unresolved ASR in LECA. The results of the two-tailed Wilcoxon tests indicate that the traits phagotrophy and phagocytosis were not present in LECA at P < 0.01. Fig. 2. Distribution of marginal probabilities for alternative trait-states in LECA across single-copy gene trees (left; without paralogs) and multi-copy gene trees (right; with paralogs). Multinucleate, phagocytosis and phagotrophy were treated as binary traits, while plastids and mitochondria were treated as traits with three states each. For plastids the states were: absence, primary plastid or secondary plastid. For mitochondriam the states were as follows: mitosome, hydrogenosome, or canonical mitochondria (see Methods and supplemental table 1, Supplementary Material online for details). The number of trees used in the analyses are show in table 1 . Trait-states with high probabilities in the trees have distributions (colored lines) that are right-shifted in the plots. GTPases, tubulins, and actins are common among eukaryotes and play key roles in phagocytosis ( Rougerie et al. 2013 ; Hall 2012 ; Lancaster et al. 2018 ); the likely presence of these genes in LECA has been interpreted as evidence for an early origin of phagocytosis relative to mitochondria. However, the origin of phagocytosis-related genes is not guaranteed to coincide with the origin of phagocytosis because the genes that precipitated the origin of phagocytosis may have been lost or replaced over the course of 1.5 billion years of evolution since eukaryotes emerged ( Betts et al. 2018 ). Contrary to phagotrophic theories for the origin of mitochondria, but in line with some earlier views ( Martin et al. 2003 ), our results show that LECA was neither phagotrophic nor phagocytic, obviating the requirement of these traits for the origin of mitochondria in eukaryotes. A previous study based on the comparative analyses of gene expression data for phagocytic-related genes ( Yutin et al. 2009 ) also suggested a late origin of phagocytosis, as did a study of microfossil evidence for the late origin of phagocytosis ( Mills 2020 ). Tree Quality, Sampling, and Conflicting Evidence in Phylogenomic Analyses The accuracy of ASR depends on the quality of the individual gene trees. Because of gene duplications and gene losses, topological discordance cannot be equated to the ever-present problem of tree reconstruction errors. Tree reconstruction strongly depends on the quality of the sequence alignments, which can be assessed using the heads or tails (HoT) analyses ( Landan and Graur, 2007 ). We investigated the grade of HoT scores across the 1789 trees by comparing the positional consistency of the original alignments (heads) against the alignments obtained from the sequences in their reversed amino-acid order (tails). Higher HoT values indicate higher positional consistency between the original and reversed alignments, which is indicative of well-aligned sequences. The distribution of HoT scores for all the 1789 gene trees, grouping trees according to trait-state outcome in LECA, for each trait separately, are shown in supplemental fig. 1, Supplementary Material online. The HoT scores indicate little difference between forward and reverse alignments. We found that the overall tree quality is high, with the majority of trees having scores above 0.6 according to the mean column score (MCS), which indicates the proportion of identically aligned site columns, and above 0.9 for the mean residue pair score (MRPS, identically aligned pairwise site comparisons). Furthermore, the distributions of alignment scores underlying trees that recovered different trait-states in LECA had no clear difference, suggesting that tree reconstruction errors are unlikely to explain different ASR results. Alignment quality does not impact our current results because the Wilcoxon-tests, using only the top 200 trees according to HoT scores, recovered the same ASR for all five traits ( supplemental fig. 3, Supplementary Material online; fig. 2 ). Another factor that may influence ASR is the position of the root within the trees. We used the minimal ancestor deviation (MAD) approach to root the trees ( Tria et al., 2017 ), which outperformed alternative approaches in independent studies ( Wade et al. 2020 ; Lamarca et al. 2022 ) and has the advantage of not requiring outgroups. Yet, MAD rooting is expected to fail for trees with high levels of molecular-clock departure, which may vary across trees. Indeed, we found microsporidians, a highly specialized group of fungal pathogens with highly relaxed functional constraints (high rates) for many genes, at the base of 10% of our gene trees, which is indicative of errors due to long branch attraction ( Brinkmann et al. 2005 ). To account for the effect of the quality of inferences, we analyzed the distribution of two root scores calculated by MAD: the ancestor deviation (AD) statistic for the inferred root position, which measures the degree of deviation from the molecular-clock associated to the inferred root, and the root ambiguity index (AI), defined as the ratio of AD scores for the inferred root over the second-best root. We found the distribution of AD and AI to be remarkably similar for trees that obtained a different trait-state in LECA ( supplemental fig. 1, Supplementary Material online), suggesting that no significant bias of ASR was caused by variable levels of root inference accuracy. Furthermore, by repeating the Wilcoxon tests with the top 200 trees with best root quality, as judged independently for AD and AI, we recovered the same ASR as obtained with all trees in the sample ( supplemental fig. 3, Supplementary Material online). It is noteworthy that the results of our ASR analyses depend upon the eukaryotic species sampled, which were limited to the species with genomic sequences in RefSeq ( O'Leary et al. 2016 ). We deliberately avoided the inclusion of metagenomic and transcriptomic sequences, because they are notoriously more susceptible to contamination (false taxon label), base-calling, and assembly errors, which bias phylogenetic reconstructions ( Garg et al. 2021 ). Nevertheless, sampling is an important factor in ASR analyses. Since the gene families we used to reconstruct the trees are not uniformly distributed across the eukaryotic genomes sampled here, we could investigate the effect of differential sampling upon our results, using the natural distribution of the genes as reference. We analyzed four sampling parameters calculated for each tree: (1) the fraction of the least frequent trait-state occurring at the tips of the trees; (2) the fraction basal lineages measured as the number of Excavates and Mycetozoa relative to Opisthokonts; (3) the total number of species; and (4) the total number of OTUs (operational taxonomic units). For each of the four sampling parameters, we ranked the trees in decreasing order, selected the top 200 trees, and repeated the Wilcoxon tests ( supplemental fig. 3, Supplementary Material online). With only one exception, these tree subsamples corroborated the results shown in fig. 1 , albeit with variable P -values due to decreased sample size. The only exception occurred for the subsample of trees with highest fraction of basal lineages, where the ASR for phagocytosis in LECA could not be resolved ( P -value > 0.05). To find out which species were enriched in the subsample of 200 trees with enriched basal lineages, we calculated the frequency of appearance for each species across the tree subsample and compared to that of the entire tree sample. We found that the three microsporidia species present in our genome set and one SAR species had the highest degree of sampling improvement, when comparing how frequently these species appeared in the tree subsample relative how frequent they appeared in all trees ( supplemental table 2, Supplementary Material online). By restricting the analyses to trees with high sampling of basal species resulted in a subsample of trees that are also rich in species with highly reduced genomes. This is a noteworthy result because the Microsporidia and SAR species, enriched in the subsample of 200 trees, are fast-evolving lineages known to introduce bias in phylogenetic analyses ( Brinkmann et al. 2005 ). Unrestricted species sampling, although theoretically desirable to cover grades of biological diversity, can hinder phylogenetic analyses by increasing heterogeneity in the data. Indeed, we found a significant negative correlation of HoT scores with the total number of species in the sequence alignments underlying the trees (rho = −0.4, P < 0.01, two-tailed Spearman-rank correlation). As in all molecular phylogenetic studies, there is conflicting evidence in the form of conflicting signals in the present data. Conflicting signals can arise as a result of fragmented or contaminated data and therefore lead to falsely constructed clades in tree topologies ( Wägele et al. 2009 ), which we avoided by excluding metagenomic and transcriptomic data. An important source of conflict in eukaryotic gene families is gene duplication and the presence of paralogs. An earlier independent study found that at least 475 genes were duplicated in LECA ( Tria et al. 2021 ). Although these duplications complicate the analysis of eukaryotic phylogenies, it is important to keep in mind that duplications are the hallmark of eukaryotic genes such that phylogeny-based analyses of eukaryote evolution have to take this into account. Eliminating genes with duplications or paralogs would eliminate almost all gene families from this or any other study of eukaryote gene or genome evolution, as nearly half of all eukaryotic protein-coding genes exist as multiple copies in at least one genome ( Tria et al. 2021 ). The “manual” removal of paralogs from individual trees would also introduce biases of effectively arbitrary nature. Nonetheless, we could rule out paralogues as a potential bias because independent analyses of multi-copy and single-copy trees rendered the same ASR for the five eukaryotic traits we analyzed ( table 1 and fig. 2 ). Whether or not paralogues actually hinder phylogenetic reconstructions is still unanswered, possibly case-dependent, and our analyses will motivate further investigations. Phagocytosis and Phagotrophy Evolved Multiple Times within the Eukaryotic Lineage A late origin of phagocytosis and phagotrophy, together with the wide distribution of these traits across eukaryotic species, raises the question of how many times these traits evolved within eukaryotes. One possibility is that phagocytosis and phagotrophy evolved only once prior to the divergence of some eukaryotic supergroups or multiple independent times within supergroups. To test the multiple origin hypothesis, we counted for each tree the number of trait-origins. The average number of trait-origins across trees is shown in table 2 , for each of the five traits investigated here. Only mitochondria showed up as a clear single origin trait, with an average of one origin per tree. By counting the number of origins for plastids, regardless of its type (that is, primary or secondary), rendered an average of four to six origins which is in line with one primary acquisition of plastids in the Archaeplastida ancestor followed by subsequent acquisitions via secondary and tertiary plastids in Hacrobia and SAR ( Gould et al. 2015 ). Table 2 Summary statistics for the number of trait origins across trees (see note below the table). Trait origin in internal and terminal nodes are distinguished. Single-copy trees (without paralogs) were distinguished from multi-copy trees (with paralogs) Single-copy gene trees \n n. origin \n \n a \n trait Terminal nodes Internal nodes All nodes \n mitochondria \n 0, 0, 0 1, 1, 0 1, 1, 0 \n plastid \n 3, 3, 1.6 1.1, 1, 0.8 4.1, 4, 1.4 \n multinucleate \n 2, 1, 2.7 1, 1, 0.6 3, 2, 2.8 \n phagocytosis \n 4.4, 4, 1.8 0.7, 1, 0.6 5.1, 5, 1.65 \n phagotrophy \n 4.1, 4, 1.7 0.6, 1, 0.6 4.7, 4, 1.4 \n Multi-copy gene trees \n \n \n n. origin \n \n \n a \n trait \n Terminal nodes \n \n Internal nodes \n \n All nodes \n \n mitochondria \n 0, 0, 0.2 1, 1, 0.2 1, 1, 0.4 \n plastid \n 2.9, 3, 1 2.9, 3, 1.3 5.8, 6, 2.4 \n multinucleate \n 3.5, 1, 4.6 3.6, 3, 2.9 7.1, 4, 7.1 \n phagocytosis \n 4.6, 4, 3.3 2.5, 2, 1.5 7.1, 7, 4.3 \n phagotrophy \n 5.8, 6, 3.3 2, 2, 1.3 7.8, 8, 3.9 \n a \n \n Note : Numbers indicate mean, median, and standard deviation across trees. Our analyses show that even though LECA was multinucleated, the trait had on average three to seven origins across trees, indicating a high turnover rate (loss with reappearance) for this trait in eukaryote evolution. Instances of multiple origins for the multinucleate state may reflect the selective trade-offs imposed by the co-existence of multiple nuclei within the same cell. It has been suggested that the existence of multiple nuclei in LECA permitted mutations, chromosomal rearrangements, and aneuploidies to occur freely during chromosomal segregation, because the eventual loss of gene function in one nucleus, arising from defective mutations, can be compensated by the proper functioning of the same gene in another nucleus ( Garg and Martin 2016 ; Skejo et al. 2021 ). While stable environmental conditions may favor individuals with few nuclei per cell, the multinucleate state offers important adaptive capacity for populations inhabiting rapidly changing environments. In that sense, the multinucleated state is a special case of polyploidy, which can postpone the effects of Muller’s ratchet in asexually reproducing eukaryotes ( Kondrashov 1994 ), which LECA was at some point during the transition from a symbiosis of prokaryotes to a nucleated cell with mitochondria. Phagocytosis originated as a trait two to five times on average in the trees. Even though some key genes for these processes were already present in LECA such as GTPases, tubulins, and actins, which also exist in prokaryotes ( Shih and Rothfield 2006 ; Verstraeten et al. 2011 ; Fletcher and Mullins 2010 ), the presence of these genes alone does not imply in the capacity to perform phagocytosis. The multiple independent origins of phagocytosis supported by our data align very well with previous observations that phagocytosis-related genes are rarely shared among distantly related eukaryotes ( Yutin et al. 2009 ). Gene expression analyses have shown thousands of genes being differentially expressed during phagocytosis ( Gotthardt et al. 2006 ; Okada et al. 2006 ; Marion et al. 2005 ; Jacobs et al. 2006 ). Among these, only about a dozen are common to phagocytic eukaryotic genomes, with the vast majority of differentially expressed genes being supergroup exclusive ( Yutin et al. 2009 ). Overall, both ASR and comparative genome analyses point to multiple origins of phagocytosis in eukaryotes. One important implication of our finding is that phagocytosis, as a process, is not homologous among eukaryotic species capable of phagocytosis. Hence, comparative analyses targeting a better understanding of phagocytosis as a process need to take process homology among species, or lack thereof, into account. One possibility is to restrict comparative genome analyses to species suspected to share phagocytic homology, which might assist the identification of currently unknown phagocytic-related genes. In a broader context, assessing trait homology using ASR as done here has the potential to improve studies aimed towards a better understanding of trait evolution across the tree of life. It also allows us to address the relative order of appearance of the eukaryotic traits investigated here, as outlined in the following. Timing the Origin of Eukaryotic Traits Relative to the Emergence of Eukaryotic Supergroups To time the origin of traits relative to the divergence of six well-known eukaryotic supergroups considered here, we identified the eukaryotic species that descend from the origin node and recorded the corresponding supergroup affiliation of descending species. We repeated this process for each trait and trait-origin independently using all origin nodes as inferred by the ASR, across all trees, and plotted the distribution supergroups descending from the origin nodes ( fig. 3 ). For each reconstructed origin, all the species (tips) descending from it in the tree were used to score an origin as a combination of supergroups so identified. In this way, we were able to estimate the approximate origin of the traits relative to the supergroups without committing to any particular eukaryotic supergroup phylogeny, which is a recognized challenge and hotly debated topic ( Burki et al. 2020 ). Furthermore, the possibility that some of the supergroups used here might not be monophyletic has no influence on our results because the species were allowed to assume any relationship in the trees, without topological constraints. The supergroups only serve the purpose of displaying the results, as higher order leaf labels regardless of underlying backbone species tree, and some traits map well to supergroup assignments used here. As it concerns the traits that originated in LECA, we only considered the ASRs that placed an origin at the root of the trees (red circles in fig. 3 ). Fig. 3. Distribution of supergroups descending from origin nodes across 1789 trees. For each internal node reconstructed as a trait origin, all the species (tips) descending from it were used to score an origin to the combination of supergroups (filled circles) to which the descending species belong. Origins at the root node (LECA) are shown in red. We distinguish origins that occurred after the root node, for which the descending species represent all six supergroups (black circles) which could also be indicative of trait origin at LECA but with some level of uncertainty since they could alternatively be the result of phylogenetic errors. The combination of supergroups with high frequency of origins across trees are likely to coincide with a true trait-origin in the underlying supergroup phylogeny, while low-frequency supergroup combinations are more likely spurious results. For mitochondria the result is very clear, for 1326 gene trees a high number of origin nodes ( n = 1199) occurred in LECA. For plastids, the highest number of origins occurred in an Archaeplastida ancestor ( n = 1705) for 1789 gene trees, followed closely by the number of origins in SAR ( n = 1240). A moderate number of plastids origins was also observed in the SAR + Hacrobia ancestor ( n = 282) and the Hacrobia exclusive ancestor ( n = 200). The multinucleate trait had the highest number of origins in LECA ( n = 1234) for 1789 gene trees, albeit a high to moderate number of origins was also observed in the ancestor of each supergroup ( fig. 3 ), indicating presence in LECA in addition to multiple lineage specific (secondary) origins for the multinucleate form. That is, the multinucleate state was likely lost several times subsequent to LECA’s divergence but recurrently reemerged within each supergroup. For phagocytosis, the highest number of origins occurred in Opisthokonta ( n = 891) for 1789 gene trees, followed by Excavata ( n = 641) and Mycetozoa ( n = 620). The natural diversity of the processes usually classified as phagocytosis across eukaryotic supergroups, together with our results, indicates that the phagocytic processes evolved independently in Opisthokonta, Mycetozoa, and Excavata. For phagotrophy, the highest numbers of origins for 1789 gene trees were found within three supergroups: Mycetozoa ( n = 805), Excavata ( n = 793), and SAR ( n = 528). For clarity, 805 origins of phagocytosis refer to the sum of origins scored across 1789 separate trees having on average 105 species, each tree containing representatives from all six eukaryotic supergroups sampled here."
} | 11,304 |
25309418 | PMC4160969 | pmc | 2,117 | {
"abstract": "To understand how the central nervous system performs computations using recurrent neuronal circuitry, simulations have become an indispensable tool for theoretical neuroscience. To study neuronal circuits and their ability to self-organize, increasing attention has been directed toward synaptic plasticity. In particular spike-timing-dependent plasticity (STDP) creates specific demands for simulations of spiking neural networks. On the one hand a high temporal resolution is required to capture the millisecond timescale of typical STDP windows. On the other hand network simulations have to evolve over hours up to days, to capture the timescale of long-term plasticity. To do this efficiently, fast simulation speed is the crucial ingredient rather than large neuron numbers. Using different medium-sized network models consisting of several thousands of neurons and off-the-shelf hardware, we compare the simulation speed of the simulators: Brian, NEST and Neuron as well as our own simulator Auryn. Our results show that real-time simulations of different plastic network models are possible in parallel simulations in which numerical precision is not a primary concern. Even so, the speed-up margin of parallelism is limited and boosting simulation speeds beyond one tenth of real-time is difficult. By profiling simulation code we show that the run times of typical plastic network simulations encounter a hard boundary. This limit is partly due to latencies in the inter-process communications and thus cannot be overcome by increased parallelism. Overall, these results show that to study plasticity in medium-sized spiking neural networks, adequate simulation tools are readily available which run efficiently on small clusters. However, to run simulations substantially faster than real-time, special hardware is a prerequisite.",
"introduction": "1. Introduction Neurons communicate with each other by short electrical pulses, called action potentials or spikes, which can be considered as unitary events. In simple neuron models of integrate-and-fire type, such events are generated by a threshold crossing process. The dynamics of a single neuron, which forms one unit of a large brain network, are therefore relatively simple. Nevertheless, the simulation of activity in large neural networks, which has been receiving increasing interest over the past years (Markram, 2006 ; Ananthanarayanan et al., 2009 ; Lang et al., 2011 ; Koch and Reid, 2012 ; Waldrop, 2012 ; Kandel et al., 2013 ), poses multiple computational challenges. First, brain networks consist of billions of neurons (Kandel et al., 2000 ). Even if each neuron is described as a relatively simple dynamic processing unit (e.g., an adaptive integrate-and-fire neuron with two or three update equations per neuron Izhikevich, 2003 ; Brette and Gerstner, 2005 ; Gerstner et al., 2014 ), the sheer number of units suggests that faster than real-time simulation of these equations will be hard to achieve on a single core. Hence parallelization of computation is desirable. Second, each unit sends and receives signals from thousands of others (DeFelipe and Fariñas, 1992 ; Kandel et al., 2000 ), such that connectivity between units is relatively high compared to classical models in the physical sciences where interactions are mainly between nearest neighbors in physical space (Anderson, 1995 ). Therefore, the communication overhead in a parallel implementation could potentially be high. Third, the synaptic contact points between two connected units are not fixed but may change (Bliss and Lømo, 1973 ; Markram et al., 1997 ; Bi and Poo, 1998 ; Zhang et al., 1998 ; Bi and Poo, 2001 ; Markram et al., 2012 ). Consequently, connections cannot be described with fixed parameters, but need further dynamic variables. Moreover, the evolution of these synaptic variables depends on activity of both the sending and the receiving neuron so that their treatment requires additional care and readily available parallelization approaches cannot be used. The changes in the dynamic values associated with the synaptic contact points are referred to as synaptic plasticity. The question therefore arises whether the scaling of parallel implementations of simulated neural networks is dominated mainly by the inter-process communication or by the dynamics of the connections. This question cannot be answered in a straightforward manner, because it depends on multiple factors. First, the communication between neurons only takes place at the moment when a spike happens, leading to event-based updating schemes (Morrison et al., 2007 , 2008 ). Accordingly, the number of events per unit of time plays a role for the communication load. Second, changes of synaptic parameters, while induced by spike events, are relatively small so that they evolve on a slower time scale. Roughly speaking, a biological neuron sends out spikes that last each about 1 ms. The rate at which these spike events are generated is a few per second. The slowest dynamics are those of synaptic plasticity which typically needs several spike events to induce a measurable change. Moreover, once changes are induced, they often persist for many hours. In the field of neuroscience, the behavioral phenomenon of learning and memory formation is thought to be intimately linked to the biological rules of synaptic plasticity (Bliss and Lømo, 1973 ; Markram et al., 1997 ; Bi and Poo, 1998 ; Zhang et al., 1998 ; Bi and Poo, 2001 ; Markram et al., 2012 ). To verify in experiments whether a stable memory has been formed it is not uncommon to follow a biological substrate for 24 h or more. If we want to simulate learning and memory formation, the simulation software has to cover time scales from milliseconds to days. To facilitate studies of learning and plasticity in network models it is therefore highly desirable to run simulations as fast as possible. While simulation packages for networks with static (i.e., non-plastic) connections are readily available (Gewaltig and Diesmann, 2007 ; Eliasmith et al., 2012 ; Hoang et al., 2013 ), simulations of plastic brain circuits have received much less attention (Gewaltig and Diesmann, 2007 ; Izhikevich and Edelman, 2008 ; Ananthanarayanan et al., 2009 ). For example, the NEST simulation environment has been released initially for fixed network connections and models of synaptic plasticity have been added later on (Gewaltig and Diesmann, 2007 ; Morrison et al., 2007 ). Recently, increasing efforts are being made to enable real-time simulations by using specialized simulation hardware (Furber and Temple, 2007 ; Schemmel et al., 2010 ) or GPUs (Yudanov et al., 2010 ; Hoang et al., 2013 ). Here we focus on networks of several thousands of neurons. These medium-sized networks are of particular practical importance because they are used in many theory and modeling labs worldwide. Since not all modeling labs have access to super computers we further limit our study to the use of general purpose computers, which can be used either individually or as clusters. In this framework we are interested in strong scaling, i.e., how fast a given network model of fixed size can be simulated. To explore what is currently achievable using standard off-the-shelf hardware and publicly available software, we compare the results and execution times of three typical network simulations—with and without plasticity. In particular we use the multi-purpose simulation frameworks NEST, Neuron and Brian and compare them with our own simulator Auryn. Auryn has specifically been optimized to study plasticity in large-timescale simulations (up to days of simulated time). To minimize run times in such simulations Auryn uses forward Euler integration and relies on single precision arithmetic. In this work we first analyze the trade-off between simulation precision and simulation speed. In particular we focus on a variant of the classic balanced network model by Vogels and Abbott ( 2005 ) and show that meaningful results can be obtained at high simulation speed when using numerical integration algorithms with a comparatively low fidelity (e.g., forward Euler method). We then turn to parallel simulations and analyze by how much network simulations can be sped up and how strong scaling is limited by multiple factors. In particular we identify inter-process communication and spike propagation as the two major limiting factors which prevent a further speed-up in simulations of medium-sized spiking neural networks. In summary we show, by using three examples of standard balanced network models, that real-time simulations are well within reach with today's off-the-shelf hardware. However, the increase of simulation speed well beyond real-time, as required for studying synaptic plasticity and learning, calls for specialized hardware with low communication latencies.",
"discussion": "4. Discussion In this paper we have shown that small or medium size recurrent networks with STDP can be simulated in, or faster than real-time if performance-optimized parallel code is used. However, we also show that the margin for speed-up through parallelization on standard hardware is limited due to finite communication delays and the deviation from strong scaling in the mechanisms of spike propagation. In particular we compared the run times of several standard simulators (Brian, NEST and Neuron) and our own simulator Auryn, when simulating a classic Vogels-Abbott benchmark network. We illustrated that the choice of integration algorithm has a considerable effect on performance. Specifically simulation speed can be increased substantially when numerical precision is not a primary concern. Moreover, in some simulators, such as Brian, the activation of additional performance options can increase simulation speed dramatically without affecting numerical precision. Which numerical precision is necessary to conduct a particular study strongly depends on the questions asked. On the one hand, high numerical precision is almost always desirable because it reduces the risk of emerging systematic errors and artifacts in the analyzed system. On the other hand, the higher associated computational cost of high-precision simulations, can render certain types of studies infeasible. The final compromise between simulation speed and simulation precision has to be taken with care and with respect to the exact nature of the scientific questions addressed. Most network simulations can be sped up even further through parallelism. In particular we compared the run times of the Vogels-Abbott benchmark simulation using NEST, Neuron and Auryn when run on a single machine with 16 cores or a small cluster of four such machines. While parallelization led to increased speed in most cases it was nevertheless difficult to speed up network simulations beyond a tenth of real-time. Since this constitutes a severe restriction in plasticity studies, we analyzed the scaling behavior of Auryn more deeply from which we concluded that this limitation cannot be simply levitated by using more computers. Specifically we found that in the realm of medium-sized network models with plasticity strong scaling ends for a relatively low number of cores. The origin of this saturation was two-fold: First, network simulations are limited by communication delays when simulated in a distributed fashion on a cluster. Second, larger and more computationally costly simulations additionally suffer from the break-down of strong scaling in each process when high numbers of cores per machine are used. In particular we observed a break-down in scaling behavior in distributed simulations using 48 or 64 out of 64 available physical cores. This effect could be due to the fact that in the latter case no dedicated core is available for the operating system. However, we observed similar saturation effects when using only 48 cores. This seems to suggest that another limiting mechanism is responsible. It is tempting to speculate that it is linked to bandwidth limits in shared memory access on a local machine. To gain a deeper insight into which parts of a typical network simulation contribute most to the break-down of strong scaling, we performed multiple profiling studies of the parallel simulations in Auryn. This study revealed that while the pure numerical integration of the neuronal differential equations scaled close to linearly in most cases, scaling of spike propagation was generally sub-linear and could become substantial for large numbers of parallel processes. Spike propagation is a memory intensive process because it generally requires to iterate over large fractions of the synaptic weight matrix in a quasi random order due to the stochastic spiking of neurons. These findings therefore seem to support the idea that memory bandwidth limits are indeed the cause behind the breakdown of strong scaling. It will be an interesting avenue for future studies to directly verify this hypothesis in a detailed memory profiling analysis. Taken together it seems as if the potential for further increase in simulation speed of medium-sized spiking network models on standard hardware is exhausted. Even if it was not for the break-down of strong scaling at the per-process level, strong scaling would still end in distributed simulations at around one tenth of real-time due to communication delays between processes. Therefore, large clusters can only be advantageous if they have extremely low latency communication capabilities. With 10 Gigabit Ethernet becoming increasingly available, a decrease of communication latencies by a factor of 5–10 seems realistic, before yet another performance threshold is reached. At this point a continuation of our study on such machines and on super computers with dedicated low latency communication hardware would be particularly insightful. With the restrictions at hand it is currently difficult to speed up typical simulations of recurrent networks much further than real-time. In particular this means that a simulation of one day of biological time takes at least several hours to complete. To circumvent performance limitations in simulations of spiking neural networks, GPUs have recently received increasing attention as an inexpensive and massively parallel alternative to distributed simulations (Yudanov et al., 2010 ; Richert et al., 2011 ; Brette and Goodman, 2012 ; Hoang et al., 2013 ). As for now, it seems as if these approaches are experiencing similar difficulties (Brette and Goodman, 2012 ) as the ones encountered for the path taken in this manuscript. While real-time simulations are feasible, at present it is not clear if a further decrease of simulation times of networks with realistic plasticity rules is possible. The evident lack of options to increase simulation speed for large- time -scale studies on learning and synaptic plasticity calls for novel ideas of how to approach this type of problem. Noteworthy are approaches addressing neural simulation at the hardware level (Furber and Temple, 2007 ; Schemmel et al., 2010 ). Although most of these projects aim at achieving large scale simulations (neuron numbers comparable to the human brain) in real-time, they might also be a good fit for much smaller network configurations where they could provide a significant speed-up. Regardless of which solutions one considers, it would be inevitable that the system brings the necessary flexibility to support a large variety of synaptic plasticity rules ideally without a notable impairment of the overall performance. Finally, it remains an open question if such specialized modeling hardware can ultimately be made available for theory labs worldwide. In summary we have shown that real-time simulations of plastic networks of point neurons are achievable with appropriate and highly optimized software. However, at the same time increasing simulation speed beyond 10× faster than real-time is challenging due to limitations in the inter-process communications."
} | 4,009 |
32046366 | PMC7074789 | pmc | 2,118 | {
"abstract": "Common mycorrhizal networks (CMNs) allow the transfer of nutrients between plants, influencing the growth of the neighboring plants and soil properties. Cleistogene squarrosa ( C. squarrosa ) is one of the most common grass species in the steppe ecosystem of Inner Mongolia, where nitrogen (N) is often a key limiting nutrient for plant growth, but little is known about whether CMNs exist between neighboring individuals of C. squarrosa or play any roles in the N acquisition of the C. squarrosa population. In this study, two C. squarrosa individuals, one as a donor plant and the other as a recipient plant, were planted in separate compartments in a partitioned root-box. Adjacent compartments were separated by 37 µm nylon mesh, in which mycorrhizal hyphae can go through but not roots. The donor plant was inoculated with arbuscular mycorrhizal (AM) fungi, and their hyphae potentially passed through nylon mesh to colonize the roots of the recipient plant, resulting in the establishment of CMNs. The formation of CMNs was verified by microscopic examination and 15 N tracer techniques. Moreover, different levels of N fertilization (N0 = 0, N1 = 7.06, N2 = 14.15, N3 = 21.19 mg/kg) were applied to evaluate the CMNs’ functioning under different soil nutrient conditions. Our results showed that when C. squarrosa–C. squarrosa was the association, the extraradical mycelium transferred the 15 N in the range of 45–55% at different N levels. Moreover, AM fungal colonization of the recipient plant by the extraradical hyphae from the donor plant significantly increased the plant biomass and the chlorophyll content in the recipient plant. The extraradical hyphae released the highest content of glomalin-related soil protein into the rhizosphere upon N2 treatment, and a significant positive correlation was found between hyphal length and glomalin-related soil proteins (GRSPs). GRSPs and soil organic carbon (SOC) were significantly correlated with mean weight diameter (MWD) and helped in the aggregation of soil particles, resulting in improved soil structure. In short, the formation of CMNs in this root-box experiment supposes the existence of CMNs in the typical steppe plants, and CMNs-mediated N transfer and root colonization increased the plant growth and soil properties of the recipient plant.",
"conclusion": "5. Conclusions This study revealed for the first time that CMNs exist between individuals of C. squarrosa in a constrained environment. Colonization of AM fungi occurred in the donor roots and subsequent CMN formation caused root colonization in the recipient plant. AM fungal inoculation increased the plant biomass, chlorophyll content, and EE-GRSP, T-GRSP, MWD, and SOC. Moreover, the CMNs originating from the donor plant also facilitated and improved plant growth and soil properties in the recipient plant. These findings suggest that AM fungal inoculation and the subsequent establishment of CMNs can play important roles in improving soil aggregation, soil fertility and plant growth. Therefore, this study confirms the existence of mycorrhizal networks in the typical steppe of Inner Mongolia. This provides a basis for understanding the mechanism of intra-plant communication in association with plant growth and development in this grassland system.",
"introduction": "1. Introduction Arbuscular mycorrhizal (AM) symbiosis between higher plant roots and the fungi belonging to the phylum Glomeromycota is one of the most common mutualistic associations in terrestrial ecosystems [ 1 , 2 ]. In AM symbionts, the fungi act as an interface between plant roots and soil, thereby helping the host plant in the acquisition of limiting soil nutrients, such as phosphorus and nitrogen (N). One key characteristic feature of AM fungi is that their hyphae can penetrate into root cortical cells to form intraradical structures and extend outside the roots to form extraradical hyphae in the rhizosphere [ 3 ]. Moreover, extensively branched extraradical mycelia can interconnect neighboring plants to form common mycorrhizal networks (CMNs) [ 4 , 5 , 6 ]. These CMNs can affect the distribution of mineral nutrients like carbon [ 7 , 8 ], N [ 9 ], and phosphorus [ 10 ] among the connected plants. This could ultimately influence the plant’s establishment [ 11 , 12 ], survival [ 13 , 14 ], growth [ 15 ] and physiology [ 16 , 17 ]. However, the underground network is very complex, and a deep understanding of CMN’s formation, existence and functioning requires microscopic or tracer element techniques. The application of an N stable isotope tracer technique has confirmed the transfer of nutrients between CMN-connected plants. For example, a CMN was established by the native AM fungi between the grasses ( Nassella pulchra , Bromus madritensis , and B. hordeaceus ) and the forbs ( Trifolium microcephalum , Sanicula bipinnata , and Madia gracilis ) and CMNs exhibited N communication between the plants [ 18 ]. Barto et al. (2011) also studied the transfer of allelochemicals from source plant to target plant of Tagetes tenuifolia with the help of CMNs [ 19 ]. It is also documented that flax (C 3 -plant) invested little carbon, but obtained N and phosphorous by up to 94% via CMNs from the sorghum (C 4 -plant) [ 20 ], revealing the high dependency of CMN-aided nutrient acquisition from the donor plant. Therefore, these below-ground mycorrhizal networks play important roles in the signal transduction and nutrient sharing between the interconnected plants [ 21 ]. Besides the improvement in plant growth and establishment, the AM fungal extraradical mycelium entangles the soil particles and facilitates their aggregation and stabilization [ 22 ], thereby improving the soil’s physical properties, such as infiltration rate, water holding capacity, and carbon storage [ 23 , 24 ]. Glomalin-related soil proteins (GRSPs), produced mainly by AM fungi, exhibited substantial functioning in cementing soil aggregates and stabilizing soil structures [ 18 ]. It has been reported that GRSPs significantly increased soil stability in the grassland ecosystems of northeast China [ 25 , 26 ]. However, no evidence is available regarding the effect of CMNs on the production of GRSPs and their functioning on soil aggregate stability in the typical steppe. The typical steppe of Inner Mongolia is the dominant vegetation type in semi-arid areas of northern China [ 27 ] and plays an essential role in providing ecological services and life necessities [ 28 , 29 ]. However, anthropogenic activities and climate change have severely degraded steppe grasslands, resulting in decreased soil quality and plant productivity [ 30 , 31 , 32 , 33 ]. This grassland system is particularly sensitive to N enrichment because N is a major limiting soil nutrient in this region [ 34 , 35 ] and even a small amount of change in soil N could have significant effects on plant growth and soil quality [ 36 ]. Therefore, N fertilization has been extensively used to increase the availability of soil N [ 37 , 38 ] enhance plant production [ 39 , 40 , 41 ], and improve soil properties [ 38 ]. These effects can be boosted by mycorrhizal networks that play an active role in ecosystem functioning and regulate N cycling [ 42 , 43 ]. It has also been observed that increased N availability often results in improved plant productivity but decreases the species diversity of the plants [ 44 , 45 ] and leads to the extinction of susceptible functional groups [ 46 ]. Additionally, N enrichment can significantly change the diversity and abundance of soil microbial communities [ 47 , 48 ], causes dormancy, decreases the diversity of the active soil microbial community [ 49 ], and weakens the plant–microbe interactions [ 49 ]. Global N enrichment is considered to be one of the major threats to the structure and functioning of the ecosystem because of its various negative effects on biotic communities [ 50 ]. Therefore, besides the importance of mycorrhizal networks for improving plant growth and soil properties, it is also important to study how the changing environment, such as an increasing amount of terrestrial N deposition, would affect the CMNs’ functioning. Filling this knowledge gap will enable better predictions of the consequence of a change in CMNs functioning under global changing scenarios. Cleistogene squarrosa is a common perennial grass species in the typical steppe of Inner Mongolia. Due to its dominance in various grassland systems, the importance of C. squarrosa has been recognized for the development of a sustainable grassland system. Moreover, mycorrhizal networks play a significant role in stabilizing the long-term dominance of plant species in an ecosystem [ 51 ]. Therefore, it is important to find the existence of CMNs in the typical steppe of Inner Mongolia and their importance for the growth and development of C. squarrosa and its neighboring plants. This research was designed to examine the existence of CMNs between different individuals of C. squarrosa species, and to evaluate the functioning of CMNs across an N gradient. We hypothesize that CMNs exist between individual plants of the same species and, if so, we further address the following key questions: How could CMNs affect the plant growth and soil properties of the neighboring plants? How would the functioning of CMNs change under different levels of N?",
"discussion": "4. Discussion In this study, the mycorrhizal network was established between the inoculated donor and non-inoculated recipient plants of C. squarrosa . Inoculation in the donor plant resulted in root AM fungal colonization and, as a result, the extraradical hyphae from a donor plant moved to the recipient plant and formed AM fungal colonization. This AM fungal colonization in the recipient plant led to the formation of hyphal connection between the donor and recipient plant, called a common mycorrhizal network. This was consistent with the results of [ 61 ], who found that the inoculation of a donor plant with AM fungi resulted in the formation of the mycorrhizal network between trifoliate orange and white clover. Different levels of N fertilization were applied to assess the efficiency of AM fungal colonization and the mycorrhizal network to observe the transfer of nutrients from the donor to recipient plants and their effects on the neighboring plants. It was found that N fertilization had a significant effect on AM fungal colonization and soil hyphal length, which is in line with the previous findings that N-fertilization plays an important role in increasing the AM fungal colonization [ 62 ]. As a result, plant shoot weight, root weight and chlorophyll content in the donor and recipient plants increased as compared to non-mycorrhizal plants. These findings are in agreement with previous studies of trifoliate orange-white clover and flax-sorghum, and Andropogon gerardii , where plant biomass was improved as a result of mycorrhizal networks [ 20 , 57 , 61 , 63 , 64 , 65 ]. This is because these extraradical radical hyphae provided the increased nutrients absorption surface for the plants. The maximum 15 N recovery was found at N2 treatment in the donor and recipient plants, suggesting that the recovery of 15 N decreased at a high level of N fertilization [ 66 ]. Thus, mycorrhizal plants used the nutrients more effectively than non-mycorrhizal plants, resulting in increased plant biomass at N2 treatment. This increase in plant biomass provided a larger sink for 15 N recovery at N2. Therefore, as a result, a lower transfer rate was found at N2 treatment as compared to other N-treatments. Another reason could be that the recovered 15 N at N2 was efficiently utilized by the donor plant to produce the higher plant biomass, and hence found a lower transfer rate at N2, which was also reported previously [ 66 ]. Therefore, AM fungal inoculation and CMNs formation increase plant biomass [ 57 , 63 ]. This effect was also observed in maize plants [ 62 ]. We found that CMNs in C. squarrosa–C. squarrosa association significantly increased the root and shoot weight of the recipient plant, suggesting that CMNs can enhance the plant growth performance of the neighboring plants. The other reason for the increase in plant biomass was due to an increase in chlorophyll content at high N level. However, the mycorrhizal plants showed higher chlorophyll content as compared to the non-mycorrhizal plants. The higher chlorophyll content might be due to more chloroplasts existing in the bundle sheath of inoculated plants [ 67 ]. It was also observed the AM fungal inoculation increased the chlorophyll content in the plants [ 67 , 68 ]. Moreover, other reasons might be the increased stomatal conductance, higher photosynthesis and transpiration rate, and improved plant growth. Therefore, in the mycorrhizal plants, significant effects of N-fertilization were found, with increased plant biomass and chlorophyll content. The mycorrhizal network significantly improved plant growth by the redistribution of nutrients to their neighboring plants. In this study, the mycorrhizal network transferred soil N from the donor to recipient plants on an average of 50%. The ability of the mycorrhizal network to transfer N from the donor to recipient plant varies from 0–80% [ 69 ]. However, much elevated N-fertilization did not increase nutrient transfer and plant biomass because plants may face other limitations, like water, light or space, that limit their growth [ 70 ]. However, at severe N deficiency, the AM fungi might consume additional N sources for their own needs, and as a result, there is little or no N transfer to the plants [ 71 ]. It is also observed that the amount of N transfer is correlated with mycorrhizal colonization level and hyphal length density [ 72 ]. Moreover, it is well documented that the fungal partner makes a significant contribution to the uptake in soil nutrients mediated by the mycelial network [ 73 , 74 , 75 , 76 ], making a significant contribution to improve the performance of the neighboring plants. Soil aggregability was improved with the help of CMNs, as the aggregation of soil particles of different sizes stabilizes the soil organic carbon [ 77 ]. Soil aggregation may be influenced by various factors, such as soil biota, clay, and soil organic carbon [ 78 ]. It is observed that AM fungi can play a significant role in the stabilization of soil particles by the GRSP contents released by the mycelial network. [ 18 ]. In this study, the average EE-GRSP and T-GRSP contents were higher in mycorrhizal treatment than non-mycorrhizal, which is consistent with the previous finding [ 79 ]. The GRSPs’ contents increased with the addition of N, and the maximum GRSP fractions were found at N2. However, after a certain increase in N, the GRSP contents decreased, and similar findings were observed by Sun et al. (2018). This is because, in the beginning, the addition of N quickly relieves the N deficiency in the soil, thereby encouraging the microbial activities to stimulate the production of GRSPs. However, after a certain increase of N results in N saturation into the soil and inhibits microbial activities, the production of GRSPs would decrease [ 80 ]. Therefore, N addition has the potential to increase the GRSPs in the soil [ 61 , 81 ]. AM fungal inoculation played a significant role in the production of GRSPs in both the donor and recipient plants. Moreover, a significant positive correlation was found with soil hyphal length ( Figure 4 ), as reported previously [ 61 , 82 ], which improved the soil structure. These findings are supported by researches showing that the percentage of water-stable aggregates increases with the increase in hyphal length in the pot experiments [ 61 , 82 , 83 ]. Moreover, the long term field experiments in grassland ecosystems also found a positive correlation of hyphal length with water-stable aggregates [ 83 ]. In our case, we also found that the percentage of water-stable aggregates (WSA) at sizes of 2.00–1.00 and 1.00–0.50 mm was significantly higher in the mycorrhizal treatment as compared to non-mycorrhizal treatment ( Table S1 ). Thus, the mycorrhizal hyphae entangle the soil particles and stabilize the macroaggregates [ 84 ] and the GRSPs help in the binding of these macroaggregates [ 85 ]. Besides the GRSPs, it has been widely accepted that AM fungi also play a key role in soil carbon storage, by either depositing organic compounds such as chitin and glomalin in the rhizosphere [ 26 , 86 ] or protecting the soil organic matter from the microbial decomposition through promoting aggregate stability [ 87 ]. Mycorrhizal plants have more SOC contents than non-mycorrhizal plants in our study, which supports previous findings [ 79 ]. Our findings also revealed that MWD significantly increased in the mycorrhizal treatments, and a significant positive correlation of MWD was found with AM fungal colonization, hyphal length, EE-GRSP, T-GRSP, and SOC ( Figure 5 ). This implies that GRSPs fractions, AM fungal colonization, hyphal length, and SOC all significantly improved the soil aggregate stability [ 82 , 88 , 89 ]. For example, the addition of N can increase the AM fungal colonization with a low N level [ 90 ], thereby increasing the GRSPs’ fractions in the soil [ 91 ]. However, N addition beyond a certain level decreases the GRSPs in the soil [ 92 ]. Interestingly, the addition of N significantly increased the SOC in the soil of mycorrhizal treatment as compared to non-mycorrhizal treatment [ 93 ]. SOC is considered an important component of soil fertility and therefore AM fungal inoculation and the subsequent formation of CMNs enhance the soil fertility and improve the growth of donor and recipient plants. This suggests that a common mycorrhizal network would persuade the well-developed mycelium and GRSP fractions (EE-GRSP and T-GRSP), and SOC in the rhizosphere of the recipient plant, resulting in the improvement of soil aggregate stability and the growth of the recipient plant."
} | 4,507 |
33889817 | PMC8050373 | pmc | 2,119 | {
"abstract": "Summary Triboelectric nanogenerator (TENG) is regarded as an equally important mechanical energy harvesting technology as electromagnetic generator (EMG). Here, the input mechanical torques and energy conversion efficiencies of the rotating EMG and TENG are systematically measured, respectively. At constant rotation rates, the input mechanical torque of EMG is balanced by the friction resisting torque and electromagnetic resisting torque, which increases with the increasing rotation rate due to Ampere force. While the input mechanical torque of TENG is balanced by the friction resisting torque and electrostatic resisting torque, which is nearly constant at different rotation rates. The energy conversion efficiency of EMG increases with the increasing input mechanical power, while that of the TENG remains nearly constant. Compared with the EMG, the TENG has a higher conversion efficiency at a low input mechanical power, which demonstrates a remarkable merit of the TENG for efficiently harvesting weak ambient mechanical energy.",
"introduction": "Introduction With the rapid development of the Internet of Things ( Borgia, 2014 ; Roman et al., 2013 ), sustainable and long-life energy supply for billions of distributed electronics has become a major issue( Chu and Majumdar, 2012 ; Gielen et al., 2016 ). Compared with the traditional energy supply by using chemical battery, energy harvesting technology is a more effective way by converting ambient mechanical energy into electric power ( Gao et al., 2020 ; Guo et al., 2016 ; Shao et al., 2018b ; Xie et al., 2014 ; Xu et al., 2019 ). Nowadays, triboelectric nanogenerator (TENG) has attracted growing attention due to its simple structure( Wang et al., 2014 ; Ye et al., 2019 ; Zhang et al., 2019a , 2019b ), high power density, low cost, high flexibility ( Liu et al., 2020 ; Wang et al., 2020a ), and abundant selection of materials ( Li et al., 2020 ; Wang et al., 2020b ; Yan et al., 2020 ; Zheng et al., 2020 ). Derived from the second term in Maxwell's displacement current ( Akande and Lowell, 1985 ; Wang, 2017 ), the TENG has been widely used in micro/nanopower sources ( Seung et al., 2015 ; Tayyab et al., 2020 ; Xiong et al., 2019 ), self-powered sensing ( Jin et al., 2017 ; Khan et al., 2017 ; Wu et al., 2016 ), blue energy ( Shao et al., 2018a ; Wang et al., 2016 ), and high voltage sources ( Bui et al., 2019 ; Li et al., 2017 ; Xia et al., 2019 ) since its first invention in 2012 ( Fan et al., 2012 ). As a new energy technology, the TENG has an equally vital role in harvesting mechanical energy compare to the traditional electromagnetic generator (EMG). The TENG has demonstrated to be comparable and symmetrical with EMG in working mechanisms, governing equations, and output characteristics ( Zhang et al., 2014 ), which has demonstrated that the TENG could be equivalently important as the EMG for harvesting mechanical energy. Moreover, the reports have demonstrated that the TENG has a better output performance than EMG at low frequency ( Zi et al., 2016 ) and small amplitude ( Zhao et al., 2019 ), which indicates a great advantages and possible killer applications of TENG in micromechanical energy harvesting and sensing. However, the applied force/torque and energy conversion efficiency of both generators have not systematically studied and compared yet, which are directly influenced by their different damping characteristics. So, it is extremely important to reveal the characteristics of applied force/torque for both generators and investigate the energy conversion efficiency dependence on input mechanical power. Here in this work, the input mechanical torque of the rotating EMG and TENG are systematically measured, respectively. At constant rotation rates, the input mechanical torque of EMG is balanced by the friction resisting torque and electromagnetic resisting torque, which increases with the increasing rotation rate due to Ampere force. While the input mechanical torque of TENG is balanced by the friction resisting torque and electrostatic resisting torque, which is nearly constant at different rotation rates. The energy conversion efficiencies of both generators are also quantified and compared. With the increase of the input mechanical power, the energy conversion efficiency of EMG is increasing, while that of the TENG remains nearly constant. The comparison results and demonstrations have shown that the TENG has a higher conversion efficiency than that of the EMG at a low input mechanical power. This work has demonstrated a remarkable merit of the TENG under a gentle-triggering, which has verified the possible applications for efficiently harvesting weak mechanical energy from human body and ambient environment.",
"discussion": "Discussion In summary, this work has investigated the applied torque and energy conversion efficiency of the EMG and TENG, which proves that the TENG has a higher conversion efficiency than that of the EMG at the low input mechanical power. The input mechanical torques of the rotating EMG and TENG are systematically measured, respectively. At constant rotation rates, the input mechanical torque of EMG is balanced by the friction resisting torque and electromagnetic resisting torque, which increases with the increasing rotation rate due to Ampere force. While the input mechanical torque of TENG is balanced by the friction resisting torque and electrostatic resisting torque, which is nearly constant at different rotation rates. The energy conversion efficiencies of both generators are also quantified and compared. With the increase of the input mechanical power, the energy conversion efficiency of EMG is increasing, while that of the TENG remains nearly constant. The comparison results and demonstrations have shown that the TENG has a higher conversion efficiency than that of the EMG at a low input mechanical power. Moreover, the fewer the number of power generation units for both generators, the larger dominant power range for the TENG. In the energy conversion demonstration, an LED powered by the TENG and EMG in the same input mechanical energy has been exhibited. The TENG rather than the EMG can light up the LED by a small mechanical energy. This work has demonstrated a remarkable merit of the TENG under a gentle-triggering, which has verified the possible applications for efficiently harvesting weak mechanical energy from human body and ambient environment. Limitations of the study The applied torque, input mechanical power and energy conversion efficiencies of the rotational EMG and TENG are systematically investigated and compared. The results have demonstrated the triboelectric nanogenerator is more suitable for harvesting weak input mechanical energy. However, this method for measuring applied torque, input mechanical power and energy conversion efficiencies is only applicable to the rotational mode. For the TENGs in other modes such as contact-separation mode, further research on measurement methods are expected in the near future. Resource availability Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Chi Zhang. Materials availability This study did not generate new unique reagents. Data and code availability We do not have any code and upon request we can provide the original data."
} | 1,843 |
38985860 | PMC11235161 | pmc | 2,121 | {
"abstract": "Self-healing ability of materials, particularly polymers, improves their functional stabilities and lifespan. To date, the designs for self-healable polymers have relied on specific intermolecular interactions or chemistries. We report a design methodology for self-healable polymers based on glass transition. Statistical copolymer series of two monomers with different glass transition temperatures ( T g ) were synthesized, and their self-healing tendency depends on the T g of the copolymers and the constituents. Self-healing occurs more efficiently when the difference in T g between two monomer units is larger, within a narrow T g range of the copolymers, irrespective of their functional groups. The self-healable copolymers are elastomeric and nonpolar. The strategy to graft glass transition onto self-healing would expand the scope of polymer design.",
"introduction": "INTRODUCTION Self-healing materials are of great advantages for preserving their original functionality and extending their lifespan ( 1 , 2 ). Viscoelastic nature of polymers, wherein the polymer networks can restore elastic energy or flow under external load, potentiates their self-healing from mechanical damage ( 3 , 4 ). The potential has led to intensive studies on developing self-healable polymers over decades ( 5 ). To achieve self-healing, polymer chains require sufficient mobility and interchain interactions ( Fig. 1A ) ( 5 ). Therefore, many studies have investigated to incorporate diverse molecular interactions, including hydrogen bonding ( 6 – 12 ), host-guest chemistry ( 13 , 14 ), metal-ligand coordinations ( 15 – 18 ), ionic ( 19 – 22 ), hydrophobic ( 23 ), ion-dipole ( 24 , 25 ), π-π ( 26 , 27 ), van der Waals ( 28 ), and dipole-dipole interactions ( 29 , 30 ), into flexible polymer networks or gels (plasticized networks). Dynamic covalent bonds ( 31 – 33 ) or hard-soft multiphase ( 34 , 35 ) has also been incorporated into the polymer networks for self-healing. Fig. 1. The design methodology for self-healable polymers based on glass transition. ( A ) Self-healing of polymers occurs by two factors: interchain interaction and chain mobility. ( B ) The two factors can be regulated by glass transition temperature ( T g ) of copolymers and their constituents. ( C ) Statistical copolymers with two types of acrylate monomer units, where one has high T g that stores the interaction energy (named R H ), and the other has low T g that gives the mobility to polymer chains (named R L ). Diverse R H and R L with nonpolar hydrocarbon functional groups are introduced to investigate the self-healing of copolymers. Bz and He are control functional groups for the investigation of molecular interaction effects on the self-healing phenomenon. Since the interchain interactions and the chain mobility were considered individual properties, the paradigm of designing self-healable polymers has been focused on finding specific chemistries or molecular interactions and applying them to flexible chains. We believed and hypothesized that a simple but holistic parameter would exist in designing self-healable polymers. In this work, we present a design methodology for self-healable polymers based on glass transition ( Fig. 1B ). We synthesized a series of statistical (random) copolymers of two nonpolar acrylic monomer units with different T g values ( Fig. 1C ). We found that the self-healing tendency can be explained by the T g values of both the copolymers and the constituent monomer units.",
"discussion": "DISCUSSION The proposed copolymer design has shown a T g -dependent self-healing behavior, where it presents as an analogous way to the existing designs for self-healing polymers. Although the microscopic behavior of self-healing in polymers is complicated, we can design the self-healable copolymers by simply regulating a macroscopic property: the glass transition. The nonpolarity of the copolymers reveals the self-healing tendency, where the self-healing does not rely on specific molecular interactions. Further, the attempt to graft glass transition onto self-healing would expand the design scope of polymer science and engineering. The simplicity of the design parameter and synthetic method may be useful for materials engineers to grant self-healability to diverse polymer types with desired features. Potential applications include the design of encapsulation materials for stretchable electronics, dielectric elastomers for actuators and sensors, polymer electrolytes, or commercial rubbers."
} | 1,128 |
34705502 | PMC8550235 | pmc | 2,122 | {
"abstract": "Anaerobic methanotrophs and methanogens synthesize amorphous carbon.",
"introduction": "INTRODUCTION Most of the elements found on Earth are processed through many different oxidation states as a result of both abiotic and biotic reactions. The oxidation state of carbon varies widely in nature, ranging from methane (−4) to carbon dioxide (+4). One of these states is pure elemental carbon (oxidation state of 0), which occurs in highly ordered crystalline forms, such as graphite and diamond, or as amorphous “black” carbon that lacks crystalline structure ( 1 , 2 ). Crystalline carbon is naturally produced under high heat and pressure in Earth’s crust and in the upper mantle, while amorphous carbon is found primarily in coal and charcoal produced by incomplete combustion of biomaterials as well as in carbonaceous chondrite meteorites ( 3 , 4 ). The different forms and amounts of elemental carbon are influenced mainly by abiotic geochemical processes. However, some evidence exists for the biological degradation of elemental carbon. For example, pyrogenic black carbon in soil is degraded on a centennial time scale facilitated by both abiotic processes and by microbial oxidation ( 5 ). The rate of microbial degradation depends on the degree of thermal alteration of the carbon to generate biochemically accessible sites ( 6 ). The bacterial aerobic catabolism of amorphous carbon was first reported by Potter in 1908 ( 7 ), and more recent studies support microbially facilitated degradation of black carbon ( 8 , 9 ), but the associated biochemical mechanism remains unclear. In addition, fungi have been shown to oxidize diamond-like carbon films ( 10 ), and the fungus Neosartorya fischeri is proposed to metabolize coal via an oxidative enzyme ( 11 ). Despite the evidence for biodegradation of elemental carbon, the biological production of elemental carbon appears to have never been reported. Methane, the most reduced form of carbon, is produced and consumed in anaerobic environments by methanogenic and methanotrophic archaea, respectively. Methanogens are widespread anaerobes that perform methanogenesis with various substrates including H 2 /CO 2 , small methylated compounds, and acetate (fig. S1) ( 12 ). Methanogenesis is one of the first forms of energy metabolisms to arise on Earth and is responsible for the production of over a billion tons of methane each year, which accounts for at least 70% of global methane emissions ( 12 , 13 ). Related to methanogens are anaerobic methanotrophic archaea (ANME) that carry out the anaerobic oxidation of methane (AOM) to CO 2 using a reverse methanogenesis pathway (fig. S1) ( 14 ). AOM is an essential component of the global methane budget and has a substantial role in controlling the amount of methane released from marine sediments ( 15 , 16 ). There are four major clades of ANME—ANME-1 ( 17 ) (subclusters a and b), ANME-2 ( 18 ) (subclusters a, b, and c), ANME-2d ( 19 ) (more recently referred to as Candidatus Methanoperedenaceae) ( 20 ), and ANME-3 ( 21 )—which inhabit different ecological niches and differ in their physiology ( 15 , 21 , 22 ). Most ANME exist in syntrophic consortia with sulfate-reducing bacteria (SRB), which use reducing equivalents obtained from AOM for the reduction of sulfate to sulfide ( 23 – 26 ). Here, we report on the identification and characterization of amorphous carbon produced by two different AOM enrichment cultures. The AOM50 culture (cultured at 50°C) derived from the gas-rich, hydrothermal vents of the Guaymas Basin and is dominated by consortia of ANME-1a and HotSeep-1 partner bacteria ( 24 , 27 , 28 ). The AOM20 culture (cultured at 20°C) derived from cold seeps at the Nile deep sea fan and is dominated by ANME-2a/c and Seep-SRB partner bacteria ( 28 , 29 ). Further, we investigated several methanogenic species and confirmed the production of amorphous carbon in Methanocaldococcus jannaschii , Methanococcus maripaludis , and Methanosarcina barkeri .",
"discussion": "DISCUSSION Structure and origin of biogenic elemental carbon Both AOM cultures and pure cultures of methanogens produce a black material that was identified as amorphous carbon. On the basis of the relative sizes and blackness of the pellets recovered from different cultures, ANME produce substantially more of this material compared with methanogens. The identity of the amorphous carbon was confirmed by Raman spectroscopy that showed the G and D bands characteristic of elemental carbon ( Figs. 3D and 6, A and B ) ( 44 , 45 ). In graphite (black spectrum in Fig. 6A ), the G band is intense and well resolved, while the D band is very low in intensity (in ideal graphite, the D band would be absent). In contrast, activated carbon, a form of amorphous carbon that lacks crystalline structure, has broad G and D bands of relatively equal intensities ( Fig. 6A ). For the black material isolated from the organisms studied here, the G and D bands are also broad with similar intensities, indicating a high level of disorder characteristic of amorphous carbon rather than crystalline graphite. The XPS data show that the carbon pellets are coated with residual proteins that contain sp 3 carbon; however, the bulk of the carbon below the surface is predominately sp 2 based on the D parameter of 19 eV in the C KLL spectra ( Figs. 4D and 6D ). The amorphous carbon from ANME-1a/HotSeep-1 cultures was highly depleted in 13 C with δ 13 C values of −60‰ ( Table 3 ). The stable isotope probing experiments with 13 CH 4 or 13 C-DIC showed that the amorphous carbon produced by ANME derived from DIC and methane-derived DIC ( Fig. 5 and table S1). Thus, the isotope values of the carbon result from a combination of the original isotopic composition of DIC (−6‰), the provided methane (−41‰), and isotope fractionation during carbon fixation ( 29 ). Considering that in marine environments AOM results in δ 13 C-DIC values often down to −50 to −80‰ ( 46 ), amorphous carbon produced by ANME will have highly characteristic isotopic compositions, similar to those reported for ANME lipids ( 17 , 40 , 47 ). In the active AOM culture, the relative production or turnover of amorphous carbon is slower than the formation of the TBC. This suggests that the carbon accumulates in our cultivation procedure because of its highly inert nature. Thus, the presence of amorphous carbon with its characteristic isotope signatures might be an ideal marker to trace AOM activity in the geological record. Summary of reactions that produce elemental carbon To begin considering the processes that could potentially be associated with biochemically formed elemental carbon, we evaluated known reactions that form elemental carbon. In nature, carbon forms abiotically via one of three different routes: 1. 2CO → C + CO 2 2. CO + H 2 → C + H 2 O 3. CO 2 → C + O 2 Reaction #1, the disproportionation of CO, is catalyzed by sulfide-oxide surfaces ( 48 ) and iron and silicon-iron single crystals ( 49 ). Reaction #2 represents the reduction of CO by hydrogen. Reaction #3 is known to proceed only when the oxygen produced is reduced by H 2 , H 2 S, Fe, Fe +2 , etc., to drive the reaction ( 49 ). All of these reactions occur at temperatures above 250°C, and one or more of them have been proposed for the generation of graphite crystals in hydrothermal vents ( 50 ). Notably, methane-rich fluids have recently been shown to generate pure carbon in the form of diamond under very high pressure (5 to 7 GA) ( 51 ). This process occurs by the removal of hydrogen from methane: 4. CH 4 → C + 2H 2 Another route for carbon formation comes from studies on soot formation, which propose that resonance-stabilized hydrocarbon-radical chain reactions are responsible for soot inception and growth ( 52 , 53 ). The first stage in this process is the production of ethylene or acetylene radicals that undergo radical chain reactions to generate polyaromatic hydrocarbons that then undergo further condensation reactions to form the final soot (carbon) particles (fig. S6) ( 54 ). This type of chemistry could be relevant in a biological system because the intermediate, ethylene, is a known biomolecule derived from S -adenosylmethionine ( 55 ), and radical biochemistry is abundant in anaerobes. However, abiotic soot formation reactions occur in flames at high combustion temperatures (1500 K). In addition, we did not observe any evidence of polycyclic aromatic hydrocarbons in samples from AOM consortia (see the Supplementary Materials). Furthermore, ethylene is an inhibitor of methanogenesis ( 56 ), further excluding this compound as a likely precursor to biogenic amorphous carbon. The reactions producing amorphous carbon in ANME and methanogens are unclear, as they lack biological precedence. AOM and methanogenesis use the same central enzymes and coenzymes (fig. S1) ( 22 , 57 ), and our results indicate that the amorphous carbon produced in both ANME and methanogens is mostly derived from CO 2 . Thus, the overall reaction could entail the reduction of carbon dioxide to generate pure carbon and water, which is thermodynamically favorable: 5. CO 2 + 2H 2 → C + 2H 2 O ∆ G ° = −79.8 kJ mol −1 CO 2 The catalytic machinery that would enable the production of elemental carbon is unknown and, thus, will be an important area for future work. Potential physiological role(s) of amorphous carbon The identification of amorphous carbon produced by methanogenic and methanotrophic archaea raises many questions surrounding its possible physiological functions. In both AOM consortia and in methanogens, the amorphous carbon may serve as a scaffolding material for the attachment and interaction with partner organisms or provide a protective barrier to increase resilience to toxic waste products. Alternatively, the carbon may facilitate the transfer of reducing equivalents between consortial partners. In sulfate-dependent AOM, the ANME oxidize methane to CO 2 and the partner bacteria perform sulfate reduction ( 26 , 28 ). Current evidence indicates that this interaction bases on direct interspecies electron transfer (DIET) ( 23 , 24 ). Similarly, some methanogens are also able to receive electrons from bacteria by DIET ( 58 , 59 ). The identification of intercellular wiring consisting of pili-like connecting structures has been reported as a mechanism for DIET in AOM consortia ( 24 ), and cytochrome-based electron transfer may also be involved ( 60 , 61 ). Considering the well-known electric conductivity of elemental carbon materials ( 62 , 63 ), the amorphous carbon may serve these archaea as a conductive element in interspecies electron transfer. This is supported by studies demonstrating that granular activated carbon promotes DIET between M. barkeri and Geobacter metallireducens ( 64 ). The biogenic amorphous carbon identified here shares similarities in chemical composition and, possibly, in function to biochar, a carbon-rich material generated from thermochemical treatment (usually pyrolysis or gasification) of organic waste in oxygen-limited conditions ( 65 ). The elemental composition and chemical characteristics of biochar can vary widely depending on the type of biomass as well as the reaction conditions for its production. Although primarily carbonaceous, there are key functional groups present on the surface of biochar that result in its various useful properties. Like activated carbon, biochar has been shown to promote DIET through a proposed electron conduction role ( 66 ). Biochar is additionally proposed to enhance methanogenesis during anaerobic digestion by serving as a pH buffer and acting as a surface for colonization ( 67 ). Other work has shown that biochar is redox active and can serve as an electron acceptor and an electron donor to drive microbial metabolism ( 68 , 69 ). Klupfel et al. ( 69 ) demonstrated that the ability of biochar to reversibly accept and donate electrons is due to the presence of quinone/quinol molecules. Most notably, biochar was recently demonstrated to serve as the sole electron acceptor for AOM by ANME-2d where the redox activity was attributed to its oxygen-based functional groups ( 70 ). Our XPS results showed that the surface of the black carbon pellets from the organisms studied here contained carbon bonded to heteroatoms including C─O/C─N and C═O/C─O─C. Some of these characteristics are due to adventitiously bound proteins, but similarly to biochar, the amorphous carbon could also be associated with or functionalized with redox active groups that act as electron sources/sinks or electron carriers. Final remarks The identification of biogenic amorphous carbon is a remarkable finding because elemental carbon formation was thought to occur only at elevated temperatures and pressures in chemical processes over geological time scales or as result of incomplete burning of organic materials. Difficulties associated with completely purifying and accurately quantifying the amorphous carbon currently preclude more detailed investigations of its function and biosynthesis. Thus, future studies will be required to determine the metabolic origin and function of amorphous carbon in the context of complex microbial consortia. Because of the wide distribution and abundance of ANME and methanogens, the biogenic formation of amorphous carbon may have important roles in carbon cycling and climate regulation. This carbon may persist in sediments over geological time scales and thus may represent a so far overlooked carbon sink in nature."
} | 3,386 |
35577982 | PMC7613029 | pmc | 2,124 | {
"abstract": "Recent studies have brought forward the critical role of emergent properties in shaping microbial communities and the ecosystems they are part of. Emergent properties - patterns or functions that cannot be deduced linearly from the properties of the constituent parts - underlie important ecological characteristics such as resilience, niche expansion, and spatial self-organisation. While it is clear that emergent properties are a consequence of interactions within the community, their non-linear nature makes mathematical modelling imperative for establishing the quantitative link between community structure and function. As the need for conservation and rational modulation of microbial ecosystems is increasingly apparent, so is the consideration of the benefits and limitations of the approaches to model emergent properties. Here we review ecosystem modelling approaches from the viewpoint of emergent properties. We consider the scope, advantages, and limitations of Lotka-Volterra, consumer-resource, trait-based, individual-based, and genome-scale metabolic models. Future efforts in this research area would benefit from capitalising on the complementarity between these approaches towards enabling rational modulation of complex microbial ecosystems.",
"introduction": "Introduction Microbial communities profoundly contribute to human and planetary health [ 1 – 5 ]. Yet, the quantitative principles underlying community composition and assembly, and the link between community diversity and function, remain largely unknown. Studying the emergent properties of microbial communities requires relating interactions across spatio-temporal scales, as well as their evolutionary dynamics. Unravelling, and ultimately predicting, community dynamics and emergent properties is a topical challenge in microbial ecology. Higher-order units - from microbial consortia to entire ecosystems - feature macroscopic properties that emerge from microscopic interactions [ 6 , 7 ]. Such emergent properties refer to any pattern or function that cannot be deduced as the sum of the properties of the constituent parts ( Fig. 1 ). Examples in microbial systems include resilience to abiotic and biotic perturbations [ 8 ], stable co-existence [ 9 , 10 ], and biochemical abilities [ 11 – 15 ]. Emergent properties typically arise when community members reach some threshold level of community size and connectivity [ 16 – 18 ]. Conditions such as high flow rates or high media viscosity can obstruct connectivity and hence limit the onset of emergent properties [ 19 , 20 ]. Finding the threshold connectivity is often difficult, even in a defined community, due to the non-linear nature of emergence. The threshold is property-specific and may also (non-linearly) depend on environmental factors [ 21 ]. Hence, descriptions of emergent properties have been largely qualitative, although synthetic biology approaches are starting to characterise emergent properties quantitatively [ 14 , 22 ]. Natural communities are difficult to study in situ due to their inherent dynamic nature and heterogeneity of habitats. Further, abiotic variables influencing the communities are difficult to measure or control in situ. Laboratory experiments, often considering interactions between 2 or 3 community members, are amenable to quantitative analysis but limited in capturing emergent properties of natural, more diverse microbial communities. Defined communities with larger membership can be constituted in vitro [ 23 ] yet remain limited to culturable microbes [ 24 ]. Mathematical models are therefore indispensable for linking community composition and connectivity to emergent functions. Models can bridge principles learned from simple laboratory systems to complex natural ecosystems; a continuity that is very difficult to achieve in an experimental setting. Complex natural systems like microbial communities are never fully closed, and more than one model will be plausible for any data available from the system of interest [ 25 , 26 ]. The choice of the model depends on the research question and the data available to estimate the parameters. When experimental data is limited, models can be set-up using first principles [ 27 ]. Such models can be used to probe system responses to perturbations that are inaccessible to experimentation or observations. In comparison to statistical analyses, models can help in generating mechanistic hypotheses and establishing causal relations. Examples include identification of bacteria conferring colonisation resistance against Clostridium difficile [ 28 ], and emergence of stability through competition [ 29 , 30 ]. Nevertheless, models are often most useful in complementation with statistical patterns, e.g., a recent study using metabolic modelling uncovered polarisation between cooperation and competition in microbial communities [ 31 ]. Two general approaches have been commonly used to model community dynamics and emergent properties: ecological models with species or cells as basic units and interactions among them as the focus, and genome-scale metabolic models that have intra-cellular reactions as main units and nutrient generation/consumption as the focus. As the use of genome scale metabolic models has been reviewed previously [ 32 – 37 ], we focus on four commonly used ecological models. We provide an overview of their advantages in capturing emergent properties, discuss adaptations that challenge their limitations, and comment on their complementarity towards predictive modelling of complex ecosystems.",
"discussion": "Discussion Since microbial community dynamics and properties can be attributed to (interaction of) various complex networks, such as gene regulatory, metabolic, cell-environment and cell-cell, identifying which of these should be included to accurately predict the emergent property of interest is key in choosing an appropriate modelling approach ( Fig. 3 , Supplementary Table 2). Depending on the temporal and spatial characteristics of the property of interest, a combination of approaches is advantageous. Each approach has advantages and limitations in terms of integrating omics data, handling missing data, generating analytical solutions, predictive power, scope of applications, and amenability for generating hypotheses that can be experimentally tested. The choice of model(s) also depends on whether a general phenomenological understanding is sufficient, or if the goal is to fit data to generate predictions in a specific system. While the top-down population-based approaches (i.e., Lotka-Volterra, MacArthurian consumer-resource and phenomenological trait-based models) are generally more suited for modelling resilience and co-existence, the bottom-up approaches (i.e., individual-based and genome-scale metabolic models) are typically better suited for modelling phenotypic trait ranges and spatial patterns. Each of these approaches can be adjusted in the number of dimensions included and units modelled. The distinction between the modelling approaches detailed in this review is not necessarily categorical. For example, adaptations of a consumer-resource model can make it fall within the category of individual-based models, whereas an individual-based model with few dimensions and low resolution can be less accurate in predicting the emergent phenotypic property of substrate utilisation than a multi-population consumer-resource model. As ecosystem stability and resilience become a pressing issue on health and environmental fronts [ 3 , 173 , 174 ], the need for developing predictive models cannot be overstated. With computational power becoming less limiting, and molecular and single-cell data becoming increasingly available, we envisage that fusion between different ecological and cellular-level models will, in the coming years, enable predictive modelling of emergent properties in complex ecosystems from the molecular to the ecosystem scale."
} | 1,987 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.